<%BANNER%>

Advances in Life Cycle Assessment and Emergy Evaluation with Case Studies in Gold Mining and Pineapple Production

Permanent Link: http://ufdc.ufl.edu/UFE0041906/00001

Material Information

Title: Advances in Life Cycle Assessment and Emergy Evaluation with Case Studies in Gold Mining and Pineapple Production
Physical Description: 1 online resource (200 p.)
Language: english
Creator: Ingwersen, Wesley
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: assessment, chain, costa, cycle, dore, emergy, engineering, environmental, evaluation, gold, impact, indicator, life, modeling, performance, peru, pineapple, range, rica, supply, sustainability, uncertainty
Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Life cycle assessment (LCA) is an internationally standardized framework for assessing the environmental impacts of products that is rapidly evolving to improve understanding and quantification of how complex product systems depend upon and affect the environment. This dissertation contributes to that evolution through the development of new methods for measuring impacts, estimating the uncertainty of impacts, and measuring ranges of environmental performance, with a focus on product systems in non-OECD countries that have not been well characterized. The integration of a measure of total energy use, emergy, is demonstrated in an LCA of gold from the Yanacocha mine in Peru in the second chapter. A model for estimating the accuracy of emergy results is proposed in the following chapter. The fourth chapter presents a template for LCA-based quantification of the range of environmental performance for tropical agricultural products using the example of fresh pineapple production for export in Costa Rica that can be used to create product labels with environmental information. The final chapter synthesizes how each methodological contribution will together improve the science of measuring product environmental performance.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Wesley Ingwersen.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Brown, Mark T.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041906:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041906/00001

Material Information

Title: Advances in Life Cycle Assessment and Emergy Evaluation with Case Studies in Gold Mining and Pineapple Production
Physical Description: 1 online resource (200 p.)
Language: english
Creator: Ingwersen, Wesley
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: assessment, chain, costa, cycle, dore, emergy, engineering, environmental, evaluation, gold, impact, indicator, life, modeling, performance, peru, pineapple, range, rica, supply, sustainability, uncertainty
Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Life cycle assessment (LCA) is an internationally standardized framework for assessing the environmental impacts of products that is rapidly evolving to improve understanding and quantification of how complex product systems depend upon and affect the environment. This dissertation contributes to that evolution through the development of new methods for measuring impacts, estimating the uncertainty of impacts, and measuring ranges of environmental performance, with a focus on product systems in non-OECD countries that have not been well characterized. The integration of a measure of total energy use, emergy, is demonstrated in an LCA of gold from the Yanacocha mine in Peru in the second chapter. A model for estimating the accuracy of emergy results is proposed in the following chapter. The fourth chapter presents a template for LCA-based quantification of the range of environmental performance for tropical agricultural products using the example of fresh pineapple production for export in Costa Rica that can be used to create product labels with environmental information. The final chapter synthesizes how each methodological contribution will together improve the science of measuring product environmental performance.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Wesley Ingwersen.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Brown, Mark T.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041906:00001


This item has the following downloads:


Full Text





ADVANCES IN LIFE CYCLE ASSESSMENT AND EMERGY EVALUATION WITH
CASE STUDIES IN GOLD MINING AND PINEAPPLE PRODUCTION




















By

WESLEY W. INGWERSEN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2010

































2010 Wesley W. Ingwersen





























To the memory of James H. Weeks and Blanche R. Ingwersen, two of my grandparents
who passed away late in the course of my Ph.D. program, but who believed in me and
forever inspire me.









ACKNOWLEDGMENTS

I first and foremost thank Dr. Mark Brown, my Ph.D. adviser, who provided me the

opportunity to complete the degree and inspired the pursuit through his teaching and

innovative work. I thank all my other committee members for the insights and criticisms

that were incorporated into this dissertation.

My studies and field work would not have been possible without the financial and

administrative support of the Department of Environmental Engineering Sciences. I

received the support of a Latin American Studies Tinker Travel grant for my research in

Peru. My research in Costa Rica was facilitated by the University of Florida University

of Costa Rica Conservation Clinic under the direction of Tom Ankersen.

Numerous persons provided me direct support for my field studies. I thank in

particular Ricardo Gallardo in Cajamarca, Peru and various employees of the

Yanacocha mine, Randall Arias from PROCOMER in Costa Rica, and Dr. Mauricio Avila

from the University of Wisconsin. I anonymously thank the pineapple companies that

agreed to participate in this study.

Through the formation and evolution of my topics, my travels, data analysis, and

final drafting, my wife Laura has been my steadfast intellectual and emotional

companion.

Finally, my family have provided me the encouragement to bring this intellectual

and personal journey to fruition a heartfelt thank you to all of you.









TABLE OF CONTENTS

page

A C KNOW LEDG M ENTS ......... ............... ............................................. ............... 4

LIST O F TA BLES .......... ..... ..... ............................................................. ........ 8

LIST OF FIGURES.................................. ......... 11

A BST RA CT ............... ... ..... ......................................................... ...... 13

CHAPTER

1 INTR O D U CT IO N ............................................................................................. 14

Measurement of Sustainable Production and Consumption ................................. 14
Life Cycle Assessment as a Measurement Tool ............................................. 15
Research Problems in Life Cycle Assessment ......... .. ............. .............. ...... 17
Life Cycle Impact Assessment (LCIA) Indicators for Resource Use ................ 17
Applications of LCA for Non-OECD Country Products .................. 23
Research Overview ...... .... .............. .......................................................... 27

2 EMERGY AS AN IMPACT ASSESSMENT METHOD FOR LIFE CYCLE
ASSESSMENT PRESENTED IN A GOLD MINING CASE STUDY...................... 29

Introduction ........................ ........ ..... ................ ...... 29
Emergy in the LCA Context .......................... ........ 29
A Case Study of Emergy in an LCA of Gold-Silver Bullion Production ............. 33
M methodology ....... ...... ................................. ........... .... ...... 36
Em ergy and Energy Calculations .......... .............................. ........... ...... 38
Uncertainty M modeling ................................ ............... 40
A llo c a tio n ......... .................. .................... ...... ............ 4 1
Data Management and Tools ................. ...... .. ............ ............... 41
R e s u lts ......... .. ............... ................ ......... ............... ................... ......... 4 2
Environmental Contribution to Gold, Silver, and Mercury in the Ground .......... 42
Environmental Contribution to Dore ............. .... ..... ...... ........... 43
Emergy by Unit Process ......... ....... ......... .............. 44
Allocation and Emergy Uncertainty .......... .......... ............. .............. 46
D discussion .............. ............................................................................... 47
Usefulness of Emergy Results ............ ...... ...... ....................... 47
Emergy in LCA: Challenges ............. ....................... .................50
Challenges of using emergy with LCI databases and software............... 51
Energy in environmental support not conventionally included in energy
e v a lu a tio n ......... ............................................... ...................... 5 3
Uncertainty in unit energy values............. .. ....... .......... ........ 54
Emergy and Other Resource Use Indicators .............. ............ ............... 55









3 UNCERTAINTY CHARACTERIZATION FOR EMERGY VALUES...................... 58

Introduction ............................................................................................................. 58
Sources of U uncertainty in U EV s ................................................................... 59
Models for Describing Uncertainty in Lognormal Distributions ...................... 60
Models for Uncertainty in UEVs ......... ............... .............. ............... 62
Selecting Appropriate Methods for Uncertainty Estimations....................... ... 62
Modeling Procedure and Analysis ...... ................................. 64
Results ........... .... .............. .................. ............... 70
D discussion and C conclusions ............................................................ .... ............... 74
How Much Uncertainty is in a UEV and Can it Be Quantified?...................... 74
Comparing the Analytical and Stochastic Solutions ............... ... ............ 75
Conclusions ......... ...................................... ... ...... 77

4 LIFE CYCLE ASSESSMENT FOR FRESH PINEAPPLE FROM COSTA RICA -
SCOPING, IMPACT MODELING AND FARM LEVEL ASSESSMENT................ 81

Introduction ............... ......... ....... .......... ........ .... .... ......... 81
Objectives ............................................... ...... 81
The Fresh Pineapple System in Costa Rica...... ........ ..... ................. 82
M methods .................. .......... .. .. .................... .........84
System Boundaries and Functional Units................................... ............ 84
Data Collection ................................................................................................. 85
Em missions and Im pact M odels................... .................... .................. 86
Estimating the Sector Range of Environmental Performance........................ 90
LC IA Indicators .... .... ......... ......... .... .............. ..... .......... 93
Soil erosion im pact.................................................. .................... 93
Cumulative energy demand ......... ..... ...... .... ........................ 94
Virtual water content and stress-weighted water footprint....................... 94
Aquatic eutrophication ........... ........ .... .......... ............. 96
Human and freshwater ecotoxicity....... ............... ... ......... ..... 97
Other indicators......................... ....... ......... ............... 98
R results ....................... ........ ...... ..... ............................. 99
Pineapple Sector Inventory ............... ......................................... .. ..... 99
Soil Erosion ......................... ....... ....... ............................... 99
Cumulative Energy Demand (CED) of Pineapple ................. ................. 100
C a rbo n F o otp rint ........................................ .. ........................... ..... 10 1
Virtual Water Content and Stress-W eighted Footprint................................ 103
A quatic Eutrophication ......................................_.......................... 105
Human and Ecological Toxicity .......... ............. ......... ................ 107
Results Summary ................................. ................. 109
Discussion ................................. .....................................109
The Significance of Regionalized Emissions and Impact Models.................. 111
Estimated Environmental Impacts .......... ....... ..... ........ .... ............. 112
Potential Impacts Not Measured ..................................................... .......... 113
Conclusions and Recommendations for Farm Level LCA of Fruit Products ........ 114



6









5 SUM M A RY A N D SY NTHESIS........................................................ ................. 118

Summary .................................................... .. ................................... 118
Chapter 2 Summary ............................................. .................. 118
Chapter 3 Summary ............... ...... .............................. 119
Chapter 4 Summary ............... ...... .............................. 120
S y n th e s is .............. ..... ............ ................. ............................................. 1 2 2

APPENDIX

A SUPPLEMENT TO CHAPTER 2: PROCESS TREE AND UNCERTAINTY
ESTIMATES ................. ........ ............... 127

B SUPPLEMENT TO CHAPTER 2: LIFE CYCLE INVENTORY OF GOLD MINED
AT YANACOCHA ......................... .......... ......... 130

C SUPPLEMENT TO CHAPTER 3: R CODE FOR STOCHASTIC UNCERTAINTY
MODELS.................................... ...... ............... 159

D SUPPLEMENT TO CHAPTER 4: ADDITIONAL TABLES AND FIGURES........ 167

LIST OF REFERENCES ....... .................. .................. 183

BIOGRAPHICAL SKETCH ............... .. ........ ................. 200









LIST OF TABLES


Table page

2-1 Summary of emergy in mine products based on two allocation rules............... 44

3-1 Elements of uncertainty in the UEV of lead in the ground ...... ....................... 60

3-2 Unit emergy value models used for parameter uncertainty calculations............. 67

3-3 Analytical uncertainty estimation for lead UEV, in ground ...... ....................... 71

3-4 Emergy summary with uncertainty of 1 kg of sulfuric acid ...... ....................... 72

3-5 UEV uncertainty estimated from the analytical solution.......................... 79

3-6 UEV Monte Carlo results and comparison of model Cl's with lognormal,
hybrid, and normal confidence intervals ...... ............. .... ........... 79

4-1 Summary table for impacts of 1 kg pineapple delivered to packing facility....... 117

A-1 Uncertainty estimates for UEVs for inputs into gold-silver bullion production... 128

A-2 Estimation of total uncertainty in gold in the ground ..................... ......... 128

A-3 Estimation of total uncertainty of silver in the ground ............. .............. 129

B-1 Inputs to process 'Dore, at Yanacocha'.............. ................... ..... ........... 134

B-2 Inputs to process 'Exploration, at Yanacocha'. ..... ..... .......... ....... ......... 135

B-3 Inputs to process 'Mine infrastructure, Yanacocha' ...................................... 136

B-4 Inputs to process 'Extraction, Yanacocha'........ ... ...... .. ........... 136

B-5 Inputs to process 'Leaching, Yanacocha' ......... .. .................................... 138

B-6 Inputs to process 'Leach Pad, Yanacocha'......... ... ..... ............. 138

B-7 Inputs to process 'Leach Pool, Yanacocha'................. ............. ............... 138

B-8 Inputs to process 'Processing, Yanacocha................................................. 139

B-9 Inputs to process 'W ater Treatment, Yanacocha'.................. .................... 140

B-10 Inputs to process 'Conventional Process Water Treatment, Yanacocha' ......... 140

B-11 Inputs to process 'Reverse Osmosis Process Water Treatment, Yanacocha'.. 141









B-12 Inputs to process 'Acid Water Treatment, Yanacocha' ................. .............. 141

B-13 Inputs to process 'Reclamation, Yanacocha'........................ ................... 142

B-14 Inputs for process 'Sediment and dust control, Yanacocha' .......................... 143

B-15 Comparison of this inventory with the equivalent Ecoinvent process ............... 146

B-16 List of processes in the 'Gold_Yanacocha' project inventory.......................... 147

B-17 Mine hauling road parameters, based on Hartman...................... ............... 149

B-20 Mine vehicle data..... ..................................... .......... 150

B-21 Mass balance of leaching, processing, and water treatment.......................... 151

B-22 Inventory of Peruvian road transport. ..... ................................. 154

B-23 Assumed origins and transport distances for inputs to mining....................... 156

B-24 System -level param eters ............................................................. ............... 157

B-25 Uncertainty estimates for inventory data using Ecoinvent method ................... 158

D-1 Inputs to one kg pineapple at the packing facility. .......... .... ................ 167

D-2 Emissions from one kg pineapple at the packing facility................................ 168

D-4 Emissions estimations for mineral-P in applied fertilizers.............................. 170

D-5 General assumptions used in the FAO CROPWAT model............................. 170

D-6 Crop water requirement variables for CROPWAT ............. ...................... 170

D-7 RUSLE2 parameters for Pineapple in Costa Rica ............. ...................... 171

D-8 Parameters modified for USETox-CR model .............................................. 173

D-9 Sensitivity analysis of the RUSLE2 model customized for pineapple in CR. .... 175

D-10 Sensitivity analysis of the FAO CROPWAT model to variables found in
pineapple cultivation. ..... ....... .......... ... ... .. ............... 176

D-11 Sensitivity analysis of PestLCI model for pineapple conditions ...... .................. 176

D-12 Recalculation of Pimentel (2009) energy demand for US oranges............... 177

D-13 Recalculation of Pimentel (2009) energy demand for US apples .................... 177

D-14 Recalculation of Coltro (2009) energy demand for BR oranges .................... 178









D-15 CED values for inputs used in recalculations of Orange BR, Orange US and
A p p le s U S .................... ................................. ................. 1 7 8









LIST OF FIGURES


Figure page

2-1 Proposed boundary expansion of LCA with emergy.................................. 32

2-2 Gold production system at Yanacocha with modeled flows and unit
p ro ce sse s ...................... .. .. ......... .. .. ....... .................................. 3 7

2-3 Environmental contribution (emergy) to dore by input type .............................. 45

2-4 Emergy and primary energy in 1 g of dore by unit process .............................. 45

2-5 Monte Carlo analysis of 1 g of dore .................................... ..... ................ 48

3-1 Conceptual approach to modeling uncertainty ............... ............. .............. 66

3-2 Published UEVs for electricity by source from Brown and Ulgiati (2002),
superimposed upon a modeled range of the oil UEV ................................. ... 80

4-1 Fresh pineapple production unit processes and boundaries for the LCA .......... 84

4-2 Contribution to CED of pineapple, at packing facility...... ......... ..... ........ 102

4-3 Non-renewable CED of one serving pineapple in comparison with
evaluations of the farming stage of other fruits....... ..... .............................. 102

4-4 Contribution to carbon footprint of pineapple, at packing facility..................... 103

4-5 Carbon footprint of one serving pineapple in comparison with evaluations of
the farming stage of other fruits .................... ........... ................. 104

4-6 Virtual water content (VWC) for pineapple in comparison with other fruits....... 105

4-7 Contribution to potential eutrophication of pineapple by emission.................... 106

4-8 Preliminary comparison of potential eutrophication effects of one serving
pineapple in comparison with evaluations of the farming stage of other fruits.. 106

4-9 Relative contribution of active ingredients of pesticides used in pineapple
production to human toxicity and freshwater ecotoxicity............................... 108

A-1 SimaPro process tree of environmental contribution (sej) to 1 g dore.............. 127

B-1 Process overview ............. ......... ......... .............. ............... 132

D-1 Emission fractions of applied pesticides in PestLCI-CR vs. the PestLCI
d e fa u lt................................................ ....... .......... ...... 1 7 9









D-2 Freshwater ecotoxicity characterization factors for pesticides in USETox-CR
vs USETox-Default ............. ..... ............. ............. .. ............... 180

D-3 Human toxicity characterization factors for pesticides in USETox-CR vs
U S ETox-Default .. ....................... .......... .......... ................ .............. 181

D-4 Human toxicity and freshwater ecotoxicity for pesticide emissions from
pineapple production in the baseline scenario............................................ 182









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

ADVANCES IN LIFE CYCLE ASSESSMENT AND EMERGY EVALUATION WITH
CASE STUDIES IN GOLD MINING AND PINEAPPLE PRODUCTION


By

Wesley W. Ingwersen

August 2010

Chair: Mark T. Brown
Major: Environmental Engineering Sciences

Life cycle assessment (LCA) is an internationally standardized framework for

assessing the environmental impacts of products that is rapidly evolving to improve

understanding and quantification of how complex product systems depend upon and

affect the environment. This dissertation contributes to that evolution through the

development of new methods for measuring impacts, estimating the uncertainty of

impacts, and measuring ranges of environmental performance, with a focus on product

systems in non-OECD countries that have not been well characterized. The integration

of a measure of total energy use, emergy, is demonstrated in an LCA of gold from the

Yanacocha mine in Peru in the second chapter. A model for estimating the accuracy of

emergy results is proposed in the following chapter. The fourth chapter presents a

template for LCA-based quantification of the range of environmental performance for

tropical agricultural products using the example of fresh pineapple production for export

in Costa Rica that can be used to create product labels with environmental information.

The final chapter synthesizes how each methodological contribution will together

improve the science of measuring product environmental performance.









CHAPTER 1
INTRODUCTION

Production of goods and services is inextricably tied to the environment. As basic

resources for modern economies are becoming more costly or less available (e.g.,

freshwater and petroleum) and impacts of productive activities have created local and

global scale environmental change (e.g., climate change), the need to understand

connections between the environment and economy has become more critical. The

delegates to the UN Conference on Environment and Development, representing over a

100 of the world's nations, acknowledged in the milestone Rio Declaration on

Environment and Development, or Agenda 21, that all productive processes in

economies are dependent upon sources of energy and materials from the environment

and sinks to absorb the pollution that they generate (principle 8, UN 1992). At the

World Summit on Sustainable Development a decade later, it was furthered

acknowledged that measurement systems are necessary to quantify these

dependencies and pollution impacts for the purposes of achieving more sustainable

development (Chapter 3, UN 2005).

Measurement of Sustainable Production and Consumption

Measurement is the first step toward effective management and protection of the

environment in the context of productive processes in economies. But the concept of

measurement of environmental impacts of production processes has been evolving with

broader understandings of what, how and where impacts occur and who in turn is

responsible for those impacts. The first generation of environmental policy in the United

States (such as the Clean Air Act of 1970), and still the dominant form of regulation in

place in the United States, is primarily based on the regulation of pollution "at the pipe",









implicitly focusing only on pollution at the point of occurrence and obligating only the

party responsible at that point to reduce or cease the pollution. This style of legislation

reflects the assumption that impacts should be measured only at the point of impact.

But the ultimate purpose and driver of a production processes is to provide for an end

product or service, and thus the impacts of productive processes can all be related to

the intermediate or end products. That product or service is demanded by a consumer,

and that consumer shares responsibility for the environmental impacts that occur along

the production chain. Shared producer and consumer responsibility was recognized in

the Rio Declaration and reinforced in international action plans such as the Marrakesh

Process launched at the World Summit on Sustainable Development (UN DESA 2008),

and is now becoming further integrated at national, regional, and local scales, especially

through voluntary public and private initiatives (e.g. Environmental Management

Systems, Extended Producer Responsibility policies, corporate greenhouse gas

accounting standards). It then becomes clear that measurement tools are needed that

relate these broader impacts to products or services in a way that accounts for impacts

along the full production chain such that management can involve both producer and

consumer, and so that no impacts associated with production processes are left out.

Life Cycle Assessment as a Measurement Tool

Life Cycle Assessment (LCA) is an established and standardized framework for

assessing impacts of production processes and for relating full life-cycle impacts to a

final product (ISO 2006c). LCA is being used globally for product systems for purposes

of design, management, and communication of environmental performance (UNEP

2007), as well as to guide environmental product policy (European Commission 2003).









LCA is an appropriate framework for measuring impacts of products because it

uses a full life cycle perspective from "cradle-to-grave" thus including all product

stages during which significant impacts might occur, including all production and

consumption stages. This begins with assessing the goal and scope of a product

system and continues with an inventory of inputs and emissions by product stage

relevant to estimation of impacts. Estimating the impacts of these emissions is done

with impact characterization factors developed from impact models. Impacts are all

related to a unit of the product serving a particular functional purpose, called a

functional unit. These impacts typically measure use of environmental sources

(resource use indicators) or stressors on environmental sinks (impact indicators).

Impact indicators depict impacts at varying points in the chain of causality from the

release of an emission to its ultimate impact (end-point) on primary areas of concern

(human health, natural environment, resources, manmade environment), depending

upon the state of the science for modeling impacts along this chain (Bare et al. 2006).

LCA is arguably the strongest framework for measuring environmental impacts of

production activities for the complex, global supply chains typical of modern products.

Ness and colleagues (2007) categorized measures of sustainability based on their focus

and their temporal aspects. In contrast with techniques such as environmental impacts

assessment, which is focused on future activity and is highly-location specific, LCA is

primarily focused on current systems (though can be used for design purposes) and is

not limited in focus to one particular site. In contrast with sustainability indices (e.g.

environmental pressure indicators), which are often retrospective indicators of larger

systems, LCA is more product specific. LCA also originates from industrial ecology and









engineering, and its quantification by particular unit processes make LCA results more

relevant for product management. In comparison with other systems-oriented

approaches such as embodied energy or emergy analysis, LCA is multi-criteria, which

provides a broader view of products and makes it less likely that important impacts are

overlooked (Ulgiati et al. 2006).

Research Problems in Life Cycle Assessment

The effectiveness of the bold intention to use LCA to relate a product to all the

damages (or benefits) that occur to the environment over the life cycle of its production,

use, and disposal depends upon detailed inventories of complex product life cycles as

well as accurate models to estimate environmental damages related to resource used

or emissions that occur with these life cycles. LCA adapts scientific theory and models

from many other fields to accurately identify and model impacts and thus is only as

advanced as the science and its application within this fields. LCAs are often limited by

incomplete or inappropriate data and absence of relevant impact models. Two focal

areas of LCA that specifically need to be addressed to better measure sustainable

production and consumption in a manner applicable to global supply chains are 1)

resource-use indicators and 2) impact models for processes occurring in non-OECD

product systems. These problems and a proposed plan for addressing them are

described in the following three sections.

Life Cycle Impact Assessment (LCIA) Indicators for Resource Use

As described above, indicators in LCA may be broadly split into resource use and

impact indicators. Resource use indicators may be based on the use of a particular

energy source or material (e.g. fossil energy use or freshwater use) or may be an

aggregate measure. Furthermore, they may focus on relating that use to ultimate









availability (e.g. mineral resource depletion) or simply just report usage. Relating

different indicators of resource use together may require use of subjective weighting

criteria when there is not a physical basis for relating the resources (Guinee 2002). But

the impact of using different resources may be related together without the need for

subjective judgment if resources can be characterized on a common physical basis with

a common unit, which is instructive for synthesizing the effects of resource use.

Various authors have argued for the need to incorporate a unified measure of resource

use into LCA to limit resource consumption associated with productive processes

(Finnveden 2005; Seager and Theis 2002; Stewart and Weidema 2005).

Single-unit measures of resource use have been extensively developed outside of

the LCA framework, but not all of these methods have been applied as indicators in

LCA. These methods typically aggregate resource use using a common biophysical

unit. Common biophysical units may be units of mass, land area, or energy. Life cycle

based methods using mass include extensions of material flow analysis (MFA) and

closely related methods including ecological rucksack and material inputs per unit

service (MIPS) (Brunner and Rechburger 2003; Schmidt-Bleek 1994). In short, these

methods associate a material intensity (g material/g product) with all inputs to a product

over the production cycle. They have been applied predominantly in studies of

dematerialization of economies (Bartelmus 2003; Matthews et al. 2000; NAS 1999) and

have not been formally integrated as an impact method in life cycle assessment. The

major weakness of using MFA-derived units of mass as a common resource use

indicator for a product is the absence of differentiation of the quality of different resource

types, as well as the difference in the use of materials that may only temporarily









sequester them (e.g. cooling water) or may completely transform them rendering them

useless for future production processes (e.g. combusted fuels) (Van Der Voet et al.

2004).

Area-based measures of resource use either measure solely direct and indirect

occupation and transformation of land or go further by using equivalence factors to

relate different land use types and symbolic land uses together to measure a broader

concept of land requirements (e.g. ecological footprint). Measures of occupation and

transformation of land use are commonly employed in LCA (Guinee 2002). A measure

that combines all types of land use in a single unit based on their biological capacity is

the ecological footprint (Wackernagel et al. 2002). Ecological footprint has been more

recently integrated as a resource-use measure in the largest commercial LCA database

(Frischknecht and Jungbluth 2007). Indicators of land occupation suffer from numerous

shortcomings. Neither direct land use nor the ecological footprint measure below-

ground resource use (non-renewable), and neither incorporate the use of hydrologic

resources. Furthermore, land itself it not expected to become a limiting resource in the

future. Although the ecological footprint already shows that total direct and indirect use

of the Earth's biocapacity has been exceeded, which is referred to as an ecological

deficit (Hails et al. 2008).

Energy-based measures are potentially more comprehensive in their inclusion of

resources than land-based and material-based measures. Energy-based measures are

derived from the laws of thermodynamics, the first of which states that energy is

consumed in every transformation process. Thus every process, both independent of

and dependent on humans, involves the consumption of energy, which makes energy









an ideal common unit for tracking total resource use (Odum 2007). Some energy-based

resource use measures have already been incorporated into LCA. Energy analysis

(Boustead and Hancock 1978), known as cumulative energy demand (CED) analysis

implemented in a life cycle framework (Frischknecht and Jungbluth 2007), measures the

total heat energy (enthalpy) in fuel and other energy carrier consumed based on their

heating values. CED does not include the contribution of non-energy sources. Surplus

energy, part of the Eco-indicator 99 methodology (Goedkoop and Spriensma 2001)

estimates the difference in the amount of energy required to extract resources now

versus at a designated point in the future. Surplus energy is also limited to energy

sources.

Another thermodynamically-based indicator already integrated into LCA that

includes a broader array of resource is exergy, which may be defined as the total of

available energies of different types in a material (primarily as pressure, kinetic,

physical, chemical) in respect to their difference from reference conditions. Raw

resources have high exergy values until processed or transformed at which time some

of their exergy is lost as entropy. The exergy losses associated with transformations of

all inputs into processes in an LCA can be measured with cumulative exergy demand,

or CExD (Bosch et al. 2007). CExD is particularly valuable as a measure of the total

thermodynamic efficiency of a process where the goal is to minimize total exergy

consumption.

None of the aforementioned energy-based methods account for the energy

required by the environment to support and recreate the resource basis of economies;

they only account for energy consumed in existing resources. Thus a critical first link in









the chain of resource provision (environment to resource) is missing in how resource

use is accounting for in product life cycles. Accounting for this first link, however, is

possible using the emergy method to relate all resources on the basis of sunlight

energy. Emergy is an energy accounting metric that may be defined as the total direct

and indirect energy used to support a system measured in a common unit of energy -

conventionally sunlight equivalents (Odum 1996). The origins of all resources, both

renewable and non-renewable, can all be directly or indirectly traced back to the primary

energy driving the biosphere, sunlight, and can thus be tracked in units of energy of this

type. Thus it becomes a biophysically legitimate way of combining different forms of

resources in a common measurement unit.

Emergy evaluation is an independently developed methodology for measuring the

environmental performance of an ecosystem or human-dominated system, which has

also been applied to evaluating product systems. Emergy has been used in conjunction

with LCA as part of a comparative or multi-criteria approach (Cherubini et al. 2008;

Pizzigallo et al. 2008). Emergy has been adapted for use in economic-based input-

output LCA by Bhakshi and colleagues, who define emergy as an extension of exergy

called ecological exergy (Hau and Bakshi 2004a) and have used it as a measure of the

contribution of ecosystem processes to sectors of the US economy (Ukidwe and Bakshi

2004) and to evaluate individual products (Baral and Bakshi 2010). Nevertheless

emergy has not been integrated into traditional process LCA in such a manner that it

can be used in conjunction with conventional life cycle inventory databases and in

comparison with other LCA metrics.









A measure of the ultimate limitations that the biosphere imposes upon economic

processes must relate these processes to the energetic limits of the biosphere (Odum

2007). While such a broad concept may not highlight the scarcity of particular

resources, it does provide a sufficiently wide context through which to compare any and

all products with our planetary resource base; in doing so it can provide insight into

absolute sustainability of economic processes in the long-term. Emergy (in sunlight

energy equivalents) can be used to measure contribution of all forms of resources and

environmental processes to a product and report them with a common unit relates each

resource back to the energy consumed in its -origin, and as such is an optimal

numeraire for measuring total resource use per unit of the product. Further clarifying

the rationale for integrating emergy into LCA a measure of total resource use and

demonstrating the means of integrating emergy into a complex process-based LCA

typical of high volume products is a primary objective of this dissertation.

An implicit requirement for integrating emergy or any other impact metric into LCA

is to quantify the uncertainty in the impact model. It has been recognized among the

LCA community that the data and models used to represent complex product life cycles

potentially have a significant amount of variation and uncertainty (Fava et al. 1994).

Reporting average scores for products can at times be misleading to the degree of

accuracy occurring. Better estimation of uncertainty in these scores is a current priority

in the LCA field (Reap et al. 2008).

Uncertainty characterization should include uncertainty in model parameters,

uncertainty to represent variation among different geographic, technological, or

alternative production scenarios that may be unknown, and uncertainty built into the









actual impact models themselves (Lloyd and Ries 2007). When emergy is incorporated

into LCA as an impact model, this should therefore include the additional model

uncertainty that is added when unit emergy values (UEVs) are used to relate inputs to

processes to the emergy that was used to make them.

In the practice of emergy evaluation, emergy results are not typically presented

with uncertainty ranges. The originator of the emergy concept, H.T. Odum, believed

that an emergy result was accurate within an order of magnitude (Brown 2009) The

lack of a more clearly defined and systematic manner of characterizing the accuracy of

emergy results has been a criticism of emergy work for decades (Rydburg 2010). A

couple notable first attempts at characterizing uncertainty in specific UEVs were

performed by Campbell (2001) and Cohen (2001). Campbell estimated the uncertainty

in the transformity of global rainfall and river chemical potential based on differences in

estimated global water flows. Cohen used a stochastic simulation technique to generate

confidence envelopes for UEVs of various soil parameters. Both of these approaches

were first-order attempts for estimating ranges of specific emergy values, but did not

fully characterize this uncertainty or propose methods of propagating this uncertainty for

use in future evaluations. A model for estimating uncertainty in emergy results would be

useful for estimating ranges in emergy results within emergy and beyond for the

estimation of the additional uncertainty related to emergy models in life cycle results that

use emergy as a unit of measurement.

Applications of LCA for Non-OECD Country Products

LCA studies have predominantly been conducted on product systems located in

the United States, EU countries, Canada, Japan, and Australia and other member of the

Organization for Economic and Co-operation and Development (OECD) (Thiesen et al.









2007). As a result there has been a geographic-bias in the development of all aspects

of LCA, including product system inventories, selection of impact categories, and LCA

impact models. This bias has resulted in two primary deficiencies in LCA: (1)

production in non-OECD countries is less well-characterized resulting in lesser capacity

to use life cycle management; and (2) comparisons with OECD products has been

hindered thus limiting ability to use LCA in OECD countries that consume products from

all over the world. Unless this gap in life cycle management capacity is closed,

increasing environmental demands on producers could marginalize non-OECD country

producers with lesser capacity (Sonnemann and de Leeuw 2006). Without improved life

cycle management, the consumer demand for increasing non-OECD country products

may increase environmental damage in non-OECD countries. Expanding the scope of

LCA to incorporate more global analysis including for products from non-OECD

countries is a priority in the current phase of the UNEP-SETAC Life Cycle Initiative

(UNEP Life Cycle Initiative 2007).

Export of products to OECD countries plays a significant role in the economy of

many non-OECD countries. For those in Latin America and Africa, these exports are

largely from the primary sectors, which include fuels, agricultural products, and minerals

(Zhang et al. 2010). Mineral and agricultural sectors are both responsible for many

direct environmental impacts that are site-specific, because they generally require

significant transformation of the land and emissions occur often in a diffuse manner into

the environment surrounding the site. As a result, both mineral and agricultural

environmental impacts are less easily characterized than impacts from more enclosed









processes with less direct interaction with the local environment (more concentrated

and controlled emissions).

Accurate characterization of diffuse emissions and their impacts in mining and

agriculture depends on models that account for the local environmental factors that

influence emissions and their potency at production sites (spatial and temporal

specificity). There have been calls for greater regionalization of impact methods in both

the mining (Yellishetty et al. 2009) and agricultural sectors (Gaillard and Nemecek

2009). In mining systems, this may include the geologic work required to create a

particular deposit, if the boundary of resource use is extended to include all

environmental resources as suggested in the previous section. In agricultural systems,

regional factors effect emissions and their impacts. This is particular relevant for

emissions such as fertilizers and pesticides and the impacts they can cause including

eutrophication and ecological and human toxicity. Local factors also effect emissions

that have just recently begun to be characterized in LCA, including water loss (Pfister et

al. 2009). Regional characterization of models based on geographic difference can

have dramatic effects on LCA outcomes (Lenzen and Wachsmann 2004).

Not all relevant environmental impacts from agricultural systems have been

characterized in LCA. Two that the UNEP taskforce has identified as extremely

relevant, particularly in non-OECD countries, are biodiversity impacts and soil erosion

(Jolliet et al. 2003b). Models to estimate impacts from biodiversity are very much in

their infancy, while some have been proposed (e.g. Maia de Souza et al. 2009; Schenck

and Vickerman 2001). Erosion is the most significant cause of land degradation

globally (Gobin et al. 2003). Soil erosion has not frequently been characterized in LCA,









but universal methods for estimating soil erosion based on geographic, climatic, soil and

management factors do exist. The most commonly applied measure of soil erosion is

probably the Universal Soil Loss Equation (USLE) and its more recent developments,

the Revised Universal Soil Loss Equation (RUSLE) and most recently, RUSLE2 (Foster

et al. 2008). Soil erosion has rarely been used in LCA, and has not been customized

for use in LCA of non-OECD countries, many of which have humid tropical

environments, where heavy rain-based erosion risks can be much greater (Lal 1983).

Without a strong demand on the part of buyers or regulation imposed by

governments, there is not a strong incentive to use LCA in non-OECD countries

(Sonnemann and de Leeuw 2006). However, because of the emerging life cycle

perspective in countries where non-OECD exports are consumed, many of which are

OECD countries, the demand for use of LCA to measure environmental performance

may come from the consumers. Yet, there needs to be a standardized mechanism

through which the LCA results can be conveyed to the consumers in a way that they

can use this information to inform decision making. One solution is to present this LCA-

based environmental performance information in the form of a product label. A Type III

environmental label or environmental product declaration (EPD), as defined by ISO

14025, is designed for this purpose (ISO 2006b). EPDs are designed to convey

information on product function and production of the product, and relate this

information to environmental performance in a manner that one product can be

compared with another product in the same category. Product category rules (PCRs)

have to be specified so that results presented in EPDs are comparable. The ISO 14025

standard recommends that PCRs be based on at least one background assessment of









a product, so that the product life cycle can be characterized and relevant impacts

determined. This aspect of PCRs present a challenge for product systems in

developing countries,- because often little life cycle data and or LCA analysis of these

systems exist. Another potential barrier to use of EPDs that applies not only to non-

OECD countries was identified by Christiansen et al. (2006) and is related to the

interpretation of EPDs. These authors note that LCA data presented in EPDs are often

not readily meaningful without reference to the relative performance of other products in

the category. This shortcoming of EPDs is another important issue to address to make

LCA more relevant for non-OECD product systems.

Research Overview

Three independent studies addressing the research problems described comprise

this dissertation. The first study proposes a means to integrate emergy as a life cycle

assessment indicator to provide a measure of long-term sustainability in LCA. This

study uses the case of the Yanacocha gold mine in northern Peru. A detailed process-

based life cycle assessment is carried out to track the emergy in all direct and indirect

inputs to the mining process, including in the ore itself. Methods of associating emergy

values with inventory data and calculating results with emergy in LCA are described.

Comparisons of emergy results are made with a commonly used measure of life cycle

energy requirement, or cumulative energy demand. Following presentation of these

results, their potential value in the regional context and the broader value of emergy

results for LCA are discussed, along with remaining questions and problems with this

integration.

The problem of statistically describing the confidence of emergy results leads

directly into the research needs addressed in the second study: estimating the









uncertainty of emergy values. In this study, sources of uncertainty in emergy are

explored and the likely forms of probability distributions of different types of emergy

calculations are suggested. The description of the sources and forms of uncertainty

lead to the proposal for a model for describing uncertainty in emergy, and two

alternative procedures for estimating confidence intervals of emergy values are

described. This study proceeds with an evaluation of the accuracy of the proposed

model and proposes a means of integrating confidence intervals into the tables

commonly used to present emergy results.

The third study shifts to addressing the problems associated with broad

characterization and application of life cycle assessment for poorly characterized or

data-poor product categories in regions where existing emissions and impacts models

are not appropriate because of differences in environmental conditions. A multi-criteria

process-based LCA is conducted of fresh pineapple for export in Costa Rica (not

previously characterized with LCA), based on data from a representative sample of

pineapple producers. Existing universally-applicable emissions and inventory models

are customized to better characterize environmental impacts. An original method for

characterizing soil erosion is integrating using the RUSLE2 model. Variation and

uncertainty in inputs and emissions among the participating producers are used to

estimate the range of environmental performance in the sector for each impact

category. This LCA is furthermore designed to contribute to creating the rules for an

environmental product declaration in a manner applicable for yet uncharacterized

product categories.









CHAPTER 2
EMERGY AS AN IMPACT ASSESSMENT METHOD FOR LIFE CYCLE ASSESSMENT
PRESENTED IN A GOLD MINING CASE STUDY

Introduction

LCA is an established and widely-utilized approach to evaluating environmental

burdens associated with production activities. Emergy synthesis has been used for

similar ends, although in an emergy synthesis one tracks a single, all encompassing

environmental aspect, a measure of embodied energy (Odum 1996). While each is a

developed methodology of environmental accounting, they are not mutually exclusive.

Emergy in the LCA Context

LCA is a flexible framework that continues to grow to integrate new and revised

indicators of impact, as determined by their relevance to the LCA purpose and the

scientific validity of the indicator sets (ISO 2006c). Other thermodynamically-based

methods, such as exergy, have been integrated into LCA (Ayres et al. 1998; Bosch et

al. 2007). Emergy synthesis offers original information about the relationship between a

product or process and the environment, not captured by existing LCA indicators,

particularly relevant to resource use and long-term sustainability, which could be

valuable for LCA. However there are differences in the conventions, systems

boundaries and allocation rules between emergy and LCA, which require adjustments

from the conventional application of emergy, to achieve a consistent integration.

From the perspective of the LCA practitioner, the first questions regarding use of

emergy would be those of its utility. Why would one select emergy, in lieu of or in

addition to other indicators of environmental impact? For what purposes defined for an

LCA study would emergy be an appropriate metric? Assuming the inclusion of emergy

as an indicator, what would be necessary for its integration into the LCA framework?

29









This paper briefly describes the utility of emergy, and through a case study evaluation of

a gold mining operation at Yanacocha, Peru, presents one example of how emergy can

be used in an LCA framework. Finally, the theoretical and technical challenges posed

by integration are discussed.

In reference to the first question, these four key points provide a theoretical

justification for the use of emergy in LCA:

Emergy offers the most extensive measure of energy requirements. System

boundaries in a cradle to gate LCA typically begin with an initial unit process in

which a raw material is acquired (e.g. extraction), and would include raw materials

entering into that process, but would not include any information on the

environmental processes1 creating those raw materials. Emergy traces energy

inputs back further into the life cycle than any other thermodynamic method,

summing life cycle energy inputs using the common denominator of the solar energy

directly and indirectly driving all biosphere processes (Figure 1).2 Other

thermodynamic methods including exergy do not include energy requirements

underlying environmental processes (Ukidwe and Bakshi 2004).

Emergy approximates the work of the environment to replace what is used.

When a resource is consumed in a production process, more energy is required to





All references to 'environmental processes' and 'environmental flows' in this paper refer to solar,
geologic, and hydrologic flows that sustain both ecosystems and human-dominated systems. This is the
essence of what is meant here by 'environmental contribution'.
2
2For example, growing corn requires the solar energy necessary to support photosynthesis of the corn
plant. This includes all the solar energy falling on the corn field, not just the amount the corn used to fix
C02. Furthermore growing corn requires fossil inputs among others, all of which were originally created
with solar energy, and thus which are included in emergy analysis.









regenerate or replenish that resource. The emergy of a resource is this energy

required to make it including work of the environment, and assuming equivalent

conditions; this is the energy that is takes to replenish it. Sustainability ultimately

requires that inputs and outputs to the biosphere or its subsystems balance out

(Gallopin 2003). As the only measure that relates products to energy inputs into the

biosphere required to create them, emergy relates consumption to ultimate limits in

the biosphere, by quantifying the additional work it would require from nature to

replace the consumed resources.

Emergy presents a unified measure of resource use. Comparing the impacts of

use of biotic vs. abiotic resources, or renewable vs. non-renewable resources,

typically necessitates some sort of weighting scheme for comparison.3 Because

there is less agreement upon characterization of biotic resources, these may not be

included despite their potential relevance (Guinee 2002). Using emergy, abiotic and

biotic resources are both included and measured with the same units. As follows

from its nature as a unified indicator, one which characterizes inputs with a single

methodology to relate them with one unit (emergy uses sejs, or solar emjoules,

which are sunlight-equivalent joules), no weighting scheme is necessary to join

different forms of resources (e.g. renewable and non-renewable; fuels and minerals)

to interpret the results.

The choice of measures of impact in an LCA follow from the goal and scope of the

study (ISO 2006c). Emergy analyses have been used for a multitude of LCA-related


3
In the IMPACT 2002+, and Eco-indicator 99 methodologies, use of non-renewable resources is included
in the damage categories of resources but renewable resources are omitted (Goedkoop and Spriensma
2001; Jolliet et al. 2003a)









purposes, including to measure cumulative energy consumption (Federici et al. 2008),

to compare environmental performance of process alternatives (La Rosa et al. 2008), to

create indices for measuring sustainability (Brown and Ulgiati 1997), to quantify the

resource base of ecosystems (Tilley 2003), to measure environmental carrying capacity

(Cuadra and Bjorklund 2007) and for non-market based valuation (Odum and Odum

2000). The incorporation of emergy in LCA could potentially enhance the ability of LCA

studies to achieve these same and other purposes.

------------------------------------------------------------------------------------


^ ^ ------, -----.----------.--------
u g- Non-renewable
N 2 n O Unit process in Pr
yi resources Products
= re e life cvcle
S- Wastes
I Renewable p
resources
a ,


Proposed LCAboundary Conventional LCAboundar
Figure 2-1. Proposed boundary expansion of LCA with emergy. Driving energies
include sunlight, rain, wind, deep heat, tidal flow, etc.

This was not the first study to attempt to combine emergy and life cycle

assessment. Earlier studies focused on contrasting the two approaches (Pizzigallo et al.

2008) or extending emergy to include disposal and recycling processes (Brown and

Buranakarn 2003). The most comprehensive approaches probably include the Eco-

LCA and SUMMA models. Although referred to as ecological cumulative exergy

consumption (ECEC) rather than emergy due some slight modifications to emergy

algebra, the Eco-LCA model is an EIO-LCA model which uses emergy as an impact

indicator (Urban and Bakshi 2009). The SUMMA model is a multi-criterion analysis tool

which uses emergy as one measure of "upstream" impact which it combines with other









measures of downstream impact (Ulgiati et al. 2006). A similar multi-criteria approach

using MFA, embodied energy, exergy and emergy is used by Cherubini et al. (Cherubini

et al. 2008).

In contrast with these previous studies, this study uses a more conventional

process LCA approach through using an common industry software (SimaPro) and

attempts to integrate emergy as an indicator within that framework as specified by the

ISO 14040/44 standards, which results in adjustments to the conventional emergy

methodology. This is also the first study to use emergy in a detailed process LCA

where flows are tracked at a unit process level. Results from the study, addressed in

the discussion, reveal insights for which emergy is suggested to be a useful metric for

LCA.

A Case Study of Emergy in an LCA of Gold-Silver Bullion Production

Metals and their related mining and metallurgical processes have been a frequent

subject of LCA and other studies using approaches from industrial ecology (e.g.

Yellishetty et al. 2009 and Dubriel 2005), which is reflective of the critical dependence of

society upon metals, as well as an acknowledgement of the potential environmental

consequences of their life cycles. While these studies have addressed both

downstream and upstream impacts, including resource consumption, none have used

tools capable of connecting the product system to the environmental processes

providing for the raw resources they require (especially because they are largely

nonrenewable). An LCA is presented here of a gold-silver mining operation that uses

emergy to quantify the dependence on environmental flows. In this case study, the

primary purpose could be succinctly stated as follows:









To quantify the total environmental contribution underlyingproduction ofgold-silver. ,. ,...:,,. at the Yanacocha

mine in Peru.

Total environmental contribution includes the total work required by the environment

(biosphere) and the human-dominated systems it supports (technosphere) to provide for

that product. As impacts in LCA are categorized as resource-related (referring to

upstream impacts) or pollution-related (referring to downstream impacts) (Bare et al.

2003), environmental contribution would be categorized with the former.

The scope of this study, following from this goal, extends from the formation of the

gold deposit (representing the work of the environment) to the production of the semi-

refined dore, a bar of mixed gold and silver.5 Emergy is chosen as the measure of

environmental contribution, to be tracked over this 'cradle to gate' study, and to be the

basis of the indicator of impact of mining. Energy is commonly used in LCA to track the

total energy supplied to drive processes in an industrial life cycle. Yet the interest here

is in how much work was done in both environmental systems and human-dominated

systems to provide for it (point 2), which is not measured by just considering available

energy used by energy carriers (e.g. cumulative energy demand) or by summing all

available energy (exergy) in all the inputs (point 1). Additionally the energy from the

environment to provide for non-energy resources (materials) is part of the environmental

contribution (point 2), so all need to be tracked. However, in order to directly compare


The Yanacocha mine is one of the largest gold mines (in terms of production) in the world. The mine
produced 3.3275 million ounces in 2005 (Buenaventura Mining Company Inc. 2006). This represented
more than 40% of Peruvian production (Peruvian Ministry of Energy and Mines 2006) and approximately
3.8% of the world's gold supply in 2005, assuming 100% recovery of gold from dor6 and using the total of
2467 tonnes reported by the World Gold Council (2006).

The system and inventory are described in detail in the appendix 'Life Cycle Inventory of Gold Mined at
Yanacocha, Peru Description'.









the environmental contribution underlying each resource input together with the others

contributing to a unit process of mining operation, the contribution should be tracked

with a single indicator, for which emergy serves as this indicator here (point 3).

Using emergy allows for the introduction of more specific questions which, when

used in an LCA context, are answerable where they are traditionally not in an emergy

evaluation, which lumps all inputs into a single system process. The ability to track unit

processes from the biosphere together with unit processes in the technosphere enables

one to ask:

Is there more environmental contribution underlying the formation of the gold or

the combined mining processes?

as well the more familiar (to LCA) comparisons of inputs and unit processes in the

product system:

Which unit processes) are the most intensive in terms of environmental

contribution? Which inputs are responsible for this?

To address long-term sustainability, the activity surrounding this life cycle can be

put in context of available resources; more specifically:

How does this relate to the availability of energy driving environmental processes

in this region?

LCA results should be presented with accompanying uncertainty quantified to the

extent feasible (ISO 2006a). To fit in the LCA framework, emergy results also need to

be presented with uncertainty estimations to explain the accuracy with which

environmental contribution can be predicted.









Gold and silver are co-products, which may be mined separately and which have

independent end-uses, so comparison of this life cycle data with alternative production

routes or for end-use requires allocating environmental contribution between them, as

well as between mercury, which is naturally associated with the ore body, separated

during the refining stage and sold as a by-product.

This LCA is not comparative, because no other alternative solutions for providing

the gold are being evaluated. Nevertheless with a universal measure of impact that

does not require normalization or weighting (point 4), results can be compared with

alternative product systems for which emergy evaluation has been done, if the

boundaries and allocation rules for these alternative products are comparable, or put in

the context of other relevant emergy flows, such as those supporting ecosystems or

economic systems in the same region.

Methodology

The functional unit chosen for the study is 1 g of dore (gold-silver bullion) at the

mine gate, consisting of 43.4% gold and 56.6% silver. For comparison with other gold,

silver, and mercury products, results are also reported in relation to 1 g of gold, 1 g of

silver, and 1 g of mercury. The inventory for these products was based on the average

of annual production in 2005, the most recent year for which all necessary data were

available. Annual production was reported by one of the mine partners (Buenaventura

Mining Company Inc. 2006). The total production for this year was approximately

9.40E+046 kg of gold and 1.23E+05 kg of silver combined as gold-silver bullion, or dore.



6 "xE+y" is the form of scientific notation used throughout this document to represent "x times 10 to the y
power".










A process-based inventory was completed in accordance with the ISO 14040

series standards (ISO 2006a, 2006b) and included direct inputs from the environment

(elementary flows), capital and nondurable goods, fuels, electricity, and transportation,

along with inputs not traditionally or commonly accounted for, including the geologic

contribution to mineral formation. Nine unit processes representing process stages

were defined, and inputs were tracked by unit process (Figure 2-2). These were divided

into background processes (deposit formation, exploration, and mine infrastructure),

production processes (extraction, leaching, and processing), and auxiliary processes

(water treatment, sediment control, and reclamation). A description of the inventory

calculations and results is in the supplemental material.


Geologic BACKGROUND PRODUCTION AUXILIARY
Energy Deposit Overburden
I w Extraction Sdit Sediment
AWR Control I
I FF,WH .E T- H
I BarrenI
Ore Solution
I I


I WaterI
I Exploration Leaching at
C, IE Treatment
IF,HM, __, w I
Pregnant AWR I
Solution
I I

I n Sediment
Mine PWW I
I Infrastructure Processing Reclamation
FF,_HM, I FF,_C,i I FF,_HM _
I II
I Mercury
Dt Dore
Gold Production at Yanacocha
Study boundary Internal & product flows External input External input included for Emergy
Mining Process
Figure 2-2. Gold production system at Yanacocha with modeled flows and unit
processes. FF = fossil fuels, HM = heavy machinery, I = infrastructure, C =
chemicals, W= precipitation and pumped water, E = electricity, AWR = acid
water runoff, PWW = process wastewater.









Emergy and Energy Calculations

All inputs were converted into emergy values either via original emergy

calculations or by using previously calculated unit emergy values which relate input

flows in the inventory to emergy values (Odum 1996). An inventory cutoff for inputs

consisting of 99% of the emergy for the process was declared, to be as comprehensive

as possible without including all minor inputs. As the emergy of some inputs was not

readily estimated prior to the inventory collection, these inputs were by default included

and, even if determined to contribute less than 1% of the total emergy, were kept in the

inventory.

The geologic emergy of gold, silver, and mercury (representing the work of the

environment in the placement of mineable deposits) were estimated using the method

of Cohen et al. (2008), who proposed a new universal model for estimating emergy in

elemental metals in the ground, based on an enrichment ratio of the element, which can

be described in the form:

UEVi = ERi* 1.68E+09 sej/g (1)

where UEV is the unit emergy value (sej/g) for this element in the ground, ER is

the enrichment ratio, and i denotes a particular element. The ER can be estimated with

the following equation:

ERi = OGCi/CCi (2)

where OGC is the ore grade cutoff of element i, which is the current minimal

mineable concentration, and CC is the crustal background concentration of that

element. This model assumes that ores with greater concentrations of metals require

greater geologic work to form, without attempting to mechanistically model the diverse

and random geological processes at work, conferring a general advantage of consistent
38









and comparable emergy estimations for all mined metals. This universal method

provides average UEVs for a particular metal in the ground, but was adapted here using

the specific concentrations of gold, silver, and mercury at Yanacocha in place of the

OGC for those elements.

Original emergy calculations were necessary for a number of mining inputs,

including mine vehicles, chemicals, mine infrastructure, and transportation. When

available, data on these inputs was adapted from a commercial life cycle inventory

database, Ecoinvent v2.0 (Ecoinvent Centre 2007), and copied into a new process.

Inputs for these processes were replaced by processes carrying UEVs calculated from

previously published emergy analyses. When the processes were adapted from

Ecoinvent, emissions, infrastructure, and transportation data were not included, the

latter of which was decided to be inappropriate for the mine location and calculated

independently or estimated to be insignificant. For chemicals not available in Ecoinvent,

synthesis processes were based on stochiometry found in literature references, and

primary material inputs as well as energy sources were included. Emergy in overseas

shipping and transportation within Peru of inputs was estimated for all materials

comprising 99% of the total mass of inputs to the process.

The global baseline (estimate of emergy driving a planet and basis of all emergy

estimates) of 15.83E+24 sej/yr was used for all original UEV calculations (Odum et al.

2000) and for updates of all existing UEVs calculated in other studies. When available,

existing UEVs were incorporated without labor or services, to be consistent with the

Ecoinvent data used which do not include labor inputs to processes. For comparison

with emergy values, primary energy was estimated by summing the total energy content









of fossil fuels and electricity consumed on site using energy values from the Cumulative

Energy Demand characterization method as implemented in SimaPro (Frischknecht and

Jungbluth 2007).

Uncertainty Modeling

Uncertainty was present at the inventory level (e.g. inputs to mining) and for the

unit emergy values (the UEVs) used to convert that data into emergy. Uncertainty data

for both direct inputs and UEV values (existing and original) were included in the life

cycle model. Quantities of direct inputs to one of the nine unit processes were assigned

a range of uncertainty based upon the same model defined for the Ecoinvent database

(Frischknecht et al. 2007). This model assumes data fit a log-normal distribution. Using

this model, the geometric variance, was estimated for each input. Calculations of

uncertainty ranges for the UEVs for inputs to the process were estimated based on a

UEV uncertainty model (Ingwersen 2010). This model produces 95% confidence

intervals for UEVs also based on a lognormal distribution, and is described in the form

of the geometric mean (median) times/divided by the geometric variance, abbreviated in

the following form:

Jgeo (X-) O2geo (3)

where Pgeo is the geometric mean or median and O2geo is the geometric variance. The

bounds of the 95% confidence interval are defined such that the lower bound is equal to

the median divided by the geometric variance, and the upper bound is the median

multiplied by the geometric variance. Original uncertainty estimations based on the

analytical method (Ingwersen 2010) were performed for gold and silver in the ground.









Allocation

Two allocation approaches were adopted: the co-product rule often used in

emergy analysis and a by-product economic allocation rule used when applicable in

LCA. The co-product rule assumes that each product, in these case gold silver, and

mercury, each require the total emergy of the mining processes for their production, and

therefore the total mining emergy is allocated to each. Economic allocation is one

method in LCA in which an environmental impact is divided among multiple products.

Economic allocation was selected here in preference to allocation by mass because it

most closely reflects the motivations of co-product metal producers (Weidema and

Norris 2002). In this case, revenue from production was used to allocate environmental

contribution, by determining the market value of the gold contained in the dore as a

percent of the total value of dore and mercury production. The resulting percentage was

used as the percentage of total mining emergy allocated to gold. The same method was

applied for silver and mercury. In both cases, geologic emergy was allocated to each

product separately, since the model used for estimating geologic emergy in the products

was element-specific.

Data Management and Tools

All inventory data was stored in SimaPro 7.1 life cycle analysis software (PRe

Consultants 2008). A new process was created for each input. Emergy was entered as

a 'substance' in the substance library, and a new unit 'sej' was defined in the unit library

and given the equivalent of 1 Joule.7 This unit was assigned to the emergy substance.

When existing UEVs were relied on (e.g. for refined oil), a 'system' process was


For purposed of functionality in SimaPro the integrity of the energy algebra was not affected.

41









created, for which emergy was the only input. A quantity of emergy in sejs was assigned

to the output that corresponded with the unit emergy value (sej/g, sej/J, etc.). For inputs

for which UEV values did not exist or were not appropriate, 'unit' processes were

created that consisted of one or more system processes or other unit processes.8 A

new impact method was defined to sum life cycle emergy of all inputs to a process. To

characterize total uncertainty (both input and UEV uncertainty) in the emergy of the

mining products, Monte Carlo simulations of 1,000 iterations were run in SimaPro for

estimates of confidence intervals of emergy in the products using both emergy co-

product and economic allocation rules.

Results

Environmental Contribution to Gold, Silver, and Mercury in the Ground

The enrichment ratio of gold was estimated as 218.8:1, based on a reported gold

concentration of 0.87 ppm (Buenaventura Mining Company Inc. 2006) and a crustal

background concentration of 4 ppb (Butterman and Amey 2005), which using Eq. 1

resulted in an unit emergy value for gold in the ground of 3.65E+11 sej/g. The silver

concentration at the mine was not reported, but was estimated based on the silver in the

product and a calculated recovery rate of gold (81.52%) to be 1.13 ppm. Using the

background concentration of 0.075 ppm (Butterman and Hilliard 2004), the enrichment

ratio of silver was estimated as 15.1:1, which resulted in an estimate of the UEV of

silver in the ground at Yanacocha to be 1.54E+10 sej/g. The emergy of mercury in the

ground was estimated to be 1.71 E11 sej/g based on concentration at the mine of 8.6

ppm (Stratus Consulting 2003) and a crustal background concentration of .085 ppm

8
'Unit' processes as defined here correspond to the SimaPro definition, not to the unit processes defined
earlier as one of the nine phases of mining.









(Ehrlich and Newman 2008). The total emergy in the amount of gold extracted and

transformed into dore in 2005, just including the geologic contribution to gold in the

ground, was 8.55E+18 (x+) 10.7 sej (median times or divided by the geometric variance,

as in Eq. 3).

Environmental Contribution to Dore

Table 2-1 shows the results of the total emergy in the mining products including for

the dore, the gold and silver separately, and the mercury by-product. The total emergy

in the all life cycle stages contributing to 1 g of dore was approximately 6.8E+12 sej,

with an approximate confidence interval of 6.2E+12 (x+) 2.0. Considering estimated

uncertainty both in the inventory data and in the unit emergy values, the emergy in dore

could with 95% confidence be predicted to be as low as 4.4 E+12 sej/g and as high as

1.3E+13 sej/g, representing an approximate range of a factor of two around the median

value.

As a portion of the contribution to the total emergy in the dore, the geologic

emergy in deposit formation contributes approximately 3% (Figure 2-3), but could be as

high as 7% if the highest value in the range is used. The largest contributors to the total

emergy of the dore include chemicals (42%) followed by fossil fuels (32%), and

electricity (14%). Capital goods (mine infrastructure and heavy equipment) contribute

5%.

Relative emergy contribution of inputs is not well associated with input mass

because of differences in the unit emergy values of inputs to the process. Chemicals

used in the process illustrate this difference.









Table 2-1. Summary of emergy in mine
units are in sej/g.


products based on two allocation rules. All


Mining
Geologic Mining Allocation Total 95% Confidence
Product Emergy Emergy % Emergy Interval
Emergy based on co-product allocation
Dore 1.7E+11 6.6E+12 100% 6.8E+12 4.4E+12 1.3E+13
Gold in dore 3.7E+11 1.5E+13 100% 1.6E+13 1.0E+13 2.7E+13
Silver in dore 2.5E+10 1.2E+13 100% 1.2E+13 7.5E+12 2.2E+13
Mercury 1.7E+11 2.4E+13 100% 2.4E+13 1.6E+13 4.5E+13
Emergy based on economic allocation1
Dore 1.7E+11 6.6E+12 99.90% 6.8E+12 4.4E+12 1.3E+13
Gold in dore 3.7E+11 1.5E+13 97.31% 1.5E+13 9.9E+12 2.5E+13
Silver in dore 2.5E+10 3.0E+11 2.61% 3.3E+11 2.2E+11 5.4E+11
Mercury 1.7E+11 2.0E+10 0.08% 1.9E+11 1.8E+11 2.1E+11
1 Based on 2005 Au and Ag price received of $12.69/g and $0.26/g (Buenaventura 2006); Hg market
price of $0.02/g (Metalprices.com)

A minor input by mass used in the processing stage, lead acetate, contributed more

emergy than did lime, whose mass input was 267 times greater.

Emergy by Unit Process

Breaking down the life cycle of a product into unit processes is not typically done

in emergy analysis, but is a common step of interpretation in a life cycle assessment.

Analyzing process contribution can help target where in the life cycle environmental

burdens are greatest. Figure 2-4 shows the breakdown of emergy and primary energy

by mining unit process.

The largest environmental contribution comes from the extraction process.

Extraction emergy is dominated by diesel fuel consumed by mine vehicles. The other

production processes are chemically-intensive processes. Together the production

processes represent 67% of the total emergy. Controlling for pollution to air, water and









Gold ore, geolog
3%
Explosives 3
4%


Heavy equipment
I 1%


Infrastructure-
4% A


Figure 2-3. Environmental contribution (emergy) to dore by input type.


* Emergy 0 Primaryenergy


1.2

1

0.8 :

0.6

0.4 *S

0.2

0


4~O


Figure 2-4. Emergy and primary energy in 1 g of dore by unit process. Primary energy
is depicted on a second axis which is adjusted so that emergy and primary
energy in extraction appear the same so relative contribution of each to
processes can be depicted.


,~"o rsSe


~100
a~:u
e-\









soil, which is the objective of the auxiliary processes, contribute about 30% of the total

emergy. Background processes contribute little (<4%) to the emergy in the dore.

Figure 2-4 reveals differences in the absolute and relative contributions to

processes as indicated by emergy and primary energy. First, the emergy for each

process is six orders of magnitude greater than the primary energy in each process.

Additionally the contributions of the non-extraction processes are relatively greater

when measured in emergy than when measured with primary energy. Primary energy

reveals no use of energy in the deposit formation process, and relatively less energy in

processes that are more chemically and materially intensive.

Allocation and Emergy Uncertainty

Relative emergy contribution of inputs is not well associated with input mass

because of differences in the unit emergy values of inputs to the process. Chemicals

used in the process illustrate this difference.

Table 2-1 presents the differences in the gold, silver, and mercury UEVs according

to the two different allocation rules used. Because of its high value, under the economic

allocation rule the gold product is allocated 97.3% of the emergy, which results in a

similar UEV to that calculated under the co-product scheme, where it is allocated 100%.

The big difference appears in the calculations of the UEVs for silver and mercury

(3E+11 and 1.9E11 sej/g ), since they are allocated small portions of the total emergy

(2.61% and 0.08%) This reduces the silver UEV to 2.8% of the co-product value, and

reduces the mercury UEV to only 0.8% of the co-product value.

Uncertainties in process inputs ranged based on uncertainty in the inventory data, but

primarily due to the uncertainty of the UEVs. The inputs with greatest range of UEV

values are the minerals and inorganic chemicals which are mineral based (see ranges

46









in Table 2 of supplement 1). In comparison, uncertainty o2geo values were between 1

and 1.5 for most inputs in the inventory. Figure 2-5 shows the results of the Monte

Carlo analysis of the emergy in 1 g of dore, illustrating the resulting uncertainty range

for the dore product. The distribution is right-skewed and resembles a log-normal

distribution. Overall the combined uncertainties in the inputs lead to less uncertainty in

the dore (a factor of 2) than some of the major inputs (e.g. gold in the ground with a

factor of 10).

Discussion

Usefulness of Emergy Results

A significant finding of this LCA is that the environmental contribution to the mining

process, dominated by fuels and chemicals, was estimated to be greater than that to the

formation of the gold itself. This result holds despite the large uncertainty associated

with quantification of the environmental contribution to gold in the ground. The

production of dore can also be interpreted to be process with a net emergy loss, with an

emergy yield ratio (EYR) of close to 1, since the emergy expended in making the

product (represented here by the mining processes) is greater than the emergy

embodied in the raw resource.9 This is unfavorable in comparison with fossil energy

sources and other primary sector products which generally have emergy yield ratios of

greater than 2 (Brown et al. 2009), but this provides no insight into the utility of the

resource in society, which is much different in function and lifetime than these other

products.



9
The EYR may be defined as the total emergy in a product divided by the emergy in purchased inputs
from outside the product system (Brown and Ulgiati 1997).
























=-.
2
- 0.1
0.1


m m m


__ __ __ __ __ __ __ II


1 OE+13


1 .3E+13


Emergy (s ej


Figure 2-5. Monte Carlo analysis of 1 g of dore, showing the tails and center of the 95% CI, along with the mean (dashed
line).


-









While primary energy would indicate that the energy in mining is heavily

dominated by fuel consumption during extraction, using emergy as in indicator shows

that the other more chemically- and capital-intensive processes weigh more

significantly, and therefore that reducing total environmental contribution to the process

would demand a broader look at the other processes and inputs. This is consistent with

the trends in the results that Franzese et al. (2009) found in their comparison of gross

energy and emergy in biomass.

Quantifying resource use in emergy units permits putting processes in the context

of the flows of available renewable resources. Emergy used in a process can be seen

as the liquidation of stocks of accumulated renewable energy in all the inputs to that

process. The limit of sustainability, in emergy terms, is such that total emergy used by

society be less than or equal to the emergy driving the biosphere during the same

period of time. Thus the liquidation of the stock of emergy should not be greater than

the flows of emergy. In this case, the amount of emergy in the dore (the stock)

produced by the mine in one year is equivalent to approximately one third of the emergy

in sunlight falling on the nation of Peru in one year, and one third of one percent of the

emergy in all the renewable resources available annually to Peru (Sweeney et al.

2009).10 While this does not represent a trade off for the current period (since the stock

of emergy in the dore was largely accumulated in a prior time-period) it puts the total

resource use in the process and the available flows of resources on the same scale,

which is a step towards quantifying the sustainability of production. The Peruvian

economy is driven on average by 35% percent renewable resources, but the mining

10 Sunlight on Peru = 5E+21 J = 5E+21 sej (Sweeney et al. 2009); since 1 sej = 1 J sunlight. 1.66E+21
sej in dor6 /5E+21 sej in average sunlight on Peru = 0.3.









process at Yanacocha itself is only approximately 3.5% renewable on a life cycle

emergy basis.11 This result should not come as a surprise since mining and other

resource extraction activities are largely using non-renewable energy sources to extract

non-renewable resources.

The emergy in 1 g of dore is on the order of E+12-13 sej/g. The eventual 'London

good' gold sold on the international market, which will be produced by further refining

the dore, will have a minimum emergy on the order of E+13 sej/g. This is hundreds of

times greater than that reported for products from other economic sectors, such as

biomass-based products, chemicals, and plastics, which have UEVs consistent with the

global emergy base used here on the order of E+8-E+11 sej/g (Odum 1996), reflecting

the high environmental contribution underlying gold products, which is consistent with

the high market value of gold.

Emergy in LCA: Challenges

The boundary, allocation and other accounting differences between emergy and

LCA were dealt with here in a progressive manner. The system boundary was

expanded beyond traditional LCA to included flows of energy underlying the creation of

resources used as inputs to the foreground and background processes.. The inventory

to the gold mining process involved a hybridization of background data from previous

emergy analyses as well data from an LCI database. Numerous challenges remain for

a theoretically and procedurally consistent integration of emergy and LCA and are

discussed here.



This includes only the portion of direct electricity use from hydropower. Energy sources for all other
inputs are assumed to be non-renewable.









Challenges of using emergy with LCI databases and software

This study revealed some of the complexities and potential inconsistencies of

integrating emergy into LCA, particularly to be able to use emergy along with other LCIA

indicators and to be consistent in use of accounting rules. The technical integration of

emergy for the characterization of some of the processes (e.g. inventories for processes

occurring off-site) implemented here in SimaPro had the shortcoming of not being able

to comparatively measure other environmental aspects from background processes in

the life cycle. For some of these inputs for which emergy evaluations already existed

(e.g. for stainless steel used in mine infrastructure and vehicles) emergy was the only

input to the item, which made computation of other full life cycle indicators for resources

use (e.g. cumulative exergy demand) impossible. A better method of integrating

emergy into a Life Cycle Inventory would be to associate emergy with substances, and

then to allow the software to track the emergy through all the processes, rather than

creating processes that store unit emergy values. Such a method would permit more

accurate cross-comparison of emergy with other impact indicators.

Emergy evaluation conventionally incorporates the emergy embodied in human

labor and services (Odum 1996). Adding labor as an input may be present in some

forms in traditional LCA, such as in worker transportation (O'Brien et al. 2006), but

energy in labor has largely been left out and its inclusion represents a potential addition

to LCA from the emergy field. However, inclusion of labor, as in a typical emergy

evaluation, is not included in processes in existing LCI databases including Ecoinvent

2.0. For this reason labor was not included here. 'Services' is the conventional means

by which the labor of background processes is included in an emergy analysis.

'Services' is the emergy in the dollars paid for process inputs, estimated using a

51









emergy:money ratio to represent the average emergy behind a unit of money, and

represents labor in background processes based on the assumption that money paid for

goods and services eventually goes back to pay for the cost of human labor, since

money never returns to the natural resources themselves (Odum 1996). Unit emergy

values are often reported as "with labor and services" or "without labor and services".

For consistent incorporation of emergy in labor in an LCA, labor would also need to be

incorporated into the background processes drawn from LCI databases. Unless

background processes can be "retro-fitted" with labor estimations, unit emergy values

used for LCA should be those "without labor and services." This will however result in

the omission of an input which is considered to be integral to holistic accounting in

emergy theory, since all technosphere products rely on human input.

Reconciling rules for allocation is another necessary step for inclusion of emergy

in LCA. In the LCA context, the emergy co-product allocation would be inconsistent and

non-additive, because the emergy in the products would be double-counted when they

become inputs in the same system (which can be as large as the global economy).

Thus results based on this allocation rule should be recalculated using an allocation rule

that divides up emergy before being used with existing LCIA calculation routines, to

avoid the potential double-counting of emergy.12 Allocation rules or alternatives to

allocation typically used in LCA can easily be applied to allocate emergy among by-

products and co-products, as was demonstrated here, but if existing UEVs for co-




12
1Emergy practioners also point out that emergy of co-products cannot be double-counted when they are
inputs to the same system. See p. 1967 of (Sciubba and Ulgiati 2005). However in LCA all impacts have
to be split according to one of the methods described in ISO 14044.









products are incorporated they will have to be recalculated with the chosen allocation

rule before incorporation.

Allocation is not just an issue among co-products but also an issue related to end-

of-life of many of the materials used. While many of the inputs to dore were

transformed in such a way that they were completely consumed (e.g. the refined oil is

combusted), others, particularly the gold itself, was not consumed in such as manner.

Gold is a material that can theoretically be infinitely recycled and is not generally

consumed in its common uses (e.g. jewelry). In emergy evaluation of recycled

products, the amount of emergy that goes into the formation of the resource would be

retained (i.e. deposit formation) for the materials each time its recycled (Brown and

Buranakarn 2003). In contrast, it has been traditional practice for systems with open

loop recycling, (like the metals industry) to split the total environmental impact between

the number of distinct uses of a material (Gloria 2009). If this approach were used it

would require splitting the emergy of resource formation as well as the emergy of mining

among the anticipated number of lifetime uses of the gold product. But allocation in

systems with recycle loops is an unresolved issue in LCA especially for products such a

metals and minerals and the problem is not limited to the context of integrating emergy

into LCA (Yellishetty et al. 2009).

Energy in environmental support not conventionally included in emergy
evaluation

While more thorough than other resource use indicators in consideration of energy

use from the environment, not all the energy required by the environment to support the

dore product is included here. Geologic emergy in the clay and gravel used as a base

layer for roads and the leach pads is not included, under the assumption that these









materials are not consumed in the process. Additionally, there are waste flows from the

mine, some of which, such as those potentially emanating from the process sludge and

residuals on the leach pads, may occur over a long period of time following mine

closure. These and contemporary emissions to air, water, and soil require energy to

absorb, but these are not quantified here, as they are not typically quantified in emergy

analysis. Other measures to quantify damage in this waste, though they may not be

numerically consistent with the analysis here, could fill in the information gap, although

unless they are consistent with emergy units and methods, they will not allow for a

single measure of impact. Traditional measures of impact used in LCA, such as global

warming potential and freshwater aquatic ecotoxicity potential (Guinee 2002), could

serve this purpose. More investigation needs to be done to relate emergy with other

environmental impact metrics within the LCA framework. The outcome of emergy and

other LCA metrics may not warrant the same management action, esp. those LCA

metrics that measure waste flows, as they are measures of effects on environmental

sinks instead of use of sources.

Uncertainty in unit emergy values

Emergy from geologic processes in scarce minerals is characterized by a high

degree of uncertainty (around a factor of 10) relative to other products largely due to the

differences in different models used to estimate emergy in minerals (Ingwersen 2010).

However there is limited analysis of uncertainty in emergy values .The largely

unquantified uncertainty associated with UEV values needs to be addressed so that use

of emergy in LCA attributes appropriate uncertainty not just to inventory data, but also to

previous UEVs. The uncertainty of UEVs contributing 90% of the emergy was

characterized in this paper using a method proposed in Ingwersen (2010). Using a

54









model to estimate UEV uncertainty to couple with inventory uncertainty will help to

better quantify uncertainty in LCA studies that use emergy, which will permit statistically-

robust comparison of emergy in products that serve the same function (e.g. comparative

LCA).

Emergy and Other Resource Use Indicators

As integrated into LCA in this analysis, emergy is suggested as one measure of

resource use, defined as environmental contribution. Although primary energy use was

the only other resource use metric that was quantitatively compared with emergy in this

study, it would be useful to see how emergy compares with other implemented and

proposed indicators of resource use in LCA, namely indicators of abiotic resource

depletion, direct material input and cumulative energy demand and cumulative exergy

demand.

Indicators of resource depletion are commonly used in LCA to represent how

much of a particular resource is consumed in reference to its availability.13 These are

resource specific indicators and depend upon information on total reserves of various

resources, which is not readily available. Emergy is not often applied to assess

reserves and it is not resource-specific. Use of emergy as proposed here is therefore

not closely comparable with indicators of resource depletion, which in cases of resource

scarcity, convey very useful information on informing material selection.

Direct material input has been used as an indicator, particularly in the mining

sector (see Giljum 2004). However it has also been argued to be of limited utility,

primarily because it doesn't account for quality differences among resources and also

13
1Resource depletion indicators are build into the most common LCIA methodologies including TRACI
and Eco-indicator 99 (Bare et al. 2003; Goodkoep and Springsma 2001).









includes resources that are not transformed or consumed in processes (like

overburden) (Gossling-Reisemann 2008b). Emergy does take into account resource

quality based on a principle that more embodied energy in creating a resource

represents higher quality (Odum 1988).

Of the resource use indicators, emergy is seen by some as closely related with

exergy (Bastianoni et al. 2007; Hau and Bakshi 2004a). This is in fact only the case

when conventional exergy analysis is expanded to include available energy in inputs

from driving energies in the environment (Figure 2-1). Otherwise the boundaries for

exergy consumption are like those in conventional LCA, and still do not account for the

energy driving environmental processes. Cumulative exergy consumption or a similar

metric, entropy production (Gossling-Reisemann 2008a), are useful measures of

efficient use of the available energy embodied in resources, and thus relative measures

of thermodynamic efficiency of systems, or ultimate measures of the depletion of a the

utility of resources in the process of providing a product or service (Bosch et al. 2007).

Because of the similarity between exergy and emergy, one might expect redundant

results by using both exergy-based indicators and emergy-based indicators. However, a

brief comparison of the result of applying the Cumulative Exergy Demand (CExD)

indicator to a product from the Ecoinvent database 'Gold, from combined gold-silver

production, at refinery/PE U'14 to the emergy results here show some significant

differences in the sources of exergy contribution in comparison with emergy

contribution. Approximately 72% of the exergy in this product comes from electricity

production and 22% from the gold ore in the ground. In comparison with the results from

14A detailed comparison between an inventory of this product with the inventory of Gold at Yanacocha is
presented in the discussion of Supplement 2.









this study (Figure 2-2), emergy shows a much higher relative role of the fuels and

chemicals used in the process15. This can be largely explained by the differences in the

information that emergy and exergy provide. Exergy and entropy production more

precisely measure embodied energy consumption whereas emergy is a measure of

energy throughput and could be better described as measuring use than consumption

(Gossling-Reisemann 2008b). Also exergy describes the available energy in

substances (including the chemical energy in minerals), which is not the same as the

amount of energy used directly and indirectly in their creation in the environment. In

summary, the use of emergy provides unique information regarding resource use that

does not make other resource use indicators like exergy irrelevant, but rather can

augment the understanding of resource use by tailoring their use to address questions

at different scales (Ulgiati et al. 2006). However, emergy is the only one of these

measures that relates resources used in product life cycles back the process in the

environment necessary to replace those resources, and hence the best potential

measure of the long-term environmental sustainability of production.















1This implementation of CExD in SimaPro is incomplete and does not provide characterization factors
for many of the chemicals used in the refining processes. The relative exergy contribution of chemicals to
total exergy in gold would likely be higher if this were the case.









CHAPTER 3
UNCERTAINTY CHARACTERIZATION FOR EMERGY VALUES16

Introduction

Emergy, a measure of energy used in making a product extending back to the

work of nature in generating the raw resources used (Odum 1996), arises from general

systems theory and has been applied to ecosystems as well as to human-dominated

systems to address scientific questions at many levels, from the understanding

ecosystem dynamics (Brown et al. 2006) to studies of modern urban metabolism and

sustainability (Zhang et al. 2009). Emergy, or one any the many indicators derived from

it (Brown and Ulgiati 1997), is not an empirical property of an object, but an estimation

of embodied energy based on a relevant collection of empirical data from the systems

underlying an object, as well as rules and theoretical assumptions, and therefore cannot

be directly measured. In the process of emergy evaluation, especially due to its

extensive and ambitious scope, the emergy in a object is estimated in the presence of

numerical uncertainty, which arises in all steps and from all sources used in the

evaluation process.

The proximate motivation for development of this model was for use of emergy

as an indicator within a life cycle assessment (LCA) to provide information regarding the

energy appropriated from the environment during the life cycle of a product. The

advantages of using emergy in an LCA framework are delineated and demonstrated

through an example of a gold mining (Ingwersen Accepted). The incorporation of





16 Reprint with permission from the publisher of Ingwersen, W. W. 2010. Uncertainty characterization for
emergy values. Ecological Modelling 221(3): 445-452.









uncertainty in LCA results is commonplace and furthermore prerequisite to using results

to make comparative assertions that are disclosed to the public (ISO 2006a).

But the utility of uncertainty values for emergy is not only restricted to emergy

used along with other environmental assessment methodologies; uncertainty

characterization of emergy values has been of increasing interest and in some cases

begun to be described by emergy practitioners (Bastianoni et al. 2009) for use in

traditional emergy evaluations. Herein lies the ultimate motivation for this manuscript,

which is to provide an initial framework for characterization of uncertainty of unit emergy

values (UEVs), or inventory unit-to-emergy conversions, which can be applied or

improved upon to characterize UEVs for any application, whether they be original

emergy calculations or drawn upon from previous evaluations.

Sources of Uncertainty in UEVs

Uncertainty in UEVs may exist on numerous levels. Classification of uncertainty

is helpful for identification of these sources of uncertainty, and for formal description of

uncertainty in a replicable fashion. The classification scheme defined by the US EPA

defines three uncertainty types: parameter, scenario, and model uncertainty (Lloyd and

Reis, 2007). This scheme is co-opted here to represent the uncertainty types associated

with UEVs. These uncertainty types are defined in Table 3-1 using the example of the

UEV for lead in the ground.

There are additional elements of uncertainty in the adoption of UEVs from

previous analyses. These occur due to the following:

Incorporation of UEVs from sources without documented methods
Errors in use of significant figures
Inclusion of UEVs with different inventory items (e.g. with or without labor &
services)









Calculation errors in the evaluation
Conflicts in global baseline underlying UEVs, which may be propagated
unwittingly
Use of a UEV for an inappropriate product or process

These bulleted errors are due to random calculation error, human error, and

methodological discrepancy, which is not well-suited to formal characterization, and can

be better addressed with more transparent and uniform methodology and critical review.

But uncertainty and variability in parameters, models, and scenarios can theoretically be

quantified.

Table 3-1. Elements of uncertainty in the UEV of lead in the ground.
Uncertainty Definition Example Explanation
Type
Parameter Uncertainty in a Flux of continental Global average
parameter used in the crust = .0024 cm/yr number. A more
model recent number is
.003cm/yr (Scholl and
Huene 2004)

Model Uncertainty regarding See model for minerals Variation exists
which model used to in Table 2 between this model
make estimations is and others proposed
appropriate for minerals

Scenario Uncertainty regarding Variation in enrichment Assumption that the
the fit of model ratio based on deposit emergy in all minerals
parameters to a given type of a given form is
geographical, temporal, equal
or technological context

Models for Describing Uncertainty in Lognormal Distributions

Different components of uncertainty in a model must be combined to estimate

total uncertainty in the result. These component uncertainties may originate from

uncertainty in model parameters. In multiple parameter models, such as emergy formula

models, each parameter has its own characteristic uncertainty. Uncertainty in

environmental variables is often assumed to be normal, although Limpert et al. (2001)

presents evidence that lognormal distributions are more versatile in application and may









be more appropriate for parameters in many environmental disciplines. This distribution

is increasingly used to characterize data on process inputs used in life cycle

assessments (Frischknecht et al. 2007; Huijbregts et al. 2003a).

A spread of lognormal variable can be described by a factor that relates the

median value to the tails of its distribution. Slob (1994) defines this value as the

dispersion factor, k, but it is also known as the geometric variance, a2geo:

U2geo of a =8 6 'ln (1)
2
a=1 + ") (2)

where a2geo for variable a is a function of Wa (Eq. (1)),17 which a simple transformation of

the coefficient of variation (Eq. (2)), 18 where oa is the sample standard deviation of

variable a and Pa is the sample mean. This can be applied to positive, normal variables

with certain advantages, because parameters for describing lognormal distributions

result in positive confidence intervals, and the lognormal distribution approximates the

normal distribution with low dispersion factor values.

The geometric variance, o2geo, (k = O2geo) is a symmetrical measure of the spread

between the median, also known as the geometric mean, Pgeo,, and the tails of the

95.5% (henceforth 95%) confidence interval (Eq. (3)).

C/95 = Pgeo (X) Ogeo2 (3)

The symbol '(x+)' represents 'times or divided by'. The geometric mean for

variable a may be defined as in the following expression (Eq. (4)):



17 Eq. (1) adapted from Slob (1994).
18 Eqs. (2)-(4) adapted from Limpert et al. (2001).










Pgeo (4)


The confidence interval describes the uncertainty surrounding a lognormal

variable, but not for a formula model that is a combination of multiplication or division of

each of these variables. The uncertainty of each model parameter has to be

propagated to estimate a total parameter uncertainty. This can be done with Eq. (5):

2 2 2
S2geo of model = e (5)

where a, b ... z are references to parameters of a multiplicative model y of the form

y = a ...z Note that parameter uncertainties are not simply summed together,

which would overestimate uncertainty. This solution (Eq. (5)) is valid under the

assumption that each model parameter is independent and lognormally distributed.

Describing the confidence interval requires the median, or geometric mean, as

well as the geometric variance. The geometric mean of a model can be estimated first

by estimating the model CV (Eq. (6)) and then with a variation of Eq. (4) (Eq. (7)).19



CVmodel = e (6)




Pgeo of model Lmd- (7)
i1l+CVmodel

Models for Uncertainty in UEVs

Selecting Appropriate Methods for Uncertainty Estimations

Numerous methods exist for computing unit emergy values20, but for uncertainty

estimation, it is import to distinguish between them according to a fundamental

19 Eqs. 5-7 adapted from Slob (1994)









difference in the way UEVs are calculated: the formula vs. the table-form model. The

formula model is used for estimation of emergy in raw materials, such as minerals, fossil

fuels and water sources (the UEV in Table 1 is of this form). The traditional table-form

evaluation procedure- is typically used for ecosystem products and products of human

activities. Formula models are generally multiplicative models using estimates of

various biophysical flows and storage in the biosphere as parameters. In order to

quantify variability within a formula model, such as an emergy calculation, the result

distribution needs to be known or at least predicted. Model parameters are generally

positive values multiplied to generate the UEVs. Such multiplicative formulas have

been shown to lead to results approximating a log-normal distribution (Hill and Hoist

2001; Limpert et al. 2001). Therefore it would be logical to assume that UEVs calculated

in this manner are distributed lognormally.

The model geometric mean and variance (Eqs. (5) and (7)), used in conjunction,

offer an analytical solution for estimating uncertainty for formula-type unit emergy

values, with some built in assumptions, foremost being that the model parameters have

a common lognormal distribution. For models with parameters of mixed and unknown

distributions and large coefficients a variation, a common method for estimating

uncertainty is to simulate a model distribution using a stochastic method such as Monte

Carlo, and estimate uncertainty based on the model distribution's confidence interval

(Rai and Krewski 1998). A notable drawback of a stochastic simulation method is that

the results obtained have some variability in themselves, which, however, can be

reduced by increasing the number of iterations.

20 See (Odum 1996) for procedure for calculating UEVs, which are also known as transformities when the
denominator is an energy unit, or specific emergy when the denominator is a mass unit.









Table-form UEV calculations would be more accurately described as sum

products, where UEVs of inputs contributing to the total emergy in an item of interest

are multiplied by the quantities of each input to get emergies in those inputs, and the

emergy in each input is then added together to get the total emergy in the item of

interest. This hybrid form operation is not readily amenable to an analytical solution (Rai

and Krewski 1998). In the absence of a readily-available analytical model for this type

of UEV, a Monte Carlo model may be adopted for modeling UEV uncertainty for table-

form calculations.

Figure 3-1 provides an conceptual overview of the proposed uncertainty model.

The analytical solution is used to model all quantifiable sources of uncertainty

(parameter, model, and scenario) while the Monte Carlo model is used only to estimate

total parameter uncertainty.

Modeling Procedure and Analysis

First the geometric variance and medians of five formula-type UEVs are

estimated with the analytical solution to describe the type of variability and distribution of

some commonly used UEVs, breaking down the uncertainty into the three classes

described. Parameter uncertainty for these same UEVs is then also estimated with the

stochastic model, along with two table-form UEVs. The modeling results are cross

compared. As the distribution of UEVs has not previously been described, the resulting

distributions from the stochastic model are tested to see how closely they fit traditional

lognormal and normal distributions, as well as a hybrid of the two. In the process of this

analysis a means of reporting UEV uncertainty for future incorporation and interpretation

of uncertainty is described.









Uncertainty was estimated for five formula-type UEVs: lead, iron, oil,

groundwater, and labor. These UEVs were chosen because they represent categories

of inputs from the biosphere (labor excepted) scarce and abundant minerals,

petroleum, water, and human input that form the basis of many product life cycles.

Models for calculating each UEV are presented in Table 3-2 along with their

sources. Parameter uncertainty was estimated as follows: ranges of values or multiple

values from distinct sources when available were taken from the literature for each

model parameter. The mean and sample standard deviation for each model parameter

was calculated. With this value, the uncertainty factor, w, corresponding to each

parameter was calculated with Eq. (2). The UEV parameter uncertainty was then

estimated for the combined parameter uncertainty factors with Eq. (4).

Model and/or scenario uncertainty was incorporated by estimation of separate

uncertainty factors for these types of uncertainty. When multiple models existed for a

UEV, the average and sample standard deviation of the UEVs produced by different

models were calculated. Model uncertainty was estimated for lead, iron, petroleum and

water. When models exist for UEVs which are specific to a set of conditions but for

which those conditions are unknown in the adoption of a UEV, scenario uncertainty can

be included. For instance if labor is an input in a process, but the country in which the

labor takes place is undefined, there is scenario uncertainty which includes the

variability of the emergy in the labor depending on which country it comes from. Two

scenario uncertainties were estimated for labor UEVs (one for US labor and one for

world labor) for purposes of example.










total uncertainty
model uncertainty

Alt. UEV Model
scenario uncertainty
parameters for (!) 2 e
alternative Alt. UEV Model
scenarios 2&3 af ( )

parameter
uncertainty

UEV MODEL



Scope for stochastic solution
<-.----------------------------- -/
Scope for analytical solution
Figure 3-1. Conceptual approach to modeling uncertainty. The parameter uncertainty
consists of uncertainty and variability in the parameters used to estimate the
UEV; the scenario uncertainty consists of the uncertainty arising from use of
parameter values for different geographic or technological scenarios; the
model uncertainty from different models.

Parameter along with either model or scenario uncertainty were combined for an

estimate of total uncertainty by combining the uncertainty factors for each parameter

and for scenario and/or model uncertainty according to Eq. (5). This can be summarized

as:

total uncertainty = parameter uncertainty + model uncertainty + scenario uncertainty (8)

In order to compare the consistency of the analytical solution for the median and

geometric variance with the confidence interval generated by the simulation, stochastic

simulation models for the lead, iron, water, and labor UEV calculations were run. A

Monte Carlo simulation was scripted in R 2.6.2 statistical software (R Development

Core Team 2008) to calculate each UEV 100 times using a randomly selected set of









Unit emergy value models used for parameter uncertainty calculations.


Category Model Source
Minerals UEVminerai = Enrichment Ratio Land Cycle UEV, sej/g Cohen et al. 2008
Enrichment Ratio = (ore grade cutoff, %)/(crustal
concentration, ppm)/(1 E6)
Land Cycle, sej/g = (Emergy base, 15.83 E24 sej/yr) / Odum 1996
(crustal turnover, cm/yr)(density of crust, g/cm3) (crustal
area, cm2)

Petroleum UEVoil, sej/J = (1.68b emergy of kerogen, sej/J)(C content, Bastianoni et al.
%)/((Conversion of kerogen to petroleum, fraction)*(Enthalpy 2000
of petroleum, 4.19E4 J/g))
UEVcarbon in kerogen, sej/g = (emergy of C in phytoplankton,
sej/g)/conversion to kerogen, fraction
UEVcarbon in phytoplankton, sej/g = (phytoplanton UEV,
sej/J)*(Phytoplankton Gibbs Energy, 1.78E4 J/g)/
(phytoplanton fraction C)

Groundwater UEVgroundwater, sej/g = (Emergy base, 15.83E24 Buenfil 2001
sej/yr)/(Annual flux, g/yr)
Annual flux, g/yr = ((Precip on land, mm/yr)/(1E6
mm/km))*(Land area, km2)*(infiltration rate, %)*(1E12
L/km3)(1000 g/L)

Labor Total annual emergy use model. Odum 1996
UEVIabor, sej/J = ((Emergy use)c/(Population)*(Per capital
calorie intake, kcal/day)(365 days/yr)(4184 J/kcal))

a Omitted when concentration is reported in %
b Included for conversion from global emergy baseline of 9.44E24 to
15.83E24 sej/yr
c Emegy use for global estimate was 1.61E26 sej/yr, or a total emergy use of the world's nations
(Cohen et al. 2008)


parameters. Randomized parameters were created with a random function using the

sample standard deviation and means of each parameter. The parameters were

assumed to be log-normally distributed.

The mean and standard deviations of the log-form of each parameter were used to

create variables with a lognormal distribution, for which the following equations (Eqs. (9)

and (10)) were used (Atchinson and Brown, 1957):


Table 3-2.










logUEV Vnl jUEV (9)

logUE = In (UEV) O.5(UogUE) (10)

The resulting set of UEV approximations (100) provide a distribution from which

the left and right sides of the confidence interval can be estimated by the 2.5 and 97.5

percentile values, respectively. In order to get a representative sample, this procedure

was executed 100 times thus generating 100 distributions (for a total of 10 000 UEV

values). From each distribution, the mean, median, and standard deviation values were

reported, and these values were averaged across the 100 distributions to arrive at

average values for each UEV. From the average mean and standard deviation, the O2geo

value for that UEV was estimated according to Eq. (1).

The stochastic simulation did not incorporate the model and scenario uncertainty

components, which could only be estimated by way of the analytical solution. The

stochastic simulation recalculates the UEV by varying the parameters, but does not

incorporate uncertainty from use of alternative models or on account of parameters from

other scenarios. Thus to compare the stochastic and analytically-derived results from

parameter uncertainty, the calculated parameter O2geo (Eq. (5)) may be compared with

the O2geo value obtained from the simulation distributions.

Uncertainty was also estimated for two UEVs calculated with the table-form

model -- electricity from oil and sulfuric acid made from secondary sulfur. The emergy

tables used to estimate these two UEVs were simplified to include only items that

contributed in total to 99% of the emergy in these items.21 Uncertainty was estimated

solely with the Monte Carlo simulation routine used for the formula UEVs, with the

21 The table for electricity from oil was adapted from Brown and Ulgiati (2002)









following change: uncertainty data in the form of a2geo values for both inventory values

(e.g. secondary sulfur in g in Table 4) and their respective UEVs (e.g. UEV for

secondary sulfur in sej/g) were used in conjunction with their means to create random

lognormal variables for use in the simulation. Estimation of the natural log-form of the

standard deviation for these variables for generating lognormal random values was

slightly different than for the formula UEV case, because it used the a2geo value instead

of the sample standard deviation (Eq. (11)).

In2gteo
OlogUEV 1 (11)

The uncertainty factors in the Ecoinvent Unit Processes library for geometric

variance were used for the O2geo values for the inventory data (Ecoinvent Centre, 2007).

For the UEVs of the inventory items, the deterministic mean and the geometric variance

of the UEV for the same item calculated with the formula model were used when

appropriate as the mean and a2geo value, respectively. This choice was based on the

assumption that the inventory items (e.g. water to make sulfuric acid) had the same

UEV as those calculated with formula UEV models (e.g. groundwater).

The 95% confidence interval of the simulation distributions for formula and the

table-form UEVs were compared with the confidence intervals predicted by a perfect

log-normal distribution (I geo (x+) O2geo), those predicted by a normal-lognormal hybrid

distribution using the arithmetic mean as the center parameter ( (x+) O2geo), and those

predicted by a normal distribution (p 1.960). Eqs. (1) (3) were used to estimate the

iIgeo and O2geo from the p and a derived from the sample distribution. The percent

difference between the predicted and model distribution tails was calculated to measure

the how accurately the predicted distributions represented the model distribution.









Results

The details of the uncertainty calculations for lead are shown in Table 3-3. For

lead, parameter and model uncertainty were estimated. The O2geo values (approximately

the upper tail of the distribution divided by the median) for the five parameters range

from 1.03 to 2.25. The total parameter uncertainty (O2geo) is larger than the largest

individual parameter a2geo value, but less than the sum of these parameter O2geo values.

The total uncertainty for lead, consisting of the combined model and parameter

uncertainty (without scenario uncertainty) is dominated by the model uncertainty, which

has a large a2geo value due to large differences in previously published estimates used

for the UEV of lead. The 95% confidence interval for the lead UEV using this analytical

form of estimation would vary across three orders of magnitude, from 4.38E+11 sej/g to

5.38E+13 sej/g. However, if the UEV model used to estimate the mean was the only

acceptable model, the interval would shrink to 1.87E+12 1.26E+13, indicating

considerably less uncertainty.

The geometric variance calculations from the analytical solution for the formula

UEVs (lead, iron, crude oil, groundwater, and labor) showed a wide range of values

presented in Table 3-5. Geometric variance values were dominated by model or

scenario variances in the cases of the minerals and labor. The total parameter

uncertainty ranged from 1.08 for labor to 3.59 for crude oil, whereas model uncertainty

was as high as 9.12 for lead. The confidence intervals estimated from the analytical

and stochastic methods were of similar breadth (for all five formula UEVs), although

they were not identical the intervals from the analytical solution were all shifted slightly

to the left.









Table 3-3. Analytical uncertainty estimation for lead UEV, in ground.
No. Parameters Ip O 02geo
1 crustal concentration (ppm) 1.50E+01 1.41 1.20
2 ore grade (fraction) 0.06 0.03 2.25
3 crustal turnover (cm/yr) 2.88E-03 6.77E-04 1.58
4 density of crust (g/cm3) 2.72 0.04 1.03
5 crustal area (cm2) 1.48E+18 2.1E+16 1.03
Models
6 Alternate Model UEVs 4.52E+11 7.25E+11 9.12
Summary
Unit emergy value, p (sej/g) 5.46E+12
Parameter Uncertainty Range (No. 1-5), lgeo (sej/g) (x-) 4.85E+12
0 qeo (X+) 2.59
Total Uncertainty Range (No. 1-6), Peo (sej/g) (x+) 0 2,eo 2.57E+12 (x+) 11.09
Sources
1 Odum (1996); Thornton and Brush (2001)
2 Gabby (2007)
3 Odum (1996); (Scholl and Huene 2004)
4 Australian Museum (2007); Odum (1996)
5 UNSTAT (2006); Taylor and McLennan (1985); Odum (1996)
6 ER method and Abundance-Price Methods (Cohen et al. 2008)

The Monte Carlo simulation of the UEVs produced largely right-skewed

distributions, as indicated by the means for UEVs (see column 3 of Table 5) being less

than the medians. Without exception the means of the simulated UEV distributions

were less than the medians.

The table-form UEV calculation for sulfuric acid appears in Table 3-4. The

geometric variance values for the inputs of secondary sulfur and diesel are those

calculated for oil in the ground22; the UEV for diesel is that calculated for oil; the UEV for

electricity from oil was calculated from an emergy table and the geometric variance is

the a2geo value from the Monte Carlo simulation; and the UEV and geometric variance

for water are those calculated above for groundwater. The Monte Carlo simulation


22 Assuming the geometric variance is the same because they share similar UEV models, which is an
assumption mentioned later in the discussion.










resulted in a median of 6.51 E7 and a O2geo value of 1.75, which, in comparison with the

formula UEVs, indicates less of a spread in the distribution for this UEV. The other

table-form UEV, electricity, also had a o2geo value less than that of its major input, crude

oil, suggesting a pattern of less breadth in the confidence intervals of table-form UEVs

than those of their most variable input.

Table 3-4. Emergy summary with uncertainty of 1 kg of sulfuric acid.a
Relative Relative
Data UEV Solar
Data Uncertainty UEV Uncertainty Emergy
No Item (units) Unit U2geo (sej/unit) geo (sej)
Secondary
1 sulfur 2.14E+02 g 1.32 5.20E+09 3.59 1.11E+12
2 Diesel 3.41E+03 J 1.34 1.21E+05 3.59 4.13E+08
3 Electricity 6.30E+04 J 1.34 3.71E+05 2.77 2.34E+10
4 Water 2.41E+05 J 1.23 1.90E+05 1.95 4.57E+10
Product
5 Sulfuric acid 1.00E+03 g 1.18E+09 3.31 1.18E+12
b 195= 8.10E+08 (x+) 3.31
Notes:
1. UEV for secondary sulfur and diesel from Hopper (2008). Uses k-value for oil since secondary
sulfur is a petroleum by-product.
4. UEV in sej/J = (UEV for global groundwater, 9.36E5 sej/g)/(4.94 J/g)
Footnotes:
a Inventory data from Ecoinvent 2.0 (Ecoinvent Centre 2007)
b Example of incorporation of a confidence interval into an emergy table assuming a lognormal
distribution.

Table 3-6 summarizes the results of the Monte Carlo simulations for all UEVs

when the parameter distributions were assumed lognormal, and compares the resulting

confidence intervals against those that would be predicted by lognormal, hybrid, and

normal distributions. A number of notable differences are present between these

results and those of the calculated uncertainty values for formula UEVs in Table 3-5.

The UEV means from the simulation are higher in all cases than the deterministic

means presented in Table 3-5, but the simulation median values are lower than the









deterministic means. The O2geo values from the simulation, which were calculated

according to Eq. (1) from the average mean and standard deviations of the Monte Carlo

distributions, are not identical to the parameter geometric variance values from Table

3-5; however, the Monte Carlo a2geo values were always 5% of the analytically

calculated geometric variances.

The lognormal confidence interval was the best fit for the simulated UEV

distributions: error of the lognormal approximation of either the lower or upper tail was

never larger than 5%. However this distribution tended to consistently overestimate the

confidence interval.23 The hybrid distribution tended to predict a distribution shifted to

the right of the model with increased error, and the normal distribution often predicted a

lower tail many orders of magnitude less than the model value. The smaller the

standard deviation relative to the mean (reflected by the a2geo value), the better all

predicted distributions fit the model interval. In the case of the two table-form UEVs,

electricity from oil and sulfuric acid, the lognormal confidence interval tended to

underpredict the model lower tail more severely (suggesting that the tail is closer to the

mean), but was still the best fit when considering the combined error in both tails. The

left tail of these model UEV distributions was more constricted, and in these cases the

quotient of the model mean and o2geo value, reflected by the hybrid model, was a closer

approximate of the lower tail.







23 This could be in part be explained by the fact that the equation (3) is more precisely for a 95.5%
confidence, rather than a 95.0%, confidence interval (Limpert et al. 2001).









Discussion and Conclusions

How Much Uncertainty is in a UEV and Can it Be Quantified?

To fully characterize uncertainty for UEVs, the sources of uncertainty need to be

identified and quantified. The classification scheme introduced by the EPA provides a

useful framework which helps in identification of quantifiable aspects of uncertainty.

However in practice, describing the uncertainty in parameters, scenarios and models

requires significant effort and must draw from previous applications of various models

and across various scenarios. In this manuscript, the data sufficient to characterize

these three types of uncertainty for each UEV was not readily available, and as a result

in no cases has a total parameter uncertainty been estimated that includes all

parameter, model, and scenario uncertainty for lack of either multiple models or

modeled scenarios from which to include that component of uncertainty. Unless one or

more of these types of uncertainty can be categorically determined to be absent for a

UEV, the uncertainty measures presented here underestimate the total uncertainty in

these UEVs.

Acknowledging this underestimate, how much uncertainty are in unit emergy

values? Parameters for describing the uncertainty ranges inherit in 7 UEVs have been

presented and analyzed here. Informally, emergy practitioners may have assumed an

implicit error range of "an order of magnitude", but this analysis reveals such a general

rule of thumb is inappropriate. As quantified here the UEVs may vary with either less or

more than one order of magnitude, but this is UEV specific. However, when UEVs have

as their basis the same underlying models, if the parameters specific to one or more of

UEVs have a similar spread, then the UEV uncertainty should be similar. Thus, as was

demonstrated here, uncertainty values for a UEV may be co-opted from an UEV









calculated with the same model (eg. minerals in the ground) with reasonable confidence

if original estimation is infeasible. Adoption of geometric variances from UEVs

calculated with the same model would provide an advantage as a reasonable estimation

of uncertainty rather than a vague or undefined measure.

Quantifying model uncertainty may have implications regarding the certainty of

comparative evaluations. Figure 3-2 shows the UEVs estimated for different types of

electricity in Brown and Ulgiati (2002) all fall within the range of confidence internal of

the UEV for oil, estimated from the mean UEV reported by the authors and the

geometric variance calculated for this electricity type in this paper (2.77), using

equations 5 and 6 to estimate the median and equation 3 to estimate the tails. Although

it appears that from this analysis the UEVs of electricity sources would be statistically

similar, this ignores the fact that many of the same UEVs are used in the inputs to these

electricity processes. Hypothetically, if the same UEVs are used as inputs to processes

being compared, relative comparisons can still be made, all of the variance due to the

UEVs of inputs is covariance. This represents a problem of applying this uncertainty

model to rank UEVs where there is strong covariance, which is not addressed here.

Comparing the Analytical and Stochastic Solutions

Multiple advantages of proceeding with an analytical solution have been listed in

the risk analysis literature. These include the ability to partition uncertainty among its

contributing factors and identify factors contributing to the greatest uncertainty in a

model (Rai and Krewski 1998) as well as the greater simplicity of calculation (Slob

1994). Further advantages suggested here in the context of UEVs are the ability to

include other sources of uncertainty which cannot be quantified in a simple Monte Carlo

analysis, and the ability to replicate the values for geometric variance.









However, because table-form UEVs are the most common form of emergy

evaluation, and the stochastic simulation method is the only method presented which is

functional for this form of unit emergy calculations, the stochastic method is likely to be

more useful to emergy practitioners.

Model and scenario uncertainty components, which were not quantified in the

Monte Carlo simulation, can be particularly significant in emergy, due to the fact that

emergy values for a product are often used across a wide breadth of scenarios,

computed with alternative models, and adopted in subsequent evaluations by other

authors without knowledge of the context in which the original UEVs were calculated.

The most desirable solution to these problems with uncertainty would be: first for model

uncertainty, to agree on the use of consistent models for a UEV type to eliminate the

discrepancy that occurs between competing models; for scenario uncertainty, to make

UEVs more scenario specific whenever possible to eliminate scenario uncertainty.

Where elimination of this model and scenario uncertainty is not possible, an alternative

would be to develop a more complex version of stochastic model that would include

estimation of model and scenario uncertainty in addition to parameter uncertainty.

Following from what is predicted mathematically, this study confirmed that

formula UEVs as multiplicative products fit a lognormal distribution better than a normal

distribution. Table-form UEVs, while they are sumproducts, also tended to be better

described by lognormal distributions than normal distributions, although the two UEVs

simulated both fits this distribution to a lesser degree than the formula UEVs. Using the

deterministic mean as the center parameter for a multiplicative confidence interval,

represented by the hybrid approach, may be a tendency of emergy practioners for









simplified description of confidence intervals, but was shown here to result in more error

than using the median, except for the estimate of the lower tail of the confidence interval

for table-form UEVs.

Conclusions

Ultimately the accuracy of UEV uncertainty measures depend upon the

representativeness of the statistics describing the model parameters. In this case a

broad but not exhaustive attempt was made to describe uncertainty and variability in the

model factors for the UEVs evaluated. For this reason, this author recommends

sources of uncertainty be further investigated and more thoroughly quantified before

they are propagated for use in future studies. The responsibility should rest with

authors to diligently seek out and to summarize the uncertainty in parameters they

adopt, and to perpetuate that uncertainty with the UEV uncertainty both to present the

uncertainty of their own work and so that it can be adopted by those that use this UEV

in the future.

By describing uncertainty associated with emergy estimates, emergy is more

likely to become adopted as a measure of cumulative resource use or for other purpose

in LCA. Description of uncertainty in parameters and across models and scenarios will

increase transparency in emergy calculations, thus answering one of the critiques which

has hindered wider adoption (Hau and Bakshi 2004b). Uncertainty descriptors, namely

the geometric variance, can be used along with inventory uncertainty data to calculate

uncertainty in estimates of total emergy in complex life cycles. It can be further be used

to compare different life cycle scenarios with greater statistical confidence. Pairing

UEVs with uncertainty data and identifying sources of uncertainty will also help emergy

practitioners understand and report the statistical confidence of their calculated emergy









values and to prioritize reduction of uncertainty as a means to improve the accuracy of

emergy values.














Table 3-5. UEV uncertainty estimated from the analytical solution.


Model Lower Upper Lower Upper
and/or UEV using UEV using UEV using UEV using
UEV UEV Parameter Parameter Scenario1 Total Total parameter parameter total total
Item Den. (sej/Den.) pgeo 0 2geo 02geo Ugeo 02geo uncertainty uncertainty uncertainty uncertainty


Lead g 5.46E+12 4.85E+12 2.59 9.12 2.57E+12 11.09
Iron g 1.06E+10 1.15E+10 2.00 6.66 7.18E+09 7.53
Crude oil J 1.21E+05 9.78E+04 3.59 1.04 9.77E+04 3.59
Groundwater g 9.36E+05 8.90E+05 1.86 1.28 8.83E+05 1.95
Labor J 6.74E+06 6.73E+06 1.08 11.43 3.11E+06 11.44
1All values represent model uncertainty, except for labor for which this is scenario uncertainty


1.87E+12
5.73E+09
2.72E+04
4.78E+05
6.26E+06


1.26E+13
2.29E+10
3.51E+05
1.66E+06
7.24E+06


4.38E+11
1.52E+09
2.72E+04
4.56E+05
5.89E+05


5.38E+13
8.63E+10
3.51E+05
1.74E+06
7.70E+07


Table 3-6. UEV Monte Carlo results and comparison of model Cl's with lognormal, hybrid, and normal confidence
intervals. 1
Monte Carlo Results Model 95% Cl Predicted 95% CIs
Lognormal Cl Hybrid Cl Normal Cl

UEV Lower Upper Lower Upper Lower Upper
Item Type2 geo Cy2geo Lower Upper error error error error error error
Lead F 5.19E+12 2.73 1.93E+12 1.38E+13 -1.5% 2.6% 12% 17% -123% -11%
Iron 1.30E+10 1.99 6.62E+09 2.53E+10 -1.8% 2.3% 4.5% 8.8% -40% -6.6%
Crude oil 1.57E+05 3.55 4.66E+04 5.44E+05 -4.5% 2.9% 18% 27% -273% -14%
Ground
H20 9.40E+05 1.92 5.06E+05 1.77E+06 -2.9% 2.4% 2.6% 8.3% -35% -5.8%
Labor 6.91E+06 1.08 6.45E+06 7.40E+06 -0.32% 0.35% -0.25% 0.42% -0.57% 0.12%
Electricity
from oil T 2.81E+05 2.77 1.16E+05 7.68E+05 -12% 2.4% 0.85% 17.3% -126% -11%
Sulfuric
Acid T 8.10E+08 3.31 2.72E+08 2.67E+09 -10% 0.50% 31% 47% -179% -96%

1
Confidence intervals defined as follows: Lognormal = Pgeo (x+) k; hybrid = p (x+) k; normal = p 1.960.
2 F = formula UEV; T = table-form UEV. UEVs are in sej/g for lead, iron, groundwater, and sulfuric acid, and sej/J for crude oil, labor, and electricity from oil..















5E+4 5E+5
sej/J
Figure 3-2. Published UEVs for electricity by source (diamonds on axis) from Brown
and Ulgiati (2002), superimposed upon a modeled range of the oil UEV, using
the geometric variance for electricity from oil (a2geo = 2.77) calculated in this
paper.









CHAPTER 4
LIFE CYCLE ASSESSMENT FOR FRESH PINEAPPLE FROM COSTA RICA -
SCOPING, IMPACT MODELING AND FARM LEVEL ASSESSMENT

Introduction

Although tropical fruits and their derivative food products make up a substantial

and increasing portion of the fruit consumption in the temperate countries of Europe and

North America24, little life cycle data or published life cycle assessments (LCA) of these

products are available. At the same time, large areas and substantial resources in

tropical countries are dedicated to growing tropical fruits, such as banana, pineapple,

and mango, primarily for export (FAO 2009). Associated local and global environmental

impacts need to be accounted for and better managed both locally and globally as these

fruits continue to grow as a proportion of temperate-climate diets. One way to

encourage better environmental management could be through LCA-based Type III

environmental product declarations (EPDs), so that quantitative environmental

information can be used to help producers make better management choices and help

buyers and consumers make informed environmental choices that take into account the

full product life cycle (Schenck 2009).

Objectives

The primary objective of this study was to conduct a background LCA of fresh

pineapple production in Costa Rica to be used as a guide for creating a product

category rule (PCR) for fresh pineapple, as specified by ISO 14025 (6.7.1 ISO 2006b).

The development of a PCR is a mandatory step toward the process of creating an EPD.

A goal of any product category rule is to enable comparative assertions of

24 Pineapple import growth (by weight) was 248% from 1996-2006 in the EU and North America while
only 56% for grapes, 33% for bananas, 27% for apples, and 14% for oranges in the same period (FAO
2009).









environmental performance between products of the same category. To create a PCR,

a background LCA can be used as a reference for establishing the environmental

impact categories and indicators for reporting, methods for conducting inventories and

estimating impacts, and calculation parameters for these inventories and impact

models. Although the objective is to create a PCR for fresh pineapple, this LCA is

scoped bearing in mind the functional use of the product, to provide nutrition through

fruit consumption, and thus is created with the wider intention of providing life cycle data

relevant to a wider range of environmental impacts of concern in fruit-product supply

chains. Impacts are estimated with methods that are as globally-valid and adaptable as

possible, to permit comparable analysis with other fruit-group food products. The LCA

should have sufficient coverage to represent the range of climatic, field, management,

and production levels so that ranges of potential impacts can be bounded with a

statistical confidence. Furthermore comparisons of environmental performance are

made between fresh pineapple and other fruits through the farm scale to provide an

initial analysis of how fresh pineapple from Costa Rica compares to production of other

fruits consumed raw or used as the basis of processed food products.

A secondary objective is to provide a model for other such background LCAs of

agricultural products, particularly for those that have yet to be performed in countries

and environments where assumptions made in emission and impacts models may not

hold and that hence require regional adaptation of these models for more accurate

impact assessment.

The Fresh Pineapple System in Costa Rica

Costa Rica is the largest provider of fresh pineapple to the EU and the US.

Approximately 85% of pineapples imported to the U.S. in 2005 were produced in Costa









Rica; in the EU 71% of fresh pineapple imports came from Costa Rica (FAO 2009).

Pineapple export has overtaken coffee to become Costa Rica's second largest

agriculture export (to bananas) in terms of international exchange. This production has

resulted in a rapid expansion of pineapple plantations in the Limon (Atlantic region),

Alajuela (North region), Heredia (North region), and Puntarenas (Pacific region)

provinces (Bach 2008). There are a number of environmental and health-related

concerns surrounding this recent expansion and the modern production process. Public

concerns include soil erosion, pesticide contamination of natural areas and water

supplies, lowering of water tables, worker exposure to agrochemicals, and impacts of

organic wastes, among others (Sandoval 2009).

Pineapples are primarily grown in three regions, hereafter referred to as the North,

Atlantic, and Pacific regions, on ultisols but also on other well-drained mineral soil

orders. Pineapples for the fresh export market in Costa Rica are a highly technical,

non-traditional cash crop. The high level of technicality has resulted in a high degree of

uniformity in production systems to meet international standards (e.g. GLOBALGAP)

and produce competitive yields and fruit quality. The variety grown almost universally

for export is the MD2, or "golden". A good description of the production process in

Costa Rica can be found in Gomez et al. (2007). Fields are prepared with adequate

drainage and raised beds. Seed materials are most often suckers (shoots from existing

plants) harvested within farms. Once established pineapples require regular fertilization

primarily through foliar application of fertilizers. Nematicides, herbicides and insecticides

are used to reduce pests and competition. Once mature (about 150 days on average)

plants are often "forced" to begin fruiting, usually by application of ethylene gas. Fruits









are ready for harvest in another six months, from where they are manually harvested

and transported to packing facilities. When plants are not left to produce a second

harvest, they are chopped and the field is prepared again for another planting.

Methods

System Boundaries and Functional Units

The LCA boundaries are the farm stage though transport to the packing facility

including all upstream processes (Figure 4-1).


|M .
lI
gro ^ Farm leri Packing Transport
Fuels -, Pineapple, attacking and -
Iuels Boxed Distbut i neapple at
Prepared s uels bu retailer
hernicals els n
nj --onsumer
IMachinery E lectncit to US/EU onsu
i Hater retailer


Figure 4-1. Fresh pineapple production unit processes and boundaries for the LCA.
The first unit process is the focus of this paper.

The primary functional unit (FU) is 1 kg of fruit delivered to the packing facility. For

comparison with other fruit products at the farm level, one serving of fruit at the packing

facility is used, because it is a more relevant unit for comparison because of its

functional equivalency. The USDA defines a serving of fruit as 1 cup of fresh fruit,

which for pineapple is 165 g (USDA 2009). In order to estimate the number of servings

that can be obtained for 1 kg of pineapple the following equation is used:

Servings/kg fresh weight fruit = (edible fraction of fruit)/(kg fruit/serving) (12)

For pineapple this results in 3.09 servings/kg fresh fruit. Life cycle inputs for all inputs of

agrochemicals and machinery and related emissions are included. Permanent farm









infrastructure (buildings and road) was judged to be environmentally insignificant and

excluded from the study.

Data Collection

A public call for producer participation in this LCA followed from a workshop

organized in San Jose, Costa Rica in July 2009 for pineapple producers, government

officials, LCA experts, and other potential stakeholders to present the concept of LCA-

based EPDs (Ingwersen et al. 2009). Participation in the LCA was anonymous to

encourage sharing of production data and evaluating environmental performance

without revealing any private producer data. Farms representing all three primary

producing regions of the country, with management schemes including conventional

and organic, and with sizes ranging from 1 to >1000 hectares were directly solicited in

order to seek a representative sample. Following agreement to participate, each

producer was sent a standardized questionnaire requesting data on historical farm area,

production inputs including fuels, fertilizers, pesticides, water use, agricultural

machinery models and use, yield, harvest schedule, distance and means of transport to

the packing facility. Data collection was supervised through in-person meetings with

producer contacts to assure common understanding of the questions for data collection.

Data were later verified through comparison of data items across the entire participant

pool to assure that input data were reasonably suited to pineapple production

requirements. To acquire site-specific data for inventory emissions models, farms were

visited and data on soils, topography, and operations were collected.

Because of the discontinuity between the non-annual production cycle and annual

data collected from producers, all annual production input data had to be adjusted with

the following equation:









Input, x/kg pineapple = (Annual input, x/yr)/(Farm area, ha)/(Harvest

kg/ha/harvest) (harvests/yr) (13)

Because of the same reasons mentioned above, yield data were collected on a per

harvest basis.

Data on all production inputs were matched with the appropriate processes in the

Ecoinvent v2.0 database (Ecoinvent Centre 2007) for inclusion in the inventory and

entered into SimaPro software (PRe Consultants 2008) after being converted into

EcoSpold XML format for validation. For pesticides reported, mass of the active

ingredient applied was determined and used as the mass of the pesticide input from

Ecoinvent of the same class (Nemecek and Kagi 2007). New processes were created

for inputs without appropriate equivalents in the Ecoinvent database by assembling their

active ingredients under a new process. N-P-K fertilizers were estimated by combining

single or double fertilizers in quantities to match the N-P-K weight ratios of the actual

fertilizers, as recommended by the Ecoinvent designers (Nemecek and Kagi 2007).

Emissions and Impact Models

Emissions and impact models were chosen based on the following criteria:

1 Universal midpoint models are used for global impacts (e.g., climate

change)

2 Regionalization of universally-applicable endpoint models are used for local

impacts of concern when available (e.g., USETox)

When appropriate characterization factors are not yet available, the measured impacts

are reported as the quantity of relevant emissions.

Recent work in the environmental evaluation of the food sector has focused

heavily on carbon footprinting, in conjunction with the development of product-level









carbon footprinting standards (Sinden 2008). Acknowledging the growing importance of

this effort, rules for carbon accounting in this LCA are set as synchronously as possible

with the PAS 2050 standard. Land transformation from forest is a potentially significant

contributor to carbon release surrounding agricultural products, especially in tropical

regions (Ebeling and Yasue 2008). Carbon loss from land transformation in kg C/ha

was estimated only when conversion from primary or secondary forest was reported.

Loss was estimated by identifying the historical Holdridge life zones that occupied the

land the farm currently occupies (Holdridge 1967) and summing the carbon in living

biomass (Helmer and Brown 2000) with the estimated soil carbon (IPCC 2007) and

dividing this carbon loss over 20 years. Emissions to air resulting from on-farm fuel

combustion were estimated based on the same fuel-specific coefficients and equations

used for agricultural data in the Ecoinvent database (Nemecek and Kagi 2007).

Estimating other emissions from farm stage processes required customization of

emissions models capable of capturing, to the extent possible, the crop and field-

specific variables that affect these emission rates. Models capable of parameterization

with site-specific inputs were used to estimate emissions of eroded soil, consumed

water, nitrogen and phosphorus in fertilizers, and active-ingredients of pesticides.

Emissions of nitrogen and phosphorus compounds to air and water are functions of

crop- and field-specific factors. Pathways considered here for N include uptake,

ammonia, dinitrogen oxide, and nitrous oxide formation and volatilization, and nitrate

leaching and runoff. Modeled pathways for P include uptake, phosphate runoff, and

loss of P bound to sediments from erosion. Uptake quantities were based on the









average N and P concentration in pineapple leaf tissue. Equations and references used

in estimating N and P emission can be found in the Appendix.

The PestLCI model (Birkveda and Hauschild 2006) was customized with site-

specific climate and soil data to quantify the fate of pesticides applied in the field to air

and water. Because drainage is present on the majority of pineapple farms, drainage

was assumed to be 100% effective in the model and thus all emissions to soil that are

either lost via direct runoff after application or after lost after leaching through the soil

column were characterized as an emission to surface water. Pesticides not present in

the default PestLCI model provided by the authors were added into the database so that

fate of all pesticides applied to the field could be characterized. Characterization was

farm-specific but application dates were unknown and thus the annual average of

climate data was used. The plant type "2", citrus, was chosen from the two plant types

available, because the thick cuticle most resembles that of pineapple (Malezieux et al.

2003). Assumed canopy cover was 75% at time of application. All other default

settings in PestLCI were maintained.

For estimating consumed water, the FAO CROPWAT model (Swennenhuis 2009)

was parameterized with site-specific climatic and soil data, and plant-specific

parameters. Actual water use from the "irrigation schedule option" was the quantity of

water reported. Irrigation water was added through the irrigation schedule for farms that

use irrigation. Farm specific climate data were taken from the FAO LocClim database

based on the geographic coordinates of the farms, and coupled with farm data on

irrigation practices from the questionnaires. Other general model assumptions and

plant-specific parameters can be found in the appendix.









Soil erosion was estimated for each farm using the most recent ARS version of the

RUSLE2 model (Foster et al. 2008), and customizing it for site-specific conditions.

RUSLE2 models rain-based erosion on overland flow paths. Not included in this model

are wind-based erosion and rain-based erosion from ditches or other concentrated flow

areas, which are less significant sources of erosion on Costa Rican pineapple farms.

Climate data required for the model were interpolated with the FAO Locclim database

from the nearest 12 weather stations, including temperature, monthly rainfall, and

number of days with rain per month (FAO 2010). R-values (rainfall intensity factors)

were adopted from maps created in an implementation of the USLE model for the

country of Costa Rica (Rubin and Hyman 2000). To parameterize the model, the

following measurements were taken in representative areas of each participating farm:

the percent slope and effective length of the slope were measured for each unique

slope in the farm segment using a clinometer and metric tape. A unique slope

consisted of a slope 2-3 % different from other slopes based on visual assessment or

with unique drainage or contouring (e.g., bed direction) elements. In each area of the

farm with a unique soil profile, the profile was described and samples were collected for

soil texture analysis (Burt 2009). Slope and soil data collected in the field were used

along with farm specific management data including production schedules and other

general data on pineapple morphology. One model was run for each unique

combination of soil, % slope, field geometry and production schedule within each farm.

Results for each farm were then averaged based on the total farm area represented by

those conditions. Erosion occurring during initial conversion of the land from previous









land use was not estimated. All general assumptions and parameters selected for the

RUSLE2 model are reported in the appendix (Table D-7).

Sensitivity analyses of the adaptations of the PestLCI, RUSLE2, FAO CROPWAT

models were conducted by selecting environmental and management scenarios

reported or assumed to exist based on expert knowledge of the sector. Analyses were

performed using the production-weighted average of sample data (described below)

and the climate variables of the North region as the default condition. Percent changes

from the default conditions were reported by sequentially varying model variables within

ranges naturally present in climate, field conditions, pineapple physiology, or ranges

reported in management and harvest schedule.

Estimating the Sector Range of Environmental Performance

In order to meet the goal of conducting an LCA representative of production in the

sector and maintaining the anonymity of producers participating in the study, a single

unit process was created from the inventories of the participating farms. This process

was used to create a distribution of environmental impacts to characterize the sector,

henceforth referred to as the sector range of environmental performance (RoEP). To

create the unit process, production-weighted average input data from the individual

farms were used as means, and parameterized with confidence intervals based on

ranges existing within and among farms, or moreover likely to exist within the sector.

For pesticide inputs and related emissions, only inputs to conventional farms were used

in the baseline because inventory data on biological control agents and their associated

environmental impacts were not available.

Each of these inventory inputs was parameterized with a standard deviation based

on the variation among the sample farms, and assumed to have a normal distribution.









A correction of uncertainty for each input had to be made to reflect the variation in yield

within and between farms. A standard deviation of yield within each farm was estimated

using the reported min, max, and mean production values. A production-weighted

combined uncertainty of the yield was estimated with a propagation of standard

uncertainty formula (NIST 2010) of the form:

C Vyield = (C V2a (PaPtotal)2 + C V2b (PbPtota 2 + .... C V2* (Pz/Ptotal 2) (14)

where CVyield is the coefficient of variation of the yield for the baseline scenario, CV2 is

the square of the coefficient of variation of the yield for a farm a, and Pa/Ptotal is the

percent of the total production of farm a from the total production of participating farms.

The uncertainty based on variation in production inputs per hectare and uncertainty

based on yield were then combined to estimate total uncertainty for each input, using

the simplified form of equation 14:

CVmod, input,i = /( C V2yield + CV2input,i,) (15)

where CVmod, input, represents the yield-modified coefficient of variation for input i. The

standard deviation used to parameterize a normal distribution for a given input, i was

then estimated by multiplying CVinput, by the sample mean value.

For the emissions inventory, log-normal distributions were assumed and extremes

from sensitivity analyses of the emissions models were assumed to represent the 2.5%

and 97.5% values of these distributions. The geometric variance (GVemission), or

measure of spread of the lognormal distribution, of the modeled emission from the

sensitivity analysis was estimated by taking the maximum positive % change from the

tested parameter values, dividing by 100% and adding 1.25 The variation based on the


25 For example, if they max percent change from the default value from the sensitivity analysis was +60%,
the estimated geometric variance = 1+60%/100% = 1.6.









sensitivity analysis was combined with variation in farm yields and in the production

input related to that emission (e.g. nitrogen fertilizers for nitrate). A variation of equation

4 for propagation of uncertainty for lognormal variables was used to combine

uncertainty from sensitivity analyses with yield uncertainty using to the follow formulas:

G Vmod,emission i = exp( /( In(G Vyield) 2 + In(GVinput, i) 2 + In(GVemission ,) 2) (16)

where GVmod,emissioni is the yield-modified GV of the emission, GV2 yield, i is again the GV

of the yield, GVinput, i is the GV of the respective input related to the emission, and

GVmod,emissioni is the GV of emission, i. For emissions related to multiple inputs, the GV

input, i used was the related input with the maximum coefficient of variation. GV for the

inputs and emissions were calculated from the coefficient of variation with the formula

(Slob 1994):

GVx = exp(1.96lIn(1+CVx2)) (17)

where GVx is either the GV of yield or input and CVx is the coefficient of variation of the

input or emission.

An exception to a production-weighted average of emissions was made for

modeling the emission of carbon dioxide potentially resulting from land-use change. For

estimation of carbon emissions, the PAS 2050 standard dictates that, for cases where

an agricultural product is from an unknown location in a country, the land use

transformation allocated to the product should be the carbon lost in conversion of the

most carbon-rich ecosystem of the country divided by the lifetime of the crop (default =

20 years) (Sinden 2008). The max potential kg C/ha loss was estimated by overlaying

the historical Holdridge life zones on current pineapple-occupied areas (Holdridge

1967), selecting the life zone with the highest storage of above ground and below-









ground carbon (Helmer and Brown 2000), adding in estimated soil carbon (IPCC 2007),

and dividing this carbon loss over 20 years. The uncertainty range of carbon loss

allocated to pineapples due to conversion from forest was then modeled with a uniform

distribution with the min equal to 0 and the max equal to the max potential carbon loss,

all in kg/ha.

Monte Carlo simulations with 1000 runs were executed in SimaPro for each impact

(described below). The final RoEP was estimated by taking the ends of the 99%

confidence intervals (0.5th and 99.5th percentiles) to represent the ends of the RoEP.

LCIA Indicators

The measures of environmental impact selected, or LCIA indicators, were chosen

both because of their precedence in existing agricultural LCA and for their

environmental relevance to both the geographically-specific human health and

environmental concerns of the regions as well as larger concerns associated with the

farm stage in production of fruit products. Characterization was done for both farm

stages and upstream processes for farm inputs (e.g., manufacture and transport of

agrochemicals to the farm). Impact categories selected were cumulative energy

demand, potential soil erosion, potential aquatic eutrophication, water footprint and

stress-weighted water footprint, human and freshwater toxicity, carbon footprint and

land use.

Soil erosion impact

Soil erosion or loss is infrequently reported as an emission and lacks a suitable

LCIA methodology to relate erosion to impacts to damage to ecosystems or human

communities. Soil erosion was one impact category with particular concern to experts

from non-OECD countries and thus recommended for further development in LCAs









studies by members of the UNEP working group on LCIA in 2003 (Jolliet et al. 2003b).

Soil loss or potential has been reported as an inventory indicator in mass of soil lost or

depleted per functional unit (Heuvelmans et al. 2005; Peters et al. 2010; Schenck 2007)

and is done as such here.

Cumulative energy demand

Energy use from non-renewable resources is often considered an indicator

appropriate for all product systems and has been shown to correlate well with other

categories of environmental impact (Huijbregts et al. 2010). Total energy life cycle use

in fuels and electricity is measured using the cumulative energy demand (CED)

indicator implemented in the Ecoinvent database (Frischknecht and Jungbluth 2007).

Only characterization of non-renewable energy from fossil sources is implemented here.

A proposed indicator (Ingwersen Accepted) based on the emergy method is potentially

a stronger indicator of resource use for agricultural systems, but, because

characterization factors were not available for the majority of the Ecoinvent processes

used in the inventory it was not applied here.

Virtual water content and stress-weighted water footprint

Freshwater consumption and its resulting impacts on water availability and quality

for ecosystems and human health is a significant environmental concern, particularly in

areas susceptible to drought or water scarcity from overuse. Food consumption is a

strong driver of water use globally (Chapagain and Hoekstra 2004). Nevertheless,

estimating freshwater consumption has only recently been developed in reference to the

water required per unit of food output, and just in the last year been integrated into LCA

as an LCIA method (Pfister et al. 2009). Here, water consumption is estimated both by

the water footprinting method (Hoekstra et al. 2009), henceforth referred to as









volumetric water footprint to reduce confusion of terms, and further extended as a

midpoint LCIA method called stress-weighted water footprint (SWWF), as described by

Ridoutt and Pfister (2010).

The volumetric water content, also known as virtual water, represents the total

consumptive water use of green water (rainwater), blue water (water stored in surface

and groundwater), and grey water (equivalent water use required to dilute polluted

water to background levels). Life cycle consumptive water use in background

processes is not included in this study for lack of appropriate background data, which

has been acknowledged as a shortcoming of existing LCI databases (Pfister et al.

2009). However, consumptive water use has thus far been shown to be heavily

dominated by agricultural processes, and upstream process are assumed not to have a

significant effects on the results. The green and blue water components in the farm

stage were estimated with the FAO CROPWAT model as described above; grey water

was estimated as the water required to dilute the nitrate emission from the farms to 10

mg/L (Hoekstra et al. 2009).

Because the effects of water use for production are very different depending on

the relationship of that use to regional water availability, the water stress index (WSI) is

applied as a characterization factor to relate use to its likelihood of depraving humans

and ecosystems of water in the region. A WSI for Costa Rica of 0.0163 calculated by

Pfister et al. (2009) as part of the creation of global characterization factors and was

applied using an equation by Ridoutt and Pfister (2010) to calculate the stress-weighted

water footprint:

SWWF = WSICR(WFproc,blue) (18)









where WFproc,blue is blue water footprint in L/kg pineapple and WSICR is the unitless

water stress index for Costa Rica. Ridoutt and Pfister (2010) also propose calculating

the SWWF by including the grey water. However, the water represented by grey water

(the water necessary for dilution) is not depriving other users of water, so it is not

included in the SWWF here.

Aquatic eutrophication

Macro-nutrient excess is a threat to both terrestrial and aquatic ecosystems,

however it is perhaps more of a threat in aquatic ecosystems. The process of

eutrophication in aquatic ecosystems (nutrient excess leading to sharp increase in

primary production and subsequent increase in microbial oxygen consumption leading

to a depletion of oxygen) is closely tied with runoff of N and P in agricultural fertilizers.

The effects of N and P nutrient influx are system-dependent, but freshwater systems are

generally P-limited and seawater, N-limited. Studies in streams on the Caribbean side

of Costa Rica have shown that P addition can have cascading ecological effects on

stream ecosystems (Rosemond et al., 2001). N escaping to the Pacific and Caribbean

estuaries is assumed here to have the same effects documented in other estuarine

environments, such as the Gulf of Mexico (Miller et al. 2006). As a result, quantification

of the effects of N and P in runoff from pineapple farms is performed here with regard to

its potential to cause eutrophication. A variation of formula has been previously used

(Gallego et al. 2010; Seppala et al. 2004) to create eutrophication characterization

factors for aquatic ecosystems:

Cfe = tfe*afe*nfe (19)

where the characterization factor for emission e is cfe (here in kg N/kg emission); ife is

the transport factor, the probability that emission e will be transported to an aquatic









environment where it will have an effect; afe is the bioavailability factor for a emission e;

nfe is the nutritive factor for emission e, which is its ability to cause eutrophication

relative to N. Because emissions to water from farms occur directly to freshwater

environments, and because land in Costa Rica is 100% exorheic (rainfall terminates in

ocean), so as for areas where this is the case in the US, as in Norris (2003), tfe is set to

1. Most of the air currents in Costa Rica move inward toward the mountains (Daly et al.

2007), with rainfall depositing airborne emissions back to the land so for emissions to air

we also set tfe to 1. Availability factors are based on the relative proportion of readily-

available inorganic forms of nutrients to organic forms in this case only emissions of

inorganic nutrients are characterized, so afe is set to 1 for all emissions. The nutritive

factors for the emissions are all based on the Redfield ratio of 116:16:1 (C:N:P) as in

Norris (2003). Because the ratio of N:P has been found to vary between 13-19 in

aquatic systems, the CV applied to each nfand propagated the final cfe is 0.09. Each cfe

is thus equivalent to the nfe since both the transport and availability factors are set to 1

here for all characterized emissions. The resulting values, especially for emissions to

air, are notably higher those in the Ecoinvent implementation of TRACI (Frischknecht

and Jungbluth 2007), which uses the average US characterization values, because they

account for transport losses assumed not to occur here.

Human and freshwater ecotoxicity

Pesticides used in pineapple farming include herbicides, insecticides, nematicides

and soil fumigants. Toxicity of these pesticides to humans and ecosystems is a function

of fate in the environment, lifetime, transport, intake and effect. Models were reviewed

that consider the fate, incidence of contact, and effect of pesticide emissions both on

ecosystems and human health. Numerous models that have been used in LCA are









available for this purpose, including USES-LCA, IMPACT 2002+, CAL-TOX, and others.

Despite their similarities in purpose and orientation, results of these models have been

shown to be widely divergent. Recognition of this divergence prompted the cooperative

development of the USEtox model (Rosenbaum et al. 2008). USEtox was therefore

selected to characterize toxicity here, in line with the intent of selecting models based

on international consensus. USEtox is, however, based on the European continent, and

the characterization factors are based on the climate, population, land use, and other

data geographically representative of Europe. Other authors have shown that

characterization scores for pesticides in multimedia fate, transport and effect models are

very sensitive to geographic variables (Huijbregts et al. 2003b), particularly soil erosion

and fraction of surface water, which are very different in Costa Rica than in the

European continent. An evaluation of sources of uncertainty in the IMPACT model

showed that the misrepresentation of geographic variables can potentially result in

errors of three orders of magnitude (Pennington et al. 2005). Thus all geographic and

demographic variables in the USEtox default model were tailored to the Costa Rican

environment, which is henceforth referred to as USEtox-CR. Results are reported in

number of disease cases for human toxicity, and potentially affected fraction of

species/m3/day for freshwater ecotoxicity.

Other indicators

The IPCC global warming potential 100-year characterization factors (IPCC 2007),

expressed in CO2-equivalents, were used as characterization factors for emissions with

a potential to cause global warming, which sum together to create the carbon footprint.

Occupation of land is described in m2/yr without impact characterization.









Results


Pineapple Sector Inventory

Pineapple field data on geographic location, topography, management and soils

were collected for areas in total representing approximately 200 ha and producing

approximately 18,000 tons pineapple/harvest or 10,000 tons/yr. Participating farms

represented all three primary production districts (North, Atlantic, Pacific) and included

both conventional and organic, respectively represented by approximately 88% and

12% by total production of the sample. Complete data on production inputs in the

questionnaires was provided for 93% of farms surveyed based on total production

volume.

The production-weight average yield among farms providing complete data was

95 36 tons/harvest with an average of 0.60 0.24 harvests/yr. The average yield

reported for the sector is 67 tons/harvest (G6mez et al. 2007). Within farm yield

variation between minimum and maximum yield/ha was up to 38 tons in one case, with

an overall minimum of 48 tons/ha and a maximum of 129 tons/ha. Inputs per kg

pineapple by category were 0.17 0.04 m2/yr of land, 0.0075 0.0030 kg fuels, 0.043 +

0.012 kg minerals in fertilizers, 7.8E-4 1.6E-4 kg pesticides and 3.3E-4 1.35E-4 kg

machinery. The inputs and standard deviations for 1 kg of pineapple at the packing

facility are presented in the Appendix.

Soil Erosion

The estimated average soil erosion for the sampled pineapple farms varied from

approximately 2.5 to 5 tons/ha/yr, which was approximately 0.05 to 0.10 kg soil/kg

pineapple. There was significant variation within individual farms with erosion estimates

for slope profiles within farms varying from less than 1 to 40 tons/ha/yr in one case,









which equated to a range of 0.05 to 0.82 kg eroded soil/kg pineapple; a maximum of 16

times the minimum that was diluted by the averaging of erosion within farms.

For the sector range of environmental performance (RoEP), the median value

was 0.02 kg eroded soil/kg pineapple with a lower confidence bound of 0.0005 and

upper bound of 0.6 kg eroded soil/kg pineapple.

The results of the sensitivity analysis show that % slope was the factor most

strongly influencing the erosion results. An increase in % slope alone from 2.5% to 30%

caused an increase in erosion in tons/ha/yr of 1680%. The sensitivity of soil texture, in

reference to percent change in erosion from the baseline (-38 to 92% of the baseline

from low to highest erodibility), along with degree of contouring of the rows (-53 to 0% of

the baseline from standard to no contouring), use of plastic mulch (-78%) and use of

double harvesting systems (-32% of the baseline) all had significant influences on the

soil erosion at the pineapple farms. Summary tables of the sensitivity analyses for the

soil erosion and other emissions inventory models can be found in the appendix.

Cumulative Energy Demand (CED) of Pineapple

The RoEP for life cycle cumulative non-renewable energy demand of pineapple

was 1.2 to 2.2 MJ/kg with a median value of 1.5 MJ/kg. Most of this energy is used to

make production inputs (77%), particularly fertilizers (see Figure 4-2). Figure 4-3 shows

a comparison with evaluations of apples (4 countries), oranges (2 countries), and

strawberries (2 countries) using a serving of fruit26 as the unit of comparison. This and


26 Servings/kg for fruits used for comparison in the results are: 1 kg pineapple = 3.09 servings; 1 kg apple
= 8.26 servings; 1 kg orange = 4.06 servings; 1 kg mango = 4.18 servings; 1 kg cantaloupe = 2.88
servings (based on formula used for pineapple in methods, ((1 kg fruit)(edible fraction))/(weight of USDA
kg/serving)). Comparisons to Pimentel and Coltro were made by calculating the CED of analogous inputs
from Ecoinvent for reported inputs rather that using originally reported energy totals. See the Appendix
for recalculations.


100









forthcoming comparisons are only preliminary, as the full ROeP of these other sectors,

with the exception of orange (BR) in this case, is not fully characterized. Nevertheless,

the median value of pineapple is higher than the values reported for apples and

oranges, although there is likely cases in production of these fruits (based on the RoEP

of Brazilian oranges), where a better performing pineapple has a lower CED. This

results differs from what is revealed in a comparison on a per kg basis, where the

median of the RoEP for pineapple (1.5 MJ/kg) is in the middle of the RoEP of CED for

the different apple sectors (1.2, 1.0, 1.67, and 2.4MJ/kg). The strawberries both show

more than double the pineapple CED/serving.

Carbon Footprint

The carbon footprint RoEP for pineapple at the packing facility was between 0.16

and 1.42 kg CO2-equivalent/kg, which is equivalent to a range of 52 to 469 g per

serving. The majority of this carbon footprint could potentially come from carbon loss

from land use change, which could add up to 1.24 kg CO2-eq./kg pineapple in the case

of conversion from tropical moist forest, which was estimated to contain 394 tons C/ha.

Of the sample farms, no land conversion from primary forest was reported by the

producers, with no resulting loss of carbon from land use change, and as this is likely

the case for many farms, RoEP is also reported without land-use change. Not including

land-use change, approximately half of the carbon footprint occurred upstream of the

farm (51%) and (49%) of the footprint occurred on the farm, with 34% being contributed

from N20 release from N-fertilizer and 15% from CO2 primarily from fuel combustion.

Fertilizer production (44%), followed by pesticide production (4%), fuel production (2%),

and machinery production (1%) dominated upstream carbon footprint (Figure 4-4).


101
























',l,-1 t -: ;,in'


-- I


I Fi,- T


F-r riii i 1 -





F-::.-i. i I-.- 1,--

[ I;^:! 111 1 -1 7- '.


Figure 4-2. Contribution to CED of pineapple, at packing facility.


.1


I,-<-
<;


, ,


Figure 4-3. Non-renewable CED of one serving pineapple in comparison with
evaluations of the farming stage of other fruits. Sources: Apple DE and Apple
ZA (Blanke and Burdick 2009); Apple NZ (Blanke and Burdick 2009; Canals
2003); Apple US and Orange US (Pimentel 2009); Strawberry ES (Blanke
and Burdick 2009; Williams et al. 2008); Strawberry UK (Lillywhite et al. 2007;
UoH 2005; Williams et al. 2008)


102


onfarm. 1


.


-









The carbon footprint of pineapple, assuming no land use change, translates to

approximately 0.03 to 0.08 kg C02-eq./serving. This is higher than reported for apples

from New Zealand and the United Kingdom, close to that reported for strawberries from

Spain but mostly lower than strawberries from the UK; noting that the full RoEP for

these other fruits is not reported (Figure 4-5).















Figure 4-4. Contribution to carbon footprint of pineapple, at packing facility. Potential
footprint from land-use change is not included.

Virtual Water Content and Stress-Weighted Footprint

Lower ET rates due to the physiological adaptations of the pineapple plants, along

with infrequent to no use of irrigation due to high and consistent annual rainfall (with the

exception of one farm) resulted in a lower evaporative portion of the virtual water

content (green + blue water) for pineapple in comparison with the farm stage for other

fruits (Figure 4-6). For pineapple, the non-evaporative, grey water component is larger

than the evaporative water, owing to the leaching of nitrate from use of N-fertilizers in

pineapple cultivation. Most of the uncertainty in the virtual water content can be

explained by the variation in the grey water footprint due to nitrate emissions; the

sensitivity analysis of the CROPWAT model for pineapple showed little regional
sensitivity analysis of the CROPWAT model for pineapple showed little regional


103









variation in estimated ET for pineapple fields; the most significant variable is the crop

coefficient (relationship of crop ET to pan ET), which has variable estimates in the

literature (Malezieux et al. 2003).

The stress-weighted water footprint (SWWF) of pineapple in the baseline scenario

is negligible; the estimated confidence interval is 0.004-0.017 L/serving, because the

water-stress index for Costa Rica is very low (0.02 on a scale of 0 to 1). In comparison

with mango grown in AU, with a stress-weighted water footprint on average of 74

L/serving, the effect on water deprivation caused by pineapple is negligible.



0.5


S14


0 :,
i "







Pineapple Fi- aple \ :-\- H Z :.I- K ytrlawberry -tIawberry
,:F:.' .'LUC) ,F:(no L.i', ES UK

Figure 4-5. Carbon footprint of one serving pineapple in comparison with evaluations of
the farming stage of other fruits. Sources: Apple NZ (Canals 2003); Apple UK
(Lillywhite et al. 2007); Strawberry ES (Williams et al. 2008); Strawberry UK
(Lillywhite et al. 2007; UoH 2005; Williams et al. 2008).


104










700

600

500

400



200

100

0


B grey H20

[ green + blue-1 H20



-" ,- "- -
^-.-.-=^.^




*/=*;** */=*=


F[.- I I I CR Apple GLO 'r ir GLO IG.i.- : 1.1i

Figure 4-6. Virtual water content (VWC) for pineapple in comparison with other fruits.
Evaporative and non-evaporative water are included for pineapple and mango
(green + blue + grey); only evaporative water is included for apples and
oranges (green + blue). Mango data is from Riddout et al. (2009); apple and
orange data from Chapagain and Hoekstra (2004).

Aquatic Eutrophication

The eutrophication RoEP was estimated to be between approximately 1 and 15 g

N-eq./kg pineapple or 0.3 to 4.8 g N-eq/serving. More than 90% of potential

eutrophication effects were related to NO3 leached from fields (53%), phosphorus

bound to eroded sediment, and leached phosphate (10%) (Figure 4-7). P in eroded soil

was a the most variable of the contributors, with a cooefficient of variation of 173%,

which relates to the high variability of erosion. The estimated percentage of P lost to

erosion of all P applied varied between 0 and 20% among participating farms; percent

of N estimated to leach from fields as N03-N varied between 10% and 34%.

While direct comparison among evaluations of fruits using different methods of

estimating eutrophication-related field emissions is very difficult, preliminary


105










comparisons can be made by multiplying emissions by the same TRACI

characterization factors used in this study. The results are shown in Figure 4-8.


N i 'Ii I






-..












Figure 4-7. Contribution to potential eutrophication of pineapple by emission.


60











1.0

I II I I

PineappleCR AppleNZ AppleUK Cantaloupe Sti ,. l-i. Strawberry
CR ES UK
Figure 4-8. Preliminary comparison of potential eutrophication effects of one serving
pineapple in comparison with evaluations of the farming stage of other fruits.
Sources: (Canals 2003); Apple UK (Lillywhite et al. 2007); Cantaloupe CR
(Hartley-B. and Diaz-P. 2008); Strawberry ES (Williams et al. 2008);
Strawberry UK (Lillywhite et al. 2007).


106









Human and Ecological Toxicity

The RoEP for human toxicity was estimated to be 1.7E-10 to 1.1 E-9 disease

cases/kg pineapple, but could be as much as 1000 times greater or less, due to the

uncertainties inherent in the USETox model. The RoEP for freshwater ecotoxicity was

0.2 to 1.4 PAF in m3/day/kg pineapple, but could be as much as 100 times up greater or

less.

The pesticides contributing the most to ecotoxicity are diuron, ametryne

(herbicide), ethoprop, and paraquat (herbicide) (Figure 4-9). Toxicity characterization

does not necessarily correspond to quantity applied in the field; half as much ethoprop

is applied as diuron and diazinon, and less of that applied is emitted from the field (5%

for ethoprop vs. 26% and 27% of diuron and paraquat), but its toxicity effects when

being transported and coming into contact with humans and freshwater ecosystems is

much stronger on a unit basis. Not all pesticides have demonstrated human toxicity

effects although they do cause damage to freshwater ecosystems, including ametryne

and bromacil.

In contrast to the temperate environment (Denmark) in which PESTLCI was

originally calibrated, the Costa Rican environment has higher average annual rainfall

and solar insolation which increases the estimated runoff and abiotic degradation of

pesticides, respectively. The PestLCI-CR model shows a greater fraction being

delivered to water, but a smaller fraction being delivered to air than in the default

PestLCI model. Total emissions of pesticides are greater overall in the default model.

The USETox-CR characterization model for the toxicity effects of these pesticides also

shows differences from the default European parameterization. The USETox-CR


107









characterization factors for ecotoxicity for emissions range from 1.5 to 6 times less than

in USETox-EU; characterization factors for human toxicity for emissions are equal for

emissions to air but 1.5 to 3 times less for emissions to water. Despite these absolute

difference, relative toxicities among these pesticides are modeled similarly.




S- rba rb l




_-- .-





























Figure 4-9. Relative contribution of active ingredients of pesticides used in pineapple







108
,,' ,; "


















..... ----'---" -- i_-_ D iazrin o n








Figure 4-9. Relative contribution of active ingredients of pesticides used in pineapple
production to (a) human toxicity and (b) freshwater ecotoxicity.





108









Results Summary

Table 4-1 presents a summary of the life cycle environmental performance of

pineapple production through transport to the packing facility. On farm processes are

responsible for the majority of impacts (given since some impacts were only modeled at

the farm stage due to assumption it contributes the majority of this type of impact) with

the exception of the cumulative energy demand and to carbon footprint; about half of

the carbon footprint occurs upstream and half on the farm. The uncertainty of each

modeled impact, as measured by the coefficient of variation, varies markedly from less

than 10% for land use, for which yield variation is the sole contributor to uncertainty, to

human toxicity, which has a high level of uncertainty due to the large uncertainty in the

toxicity characterization factors.

Discussion

The data underlying the inventory represent medium to large size farms in the

three primary geographic zones in Costa Rica. Sufficient input data from the smallest

producers (<10 ha) was solicited but not acquired, likely due to less stringent

bookkeeping practices and also heavier reliance upon larger producer associations for

tasks, managements, and equipment. The other end of the spectrum of producers, the

largest national and multi-national companies with farms >250 ha, is neither directly

represented. Although solicited, none of the four largest companies agreed to provide

primary data for this study.

All emissions and inventory results reveal the importance of yield in impact

estimations, confirming recent findings in agricultural LCA (Roos et al. 2010). With

higher yields and an equal amount of impact/area, impacts are diluted across more

product, representing higher environmental efficiency. The average yield reported for


109









the sector (67 tons/ha) falls at the 9th percentile of the yield distribution of the sample

farms that contributing production data, indicating a bias toward more productive farms

in the sample used to create the baseline scenario. However, because the reported

average sector yield falls within the confidence intervals for yield varied here, this

national average pineapple falls within the distribution modeled. It is necessary to

reiterate here that the objective was to model the expected range of environmental

performance in the sector, and that the range rather than the median or mean values

should be the focus of the results.

The wide ranges of performance evident for all impacts categories indicate the

importance of farm-level assessment to differentiate environmental performance of

pineapple production among farms. In the initial comparisons of environmental

performance between farm stage production of pineapple and other fruits, where such

comparisons were possible, pineapples perform within a similar range, seemingly better

in some categories and worse in others, but the full RoEP for the other fruits was not

published nor calculable in most cases, limiting the ability of comparison. The

estimated RoEP for energy demand for pineapple showed it to be higher in energy

demand than apples and oranges on a per serving basis, but lower than Spanish and

British strawberries. The carbon footprint reflected a similar patterns with less of a

relative difference between pineapples and other fruits. Pineapple was lower in

consumptive water use than apples, oranges and mangos, but higher than mangos in

its gray water requirement. Without the need for irrigation in most areas and because of

its physiological adaptations to water stress, water use impacts were minimal in

comparison with other fruits. The broad RoEP of eutrophication for pineapple indicates


110









the relatively higher degree of uncertainty for this category, and considerable potential

overlap in this respect with other fruits.

Because production inputs dominate energy demand and carbon footprint, the

relatively high-agrochemical input intensity of pineapple cultivation (FAO 2006; Su

1968) may explain in part why these indicators are higher for pineapple in relation to

other fruit. Additional explanation is provided by the fact that there are less servings of

pineapple per kg than the fruits compared here, largely because of the higher non-

edible potion of pineapple (about 50%).

The Significance of Regionalized Emissions and Impact Models

The significance that climatic, geographic, crop, and field-specific factors have in

emissions and impact models is supported by the differences in outcomes of the

regionalized and the original versions of models used here. Water loss estimates from

CROPWAT are dependent on water balance calculations based on climatic, soil, and

plant conditions, and estimated will differ greatly among different climate zones and by

crop. The PESTLCI model showed great variation in emissions between the default

conditions (Denmark) and Costa Rica. Characterization factors for pesticides differed

by up to 70 times for toxicity factors between the default USETox and the USETox-CR

model. Using regionalized models will likely have significant effects on LCA outcomes,

and should be applied with careful attention to the capacity to accurately describe

conditions, but is essential for more accurate characterization of local and regional

impacts.

Although regional data was incorporated into these models, all those adapted here

operate independently and use a unique set of field parameters. Attempt was made to

use consistent parameterization of these models, but there is no guarantee of


111









consistency of model calculations of common parameters (e.g. runoff is estimated in

PestLCI, CROPWAT, and RUSLE2). Some models achieve a higher degree of

specificity (RUSLE2) than others (CROPWAT) and thus some do not utilize all data that

could theoretically influence results. However, the use of freely, publically-available

models adaptable to a wide range of conditions is of high utility for likelihood of use and

for comparability. The N and P fertilizers emissions model was adapted based on

average pineapple nutrient uptake rates, but otherwise did not account for regional

climatic conditions or soil properties. The model presented here is an improvement

upon solely arbitrary designation of emissions fractions of all forms of N and P (e.g.

35% of N leaches to soil), some of which, including N leaching, has been estimated to

vary between 10 and 80% of applied N (Miller et al. 2006), and may be sufficient for

relative comparison among farms, but could be replaced with a more detailed process-

based model as is used here for soil erosion, water use and pesticide emissions. These

models could all be improved with better parameterization based on data collection on

pineapple farms in Costa Rica for variables including pineapple biomass, nutrient

uptake, water use, and leaf permeability to pesticides.

Estimated Environmental Impacts

All estimates of environmental impacts need to be considered in light of the

accuracy of their characterization and of the inputs data underlying this characterization.

Experimental quantification of soil erosion is typically marked by high variability,

usually because erosion is strongly event-based and the difficulty of capturing a

representative sample of eroded sediment. Data from experimental measurement of

soil loss in CR are no exception to this (see Table 15-1, Rubin and Hyman 2000). In

consequences models based on long-term climatic and management data may be


112









preferable and yield more comparable results for quantification of soil erosion in LCA.

However they should still be validated with existing data. The RoEP of 0.02 to 32

tons/ha soil erosion tons/ha/yr found here does confer with existing estimates of erosion

of mineral soils under pineapple cultivation in Hawaii and Australia.

Land use, energy use and carbon footprint were estimated with the lowest

uncertainty, however the latter two are both heavily dependent upon the quality of the

input data for upstream processes. Carbon loss through land transformation has been

calculated to be a dominant factor in the carbon footprint of crops occupying former

tropical forest (Fargione et al. 2008), and that could possibly occur for pineapple

cultivation, if it replaces tropical forest. There is, however, little evidence to suggest that

pineapple expansion in Costa Rica has been a direct cause of deforestation since 1990

(Joyce 2006). Nevertheless conversion from other types of land use, including

secondary forest and pasture, could also result in carbon loss but is not quantified here.

As far as eutrophication and toxicity impacts are concern, which are impacts based on

potentially long-range transport, persistence and availability in environmental media, the

effects on ecosystems (freshwater ecotoxicity) and humans (human toxicity) should be

read with appropriate skepticism of the capacity of generic models to make accurate

estimations without explicit spatial data; nevertheless because these aspects (fate,

transport, toxicity effects) are all relevant to their ultimate effect, they should be

considered superior to just reporting quantities of pesticides released.

Potential Impacts Not Measured

The scope of this LCA was strictly limited to environmental impacts, and did not

include any evaluation of social or economic impacts. Both of these impacts can


113









potentially be accounted for in LCA, with the related tools of Life Cycle Costing (LCC)

and the newly developed Social Life Cycle Assessment (SLCA).

Aside from loss of stored carbon, land use conversion and occupation can have

ecosystem consequences on biodiversity across multiple scales (ME Assessment

2005), and this should be accounted for in the LCA, and has been recommended for

consideration and methods are under development, but none were judged to be

sufficient to capture effects on biodiversity of pineapple production in the studied

environment.

Handling and application of pesticides in the field could have direct impacts on

worker health, but no suitable methodology exists for measuring this in LCA. However

all farms sampled reported use of protective equipment among workers in the field to

reduce this risk.

Residual organic waste on pineapple fields has been blamed for ecological

consequences such as providing the substrate for the larval stage development of biting

flies (Sandoval 2009), which have potential consequences for local livestock. Such

consequences have not been addressed here.

Conclusions and Recommendations for Farm Level LCA of Fruit Products

The development of inventories of agricultural processes and the characterization

of their impacts are two separate but interdependent stages of the LCA. Since fruit

products depend on further downstream processes before reaching the final consumer,

inventories should include sufficient information that impacts can be characterized for

their entire farm-to-disposal life cycle stages. Yet particular attention should be paid to

those inventory items that need to be recorded in the farm stage because of their


114









likelihood to dominant full life cycle impacts: these include water use, eutrophication,

toxicity, and soil erosion.

Evidence here shows that it is essential to include upstream processes to fully

characterize energy use for farm LCA, because energy use in agricultural inputs such

as fertilizers may dominate cumulative energy use through the life cycle stage.

Acknowledging this importance, life cycle data on farm input production adapted from

LCI databases with a EU-focus such as Ecoinvent used here needs to be validated for

its application in other world regions. Because actual farm level energy use is

dominated by liquid fuels for farm equipment such as tractors, energy use is likely to be

strongly correlated with other impacts during the farm stage dominated by fuel

combustion, including greenhouse gas production, acidification, and photochemical

oxidant production. Emissions to air causing these impacts should be included in

agricultural inventories for use in full life cycle studies, but for sake of brevity and

increased interpretability of LCA users, characterization of these impacts at the farm

level is likely to be unnecessary because of its redundancy. This may not be the case if

other energy sources (e.g. biofuels or electricity) comprise a substantial proportion of

farm stage energy use.

Use of LCIA indicators should be based both on environmental relevancy and

sufficient characterization models and uncertainty estimation. In this case we

recommend use of a measure of cumulative energy consumption, such as CED. Use of

other broader measures of energy use, such as emergy, would present a richer picture

of energy use that is more informative for measurement of long-term sustainability, but

should only be used if accurately integrated into the life cycle inventory and for which


115









model uncertainty is described. Energy use also is characterized by relatively low

model uncertainty, which increases comparability of different products.

Local and regional environmental impacts related to soil erosion, water stress,

eutrophication, and ecological and human toxicity are particularly relevant for farm level

process and require characterization adapted to the region of production. Soil erosion

is a particularly localized indicator requiring a large amount of field-specific information

to accurately model. It is highly relevant for areas with sloped terrain and high rainfall.

The direct downstream impact of soil erosion on water quality though sedimentation,

was not quantified here but is a relevant environmental impact that deserves future

investigation for LCA characterization. And as demonstrated here, accurate

quantification of soil erosion can be particularly relevant for other impacts, including

eutrophication, due to loss of nutrients bound to soil in erosion, and potentially for

toxicity impacts, although the contribution of eroded sediments to those impacts was not

quantified here. Farm level emissions are marked by high levels of variability,

especially related to yields, and uncertainty due to complex and site-specific fate,

transport, and effect processes of agricultural emissions. We recommended that farm-

stage LCAs reported data along with sufficient range parameters to quantify uncertainty

in input data related to those emissions, uncertainty in the emissions themselves, and if

characterized, uncertainty in the characterization factors. Finally, farm stage

assessment data must be coupled with data on downstream life cycle stages before

being fully evaluated by the end-consumer.


116









Table 4-1. Summary table for impacts of 1 kg pineapple delivered to packing facility.
RoEP Contribution to Impact Variance of Impact
%
contribution Most significant Factor most responsible
Indicator Unit Min Max of farm stage contributor CV for variance
Land occupation m2/yr 0.14 0.21 100% yield 9% yield
Soil erosion kg eroded 0.0005 0.6 100% farm slope 165% farm slope
soil
NR cumulative energy MJ 1.2 2.2 23% fertilizer production 25% yield
demand
Carbon footprint (with kg C02-eq. 0.16 1.4 89% land use change 48% carbon loss from land-use
LUC) change
Carbon footprint (no kg C02-eq. 0.10 0.3 49% fertilizer production 19% yield
LUC)
Virtual water content L 124 1450 100% water for dilution of 21% nitrate emission
pollution
Stress-weighted water L 0.0044 0.017 100% water for application of 21% yield
footprint fert./pest.
Aquatic eutrophication kg N-eq. 0.0008 0.015 96% nitrate emission to 62% P in soil eroded
6 water
Human toxicity disease 1.7E- 1.1E- 100% Ethoprop (nematicide) 46% amount of ethoprop
cases 10 09 applied
Freshwater ecotoxicity PAF/m3/da 0.2 1.4 100% Diuron (herbicide) 44% fraction of diuron emitted
y to water
Notes
a Based on the largest CV for related inventory item among yield, associated input, or emission model. If this was the
emissions model, the most sensitive variable in the sensitivity analysis was used.









CHAPTER 5
SUMMARY AND SYNTHESIS

Summary

The primary objectives of this dissertation were to better equip life cycle

assessment to relate the production of goods and services to their associated

environmental impacts by means of the following tasks: to provide a new process-based

life cycle assessment (LCA) impact method for quantifying impacts of resource use with

emergy; to provide this method with an accompanying method of uncertainty analysis;

and to create a method for establishing the range of environmental performance for

agricultural products with a set of indicators adapted to a tropical environment. To do

so, this dissertation included two original LCA studies, one of gold-silver bullion from the

Yanacocha mine and one of pineapple production in Costa Rica, and an original

uncertainty model for use with emergy results. The major conclusions that can be drawn

from these studies are first listed by chapter and followed by a general synthesis of the

dissertation along with the ramifications of the findings.

Chapter 2 Summary

Emergy is an ideal measure of total resource use because it traces energy
directly and indirectly used in creation of products back to the driving energies of
the biosphere (sunlight, tides, and deep heat) and can be used to measure
environmental contribution to raw and processed resources and materials as
well as direct environmental flows (e.g. sunlight, wind, rain). All indirect and
direct energy can then be aggregated as emergy in sunlight energy equivalents
for a single numeric value of resource use.

Emergy can be integrated into conventional process-based LCA databases to
track direct and indirect energy flows associated with complex process chains
and in this manner is compatible with process-based LCA.

In order to characterize resource use with emergy for a mining product, an LCA
of the gold-silver mining operation at the Yanacocha mine in Peru was conducted
using an boundary that extended from the environmental contribution to the


118









inputs to mining (permitted by emergy) to the creation of gold-silver bullion. A
gram of gold-silver bullion was used as the functional unit.

Total emergy in 1 gram of gold-silver bullion is in the range of 4.4E+11 to
1.3E+13 sunlight equivalent joules (sejs), which is orders of magnitude higher
than most common resources, including other minerals, fuel sources, foods, and
ecosystem products. 95% of the emergy in gold-silver bullion comes from inputs
to mining processes rather than gold formation (and thus is based on
environmental contribution that occurs off-site), despite the millennia of
environmental work used to form gold deposits. The contribution of emergy to
chemicals and fuels used in the mining and refining processes dominate the
emergy contributing to the bullion.

The breakdown of emergy used to make gold-silver bullion does not reflect the
same pattern as cumulative energy demand, indicating the failure of the latter to
characterize all indirect environmental flows to processes, and reinforcing the
role of emergy in LCA to quantify these flows for a more complete measure of
resource use.

Use of allocation rules from LCA for allocating impacts among by-products and
those traditionally used in emergy result in drastically divergent outcomes;
allocation rules from LCA are more consistent with LCA data and should be used
if results are to be adopted in future downstream LCAs (e.g., for a product that
uses gold-silver bullion as an input).

Tracking labor and information inputs into processes is not typically done in LCA
and thus integrating emergy into life cycle assessment databases will not permit
the quantification of emergy in labor or information which is a shortcoming to
using emergy in LCA because it arguably omits important environmental
contributions to final products that should be accounted for.

Chapter 3 Summary

The range of accuracy, or uncertainty, of emergy values should be quantified so
that the model uncertainty of using emergy is quantified in an LCA study, as this
could be the dominate form of uncertainty present in the LCA results that use
emergy as an indicator.

Two options are demonstrated for estimating uncertainty of unit emergy values
including an analytical model based on mathematical rules for propagation of
uncertainty and a stochastic model using Monte Carlo analysis. Results of either
approach show that unit emergy values have confidence intervals that resemble
lognormal distributions and that these confidence intervals can be represented
mathematically with the median value times or divided by the geometric
variance.


119









Three forms of uncertainty are present in emergy calculations, including
parameter, scenario, and model uncertainty. All three components can be
combined using the propagation of uncertainty approach to result in the broadest
estimation of potential uncertainty, but depend upon the estimation of the
uncertainty in parameters and existing models.

Unit emergy value confidence intervals for table-form unit emergy value
calculations, the most common calculation approach, are only renderable with a
Monte Carlo model approach because there is no simplified mathematical form
for estimating them analytically; thus the stochastic approach is suggested to be
the most valuable of the approaches introduced.

The estimated factor of uncertainty for emergy values does not always
correspond to the presumed range of an order of magnitude. The uncertainty is
variable but will be smaller than the uncertainty factor of the largest contributing
input, demonstrating that uncertainty is not infinitely compounded in more highly-
transformed products.

Issues remain with using uncertainty factors for comparison of unit emergy
values that share common parameters. The method requires further adaptation
for handling the issue of covariance.

Chapter 4 Summary

LCA-based environmental performance of tropical fruit production in non-OECD
countries is largely uncharacterized in comparison with agricultural activities in
temperate countries, yet the production of fruit has growing importance in the diet
of North Americans and Europeans, and occupies increasing area in the tropics.
Environmental product declarations provide one means of providing both LCA-
based information and a market-based mechanism for reduction of impacts
associated with production activities. A host of LCA methods need to be
developed or adapted to account for the potential environmental impacts that are
very relevant especially in humid tropical environments. Fresh pineapple from
Costa Rica is a crop of both growing export importance and increasing
environmental concerns with production.

A farm-to-gate LCA was designed to sample representative production systems
and conditions present in the Costa Rican pineapple sector. A statistical method
was used to combine variability in yield, production inputs, and emissions models
to estimate a range of inputs and emissions relevant to energy use, water
consumption, soil erosion, land use, carbon footprint, eutrophication, and toxicity.
Combined with impact characterization methods, this variability in inputs and
emissions was used to create ranges of environmental performance for the
sector.


120









* In addition to a functional unit of mass (1 kg), the functional unit of 1 USDA
serving was used in order to compare LCA results with those generated for
products that serve the same function providing 1 serving of fruit.

* Soil erosion is a primary environmental concern associated with pineapple
production in Costa Rica because of exposure of topsoils, sometimes on steep
slopes, to high rainfall, but no commonly-used LCA method incorporates soil
erosion as an indicator. USDA's RUSLE2 soil erosion model was adapted to the
climate conditions and observed field parameters in Costa Rican pineapple
plantations for estimating soil erosion.

* Methods developed for characterization of pesticide emissions (PestLCI), toxicity
assessment (USETox), and crop water consumption (water footprint using FAO's
CROPWAT), and were each adapted to the extent possible to account for the
local conditions. The result of these adaptations were significant differences in
characterization of impacts occurring in Costa Rica from the same
characterization in the default models (developing mainly in Europe), suggesting
the importance of adaptation of emissions and impacts models to the
environments in which the emissions occur.

* The ranges of environmental performance for pineapple, described by the
coefficients of variation, ranged from 9% for land-use to 165% for soil erosion,
demonstrating significant variation within the sector, with range of performance
for impacts where models incorporated local conditions being the most variable.

* The largest contributor to farm-to-gate energy use and carbon footprint was
fertilizer production, thus stemming from upstream processes. On the farm level,
greenhouse gas emission were dominated by N20. Water consumption was low
because of the low water requirement of pineapple and sufficient precipitation.
Soil erosion was highest (close to 0.5 kg soil/kg pineapple) in areas with steep
slopes, no contouring, and erodible soils, but is potentially as low as 0.005 kg
soil/kg pineapple in flat sites with good drainage, erosion-resistant soils, and
management practices that involve contouring, use of plastic mulch, and minimal
exposure time. Eutrophication was dominated by nitrate emissions but was
highly variable due to variable emission of P in eroded sediment. The pesticides
contributing the most toward human toxicity and freshwater ecotoxicity were
complex factors related to transport, degradation, as well as potential health
effects, and could not be predicted simply by mass applied or their specific
toxicity, supporting the importance of using emissions and impact models.

* Because of the variability within the sector, as well as potential model differences
for some indicators, comparisons with other fruits were inconclusive. However,
pineapple largely had a higher energy use and carbon footprint than apples and
oranges, but lower values than greenhouse-grown strawberries. Pineapple had a
lower virtual water content than apples, oranges, and mango although it had a
large grey water footprint (pollution dilution water requirement) because of
significant nitrate losses.


121









Synthesis

LCA serves the purpose of providing a measureable link between production of

goods and services and pressure on environmental resources but requires further

methodological expansion and refinement to provide more relevant and accurate

information for environmental decision making regarding production and consumption.

Expanded methods described here include the use of emergy as an indicator of

resource use accompanied with an uncertainty model for emergy, and a unique

combination of LCIA indicators applied with an original method of describing the range

of environmental performance of a tropical agricultural product across the product

sector in a country.

The use of LCA as a tool to inform and direct sustainable production and

consumption depends both on its methodological capacity to describe impacts

accurately and a means of conveying complex environmental information in a form that

is useful when making product design, management, or purchasing decisions as well as

for informing policy making. In reference to its current methodological capacity, the

incorporation of a measure of resource use that characterizes all process inputs in a

common form using emergy provides a way of measuring the environmental

contribution to products in the context of the availability of resources. Integration of

emergy in LCA as a measure of resource use is not limited to mining products but is

applicable to all products and services for which process data is available to

characterize them.

The incorporation of additional or modified methodologies for impact assessment

in LCA is also needed for cases where relevant indicators do not exist or are not

accurate in the environments in which production processes take place. Soil erosion is


122









a relevant environmental concern with agricultural production in Costa Rica, but no

adequate methodology for characterizing it existed in LCA. To this end the RUSLE2

model was incorporated into LCA as a measure of rain-based soil erosion. Furthermore

other impacts models were adapted for use in the Costa Rican environment, including

PestLCI and USETox, that had been developed for the Danish and European

environments, respectively. Customization of impacts models for local conditions are

not commonly performed in LCA, however, they are essential for characterizing impacts

of emissions that have local or regional effects, such as eutrophication and toxicity

associated with nutrient and pesticide emissions from farms. Through adapting model

parameters to the local environment, strong comparability can still be achieved between

products in different environments because the same model structure and its underlying

physical assumptions are used. This is particularly necessary when environmental

conditions are sufficiently different than those in which a baseline model was developed,

to the extent that they create meaningful differences in model outcomes.

Despite the increased accuracy of LCA emissions and impact models that may be

conveyed with greater environmental customization, uncertainty in the results will

always be an issue because of lack of full primary data availability and because of

model uncertainty. The uncertainty associated with model results can be large, as

shown in some of the impact assessment results for pineapple and those for gold, which

can complicate the task of selecting environmentally preferable products. Incorporation

of uncertainty information should be associated with any LCA impact method. Emergy

was adapted as a new impact method, but there was no associated method of

measuring model uncertainty in emergy. The new methodology developed now


123









provides a means of characterizing the uncertainty of the emergy of a product that can

accompany the use of emergy in LCA, as was demonstrated with the LCA of gold.

One of the most valuable uses of LCA is for direct comparison of the

environmental performance of products that serve the same function. By incorporating

data and model uncertainty into LCA results, comparisons can be more realistic and

subject to statistical tests that are not possible by comparing averages. Because

incorporation of uncertainty may hinder interpretability and the purposes of comparison

(because comparing ranges is less intuitive than comparing points), further

methodological detail and guidance can assist in the consistency of uncertainty

application and its practical usage. This is the very purpose of a current UNEP working

group on uncertainty management in LCA.

Another challenge related to the use of LCA for promoting sustainable production

and consumption, other than the challenge of improving the accuracy of LCA data and

models, is the meaningful presentation of LCA results in a form that both producers and

consumers can interpret to make informed production and consumption decisions. One

way to improve interpretability is to permit direct comparisons between a product and

others in the same product category. Environmental performance for products within a

category may include a high degree of variability that comes from differences in

production practices and production site characteristics. Describing this variability can

be potentially used to situate the environmental performance of individual product

supply chains within the product category as an improved means of interpreting its

environmental performance in addition to describing the performance range present

across the sector for comparison with other products that serve the same function (e.g.


124









a serving of fruit). The construction of such a range is often hindered by lack of

sufficient production data to describe an entire product sector. However this range can

be approximated by sampling representative production processes and for accounting

for variability of environmental conditions that could occur based on differences in

production location, which are particularly relevant for agricultural production activities.

A method for combining variation in production practice and environmental conditions to

describe the range of environmental performance was developed here for pineapple

production in Costa Rica, with the outcome being ranges of environmental performance

for farm-level pineapple production for Costa Rica.

The production systems modeled in this dissertation were two primary sector

products, gold-silver bullion and fresh pineapple, from two different non-OECD

countries, Peru and Costa Rica. Product supply chains in non-OECD countries,

particularly those which are largely located in the tropics, have been poorly

characterized thus far through LCA. Implementation of LCA in non-OECD countries

requires adaptation of data and impact assessment (LCIA) methodologies for

measuring the environmental impact associated with production chains in these

countries for exports that are consumed in OECD countries. LCA is uniquely

appropriate for quantifying the environmental burden of this production-consumption

pattern because it is only by accounting for impacts over the full life cycle that the

responsibility of OECD consumers for environmental burdens in non-OECD countries is

quantifiable and thus can be addressed by associated market-based or policy measures

to reduce these burdens. This was demonstrated through two unique adaptations of

LCA for one Peruvian and one Costa Rican export product, with implications in each


125









case for improved environmental management from both producer and consumer

perspectives.

The availability of process data and technical capacity to use LCA in non-OECD

countries is likely to be less than in OECD countries. Where the data or capacity does

not exist, incentives and new mechanisms for use of LCA in non-OECD countries,

particularly for products exported to OECD countries where there is a demand for

environmental information, are required. The use of LCA-based labels called

environmental product declarations (EPDs) could be a market-based mechanism for

improvement of export production by providing information for buyers and consumers in

importing nations that could be used to select the products that have the lesser impact.

EPD programs for these product labels exist in a number of EU and Asian countries and

are emerging in the US. Products from non-OECD countries could be registered in

these programs and used to inform purchasing decisions of buyers. Alternatively or in

concert with OECD programs, EPD programs could be developed in non-OECD

countries as a way to gauge, publish, and promote environmental performance of export

production. Thus EPDs are an application of life cycle assessment that could promote

trade of more environmentally benign products by influencing both the production and

consumption aspects of the supply chain. This particular application of life cycle

assessment should improve LCA interpretability and function to broaden use into

international markets.


126










APPENDIX A
SUPPLEMENT TO CHAPTER 2: PROCESS TREE AND UNCERTAINTY ESTIMATES

Dore, at
Yanacocha
Q1ZT7 j


Figure A-1. SimaPro process tree of environmental contribution (sej) to 1 g dore.
Inputs contributing 5% or more of the total emergy visible.


127









Table A-1. Uncertainty estimates for UEVs for inputs into gold-silver bullion production.
Item for which uncertainty
estimated Uncertainty estimate used for a2geoReference
Electricity, from oil Electricity from all sources in mix 2.8 1


Gold, in ground
Groundwater, global
Iron, in ground

Lead, in ground
Oil, crude

Silver, in ground
Sulfuric acid
Sources
1 (Ingwersen 2010)


Gold, in ground
All process water
Pig iron, steel
Pb in lead acetate and Zn in zinc
powder
Crude and refined oil, natural gas

Silver, in ground
sulfuric acid, HCI, general acids


Table A-2. Estimation of total uncertainty in gold in the ground.
No. Parameters p 2 2geo
1 crustal concentration (ppm) 4.00E-03 0.001 1.96
2 ore grade (ppm) 0.87 0.04 1.10
3 crustal turnover (cm/yr) 2.88E-03 6.77E-04 1.58
4 density of crust (g/cm3) 2.72 0.04 1.03
5 crustal area (cm2) 1.48E+18 2.1E+16 1.03
Models
6 Alternate Model UEVs 5.68E+14 9.22E+14 9.28
Summary
Unit emergy value, p (sej/g) 3.65E+11
Parameter Uncertainty Range (No. 1-
5)
Pgeo (sej/g) (x+) o2geo 3.35E+11 (x+) 2.27
Total Uncertainty Range (No. 1-6),
Pgeo (sej/g) (x-) o geo 1.75E+11 (x+) 10.74
Sources
1 Butterman and Amey (2005)
2 Newmont (2006c)
3 Odum (1996); Scholl and von Huene (2004)
4 Australian Museum (2007); Odum (1996)
5 UNSTAT (2006); Taylor and McLennan (1985); Odum (1996)
6 ER method and Abundance-Price Methods (Cohen et al. 2008), Odum
(1991)


128


10.7
2.0
7.5

11.1
3.6

10.6
3.3


Table
A-2
1
1

1
1
Table
A-3
1









Table A-3. Estimation of total uncertainty of silver in the ground.
No. Parameters p O2geo
1 crustal concentration (ppm) 7.50E-02 0.007 1.20
2 ore grade (ppm) 1.13 0.06 1.10
3 crustal turnover (cm/yr) 2.88E-03 6.77E-04 1.58
4 density of crust (g/cm3) 2.72 0.04 1.03
5 crustal area (cm2) 1.48E+18 2.1E+16 1.03
Models
6 Alternate Model UEVs 4.97E+14 8.60E+14 10.03
Summary
Unit emergy value, p (sej/g) 2.54E+10
Parameter Uncertainty Range (No. 1-
5)
Pgeo (sej/g) (x+) 2 geo 2.46E+10 (x+) 1.65
Total Uncertainty Range (No. 1-6),
Pgeo (sej/g) (x-) a geo 1.23E+10 (x+) 10.59
Sources
1 Butterman and Hillard (2004)
2-6 See Table 1 sources


129









APPENDIX B
SUPPLEMENT TO CHAPTER 2: LIFE CYCLE INVENTORY OF GOLD MINED AT
YANACOCHA

Background
The gold mine at Yanacocha, Peru operated by Minera Yanacocha, S.R.L, is the
largest gold mine in South America, and the second largest in the world in terms of
production volume. Yanacocha is co-owned by Newmont Mining Company(US),
Buenaventura (Peru), and the International Finance Corporation. The Yanacocha mine
is one of the largest gold mines (in terms of production) in the world. The mine produced
3.3275 million ounces of gold in 2005 (Buenaventura Mining Company Inc. 2006). This
represented more than 40% of Peruvian production (Peruvian Ministry of Energy and
Mines 2006) and approximately 3.8% of the world's gold supply in 2005, assuming
100% recovery of gold from dore and using the total of 2467 tonnes reported by the
World Gold Council (World Gold Council 2006).
Yanacocha is an open pit mine. Ore is obtained through surface extraction. Gold
and silver are extracted from ore through cyanide heap leaching and further refined
through a series chemo- and pyrometallurgical processes. The output of the Yanacocha
mine is a gold-silver bullion called dore, with a mercury by-product. The dore is shipped
overseas for further refining.

Methodology

Scope
The scope of the life cycle inventory (LCI) included gold mining and processing
from the stage of the deposit formation to the overseas export of a semi-refined gold
product (dore). The purpose was to include every critical link in the mining process,
including background and auxiliary processes, with the exception of administrative,
community, and information and other mine support services. The choice to include all
mine operations, described later, is based on the supposition that are all these
operations are necessary for gold mining to occur within the current regulatory and
business contexts. The scope is consistent with a cradle-to-gate LCI but extends further
upstream to encompass both pre-mining activity of the company and geologic work of
the environment. The downstream life cycle of gold production was not included. The
inventory is based on total reported production in year 2005. This a source-side LCI -
accounting for all the inputs to the process but not the emissions and wastes. Therefore
this inventory would not be sufficient for characterizing pollution impacts such as air,
water, or soil contamination.

Purpose
This LCI was constructed to provide a measure of total environmental
contribution to mining. Total environmental contribution was measured as the total
energy used to supply all inputs tracing back to the energies that drive the biosphere
(e.g. solar, tidal, deep heat). This energy, a form of embodied energy which includes
environmental inputs, was estimated following the emergy methodology (Brown and
Ulgiati 2004; Odum 1996)


130









The aim of this LCI is generally descriptive, rather than decision-oriented
(Frischknecht 1997). Neither was it completed for specific comparison. As a
consequence, no inputs or processes were omitted because of redundancy with similar
products or systems.
Furthermore, the purpose was to complete a detailed LCI, rather than a
screening LCA. Therefore rather than relying on existing LCI data, primary data from
Yanacocha was used or original calculations specific to processes at Yanacocha were
performed in all main unit processes and significant27 indirect processes.

Inventory Contents and Organization
As is customary in LCI, the inventory was grouped into a series of unit processes
(National Renewable Energy Laboratory 2008). Nine primary unit processes were
identified and grouped into three unit process types. These unit processes and types
are identified in Figure B-1 Background and auxiliary processes are not always
included in mining LCIs, but are both essential to the mining process. A generic mining
LCI model called LICYMIN includes auxiliary processes (Durucan et al. 2006). This
inventory is unique among mining LCIs, in that background processes, including natural
processes, are included.
Data for the mining activities are grouped by nine units processes, except in
cases where data was available only at the mine level, which was the case for labor.
This item is only tracked at the system level.
Water included in the inventory was water used and evaporated in the process.
Other water used that is recycled or released downstream was not included, as it was
not considered to be consumed.
Both raw materials inputs and core capital goods are included in the inventory.
Core capital goods are defined as installations and heavy equipment critical to
processes at Yanacocha. These include heavy vehicles, processing units such as
ovens and reaction tanks, primary pipes, and large storage tanks. Auxiliary equipment
such as connector pipes, structural skeletons, monitoring equipment are not included.
The omission of small auxiliary capital is justified in the Section 'Inventory Cutoffs'.
Capital goods included elements of process infrastructure such as pad and pool
geomembranes, pipes conveying process material and waste between units, and
earthen materials supporting pads and used in restoration. Earthwork was not included.
Elements of non-process mine infrastructure included in the inventory are roads,
steel buildings, water supply, electricity transmission line, and dams. Equipment used
in mine administration and maintenance such as small trucks, computers, protective
clothing, were omitted. Employee support services such as food, medical, and housing
services were not included due to lack of data. Infrastructure and management of the
San Jose reservoir, a reservoir for mine and community water storage within the mine
boundary, was not included.





27 'Significance' indicates that a process falls within the inventory cutoff as described below.


131











Geologic BACKGROUND PRODUCTION AUXILIARY

Energy I- Deposit Overburden
I Extraction Sedimet It Sediment
AWR Control
I F wT-F, HM, W _
I Barren
Ore Solution
I I A Water
F I__ I


Exploration Leaching -W Water
F,Treatment
FE_H M, C, IW_ CV, E

Pregnant AWR I
Solution


Mine Sediment -W
Mine PWW I
I Infrastructure Processing Reclamation
FFJHM_, FF,fC,I FF, HM I
I I Mercury
Dore
Gold Production at Yanacocha
Study boundary Internal & product flows External input External input included for Emergy
-----------------
Mining Process
Figure B-1. Process overview. Nine unit processes (boxes) are grouped by three
process types: background, production and auxiliary. Geologic processes led
to deposit formation. Deposit discovery occurs during exploration. Before a
deposit can be mined the necessary infrastructure such as roads, electricity
and water supply, and office facilities are put in place.Mining itself begins with
extraction which requires drilling and blasting away surface rock, and loading
and hauling ore to leach pad. Leach pads and pools are prepared to contain
and extracted ore and capture gold in solution in the leaching process. The
leached solution is further refined in multiple stages, including a retort process
in which the mercury is separated. Pouring into dore bars completes the
processing steps that occur at the mine. Excess water from processing and
acid runoff from pit is treated before release at water treatment plants. To
prevent degradation of stream function sediment control structures are used
to capture sediments. Once an area becomes inactive it is filled with waste
rock, covered with top soil and in cases other protective layers, and replanted
during reclamation.

Data Collection
The mining process was modeled based on written and graphic descriptions in
corporate literature from primary sources. The model was corrected and/or confirmed
through visits to the mine in July 2007 and in conversations with mine employees.
Primary, public data from Newmont and partners were used as the source whenever
possible. When primary data was missing, inputs were calculated or 'back-calculated'
based on stoichometric formulas (for chemical reactions), equations in reference books


132









(for mine equipment, operations and infrastructure), or calculated using, when
necessary, generic industry data. Areas and distances utilized in calculations, when not
published in primary data, were estimated by delineating polygons of pertinent process
footprints from satellite imagery in Google Earth software, saving them as KML files,
and using a freely available web-based KML-polygon area calculator (GeoNews 2008).

Inventory Cutoffs
Rather than choosing a strict material, energetic or economic cutoff for data
collection, inventory cutoff was based on contribution to final measure of resource
impact from mining, measured in emergy. Inputs estimated to contribute to 99% of all
emergy were included. In many cases items with less than 1% of contribution to impact
were included, because lack of significance could not be assumed prior to calculation.
Many of these inputs were left in the inventory both to demonstrate their lack of
significance and to make the inventory more complete for use with other measures of
impact, for which relative impact would vary.

Data Management
The inventory data was managed in SimaPro 7.1 software (PRe Consultants
2008). Original processes and product stages were created for the primary unit
processes identified (Figure B-1) as well as for direct and indirect inputs to those
processes. For some input data was replicated from processes available in the
Ecoinvent database version 2.0 (Ecoinvent Centre 2007). The Ecoinvent database was
the only third-party data used to avoid boundary issues that would result from
incorporation of processes from other LCI databases available in SimaPro. Data
underlying Ecoinvent processes were altered in some cases, such as for heavy
vehicles, where the most analogous Ecoinvent process (e.g. lorry, 40 ton) was modified
with manufacturer data on weight to make it applicable to the mining process at
Yanacocha (e.g. rear dump truck). Only Ecoinvent data corresponding to 'Inputs from
Nature' or 'Inputs from Technosphere' were included, since these were relevant to the
scope of this LCI. Transport and excavation inputs were omitted for infrastructure items
adapted from Ecoinvent.
Processes were stored either as unit processes or system processes. Unit
processes were used in all cases except for those indirect processes (e.g. fabrication of
infrastructure) for which emergy values already existed, in which cases system
processes were used.
The process were named according to the following scheme: processes based
on primary data the name 'Yanacocha' was attached to the end. For processes based
on general estimates or calculation from the mining literature or other mines, no
additional ending was attached to the name. When inputs were prepared off site but
transportation to Yanacocha from their origin is included, the ending 'at Yanacocha' is
used. For processes that only stored unit emergy values, the name 'emergy' was
added to the end and if this unit emergy value did or did not include labor and services
'w/labor and services' or 'wout/labor and services' was attached to the names.


133










Results
The LCI consists of 164 SimaPro processes (Table B-16). 'Dore, at Yanacocha'
is the process for the final product (Table B-1), and 'Mercury, at Yanacocha' for the by-
product. All results are presented relative to the total production at the mine in 2005 of
2.17E+08 g dore which comprised 9.43E+07 g of gold, 1.23E+08 g of silver, and had
by-product of 5.99E+07 g of mercury. 'Mercury, at Yanacocha' is represented by an
identical process list except 'Processing, Yanacocha' is replaced with 'Processing,
without smelting, Yanacocha' sincercury is removed prior to smelting, and the 'Gold at
Yanacocha, geologic emergy' and 'Silver at Yanacocha, geologic emergy' processes
are replaced by the 'Mercury at Yanacocha, geologic emergy' process. 100% of all
mining inputs are allocated to both the dore and mercury by-products.

Table B-1. Inputs to process 'Dore, at Yanacocha'. Output is 2.17E+08 g dore.
No Process Amount Unit2"
1 Processing, Yanacocha 1 yr
2 Water Treatment, Yanacocha 1 yr
3 Gold at Yanacocha, geologic 9.43E+07 g
emergy
4 Silver at Yanacocha, geologic 1.23E+08 g
emergy
5 Exploration, Yanacocha 1 year
6 Mine infrastructure, Yanacocha 1/mine_lifetime p
7 Extraction, Yanacocha 1.33E+11 kg
8 Leaching, Yanacocha 1.20E+14 g
9 Sediment and dust control, 1 year
Yanacocha
10 Reclamation, Yanacocha (6.56E+10*waste_to_reclam)+8.3E+07 kg
11 Labor, total, Yanacocha 1 p
Notes
All variables with their default values are listed in Table B-24

Descriptions of the nine primary unit processes depicted in Figure B-1 and
procedures for collection of data associated with these process are presented by
process below.

Deposit Formation
The gold deposits at Yanacocha were formed by the flux of hydrothermal fluids
containing Au and other minerals from deeper within the crust. These fluids pushed up
and crystallized on near-surface rock that had been previously altered by flows of
magma. At Yanacocha, periods of volcanic activity producing magmatic flows alternated
with hydrothermal flows over approximately 5.4 million years created the deposits.
Greater depth and detail on the formation of gold deposits at Yanacocha is provided by

28 All symbols for units are the same as those used in SimaPro 7.1.


134









Longo (2005). The inventory for this process only contains the estimated mass of gold,
silver, and mercury in the final products.

Exploration
The exploration model consists of land-based exploration with a drill rig.
Inventory data is presented in Table B-2. Drill rig use is based on Newmont worldwide
ratio of oz reserve added to meters drilled, and reported reserve oz added at
Yanacocha (Newmont 2006b). This results in 0.8 m drilled/oz reserve added. Drilling
includes a diamond drill rig, diamond drills bits, and and water and diesel use for
operation. Drilling calculations are based on Hankce (1991). Water use is reported by
the company (Minera Yanacocha S.R.L. 2005). Initial exploration is done though aerial
surveys and remote sensing techniques, but this phase was not accounted for due to
lack of data. Support for exploration teams and sample processing was also omitted.

Table B-2. Inputs to process 'Exploration, at Yanacocha'. Output is 1 yr of exploration.


No. Process


Amount


Unit


0' 2eo


1 Process water, at Yanacocha 1.37E+11 g 1.2
Diamond exploration drill,
2 Yanacocha 50665 hr 1.3
3 Diamond drill bit 2.00E+02 p 1.3
4 Oil, refined, at Yanacocha 5.67E+13 J 1.3

Infrastructure
Inputs to mine infrastructure are presented in
Table B-3. Land use prior to mining was predominately pasture (Montgomery
Watson 2004). Loss of aboveground biomass due to clearing for mining is included.
Mine roads, water and electricity supply, and buildings were included in the inventory.
Total length and width of mine roads was estimated using satellite imagery. Models for
road materials and constructions were created for three roads types: (1) hauling roads
for use by heavy mine vehicles (approx 25m in width), (2) service roads (approx. 10 m
in width), and a provincial highway connecting Cajamarca and the mine which was
improved by the mining company for support of increased traffic and weight (Minera
Yanacocha S.R.L. 2007). Road models were based on standards in accordance for
support of vehicle weight and material type, based on California Bearing Ratios
obtained from Hartman (1992). Table B-17 provides assumed road layer depths. Road
materials and diesel used in transport of materials in road construction was included.
Materials were assumed to be gathered on site, at an average distance of 2.5 km,
based on visual estimate. Equations for transport of mine dump trucks (CAT 777C)
were used to estimated trips and fuel use (see next section). Material and fuel use for
the provincial highway were based on the 'Road/CH/I U' model in Ecoinvent (Spielmann
et al. 2004).
Estimations for an electricity supply network were based on Ecoinvent's
'Transmission network, electricity, medium voltage/km/CH/I' process (Dones et al.
2003). Water supply and a pump station were also based on Ecoinvent 'Pumpstation'
and 'Water supply network' processes (Althaus et al. 2004). Distance for electricity and


135









water supply networks were assumed equal to major mine road length (hauling road),
and total water supply was reported by the company (Newmont 2006a).
Total mine building area was estimated from satellite photos to the nearest 10000
m2. Inputs for process buildings were based on 'Building, hall, steel
construction/m2/CH/I' from Ecoinvent (Althaus et al. 2004).

Table B-3. Inputs to process 'Mine infrastructure, Yanacocha'. Output is 1p. *
No. Process Amount Unit O2 eo
1 Hauling Road, Yanacocha 44 km 1.5
2 Service Road, Yanacocha 110 km 1.5
3 Highway, provincial 3.60E+06 my 1.5
4 Building, hall, steel 3.00E+04 m2 1.5
5 Pump station 6.21 p 1.2
6 Water supply network 44 km 1.2
Transmission network, electricity, medium
7 voltage 44 km 1.5
Standing biomass before mining,
8 Yanacocha 7895 acre 1.5
'p' is the symbol for 1 item or unit in
SimaPro.

Extraction
The extraction phase model is based on a process descriptions reported by the
mining company (Minera Yanacocha S.R.L. 2005, 2006, 2007) and third parties
(Infomine 2005; International Mining News 2005; Mining Technology 2007). The
extraction phase commences with the removal and onsite storage of topsoil. Drill rigs
drill bore holes for placement of ANFO explosives for loosening overburden. Explosives
are assumed to be ANFO type (Newmont 2006a). Large mining shovels scrape
overburden and ore into large dump trucks. Overburden is transferred into waste rock
storage piles. Gold-bearing ore is transported and stacked on heap leach pads. The
total amount of ore mined, explosives used, percentage waste rock, and water used are
reported by Newmont (Minera Yanacocha S.R.L. 2005; Newmont 2006a). Inputs are
presented in Table B-4.

Table B-4. Inputs to process 'Extraction, Yanacocha'. Output is 1.99E+11 kg extracted
material.
No. Process Amount Unit O2geo
1 Scraper, Yanacocha' 596 hr 1.3
2 Drill rig, Yanacocha 2273 hr 1.3
3 Explosives (ANFO), at Yanacocha 7.71E+03 tn.sh 1.0
4 Mining shovel, Yanacocha 4.60E+04 hr 1.3
5 Rear dump truck, at Yanacocha 2.1 +E+05 hr 1.3
6 Oil, refined, at Yanacocha 2.83E+15 J 1.3
7 Process water, at Yanacocha 3E+11 g 1.2


136











Transport of Ore and Waste Rock
Models and makes of mine vehicles were confirmed from the primary and
secondary sources listed in the previous paragraph. Weight and capacity specifications
for these vehicles were acquired from vehicle manufacturers. Fuel economy was
estimated from data for another Newmont mine (Newmont Waihi Gold 2007). These
specifications were used as parameters for vehicle production equations from the SME
Mining Engineering Handbook (Lowrie 2002), for estimating total hours of use for
scrapers, mechanical shovels, dump trucks, and stackers (see Table B-19). The
estimated number of hours of use of each vehicle was then used to estimate fuel
consumption.

Mine Vehicle Model
Fabrication and transport of mine vehicles was included in the inventory.
Material composition, electricity and gas used in fabrication of mine vehicles were
scaled up from a simplified version of the 'Lorry 40t/RER/I U' process in Ecoinvent v1.3
((Spielmann et al. 2004), based upon the difference in weight. Only mass inputs into
the 'Lorry 40t/RER/I U' that comprised at least 1% of the total input weight were
included, with the addition of copper, lead, electricity, and natural gas. Materials were
aggregated together in the case of iron (e.g. weights of wrought iron and pig iron were
combined under the input 'iron'). A set percentage of the weight increase from
manufacturer of larger vehicles was attributed to steel for all vehicles (40% of weight)
and rubber for vehicles (7% of the weight) with larger tires including the rear dump truck
and scraper. Remaining additional weight was assumed to have the same composition
as the 40 ton lorry. Vehicle models including weights and lifetimes and equations for
scaling weights of materials and energy in vehicle fabrication are given in Table B-20.

Leaching
The leaching process at Yanaococha is a hydrometallurgical process whereby a
dissolved cyanide solution is dripped through gold and silver-bearing ore to strip these
metals and collect them in lined pool before being pumped out for further processing.
Total leached solution processed in 2005 was 1.21 E+14 g (Condori et al. 2007). The
leaching process is a circular process whereby barren solution (from CIC plant) is
recycled after replenishment with cyanide. A stacker is used to stack the extracted and
delivered ore on the leach pads. Estimated use is based on ore quanity and SME
Reference Handbook equations (see Table B-19). A total of 4845.5 tons as of sodium
cyanide as CN were consumed in this process in 2005 (Newmont 2006a). This was
multiplied by molecular weight ratio of NaCN:CN to get estimated NaCN used. Calcium
hydroxide, or lime, is added to raise the pH for optimal leaching. The estimated quantity
of lime is based on an addition of .38 g CaOH:kg ore, which matches the total use
reported by Newmont (Newmont 2006a) and is consistent with the range of 0.15-0.5
gCaOH:kg ore reported in Marsden and House (2006). Use of the leachpads and pool
were based on a ratio of ore capacity to total pad area (Buenaventura Mining Company
Inc. 2006). Details on leach pad and pool facilities were obtained from a mine tour and
primary sources (Minera Yanacocha S.R.L. 2007; Montgomery Watson 1998). Leach
pads consists of a clay layer, two layers of geomembranes, a gravel layer and collection

137









and conveyance pipes. These inputs were estimated based on area and specifications.
Total leach pad and pool areas in 2005 were reported by Buenaventura Mining
Company Inc. (2006). The leach pad process is based on the largest pad at La Quinua.
Fuel used in transport of the gravel from China Linda lime plant (12 km) and of the clay
from borrow pits within the mine (2.5 km) was estimated assuming dump truck
equations (Table B-19), assuming use of a CAT 777C with a fuel economy of 129L/hr.
Pipe network for leachate irrigation was not included. Leach pools for collecting
leachate prior to processing consist of three layers of geomembranes, a geotextile,
pipes for collection and pumping to treatment, and storage tanks for NaCN and mixing.

Table B-5. Inputs to process 'Leaching, Yanacocha'. Output is 1.21E+14 g leachate.
No. Process Amount Unit O2geo
1 Stacker, Yanacocha 1.54E+05 hr 1.3
2 Sodium cyanide, at Yanacocha 6.74E+09 g 1
3 Lime, loose, hydrated, at Yanacocha 4.6E+10 g 1.2
4 Process water, at Yanacocha 4.23E+12 g 1.2
5 Leach Pad, Yanacocha 6.69E+05 m2
6 Leach Pool, Yanacocha 3.28E+04 m2
7 Recycled leach solution 1.25E+14 g


Table B-6. Inputs to process 'Leach Pad, Yanacocha'. Output is a 2.1 E+6 m2 leachpad.
No. Process Amount Unit
1 Geomembrane, HPDE, 2mm thickness 2.10E+06 m2
2 Scraper, Yanacocha' 1.86E+03 hr
3 Geomembrane, LLPDE, 2mm thickness 2.10E+06 m2
4 HDPE Pipe, 40" dia. 6.67E+04 m
5 Fill material, Yanacocha 8.00E+08 kg
6 Gravel, crushed and washed, Peru 1.12E+09 kg
7 Oil, refined, at Yanacocha 1.63E+15 J


Table B-7. Inputs to process 'Leach Pool, Yanacocha'.
leachpool.


Output is a 1.03E+05 m2


No. Process Amount Unit
1 Geomembrane, HPDE, 2mm thickness 4.81E+04 m2
2 Geomembrane, LLPDE, 1mm thickness 1.03E+05 m2
3 Geomembrane, HPDE, 1.5mm thickness 2.06E+05 m2
4 Steel Pipe, 36" dia., at Yanacocha 2.74E+04 m
5 Geotextile, 8 oz. 3.09E+05 m2
6 Steel Pipe, 36" dia., at Yanacocha 1.70E+04 m
7 Storage tank, steel 1.50E+04 kg


138









Processing
Gold-bearing leachate is further processed and refined on site into dore. The
process train includes carbon-in-column adsorption and stripping, Merrill-Crowe
precipitation, retorting, and smelting (Mimbela 2007). Wastes from these various stages
go into process water treatment. These stages are aggregated together in an inventory
process called 'Processing, Yanacocha' (
Table B-8). Processing is assumed to be the major consumer of electricity.
Electricity is purchased by the mine from the national grid. Provision of electricity was
modeled after the national feedstock mix for Peru (Energy Information Administration
2007).

Table B-8. Inputs to process 'Processing, Yanacocha'. Ouput is 1 yr of processing.
No. Process Amount Unit
1 CIC process solution, Yanacocha 1.06E+13 g
2 Merrill Crowe process, Yanacocha 1.16E+13 g
3 Smelting, Yanacocha 2.17E+08 g
3 Retort process, Yanacocha 1.16E+13 g
4 Electricity, at powerplant, Peru 1.07E+06 GJ

The inputs included for the CIC process was activated carbon and the CIC plant
infrastructure. A ration of 4 g Au: 1000g activated carbon with a reuse rate of 90% of
the carbon was assumed ('Carbon in pulp', 2008). For the Merrill Crowe process,
1.89E+08 g of zinc powder and 4.45E+08 g of lead acetate are assumed to be included.
Estimates are based on ratios from Lowrie (2002). The retort process is merely an
empty place holder. The smelting process includes two smelters in addition to 1.68E+03
GJ natural gas, an amount based on a calculation of the energy necessary to heat gold
to its melting point of 1337K, assuming a heat capacity of 25.4 J mol-1 K1, and the
operational parameters of the smelter (see below).

Mass Balance Model
A dynamic mass balance model was used to track the fate of core species
through the process train (see Table B-21). Company reported concentrations of
elements in the feedstock at various stages and concentrations of reagents used were
set as constants in the model (e.g. Water used in process; cyanide used; ppm CN in the
leachate; gold and silver in final product). Other ranges of concentrations not reported
were gathered from the literature and upper and lower limits were used as constraints.
Recycle loops back to the leaching process exists at each stage, as the solution is
reused in the process. Values for unknown quantities were manipulated within upper
and lower limits until all mass balance conditions were satisfied, within an error of 2%
for water flows, and up to 5% for constituents.
The following species were tracked through the processing stages: H20
(including pumped water and precipitation), CN, Au, Ag, Hg, and Cu, primarily to
account for the various reagents used in the treatment chain, including activated
carbon, zinc and lead acetate (for precipitation in the presence of lead acetate), and to
account for the quantities of reagents used in treatment of the process water.


139









Process Infrastructure
Significant components of processing and water treatment infrastructure were
included based on estimates during a site visit and through measurements of geo-
referenced aerial photographs (Google 2008). Infrastructure includes storage and
processing tanks and steel buildings. Tanks were assumed to be steel and weights
were estimated from formulas from The Tank Shop (2007). Other process capital
components included in the inventory were 2 tilting electric-arc furnaces for smelting
and a reverse osmosis membrane treatment system for process water. The tilting
furnace was based on the Lindberg 61-MNP-1000 model.29 For simplicity the furnace
was assumed to be 100% steel.


Water Treatment
Water treatment at Yanacocha consists of treatment of process water and
treatment of acid water from previously mined open pits and reclaimed pits. Treatment
occurs in separate facilities. The process 'Water treatment' aggregates the treatment
type, plus includes reported additional acid use in excess of the modeled requirements
from the mass balance model (Table B-9).

Table B-9. Inputs to process 'Water Treatment, Yanacocha'. Output is 1 yr of water
treatment.
No. Process Amount Unit
1 Acid Water Treatment, Yanacocha 1.42E+13 g
Conventional Process Water Treatment,
2 Yanacocha 7.02E+12 g
Reverse Osmosis Process Water Treatment,
3 Yanacocha 4.68E+12 g
3 Acid,Yanacocha, unaccounting for 1.08E+09 g

Table B-10. Inputs to process 'Conventional Process Water Treatment, Yanacocha'.
Output is 3.1E+12g treated water.
No. Process Amount Unit O2geo
1 Chlorine, at Yanacocha 1.17E+10 g 1.2
2 Iron(lll) Chloride 3.02E+08 g 1.2
3 Sodium hydrosulfide, 100% 3.62E+07 g 1.2
4 Polyacrylamide (PAM) 3.00E+08 g 1.2
5 Sulfuric acid, 98%, emergy w/out L&S 4.91E+04 g 1.2
6 Electricity, at powerplant, Peru 1.16E+06 kWh 1.31
Conventional Process Water Treatment Plant,
7 Yanacocha 0.05 p


29 Approx. weight 8000 Ibs empty. Uses maximum of 3,100 cf per hr of natural gas based on 1,000 Btu/cf
natural gas. Max load 2,800 Ibs. Melt time for this load about 3 hrs (Hosier 2008).


140









Table B-11. Inputs to process 'Reverse Osmosis Process Water Treatment,
Yanacocha'. Output is 5.55E+12 g treated water.
No. Process Amount Unit O2geo
1 Chlorine, at Yanacocha 2.09E+10 g 1.2
2 Sulfuric acid, 98%, emergy w/out L&S 5.40E+04 g 1.2
3 Electricity, at powerplant, Peru 1.20E+14 J 1.31
4 RO System 1.71 p

Table B-12. Inputs to process 'Acid Water Treatment, Yanacocha'. Ouput is 1.42 E+13g
treated water.
No. Process Amount Unit O2geo
1 Lime, loose, at Yanacocha 7.96E+09 g 1.2
2 Iron(lll) Chloride 7.10E+08 g 1.2
3 Polyacrylamide (PAM) 9.22E+08 g 1.2
4 Sulfuric acid, 98%, emergy w/out L&S 2.24E+04 g 1.2
5 Electricity, at powerplant, Peru 2.74E+06 kWh 1.31
6 Acid Water Treatment Plant, Yanacocha 0.05 p

Water treatment process models are based on site visits and personal communication
with engineers at Yanacocha. Process water treatment included both conventional and
reverse osmosis systems. Allocation between these systems is based on installed
capacity in 2005. Chemical reagents used in these processes are included. Reagents
quantities are based on reported quantities used when available or calculated based on
total water treated and requirements specified in water treatment literature. Sludge
waste from treatment is slurried and pumped back to the leach pads no additional
long-term management for sludge is included other than leach pad reclamation, as none
is planned.
Conventional process water treatment inputs were based on the following.
Chlorine calculations were based on the stochiometric calculation of 4 mol Cl per mol
CN, with an excess ratio of 1.1 mol Cl (National Metal Finishing Resource Center
2007). NaSH is added to release cyanide bound to copper. Inputs is based on the
stochiometric equation from Coderre and Dixon (Coderre and Dixon 1999). PAM added
is based on an optimal concentration of 65 ppm (Wong et al. 2006). The sulfuric acid
addition is based on a stochiometric requirement to adjust the pH of the water.
Electricity of 0.193 kWh/ m3 of process water is adapted from Ecoinvent 'Treatment,
Sewage to Wastewater'. Iron chloride added is based on a concentration of 55 ppm
(Abou-Elela et al. 2008).
The reverse osmosis process only requires the addition of CN to destroy cyanide
and sulfuric acid to adjust the pH after treatment. It does require additional electricity.
The assumed electricity requirement was 6 kWh/m3 treated water.
Acid water treatment is assumed similar to process water treatment, without the
addition of chlorine for cyanide destruction, and with the addition of additional lime for
pH treatment. Lime added is based on the lime needed to adjust the pH of the influent
from 2-11.


141









Reclamation
Reclamation models are based on primary data on restoration methods and long-
term mine closure plans (Montgomery Watson 2004; Montoya and Quispe 2007). Total
reclamation amount is based on the total amount of waste rock (material extracted),
which is the difference between total extraction and total ore to leachpads. Inputs are
all estimated relative to the mass of overburden returned to mining pits. All waste rock
was assumed to be loaded from waste rock piles, transported and backfilled in pits, and
limed at a ratio of 1gCaOH:1 kg fill. Fuel consumption for mining shovels and dump
trucks is included and based on mining equations (Table B-19). Protective layering,
capping, seeding/planting and reclamation maintenance activities were not included
due to assumption of insignificance to entire process (< 1%). Inputs to reclamation are
shown in Table B-13.

Table B-13. Inputs to process 'Reclamation, Yanacocha'. Output is 1 kg of returned
overburden.
No. Process Amount Unit o2geo
1 Lime, loose, at Yanacocha 1 g 1.2
2 Rear dump truck, at Yanacocha 1.32E-06 hr 1.3
3 Mining shovel, Yanacocha 2.33E-07 hr 1.3
4 Oil, refined, at Yanacocha 9.79E+03 J 1.3


Sediment and Dust Control
The primary measures taken at Yanacocha to reduce sediment in runoff are serpentine
structures immediately adjacent to mine facilities and three large sediment dams.
Sediment runoff is based on sediment storage capacity in dams and dam lifetime.
Thirteen serpentines are reported (Campos 2007). Dimensions of a representative
serpentine were estimated from satellite imagery (Google 2008).Serpentines were
assumed to be constructed of 1540 m3 reinforced concrete. Flocculants to cause
sediments to drop out of the water column were not included. Reinforced concrete was
also the only input included in sediment dams. Total concrete volume was reported as
7000 and 3000 m3 for the Grande and Rejo dams, respectively (Newmont 2004).
Concrete for the Azufre dam, not reported, was estimated as the average of the
aforementioned dams. The contribution of these structures is annualized over the
assumed mine lifetime of 25 years.

Mine roads are regularly watered to reduce particulates in the air. The amount of water
used by the mine in dust control was reported (Minera Yanacocha S.R.L. 2005). An
evaporation rate of 50% was assumed for water spayed on roads, and only this water, a
total of 1.34 E+11 g, was included.


142










Table B-14. Inputs for process 'Sediment and dust control, Yanacocha'. Output is 1 yr.
No. Process Amount Unit
1 Sediment control structures, Yanacocha 0.04 p
2 Dust control, Yanacocha 1 year


System Level Inputs
Because labor was not reported by unit process, it was included as a system level input,
and appears in the 'Dore, at Yanacocha' process (see Table B-1. Inputs to process
'Dore, at Yanacocha'. Output is 2.17E+08 g dore.Table B-1).

Labor
Energy in labor was included based on the total hours worked and average human
energetic consumption. Total hours worked by employees and contractors is reported
by the company (Newmont 2006a). Total J of energy in human labor at Yanacocha was
calculated as:

(3.82E+09 J/yr avg human consumption)/(365*8 working hrs/yr)(2.3E+07 hrs worked at
Yanacocha) = 3.01E+13 J/yr (1)

A year's calorie intake is assumed necessary to support 8 hours of work daily for 365
days a year.

Transport
Transport of materials and capital goods making up 99% of the mass of all inputs
was considered. Sea, land, and air transport were all included. Inputs to transport
included transport infrastructure construction and operation.
Transport distance was based on origin of the item if known. If unknown, origin
was first determined to be domestic or foreign by consultation of the Peru statistical
companion for domestic production data and United Nation trade data for import-export
data (Instituto Nacional Estadistica y Informacion 2006; United Nations 2008). If the
item was produced or exported in quantities sufficient to supply the usage at
Yanacocha, origin was assumed domestic and assumed to originate in Lima. If item
was assumed to be of foreign origin, a sea distance of 5900 km was assumed (Los
Angeles to Lima) in addition to road transport from Lima. Top ten items, mass inputs,
and transport distances are given in Table B-23.
Inputs for sea and air transport were based on the Ecoinvent processes
'Transport, transoceanic freight ship/OCE U', 'Transport, transoceanic tanker/OCE U',
and 'Transport, aircraft, freight, intercontinental/RER U' (Spielmann et al. 2004). An
inventory of US truck transport from Buranakarn (1998) was adapted with data from
Spielman and data on the Peruvian truck fleet (Instituto Peruano de Economia 2003).
Data and notes are given in Table B-22. Due to complex geography, an older fleet, and
significantly less transport, ton-km efficiency was assuming to be 50% of that of the
United States.


143









Life Cycle Model Parameters
Various life cycle parameters can be switched to include or exclude input of
geologic emergy of ore, to clay and gravel construction material. By default these inputs
are switched to '0', indicating they are not included. Lifetime of all mine-infrastructure
and long-term activities such as reclamation are based on the 'mine_lifetime' variable,
which is set to 25 years, representing the time the mine area is occupied and run by the
company. The 'process_lifetime' variable is used for capital goods used processes, and
represents the time of active mining and processing at the mine, and is set by default to
20 yrs. 'Waste_to_reclam' is the fraction of waste rock backfilled in reclamation and is
by default set to '1', representing 100%. Other parameters are (1) related to the size of
leach pad and carrying capacity and are used for leach pad capital estimations; (2)
related to the mine vehicle models; (3) the ore grade at Yanacocha (Au_ore_grade); (4)
the percent of process water treated with reverse osmosis (per_RO_treat); and (5) the
way that emergy of labor is included. Parameters are given in Table B-24.

Uncertainty
The inventory estimates were complemented with uncertainty ranges for direct
inputs to the nine primary unit processes. For these inputs, uncertainty range was
estimated using the same model specified for the Ecoinvent v2.0 database
(Frischknecht et al., 2007). This model assumes inventory data fit a log-normal
distribution, and that uncertainty can be estimated according to six factors: reliability,
completeness, temporal correlation, geographic correlation, technological correlation,
and sample size. The uncertainty is reported as the square of the geometric standard
distribution, 02. Uncertainty estimates are presented in Table B-25. Model parameters
related to lifetime of operations were also assigned ranges. Parameters for mine
infrastructure, transport distances, and mine vehicle models were estimated with the
Ecoinvent method. For processes based on Ecoinvent data, uncertainty data was
perpetuated from Ecoinvent processes.

Emergy Conversions
All system processes containing in their name 'emergy' consisted solely of an
emergy input, listed as an 'Input from Nature', estimated in units of solar emjoules (sej).
These processes served as conversion factors between inventory units and emergy
values (e.g. 1.1 E+05 sej per J of refined oil), commonly called unit emergy values
(UEVs). The UEVs were applied in order to calculate total environmental contribution as
energy in sunlight equivalents. Sources for emergy values per unit input were based on
previous emergy evaluations of an identical or similar product.
Like inventory values, UEVs were assigned an error range, due to uncertainty in
the equivalence of the product, uncertainty in processes in nature, or due to
methodological differences in emergy calculations. A log-normal distribution is
assumed for the UEVs.

Discussion
This inventory may be directly compared with an existing process in the
Ecoinvent database 'Gold, from combined gold-silver production, at refinery/PE U'
(henceforth 'Gold .... /PE U') and its accompanying description (Classen et al. 2007),
which is also based on production at the Yanacocha mine.


144









This study reports a total production of 9.43E+07 g of gold in dore while the 'Gold
.... /PE U' process assumes 1.03E+08 g gold in dore. In the 'Gold .... /PE U' process,
the inventory data has already been allocated between gold and silver in the dore. This
process assumes an additional inputs for separating the gold from the silver in the dore.
In this study, the inventory data has not been pre-allocated between gold and silver.
The structure of this inventory is much more elaborate than that of the 'Gold ....
/PE U' process in Ecoinvent. The Ecoinvent process is essentially a system process,
where inputs to dore production are all grouped under the aforementioned process.
This inventory is based on nine unit processes, each of which have additional unit
processes contributing to them.
The 'Gold .... /PE U' process does not consider any inputs into deposit formation, or
exploration. Mine infrastructure in the Ecoinvent process is based on a generic
Swedish mine. In this study major infrastructure, such as mine building, roads, and
processing structures, are based on original analysis of the mine site. The remaining
infrastructural components, included power delivery and water supply, are based on
generic Ecoinvent processes. For extraction, the 'Gold .... /PE U' process does not
estimate the contribution of mine vehicles. For leaching, the 'Gold .... /PE U' process
does not include the leach pad and pool architecture or its construction. For processing,
the 'Gold .... /PE U' process does not include the leach pad and pool architecture or its
construction. In this inventory, reagents added during processing and water treatment
are based on mass balance calculations of the process. This inventory explicitly
includes some of the major components of the process, water treatment, and sediment
control infrastructure at Yanacocha, which are missing from the 'Gold .... /PE U'
process. There are other notable differences in the inventories. Land use and
transformation are not included as inputs in this study, but are included in the 'Gold ....
/PE U' process. Standing biomass from land transformation, however, is included in
this inventory. This is only a source-side LCI, but the 'Gold .... /PE U' process includes
estimates of emissions to air and water.
The electricity mix in the 'Gold .... /PE U' process is based on the Brazilian electricity
mix. In this study a new electricity mix process specific to Peru was created. The
assumed mine lifetime presents a significant difference between the inventories, which
effects the contribution of all capital goods and infrastructure. The 'Gold .... /PE U'
process assumes a mine lifetime of 50 years; this study only 25 years. A comparison of
the outputs and direct non-durable inputs to mining in reference to output of 1 g of dore
is presented in Table B-15.


145









Table B-15. Comparison of this inventory with the equivalent Ecoinvent process
this inventory Ecoinvent v2.0
No Item 'Dore, at Yanacocha' 'Gold .... /PE U' Unit
Total production
1 Gold 9.43E+04 1.03E+05kg
2Silver 1.23E+05 3.67E+04kg
3Dore 2.17E+05 1.40E+05kg
Rel. to dore production 100% 64%
Direct non-durable inputs to 1 g of dore
4 Electricity 6.77 12.3MJ
5Diesel 18.4 47.7MJ
6Sodium Cyanide 30.8 42.9MJ
7Lime 0.55 1.17g
8Sodium hydroxide 0 52.6g
9Activated carbon 6.73 17.1 g
10Zinc 0.873 3.33g
11 Sulfuric acid 6.74 7.67g
12 Hydrochloric acid 6.75 Og
13Transport, truck 0.352 1.92tkm
14 Explosives 0.032 0.416kg
15 Water 0.022 0.016m3
16 Lead acetate 2.05 Og
17Chlorine 0.203 0kg
18Sodium hydrosulfide 0.378 Og
19 Iron chloride 6.430 Og
20Polyacrylamide 7.38 Og
Notes
4.61E-9 p of 'Dor6, at Yanacocha' (=1/annual production, g) and 1.006 g of'Gold .... /PE U'
(=1/99.4 % allocation to gold) were compared here as each represent 1 g of dore. Post-dor6
electricity and transport included in 'Gold .... /PE U' are omitted for comparison.
Item references (format: this inventory; Ecoinvent)
4'Electricity, at powerplant, Peru'; 'Electricity Mix/BR' from Ecoinvent
5'Oil, refined, at Yanacocha'; 'Diesel, burned in building machine /GLO U'
6'Sodium cyanide, at Yanacocha'; 'Sodium cyanide, at plant/RER U'
7'Limestone, loose and hydrated, at Yanacocha'; 'Lime, milled, packed, at plant'
8NA; 'Sodium hydroxide, 50% in H20, production mix, at plant/RER U'
9'Activated carbon'; 'Charcoal, at plant/GLO U'
10'Zinc, geologic emergy'; 'Zinc, primary, at regional storage/RER U'
11 'Sulfuric acid, 98%, emergy w/out L&S', 'Sulphuric acid, liquid, at plant/RER U'
12 NA, 'Hydrochloric acid, liquid, at plant/RER U'
13'Transport, truck, Peru';'Transport, lorry >16t, fleet average/RER U'
14'Explosives (ANFO), at Yanacocha'; 'Blasting/RER U'
15'Process Water, Yanacocha'; 'Water, river' + 'Water, well, in ground'
16'Lead Acetate'; NA
17'Chlorine, at Yanacocha'; NA


146










18'Sodium hydrosulfide, 100%'; NA
19'lron chloride'; NA
20'Polyacrylamide'; NA

Due to the difference in output one would expect the values in the 'Gold .... /PE U'
process to be 1.58 times greater than those in this inventory, but there are still
discrepancies beyond this difference. Electricity, diesel, lime, activated carbon, zinc,
truck transport and explosives are all greater in the Ecoinvent inventory than expected.
Sodium cyanide, sulfuric acid, and water use are less than the expected difference.


Appendix
Table B-16. List of processes in the 'Gold


Yanacocha' project inventory.


No. Process Unit No. Process Unit


Acid Water Treatment Plant, Yanacocha
Acid Water Treatment, Yanacocha
Acid,Yanacocha, unaccounting for
Activated carbon
Aircraft, long haul
Airport


7 Aluminum ingot, emergy w/out labor & g
services
8 Ammonium nitrate, emergy w/out labor & g
services
9 Ammonium, emergy w/out labor and services g


10 Antifreeze


11 Azufre Dam, Yanacocha
12 Bitumen, emergy w/out labor and services
13 Brass, emergy w/out labor & services
14 Brick, emergy w/out labor and services
15 Bronze, emergy w/out labor & services
16 Building, hall, steel

17 Cement, emergy w/out labor and services

18 Chlorine, at Yanacocha
19 Chlorine, emergy w/out labor and services

20 CIC plant, Yanacocha


CIC process solution, Yanacocha
Clay, in ground, geologic emergy
Concrete, at Yanacocha


24 Concrete, emergy w/out labor and services
25 Conventional Process Water Treatment Plant,
Yanacocha
26 Conventional Process Water Treatment,
Yanacocha
27 Copper, emergy w/out labor & services


83 Mercury, in ground, geologic emergy g
84 Merrill Crowe plants, Yanacocha p
85 Merrill Crowe process, Yanacocha g
86 Mine infrastructure, Yanacocha p
87 Mining shovel, Yanacocha hr
88 Natural gas, emergy w/out labor & J
services
89 Oil, crude, emergy w/out labor & J
services
90 Oil, refined, at Yanacocha J

91 Oil, refined, emergy wout/labor & J
services
92 Operation, aircraft, freight, tkm
intercontinental
93 Operation, maintenance, airport p
94 Operation, maintenance, port p
95 Operation, transoceanic freight ship tkm
96 Operation, transoceanic tanker tkm
97 Paint, emergy w/out labor and services g
98 Pesticide, orthophosphate, emergy g
w/out labor and services
99 Pig iron, emergy w/out labor and g
services
100 Polyacrylamide g
101 Polybutadeine rubber, emergy w/out g
labor & services
102 Polystyrene, emergy w/out labor and g
services
103 Polyurethane g
104 Port Facilities p
105 Primary steel, emergy wout/labor & g
services
106 Process water, at Yanacocha g
107 Processing without smelting, year
Yanacocha
108 Processing, Yanacocha year

109 Pump station p


147










28 Diamond drill bit
29 Diamond exploration drill, Yanacocha


30 Diamond, in ground, geologic emergy g
31 Dore from Yanacocha PE, at CH g
32 Dor6, at Yanacocha g
33 Drill rig, Yanacocha hr
34 Dust control, Yanacocha year
35 Electricity from coal, emergy w/out labor and J
services
36 Electricity from hydro, emergy w/out labor and J
services
37 Electricity from natural gas, emergy w/out J
labor & services
38 Electricity from nuclear, emergy w/out labor J
and services
39 Electricity from oil, emergy w/out labor and J
services
40 Electricity, at powerplant, Peru J
41 Electricity, at powerplant, USA J


42 Emergy in dollar, Peru, 2004

43 Ethylene-propylene rubber (EBR), emergy
w/out labor and services
44 Exploration, Yanacocha
45 Explosives (ANFO), at Yanacocha
46 Extraction, Yanacocha


USD

g

year
kg
kg


47 Fill material, Yanacocha g
48 Generic inorganic acid, 100%, emergy w/out g
labor and services
49 Generic organic chemical, emergy w/out labor g
and services
50 Geomembrane, HPDE, 1.5mm thickness m2
51 Geomembrane, HPDE, 2mm thickness m2


52 Geomembrane, LLPDE, 1mm thickness
53 Geomembrane, LLPDE, 2mm thickness
54 Geotextile, 8 oz.

55 Glass, emergy w/out labor and services
56 Gold in dor6, at Yanacocha

57 Gold, in ground, at Yanacocha, geologic
emergy
58 Grande Dam, Yanacocha
59 Gravel, crushed and washed, Peru

60 Ground water, emergy

61 Hauling Road, Yanacocha
62 HDPE Pipe, 40" dia.
63 HDPE, emergy w/out labor & services


m2
m2
sq.y


110 PVC, emergy w/out labor and services
111 Quicklime, emergy w/out labor and
services
112 Rear dump truck, at Yanacocha
113 Reclamation, Yanacocha
114 Recycled leach solution
115 Reinforced concrete, at Yanacocha
116 Rejo Dam, Yanacocha
117 Retort process, Yanacocha

118 Reverse Osmosis Process Water
Treatment, Yanacocha
119 RO membrane

120 RO System


121 Road construction, Peru


122 Road operation, Peru
123 Rock wool, emergy w/out labor and
services
124 Salt, NaCI 100%, emergy w/labor and
services
125 Sand, in ground, geologic emergy

126 Scraper, Yanacocha'
127 Sediment and dust control, Yanacocha
128 Sediment control structures,
Yanacocha
129 Serpentine, Yanacocha
130 Service Road, Yanacocha

131 Silt, in ground, geologic emergy

132 Silver in dor6, at Yanacocha
133 Silver, in ground, at Yanacocha,
geologic emergy
134 Smelters, Yanacocha
135 Smelting, Yanacocha
136 Sodium cyanide, at Yanacocha

137 Sodium hydrosulfide, 100%
138 Sodium hydroxide, 100%, at
Yanacocha
139 Sodium hydroxide, 100%, emergy
wout/labor and services
140 Stacker, Yanacocha
141 Standing biomass before mining,
Yanacocha
142 Standing biomass, tropical savannah,
emergy
143 Steel Pipe, 36" dia., at Yanacocha
144 Storage tank, steel
145 Sulfuric acid, 98%, emergy w/out labor


148


kmy

kmy
g

g

g

hr
year
p

p
km

g

g
g

p
g
kg

kg
g

g

hr
m2

g

ft
g
g











64 Heavy Vehicle
65 Highway, provincial
66 Hydrochloric acid, 100%, emergy w/out labor
and services
67 Hydrogen cyanide
68 Hydrogen sulfide, emergy w/out L&S

69 Iron ore, emergy w/out labor and services
70 Iron(lll) Chloride
71 Labor, Peru, emergy

72 Labor, total, Yanacocha
73 Leach Pad, Yanacocha

74 Leach Pool, Yanacocha
75 Leaching, Yanacocha
76 Lead acetate
77 Lead, in ground, geologic emergy
78 Lime, loose and hydrated, at Yanacocha

79 Limestone, in ground, geologic emergy
80 Lumber, emergy w/out labor and services
81 Mercury, at Yanacocha
82 Mercury, in ground, at Yanacocha, geologic
emergy


and services
146 Sulphur hexaflouride
147 Surface water, emergy
148 Tetrafluoroethylene

149 Tilting Furnace
150 Transmission network, electricity,
medium voltage
151 Transoceanic freight ship
152 Transoceanic tanker
153 Transport of Dore, Yanacocha to
Switzerland
154 Transport truck, operation, Peru
155 Transport, aircraft, freight,
intercontinental
156 Transport, aircraft, freight, Peru
157 Transport, transoceanic freight ship
158 Transport, transoceanic tanker
159 Transport, truck, Peru
160 Transport, truck, USA, emergy w/out
labor and services
161 Water supply network
162 Water Treatment, Yanacocha
163 Wood preservative
164 Zinc, in ground, geologic emergy


Table B-17. Mine hauling road parameters, based on Hartman (1992).
Thickness Cross-sectional
Course (m) Material area (m2)
Surface 0.1 Gravel 2.5
Base 0.1 Clay-sand-silt 2.5
Subbase 0.5 Clay-sand-silt 12.5

Table B-18. Mine service road parameters, based on Hartman (1992).
Thickness Cross-sectional
Course (m) Material area (m2)
Surface 0.1 Gravel 2.5
Base 0.1 Clay-sand-silt 2.5

Table B-19. Mining equations
Equation Reference1
Shovel and stacker loading production, loose m3/hr = SME, Equation
3600(Bucket capacity, loose m3)(efficiency)(fill factor)(propel 12.21
time factor)/(load cycle time, seconds)
Total shovel and stacker use, hrs = (m 'mine/yr/ loose m /hr) NA
Scraper load, m3 = (capacity, m )(swell factor, ratio of bank m3 SME, Equation 12.9
to loose m3)


149


p
km

p
p
g

km
tkm

tkm
tkm
tkm
tkm
tkm

km
year
g
g









Scraper travel time, min = (distance to soil storage, m)/(speed, SME, Equation
km/hr)(16.7 m-h/km-min) 12.18
Scraper cycle time, min = (load time,min)+(travel time,min*2)+ SME, Equation
(spread time,min) 12.19
Scraper production, m3/hr= (60)(bucket capacity, m )(operating SME, Equation
efficiency)/cycle time (hrs) 12.21
Scraper use, hrs (Topsoil to be moved, annualized)/(scraper NA
production)
Dump truck spot and load time, min = (spot time, min)+(passes- SME, Equation
1)(loading cycle time) 12.15
Travel time to dump point, min = (Distance,m)/(speed, SME, Equation
km/h)(16.7 m-h/km-min) 12.18
Dump truck cycle time, min= (load time) + (travel time) + (travel SME, Equation
time) + (dump time) 12.19
Dump truck production, m3/hr =(60)(haulage units)(load, bank SME, Equation
m3)(efficiency)/(cycle time,min) 12.21
Dump truck use, hrs = (ore mined, m /yr/ haulage production, NA
m3/hr)
Drill rig use, hrs/yr = (holes/layer)(layers/year)(digging, NA
hrs/hole+travel time, hrs/hole)
1All references with SME refer to the SME Handbook (Lowrie 2002).

Table B-20. Mine vehicle data
Lifetime
Type Manufacturer/Model Weight (kg)1 (hrs)2
Rear Dump Truck CAT 793D 166866 30000
Stacker CAT 325D w/boom 29240 14000
Scraper CAT 651 E 62000 14000
Mining shovel Hitachi EX5500 518000 90000
Drill rig Atlas Copco Simba 1250 11830 14000
1From manufacturer specifications
2 Estimated from (Lowrie 2002)


150












Table B-21.
STAGE

Input
Primary
H20
CN
Au
Ag
Hg
Cu
ppm Au
ppm CN
% CN solution
pH
Ag:Au ratio
Dore %Au
Dore %Ag


STAGE

Input
Primary
H20
CN
Au
Ag
Hg
Cu
ppm Au
ppm CN
% CN solution
pH
Ag:Au ratio
Dore %Au
Dore %Ag


Mass balance of leaching, processing, and water treatment.
1 LEACH
EXTERNAL INPUT RECYCLED INPUT EXTERN + RECYC
Mass (g) Mass (g) Mass (g)


1.42E+12
4.40E+09


Check ext+int H20
Recycled water needed to bal
H20 Recycle rate


TO AIR
Mass frac Mass (g)


0.03
0.03
0
0
0
0


1.25E+14
1.71E+09
2.31E+06
4.70E+07
1.50E+07


Ext+Rec CN (mass)

CN check
H20 check
1.26E+14
1.26E+14
98.89%


TO CARBON COL
Mass frac


4.23E+12
1.91E+08
0.00E+00
0.00E+00
0.00E+00
0.00E+00


0.75
0.75
0.50
0.22
0.04
0.60


check CN:Au ratio


Water check
Reported H20
Water Difference


PRECIP
Mass (g)


1.27E+14
6.35E+09


TOTAL INPUT
Mass (g)


1.38E+13


6.11E+09
11
96.18%
99.57%


TO MERRILL CROWE
Mass (g) Mass frac Mass (g)


1.06E+14
4.76E+09
5.82E+07
1.01E+08
4.23E+07
3.09E+09
0.55
45
82


1.20E+14
1.21E+14
98.94%


0.1
0.1
0.33
0.14
0.03
0.40


1.41E+13
6.35E+08
3.82E+07
6.64E+07
2.78E+07
2.03E+09
2.71


TO LEACH
Mass frac Mass (g)

0.12 1.69E+1
0.12 7.62E+C


1.41E+14
6.35E+09
1.16E+08
4.66E+08
1.00E+09
5.12E+09


Recycle Frac


98.46%
26.96%
1.98%
10.10%
1.50%
0.00%


4.00E+00





RESIDUAL
Mass frac Mass (g)


13
)8


0
0
0.172
0.64
0.93
0.7


0.00E+00
0.00E+00
2.00E+07
2.98E+08
9.31E+08
3.59E+09


151


KEY
Reported or calculated from reported Constrained Value
value
Check











STAGE

Input
Primary
H20
CN
Au
Ag
Hg
Cu
ppm Au
ppm CN
% CN solution
pH
Ag:Au ratio
Dore %Au
Dore %Ag


2 CARBON COLUMNS
INPUT TO MERRILL CROWE
Mass (g) Mass frac Mass (g)


1.06E+14
4.76E+09
5.82E+07
1.01E+08
4.23E+07
3.09E+09
0.551
45
4.00E+00


0.1
0.69
0.98
0.55
0.71
0.1


1.06E+13
3.29E+09
5.70E+07
5.57E+07
3.00E+07
3.09E+08
5.399
311


TO LEACH
Mass frac Mass (g)


0.9
0.31
0.02
0.45
0.29
0.9


9.51E+13
1.48E+09
1.16E+06
4.55E+07
1.23E+07
2.78E+09
0.012
16


3 MERRILL CROW
INPUT TO RETORT
Mass (g) Mass frac


2.46E+13
3.92E+09
9.52E+07
1.22E+08
5.78E+07
3.09E+08
4
159


0.47
0.94
0.988
0.988
0.988
0.988


TO LEACH
Mass (g) Mass frac Mass (g)


1.16E+13
3.69E+09
9.41E+07
1.21E+08
5.71E+07
3.06E+08
8
318


0.53
0.06
0.012
0.012
0.012
0.012


1.31E+13
2.35E+08
1.14E+06
1.46E+06
6.94E+05
3.71E+06
0.09
18


C (as activated carbon
Zn
Pb (as lead acetate)


1.46E+10


0 0.00E+00


1.42E+08
4.45E+08


0.66 9.34E+07
1 4.45E+08


0.33 4.67E+07


STAGE

Input
Primary
H20
CN
Au
Ag
Hg
Cu
ppm Au
ppm CN
% CN solution
pH
Ag:Au ratio
Dore %Au
Dore %Ag


4- RETORT
INPUT
Mass (g)

1.16E+13
3.69E+09
9.41E+07
1.21E+08
5.71E+07
3.06E+08
8


TO HG-PRODUCT
Mass frac Mass (g)


0
0
0
0
0.95
0

Check Hg


TO WWT
Mass frac Mass (g)


0.00E+00
0.00E+00
0.00E+00
0.00E+00
5.99E+07
0.00E+00

104.84%


1
1
0
0
0.01
0


1.16E+13
3.69E+09
0.00E+00
0.00E+00
5.71E+05
0.OOE+00


TO SMELT
Mass frac Mass (g)


0
0
1
1
0.04
1


0.00E+00
0.00E+00
9.41E+07
1.21E+08
2.07E+06
3.06E+08


5.26E+08


C (as activated carbon
Zn
Pb (as lead acetate)


9.34E+07
4.45E+08


1 9.34E+07
1 4.45E+08


152











STAGE


5 SMELT
INPUT


TO DORE-PRODUCT


Input Mass (g) Mass frac
Primary
H20 0.00E+00
CN 0.00E+00
Au 9.41E+07 1.
Ag 1.21E+08 1.
Hg 2.07E+06
Cu 3.06E+08
ppm Au
ppm CN
% CN solution Check recovery % Au
pH Check recovery % Ag
Ag:Au ratio
Dore %Au Percent Au in dore
Dore %Ag Percent Ag in dore
KEY
Reported or calculated from reported Constrained Value
value
Check


TO LEACH


TO WWT


Mass (g) Mass frac Mass (g) Mass frac Mass (g)

0 0.00E+00 0 0.00E+00 1 0.00E+00
0 0.00E+00 0 0.00E+00 1 3.92E+09
.00 9.43E+07 0 0.00E+00 -0.00273 -2.60E+05
.02 1.23E+08 0 0.00E+00 -0.02067
0 0.00E+00 1 2.07E+06 0 0.00E+00
0 0.00E+00 0 0.00E+00 1 3.09E+08


81.04%
26.44%

43.38%
56.62%


153









Table B-22. Inventory of Peruvian road transport.


No. Item
1 Trucks
Road Construction
2 Concrete
3 Bitumen
4 Gravel
5 Electricity
6 Diesel
Road operation
7 Electricity
8 Paint
9 Herbicide
Transport
10 Diesel consumption
Product


Flow
4.44E+10

6.00E+09
1.75E+10
2.42E+11
4.92E+11
1.18E+12

7.31E+09
6.04E+03
3.37E+02

8.90E+15


Unit
g

g
g
g
J
J

J
g
g

J


11 Annual yield of trucks 1.50E+09 ton-km
NOTES
Input references from Spielman et al. (2004)
Trucks
1 (Class 8 weight Ib)(class 8 trucks)*(Class 6 weight Ib)(class 6 trucks)*( 454 g/lb) / (10 yr lifetime)
4.44E+10 g Truck weights from Buranakarn (1998)
UEV from heavy mine vehicle
model
Highway construction
Demand by trucks of infrastructure creation
Good transport percent road
wear 0.424 Based on Swiss situation. Table 5-117.
road length=(length of road network, km)(14.4% paved)
(Economic Commission of Latin
Highway km 11351 American and the Carribbean 2006)
Improved unpaved km 18634
Concrete kg/ (m*yr) 37
Bitumen kg/ (m*yr) 15.4
Gravel for highway subbase kg/ (m*yr) 470
Gravel for unpaved road surface kg/ (m*yr) 101.25
Lifetime
Concrete yr 70
Bitumen yr 10
Gravel for highway subbase yr 100
Gravel for unpaved road surface yr 10
Standard Equation for road materials
(Good transport percent road wear)(material kg/m*yr)(road length km) (1000m/km)
(1000g/kg) / (material lifetime yr)
2 Concrete g 6.00E+09
3 Bitumen g 1.75E+10
4 Gravel g 2.42E+11


154










Electricity for highway constr. MJ/m*yr 98.7 Motorway. Table 5-94.
Electricity for unpaved road
constr. MJ/m*yr 2.18 2nd class road. Table 5-94.
(Good transport percent road wear)(energy MJ/m*yr)(road length km) (1000m/km) (1E+6 J/MJ)

5 Electricity for construction J 4.92E+11
Diesel for highway construction MJ/m*yr 192 Motorway. Table 5-94.
Diesel for unpaved road
construction MJ/m*yr 33 2nd class road. Table 5-94.
(Good transport percent road wear)(energy MJ/m*yr)(road length km) (1000m/km) (1 E+6 J/MJ)
6 Diesel J 1.18E+12


Operation
Demand by trucks of infrastructure operation
Good transport percent road use


0.103 Based on Swiss situation. Table 5-117.
Motorway. Table 5-


Electricity for highway operation KWH/m*yr 0.67 101.
Electricity for unpaved road
operation KWH/m*yr 3.4 2nd class road. Table 5-101.
(Good transport percent road use)(electricity use KWH/m*yr)(road length km) (3600000 J/KWH)
7 Electricity for operation J 7.31E+09
Paint for highway operation kg/m*yr 0.00517
(Good transport percent road use)(paint usekg/m*yr)(road length km) (1000 kg/g)
8 Paint g 6.04E+03
Herbicide for highway operation kg/m*yr 2.88E-04
(Good transport percent road use)(herbicide usekg/m*yr)(road length km) (1000
kg/g)
9 Herbicide g 3.37E+02
UEV for orthophosphate from Nepal (2008)
Transport
Mid-size truck fuel economy diesel kg/vkm 0.25 (Kodjak 2004)
Tractor trailer truck fuel economy diesel kg/vkm 0.37 (Kodjak 2004)
Mid-size truck vkm/ton-km vkm/ton-km 0.62 Lorry 3.5-16t. Table 5-119.
Lorry >16t. Table 5-
Tractor trailer vkm/ton-km vkm/ton-km 0.12 119.
Tractor trailer ton-km percentage 0.88 Table 5-119.
Lorry >16t. Table 5-
Mid-size truck ton-km ton-km 1.75E+08 119.
Tractor trailer ton-km ton-km 1.32E+09 Lorry 3.5-16t. Table 5-119.
Truck fuel use = (Truck ton-km)(ton-km/vkm)(diesel kg/vkm) (4.36E+07 J/kg)
Mid-size truck fuel use J 1.20E+15 1.08E+08
Tractor trailer fuel use J 2.53E+15 1.56E+08
10 Total diesel fuel use J 3.73E+15 2.64E+08
No. trucks= total vehicles* portion of trucks in import data (Economic Commission
11 of Latin American and the Carribbean 2006; United Nations 2008)
(5.04E+04 Ton-km/truck/yr USA)(.5 Peru/US productivity)(142872 trucks in Peru
fleet)
Annual truck transport ton-km 1.50E+09


155









Table B-23. Assumed origins and transport distances for inputs to mining.
Sea Road
Mass Data Distance Distance
Input (kg) Assumed Origin Source (km) (km)
Refined Oil 9.75E+07
Imported 2.34E+07 Balao, Ecuador 1 1148 250
Domestic 7.41E+07 Chimbote 1 0 250
Lime 7.36E+07 China Linda 2 0 12
Chlorine 4.41 E+07 Lima 3 0 850
Caustic soda 2.52E+07 Lima 1 0 850
Explosives
(ANFO) 7.00E+06 Lima 3 0 850
Sodium cyanide 6.69E+06 US 3 5900 850
Concrete 4.68E+06 China Linda 2 0 12
Steel pipe 2.97E+06 US 3 5900 850
Other 1.27E+07 Local NA 0 0
TOTAL 2.74E+08
Notes
Only inputs comprising 1% of total mass input are listed.
Data Sources
1. (Instituto Nacional Estadistica y Informacion 2006))
2. (Buenaventura Mining Company Inc. 2006)
3. (United Nations 2008)


156









Table B-24. System-level parameters.


Parameter
include_geo

include_clay_em
include_grav_em

mine lifetime
process_lifetim
waste to reclam
lima_yan_distan
Au_output
Hg_output
veh add steel

veh add rubber
veh_weight
kgore_topadarea
kgoretopoolarea
per_RO_treat
tot excess wat
Au_ore_grade

labor use J
labor use dol

seatransport


Default
Value a2geo Units and Comments
1 NA 1=Include geologic emergy of gold ore; 0=do not include
1 =Include geologic emergy of clay for roads and leach pads; 0=do not
0 NA include
0 NA 1=Include geologic emergy of gravel for roads and leach pads; 0=do not include
yrs. 1993-2018. End date estimate from
25 1.3 http://www.newmont.com/csr05/protest_yanacocha/l.html
20 1.3 yrs. Avg process lifetime for all processing facilities. Less than mine_lifetime
1 NA Fraction of waste rock used to refill pits. 1=All waste rock used for backfilling
850 1.1 km. (1.05,1,1,1.01,1,NA)
3327500 1 oz/yr, Buenaventura 2006
5.5 1 short tons/month, Newmont 2006a
0.4 1.2 Additional fraction steel for heavy vehicles. (1.2,1,1.03,1,1,NA)
Additional fraction rubber for heavy vehicles. This is substituted with steel for track
0.07 1.2 vehicles. (1.2,1,1.03,1,1,NA)
15500 1.2 kg. Based on 40ton Lorry (Ecoinvent). (1.2,1,1.03,1,1,NA)
198891 1.5 kg/m2. Based on avg of 5 leach pad areas and capacities. Actual SD*2
4057275 1.5 kg/m2. Based on avg of 5 leach pad areas and capacities. Actual SD*2
0.4 1 Fraction of excess water treatment using reverse osmosis
1.2E+13 1 G
0.028 1 oz/ton
1 = include labor by using sej/J emergy in labor. See emergy in labor process. 0= Do
0 NA not use
0 NA 1 = include labor using emergy/$ ratio. 0=do not include.
km. Los Angeles to Lima sea distance. Used for generic sea transport distance.
5900 1.1 (1.05,1,1,1.01,1,NA)


157









Table B-25. Uncertainty estimates for inventory data using Ecoinvent method (Frischknecht and Jungbluth 2007)


Unit Process(es)
Exploration,
Extraction,
Reclamation
Exploration,
Extraction, Sed. &
Dust control
Extraction,
Reclamation, Mine
Infrastructure
Mine infrastructure

Extraction

Leaching
Processing

Water treatment,
Reclamation


Variables

Variables


Input or Variable
Oil, refined


Water for process


Heavy Vehicle Use


Infrastructure based on
visual estimates
Explosives

CN
Natural gas

Chemicals for water
treatment (CaOH, CI,
FeC13, PAM, H2SO4);
and reclamation (CaOH)
Distance variables


Mine vehicle model
variables


1 1


1 1


1 1.1


1 1

1 1


o o Uncertainty
U 0 oU 0) score
1.1 1. NA 1.3
2

1 1 NA 1.2


1.1 1 NA 1.3


1 1.
5
1 1


1 1 NA
1 1. NA
2
1 1 NA


1 1 1.0
1
1 1.03 1


1 NA

1 NA


158










APPENDIX C
SUPPLEMENT TO CHAPTER 3: R CODE FOR STOCHASTIC UNCERTAINTY
MODELS

The following sections contains code for stochastic uncertainty models for both the

formula and table-form uncertainty models, as described in chapter 2. This code can be

run in R statistical software.


Code for Formula UEV Uncertainty Estimation
#A script for a Monte Carlo simulations of formula-type unit emergy values to estimate uncertainty
#Author: Wes Ingwersen, wwi@ufl.edu
##Do a Monte Carlo simulation for a formula UEV calculation, with uncertainty expressed for all variables

""""############/ Instructions################
#Prepare a tab separated table of items in your emergy table in the form of:
#variable_name average standard deviation
#the following is a sample for the lead UEV this can be copied and pasted into a new .txt file
crust_conc_ppm 15 1.41
ore_grad_frac 0.06 0.03
crust_turn_cm_yr-1 2.88E-03 6.77E-04
den_crustg_cc-1 2.72 0.04
crustal area_sqcm 1.48E+18 2.1E+16
#This file has to be saved at C:\RData\UEV\ directory unless the path name is changed in the script for
the script to function.

####.##############.Import Data#####################
#Input data in the form of a tab-delimited txt file with var name, mean, sd, on 1 line
#Uncomment lines related to UEV of interest
#To see the table that translates into this format, see Table 3 in Ingwersen (2009)

#UEV for lead
#fname <- "C:\\RData\\UEV\\lead.txt"
#item <- "lead"
#fractions <- c(1,2)
#den_unit <- "g"
#mag <- 12 #Order of mag of deterministic mean UEV

#UEV of iron
#fname <- "C:\\RData\\UEV\\iron.txt"
#item <- "iron"
#fractions <- c(1,2)
#den_unit <- "g"
#mag <- 10 #Order of mag of deterministic mean UEV

#UEV of oil
#fname <- "C:\\RData\\UEV\\oil.txt"
#item <- "oil"
#fractions <- c(2,3,4,5)
#den unit <- "J"
#mag <- 5 #Order of mag of deterministic mean UEV


159










#UEV of groundwater
#fname <- "C:\\RData\\UEV\\gw.txt"
#item <- "gw" #Groundwater
#fractions <- c(2)
#den_unit <- "g"
#mag <- 5 #Order of mag of deterministic mean UEV

#UEV of labor
#fname <- "C:\\RData\\UEV\\labor.txt"
#item <- "labor"
#fractions <- cO
#den unit <- "J"
#mag <- 6 #Order of mag of deterministic mean UEV

#Loads the text file, stores it in a data frame
cols <- c("var","mu","sig")
df <- read.delim(fname,header=FALSE,strip.white=TRUE,row.names=1, col.names=cols)
df

#Verify that the data loaded properly

#########Set Initial Parameters""###### ###########
#Run the following code
##Number of MC results
n <-100

#Number of MC's to run from which to calculate the uncertainty
j<- 100

##Case 1: Assume variables are normally distributed
##Case 2: Assume varibales are log-normally distributed
case <- 2
#Note Model only stable using case 2

###########tFunctions for MC just load on first use#ttttlt#############

##Function to return logforms of mean and standard dev
returnlogforms <- function(mu,sig) {
lamda <- 1+(sig/mu)^2
logformsig <- sqrt(log(lamda))
logformmu <- log(mu)-0.5*logformsig
return (c(logformmu,logformsig))
}

#n will also be the number of replicates of each variable in the model chosen

#Make a matrix to hold n of each model parameter)
make_params <- function
{
mc_vars <- matrix(nrow=nrow(df),ncol=n)
for (x in 1:nrow(df))
{
#Put the mean and sd in a matrix
m <- df[[1]][x]
s <- df[[2]][x]
if(case==2)


160










{
logforms <- returnlogforms(m,s)
mc_vars[x,] <- rlnorm(n,meanlog=logforms[1],sdlog=logforms[2])

} else {
mc_vars[x,] <- rnorm(n,mean=m,sd=s)

}
}
return(mc_vars)
}


clean <- function(parameters) {
a <-0
b <-0
for (a in 1 :length(fractions)) {
ind <- fractions[a]

for (b in 1:n) {
if ((parameters[ind,b]<=0 || parameters[ind,b]>=1) && !is.na(parameters[1,b])) {
parameters[,b] <- NA
}
}
}
}

####.##################Unit energy value model######tltlttttttlttttt#####tltl####
#Run the desired model, or enter your own model

#Model for land cycle is
#ER <- 2.ore_grad_frac/(1.crust_conc_ppm/1E6)
#ER
#Land_UEV <- 15.83E24/(3.crust_turn_cm_yr-1)*(4.den_crust_g_cc-1)*(5.crust_area_sqcm)
#Mineral UEV <- ER*Land UEV

#Model for water = UEVwater, sej/g = (global emergy base, 15.83E24 sej/yr)/Annual Flux, g/yr)
#turnover time = (Global groundwater resevoir)/
#(Global precip on land, mm/day)(365days/yr)/(1E6 mm/km)*(global land area (km2)*(infiltration rate)

#Function to do the model calculation
mod <- function (mat)
{
res_vec <- cO #Result vector
for (i in 1:n)
{
UEV <- NA
if ((item=="lead" I| item=="iron")&& !is.na(mat[1,i])) {
pred <- 2.64 # Predicted sq_sig_geo for lead_UEV
pred <- 2.03 # Predicted for iron_UEV
#Formula for Mineral UEV calc
if (item=="lead") {
er <- mat[2,i]/(mat[1 ,i]/1E6)#when conc is in ppm
} else {
er <- mat[2,i]/(mat[1 ,i]) #when conc is a frac
}


161










land_UEV <- 15.83E24/(mat[3,i]*mat[4,i]*mat[5,i])
UEV <- er*landUEV #For mineral calcs
}
if (item=="oil" && !is.na(mat[1 ,i])) {
#Formula for oil
#Deterministic solution
#mat <- df
#i<-1
ep_c <- (mat[1 ,i]*1.78E4)/mat[2,i]
ek <- ep_c/mat[3,i]
UEV <- (1.68*ek*mat[5,i])/(mat[4,i]*4.19E4)
#UEV
#If UEV is negative take absolute value
}
if (item=="gw" && !is.na(mat[1 ,i])) {
#Formula for groundwater
#Deterministic solution
#mat <- df
#i<-1
global_land_area <- mat[3,i] #km2
precip <- mat[1,i] #mm/yr
infiltration <- mat[2,i]
annual_flux <- ((precip/1 E6)*global_land_area*infiltration*1E 12*1000)
UEV <- 15.83E24/annualflux
}
if (item=="labor") {
#Formula for labor
#(Global emergy use per yr/global population)/(Daily per capital calorie
intake*365 days* 4184J/kcal)
#mat <- df
#i<-1
UEV<-(1.61 E26/mat[1 ,i])/(mat[2,i]*365*4184)
}
if (UEV<0) {
UEV <- NA
}
res_vec[i] <- UEV
}
return(res_vec)
}

,"#######tt##lt#t RUN SIMULATION####################
#Hightlight and run the following code

#Run the Monte Carlo, j times
mc <- cO #Store the results of one Monte Carlo here
Quot_upper_by_med <- cO #Store the results of the upper limimit divided by the median for each MC
upperlims <- cO
lowerlims <- cO
medians <- cO
means <- cO
sds <- cO
all_mc <- matrix(nrow=j,ncol=n)#Store each mc result in a row for graphing later
for (a in 1:j)
{
params <- makeparams0


162










if (length(fractions)) (clean(params)) #Removes values <0 or >1 for fractions
mc <- mod(params)
all_mc[a,] <- mc
med <- median(mc,na.rm=TRUE)
std <- sd(mc,na.rm=TRUE)
Cls <- format(quantile(mc, probs = c(0.025,0.975),na.rm=TRUE, digits=3, scientific=TRUE))
upperlim <- as.double(as.vector(Cls["97.5%"]))
lowerlim <- as.double(as.vector(Cls["2.5%"]))
up <- upperlim/med
#low <- med/lowerlim
upperlims[a] <- upperlim
lowerlims[a] <- lowerlim
medians[a] <- med
medians[a] <- med
sds[a] <- std
Quot_upper_by_med[a] <- up

}

#Take averages of medians of distributions and geometric variances
med <-mean(medians)
geo_var <- mean(Quot_upper_by_med)
lower_bound <- mean(lowerlims)
upper_bound <- mean(upperlims)
#Print the results
c('Median=',med)
c('Geometric variance=',geo_var)
c('Lower bound',lower_bound)
c('Upper bound', upper_bound)

Code for Table-form UEV Uncertainty Estimation
#A script for a Monte carlo simulations of table-form unit emergy values to estimate uncertainty
#Author: Wes Ingwersen, wwi@ufl.edu
##Do a Monte Carlo simulation for a table-form UEV calculation, with uncertainty expressed for all
variables

""""############/ Instructions#### ## ### #####
#Input data in the form of a tab-delimited txt file with
var-name flow_quanity_mean flow_quanity_geo_var UEV_mean UEV_geo_var
#the following is a sample for the sulfuric acid UEV this can be copied and pasted into a new .txt file
Secondary_sulfur 214 1.32 5200000000 3.59
Diesel 3410 1.34 121000 3.59
Electricity 63000 1.34 371000 2.77
Water 241000 1.23 189572.5914 1.95
#This file has to be saved at C:\RData\UEV\ directory unless the path name is changed in the script for
the script to function.

##################Import Data#####################
#UEV for electricity
#fname <- "C:\\RData\\UEV\\electricity.txt"
#item <- "electricity"
#den <- 3.6E6 #Joules of electricity This is the denominator for the UEV calculation
#den unit <- "J"
#mag <- 5 #Order of mag of deterministic mean UEV


163










#UEV for sulfuric acid
fname <- "C:\\RData\\UEV\\sulfuricacid.txt"
item <- "sulfuric acid"
den <- 1000 #g of H2S04 This is the denominator for the UEV calculation
den_unit <- "g"
mag <- 7 #Order of mag of deterministic mean UEV


cols <- c("param","value","k_value","UEV","k_UEV")
df <- read.delim(fname,header=FALSE,strip.white=TRUE,row.names=1, col.names=cols)
df

#########Set Initial Parameters#######"m#### l####

##Number of MC results
n <-100

#Number of MC's to run from which to calculate the uncertainty
j<- 100

##Case 2: Assume varibales are log-normally distributed
#Now it only works for log-normally distributed variables
case <- 2

###########tFunctions for MC just load on first use###########.#######

##Function to return logforms of mean and standard dev #Only used for formula UEVs copied here for
reference
returnlogforms <- function(mu,sig) {
lamda <- 1+(sig/mu)^2
logformsig <- sqrt(log(lamda)) #Source: Wikipedia, "Lognormal distribution"
logformmu <- log(mu)-0.5*logformsig #Wikipedia
return (c(logformmu,logformsig))
}

##Function to return logforms of with determininstic mean and k value (ref: Slob (1994))
returnlogforms_withKvalue <- function(mu,k) {
logformsig <- sqrt((log(k)/1.96)^2)
logformmu <- log(mu)-0.5*logformsig
logformsig
logformmu
return (c(logformmu,logformsig))
}

#Make a matrix to hold n of each model parameter)
make_params <- function
{
#Create a matrix to store n random values(3rd dimension) of both the value and UEV (2nd dimension) of
each variable (1st dim)
mc_vars <- mc_vars <- array(NA,dim=c(nrow(df),2,n))
for (x in 1:nrow(df))
{
#Gets the values from the input matrix
val <- df[[1]][x]
k_val <- df[[2]][x]
uev <- df[[3]][x]


164










k_uev <- df[[4]][x]
#Call the script to get the logforms of mu and sigma
val_logforms <- returnlogforms_withKvalue(val,k_val)
uev_logforms <- returnlogforms_withKvalue(uev,k_uev)
#Use the log-forms in a lognormal distribution random generator function
mc_vars[x,1,] <- rlnorm(n,meanlog=val_logforms[1],sdlog=val_logforms[2])
mc_vars[x,2,] <- rlnorm(n,meanlog=uev_logforms[1],sdlog=uevlogforms[2])

}
return(mc_vars)
}


#Function to do the model calculation
mod <- function (mat)
{
res_vec <- cO #Result vector
for (i in 1:n)
{
UEV <- NA
#Calculate the UEV for that random set of params

em <- 0
for (r in 1:nrow(df)){
#Multiply the value and UEV
var_em <- mat[r,1,i]*mat[r,2,i]
#Add the emergy to the sum
em <- em + varem
}
UEV <- em/den #UEV is sum of emergy divided by denominator (usu. J or g)
res_vec[i] <- UEV
}
return(res_vec)
}

#################RUN SIMULATION###################


#Run the Monte Carlo, j times
mc <- cO #Store the results of one Monte Carlo here
Quot_upper_by_med <- cO #Store the results of the upper limimit divided by the median for each MC
upperlims <- cO
lowerlims <- cO
medians <- cO
means <- cO
sds <- cO
all_mc <- matrix(nrow=j,ncol=n)#Store each mc result in a row for graphing later
for (a in 1:j)
{
params <- make_params0
mc <- mod(params)
all_mc[a,] <- mc
med <- median(mc,na.rm=TRUE)
m <- mean(mc,na.rm=TRUE)
std <- sd(mc,na.rm=TRUE)
Cls <- format(quantile(mc, probs = c(0.025,0.975),na.rm=TRUE, digits=3, scientific=TRUE))


165










upperlim <- as.double(as.vector(Cls["97.5%"]))
lowerlim <- as.double(as.vector(Cls["2.5%"]))
up_by_med <- upperlim/med
upperlims[a] <- upperlim
lowerlims[a] <- lowerlim
medians[a] <- med
means[a] <- m
sds[a] <- std
Quot_upper_by_med[a] <- up_by_med
}



#Take averages of medians of distributions and geometric variances
med <-mean(medians)
geo_var <- mean(Quot_upper_by_med)
lower_bound <- mean(lowerlims)
upper_bound <- mean(upperlims)
#Print the results
c('Median=',med)
c('Geometric variance=',geo_var)
c('Lower bound',lower_bound)
c('Upper bound', upper_bound)


166










APPENDIX D
SUPPLEMENT TO CHAPTER 4: ADDITIONAL TABLES AND FIGURES

Table D-1. Inputs to one kg pineapple at the packing facility.


Category Input name
Energy Diesel, at regional storage
Petrol, unleaded, at regional storage
Ammonium nitrate, as N, at regional
Fertilizer storehouse
Boric acid, anhydrous, powder, at
plant
Calcium nitrate, as N, at regional
storehouse
Compost, at plant
Dolomite, at plant
Fosfomax (0,30,0) fertilizer
Iron sulphate, at plant
Kaolin, at plant
Lime, hydrated, packed, at plant
Magnesium ammonium nitrate,
(22,0,0,0,7)
Magnesium sulphate, at plant
NPK (12,24,12) fertilizer
NPK (18,5,15) fertilizer
NPK (2,10,10) fertilizer
Potassium chloride, as K20, at
regional storehouse
Potassium sulphate, as K20, at
regional storehouse
Single superphosphate, as P205, at
regional storehouse
Sugar, from sugarcane, at sugar
refinery
Urea, as N, at regional storehouse
Zinc monosulphate, ZnSO4.H20, at
plant
benzoic-compounds, at regional
fungicide storehouse
pesticide unspecified, at regional
storehouse

triazine-compounds, at regional
storehouse
triazine-compounds, at regional
storehouse
organophosphorus-compounds, at
growth regional storehouse
diphenylether-compounds, at regional
herbicide storehouse
diuron, at regional storehouse
glyphosate, at regional storehouse
pesticide unspecified, at regional


Country
RER e
RER e


Unit Amount SD Active Ing.
kg 7.29E-03 2.97E-03n/a
kg 2.40E-04 2.20E-04n/a


RER e kg 1.92E-03 1.08E-03n/a

RER e kg 1.73E-04 1.89E-04n/a


RER
CH
RER
CR
RER
RER
CH

RER
RER
RER
RER
RER


1.72E-04 4.66E-05n/a
4.33E-03 2.43E-03n/a
2.03E-04 4.58E-05n/a
4.51E-04 3.67E-04n/a
2.97E-04 2.45E-04n/a
8.20E-04 6.74E-04n/a
1.63E-03 3.68E-04n/a

2.11E-03 1.19E-03n/a
2.03E-03 2.09E-03n/a
1.18E-02 9.63E-03n/a
2.11E-03 1.72E-03n/a
7.93E-05 6.46E-05n/a


RER e kg 5.82E-03 4.74E-03n/a

RER e kg 4.33E-03 3.52E-03n/a

RER e kg 5.54E-05 4.51E-05n/a

BR e kg 2.51E-04 5.67E-05n/a
RER e kg 3.62E-03 2.04E-03n/a

RER e kg 2.74E-04 7.58E-05n/a

RER e kg 5.63E-05 3.55E-05Metalaxil

RER e kg 1.49E-04 9.40E-05Fosetyl-aluminium
Thiazole, 2-
(thiocyanatemethylth
RER e kg 1.20E-06 7.54E-07io)benzo-

RER e kg 6.58E-06 4.15E-06Triadimefon

RER e kg 2.58E-05 3.69E-05Ethephon


RER
RER
RER
RER


6.58E-06 3.43E-06Fluazifop-p-butyl
1.12E-04 5.83E-05Diuron
3.76E-05 1.96E-05Glyphosate
6.60E-05 3.44E-05Bromacil


167










storehouse
phenoxy-compounds, at regional
storehouse
triazine-compounds, at regional
storehouse
insecticid [thio]carbamate-compounds, at
e regional storehouse
organophosphorus-compounds, at
regional storehouse
nematicid organophosphorus-compounds, at
e regional storehouse
Machiner
y tractor, production


RER e kg 1.38E-06 7.21 E-07Quizalofop-P

RER e kg 7.96E-05 4.14E-05Ametryn

RER e kg 3.08E-05 1.60E-05Carbaryl

RER e kg 1.24E-04 7.84E-05Diazinon

RER e kg 6.80E-05 5.47E-05Ethoprop

CH e kg 3.13E-04 1.35E-04


Table D-2. Emissions from one kg pineapple at the packing facility.
Substance To Amount GV Note
Ametryn air 4.90E-06 3.0 from pesticide application. Includes yield
and pesticide input uncertainty.


Ametryn
Ammonia
Ammonia
Benzene
Benzo(a)pyrene
Bromacil


Bromacil
Cadmium
Carbaryl

Carbaryl
Carbon dioxide, fossil


Carbon dioxide, fossil
Carbon dioxide, land transformation
Carbon monoxide, fossil
Chromium
Copper
Diazinon

Diazinon
Dinitrogen monoxide
Dinitrogen monoxide
Diuron


water
air
air
air
air
air

water
air
air


9.87E-06
1.55E-07
1.10E-04
2.33E-06
2.28E-10
9.62E-06

5.42E-06
7.53E-11
5.16E-06


water 1.78E-07
air 2.35E-02


air
air
air
air
air
air

water
air
air
air


6.45E-04
1.00E-10
2.05E-04
3.77E-10
1.28E-08
6.01E-06

3.60E-07
9.06E-07
1.78E-04
6.55E-06


water 2.20E-05 4.5
air 1.51E-05 8.4


5.9 "
2.3 from fuel combustion
2.9 volatilized from N fertilizers
2.3 from fuel combustion
2.3 from fuel combustion
2.1 from pesticide application. Includes yield
and pesticide input uncertainty.
4.8 "
5.9 from fuel combustion
4.3 from pesticide application. Includes yield
and pesticide input uncertainty.
7.5 "
2.1 from fuel combustion. Combines
uncertainty of diesel input, diesel
emission factor, and yield
2.8 from urea application
from land use change
5.9 from fuel combustion
5.9 from fuel combustion
5.9 from fuel combustion
3.0 from pesticide application. Includes yield
and pesticide input uncertainty.
r n


from fuel combustion
from N fertilizers
from pesticide application. Includes yield
and pesticide input uncertainty.

from pesticide application. Includes yield
and pesticide input uncertainty.


water 2.27E-07 12.8 "
air 2.25E-06 6.6 from pesticide application. Includes yield
and pesticide input uncertainty.
water 1.13E-06 10.4"


168


Diuron
Ethephon

Ethephon
Ethoprop

Ethoprop










Fluazifop-p-butyl

Fosetyl-aluminium
Glyphosate


Glyphosate
Lead
Metalaxil


Metalaxil
Methane, fossil
Nickel
Nitrate
Nitrogen oxides
Nitrogen oxides
NMVOC, non-methane volatile organic
compounds, unspecified origin
PAH, polycyclic aromatic hydrocarbons
Paraquat

Paraquat
Particulates, < 2.5 um
Phosphate
Phosphorus


Quizalofop-P

Quizalofop-P
Sediment, eroded

Selenium
Sulfur dioxide
Triadimefon

Triadimefon
Water

Zinc


air 1.73E-06 8.4 from pesticide application. Includes yield
and pesticide input uncertainty.
water 1.59E-05 2.8 "
air 2.17E-05 8.4 from pesticide application. Includes yield
and pesticide input uncertainty.


water
air
water

air
air
water
air
air
water
air

air
air


2.95E-06
3.51E-08
1.82E-06

4.92E-07
1.64E-06
5.27E-10
6.84E-03
2.79E-04
8.57E-08
1.87E-05


12.8
5.9
2.4

5.1
2.3
5.9
3.0
2.3
2.8
2.3


2.31E-08 3.8
4.44E-07 8.4


from fuel combustion
from pesticide application. Includes yield
and pesticide input uncertainty.

from fuel combustion
from fuel combustion
leached from N fertilizers
from fuel combustion
from N fertilizers
from fuel combustion

from fuel combustion
from pesticide application. Includes yield
and pesticide input uncertainty.


air 2.17E-06 12.8"
air 1.27E-05 3.8 from fuel combustion
water 1.15E-04 4.3 runoff of P fertilizers
air 1.17E-04 18.7 P in eroded soil. Uncertainty includes soil
erosion, P content in soil, and yield
uncertainty
water 6.88E-08 8.4 from pesticide application. Includes yield
and pesticide input uncertainty.
water 1.48E-07 12.8"
air 6.28E-02 18.2 estimated with RUSLE2 model. Includes
yield and emission uncertainty
water 7.53E-11 5.9 from fuel combustion
water 7.38E-06 2.1 from fuel combustion
air 1.16E-06 8.4 from pesticide application. Includes yield
and pesticide input uncertainty.
air 3.38E-08 12.8"
air 1.62E+00 1.5 evaporated blue water. Includes yield and
emission uncertainty
air 7.53E-09 5.9 from fuel combustion


169









Table D-3. Emissions estimations for mineral-N in applied fertilizers.


No Pathway Equation Source

1 Uptake 0.018 dry biomass Su (1968)
2 NH3-N to air 1-15 % N applied Brentrup and Kusters (2000)
3 N20-N to air 1.25 % N applied IPCC 2007, for estimating direct
N20 emissions
4 N02-N to air 0.001 % N20-N Nemecek and Kagi (2007)
5 N03-N to water 1 residual N in soil Brentrup and Kusters (2000)
Item notes
1 Based on percent concentration of N in dry pineapple biomass
5 Assuming exchange ratio (rainfall/field capacity) = 1, all residual N leaches


Table D-4. Emissions estimations for mineral-P in applied fertilizers.

Item Pathway Equation Source
1 Uptake 0.18% *biomass Su (1968)
2 P205-P to water 2.5% of applied Powers (2007)
kg P/kg
3 P in erodible sediment 0.00186 soil Nemecek and Kagi(2007)


Table D-5. General assumptions used in the FAO CROPWAT model.
CROPWAT Component Assumption
Penman ET based on geographically specific data from
Climate LocClim
Rain Rainfall from LocClim; calculation with USDA S.C. Method
Soil Medium (loam) from CROPWAT database
Crop water requirement See Table X
Schedule Irrigate at user-defined intervals; 70% (default) efficiency


Table D-6. Crop water requirement variables for CROPWAT.
Parameter Value Source


Kc init
Kc mid and end
Kc initl
Kc mid
Stage initial, days
Stage development,
days


0.9 Bartholomew (2003) p. 95, for non-mulched system
0.4 Bartholomew (2003) p. 95, for non-mulched system
0.5 Allen et al. (1998), refers to plastic mulched system
0.3 Allen et al. (1998), refers to plastic mulched system
90 Based on average reported harvest schedule

180 "


170










Stage mid-season,
days
Stage late season,
days
Yf (all stages)
Rooting depth, m

Critical depletion, p
Crop height, m


120

180
1.0 Estimated based on crops with similar critical depletion
0.45 Smith (1992), p. 61,
Smith (1992), referred to as fraction of available soil water
0.5 p. 61
0.9 Bartholomew (2003), average height


Table D-7. RUSLE2 parameters for Pineapple in Costa Rica
RUSLE2 Parameter Value/Setting
Componen
t


Introductio Template
n
Profile Horiz. overland flow path length, m
Avg. slope steepness, %
Contouring
Strips/barriers
Diversion/terrace, sediment basin
Subsurface drainage
Adjust res. burial level
Climate How to get erosivity

R factor, US units
Standard El






How determine runoff?

10-yr, 24-hr rain (mm)
Annual precip
Soils Erodibility

Erodibility, SI

Hydrologic class
Hydrologic class w/subsurface drainage
Rock cover%
Cal. Consolidation from precip
Normal consolidation time, yrs
Manageme Rel. row grade %
nt


Notes


ARS Basic Uniform
Slope
16.1 production-weighted average
2.1 production-weighted average
Up and down slope
None
None
none
Normal
Enter R & choose
El zone
450 for North zone
Enter half-montly El
based on relative
intensity of storm
events during the
month


based on 10-yr 24
hr ppt
from FAO Clim
from FAO Clim
get from standard
nomograph
0.036 production-weighted average
of samples
C mod. high runoff
B mod. low runoff drainage decreases runoff
0
Yes
7 Default


171










Long-term natural rough, mm
Normally used as a rotation?
Duration, yr
Operations Dates for 1st complete cycle
Cropland\disks\disk, tandem heavy
primary op.
Cropland\bedders/hippers\hipper
Add mulch
basic/general\begin growth
basic/general\harvest pineapple
basic/general\harvest pineapple
basic/general\kill vegetation
Operation: Portion of total biomass effected
harvest
pineapple





Portion of effected left on surface


6
No
NA


1/1/2000

2/1/2000
NA
3/1/2000
5/1/2001
NA
11/1/2002
0.4975







0.17


Portion of effected left as standing residue 0


Assume all pineapples
harvested at once, with fruit
and 25% of plant being
effected. Assuming 1.5 green
biomass:fruit weight, 33% is
the removed fruit. Of the
remaining 66%, assumed
25% is chopped. 33%*22%
Portion of biomass effected -
removed fruit
0


Vegetation First yield for biomass conversion (kg/ha) 67000

1st above ground biomass at max canopy 16000
(kg/ha)
Biomass-yield ratio 0.097
Develop growth chart for a production Yes
(yield) level other than base level


Adjust fall height based on canopy shape? NA
Adjust biomass-yield relationship NA
Adjust senescence relationship see
rela
Adjust yield/flow-retardance relationship see
Veg
nce
Setup long-term veg NA
Residue Responds to tillage like non
(cor
Decomp. half-life, days 130



Weight required for area covered, 60%, 400
kg/hec


Senescence
tionship

etation_retarda


-fragile-med
n)




0


default

Use exponential decay
equation with average lifetime
from Bartholomew (2003)
mean life In(2)
calculation (above ground
biomass) (percent chopped)


172














Decomposition half-life
Time until decay, weeks


Halflife, weeks
Senescenc Above ground biomass subject to
e senesence, %
Relationship
p
Vegetation Type of row spacing
retardance
Max. expected retardance
Avg. yield for this expect. Retardance
Does 'no retardance' apply for yields >0
Retard class at zero yield


18.02182669
0


Assume of plant mass 25% is
chopped and used to cover
60% of area. The mass can
be related to the harvest

from Bartholomew (2003)

mean life*ln(2)
Plant continues growing until
killed


Veg. on ridges

High
67000
No
Low


Table D-8. Parameters modified for USETox-CR model.
USETox-
Item Name USETox-CR Default Source
1 Continent, Area land, km2 5.11E+04 9.01E+06 INEC 2009
2 Continent, Area sea, km2 5.00E+04 9.87E+05 Humbert et al. 2006
3 Continent, Areafrac, freshwater 8.61E-03 3.00E-02 Humbert et al. 2006
4 Continent, Areafrac, natural soil 4.60E-01 4.85E-01 INEC 2009
5 Continent, Areafrac, ag soil 5.29E-01 4.85E-01
6 Continent, Areafrac, other soil 9.78E-03 1.00E-20
7 Continent, Temperature, C 2.50E+01 1.20E+01 Humbert et al. 2006
8 Continent, Rain rate, mm/yr 3.24E+03 7.00E+02 Humbert et al. 2006
Rubin and Hyman
9 Continent, Soil erosion, mm/yr 4.20E-01 3.00E-02 2000
10 Human pop., Continent 4.45E+06 9.98E+08 INEC 2009
11 Human pop., Urban 2.80E+06 2.00E+06 INEC 2009
Exposed produce, continent,
kg/day/capita
12 2.38E+00 7.53E-01 Humbert et al. 2006
Unexposed produce, continent,
kg/day/capita
13 8.62E-01 2.35E-01 Humbert et al. 2006
14 Meat, continent, kg/day/capita 1.11E-01 8.39E-02 Humbert et al. 2006
Dairy products continent,
kg/day/capita
15 aaa4.25E-01 2.50E-01 Humbert et al. 2006
16 Fish freshwater, kg/day/capita 5.43E-03 1.26E-02 Humbert et al. 2006
17 Fish marine, kg/day/capita 1.76E-03 3.57E-02 Humbert et al. 2006
Item Notes
4 Based on total protected area


173










5 Remainder of other land area fractions
6 Assume 500 km2 of urban area + semi-urban
11 63% of population lives in urban areas
14 Pork+beef+chicken+goat


174










Table D-9. Sensitivity analysis of the RUSLE2 model customized for pineapple in CR.

New Parameter Erosion
Category Variable changed Set/Value (MT/ha/yr) % Change Note
Baseline NA NA 7.3 NA


Climate Geographic location


% slope
% slope
% slope
% slope*
% slope*

Soil erodibility, SI



Soil erodibility, SI


Pacific climate
Atlantic climate


30

0.071



0.022


Contouring Cross-slope moderate
Contouring Standard contouring


7.5 3%
7.1 -3%
3.3 -55% Smallest slope among sites
20 174%
41 462%
89 1119%
130 1681% Approximately largest
slope among visited sites
9.4 29% Silt loam with 80% silt.
Estimate of most highly
erodible soils present in
pineapple zone
3.0 -59% Loamy sand with 10% silt.
Estimate of least erodible
soils present in pineapple
zone
4.3 -41%
4.0 -45%


Management
schedule
Management
schedule


Mulch
Vegetation Residue half-life,
days
Residue half-life,
days
Yield, tons/ha


Yield, tons/ha

Above ground dry
biomass: harvest
weight ratio


Above ground dry
biomass: harvest
weight ratio


Double harvest

Initial preparation
during rainy season-
rainfall
Add plastic mulch
260

65

33.5


110

0.0647




0.144


-33%

27%


-78%
-29%

22%


Double harvest


Typical in organic practice


14 92% Half of average yield,
assume limits of
competitive production
4.1 -44% Max yield reported, Gomez
et al. 2007
7.4 1% Lowest plant biomass;
based on highest
fruit:biomass fresh weight
ratio of 1 (Bartholomew
2003)
6.8 -7% Highest plant biomass;
based on lowest
fruit:biomass fresh weight
ratio of 0.45 (Bartholomew
2003)


Max for farm of unknown origin 1681%


Geometric variance 17.8


175


Profile


Management









Table D-10. Sensitivity analysis of the FAO CROPWAT model to variables found in
pineapple cultivation.
ET
(mm/
crop %
Category Variable changed New Parameter Set/Value cycle) Change
Baseline NA NA 767.6 NA
Climate Geographic location Pacific climate 811.7 6%
Atlantic climate 712.8 -7%
Field Soil texture Clay 763.0 -1%
Sand 723.0 -6%
Add plastic mulch kc init= 0.6, kc mature=0.3 565.0 -26%
Vegetation Higher relative crop transpiration kc init= 0.9, kc mature=0.74 1335.0 74%
Root depth depth = 1m 770 0.3%
Critical depletion, p High (p=0.75) 768 0.1%
Yield in response to water High (Yf = 1.25) 765 -0.3%
High (Yf= .75) 767 -0.1%
Crop height, m Tall (height = 1.25 m) 767.7 0.0%
Max 74%
Geo var 1.74



Table D-11. Sensitivity analysis of PestLCI model for pineapple conditions.
New % %
Parameter Sensitivity Change, Sensitivity Change,
Category Variable changed Set/Value ratio fair ratio fsw
Climate Solar radiation, MJ/m2/yr 6595 -1.32 -5.8% n/a n/a
Solar radiation, MJ/m2/yr 6271 -1.32 1.0% n/a n/a
Field farm average % slope 1 n/a 1.1 -66.0%
farm average % slope 5 n/a 1.1 110.0%
farm average % slope 10 n/a 1.1 330.0%
Sand content (top layer) % 82 n/a -2.0 -177.8%
Sand content (top layer) % 10 n/a -2.0 150.4%

% canopy cover when
applied 20% -0.75 55.3% 3.01 -220.5%
% canopy cover when
applied 97% -0.8 -22.1% 3.0 88.2%


MAX
Geo var


55%
1.55


330%
4.30


176














Table D-12. Recalculation of Pimentel (2009) energy demand for US oranges.
Input Quantity Unit CED (1 E3 kcal)
Machinery 50 kgc 1206.172
Diesel 337 La 3739.454
Nitrogen 196 kga 2687.112
Phosphorus 98 kga 374.5104
Potassium 196 ka 337.0593
Lime 1,120 kga 1051.304
Herbicides 0.8 kga 34.39381
Insecticides 0.3 kga 12.89768
Fungicides 1.5 kga 64.48839
Electricity 40 kWha 86.36668
Transport 228 kga 0
Yield 48,000 kg
Total without labor kcal 9593758


MJ
MJ/kg


40167.15
0.84


Table D-13. Re
Input Q
Machinery
Diesel
Nitrogen
Phosphorus
Potassium
Lime
Herbicides
Insecticides
Fungicides
Electricity
Transport
Yield
Total without labor


calculation of Pimentel (2009) energy demand for US apples.
quantity Unit CED (1E3 kcal)
88 kga 2123
2,000 Ld 22193
50 kge 685.5
114 kga 435.7
114 kga 196
682 kga 640.2
6 kgi 258
47 kgi 2021
49 kga 2107
40 kWh 86.37
3,000 kgk 0
54,000 kg
kcal 3E+07
MJ 1E+05
MJ/kg 2.4


177









Table D-14. Recalculation of Coltro (2009) energy demand for BR oranges.
Input Quantity Unit CED (MJ/kg)
Diesel 4.19 kg 53.4
Fertilizers (NPK) 11.75 kg 48.4
Bactericide 0.017 kg 180
Acaricide 1.12 kg 180
Fungicide 0.049 kg 180
Herbicide 0.149 kg 180
Insecticide 0.0093 kg 180
Lime 17.75 kg 3.93
Yield 1000 kg


MJ
MJ/kg


1005.73
1.0


Table D-15. CED values for
and Apples US.


inputs used in recalculations of Orange BR, Orange US


Process
pesticide, unspecified
ammonium nitrate, as N
diesel, at regional storage
single superphoshate, as
P205
tractor, production
lime, hydrated, packed, at
plant
electricity, US
potassium chloride, as K20


Amount
1
1
1

0.436
1


Unit
kg
kg
kg


NR fossil
CED (MJ)
180
57.4
53.4


1 kg
1 kWH
0.837 kg


178


16.0
101


--










90%


- f 7IL I--*
-; I f -. I:
-,r: I: ( -rL. I. I:*


70Q% T -i: i L. l r Li
70 QT -.(* r-..Ir- *LI I


20%

10%
Ha

Uo


Figure D-1. Emission fractions of applied pesticides in
default.


PestLCI-CR vs. the PestLCI


179













l OE+06


S.OE+06




1,OE+' J




I.OE+03





1.OE+02




I.OE+01


0 E.:r:
nE.r:
- E :r:
rE :r:


* r EmiIsion to
. l i E i i.'-.' I l, r .
* r i E llii. .l: 1 t.

. r i i E ll i .- i :i r


Water
11 I I

' ti: i
n -i1 '


- LSEtox-C
Er: -.F
I E : -E
Er: -EI'


I.OE+0W :'.' f:' I1. til t I I:I 1:Ir I:I t:

.. .


Figure D-2. Freshwater ecotoxicity characterization factors for pesticides in USETox-
CR vs USETox-Default


180


I


















I.OE-03


I .E-09


2'


l Human Tox for Emission to Water USEtox-CR
ElHumanTox for Emission to Air -1.1 Er: -CR
SH '.IIIII IT : r: E II:.I r: r I Er .-E'
'O HH i n ..i T : ri: Elli-.li: r: i I. Et: -El'


I.OE-0N


1IOE-O5


1.OE-06





1.OE-07


1, OE-08


Figure D-3. Human toxicity characterization factors for pesticides in USETox-CR vs
USETox-Default















181


''
'"













1.OE-06 a-I I ai l urnan I 1:ir. ri II. emissions [cases Kg pineapple. .OE-01

rTotal Freshwi after E-:.r. i:r 'i e emissions [PAIF.m3. I I
pineapple
1.E-07




a. I .OE-O L '
1 .1EE-0
1. .E-01



L I.E-Ii II -



C* --.


V : 1.OE-04
*. .. ..* .. -

-1 .. .. ..--.., ,- -. -"
e e...-.



















: ." ,,, -' L
..igE- -, -fo -
7. :.. 1.-E-04



1 E 1 ,, : C: ... -
.........................*..... ....... C




*.. .., .-
E1e.e. C. 'C' 1.0 E-0 5


.: ....* ...e C.p C. ..
*... g .. C.
.... -. P..V
C. .*%. .-. 1.* E-06




1. E-15 7* 7 1.7E-07









Figure D-4. Human toxicity and freshwater ecotoxicity for pesticide emissions from pineapple production in the baseline
scenario.


182









LIST OF REFERENCES


Abou-Elela, S., H. Ibrahim, and E. Abou-Taleb. 2008. Heavy metal removal and cyanide
destruction in the metal plating industry: an integrated approach from Egypt. The
Environmentalist 28(3): 223-229.

Althaus, H.-J., M. Chudacoff, R. Hischier, N. Jungbluth, M. Osses, and A. Primas. 2004.
Life cycle inventories of chemicals. Final report ecoinvent 2000. No. 10.
Dubendorf, CH: Swiss Centre for LCI, EMPA-DU.

Australian Museum. 2007. Structure and composition of the Earth.
www.amonline.net.au/geoscience/earth/structure.htm. Accessed 9 September
2008.

Ayres, R. U., L. W. Ayres, and K. Martinas. 1998. Exergy, waste accounting, and life-
cycle analysis. Energy 23(5): 355-363.

Bach, O. 2008. Agriculture e implicaciones ambientales con 6nfasis en algunas
cuencas hidrogr6ficas principles [Agriculture and environmental implications
with emphasis on selected principal watersheds]. Decimotercer Informe de
Estado de la Nacion en Desarollo Sostenible. San Jose: Consejo de Rectores.

Baral, A. and B. R. Bakshi. 2010. Thermodynamic methods for aggregation of natural
resources in life cycle analysis: Insight via application to some transportation
fuels. Environmental Science & Technology 44: 800-807.

Bare, J., T. Gloria, and G. Norris. 2006. Development of the method and U.S.
normalization database for life cycle impact assessment and sustainability
metrics. Environmental Science & Technology 40: 5108-5115.

Bare, J., G. A. Norris, D. W. Pennington, and T. McKone. 2003. TRACI The tool for the
reduction and assessment of chemical and other environmental impacts. Journal
of Industrial Ecology 6: 49-78.

Bartelmus, P. 2003. Dematerialization and capital maintenance: two sides of the
sustainability coin. Ecological Economics 46: 61-81.

Bastianoni, S., A. Facchini, L. Susani, and E. Tiezzi. 2007. Emergy as a function of
exergy. Energy 32: 1158-1162.

Bastianoni, S., D. E. Campbell, R. Ridolfi, and F. M. Pulselli. 2009. The solar
transformity of petroleum fuels. Ecological Modelling 220(1): 40-50.

Birkveda, M. and M. Z. Hauschild. 2006. PestLCI-A model for estimating field
emissions of pesticides in agricultural LCA. Ecological Modelling 198: 433-451.


183









Blanke, M. M. and B. Burdick. 2009. An energy balance (as part of an LCA) for home-
grown (apple) fruit versus those imported from South Africa or New Zealand.
Paper presented at Joint North American LCA Conference, 2 October, Boston.

Bosch, M. E., S. Hellweg, M. A. J. Huijbregts, and R. Frischknecht. 2007. Applying
Cumulative Exergy Demand (CExD) indicators to the ecoinvent database. Int J
LCA 12(3): 181-190.

Boustead, I. and G. F. Hancock. 1978. Handbook of industrial energy analysis. New
York: Ellis Horwood Ltd.

Brentrup, F. and J. KCsters. 2000. Methods to estimate potential N emissions related to
crop production. In Agricultural data for Life Cycle Assessments, edited by B. P.
Weidema and M. J. G. Meeusen. The Hague: Agricultural Economics Research
Institute (LEI).

Brown, M. T. 2009. Personal Communication with Brown, M. T., Professor of
Environmental Engineering Sciences. Gainesville, FL, 10 September 2009.

Brown, M. T. and S. Ulgiati. 1997. Emergy based indices and ratios to evaluate
sustainability: monitoring economies and technology to toward environmentally
sound innovation. Ecological Engineering 9: 51-69.

Brown, M. T. and S. Ulgiati. 2002. Emergy evaluations and environmental loading of
electricity production systems. Journal of Cleaner Production 10(4): 321-334.

Brown, M. T. and V. Buranakarn. 2003. Emergy indices and ratios for sustainable
material cycles and recycle options. Resources, Conservation and Recycling 38:
1-22.

Brown, M. T. and S. Ulgiati. 2004. Emergy and environmental accounting. In
Encyclopedia of Energy, edited by C. Cleveland. New York: Elsevier.

Brown, M. T., M. J. Cohen, and S. Sweeney. 2009. Predicting national sustainability:
The convergence of energetic, economic and environmental realities. Ecological
Modelling 220(23): 3424-3438.

Brown, M. T., M. J. Cohen, E. Bardi, and W. W. Ingwersen. 2006. Species diversity in
the Florida Everglades, USA: A systems approach to calculating biodiversity.
Aquatic Sciences 68(3): 254-277.

Brunner, P. and H. Rechburger. 2003. Practical handbook of material flow analysis.
Vero Beach, FL: CRC Press.

Buenaventura Mining Company Inc. 2006. Form 20-F for fiscal year 2005., edited by
SEC.


184









Buranakarn, V. 1998. Evaluation of Recycling and Reuse of Building Materials Using
the Emergy Analysis Method. Ph.D. thesis, University of Florida, Gainesville.

Burt, R. 2009. Soil survey field and laboratory methods manual. Soil Survey
Investigations Report No. 51. Lincoln, Nebraska: National Soil Survey Center,
Natural Resources Conservation Service, U.S. Department of Agriculture.

Butterman, W. C. and H. E. Hilliard. 2004. Silver. Mineral Commodity Profiles. Reston,
VA: U.S. Geological Survey.

Butterman, W. C. and E. B. Amey. 2005. Gold. Mineral Commodity Profiles. Reston,
Virginia: U.S. Geological Survey.

Campbell, D. 2001. A note on uncertainty in estimates of transformities based on global
water budgets. In Proceedings of the Second Biennial Emergy Analysis
Research Conference. Gainesville, FL: Center for Environmental Policy,
University of Florida.

Campos, L. 2007. Gestion de los recursos hidricos en las quencas con localizacion
minera: Caso Yanacocha [Management of hydrologic resources in watersheds
located in mining areas: The Case of Yanacocha]. Cajamarca, Peru: Minera
Yanacocha, S.R.L.

Canals, L. M. 2003. Contributions to LCA methodology for agricultural systems: Site
dependency and soil impact assessment. Ph.D. thesis, Universidad Autonoma,
Barcelona.

Chapagain, A. K. and A. Y. Hoekstra. 2004. Water footprints of nations. Vol. 1, Value of
Water Research Report Series No. 16. Delft, The Netherlands: UNESCO-IHE
Delft.

Cherubini, F., M. Raugei, and S. Ulgiati. 2008. LCA of magnesium production -
Technological overview and worldwide estimation of environmental burdens.
Resources Conservation and Recycling 52(8-9): 1093-1100.

Christiansen, K., M. Wesnaes, and B. P. Weidema. 2006. Consumer demands on Type
III environmental declarations. Copenhagen: 2.0 LCA consultants.

Classen, M., H.-J. Althaus, S. Blaser, G. Doka, N. Jungbluth, and M. Tuchschmid. 2007.
Life cycle inventories of metals. Final report ecoinvent data v2.0. Dcbendorf, CH:
Swiss Centre for LCI, Empa TSL.

Coderre, F. and D. G. Dixon. 1999. Modeling the cyanide heap leaching of cupriferous
gold ores Part 1: Introduction and interpretation of laboratory column leaching
data. Hydrometallurgy 52: 151-175.


185









Cohen, M., S. Sweeney, and M. T. Brown. 2008. Computing the unit emergy value of
crustal elements. In Proceedings of the 4th Biennial Emergy Conference, edited
by M. T. Brown. Gainesville, FL: Center for Environmental Policy, University of
Florida.

Cohen, M. J. 2001. Dynamic emergy simulation of soil genesis and techniques for
estimating transform ity confidence envelopes. In Proceedings of the Second
Biennial Emergy Analysis Research Conference. Gainesville, FL: Center for
Environmental Policy, University of Florida.

Coltro, L., A. Mourad, R. Kletecke, T. Mendonga, and S. Germer. 2009. Assessing the
environmental profile of orange production in Brazil. The International Journal of
Life Cycle Assessment 14(7): 656-664.

Condori, P., S. Garcia, and C. Ramon. 2007. Administration y optimizacion de
operaciones de heap leaching haciendo uso de un simulador de process en
Minera Yanacocha. In 28th Convencion Minera. 10-14 September: Instituto de
Ingenieros de Minas de Peru.

Cuadra, M. and J. Bjorklund. 2007. Assessment of economic and ecological carrying
capacity of agricultural crops in Nicaragua. Ecological Indicators 7: 133-149.

Daly, G. L., Y. D. Lei, C. Teixeira, D. C. G. Muir, L. E. Castillo, and F. Wania. 2007.
Accumulation of Current-Use Pesticides in Neotropical Montane Forests.
Environmental Science & Technology 41(4): 1118-1123.

Dones, R., B. Bauer, R. Bolliger, B. Burger, M. Faist Emmenegger, R. Frischknecht, T.
Heck, N. Jungbluth, and A. R6der. 2003. Sachbilanzen von Energiesystemen.
Final report ecoinvent 2000. Volume: 6. DCbendorf and Villigen, CH: Swiss
Centre for LCI, PSI.

Durucan, S., A. Korre, and G. Munoz-Melendez. 2006. Mining life cycle modelling: a
cradle-to-gate approach to environmental management in the minerals industry.
Journal of Cleaner Production 14(12-13): 1057-1070.

Ebeling, J. and M. Yasue. 2008. Generating carbon finance through avoided
deforestation and its potential to create climatic, conservation and human
development benefits. Philosphical Transactions of the Royal Society B 363:
1917-1924.

Ecoinvent Centre. 2007. Ecoinvent data v2.0. DCbendorf, CH: Swiss Centre for Life
Cycle Inventories.

Economic Commission of Latin American and the Carribbean. 2006. Statistical
yearbook for Latin America and the Caribbean, 2006. Santiago, Chile: Economic
Commission of Latin American and the Carribbean.


186









Ehrlich, H. and D. Newman. 2008. Geomicrobiology. 5th ed. Boca Raton, FL: CRC
Press.

Energy Information Administration. 2007. Peru energy data, statistics and analysis oil,
gas, electricity, coal. Washington, DC.: Department of Energy.

European Commission. 2003. Communication of the (European) Commission to the
Council and the European Parliament on Integrated Product Policy. COM(2003)
302 final.

FAO. 2006. Fertilizer use by crop. FAO Fertilizer and Plant Nutrition Bulletin. Rome:
Food and Agricultural Organization of the United Nations.

FAO. 2009. FAOSTAT Trade Database. http://faostat.fao.org/site/342/default.aspx.
Accessed 8 July 2009.

FAO. 2010. Web LocClim, local monthly climate estimator.
http://www.fao.org/sd/locclim/srv/locclim.home. Accessed 7 January 2010.

Fargione, J., J. Hill, D. Tilman, S. Polasky, and P. Hawthorne. 2008. Land clearing and
the biofuel carbon debt. Science 319(5867): 1235-1238.

Fava, J. A. A. A. J., L. Lindfors, S. Pomper, B. d. Smet, J. Warren, and B. Vigon. 1994.
Lifecycle assessment data quality. A conceptual framework. Pensacola, FL:
SETAC.

Federici, M., S. Ulgiati, and R. Basosi. 2008. A thermodynamic, environmental and
material flow analysis of the Italian highway and railway transport systems.
Energy 33(5): 760-775.

Finnveden, G. 2005. The resource debate needs to continue. The International Journal
of Life Cycle Assessment 10(5): 372-372.

Foster, G. R., D. Yoder, and S. Dabney. 2008. Revised Universal Soil Loss Equation 2
(RUSLE2) ARS Version May 20, 2008. USDA Agricultural Research Service,
Oxford, MS.

Franzese, P. P., T. Rydberg, G. F. Russo, and S. Ulgiati. 2009. Sustainable biomass
production: A comparison between gross energy requirement and emergy
synthesis methods. Ecological Indicators 9(5): 959-970.

Frischknecht, R. 1997. Goal and scope definition and inventory analysis. In Life Cycle
Assessment: State of the Art and Research Priorites, edited by H. U. d. Haes and
N. Wrisberg. Bayreuth: Ecomed Publishers.

Frischknecht, R. and N. Jungbluth. 2007. Implementation of Life Cycle Impact Methods.
Data v2.0 (2007). Ecoinvent report No. 3. DCbendorf, CH: Swiss Centre for Life
Cycle Inventories.


187









Frischknecht, R., N. Jungbluth, H.-J. Althaus, G. Doka, R. Dones, T. Heck, S. Hellweg,
R. Hischier, T. Nemecek, G. Rebitzer, M. Spielmann, and G. Wernet. 2007.
Ecoinvent report No. 1: Overview and methodology. Dubendorf: Swiss Centre for
Life Cycle Inventories.

Gabby, P. N. 2007. Lead. In U.S. Geological Survey Minerals Yearbook 2005, edited
by USGS. Washingon, DC: USGS.

Gaillard, G. and T. Nemecek. 2009. Editorial. Paper presented at 6th International
Conference on LCA in the Agri-Food Sector, November 12-14, Zurich.

Gallego, A., L. Rodriguez, A. Hospido, M. T. Moreira, and G. Feijoo. 2010. Development
of regional characterization factors for aquatic eutrophication. International
Journal of Life Cycle Assessment 15(1): 32-43.

Gallopin, G. 2003. A systems approach to sustainability and sustainable development.
Santiago, Chile: Sustainable Development and Human Settlements Division,
United Nations Economic Commission for Latin America and the Caribbean.

GeoNews. 2008. KML Polygon Area Tool. http://www.geo-
news.net/index_area_poligono.php. Accessed August 12, 2008.

Giljum, S. 2004. Trade, materials flows, and economic development in the South: The
example of Chile. Journal of Industrial Ecology 8: 241-263.

Gloria, T. 2009. Determination of empirical allocation measures for non-ferrous metals.
Paper presented at Joint North American Life Cycle Conference, 1 October,
Boston.

Gobin, A., G. Govers, R. Jones, M. Kirkby, and C. Kosmas. 2003. Assessment and
reporting on soil erosion Background and workshop report Copenhagen:
European Environment Agency.

Goedkoop, M. and R. Spriensma. 2001. The Eco-indicator 99: A damage-oriented
method for LCA. Amersfoort, NL: PRe consultants.

G6mez, M. P., P. P. Quesada, and K. M. Bucheli. 2007. Implementation of good
practices in the production of fresh pineapples for export: Case study of the
Huetar Norte region, Costa Rica. In Implementing programmes to improve safety
and quality in fruit and vegetable supply chains: benefits and drawbacks. Latin
American case studies, edited by L. B. D. R. Maya PiAeiro. Rome: Food and
Agriculture Organization of the United Nations.

Google. 2008. Google Earth 4 software. Palo Alto, CA: Google.

Gossling-Reisemann, S. 2008a. What is resource consumption and how can it be
measured? Theoretical considerations. Journal of Industrial Ecology 12(1): 10-
25.


188









Gossling-Reisemann, S. 2008b. What is resource consumption and how can it be
measured? Application of entropy analysis to copper production. Journal of
Industrial Ecology 12(4): 570-582.

Guinee, J. B., ed. 2002. Handbook on life cycle assessment: operational guide to the
ISO standards. Vol. 7, Eco-efficiency in industry and science. Doldrecht, The
Netherlands: Kluwer Academic.

Hails, C., S. Humphrey, J. Loh, and S. Goldfinger, eds. 2008. Living Planet Report
2008. Gland, Switzerland: WWF International.

Hankce, G. 1991. The effective control of a deep hole diamond drill. Paper presented at
Industry Applications Society Annual Meeting, 28 Sep-4 Oct, Dearborn, MI.

Hartley-B., M. and R. Diaz-P. 2008. Mejoras ambientales para el desarollo de la
competitividad en tres cadenas agroalimentarias costarricenses [Better
environments for competitive development of three Costa Rican agro-food
chains]. Heredia, Costa Rica: Centro Internacional de Politica Economica.

Hartman, H. L. 1992. SME mining engineering handbook. 2nd ed. Vol. 2. Littleton, CO:
Society for Mining, Metallurgy and Exploration.

Hau, J. L. and B. R. Bakshi. 2004a. Expanding exergy analysis to account for
ecosystem products and services. Environmental Science and Technology
38(13): 3768-3777.

Hau, J. L. and B. R. Bakshi. 2004b. Promise and problems of emergy analysis.
Ecological Modelling 178(1-2): 215-225.

Helmer, E. H. and S. Brown. 2000. Gradient analysis of biomass in Costa Rica and a
first estimate of countrywide emissions of greenhouse gases from biomass
burning. In Quantifying sustainable development the future of tropical economies,
edited by C. A. S. Hall, et al. San Diego: Academic Press.

Heuvelmans, G., J. F. Garcia-Qujano, B. Muys, J. Feyen, and P. Coppin. 2005.
Modelling the water balance with SWAT as part of the land use impact evaluation
in a life cycle study of CO2 emission reduction scenarios. Hydrological
Processes 19(3): 729-748.

Hill, A. R. and C. V. Hoist. 2001. A comparison of simple statistical methods for
estimating analytical uncertainty, taking into account predicted frequency
distributions. Analyst(126): 2044-2052.

Hoekstra, A. Y., A. K. Chapagain, M. M. Aldaya, and M. M. Mekonnen. 2009. Water
footprint manual State of the art 2009. Enschede, The Netherlands: Water
Footprint Network.

Holdridge, L. R. 1967. Life zone ecology. San Jose, CR: Tropical Science Center.


189









Hopper, R. 2008. Emergy synthesis of sulfuric acid. In EES5306 Energy Analysis class,
Spring 2008. Gainesville, FL: University of Florida.

Hosier, B. 2008. Personal Communication with Hosier, B., Phone conversation with
representative from Lindberg/MPH. November 3, 2007 2008.

Huijbregts, M. A. J., W. Gilijamse, A. M. J. Ragas, and L. Reijnders. 2003a. Evaluating
uncertainty in environmental life-cycle assessment. A case study comparing two
insulation options for a Dutch one-family dwelling. Environmental Science &
Technology 37(11): 2600-2608.

Huijbregts, M. A. J., S. Lundi, T. E. McKone, and D. v. d. Meent. 2003b. Geographical
scenario uncertainty in generic fate and exposure factors of toxic pollutants for
life-cycle impact assessment. Chemosphere 51: 501-508.

Huijbregts, M. A. J., S. Hellweg, R. Frischknecht, H. W. M. Hendriks, K. Hungerbui^hler,
and A. J. Hendriks. 2010. Cumulative Energy Demand as a predictor for the
environmental burden of commodity production. Environmental Science &
Technology 44(6): 2189-2196.

Infomine. 2005. Yanacocha Minesite. http://yanacocha.infomine.com. Accessed Sept. 9,
2007.

Ingwersen, W. W. 2010. Uncertainty characterization for emergy values. Ecological
Modelling 221(3): 445-452.

Ingwersen, W. W. Accepted. Emergy as a impact assessment method for life cycle
assessment presented in a gold mining case study. Journal of Industrial Ecology.

Ingwersen, W. W., S. A. Clare, D. Acuia, M. J. Charles, C. Koshal, and A. Quiros.
2009. Environmental Product Declarations: An introduction and
recommendations for their use in Costa Rica. Gainesville, FL: University of
Florida Levin College of Law Conservation Clinic.

Institute Nacional Estadistica y Informacion. 2006. Peru compendio estadistico 2006.
Lima, Peru: Instituto Nacional Estadistica y Informacion.

Institute Peruano de Economia. 2003. La brecha en infraestructura: Servicios publicos,
productividad, y crecimiento en el Peru. Lima:

International Mining News. 2005. The Yanacocha Seven. International Mining News
[Hertfordshire, UK].

IPCC. 2007. Climate change 2007. IPCC fourth assessment report. The physical
science basis. Geneva: International Panel on Climate Change.


190









ISO. 2006a. 14044: Environmental management -- Life cycle assessment --
Requirements and guidelines. Geneva: International Organization for
Standardization.

ISO. 2006b. 14025: Environmental labelling and declarations Type III environmental
declarations Principles and procedures. International Standard. Geneva:
International Organization for Standardization.

ISO. 2006c. 14040: Environmental management -- Life cycle assessment -- Principles
and framework. Geneva: International Organization for Standardization.

Jolliet, O., M. Margni, R. Charles, S. Humbert, J. Payet, G. Rebitzer, and R.
Rosenbaum. 2003a. IMPACT 2002+: A new life cycle impact assessment
methodology. International Journal of Life Cycle Assessment 8(6): 324-330.

Jolliet, O., A. Brent, M. Goedkoop, N. Itsubo, R. Mueller-Wenk, C. Peha, R. Schenk, M.
Stewart, and B. Weidema. 2003b. Final report of the LCIA Definition study.
UNEP/SETAC Life Cycle Initiative. United National Environmental Program.

Joyce, A. 2006. Land use change in Costa Rica 1966-2006 as influenced by social,
economic, political and environmental factors. San Jose: Litografia e Imprenta
LIL.

Kodjak, D. 2004. Policy discussion Heavy-duty truck fuel economy. Paper presented
at 10th Diesel Engine Emissions Reduction (DEER) Conference, 29 August -2
September, Coronado, CA.

La Rosa, A. D., G. Siracusa, and R. Cavallaro. 2008. Emergy evaluation of Sicilian red
orange production. A comparison between organic and conventional farming.
Journal of Cleaner Production 16(17): 1907-1914.

Lal, R. 1983. Soil erosion in the humid tropics with particular reference to agricultural
land development and soil management. Paper presented at Hydrology of Humid
Tropical Regions with Particular Reference to the Hydrological Effects of
Agriculture and Forestry Practice, 15 October, Hamburg.

Lenzen, M. and U. Wachsmann. 2004. Wind turbines in Brazil and Germany: an
example of geographical variability in life-cycle assessment. Applied Energy 77:
119-130.

Lillywhite, R., D. Chandler, W. Grant, K. F. Lewis, C., U. Schmutz, and D. Halpin. 2007.
Environmental footprint and sustainability of horticulture (including potatoes) A
comparison with other agricultural sectors. UK: DEFRA.

Limpert, E., W. A. Stahel, and M. Abbt. 2001. Log-normal distributions across the
sciences: Keys and clues. Bioscience 51(5): 341-352.


191









Lloyd, S. and R. Ries. 2007. Characterizing, propagating, and analyzing uncertainty in
life cycle assessment. Journal of Industrial Ecology 11(1): 161-179.

Longo, A. 2005. Evolution of volcanism and hydrothermal activity in the Yanacocha
Mining District, northern Peru. Ph.D. thesis, Oregon State University.

Lowrie, R. L. 2002. SME mining reference handbook. Littleton, CO: Society for Mining,
Metallurgy and Exploration.

Maia de Souza, D., R. Rosenbaum, L. Deschenes, and H. Lisboa. 2009. Crucial
improvements needed for land use impact assessment modeling concerning
biodiversity indicators. Paper presented at Life Cycle Assessment IX Joint North
American Life Cycle Conference, 29 September 2 October, Boston.

Malezieux, E., F. CBte, and D. P. Bartholomew. 2003. Crop environment, plant growth
and physiology. In The pineapple: Botany, production, and uses, edited by D. P.
Bartholomew, et al. Oxon, UK: CABI Pub.

Marsden, J. and I. House. 2006. The chemistry of gold extraction. 2nd ed: SME.

Matthews, E., C. Amman, S. Bringezu, M. Fischer-Kowalski, W. Huttler, R. Kleijn, Y.
Moriguichi, C. Ottke, E. Rodenburg, D. Rogich, H. Schandl, H. Schutz, E. V. d.
Voet, and H. Weisz. 2000. The weight of nations: Material outflows from
industrial economies. 1st ed. Washington, DC: World Resources Institute.

ME Assessment. 2005. Ecosystems & human well-being: Biodiversity synthesis,
Millineum Ecosystem Assessment. Washington, DC: World Resources Institute.

Miller, S. A., A. E. Landis, and T. L. Theis. 2006. Use of monte carlo analysis to
characterize nitrogen fluxes in agroecosystems. Environmental Science &
Technology 40(7): 2324-2332.

Mimbela, R. 2007. Filosofia y gestion de agua [Philosophy and management of water].
Lima: Minera Yanacocha S.R.L.

Minera Yanacocha S.R.L. 2005. Procidimiento: Plan Integral de Control de Polvo
[Procedure: Integrated plan to control dust]. MA-PA-026. Lima, Peru: Minera
Yanacocha S.R.L.

Minera Yanacocha S.R.L. 2006. La production del oro en Yanacocha [Gold production
at Yanacocha]. Informes de Centro de Informaccion. Cajamarca, Peru: Minera
Yanacocha S.R.L.

Minera Yanacocha S.R.L. 2007. Mine Tour. Cajamarca, Peru.

Mining Technology. 2007. Minera Yanacocha Gold Mine, Peru. http://www.mining-
technology.com. Accessed October 1, 2007.


192









Montgomery Watson. 1998. Estudio de impact ambiental: Proyecto La Quinua
[Environmental impact study: La Quinua project]. Santiago, Chile:

Montgomery Watson. 2004. Plan de cierre conceptual: La Quinua [Conceptual mine
closing plan: La Quinua]. Lima, Peru:

Montoya, P. and J. Quispe. 2007. Maqui maqui: Ejemplo de cierre exitoso [Maqui
maqui: Example of a successful mine closing]. DDC-Fabrica de Ideas.

NAS. 1999. Nature's numbers. Edited by W. N. a. E. Kokkelenburg. Washington, DC:
National Academy of Sciences.

National Metal Finishing Resource Center. 2007. Pollution prevention and control
technologies for plating operations. http://www.nmfrc.org/. Accessed October 20,
2007.

National Renewable Energy Laboratory. 2008. Notes regarding transparency, data
publishing (unit processes) and data exchange. Life Cycle Assessment Working
Paper No. 7. Golden, Colorado:

Nemecek, T. and T. Kagi. 2007. Life cycle inventory of agricultural production systems.
Dubendorf: Ecoinvent Centre.

Ness, B., E. Urbel-Piirsalu, S. Anderberg, and L. Olsson. 2007. Categorising tools for
sustainability assessment. Ecological Economics 60(3): 498-508.

Newmont. 2004. Social and environmental responsibility. Denver, CO: Newmont
External Affairs and Communication Department.

Newmont. 2006a. Now and beyond 2005 sustainability report: Minera Yanacocha, Peru.
Denver, Colorado: Newmont.

Newmont. 2006b. 2005 annual report. Denver, Colorado:

Newmont. 2006a. Now & Beyond 2005 Corporate Sustainability Report. Denver,
Colorado:

Newmont. 2006c. Form 10-K for fiscal year 2005, edited by SEC.

Newmont Waihi Gold. 2007. Equipment at the Martha mine.
http://www.newmont.com/en/operations/australianz/waihigold/mining/index.asp.
Accessed November 1, 2007.

NIST. 2010. The NIST reference on constants, units, and uncertainty.
http://physics.nist.gov/cuu/Uncertainty/combination.html. Accessed 26 January
2010.


193









Norris, G. A. 2003. Impact characterization in the tool for the reduction and assessment
of chemical and other environmental impacts (TRACI) Methods for acidification,
eutrophication, and ozone formation. Journal of Industrial Ecology 6(3-4): 79-101.

O'Brien, E., B. Guy, and A. S. Lindner. 2006. Life cycle analysis of the deconstruction of
military barracks: Ft. McClellan, Anniston, AL. Journal of Green Building 1(4):
166-183.

Odum, H. T. 1988. Self organization, transformity, and information. Science 242: 1132-
1139.

Odum, H. T. 1996. Environmental Accounting. New York: John Wiley & Sons.

Odum, H. T. 2007. Environment, power and society for the twenty-first century: The
hierarchy of energy. New York: Columbia University Press.

Odum, H. T. 1991. Emergy of South African gold. In Ecological Physical Chemistry.
Proceeding of a Conference, edited by C. Rossi and E. Tiezzi. Siena, Italy:
Elsevier.

Odum, H. T., M. T. Brown, and S. Brandt-Williams. 2000. Handbook of emergy
evaluation folio #1: Introduction and global budget. Gainesville: Center for
Environmental Policy, University of Florida.

Pennington, D. W., M. Margni, C. Ammann, and 0. Jolliet. 2005. Multimedia fate and
human intake modeling: Spatial versus nonspatial insights for chemical
emissions in Western Europe. Environmental Science & Technology 39(4): 1119-
1128.

Peruvian Ministry of Energy and Mines. 2006. Annual Production 2005: Gold. Lima,
Peru:

Peters, G. M., H. V. Rowley, S. Wiedemann, R. Tucker, M. D. Short, and M. Schulz.
2010. Red meat production in Australia: Life cycle assessment and comparison
with overseas studies. Environmental Science & Technology 44(4): 1327-1332.

Pfister, S., A. Koehler, and S. Hellweg. 2009. Assessing the environmental impacts of
freshwater consumption in LCA. Environmental Science & Technology 43(11):
4098-4104.

Pimentel, D. 2009. Energy inputs in food crop production in developing and developed
nations. Energies 2: 1-24.

Pizzigallo, A. C. I., C. Granai, and S. Borsa. 2008. The joint use of LCA and emergy
evaluation for the analysis of two Italian wine farms. Journal of Environmental
Management 86(2): 396-406.


194









Powers, S. E. 2007. Nutrient loads to surface water from row crop production.
International Journal of Life Cycle Assessment 12(6): 399-407.

PRe Consultants. 2008. SimaPro 7.1. Ph.D. Version., Amsfoort, NL.

Rai, S. N. and D. Krewski. 1998. Uncertainty and variability analysis in multiplicative risk
models. Risk Analysis 18(1): 37-45.

Reap, J., F. Roman, S. Duncan, and B. Bras. 2008. A survey of unresolved problems in
life cycle assessment: Part 2: impact assessment and interpretation. International
Journal of Life Cycle Assessment 13(5): 374-388.

Ridoutt, B. G. and S. Pfister. 2010. A revised approach to water footprinting to make
transparent the impacts of consumption and production on global freshwater
scarcity. Global Environmental Change 20(1): 113-120.

Ridoutt, B. G., P. Juliano, P. Sanguansri, and J. Sellahewa. 2009. Consumptive water
use associated with food waste. Hydrology and Earth System Sciences
Discussions 6: 5085-5114.

Roos, E., C. Sunderberg, and P.-A. Hansson. 2010. Uncertainties in the carbon footprint
of food products: a case study on table potatoes. International Journal of Life
Cycle Assessment 15(5): 478-488.

Rosenbaum, R. K., T. M. Bachmann, L. S. Gold, M. A. J. Huijbregts, O. Jolliet, R.
Juraske, A. Koehler, H. F. Larsen, M. MacLeod, M. Margni, T. E. McKone, J.
Payet, M. Schuhmacher, D. van de Meent, and M. Z. Hauschild. 2008. USEtox-
the UNEP-SETAC toxicity model: Recommended characterisation factors for
human toxicity and freshwater ecotoxicity in life cycle impact assessment.
International Journal of Life Cycle Assessment 13(7): 532-546.

Rubin, B. D. and G. G. Hyman. 2000. The extent and economic impacts of soil erosion
in Costa Rica. In Quantifying sustainable development the future of tropical
economies, edited by C. A. S. Hall, et al. San Diego: Academic Press.

Rydburg, T. 2010. Personal Communication with Rydburg, T., Professor of
Environmental Science. Gainesville, FL 2010.

Sandoval, A. C. C. 2009. Insensatez piiera [Foolish pineapple production]. El
Financiero [San Jose, CR], July 5, section En Portada.

Schenck, R. 2007. Canning green beans Ecoprofile of Truitt Brothers process.
Vashon, WA: Institute for Environmental Research and Education.

Schenck, R. 2009. The outlook and opportunity for Type III environmental product
declarations in the United States of America. White Paper. Vashon, WA: Institute
for Environmental Research and Education.


195









Schenck, R. C. and S. Vickerman. 2001. Developing a land use/biodiversity indicator for
agricultural product LCAs. In Proceedings of the First International Conference
on LCA in Foods. Gothenburg, Sweden.

Schmidt-Bleek, F. 1994. Wieviel Umwelt braucht der Mensch? MIPS, das Mass fur
okologisches Wirtschaften [How much environment do we need? MIPS, the
measure for ecologically sound economic performance]. Berlin: Birkhauser.

Scholl, D. and v. Huene. 2004. Crustal recycling at ocean margin and continental
subduction zones and the net accumulation of continental crust. EOS
Transactions 88(52).

Sciubba, E. and S. Ulgiati. 2005. Emergy and exergy analyses: Complementary
methods or irreducible ideological options? Energy 30(10): 1953-1988.

Seager, T. P. and T. L. Theis. 2002. A uniform definition and quantitative basis for
industrial ecology. Journal of Cleaner Production 10: 225-235.

Seppala, J., S. Knuuttila, and K. Silvo. 2004. Eutrophication of aquatic ecosystems A
new method for calculating the potential contributions of nitrogen and
phosphorus. International Journal of Life Cycle Assessment 9(2): 90-100.

Sinden, G. 2008. PAS 2050:2008 Specification for the assessment of the life cycle
greenhouse gas emissions of goods and services. London: British Standards
Institute.

Slob, W. 1994. Uncertainty analysis in multiplicative models. Risk Analysis 14(4): 571-
576.

Sonnemann, G. and B. de Leeuw. 2006. Life cycle management in developing
countries: State of the art and outlook. International Journal of Life Cycle
Assessment 11 (Special Issue 1): 123-126.

Spielmann, M., T. Kagi, P. Stadler, and 0. Tietje. 2004. Life cycle inventories of
transport services. Final report ecoinvent 2000. Volume: 14., UNS. Dubendorf,
CH: Swiss Centre for LCI.

Stewart, M. and B. P. Weidema. 2005. A consistent framework for assessing the
impacts from resource use A focus on resource functionality. The International
Journal of Life Cycle Assessment 10(4): 240-247.

Stratus Consulting. 2003. Report on the independent assessment of water quantity and
quality near the Yanacocha mining district, Cajamarca, Peru. Washington, DC:
IFC/MIGA Compliance Advisor.

Su, N. R. 1968. Pineapple (Ananas comosus (L) Merr.) nutritional requirements. Taipei:
Taiwan Council of Agriculture.


196









Sweeney, S., M. Cohen, D. King, and M. Brown. 2009. National Environmental
Accounting Database.
http://sahel.ees.ufl.edu/frame_database_resources_test.php. Accessed 10 May
2009.

Swennenhuis, J. 2009. CROPWAT version 8.0. Water Resources Development and
Management Service, FAO, Rome.

Taylor, S. R. and S. M. McLennan. 1985. The continental crust: its composition and
evolution : an examination of the geochemical record preserved in sedimentary
rocks. Palo Alto, CA: Blackwell Scientific.

The Tank Shop. 2007. Tank Weight Calculator Spreadsheet Tool.
http://www.thetankshop.ca/private/admin/upload/xls/Tank%20Weight%20Calcula
tor.xls. Accessed October 10, 2008.

Thiesen, J., S. Valdivia, G. Sonnemann, J. Fava, T. Swarr, A. A. Jensen, and E. Price.
2007. Understanding challenges and needs: A stakeholder consultation on
business' applications of life cycle approaches. In CICLA 2007. Sao Paolo,
Brazil.

Thornton, I. and S. Brush. 2001. Lead: The facts. London: IC Consultants Ltd.

Tilley, D. R. 2003. Industrial ecology and ecological engineering: Opportunities for
symbiosis. Journal of Industrial Ecology 7(2): 13-32.

Ukidwe, N. and B. R. Bakshi. 2004. Thermodynamic accounting of ecosystem
contribution to economic sectors with application to 1992 U.S. economy.
Environmental Science & Technology 38: 4810-4827.

Ulgiati, S., M. Raugei, and S. Bargigli. 2006. Overcoming the inadequacy of single-
criterion approaches to Life Cycle Assessment. Ecological Modelling 190(3-4):
432-442.

UN. 1992. Declaration on environment and development. Rio de Jainero, Brazil: United
Nations.

UN. 2005. Johannesburg Plan of Implementation. Johannesburg, SA: United Nations.

UN DESA. 2008. The Marrakech Process. http://esa.un.org/marrakechprocess/.
Accessed 19 May 2010.

UNEP. 2007. Life cycle management A business guide to sustainability. Paris: United
Nations Environment Programme.

UNEP Life Cycle Initiative. 2007. Life Cycle Initiative Phase 2 2007-2010. UNEP DTIE
Project Brief. Paris, France: United Nations Environment Programme, Division of
Technology, Industry & Economics.


197









United Nations. 2008. UN Comtrade Database. http://comtrade.un.org. Accessed March
22, 2008.

UNSTAT. 2006. Demographic Yearbook-Table 3: Population by sex, rate of population
increase, surface area and density.
http://unstats.un.org/unsd/demographic/products/dyb/dyb2006/Table03.pdf.
Accessed 13 August 2008.

UoH. 2005. Sustainability of UK Strawberry Crop. University of Hertfordshire.

Urban, R. A. and B. R. Bakshi. 2009. 1,3-Propanediol from fossils versus biomass: A life
cycle evaluation of emissions and ecological resources. Industrial & Engineering
Chemistry Research 48(17): 8068-8082.

USDA. 2009. National Nutrient Database for Standard Reference, Release 22.
http://www.nal.usda.gov. Accessed November 2, 2009.

Van Der Voet, E., L. Van Oers, and I. Nikolic. 2004. Dematerialization: Not just a matter
of weight. Journal of Industrial Ecology 8(4): 121-137.

Wackernagel, M., N. B. Schulz, D. Deumling, A. C. Linares, M. Jenkins, V. Kapos, C.
Monfreda, J. Loh, N. Myers, R. Norgaard, and J. Randers. 2002. Tracking the
ecological overshoot of the human economy. Proceedings of the National
Academy of Sciences 19(14): 9266-9271.

Weidema, B. and G. Norris. 2002. Avoiding co-product allocation in the metals sector. In
Life cycle assessment of metals: Issues and research directions, edited by A.
Dubriel. Pensacola, FL: Society of Environmental Toxicology and Chemistry.

Williams, A., E. Pell, J. Webb, E. Moorhouse, and E. Audsley. 2008. Strawberry and
tomato production for the UK compared between the UK and Spain. Paper
presented at International Conference on LCA in the Agri-Food Sector,
November 12-14, Zurich.

Wong, S. S., T. T. Teng, A. L. Ahmada, A. Zuhairi, and G. Najafpour. 2006. Treatment
of pulp and paper mill wastewater by polyacrylamide (PAM) in polymer induced
flocculation. Journal of Hazardous Materials B135: 378-388.

World Gold Council. 2006. Mine Production.
www.gold.org/value/markets/supply_demand/mine_production.html. Accessed
12 October 2007.

Yellishetty, M., P. G. Ranjith, A. Tharumarajah, and S. Bhosale. 2009. Life cycle
assessment in the minerals and metals sector: A critical review of selected
issues and challenges. International Journal of Life Cycle Assessment 14(3):
257-267.


198









Zhang, Y., Z. Yang, and X. Yu. 2009. Ecological network and emergy analysis of urban
metabolic systems: Model development, and a case study of four Chinese cities.
Ecological Modelling 220(11): 1431-1442.

Zhang, Y., S. Singh, and B. R. Bakshi. 2010. Accounting for ecosystem services in life
cycle assessment, part I: A critical review. Environmental Science & Technology
44(7): 2232-2242.


199









BIOGRAPHICAL SKETCH

Wesley W. Ingwersen was born in Atlanta, GA in 1977 and grew up in the Stone

Mountain area. He went to secondary school at Woodward Academy in College Park,

GA, where he developed a keen interest in environmental science. After a year at

Wake Forest University he transferred to Georgetown University (Washington, DC)

where he completed a B.A. in 1999. Wesley worked for an e-commerce company,

enews.com, and a software development company, Lokitech, as a web designer and

Internet applications developer until 2002. While in the DC area and volunteering with

the National Park Service and the Casey Tree Foundation, he became determined to

work toward greater scientific understanding of the dependence of human systems

upon nature and the value it provides, and returned to graduate school to pursue an

M.S. in Environmental Engineering at the University of Florida. His M.S. thesis was an

evaluation of long-term term success of wetland reclamation efforts on phosphate-

mined lands. Following the completion of his M.S. degree, Wesley joined Ecologic, and

environmental policy think-tank in Berlin as a Transatlantic Fellow, and at the end of

2006 returned to UF to pursue a Ph.D. under his M.S. adviser, Mark T. Brown.

Wesley is a Life Cycle Assessment Certified Professional. In addition to the LCA

work in this dissertation, he contributed to a study of future transportation-related GHG

emissions for the state of Florida, led a feasibility study of environmental product

declarations (EPDs) in Costa Rica, and is involved in the development of national

guidance standards for EPDs in the US. He has published book chapters, peer-

reviewed journal articles, and presented papers for conferences on issues of trade and

the environment, environmental assessment, life cycle assessment, uncertainty

modeling, and emergy analysis.


200





PAGE 1

1 A DVANCES IN LIFE CYCL E ASSESSMEN T AND EMERGY EVALUATIO N WITH CASE STUDIES IN GOLD MINING AND PINEAPPLE PRODUCTION By WESLEY W. INGWERSEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PAR TIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Wesley W. Ingwersen

PAGE 3

3 To the memory of James H. Weeks and Blanche R. Ingwersen, two of my grandparents who passed away late in the cou rse of my Ph.D. program, but who believed in me and forever inspire me

PAGE 4

4 ACKNOWLEDGMENTS I first and foremost thank Dr. Mark Brown, my Ph.D. adviser who provided me the opportunity to complete the degree and inspir ed the pursuit through his teaching a nd innovative work. I thank all my other committee members for the insights and criticisms that were incorporated into this dissertation. My studies and field work would not have been possible without the financial and administrative support of the Dep artment of Environmental Engineering Sciences. I received the support of a Latin American Studies Tinker Travel grant for my research in Peru. My research in Costa Rica was facilitated by the University of Florida University of Costa Rica Conservation Clinic under the direction of Tom Ankersen. Numerous persons provided me direct support for my field studies. I thank in particular Ricardo Gallardo in Cajamarca, Peru and various employees of the Yanacocha mine Randall Arias from PROCOMER in Costa Rica, and Dr. Mauricio Avila from the University of Wisconsin. I anonymously thank the pineapple companies that agreed to participate in this study. Through the formation and evolution of my topics, my travels, data analysis, and final drafting, my wife Laura has been my steadfast intellectual and emotional companion. Finally, m y family have provided me the encourage ment to bring this intellectual and personal journey to fruition a heartfelt thank you to all of you.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTR ACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 Measurement of Sustainable Production and Consumption ................................ ... 14 Life Cycle Assessment as a Measurement Tool ................................ ..................... 15 Research Problems in Life Cycle Assessment ................................ ....................... 17 Life Cycle Impact Assess ment (LCIA) Indicators for Resource Use ................. 17 Applications of LCA for N on OECD C ountry P roducts ................................ ..... 23 Research Overview ................................ ................................ ................................ 27 2 EMERGY AS AN IMPACT ASSESSMENT METHOD FOR LIFE CYCLE ASSESSMENT PRESENTED IN A GOLD MINING CASE STUDY ........................ 29 Introduction ................................ ................................ ................................ ............. 29 Emergy in the LCA Context ................................ ................................ .............. 29 A Case Study of Emergy in an LCA of Gold Silver Bullion Production ............. 33 Me thodology ................................ ................................ ................................ ........... 36 Emergy and Energy Calculations ................................ ................................ ..... 38 Uncertainty Modeling ................................ ................................ ........................ 40 Al location ................................ ................................ ................................ .......... 41 Data Management and Tools ................................ ................................ ........... 41 Results ................................ ................................ ................................ .................... 42 Environmental Contrib ution to Gold, Silver, and Mercury in the Ground .......... 42 Environmental Contribution to Dor ................................ ................................ 43 Emergy by Unit Process ................................ ................................ ................... 44 Allocation and Emergy Uncertainty ................................ ................................ .. 46 Discussion ................................ ................................ ................................ .............. 47 Usefulness of Emergy Results ................................ ................................ ......... 47 Emergy in LCA: Challenges ................................ ................................ ............. 50 Challenges of using emergy with LCI databases and software .................. 51 Energy in environmental support not conventionally included in emergy evaluation ................................ ................................ ............................... 53 Uncertainty in unit emergy values ................................ .............................. 54 Eme rgy and Other Resource Use Indicators ................................ .................... 55

PAGE 6

6 3 UNCERTAINTY CHARACTERIZATION FOR EMERGY VALUES ......................... 58 Introduction ................................ ................................ ................................ ............. 58 Sources of Uncertainty in UEVs ................................ ................................ ....... 59 Models for Describing Uncertainty in Lognormal Distributions ......................... 60 Models for Uncertainty in UEVs ................................ ................................ .............. 62 Selecting Appropriate Methods for Uncertainty Estimations ............................. 62 Modeling Procedure and Analysis ................................ ................................ .... 64 Results ................................ ................................ ................................ .................... 70 Discussion and Conclusions ................................ ................................ ................... 74 How Much Uncertainty is i n a UEV and Can it Be Quantified? ......................... 74 Comparing the Analytical and Stochastic Solutions ................................ ......... 75 Conclusions ................................ ................................ ................................ ............ 77 4 LIFE CYCLE ASSESSMENT FOR FRESH PINEAPPLE FROM COSTA RICA SCOPING, IMPACT MODELING AND FARM LEVEL ASSESSMENT ................... 81 Introduction ................................ ................................ ................................ ............. 81 Objectives ................................ ................................ ................................ ......... 81 The Fresh Pineapple System in Costa Rica ................................ ..................... 82 Methods ................................ ................................ ................................ .................. 84 System Boundaries and Functional Units ................................ ......................... 84 Data Collection ................................ ................................ ................................ 85 Emissions and Impact Models ................................ ................................ .......... 86 Estimating the Sector Range of Environmental Performance ........................... 90 LCIA Indicators ................................ ................................ ................................ 93 Soil erosion impact ................................ ................................ ..................... 93 Cumulative e nergy d emand ................................ ................................ ....... 94 Virtual water content and stress weighted water footprint .......................... 94 Aquatic eutrophication ................................ ................................ ............... 96 Human and freshwater ecotoxicity ................................ ............................. 97 Other indicators ................................ ................................ .......................... 98 Results ................................ ................................ ................................ .................... 99 Pineapple Sector Inventory ................................ ................................ .............. 99 Soil Erosion ................................ ................................ ................................ ...... 99 Cumulative E nergy Demand (CED) of Pineapple ................................ ........... 100 Carbon Footprint ................................ ................................ ............................ 101 Virtual Water Content and Stress Weighted Footprint ................................ .... 103 Aquatic Eutrophication ................................ ................................ ................... 105 Human and Ecological Toxicity ................................ ................................ ...... 107 Results Su mmary ................................ ................................ ........................... 109 Discussion ................................ ................................ ................................ ............ 109 The Significance of Regionalized Emissions and Impact Models ................... 111 Estimated Environmental Impacts ................................ ................................ .. 112 Potential Impacts Not Measured ................................ ................................ .... 113 Conclusions and Recommendations for Farm Level LCA of Fruit Products ......... 114

PAGE 7

7 5 SUMMARY AND SYNTHESIS ................................ ................................ .............. 118 Summary ................................ ................................ ................................ .............. 118 Chapte r 2 Summary ................................ ................................ ....................... 118 Chapter 3 Summary ................................ ................................ ....................... 119 Chapter 4 Summary ................................ ................................ ....................... 120 Synthesis ................................ ................................ ................................ .............. 122 APPENDIX A SUPPLEMENT TO CHAPTER 2: PROCESS TREE AND UNCERTAINTY ESTIMATES ................................ ................................ ................................ ......... 127 B SUPPLEMENT TO CHAPTER 2: LIFE CYCLE INVENTORY OF GOL D MINED AT YANACOCHA ................................ ................................ ................................ 130 C SUPPLEMENT TO CHAPTER 3: R CODE FOR STOCHASTIC UNCERTAINTY MODELS ................................ ................................ ................................ ............... 159 D SUPPLEMENT TO CHAPTER 4: ADDITIONAL TABLES AND FIGURES ........... 167 LIST OF REFERENCES ................................ ................................ ............................. 183 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 200

PAGE 8

8 LIST OF TABLES Table page 2 1 Summary of emergy in mine products based on two allocation rules ................. 44 3 1 Elements of uncertainty i n the UEV of lead in the g round ................................ .. 60 3 2 Unit emergy value models used for parameter uncertainty calculations ............. 67 3 3 Analytical uncertainty estimation for lead UE V, in ground ................................ .. 71 3 4 Emergy summary with uncertainty of 1 kg of sulfuric acid ................................ .. 72 3 5 UEV uncertainty estimated from the analytical solu tion ................................ ...... 79 3 6 UEV Monte Carlo results and comparison of model CI's with lognormal, hybrid, and normal confidence intervals ................................ ............................. 79 4 1 Summar y table for impacts of 1 kg pineapple delivered to packing facility ....... 117 A 1 Uncertainty estimates for UEVs for inputs into gold silver bullion production ... 128 A 2 Estimation of total uncertainty in gold in the ground ................................ ......... 128 A 3 Estimation of total unce rtainty of silver in the ground ................................ ....... 129 B 1 ................................ ........................... 134 B 2 Inputs to process 'Exploration, at Yanacocha' ................................ .................. 135 B 3 Inputs to pr ocess 'Mine infrastructure, Yanacocha' ................................ .......... 136 B 4 Inputs to process 'Extraction, Yanacocha' ................................ ........................ 136 B 5 Inputs to process 'Leaching, Yanacoch a' ................................ ......................... 138 B 6 Inputs to process 'Leach Pad, Yanacocha' ................................ ....................... 138 B 7 Inputs to process 'Leach Pool, Yanacocha' ................................ ...................... 138 B 8 Inputs to process 'Processing Yanacocha ................................ ....................... 139 B 9 ................................ ............. 140 B 10 Inputs to process 'Conventional Process Water Treatment, Yanacocha' ......... 140 B 11 Inputs to process 'Reverse Osmosis Process Water Treatment, Yanacocha' .. 141

PAGE 9

9 B 12 Inputs to process 'Ac id Water Treatment, Yanacocha' ................................ ..... 141 B 13 Inputs to process 'Reclamation, Yanacocha' ................................ .................... 142 B 14 Inputs for process 'Sediment and dust control, Yanacocha' ............................. 143 B 15 Comparison of this inventory with the equivalent Ecoinvent process ............... 146 B 16 ............................ 147 B 17 Mine hauling road parameters, based on Hartman ................................ ........... 149 B 20 Mine vehicle data ................................ ................................ .............................. 150 B 21 Mass balance of leaching, processing, and water treatment ............................ 151 B 22 Inventory of Peruvian road tra nsport. ................................ ............................... 154 B 23 Assumed origins and transport distances for inputs to mining .......................... 156 B 24 System level parameters ................................ ................................ .................. 157 B 25 Uncertainty estimates for inventory data using Ecoinvent method ................... 158 D 1 Inputs to one kg pineapple at the packing facility. ................................ ............ 167 D 2 Emissions from one kg pi neapple at the packing facility ................................ ... 168 D 4 Emissions estimations for mineral P in applied fertilizers ................................ 170 D 5 General assumptions used in the FAO CROPWAT model ............................... 170 D 6 Crop water requirement variables for CROPWAT ................................ ............ 170 D 7 RUSLE2 parameters for Pineapple in Costa Rica ................................ ............ 171 D 8 Parameters modified for USETox CR model ................................ .................... 173 D 9 Sensitivity a nalysis of the RUSLE2 model customized for pineapple in CR. .... 175 D 10 Sensitivity analysis of the FAO CROPWAT model to variables found in pineapple cultivation. ................................ ................................ ........................ 176 D 11 Sensitivity analysis of PestLCI model for pineapple conditions ........................ 176 D 12 Recalculation of Pimentel (2009) energy demand for US oranges ................... 177 D 13 Recalculation of Pimentel (2009) energy demand for US apples ..................... 177 D 14 Recalculation of Coltro (2009) energy demand for BR oranges ...................... 178

PAGE 10

10 D 15 CED values for inputs used in recalculations of Orange BR, Orange US and Apples US ................................ ................................ ................................ ......... 178

PAGE 11

11 LIST OF FIGURES Figure page 2 1 Proposed boundary expansion of LCA with emergy ................................ ........... 32 2 2 Gold production system at Yanacocha with modeled flows and unit processes ................................ ................................ ................................ ........... 37 2 3 Environmental contribution (emergy) to dor by input type ................................ 45 2 4 Emergy and primary energy in 1 g of dor by unit process ................................ 45 2 5 Monte Carlo analysis of 1 g of dor. ................................ ................................ ... 48 3 1 Conceptual approach to modeling uncertainty ................................ .................... 66 3 2 Published UEVs for elect ricity by source from Brown and Ulgiati (2002), superimposed upon a modeled range of the oil UEV ................................ ......... 80 4 1 Fresh pineapple production unit processes and boundaries for the LCA ........... 84 4 2 Contribution to CED of pineapple, at packing facility ................................ ........ 102 4 3 Non renewable CED of one serving pineapple in comparison with evaluations of the farmin g stage of oth er fruits ................................ ................. 102 4 4 Contribution to carbon footprint of pineapple, at packing facility ....................... 103 4 5 Carbon footprint of one ser ving pineapple in comparison with evaluations of the farming stage of other fruits ................................ ................................ ........ 104 4 6 Virtual water content (VWC) for pineapple in comparison with other fruits ....... 105 4 7 Contribution to potential eutrophication of pineapple by emission .................... 106 4 8 Preliminary comparison of potential eutrophication effects of one serving pineappl e in comparison with evaluations of the farming stage of other fruits. 106 4 9 Relative contribution of active ingredients of pesticides used in pineapple production to human toxicity and freshwa ter ecotoxicity ................................ ... 108 A 1 SimaPro process tree of environmental contribution (sej) to 1 g dor .............. 12 7 B 1 Process overview ................................ ................................ ............................. 132 D 1 Emission fractions of applied pesticides in Pes tLCI CR vs. the PestLCI default ................................ ................................ ................................ ............... 179

PAGE 12

12 D 2 Freshwater ecotoxicity characterization factors for pes ticides in USETox CR vs USETox Default ................................ ................................ ........................... 180 D 3 Human toxicity characterization factors for pesticides in USETox CR vs USETox Default ................................ ................................ ................................ 181 D 4 Human toxicity and freshwater ecotoxicity for pesticide emissions from pineapple prod uction in the baseline scenario ................................ .................. 182

PAGE 13

13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in P artial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ADVANCES IN LIFE CYCLE ASSESSMENT AND EMERGY EVALUATION WITH CASE STUDIES IN GOLD MINING AND PINEAPPLE PRODUCTION By Wesley W. Ingwersen August 2010 Chair: Ma rk T. Brown Major: Environmental Engineering Sciences Life cycle assessment (LCA) is an internationally standardized framework for assessing the environmental impacts of products that is rapidly evolving to improve understanding and quantification of how complex product systems depend upon and affect the environment. This dissertation contributes to that evolution through the development of new methods for measuring impacts, estimating the uncertainty of impacts, and measuring ranges of environmental perf ormance, with a focus on product systems in non OECD countries that have not been well characterized. The integration of a measure of total energy use, emergy, is demonstrated in an LCA of gold from the Yanacocha mine in Peru in the second chapter. A mod el for estimating the accuracy of emergy results is proposed in the following chapter. The fourth chapter presents a template for LCA based quantification of the range of environmental performance for tropical agricultural products using the example of fr esh pineapple production for export in Costa Rica that can be used to create product labels with environmental information. The final chapter synthesizes how each methodological contribution will together improve the science of measuring product environme ntal performance.

PAGE 14

14 1 CHAPTER 1 INTRODUCTION Production of goods and services is inextricably tied to the environment. As basic resources for modern economies are becoming more costly or less available (e.g., freshwater and petroleum) and impacts of p roductive activities have created local and global scale environmental change (e.g., climate change), the need to understand connections between the environment and economy has become more critical. The delegates to the UN Conference on Environment and De velopment, representing over a Environment and Development, or Agenda 21, that all productive processes in economies are dependent upon sources of energy and materials from the en vironment and sinks to absorb the pollution that they generate (principle 8, UN 1992) At the World Summit on Sustainable Development a decade later, it was furthered acknowledged that measurement systems are necessary to quantify these dependencies and pollution impacts for the purposes of achieving more sustainable development (Chapter 3, UN 2005) Measurement of Su stainable Production and Consumption Measurement is the first step toward effective management and protection of the environment in the context of productive processes in economies. But the concept of measurement of environmental impacts of production pr ocesses has been evolving with broader understandings of what, how and where impacts occur and who in turn is responsible for those impacts. The first generation of environmental policy in the United States (such as the Clean Air Act of 1970), and still t he dominant form of regulation in

PAGE 15

15 implicitly focusing only on pollution at the point of occurrence and obligating only the party responsible at that point to reduc e or cease the pollution. This style of legislation reflects the assumption that impacts should be measured only at the point of impact. But the ultimate purpose and driver of a production processes is to provide for an end product or service, and thus th e impacts of productive processes can all be related to the intermediate or end products. That product or service is demanded by a consumer, and that consumer shares responsibility for the environmental impacts that occur along the production chain. Shar ed producer and consumer responsibility was recognized in the Rio Declaration and reinforced in international action plans such as the Marrakesh Process launched at the World Summit on Sustainable Development (UN DESA 2008) and is now becoming further integrated at national, regional, and local scales, especially through voluntary public and private initiatives (e.g. Environmenta l Management Systems, Extended Producer Responsibility policies, corporate greenhouse gas accounting standards). It then becomes clear that measurement tools are needed that relate these broader impacts to products or services in a way that accounts for i mpacts along the full production chain such that management can involve both producer and consumer, and so that no impacts associated with production processes are left out. Life Cycle Assessment as a Measurement Tool Life Cycle Assessment (LCA) is an esta blished and standardized framework for assessing impacts of production processes and for relating full life cycle impacts to a final product (ISO 2006c) LCA is being used globally for product systems for purposes of design, management, and communication of environmental performance (UNEP 2007) as wel l as to guide environmental product policy (European Commission 2003)

PAGE 16

16 LCA is an appropriate frame work for measuring impacts of products because it uses a full life cycle perspective to thus including all product stages during which significant impacts might occur, including all production and consumption stages. This begins wi th assessing the goal and scope of a product system and continues with an inventory of inputs and emissions by product stage relevant to estimation of impacts. Estimating the impacts of these emissions is done with impact characterization factors develope d from impact models. Impacts are all related to a unit of the product serving a particular functional purpose, called a functional unit. These impacts typically measure use of environmental sources (resource use indicators) or stressors on environmental sinks (impact indicators). Impact indicators depict impacts at varying points in the chain of causality from the release of an emission to its ultimate impact (end point) on primary areas of concern (human health, natural environment, resources, manmade environment), depending upon the state of the science for modeling impacts along this chain (Bare et al. 2006) LCA is arguably the strongest framework for measuring environmental impacts of production activities for the complex, global supply chains typical of modern products. Ness and colleagues (2007) categorized measures of sustainability based on their focus and their temporal aspects. In contrast with techniques such as environmental impacts assessment, which is focused on future activity and is highly location specific, LCA is primarily focused on current systems (though can be used for design purposes) and is not limited in focus to one particular site. In contrast with sustainability indices (e.g. environmental pressure indicators), which are often retrospective indicators o f larger systems, LCA is more product specific. LCA also originates from industrial ecology and

PAGE 17

17 engineering, and its quantification by particular unit processes make LCA results more relevant for product management. In comparison with other systems orien ted approaches such as embodied energy or emergy analysis, LCA is multi criteria, which provides a broader view of products and makes it less likely that important impacts are overlooked (Ulgiati et al. 2006) R esearch Problems in Life Cycle Assessment The effectiveness of the bold intention to use LCA to relate a product to all the damages (or benefits) that occur to the environment over the life cycle of its production, use, and disposal depends upon detailed i nventories of complex product life cycles as well as accurate models to estimate environmental damages related to resource used or emissions that occur with these life cycles. LCA adapts scientific theory and models from many other fields to accurately id entify and model impacts and thus is only as advanced as the science and its application within this fields. LCAs are often limited by incomplete or inappropriate data and absence of relevant impact models. Two focal areas of LCA that specifically need t o be addressed to better measure sustainable production and consumption in a manner applicable to global supply chains are 1) resource use indicators and 2) impact models for processes occurring in non OECD product systems. These problems and a proposed p lan for addressing them are described in the following three sections. Life Cycle Impact Assessment (LCIA) Indicators for Resource Use As described above, indicators in LCA may be broadly split into resource use and impact indicators. Resource use indicat ors may be based on the use of a particular energy source or material (e.g. fossil energy use or freshwater use) or may be an aggregate measure. Furthermore, they may focus on relating that use to ultimate

PAGE 18

18 availability (e.g. mineral resource depletion) or simply just report usage. Relating different indicators of resource use together may require use of subjective weighting criteria when there is not a physical basis for relating the resources (Guine 2002) But the impact of using different resources may be related together without the need for subj ective judgment if resources can be characterized on a common physical basis with a common unit, which is instructive for synthesizing the effects of resource use. Various authors have argued for the need to incorporate a unified measure of resource use into LCA to limit resource consumption associated with productive processes (Finnveden 2005; Seager and Theis 2002; Stewart and Weidema 2005) Single unit measures of resource use have been extensively developed outside of the LCA framework, but not all of these methods have been applied as indicators in LCA. These methods typically aggregate resource use using a common biophysical unit. Common biophysical units may be units of mass, land area, or energy. Life cycle based methods using mass include extensions of material flow analysis (MFA) and closely related methods including ecological rucksack and material inputs per unit service (MIPS) (Brunner and Rechburger 2003; Schmidt Bleek 1994) In short, these met hods associate a material intensity (g material/g product) with all inputs to a product over the production cycle. They have been applied predominantly in studies of dematerialization of economies (Bartelmus 2003; Matthews et al. 2000; NAS 1999) and have not been formally integrated as an impact method in life cycle assessment. The major weakness of using MFA derived units of mass as a common resource use indicator for a product is the absence of differentiation o f the quality of different resource types, as well as the difference in the use of materials that may only temporarily

PAGE 19

19 sequester them (e.g. cooling water) or may completely transform them rendering them useless for future production processes (e.g. combust ed fuels) (Van Der Voet et al. 2004) Area based measures of resource use either measure solely direct and indirect occupation and transformation of land or go further by using equivalence factors to relate different land use types and symbolic land uses together to measure a broader concept of land requirements (e.g. ecological footprint). Measures of occupation and transformation of land use are commonly empl oyed in LCA (Guine 2002) A measure that comb ines all types of land use in a single unit based on their biological capacity is the ecological footprint (Wackernagel et al. 2002) Ecological footprint has been more recently integrated as a resource use measure in the largest commercial LCA database (Frischknech t and Jungbluth 2007) Indicators of land occupation suffer from numerous shortcomings. Neither direct land use nor the ecological footprint measure below ground resource use (non renewable), and neither incorporate the use of hydrologic resources. Furt hermore, land itself it not expected to become a limiting resource in the future. Although the ecological footprint already shows that total direct and indirect use defici t (Hails et al. 2008) Energy based measures are potentially more comprehensive in their inclusion of resources th an land based and material based measures. Energy based measures are derived from the laws of thermodynamics, the first of which states that energy is consumed in every transformation process. Thus every process, both independent of and dependent on huma ns, involves the consumption of energy, which makes energy

PAGE 20

20 an ideal common unit for tracking total resource use (Odum 2007) Some energy based resource use measure s have already been incorporated into LCA. Energy analysis (Boustead and Hancock 1978) known as cumulative energy demand (CED) analysis implemented in a life cycle framework (Frischknecht and Jungbluth 2007) measures the total heat energ y (enthalpy) in fuel and other energy carrier consumed based on their heating values. CED does not include the contribution of non energy sources. Surplus energy, part of the Eco indicator 99 methodology (Goedkoop and Spriensma 2001) estimates the difference in the amount of energy required to extract resources now versus at a designated point in the future. Surplus energy is also limited to energy sources. Another thermodynamically based indicator already integrated into LCA that includes a broader array of resource is exergy which may be defined as the total of available energies of different types in a material (primarily as pressure, kinetic, physical chemical) in respect to their difference from reference conditions. Raw resources have high exergy values until processed or transformed at which time some of their exergy is lost as entropy. The exergy losses associated with transformations of all inp uts into processes in an LCA can be measured with cumulative exergy demand, or CExD (Bsch et al. 2007) CExD is particularly valuable as a measure of the total thermodynamic efficiency of a process where the goal is to minimize total exergy consumption. None of the aforementioned energy b ased methods account for the energy required by the environment to support and recreate the resource basis of economies; they only account for energy consumed in existing resources. Thus a critical first link in

PAGE 21

21 the chain of resource provision (environment to resource) is missing in how resource use is accounting for in product life cycles. Accounting for this first link, however, is possible using the emergy method to relate all resources on the basis of sunlight energy. Emergy is an energy accounting me tric that may be defined as the total direct and indirect energy used to support a system measured in a common unit of energy conventionally sunlight equivalents (Odum 1996) The origins of all resources, both renewable and non ren ewable, can all be directly or indirectly traced back to the primary energy driving the biosphere, sunlight, and can thus be tracked in units of energy of this type. Thus it becomes a biophysically legitimate way of combining different forms of resources in a common measurement unit. Emergy evaluation is an independently developed methodology for measuring the environmental performance of an ecosystem or human dominated system, which has also been applied to evaluating product systems. Emergy has been use d in conjunction with LCA as part of a comparative or multi criteria approach (Cherubini et al. 2008; Pizzigallo et al. 2008) Emergy has been adapted for use in economic based input output LCA by Bhakshi and coll eagues, who define emergy as an extension of exergy called ecological exergy (Hau and Bakshi 2004a) and have used it as a measure of the contribution of ecosystem processes to sectors of the US economy (Ukidwe and Bakshi 2004) and to evaluate individual products (B aral and Bakshi 2010) Nevertheless emergy has not been integrated into traditional process LCA in such a manner that it can be used in conjunction with conventional life cycle inventory databases and in comparison with other LCA metrics.

PAGE 22

22 A measure of the ultimate limitations that the biosphere imposes upon economic processes must relate these processes to the energetic limits of the biosphere (Odum 2007) While such a broad concept may not highlight the scarcity of particular resources, it does provide a sufficiently wide context through which to compare any and all products with our planetary resource base; in doing so it can provide insight into absolute susta inability of economic processes in the long term. Emergy (in sunlight energy equivalents) can be used to measure contribution of all forms of resources and environmental processes to a product and report them with a common unit relates each resource back to the energy consumed in its origin, and as such is an optimal numeraire for measuring total resource use per unit of the product. Further clarifying the rationale for integrating emergy into LCA a measure of total resource use and demonstrating the mea ns of integrating emergy into a complex process based LCA typical of high volume products is a primary objective of this dissertation. An implicit requirement for integrating emergy or any other impact metric into LCA is to quantify the uncertainty in th e impact model. It has been recognized among the LCA community that the data and models used to represent complex product life cycles potentially have a significant amount of variation and uncertainty (Fava et al. 1994) Reporting average scores for products can at times be misleading to the degree of accur acy occurring. Better estimation of uncertainty in these scores is a current priority in the LCA field (Reap et al. 2008) Uncertainty characterization should i nclude uncertainty in model parameters, uncertainty to represent variation among different geographic, technological, or alternative production scenarios that may be unknown, and uncertainty built into the

PAGE 23

23 actual impact models themselves (Lloyd and Ries 2007) When emergy is incorporated into LCA as an impact model, this should therefore include the additional model uncertainty that is added when unit emergy values (UEVs) are used to relate inputs to processes to the emergy that was used to make them. In the practice of emergy evaluation, emergy results are not typically presented with uncertaint y ranges. The originator of the emergy concept, H.T. Odum, believed that an emergy result was accurate within an order of magnitude (Brown 2009) The lack of a more cle arly defined and systematic manner of characterizing the accuracy of emergy results has been a criticism of emergy work for decades (Rydburg 2010) A couple notable first attempts at characterizing uncertainty in specific UEVs wer e performed by Campbell (2001) and Cohen (2001) Campbell estimated the uncertainty in the transformity of global rainfall and river chemical potential based on differences in estimated global water flows. Co hen used a stochastic simulation technique to generate confidence envelopes for UEVs of various soil parameters. Both of these approaches were first order attempts for estimating ranges of specific emergy values, but did not fully characterize this uncert ainty or propose methods of propagating this uncertainty for use in future evaluations. A model for estimating uncertainty in emergy results would be useful for estimating ranges in emergy results within emergy and beyond for the estimation of the additio nal uncertainty related to emergy models in life cycle results that use emergy as a unit of measurement. Applications of LCA for N on OECD C ountry P roducts LCA studies have predominantly been conducted on product systems located in the United States, EU co untries, Canada, Japan, and Australia and other member of the Organization for Economic and Co operation and Development (OECD) (Thiesen et al.

PAGE 24

24 2007) As a result there has been a geographic bias in th e development of all aspects of LCA, including product system inventories, selection of impact categories, and LCA impact models. This bias has resulted in two primary deficiencies in LCA: (1) production in non OECD countries is less well characterized r esulting in lesser capacity to use life cycle management; and (2) comparisons with OECD products has been hindered thus limiting ability to use LCA in OECD countries that consume products from all over the world. Unless this gap in life cycle management c apacity is closed, increasing environmental demands on producers could marginalize non OECD country producers with lesser capacity (Sonnemann and de Leeuw 2006) Witho ut improved life cycle management, the consumer demand for increasing non OECD country products may increase environmental damage in non OECD countries. Expanding the scope of LCA to incorporate more global analysis including for products from non OECD co untries is a priority in the current phase of the UNEP SETAC Life Cycle Initiative (UNEP Life Cycle Initiative 2007) Export of products to OECD countries plays a significant role in the economy of many non OECD countries. For those in Latin America and Africa, these exports are largely from the primary sectors, which include fuels, agricultural products, and minerals (Zhang et al. 2010) Mineral and agricultural sectors are both responsible for many direct environmental impacts that are site specific, because they gen erally require significant transformation of the land and emissions occur often in a diffuse manner into the environment surrounding the site. As a result, both mineral and agricultural environmental impacts are less easily characterized than impacts from more enclosed

PAGE 25

25 processes with less direct interaction with the local environment (more concentrated and controlled emissions). Accurate characterization of diffuse emissions and their impacts in mining and agriculture depends on models that account for the local environmental factors that influence emissions and their potency at production sites (spatial and temporal specificity). There have been calls for greater regionalization of impact methods in both the mining (Yellishetty et al. 2009) and agricultural sectors (Gaillard and Nemecek 2009) In mining systems, this may include the geologic work required to create a particular deposit, if the boundary of resource use is extended to include all environmental resources as suggested i n the previous section. In agricultural systems, regional factors effect emissions and their impacts. This is particulary relevant for emissions such as fertilizers and pesticides and the impacts they can cause including eutrophication and ecological a nd human toxicity. Local factors also effect emissions that have just recently begun to be characterized in LCA, including water loss (Pfister et al. 2009) Regional characterization of models based on geographic difference can have dramatic effects on LCA outcomes (Lenzen and Wachsmann 2004) Not all relevant environmenta l impacts from agricultural systems have been characterized in LCA. Two that the UNEP taskforce has identified as extremely relevant, particularly in non OECD countries, are biodiversity impacts and soil erosion (Jolliet et al. 2003b) Models to estimate impacts from biodiversity are very much in their infancy, while some have been proposed (e.g. Maia de Souza et al. 2009; Schenck and Vickerman 2001) Erosion is the most significant cause of land degradation globally (Gobin et al. 2003) Soil erosion has not frequently been characterized in LCA,

PAGE 26

26 but universal methods for estimating soil erosion based on geographic, climatic, soil and managem ent factors do exist. The most commonly applied measure of soil erosion is probably the Universal Soil Loss Equation (USLE) and its more recent developments, the Revised Universal Soil Loss Equation (RUSLE) and most recently, RUSLE2 (Fost er et al. 2008) Soil erosion has rarely been used in LCA, and has not been customized for use in LCA of non OECD countries, many of which have humid tropical environments, where heavy rain based erosion risks can be much greater (Lal 1983) Without a strong demand on the part of buyers or regulation imposed by governments, there is not a strong incentive to use LCA in non OECD countries (Sonnemann and de Leeuw 2006) However, because of the emerging life cycle perspective in countries where non OECD exports are consumed, many of which are OECD countries, the demand for use of LCA to measure environmental perfor mance may come from the consumers. Yet, there needs to be a standardized mechanism through which the LCA results can be conveyed to the consumers in a way that they can use this information to inform decision making. One solution is to present this LCA b ased environmental performance information in the form of a product label. A Type III environmental label or environmental product declaration (EPD), as defined by ISO 14025, is designed for this purpose (ISO 2006b) EPDs are designed to convey information on product function and production of the product, and relate this information to environmental performance in a manner that one product can be compared with another product in the same category. Product category rules (PCRs) have to be specified so that results presented in EPDs are comparable. The ISO 14025 standard recommends that PCRs be based on at least one background assessment of

PAGE 27

27 a product, so that the product life cycle can be characterized and relevant impacts determined. This aspect of PCRs present a challenge for product systems in developin g countries, because often little life cycle data and or LCA analysis of these systems exist. Another potential barrier to use of EPDs that applies not only to non OECD countries was identified by Christiansen et al. (2006) and is related to the interpretation of EPDs. These authors note that LCA data presented in EPDs are often not readily meaningful without reference to the relative performance of other products in the category. This shortcoming of EPDs is another important issue to address to make LCA more relevant for non OECD product systems. Research Overview Three independent studies addressing the research problems described comprise this dissertation. The first study proposes a means to integrate emergy as a life cycle assessment indicator to provide a measure of long term sustainability in LCA. This study uses the case of the Yanacocha gold mine in northern Peru. A detailed process based life cycle assessment is carried out to track the emergy in all direct and indirect inputs to the mining process, including in the ore itself. Methods of associating emergy values with inventory data and ca lculating results with emergy in LCA are described. Comparisons of emergy results are made with a commonly used measure of life cycle energy requirement, or cumulative energy demand. Following presentation of these results, their potential value in the r egional context and the broader value of emergy results for LCA are discussed, along with remaining questions and problems with this integration. The problem of statistically describing the confidence of emergy results leads directly into the research ne eds addressed in the second study: estimating the

PAGE 28

28 uncertainty of emergy values. In this study, sources of uncertainty in emergy are explored and the likely forms of probability distributions of different types of emergy calculations are suggested. The de scription of the sources and forms of uncertainty lead to the proposal for a model for describing uncertainty in emergy, and two alternative procedures for estimating confidence intervals of emergy values are described. This study proceeds with an evaluat ion of the accuracy of the proposed model and proposes a means of integrating confidence intervals into the tables commonly used to present emergy results. The third study shifts to addressing the problems associated with broad characterization and applica tion of life cycle assessment for poorly characterized or data poor product categories in regions where existing emissions and impacts models are not appropriate because of differences in environmental conditions. A multi criteria process based LCA is con ducted of fresh pineapple for export in Costa Rica (not previously characterized with LCA), based on data from a representative sample of pineapple producers. Existing universally applicable emissions and inventory models are customized to better characte rize environmental impacts. An original method for characterizing soil erosion is integrating using the RUSLE2 model. Variation and uncertainty in inputs and emissions among the participating producers are used to estimate the range of environmental perfo rmance in the sector for each impact category. This LCA is furthermore designed to contribute to creating the rules for an environmental product declaration in a manner applicable for yet uncharacterized product categories.

PAGE 29

29 2 CHAPTER 2 EMERGY AS AN IMP ACT ASSESSMENT METHO D FOR LIFE CYCLE ASS ESSMENT PRESENTED IN A GOLD MINING CASE STUDY Introduction LCA is an established and widely utilized approach to evaluating environmental burdens associated with production activities. Emergy synthesis has been used for similar ends, although in an emergy synthesis one tracks a single, all encompassing environmental aspect, a measure of embodied energy (Odum 1996). While each is a developed methodology of environmental accounting, they are not mutually exclusive. Emer gy in the LCA Context LCA is a flexible framework that continues to grow to integrate new and revised indicators of impact, as determined by their relevance to the LCA purpose and the scientific validity of the indicator sets (ISO 2006c) Other thermodynamically based methods, such as exergy, have been integrated into LCA (Ayres et al. 1998; Bsch et al. 2007) Emergy synthesis offers original information about the relationship between a product or process and the environmen t, not captured by existing LCA indicators, particularly relevant to resource use and long term sustainability, which could be valuable for LCA. However there are differences in the conventions, systems boundaries and allocation rules between emergy and LC A which require adjustments from the conventional application of emergy to achieve a consistent integration From the perspective of the LCA practitioner, the first questions regarding use of emergy would be those of its utility. Why would one select eme rgy, in lieu of or in addition to other indicators of environmental impact? For what purposes defined for an LCA study would emergy be an appropriate metric? A ssuming the inclusion of emergy as an indicator, what would be necessary for its integration into the LCA framework?

PAGE 30

30 This paper briefly describes the utility of emergy, and through a case study evaluation of a gold mining operation at Yanacocha, Peru, p resents one example of how e mergy can be used in an LCA framework. Finally, the theoretical and tec hnical challenges posed by integration are discussed. In reference to the first question, these four key points provide a theoretical justification for the use of emergy in LCA: 1 Emergy offers the most extensive measure of energy requirements System bounda ries in a cradle to gate LCA typically begin with an initial unit process in which a raw material is acquired (e.g. extraction), and would include raw materials entering into that process, but would not include any information on the environmental processe s 1 creating those raw materials. Emergy traces energy inputs back further into the life cycle than any other thermodynamic method, summing life cycle energy inputs using the common denominator of the solar energy directly and indirectly driving all biosphe re processes (Figure 1) 2 Other thermodynamic methods including exergy do not include energy requirements underlying environmental proc esses (Ukidwe and Bakshi 2004) 2 Emergy approximates the work of the environment to replace what is used When a resource is consumed in a production process, more energy is required to 1 geologic, and hydrologic flows that sustain both ecosystems and human dominated systems. This is the essence of what is meant here b 2 For example, growing corn requires the solar energy necessary to support photosynthesis of the corn plant. This includes all the solar energy falling on the corn field, not just the amount the corn used to fix CO2. Further more growing corn requires fossil inputs among others, all of which were originally created with solar energy, and thus which are included in emergy analysis.

PAGE 31

31 regenerate or replenish that resource. The emergy of a resource is this energy required to make it including work of the environment, and assuming equivalent conditio ns; this is the energy that is takes to replenish it. Sustainability ultimately requires that inputs and outputs to the biosphere or its subsystems balance out (Gallopin 2003) As the only measure that relates products to energy inputs into the biosphere required to create them, emergy relates consumption to ultimate limits in the biosphere by quantifying the additional work it would require from nature to replace the consumed resources. 3 Emergy presents a unified measure of resource use Comparing the impacts of use of biotic vs. abiotic resources, or renewable vs. non renewable resources, typically necessitates some sort of weighting scheme for comparison. 3 Because there is less agreement upon characterization of biotic resources, these may not be included despite their potential relevance (Guine 2002) Using emergy, abiotic and biotic resources are both included and measu red with the same units. As follows from its nature as a unified indicator, one which characterizes inputs with a single methodology to relate them with one unit (emergy uses sejs, or solar emjoules, which are sunlight equivalent joules), no weighting sch eme is necessary to join different forms of resources (e.g. renewable and non renewable; fuels and minerals) to interpret the results. The choice of measures of impact in an LCA follow from the goal an d scope of the study (ISO 2006c) Emergy analyses have been used for a multitude of LCA related 3 In the IMPACT 2002+, and Eco indicator 99 methodologies, use of non renewable resources is included in the damage categories of resources but renewable resources are omitted (Goedkoop and Spriensma 2001; Jolliet et al. 2003a)

PAGE 32

32 purposes, including to measure cumulative energy consumption (Federici et al. 2008) to compare environmental performance of process alternatives (La Rosa et al. 2008) to create indices for measuring sustainability (Brown and Ulgiati 1997) to quantify the resource base of ecosystems (Tilley 2003) to measure environmental carrying capacity (Cuadra and Bjrklund 2007) and for non market based valuation (Odum and Odum 2000). The incorporation of emergy in LCA could potentially enhance the ability of LCA studies to achieve these same and other purposes. Figure 2 1 Proposed boundary expansion of LCA with emergy. Driving ener gies include sunlight, rain, wind, deep heat, tidal flow, etc. This was not the first study to attempt to combine emergy and life cycle assessment. Earlier studies focused on contrasting the two approaches (Pizzigallo et al. 2008) or extending emergy to include disposal and recycling processes (Brown and Buranakarn 2003) The most comprehensive approaches probably include the Eco LCA and S UMMA models. Although referred to as ecological cumulative exergy consumption (ECEC) rather than emergy due some slight modifications to emergy algebra, the Eco LCA model is an EIO LCA model which uses emergy as an impact indicator (Urban and Bakshi 2009) The SUMMA model is a multi criterion analysis tool

PAGE 33

33 measures of downstream impa ct (Ulgiati et al. 2006) A similar multi criteria approach using MFA, embodied energy, exergy and emergy is used by Cherubini et al. (Cherubini et al. 2008) In contrast with these previous studies, this study use s a more conventional process LCA approach through using an common industry software (SimaPro) and attempts to integrate emergy as an indicator within that framework as specified by the ISO 14040/44 standards, which results in adjustments to the conventio nal emergy methodology. This is also the first study to use emergy in a detailed process LCA where flows are tracked at a unit process level. Results from the study, addressed in the discussion, reveal insights for which emergy is suggested to be a usefu l metric for LCA. A C ase S tudy of E mergy in an LCA of G old S ilver B ullion Production Metals and their related mining and metallurgical processes have been a frequent subject of LCA and other studies using approaches from industrial ecology (e.g. Yellishett y et al. 2009 and Dubriel 2005), which is reflective of the critical dependence of society upon metals, as well as an acknowledgement of the potential environmental consequences of their life cycles. While these studies have addressed both downstream and upstream impacts, including resource consumption, none have used tools capable of connecting the product system to the environmental processes providing for the raw resources they require (especially because they are largely nonrenewable). An LCA is presen ted here of a gold silver mining operation that uses emergy to quantify the dependence on environmental flows. In th is case study, the primary purpose could be succinctly stated as follows:

PAGE 34

34 To quantify the total environmental contribution underlying produ ction of gold silver bullion at the Yanacocha mine in Peru. 4 Total environmental contribution includes the total work required by the environment (biosphere) and the human dominated systems it supports (technosphere) to provide for that product. As impact s in LCA are categorized as resource related (referring to upstream impacts) or pollution related (referring to downstream impacts) (Bare et al. 2003) environmental contributi on would be categorized with the former. The scope of this study, following from this goal, extends from the formation of the gold deposit (representing the work of the environment) to the production of the semi refined dor, a bar of mixed gold and silver 5 Emergy is chosen as the measure of basis of the indicator of impact of mining. Energy is commonly used in LCA to track the total energy supplied to drive process es in an industrial life cycle. Yet the interest here is in how much work was done in both environmental systems and human dominated systems to provide for it (point 2 ), which i s not measured by just considering available energy used by energy carriers (e .g. cumulative energy demand ) or by summing all available energy (exergy) in all the inputs (point 1). Additionally the energy from the environment to provide for non energy resources (materials) is part of the env ironmental contribution (point 2 ), so all need to be tracked. However, in order to directly compare 4 The Yanacocha mine is one of the largest gold mines (in terms of production) in the world. The mine produced 3.3275 million ounces in 2005 (Buenaventura Mining Comp any Inc. 2006) This represented more than 40% of Peruvian production (Peruvian Ministry of Energy and Mines 2006) and approximately d from dor and using the total of 2467 tonnes reported by the World Gold Council (2006) 5 The system and inventory are described in detail in the appendix Life Cycle Inventory of Gold Mined at Yanacocha, Peru

PAGE 35

35 the environmental contribution underlying each resource input t ogether with the others contributing to a unit process of mining operation, the contribution should be tracked with a single indicator for which emergy serves as this indicator here (point 3 ). Using emergy allows for the introduction of more specific questions which, when used in an LCA context, are answerable where they are traditionally not in an emergy evaluation, which lumps all inp uts into a single system process The ability to track unit processes from the biosphere together with unit processes in the technosphere enables one to ask: Is there more environmental contribution underlying the formation of the gold or the combined min ing processes? as well the more familiar (to LCA) comparison s of inputs and unit processes in the product system: Which unit process(es) are the most intensive in terms of en vironmental contribution? Which inputs are responsible for this? To address long term sustainability, the activity surrounding this life cycle can be put in context of available resources; more specifically: How does this relate to the availability of energy driving environmental processes in this region? LCA results should be presente d with accompanying uncertainty quantified to the extent feasible (ISO 2006a) To fit in the LCA framework, emergy results also need t o be presented with uncertainty estimations to explain the accuracy with which environmental contribution can be predicted.

PAGE 36

36 Gold and silver are co products, which may be mined separately and which have independent end uses, so comparison of this life cycle data with alternative production routes or for end use requires allocating environmental contribution between them, as well as between mercury, which is naturally associated with the ore body, separated during the refining sta g e and sol d as a by product. This LCA is not comparative, because no other alternative solutions for providi ng the gold are being evaluated. Nevertheless with a universal measure of impact that does not require nor malization or weighting (point 4 ), results can be c ompared with alternative product system s for which emergy evaluation has been done, if the boundaries and allocation rules for these alternative products are comparable, or put in the context of other relevant emergy flows, such as those supporting ecosyst ems or economic systems in the same region. Methodology The functiona l unit chosen for the study is 1 g of dor (gold silver bullion) at the mine gate, consisting of 43.4% gold and 56.6% silver For comparison with other gold, silver, and mercury products results are also reported in relation to 1 g of gold, 1 g of silver, and 1 g of mercury. The inventory for these products was based on the average of annual production in 2005, t he most recent year for which all necessary data were available. Annual pro duction was reported by one of the mine partners (Buenaventura Mining Company Inc. 2006) The total production for this year was approximately 9.40 E+04 6 kg of gold and 1.23 E+05 kg of silver combined as gold silver bullion, or dor. 6

PAGE 37

37 A process based invento ry was completed in accordance with the ISO 14040 series standards (ISO 2006a, 2006b) and included direct inputs from the environment (elementary flows), capital and nondurable goods, fuels, electricity, and transportation, along with inputs not traditiona lly or commonly accounted for, including the geologic contribution to mineral formation. Nine unit processes representing process stages were defined, and inputs were tracked by unit process ( Figure 2 2 ). These we re divided into background processes (deposit formation, exploration, and mine infrastructure), production processes (extraction, leaching, and processing), and auxiliary processes (water treatment, sediment control, and reclamation). A description of the inventory calculations and results is in the supplemental material. Figure 2 2 Gold production system at Yanacocha with modeled flows and unit processes. FF = fossil fuels, HM = heavy machinery, I = infras tructure, C = chemicals, W= precipitation and pumped water, E = electricity, AWR = acid water runoff, PWW = process wastewater.

PAGE 38

38 Emergy and Energy Calculations All inputs were converted into emergy values either via original emergy calculations or by using previously calculated unit emergy values which relate input flows in the inventory to emergy values (Odum 1996) An inventory cutoff for inputs consisting of 99% of the emergy for the process was declared, to be as comprehensive as p ossible without including all minor inputs. As the emergy of some inputs was not readily estimated prior to the inventory collection, these inputs were by default included and, even if determined to contribute less than 1% of the total emergy, were kept i n the inventory. The geologic emergy of gold, silver, and mercury (representing the work of the environment in the placement of mineable deposits) were estimated using the method of Cohen et al. (2008) who proposed a new universal model for estimating emergy in elemental metals in the ground, based on an enrichment ratio of the element, which can be described in the form: UEV i = ER i 1.68E+09 sej/g (1) where UEV is the u nit emergy value (sej/g ) for this element in the ground ER is the enrichment ratio, and i denotes a particular element. The ER can be estimated with the following equation: ER i = OGC i /CC i (2) w here OGC is the ore grade cutoff of element i which is the current minimal mineable concentration and CC is the crustal background concentration of that element. This model assumes that ores with greater concentrations of metals require greater geologic work to form, without attempting to mechanistically mod el the diverse and random geological processes at work, conferring a general advantage of consistent

PAGE 39

39 and comparable emergy estimations for all mined metals. This universal method provides average UEVs for a particular metal in the ground, but was adapted here using the specific concentrations of gold, silver, and mercury at Yanacocha in place of the OGC for those elements. Original emergy calculations were necessary for a number of mining inputs, including mine vehicles, chemicals, mine infrastructure, an d transportation. When available, data on these inputs was adapted from a commercial life cycle i nventory database, Ecoinvent v2.0 (Ecoinvent Centre 2007) and copied into a new process. Inputs for these processes were replaced by processes carrying UEVs calculated from previous ly published emergy analyses. When the processes were adapted from Ecoinvent, emissions, infrastruct ure, and transportation data were not included, the latter of which was decided to be inap propriate for the mine location and calculated independently or estimated to be insignificant. For chemica ls not available in Ecoinvent synthesis processes were base d on stochiometry found in literature references and primary material inputs as well as energy sources were included. Emergy in overseas shipping and transportation within Peru of inputs was estimated for all materials comprising 99% of the total mass of inputs to the process. The global base line (estimate of emergy driving a planet and basis of all emergy estimates) of 15.83 E+2 4 sej/yr was used for all original UEV calculations (Odum et al. 2000) and for updates of all existing UEVs calculated in other studies. When available, existing UEVs were incorpora ted without labor or services, to be consistent with the Ecoinvent data used which do not include labor inputs to processes. For comparison with emergy values, primary energy was estimated by summing the total energy content

PAGE 40

40 of fossil fuels and electricity consumed on site using energy values from the Cumulative Energy Demand characterization method as implemented in SimaPro (Frischknecht and Jungbluth 2007) Uncertainty Modeling Uncertainty was present at the inventory level (e.g. inputs to mining) and for the unit emergy values (the UEVs) used to convert that data in to emergy. Uncertainty data for both direct inputs and UEV values (existing and original) were included in the life cycle model. Quantities of direct inputs to one of the nine unit processes were assigned a range of uncertainty based upon the same model d efined for the Ecoinvent database (Frischknecht et al. 2007). This model assumes data fit a log normal distribution. Using this model, the geometric variance, was estimated for each input. Calculations of uncertainty ranges for the UEVs for inputs to the p rocess were estimated based on a U EV uncertainty model (Ingwersen 2010). This model produces 95% confidence intervals for UEVs also based on a lognormal distribution, and is described in the form of the geometric mean (median) times/divided by the geometr ic variance, abbreviated in the following form : geo 2 geo (3) geo is the geometric mean or median and 2 geo is the geometric variance. The bounds of the 95% confidence interval are defined such that the lower bound is equal to the median divided by the geometric variance, and the upper bound is the median multiplied by the geometric variance. Original uncertainty estimations based on the analytical method (Ingwersen 2010) were performed for gold and silver in the ground.

PAGE 41

41 Allocation Two a llocation approaches were adopted: the co product rule often used in emergy analysis and a by product economic allocation rule used when applicable in LCA. The co product rule assumes that each product, in these case gold silver, and mercury, each require the total emergy of the mining process es for their production, and therefore the total mining emergy is allocated to each. Economic allocation is one method in LCA in which an environmental impact is divided among multiple products Economic allocation w as selected here in preference to allocation by mass because it most closely reflects the motivations of co product metal producers (Weidema and Norris 2002) In this case, revenue from production was used to allocate en vironmental contribution by determining the market value of the gold contained in the dor as a percent of the total value of dor and mercury production. The resulting percentage was used as the percentage of total mining emergy allocated to gold. The sa me method was applied for silver and mercury. In both cases, geologic emergy was allocated to each product separately, since the model used for estimating geologic emergy in the products was element specific. Data Management and Tools All inventory data w as stored in SimaPro 7.1 life cycle analysis software (PR Consultants 2008) A new process was created for each input. Emergy was entered as and given the equivalent of 1 Joule. 7 This unit was assigned to the emergy substance. 7 For purposed of functionality in SimaPro the integrity of the emergy algebra was not affected.

PAGE 42

42 created, for which emergy was the only input. A quantity of emergy in sejs was assigned to the output that corresponded with the unit emergy value (sej/g, sej/J, etc.). For inputs created that consisted of one or more system processes or other unit processes. 8 A new impact method was defined to sum life cycle emergy of all inputs to a process. To characterize total uncertainty (both input and UEV uncertainty) in the emergy of the mining products, Monte Carlo simulations of 1,000 iterations were run in SimaPro for estimates of confidence intervals of emergy in the products using both emergy co product and economic allocation rules. Results Environmental Contribution to Gold, Silver, and Mercury in the Ground The enrichment ratio of gold was estimated as 218.8:1, b ased on a reported gold concentration of 0.87 ppm (Buenaventura Mining Company Inc. 2006) and a crustal background concentration of 4 ppb (Butterman and Amey 2005) which using Eq. 1 resulted in an unit emergy value for gold in the ground of 3.65E+11 sej/g. The silver concen tration at the mine was not reported, but was estimated based on the silver in the product and a calculated recovery rate of gold (81.52%) to be 1.13 ppm. Using the background concentration of 0.075 ppm (Butterman and Hilliard 2004) the enrichment ratio of silver was estimated as 15.1:1, which resulted in an estimate of the UEV of silver in the ground at Yanacocha to be 1.54E+10 sej/g. The em ergy of mercury in the ground was estimated to be 1.71E11 sej/g based on concentration at the mine of 8.6 ppm (Stratus Consulting 2003) and a crustal background concentration of .085 ppm 8 e unit processes defined earlier as one of the nine phases of mining.

PAGE 43

43 (Ehrlich and Newman 2008) The total emergy in the amount of gold extracted and transformed into dor in 2005, just including the geologic contribution to gold in the ground, was 8.55E+18 (x) 10.7 sej (median times or divided by the geometric variance, as in Eq. 3). Environmental Contribution to Dor Table 2 1 shows the results of the total emergy in the mining products including for the dor, the gold and silver separately, and the mercury by product. The total emergy in the all life cycle stages contributing to 1 g of dor was approximately 6.8E+12 sej, w ith an approximate confidence interval of 6.2E+12 (x) 2.0. Considering estimated uncertainty both in the inventory data and in the unit emergy values, the emergy in dor could with 95% confidence be predicted to be as low as 4.4 E+12 sej/g and as high as 1.3E+13 sej/g, representing an approximate range of a factor of two around the median value. As a portion of the contribution to the total emergy in the dor, the geologic emergy in deposit form ation contributes approximately 3% ( Figure 2 3 ), but could be as high as 7% if the highest value in the range is used. The largest contributors to the total emergy of the dor include chemicals (42%) followed by fossil fuels (32%), and electricity (14%). Capital goods (mine infrastructure and heavy equipment) contribute 5%. Relative emergy contribution of inputs is not well associated with input mass because of differences in the unit emergy values of inputs to the process. Chemicals used in the process illustrate this dif ference.

PAGE 44

44 Table 2 1 Summary of emergy in mine products based on two allocation rules. All units are in s ej/g. A minor input by mass used in the processing stage, lead acetate, contributed more em ergy than did lime, whose mass input was 267 times greater. Emergy by Unit Process Breaking down the life cycle of a product into unit processes is not typically done in emergy analysis, but is a common step of interpretation in a life cycle assessment. A nalyzing process contribution can help target where in the life cycle environmental burdens are greatest. Figure 2 4 shows the breakdown of emergy and primary energy by mining unit process. The largest environmen tal contribution comes from the extraction process. Extraction emergy is dominated by diesel fuel consumed by mine vehicles. The other production processes are chemically intensive processes. Together the production processes represent 67% of the total e mergy. Controlling for pollution to air, water and Product Geologic Emergy Mining Emergy Mining Allocation % Total Emergy 95% Confidence Interval Emergy based on co product allocation Dor 1.7E+11 6.6E+12 100% 6.8E+12 4.4E+12 1.3E+13 Gold in dor 3.7E+11 1.5E+13 100% 1.6E+13 1.0E+13 2.7E+1 3 Silver in dor 2.5E+10 1.2E+13 100% 1.2E+13 7.5E+12 2.2E+13 Mercury 1.7E+11 2.4E+13 100% 2.4E+13 1.6E+13 4.5E+13 Emergy based on economic allocation 1 Dor 1.7E+11 6.6E+12 99.9 0 % 6.8E+12 4.4E+12 1.3E+13 Gold in dor 3.7E+11 1.5E+13 97.31% 1.5E+ 13 9.9E+12 2.5E+13 Silver in dor 2.5E+10 3.0E+11 2.61% 3.3E+11 2.2E+11 5.4E+11 Mercury 1.7E+11 2.0E+10 0.08% 1.9E+11 1.8E+11 2.1E+11 1 Based on 2005 Au and Ag price received of $12.69/g and $0.26/g (Buenaventura 2006); Hg market price of $0.02/g (Metalprices.com)

PAGE 45

45 Figure 2 3 Environmental contribution (emergy) to dor by input type. Figure 2 4 Emergy and primary energy in 1 g of dor by unit process. Primary energy is depicted on a second axis which is adjusted so that emergy and primary energy in extraction appear the same so relative contribution of each to processes can be depicted.

PAGE 46

46 soil, which is the object ive of the auxiliary processes, contribute about 30% of the total emergy. Background processes contribute little (<4%) to the emergy in the dor. Figure 2 4 reveals differences in the absolute and relative contrib utions to processes as indicated by emergy and primary energy. First, the emergy for each process is six orders of magnitude greater than the primary energy in each process. Additionally the contributions of the non extraction processes are relatively gre ater when measured in emergy than when measured with primary energy. Primary energy reveals no use of energy in the deposit formation process, and relatively less energy in processes that are more chemically and materially intensive. Allocation and Emergy Uncertainty Relative emergy contribution of inputs is not well associated with input mass because of differences in the unit emergy values of inputs to the process. Chemicals used in the process illustrate this difference. Table 2 1 presents the differences in the gold, silver, and mercury UEVs according to the two different allocation rules used. Because of its high value, under the economic allocation rule the gold product is allocated 97.3% of the emergy, which results in a si milar UEV to that calculated under the co product scheme, where it is allocated 100%. The big difference appears in the calculations of the UEVs for silver and mercury (3E+11 and 1.9E11 sej/g ), since they are allocated small portions of the total emergy (2.61% and 0.08%) This reduces the silver UEV to 2.8% of the co product value, and reduces the mercury UEV to only 0.8% of the co product value. Uncertainties in process inputs ranged based on uncertainty in the inventory data, but primarily due to the unc ertainty of the UEVs. The inputs with greatest range of UEV values are the minerals and inorganic chemicals which are mineral based (see ranges

PAGE 47

47 between 1 and 1.5 for most inputs i n the inventory. Figure 2 5 shows the results of the Monte Carlo analysis of the emergy in 1 g of dor, illustrating the resulting uncertainty range for the dor product. The distribution is right skewed and rese mbles a log normal distribution. Overall the combined uncertainties in the inputs lead to less uncertainty in the dor (a factor of 2) than some of the major inputs (e.g. gold in the ground with a factor of 10). Discussion Usefulness of Emergy Results A significant finding of this LCA is that the environmental contribution to the mining process, dominated by fuels and chemicals, was estimated to be greater than that to the formation of the gold itself. This result holds despite the large uncertainty asso ciated with quantification of the environmental contribution to gold in the ground. The production of dor can also be interpreted to be process with a net emergy loss, with an emergy yield ratio (EYR) of close to 1, since the emergy expended in making th e product (represented here by the mining processes) is greater than the emergy embodied in the raw resource. 9 This is unfavorable in comparison with fossil energy sources and other primary sector products which generally have emergy yield ratios of greate r than 2 (Brown et al. 2009) but this provides no insight into the utility of the resource in society, which is much different in function and lifetime than these other products. 9 The EYR may be defined as the total emergy in a product divided by the emergy in purchased inputs from outside the product system (Brown and Ulgiati 1997)

PAGE 48

48 Figure 2 5 Monte Carlo analysis of 1 g of dor, showing the tails and center of the 95% CI, along with the mean (dashed line).

PAGE 49

49 While primary energy would indicate that the energy in mining is heavily dominated by fuel consum ption during extraction, using emergy as in indicator shows that the other more chemically and capital intensive processes weigh more significantly, and therefore that reducing total environmental contribution to the process would demand a broader look at the other processes and inputs. This is consistent with the trends in the results that Franzese et al. (2009) found in their comparison of gross energy and emergy in biomass. Quantifying resource use in emergy un its permits putting processes in the context of the flows of available renewable resources. Emergy used in a process can be seen as the liquidation of stocks of accumulated renewable energy in all the inputs to that process. The limit of sustainability, i n emergy terms, is such that total emergy used by society be less than or equal to the emergy driving the biosphere during the same period of time. Thus the liquidation of the stock of emergy should not be greater than the flows of emergy. In this case, the amount of emergy in the dor (the stock) produced by the mine in one year is equivalent to approximately one third of the emergy in sunlight falling on the nation of Peru in one year, and one third of one percent of the emergy in all the renewable reso urces available annually to Peru (Sweeney et al. 2009) 10 While this does not represent a trade off for the current period (since the stock of emergy in the dor was largely accumulated in a prior time period) it puts the total resource use in the process and the available flows of resources on the same scale, which is a step towards quantifying the sustainability of production. The Peruvian economy is driven on average by 35% percent rene wable resources, but the mining 10 Sunlight on Peru = 5E+21 J = 5E+21 sej (Sweeney et al. 2009) ; since 1 sej = 1 J sunlight. 1.66E+21 sej in dor /5E+21 sej in average sunlight on Peru = 0.3.

PAGE 50

50 process at Yanacocha itself is only approximately 3.5% renewable on a life cycle emergy basis. 11 This result should not come as a surprise since mining and other resource extraction activities are largely using non renewable energy sources to extract non renewable resources. The emergy in 1 g of dor is on the order of E+12 13 sej/ the dor, will have a minimum emer gy on the order of E+13 sej/g. This is hundreds of times greater than that reported for products from other economic sectors, such as biomass based products, chemicals, and plastics, which have UEVs consistent with the global emergy base used here on the order of E+8 E+11 sej/g (Odum 1996) reflecting the high environmental contribution underlying gold products, which is consistent with the high market value of gold. Emergy in LCA: Challenges The boundary, allocation and other accoun ting differences between emergy and LCA were dealt with here in a progressive manner. The system boundary was expanded beyond traditional LCA to included flows of energy underlying the creation of resources used as inputs to the foreground and background processes.. The inventory to the gold mining process involved a hybridization of background data from previous emergy analyses as well data from an LCI database. Numerous challenges remain for a theoretically and procedurally consistent integration of em ergy and LCA and are discussed here. 11 This includes only the portion of direct electricity use from hydropower. Energy sources for all other inputs are a ssumed to be non renewable.

PAGE 51

51 Challenges of using emergy with LCI databases and software This study revealed some of the complexities and potential inconsistencies of integrating emergy into LCA, particularly to be able to use emergy along with othe r LCIA indicators and to be consistent in use of accounting rules. The technical integration of emergy for the characterization of some of the processes (e.g. inventories for processes occurring off site) implemented here in SimaPro had the shortcoming of not being able to comparatively measure other environmental aspects from background processes in the life cycle. For some of these inputs for which emergy evaluations already existed (e.g. for stainless steel used in mine infrastructure and vehicles) eme rgy was the only input to the item, which made computation of other full life cycle indicators for resources use (e.g. cumulative exergy demand) impossible. A better method of integrating emergy into a Life Cycle Inventory would be to associate emergy wit h substances, and then to allow the software to track the emergy through all the processes, rather than creating processes that store unit emergy values. Such a method would permit more accurate cross comparison of emergy with other impact indicators. Emer gy evaluation conventionally incorporates the emergy embodied in human labor and services (Odum 1996) Adding labor as an input may be present in some forms in traditional LCA, such as in worker transportation (O'Brien et al. 2006) but energy in labor has largely been left out and its inclusion represents a potential addition to LCA from the emergy field. However, inclusion of labor, as in a typical emergy evaluation, is not included in processes in existi ng LCI databases including Ecoinvent by which the labor of background processes is included in an emergy analysis. cess inputs, estimated using a

PAGE 52

52 emergy:money ratio to represent the average emergy behind a unit of money, and represents labor in background processes based on the assumption that money paid for goods and services eventually goes back to pay for the cost o f human labor, since money never returns to the natural resources themselves (Odum 1996) Unit emergy For consistent incorporation of emergy in labor in an LCA, labor would also need to be incorporated into the background processes drawn from LCI databases. Unless the omission of an input which is considered to be integral to holistic accounting in emergy theory, since all technosphere products rely on human input. Reconciling rules for allocation is another necessary step for inclusion of emergy in LCA. In the LCA context, the emergy co product allocation would be inconsistent and non additive, because the emergy in the products would be double counted when they become inputs in the same system (which can be as large as the gl obal economy). Thus results based on this allocation rule should be recalculated using an allocation rule that divides up emergy before being used with existing LCIA calculation routines, to avoid the potential double counting of emergy. 12 Allocation rule s or alternatives to allocation typically used in LCA can easily be applied to allocate emergy among by products and co products, as was demonstrated here, but if existing UEVs for co 12 Emergy practioners also point out that emergy of co products cannot be double counted when they are inputs to the same system. See p. 1967 of (Sciubba and Ulgiati 2005) However in LCA all impacts have to be split according to one of the methods described in ISO 14044.

PAGE 53

53 products are incorporated they will have to be recalculated with the cho sen allocation rule before incorporation. Allocation is not just an issue among co products but also an issue related to end of life of many of the materials used. While many of the inputs to dor were transformed in such a way that they were completely c onsumed (e.g. the refined oil is combusted), others, particularly the gold itself, was not consumed in such as manner. Gold is a material that can theoretically be infinitely recycled and is not generally consumed in its common uses (e.g. jewelry). In em ergy evaluation of recycled products, the amount of emergy that goes into the formation of the resource would be retained (i.e. deposit formation) for the materials each time its recycled (Brown and Buranakarn 2003) In contrast, it has been traditional practice for systems with open loop recycling, (like the metals industry) to split the total enviro nmental impact between the number of distinct uses of a material (Gloria 2009) If this approach were used it would require splitting the emergy of resource formation as well as the emergy of mining among the anticipated number of lifetime uses of the gold product. But allocation in systems with recycle loops is an unresolved issue in LCA especially for products such a metals and minerals and the problem is not limited to the context of integr ating emergy into LCA (Yellishetty et al. 2009) Energy in environmental support not conventional ly included in emergy evaluation While more thorough than other resource use indicators in consideration of energy us e from the environment, n ot all the energy required by the environment to support the dor product is included here. Geologic emergy in the clay and gravel used as a base layer for roads and the leach pads is not included, under the assumption that these

PAGE 54

54 materials are not consumed in the process. Additionally, t here are waste flows from the mine, some of which, such as those potentially emanating from the process sludge and residuals on the leach pads, ma y occur over a long period of time following mine c losure. These and contemporary emissions to air, water, and soil require energy to absorb, but these are not quantified here, as they are not typically quantified in emergy analysis. Other measures to quantify damage in this waste, though they may not be n umerically consistent with the analysis here, could fill in the information gap although unless they are consistent with emergy units and methods, they will not allow for a single measure of impact Traditional measures of impact used in LCA, such as glob al warming potential and freshwater aquatic ecotoxi city potential (Guine 2002) could serve this purpose. More investigation needs to be done to relate emergy with other environmental impact metrics within the LCA framework. The outcome of emergy and other LCA metrics may not warrant the same manageme nt action, esp. those LCA metrics that measure waste flows, as they are measures of effects on environmental sinks instead of use of sources. Uncertainty in unit emergy values Emergy from geologic processes in scarce minerals is characterized by a high deg ree of uncertainty (around a factor of 10) relative to other products largely due to the differences in different models used to estimate emergy in minerals (Ingwersen 2010) However there is limited analysis of uncertainty in emergy values .The largely unquanti fied uncertainty associated with UEV values needs to be addressed so that use of emergy in LCA attributes appropriate uncertainty not just to inventory data, but also to previous UEVs. The uncertainty of UEVs contributing 90% of the emergy was characteriz ed in this paper using a method proposed in Ingwersen (2010) Using a

PAGE 55

55 model to estimate UEV uncertainty to couple with inventory uncertainty will help to better quantify uncertainty in LCA studies that use emergy, which will permit statistically robust comparison of emergy in products that serve the same function (e.g. comparative LCA). Emergy and Other Resource Use Indicators As integrated into LCA in this analysis, emergy is suggested as one measure of resource use, defined as environmental cont ribution. Although primary energy use was the only other resource use metric that was quantitatively compared with emergy in this study, it would be useful to see how emergy compares with other implemented and proposed indicators of resource use in LCA, n amely indicators of abiotic resource depletion, direct material input and cumulative energy demand and cumulative exergy demand. Indicators of resource depletion are commonly used in LCA to represent how much of a particular resource is consumed in referen ce to its availability. 13 These are resource specific indicators and depend upon information on total reserves of various resources, which is not readily available. Emergy is not often applied to assess reserves and it is not resource specific. Use of e mergy as proposed here is therefore not closely comparable with indicators of resource depletion, which in cases of resource scarcity, convey very useful information on informing material selection. Direct material input has been used as an indicator, part icularly in the mining sector (see Giljum 2004) However it has also been argued to be of limited utility, 13 Resource depletion indicators are b uild into the most common LCIA methodologies including TRACI and Eco indicator 99 (Bare et al. 2003; Goodkoep and Springsma 2001).

PAGE 56

56 includes resources that are not transformed or consumed in proc esses (like overburden) (Gossling Reisemann 2008b) Emergy does take into account resource quality based on a principle that more embodied energy in creating a resource represents higher quality (Odum 1988) Of the resource use indicators, emergy is seen by some as closely related with exergy (Bastianoni et al. 2007; Hau and Bakshi 2004a) This is in fact only the case when conventional exergy analysis is expand ed to include available energy in inputs from driving energies in the environment ( Figure 2 1 ). Otherwise the boundaries for exergy consumption are like those in conventional LCA, and still do not account for the energy driving e nvironmental processes. Cumulative exergy consumption or a similar metric, entropy production (Gossli ng Reisemann 2008a) are useful measures of efficient use of the available energy embodied in resources, and thus relative measures of thermodynamic efficiency of systems, or ultimate measures of the depletion of a the utility of resources in the process of providing a product or service (B sch et al. 2007) Because of the similarity between exergy and emergy, one might expect redundant results by using both exergy based indicators and emergy based indicators. However, a brief comparison of the result of applying the Cumulative Exergy Dema nd (CExD) silver 14 to the emergy results here show some significant differences in the sources of exergy contribution in comparison with emergy contri bution. Approximately 72% of the exergy in this product comes from electricity production and 22% from the gold ore in the ground. In comparison with the results from 14 A detailed comparison between an inventory of this product with the inventory of Gold at Yanacocha is presented in the dis cussion of Supplement 2.

PAGE 57

57 this study ( Figure 2 2 ), emergy shows a much hi gher relative role of the fuels and chemicals used in the process 15 This can be largely explained by the differences in the information that emergy and exergy provide. Exergy and entropy production more precisely measure embodied energy consumption where as emergy is a measure of energy throughput and could be better described as measuring use than consumption (Gossling Reisemann 2008b) Also exergy describes the available energy in substances (including the chemical energy in minerals), which is not the same as the amount of energy used directly and indirectly in their cre ation in the environment. In summary, the use of emergy provides unique information regarding resource use that does not make other resource use indicators like exergy irrelevant, but rather can augment the understanding of resource use by tailoring their use to address questions at different scales (Ulgiati et al. 2006) However, emergy is the only one of these measures that relates resources used in product life cycles back the process in the environment necessary to replace tho se resources, and hence the best potential measure of the long term environmental sustainability of production. 15 This implementation of CExD in SimaPro is incomplete and does not provide characterization factors for many of the chemicals used in the refining processes. The relative exergy contribution of chemicals to total exergy in gold w ould likely be higher if this were the case.

PAGE 58

58 3 CHAPTER 3 UNCERTAINTY CHARACTE RIZATION FOR EMERGY VALUES 16 Introduction Emergy, a measure of energy used in making a product extending back to the work of nature in generating the raw resources used (Odum 1996) arises from general systems theory and has been applied to ecosystems as well as to human dominated systems to address scientific questions at many levels, from the understanding ecosystem dynamics (Brown et al. 2006) to studies of modern urban metabolism and sustainability (Zhang et al. 2009) Emergy, or one any the many indicators derived from it (Brown and Ulgiati 1997) is not an empirical property of an object, but an estimation of embodied energy b ased on a relevant collection of empirical data from the systems underlying an object, as well as rules and theoretical assumptions, and therefore cannot be directly measured. In the process of emergy evaluation, especially due to its extensive and ambitio us scope, the emergy in a object is estimated in the presence of numerical uncertainty, which arises in all steps and from all sources used in the evaluation process. The proximate motivation for development of this model was for use of emergy as an ind icator within a life cycle assessment (LCA) to provide information regarding the energy appropriated from the environment during the life cycle of a product. The advantages of using emergy in an LCA framework are delineated and demonstrated through an exa mple of a gold mining (Ingwersen Accepted) The incorporation of 16 Reprint with permission from the publisher of Ingwersen, W. W. 2010. Uncertainty characterization for emergy values. Ecological Modelling 221(3): 445 452.

PAGE 59

59 uncertainty in LCA results is commonplace and futhermore prerequisite to u sing results to make comparative assertions that are disclosed to the public (ISO 2006a) But the utility of uncer tainty values for emergy is not only restricted to emergy used along with other environmental assessment methodologies; uncertainty characterization of emergy values has been of increasing interest and in some cases begun to be described by emergy practiti oners (Bastianoni et al. 2009) for use in traditional emergy e valuations. Herein lies the ultimate motivation for this manuscript, which is to provide an initial framework for characterization of uncertainty of unit emergy values (UEVs), or inventory unit to emergy conversions, which can be applied or improved upon to characterize UEVs for any application, whether they be original emergy calculations or drawn upon from previous evaluations. Sources of U ncertainty in UEVs U ncertainty in UEVs may exist on numerous levels. Classification of uncertainty is helpful for identification of these sources of uncertainty, and for formal description of uncertainty in a replicable fashion. The classification scheme defined by the US EPA defines three uncertainty types: parameter, scenario, and model u ncertainty (Lloyd and Reis, 2007). This scheme is co opted here to represent the uncertainty types associated with UEVs. These uncertainty types are defined in Table 3 1 using the example of the UEV for lead in the ground. There are additional element s o f uncertainty in the adoption of UEVs from previous analyses. These occur due to the following: Incorporation of UEVs from sources without documented methods Errors in use of significant figures Inclusion of UEVs with different inventory items (e.g. with o r without labor & services)

PAGE 60

60 Calculation errors in the evaluation Conflicts in global baseline underlying UEVs, which may be propagated unwittingly Use of a UEV for an inappropriate product or process These bulleted errors are due to random calculation er ror, huma n error, and methodological discrepancy, which is not well suited to formal characterization and can be better addressed with more transparent and uniform methodology and critical review But uncertainty and variability in parameter s model s an d scenario s can theoretically be quantified. Table 3 1 Elements of uncertainty in the UEV of lead in the ground. Uncertainty Type Definition Example Explanation Parameter Uncertainty in a parameter used in t he model Flux of continental crust = .0024 cm/yr Global average number. A m ore recent number is .003cm/yr (Scholl and Huene 2004) Model Uncertainty regarding which model used to make estimations is appropriate See model for minerals in Table 2 Variation exists between this model and others proposed for minerals Scenario Uncertainty regarding the fit of model parameters to a given geographical, temporal, or technological context Variation in enrichment ratio based on deposit type Assumption that the emergy in all minerals of a given form is equal Models for D escribing U ncertainty in Log normal Distributions Different components of uncertainty in a model must be combined to estimate total uncertainty in the result. These component uncertainties may originate from uncertainty in model parameters. In multiple parameter models, such as emer gy formula models, each parameter has its own characteristic uncertainty. Uncertainty in environmental variables is often assumed to be normal, although Limpert et al. (2001) presents evidence that lognormal distributions are more versatile in application and may

PAGE 61

61 be more approp riate for parameters in many environmental disciplines. This distribution is increasingly used to characterize data on process inputs used in life cycle assessments (Frischknecht et al. 2007; Huijbregts et al. 2003a ) A spread of lognormal variable can be described by a factor that relates the median value to the tails of its distribution. Slob (1994) defines this value as the dispersion factor, k 2 geo : 2 geo of a = ( 1 ) a = 1 + ( 2 ) 2 geo for variable a is a function of a (Eq. ( 1 ) ) 17 which a simple transformation of the coefficient of variation ( Eq. (2)), 18 where a is the sample standard deviation of variable a a is the sample mean. This can be applied to positive, nor mal variables with certain advantages, because parameters for describing lognormal distributions result in positive confidence intervals, and the lognormal distribution approximates the normal distribution with low dispersion factor values. The geometric 2 geo 2 geo ) is a symmetrical measure of the spread geo ,, and the tails of the 95.5% (henceforth 95%) confidence interval (Eq. (3)). CI 95 = geo (x) geo 2 ( 3 ) variable a may be defined as in the following expression (Eq. (4)): 17 Eq. (1) adapted from Slob (1994) 18 Eqs. (2) (4) adapted from Limpert et al. (2001)

PAGE 62

62 geo = ( 4 ) The confidence interval describes the uncertainty surrounding a lognormal variable, but not for a formula model that is a combination of multiplication or division of each of these variables. The uncertainty of each model parameter has to be propagated to estimate a total parameter uncertainty. This can be done with Eq. (5): 2 geo of model = ( 5 ) where a b z are references to parameters of a multiplicative model y of the form y = Note that parameter uncertainties are not simply summed together, which would overes timate uncertainty. This solution ( Eq. (5) ) is valid under the a ssum ption that each model parameter is independent and lognormally distributed. Describing the confidence interval requires the median, or geometric mean, as well as the geometric variance. The geometric mean of a model can be estimated first by estimating the model CV (Eq. (6)) and then with a variation of Eq. (4) (Eq. (7)). 19 CV model = ( 6 ) geo of model = ( 7 ) Models for U ncertainty in UEVs Selecting Appropriate Methods for Uncertainty Estimations Numerous methods exist for computing unit emergy values 20 but for uncertainty estimation, it is impo rt to distinguish between them according to a fun damental 19 Eqs. 5 7 adapted from Slob (1994)

PAGE 63

63 difference in the way UEVs are calculated: the formula vs. the table form model. The formula model is used for estimation of emergy in raw materials such as minerals, fossil fuels and water source s (the UEV in Table 1 is of this form) The traditional table form evaluation procedure is typically used for ecosystem products and products of human activities. Formula models are generally multiplicative models using estimates of various biophysical f lows and storages in the biosphere as parameters. In order to quantify variability within a formula model, such as an emergy calculation, the result distribution needs to be known or at least predicted. M odel parameters are generally positive values mul t iplied to generate the UEVs. Such multiplicative formulas have been shown to lead to results approximating a log normal distributio n (Hill and Holst 2001; Limpert et al. 2001) Therefore it would be logical to ass ume that UEVs calculated in this manner are distributed lognormally. The model geometric mean and variance (Eq s ( 5 ) and ( 7 )) used in conjunction, offer an analytical solution for estimating uncertainty for formula type unit emergy values, with some bu ilt in assumptions, foremost being that the model parameters have a common lognormal distribution. For models with parameters of mixed and unknown distributions and large coefficients a variation, a common method for estimating uncertainty is to simulate a model distribution using a stochastic method such as Monte (Rai and Krewski 1998) A notable drawback of a stochastic simulation method is that the results obtained have some variability in themselves, which, however, can be reduced by increasing the number of iterations 20 See (Odum 1996) for procedure for calculating UEVs, which are also known as transformities when the denominator is an energy unit, or specific emerg y when the denominator is a mass unit.

PAGE 64

64 Table form UEV calculations would be more accurately described as sum products, where UEVs of inputs contribut ing to the total emergy in an item of interest are multiplied by the quantities of each input to get emergies in those inputs, and the emergy in each input is then added together to get the total emergy in the item of interest. This hybrid form operation i s not readily amenable to an analytical solution (Rai and Krewski 1998) In the absence of a readily available analytical model for this type of UEV, a Monte Carlo model may be adopted for modeling UEV uncertainty for table form calculations. Figure 3 1 provides an conceptual overview of the proposed uncertainty model. The analytical solution is used to model all quantifiable sources of uncertainty (parameter, model, and scenario) while th e Monte Carlo model is used only to estimate total parameter uncertainty. M odeling P rocedure and Analysis First the geometric variance and medians of five formula type UEVs are estimated with the analytical solution to describe the type of variability and distribution of some commonly used UEVs, breaking down the uncertainty into the three classes described. Parameter uncertainty for these same UEVs is then also estimated with the stochastic model, along with two table form UEVs. The modeling results are cross compared. As the distribution of UEVs has not previously been described, the resulting distributions from the stochastic model are tested to see how closely they fit traditional lognormal and normal distributions, as well as a hybrid of the two. In the process of this analysis a means of reporting UEV uncertainty for future incorporation and interpretation of uncertainty is described.

PAGE 65

65 Uncertainty was estimated for five formula type UEVs: lead, iron, oil, groundwater, and labor. These UEVs were chos en because they represent categories of inputs from the biosphere (labor excepted) scarce and abundant minerals, petroleum, water, and human input that form the basis of many product life cycles. Models for calculating e ach UEV are presented in Table 3 2 along with their sources. Parameter uncertainty was estimated as follows: ranges of values or multiple values from distinct sources when available were taken from the literature for each model parameter. The mean and sample standard deviation for each model parameter was calculated. With this value, the uncertainty factor, esponding to each parameter was calculated with Eq. (2). The UEV parameter uncertainty was then estimated for the combined parameter uncertainty factors with Eq. (4). Model and/or scenario uncertainty was incorporated by estimation of separate unc ertainty factors for these types of uncertainty. When multiple models existed for a UEV, the average and sample standard deviation of the UEVs produced by different models were calculated. Model uncertainty was estimated for lead, iron, petroleum and wa ter. When models exist for UEVs which are specific to a set of conditions but for which those conditions are unknown in the adoption of a UEV, scenario uncertainty can be included. For instance if labor is an input in a process, but the country in which the labor takes place is undefined, there is scenario uncertainty which includes the variability of the emergy in the labor depending on which country it comes from. Two scenario uncertainties were estimated for labor UEVs (one for US labor and one for wo rld labor) for purposes of example.

PAGE 66

66 Figure 3 1 Conceptual approach to modeling uncertainty. The parameter uncertainty consists of uncertainty and variability in the parameters used to estimate the UEV; t he scenario uncertainty consists of the uncertainty arising from use of parameter values for different geographic or technological scenarios; the model uncertainty from different models. Parameter along with either model or scenario uncertainty were combi ned for an estimate of total uncertainty by combining the uncertainty factors for each parameter and for scenario and/or model uncertainty according to Eq. (5). This can be summarized as: total uncertainty = parameter uncertainty + model uncertainty + scen ario uncertainty ( 8 ) In order to compare the consistency of the analytical solution for the median and geometric variance with the confidence interval generated by the simulation, stochastic simulation models for the lead, iron water, and labor UEV calculations were run. A Monte Carlo simulation was scripted in R 2.6.2 statistical software (R Development Core Team 2008) to calculate each UEV 100 times using a randomly selected set of

PAGE 67

67 Table 3 2 Unit emergy value models used for parameter uncertainty calculations. Category Model Source Minerals UEV mineral = Enrichment Ratio Land Cycle UEV, sej/g Cohen et al. 2008 Enrichment Ratio = (ore grade cutoff, %)/(crustal concentration ppm)/(1E6) a Land Cycle, sej/g = (Emergy base, 15.83 E24 sej/yr) / (crustal turnover, cm/yr)(density of crust, g/cm3) (crustal area, cm2) Odum 1996 Petroleum UEV Oil sej/J = (1.68 b emergy of kerogen, sej/J)(C content, %)/((Conversion of kerogen t o petroleum, fraction)*(Enthalpy of petroleum, 4.19E4 J/g)) Bastianoni et al. 200 0 UEV carbon in kerogen, sej/g = (emergy of C in phytoplankton, sej/g)/conversion to kerogen, fraction UEV Carbon in phytoplankton, sej/g = (phytoplanton UEV, sej/J)*(Phyt oplankton Gibbs Energy, 1.78E4 J/g)/ (phytoplanton fraction C) Groundwater UEV groundwater sej/g = (Emergy base, 15.83E24 sej/yr)/(Annual flux, g/yr) Buenfil 2001 Annual flux, g/yr = ((Precip on land, mm/yr)/(1E6 mm/km))*(Land area, km2)*(infiltra tion rate, %)*(1E12 L/km3)(1000 g/L) Labor Total annual emergy use model. UEV labor sej/J = ((Emergy use) c /(Population)*(Per capita calorie intake, kcal/day)(365 days/yr)(4184 J/kcal)) Odum 1996 a Omitted when concentration is reported in % b Includ ed for conversion from global emergy baseline of 9.44E24 to 15.83E24 sej/yr c Emegy use for global estimate was 1.61E26 sej/yr, or a total emergy use of the world's nations (Cohen et al. 2008) parameters. Randomized parameters were created with a rando m function using the sample standard deviation and means of each parameter. The parameters were assumed to be log normally distributed. The mean and standard deviations of the log form of each parameter were used to create variables with a lognormal distri bution, for which the following equations (Eqs. (9) and (10)) were used (Atchinson and Brown, 1957):

PAGE 68

68 logUEV = ( 9 ) logUEV = ln (UEV) O.5( logUEV ) ( 10 ) The resulting set of UE V approximations (100) provide a distribution from which the left and right sides of the confidence interval can be estimated by the 2.5 and 97.5 percentile values, respectively. In order to get a representative sample, this procedure was executed 100 tim es thus generating 100 distributions (for a total of 10 000 UEV values). From each distribution, the mean, median, and standard deviation values were reported, and these values were averaged across the 100 distributions to arrive at average values for each UEV. From the average mean and standard deviation, the 2 geo value for that UEV was estimated according to Eq. (1). The stochastic simulation did not incorporate the model and scenario uncertainty components, which could only be estimated by way of the a nalytical solution. The stochastic simulation recalculates the UEV by varying the parameters, but does not incorporate uncertainty from use of alternative models or on account of parameters from other scenarios. Thus to compare the stochastic and analyti cally derived results from parameter uncertainty, the calculated parameter 2 geo (Eq. (5)) may be compared with the 2 geo value obtained from the simulation distributions. Uncertainty was also estimated for two UEVs calculated with the table form model -electricity from oil and sulfuric acid made from secondary sulfur. The emergy tables used to estimate these two UEVs were simplified to include only items that contributed in total to 99% of the emergy in these items. 21 Uncertainty was estimated solely with the Monte Carlo simulation routine used for the formula UEVs, with th e 21 The table for electricity from oil was adapted from Brown and Ulgiati (2002)

PAGE 69

69 following change: uncertainty data in the form of 2 geo values for both inventory values (e.g. secondary sulfur in g in Table 4) and their respective UEVs (e.g. UEV for secondary sulfur in sej/g) were used in conjunction with their means to create random lognormal variables for use in the simulation. Est imation of the natural log form of the standard deviation for these variables for generating lognormal random values was 2 geo value instead of the sample standard deviation (Eq. (11)). logUEV = ( 11 ) The uncertainty factors in the Ecoinvent Unit Processes library for geometric variance were used for the 2 geo values for the inventory data (Ecoin vent Centre, 2007). For the UEVs of the inventory items, the deterministic mean and the geometric variance of the UEV for the same item calculated with the formula model were used when 2 geo value, respectively. This choice wa s based on the assumption that the inventory items (e.g. water to make sulfuric acid) had the same UEV as those calculated with formula UEV models (e.g. groundwater). The 95% confidence interval of the simulation distributions for formula and the table fo rm UEVs were compared with the confidence intervals predicted by a perfect log normal distribution ( geo (x) 2 geo ), those predicted by a normal lognormal hybrid distribution using the arithmetic mean as the center parameter ( (x) 2 geo ), and those predicted by a normal distribution Eqs. (1) (3) were used to estimate the geo 2 g eo from the derived from the sample distribution. The percent difference between the predicted and model distribution tails was calculated to measure the how accurately the predicted distributions represented the model distribution.

PAGE 70

70 Results The de tails of the uncertainty calculations for lead are shown in Tab le 3 3 For lead, parameter and model uncertainty were estimated. The 2 geo values (approximately the upper tail of the distribution divided by the median) for the fiv e parameters range 2 geo ) is larger than the largest 2 geo 2 geo values. The total uncertainty for lead, consisting of the combined m odel and parameter uncertainty (without scenario uncertainty) is dominated by the model uncertainty, which has a large 2 geo value due to large differences in previously published estimates used for the UEV of lead. The 95% confidence interval for the lea d UEV using this analytical form of estimation would vary across three orders of magnitude, from 4.38E+11 sej/g to 5.38E+13 sej/g. However, if the UEV model used to estimate the mean was the only acceptable model, the interval would shrink to 1.87E+12 1. 26E+13 indicating considerably less uncertainty. The geometric variance calculations from the analytical solution for the formula UEVs (lead, iron, crude oil, groundwater, and labor) showed a wide range of values presented in Table 3 5 Geometric variance values were dominated by model or scenario variances in the cases of the minerals and labor. The total parameter uncertainty ranged from 1.08 for labor to 3.59 for crude oil, whereas model uncertainty was as high as 9.12 for lea d. The confidence intervals estimated from the analytical and stochastic methods were of similar breadth (for all five formula UEVs), although they were not identical the intervals from the analytical solution were all shifted slightly to the left.

PAGE 71

71 Tab le 3 3 Analytical uncertainty estimation for lead UEV, in ground. No. Parameters 2 geo 1 crustal concentration (ppm) 1.50E+01 1.41 1.20 2 ore grade (fraction) 0.06 0.03 2.25 3 crustal turnover (cm/yr) 2.88E 03 6.77E 04 1.58 4 density of crust (g/cm3) 2.72 0.04 1.03 5 crustal area (cm2) 1.48E+18 2.1E+16 1.03 Models 6 Alternate Model UEVs 4.52E+11 7.25E+11 9.12 Summary 5.46E+12 Parameter Uncertainty Range (N o. 1 geo (sej/g) (x) 2 geo 4.85E+12 (x) 2.59 Total Uncertainty Range (No. 1 geo (sej/g) (x) 2 geo 2.57E+12 (x) 11.09 Sources 1 Odum (1996) ; Thornton and Brush (2001) 2 Gabby (2007) 3 Odum (1996) ; (Scholl and Huene 2004) 4 Australian Museum (2007) ; Odum (1996) 5 UNSTAT (2006) ; Taylor and McLennan (1985) ; Odum (1996) 6 ER method and Abun dance Price Meth ods (Cohen et al. 2008) The Monte Carlo simulation of th e UEVs produced largely right skewed distributions, as indicated by the means for UEVs (see column 3 of Table 5) being less than t he medians. Without exception the means of the simulated UEV distributions were less than the medians. The table form UEV calculation for sulfuric acid appears in Table 3 4 The geometric variance values for the inputs of second ary sulfur and diesel are those calculated for oil in the ground 22 ; the UEV for diesel is that calculated for oil; the UEV for electricity from oil was calculated from an emergy table and the geometric variance is the 2 geo value from the Monte Carlo simula tion; and the UEV and geometric variance for water are those calculated above for groundwater. The Monte Carlo simulation 22 Assuming the geometric variance is the same because they share similar UEV models, which is an assumption mentioned later in the discussion.

PAGE 72

72 resulted in a median of 6.51E7 and a 2 geo value of 1.75, which, in comparison with the formula UEVs, indicates less of a spread in t he distribution for this UEV. The other table form UEV, electricity, also had a 2 geo value less than that of its major input, crude oil, suggesting a pattern of less breadth in the confidence intervals of table form UEVs than those of their most variable input. Table 3 4 Emergy s umma ry with uncertainty of 1 kg of s ulfur ic acid. a Relative Data Relative UEV Solar Data Uncertainty UEV Uncertainty Emergy No Item (units) Unit 2 geo (sej/unit) 2 geo (sej) 1 Secondary sulfur 2.14E+02 g 1.32 5.20E+09 3.59 1.11E+12 2 Diesel 3.41E+03 J 1.34 1.21E+05 3.59 4.13E+08 3 Electricity 6.30E+04 J 1.34 3.71E+05 2.77 2.34E+10 4 Water 2.41E+05 J 1.23 1.90E+05 1.95 4.57E+10 Product 5 Sulfuric acid 1.0 0E+03 g 1.18E+09 3.31 1.18E+12 b CI 95 = 8.10E+08 (x) 3.31 Notes: 1. UEV for secondary sulfur and diesel from Hopper (2008) Uses k value for oil since secondary sulfur is a petroleum by product 4. UEV in sej/J = (UEV for global groundwater, 9.36E5 sej/g)/(4.94 J/g) Footnotes: a Inventory data from Ecoinvent 2.0 (Ecoinvent Centre 2007) b Example of incorporation of a confidence interval into an emergy table assuming a lognormal distribution Table 3 6 summarizes the results of the Monte Carlo simulations for all UEVs when the parameter distributions were assumed lognormal, and compares the resulting confidence intervals against those that would be predicted by lognor mal, hybrid, and normal distributions. A number of notable differences are present between these results and those of the calculated uncertainty values for formula UEVs in Table 3 5 The UEV means from the simulation are higher in all cases than the deterministic means presented in Table 3 5 but the simulation median values are lower than the

PAGE 73

73 2 geo values from the simulation, which were calculated according to Eq. (1) from the average mean and standard deviations of the Monte Carlo distributions, are not identical to the parameter geometric variance values from Table 3 5 2 geo values were always 5% of the analytically calcul ated geometric variances. The lognormal confidence interval was the best fit for the simulated UEV distributions: error of the lognormal approximation of either the lower or upper tail was never larger than 5%. However this distribution tended to consist ently overestimate the confidence interval. 23 The hybrid distribution tended to predict a distribution shifted to the right of the model with increased error, and the normal distribution often predicted a lower tail many orders of magnitude less than the m odel value. The smaller the 2 geo value ), the better all predicted distributions fit the model interval. In the case of the two table form UEVs, electricity from oil and sulfuric acid, the lognorm al confidence interval tended to underpredict the model lower tail more severely (suggesting that the tail is closer to the mean), but was still the best fit when considering the combined error in both tails. The left tail of these model UEV distributions was more constricted, and in these cases the 2 geo value, reflected by the hybrid model, was a closer approximate of the lower tail. 23 This could be in part be explained by the fact that the equation (3) is more precisely for a 95.5% confidence, rather than a 95.0%, confidence interval (Limpert et al. 2001)

PAGE 74

74 Discussion and Conclusions How Much Uncertainty is i n a UEV and Can it Be Quantified? T o fully characterize uncertainty for UEVs, the sources of uncertainty need to be identified and quantified. The classification scheme introduced by the EPA provides a useful framework which helps in identification of quantifiable aspects of uncertainty. How ever in practice, describing the uncertainty in parameters, scenarios and models requires significant effort and must draw from previous applications of various models and across various scenarios. In this manuscript, the data sufficient to characterize t hese three types of uncertainty for each UEV was not readily available, and as a result in no cases has a total parameter uncertainty been estimated that includes all parameter, model, and scenario uncertainty for lack of either multiple models or modeled scenarios from which to include that component of uncertainty. Unless one or more of these types of uncertainty can be categorically determined to be absent for a UEV, the uncertainty measures presented here underestimate the tot al uncertainty in these UE Vs. Acknowledging this underestimate, how much uncertainty are in unit emergy values? Parameters for describing the uncertainty ranges inherit in 7 UEVs have been presented and analyzed here. Informally, emergy practitioners may have assumed an implicit rule of thumb is inappropriate. As quantified here the UEVs may vary with either less or more than one order of magnitude, but this is UEV specific. However, when UEVs have as their basis the same underlying models, if the parameters specific to one or more of UEVs have a similar spread, then the UEV uncertainty should be similar. Thus, as was demonstrated here, uncertainty values for a UEV may be co opted from an UEV

PAGE 75

75 calcul ated with the same model (eg. minerals in the ground) with reasonable confidence if original estimation is infeasible. Adoption of geometric variances from UEVs calculated with the same model would provide an advantage as a reasonable estimation of uncerta inty rather than a vague or undefined measure. Quantifying model uncertainty may have implications regarding the certainty of comparative evaluations. Figure 3 2 shows the UEVs estimated for different types of electricity in Bro wn and Ulgiati (2002) all fall within the range of confidence interyal of the UEV for oil, estimated from the mean UEV reported by the authors and the geometric variance calculated for this electricity type in this paper (2.77), using equations 5 and 6 to estimate the median and equation 3 to estimate the tails. Although it appears that from this analysis the UEVs of electricity sources would be statistically similar, this ignores the fact that many of the same UEVs are used in the inputs to these elect ricity processes. Hypothetically, if the same UEVs are used as inputs to processes being compared, relative comparisons can still be made, all of the variance due to the UEVs of inputs is covariance. This represents a problem of applying this uncertainty model to rank UEVs where there is strong covariance, which is not addressed here. Comparing the Analytical and Stochastic Solutions Multiple advantages of proceeding with an analytical solution have been listed in the risk analysis literature. These incl ude the ability to partition uncertainty among its contributing factors and identify factors contributing to the greatest uncertainty in a model (Rai and Krewski 1998) as well as the greater simplicity of calculation (Slob 1994) Further advantages suggested here in the context of UEVs are the ability to include other sources of uncertaint y which cannot be quantified in a simple Monte Carlo analysis, and the ability to replicate the values for geometric variance.

PAGE 76

76 However, because table form UEVs are the most common form of emergy evaluation, and the stochastic simulation method is the on ly method presented which is functional for this form of unit emergy calculations, the stochastic method is likely to be more useful to emergy practitioners. Model and scenario uncertainty components, which were not quantified in the Monte Carlo simulati on, can be particularly significant in emergy, due to the fact that emergy values for a product are often used across a wide breadth of scenarios, computed with alternative models, and adopted in subsequent evaluations by other authors without knowledge of the context in which the original UEVs were calculated. The most desirable solution to these problems with uncertainty would be: first for model uncertainty, to agree on the use of consistent models for a UEV type to eliminate the discrepancy that occurs between competing models; for scenario uncertainty, to make UEVs more scenario specific whenever possible to eliminate scenario uncertainty. Where elimination of this model and scenario uncertainty is not possible, an alternative would be to develop a mo re complex version of stochastic model that would include estimation of model and scenario uncertainty in addition to parameter uncertainty. Following from what is predicted mathematically, this study confirmed that formula UEVs as multiplicative products fit a lognormal distribution better than a normal distribution. Table form UEVs, while they are sumproducts, also tended to be better described by lognormal distributions than normal distributions, although the two UEVs simulated both fits this distributi on to a lesser degree than the formula UEVs. Using the deterministic mean as the center parameter for a multiplicative confidence interval, represented by the hybrid approach, may be a tendency of emergy practioners for

PAGE 77

77 simplified description of confidenc e intervals, but was shown here to result in more error than using the median, except for the estimate of the lower tail of the confidence interval for table form UEVs. C onclusions Ultimately the accuracy of UEV uncertainty measures depend upon the repre sentativeness of the statistics describing the model parameters. In this case a broad but not exhaustive attempt was made to describe uncertainty and variability in the model factors for the UEVs evaluated. For this reason, this author recommends sources of uncertainty be further investigated and more thoroughly quantified before they are propagated for use in future studies. The responsibility should rest with authors to diligently seek out and to summarize the uncertainty in parameters they adopt, and to perpetuate that uncertainty with the UEV uncertainty both to present the uncertainty of their own work and so that it can be adopted by those that use this UEV in the future. By describing uncertainty associated with emergy estimates, emergy is more li kely to become adopted as a measure of cumulative resource use or for other purpose in LCA. Description of uncertainty in parameters and across models and scenarios will increase transparency in emergy calculations, thus answering one of the critiques whi ch has hindered wider adoption (Hau and Bakshi 2004b) Uncertainty descriptors, namely the geometric variance can be used along with inventory uncertainty data to calculate uncertainty in estimates of total emergy in complex life cycles. It can be further be used to compare different life cycle scenarios with greater statistical confidence. Pairing UEVs with un certainty data and identifying sources of uncertainty will also help emergy practitioners understand and report the statistical confidence of their calculated emergy

PAGE 78

78 values and to prioritize reduction of uncertainty as a means to improve the accuracy of e mergy values.

PAGE 79

79 Table 3 5 UEV uncertainty estimated from the analytical solution. Item UEV Den. UEV (sej/Den.) Parameter geo Parameter 2 geo Model and/or Scenario 1 2 geo Total geo Total 2 geo Lower UEV using parameter uncertainty Upper UEV using parameter uncertainty Lower UEV using total uncertainty Upper UEV using total uncertainty Lead g 5.46E+12 4.85E+12 2.5 9 9.12 2.57E+12 11.09 1.87E+12 1.26E+13 4.38E+11 5.38E+13 Iron g 1.06E+10 1.15E+10 2.00 6.66 7.18E+09 7.53 5.73E+09 2.29E+10 1.52E+09 8.63E+10 Crude oil J 1.21E+05 9.78E+04 3.59 1.04 9.77E+04 3.59 2.72E+04 3.51E+05 2.72E+04 3.51E+05 Groundwater g 9.36E+ 05 8.90E+05 1.86 1.28 8.83E+05 1.95 4.78E+05 1.66E+06 4.56E+05 1.74E+06 Labor J 6.74E+06 6.73E+06 1.08 11.43 3.11E+06 11.44 6.26E+06 7.24E+06 5.89E+05 7.70E+07 1 All values represent model uncertainty, except for labor for which this is scenario uncertai nty Table 3 6 UEV Monte Carlo results and comparison of model CI's with lognormal, hybrid, and normal confidence intervals. 1 Item Monte Carlo Results Model 95% CI Predicted 95% CIs Lognormal CI Hybrid CI Normal CI UEV Type 2 geo 2 geo Lower Upper Lower error Upper error Lower error Upper error Lower error Upper error Lead F 5.19E+12 2.73 1.93E+12 1.38E+13 1.5% 2.6% 12% 17% 123% 11% Iron 1.30E+10 1.99 6.62E+09 2.53E+10 1.8% 2.3% 4.5% 8.8% 40% 6.6% Crude oil 1.57E+ 05 3.55 4.66E+04 5.44E+05 4.5% 2.9% 18% 27% 273% 14% Ground H2O 9.40E+05 1.92 5.06E+05 1.77E+06 2.9% 2.4% 2.6% 8.3% 35% 5.8% Labor 6.91E+06 1.08 6.45E+06 7.40E+06 0.32% 0.35% 0.25% 0.42% 0.57% 0.12% Electricity from oil T 2.81E+05 2.77 1.16 E+05 7.68E+05 12% 2.4% 0.85% 17.3% 126% 11% Sulfuric Acid T 8.10E+08 3.31 2.72E+08 2.67E+09 10% 0.50% 31% 47% 179% 96% 1 geo 2 F = formula UEV; T = table form UEV. UEVs are in sej/g for lead, iron, groundwater, and sulfuric acid, and sej/J for crude oil, labor, and electric ity from oil

PAGE 80

80 Figure 3 2 Published UEVs for electricity by source (diamonds on axis) from Brown and Ulgiati (2002), superimposed upon a modeled range of the oil UEV, using the geometric variance for elec 2 geo = 2.77) calculated in this paper.

PAGE 81

81 4 C HAPTER 4 LIFE CYCLE ASSESSMEN T FOR FRESH PINEAPPL E FROM COSTA RICA SCOPING, IMPACT MODE LING AND FARM LEVEL ASSESSMENT Introduction Although tropical fruits and their derivative food products make up a substantial and increasing portion of the fruit consumption in the temperate countries of Europe and North America 24 little life cycle data or published life cycle assessments (LCA) of these products are available. At the same time, large areas and substantial resources in tropical countries are dedicated to growing tropical fruits, such as banana, pineapple, and mango, primarily for export (FAO 2009) Associated local and global environmental impacts need to be accounted for and better managed both locally and globally as these fruits continue to grow as a proportion of temperate climate diets. One way to encourage better environmental management could be through LCA based Type III environmental product declarations (EPDs), so that quantitative environmental information can be used to help producers make better management choices and help buyers and consumers make informed environmental choices that take into account t he full product life cycle (Schenck 2009) Objectives The primary objective of this study was to conduct a background LCA of fresh pineapple production in Costa Rica to be used as a guide for creating a product category rule (PCR) for fresh pineapple, as specified by ISO 14025 (6.7.1 ISO 2006b) The development of a PCR is a mandatory step toward the process of creating an EPD. A goal of any product category rule is to enable comparative assertions of 24 Pineapple import growth (by weight) was 248% from 1996 2006 in the EU and North America while only 56% for grapes, 33% for bananas, 27% for apples, and 14% for oranges in the same period (FAO 2009)

PAGE 82

82 environmental performance between products of the same category. To create a PCR, a background LCA can be used as a reference for establish ing the environmental impact categories and indicators for reporting, methods for conducting inventories and estimating impacts, and calculation parameters for these inventories and impact models. Although the objective is to create a PCR for fresh pineap ple, this LCA is scoped bearing in mind the functional use of the product, to provide nutrition through fruit consumption, and thus is created with the wider intention of providing life cycle data relevant to a wider range of environmental impacts of conce rn in fruit product supply chains. Impacts are estimated with methods that are as globally valid and adaptable as possible, to permit comparable analysis with other fruit group food products. The LCA should have sufficient coverage to represent the range of climatic, field, management, and production levels so that ranges of potential impacts can be bounded with a statistical confidence. Furthermore comparisons of environmental performance are made between fresh pineapple and other fruits through the far m scale to provide an initial analysis of how fresh pineapple from Costa Rica compares to production of other fruits consumed raw or used as the basis of processed food products. A secondary objective is to provide a model for other such background LCAs of agricultural products, particularly for those that have yet to be performed in countries and environments where assumptions made in emission and impacts models may not hold and that hence require regional adaptation of these models for more accurate impac t assessment. The F resh P ineapple S ystem in Costa Rica Costa Rica is the largest provider of fresh pineapple to the EU and the US. Approximately 85% of pineapples imported to the U.S. in 2005 were produced in Costa

PAGE 83

83 Rica; in the EU 71% of fresh pineapple imports came from Costa Rica (FAO 2009) agriculture export (to bananas) in terms of international exc hange. This production has resulted in a rapid expansion of pineapple plantations in the Limon (Atlantic region), Alajuela (North region), Heredia (North region), and Puntarenas (Pacific region) provinces (Bach 2008) There are a number of environmental and health related concerns surrounding this recent expansion and the modern production process. Public concerns include soil erosion, pesticide contamination of natural areas and water supplies, lowering of water tables, worker exposure to agrochemicals, and impacts of organic wastes, among others (Sandoval 2009) Pineapples are primarily grown in three regions, hereafter referred to as the North, Atlantic, and Pacific regions, on ultisols but also on other well drained mineral soil orders. Pineapples for the fresh export market in Costa Rica are a highly technical, non traditional cash crop. The high level of technicality has resulted in a high degree of uniformity in production systems to meet international standards (e.g. GLOBALGAP) and produce competitive yield s and fruit quality. The variety grown almost universally Costa Rica can be found in Gomez et al. (2007) Fields are prepared with adequate dra inage and raised beds. Seed materials are most often suckers (shoots from existing plants) harvested within farms. Once established pineapples require regular fertilization primarily through foliar application of fertilizers. Nematicides, herbicides and insecticides are used to reduce pests and competition. Once mature (about 150 days on average)

PAGE 84

84 are ready for harvest in another six months, from where they are man ually harvested and transported to packing facilities. When plants are not left to produce a second harvest, they are chopped and the field is prepared again for another planting. Methods System Boundaries and Functional Units The LCA boundaries are the f arm stage though transport to the packing facility including all upstream processes ( Figure 4 1 ) Figure 4 1 Fresh pineapple production unit processes and boundaries for the LC A. The first unit process is the focus of this paper. The primary functional unit (FU) is 1 kg of fruit delivered to the packing facility. For comparison with other fruit products at the farm level, one serving of fruit at the packing facility is used, b ecause it is a more relevant unit for comparison because of its functional equivalency. The USDA defines a serving of fruit as 1 cup of fresh fruit, which for pineapple is 165 g (USDA 2009) In order to estimate the number of servings that can be obtained for 1 kg of pineapple the following equation is used: Serving s /kg fresh weight fruit = (edible fraction of fruit)/(kg fruit /serving) ( 12 ) For pineapple this results in 3.09 servings/kg fresh fruit. Life cycle inputs for all inputs of agrochemicals and machinery and related emissions are included. Permanent farm

PAGE 85

85 infrastructure (buildings and road) w as judged to be environmentally insignificant and excluded from the study. Data Collection A public call for producer participation in this LCA followed from a workshop organized in San Jose, Costa Rica in July 2009 for pineapple producers, government offi cials, LCA experts, and other potential stakeholders to present the concept of LCA based EPDs (Ingwersen et al. 2009) Participation in the LCA was anonymous to encourage s haring of production data and evaluating environmental performance without revealing any private producer data. Farms representing all three primary producing regions of the country, with management schemes including conventional and organic, and with siz es ranging from 1 to >1000 hectares were directly solicited in order to seek a representative sample. Following agreement to participate, each producer was sent a standardized questionnaire requesting data on historical farm area, production inputs inclu ding fuels, fertilizers, pesticides, water use, agricultural machinery models and use, yield, harvest schedule, distance and means of transport to the packing facility. Data collection was supervised through in person meetings with producer contacts to as sure common understanding of the questions for data collection. Data were later verified through comparison of data items across the entire participant pool to assure that input data were reasonably suited to pineapple production requirements. To acquir e site specific data for inventory emissions models, farms were visited and data on soils, topography, and operations were collected. Because of the discontinuity between the non annual production cycle and annual data collected from producers, all annual production input data had to be adjusted with the following equation :

PAGE 86

86 Input, x/kg pineapple = (Annual input, x/yr)/(Farm area, ha)/(Harvest kg/ha/harvest)(harvests/yr) ( 13 ) Because of the same reasons mentioned above, yield data were collected on a per harvest basis. Data on all production inputs were matched with the appropriate processes in the Ecoinvent v2.0 database (Ecoinvent Centre 2007) for inclusion in the inventory and entered into SimaPro software (PR Consultants 2008) after being converted into EcoSpold XML format for validation. For pesticides reported, mass of the active ingredient applied was determined and used as the mass of the pesticide input from Ecoinvent of the same class (Nemecek and Kagi 2007) New processes were c reated for inputs without appropriate equivalents in the Ecoinvent database by assembling their active ingredients under a new process. N P K fertilizers were estimated by combining single or double fertilizers in quantities to match the N P K weight ratio s of the actual fertilizers, as recommended by the Ecoinvent designers (Nemecek and Kagi 2007) Emissions and Impact Models Emissions and impact models were chosen based on the following criteria: 1 Universal midpoint models are used for global impacts (e.g., climate change) 2 Regionalization of universally applicable endpoint models are used for local impacts of concern when available (e.g., USETox) When appropriate ch aracterization factors are not yet available, the measured impacts are reported as the quantity of relevant emissions. Recent work in the environmental evaluation of the food sector has focused heavily on carbon footprinting, in conjunction with the develo pment of product level

PAGE 87

87 carbon footprinting standards (Sinden 2008) Acknowledging the growing importance of this effort, rules for car bon accounting in this LCA are set as synchronously as possible with the PAS 2050 standard. Land transformation from forest is a potentially significant contributor to carbon release surrounding agricultural products, especially in tropical regions (Ebeling and Yasue 2008) Carbon loss from land transformatio n in kg C/ha was estimated only when conversion from primary or secondary forest was reported. Loss was estimated by identifying the historical Holdridge life zones that occupied the land the farm currently occupies (Hold ridge 1967) and summing the carbon in living biomass (Helmer and Brown 2000) with the estimated soil carbon (IPCC 2007) and dividing this car bon loss over 20 years. Emissions to air resulting from on farm fuel combustion were estimated based on the same fuel specific coefficients and equations used for agricultural data in the Ecoinvent database (Nemecek and Kagi 2007) Estimating other emissions from farm stage processes required customization of emissions models capable of capturing, to the extent possible, the crop and field specific variables that af fect these emission rates. Models capable of parameterization with site specific inputs were used to estimate emissions of eroded soil, consumed water, nitrogen and phosphorus in fertilizers, and active ingredients of pesticides. Emissions of nitrogen an d phosphorus compounds to air and water are functions of crop and field specific factors. Pathways considered here for N include uptake, ammonia, dinitrogen oxide, and nitrous oxide formation and volatilization, and nitrate leaching and runoff Modeled pathways for P include uptake, phosphate runoff, and loss of P bound to sediments from erosion. Uptake quantities were based on the

PAGE 88

88 average N and P concentration in pineapple leaf tissue. Equations and references used in estimating N and P emission can be found in the Appendix. The PestLCI model (Birkveda and Hauschild 2006) was customized with site specific climate and soil data to quantify the fate of pesticides applied in the field to air and water. Because drainage is present on the majority of pineapple farms, drainage was ass umed to be 100% effective in the model and thus all emissions to soil that are either lost via direct runoff after application or after lost after leaching through the soil column were characterized as an emission to surface water. Pesticides not present i n the default PestLCI model provided by the authors were added into the database so that fate of all pesticides applied to the field could be characterized. Characterization was farm specific but application dates were unknown and thus the annual average of available, because the thick cuticle most resembles that of pineapple (Malzieux et al. 2003) Assumed canopy cover was 75% at time of application. All other default settings in PestLCI were maintained. For estimating consumed water, the FAO CROPWAT model (Swennenhuis 2009) was parameterized with site specific climatic and soil data, and plant specific water reported. Irrigation water was added through the irrigation schedule for farms that use irrigation. Farm specific climate data were taken from the FAO LocClim database based on the geographic coordinates of the farms, and coupled with farm data on irrigation practices from the questionnaires. Other general model assumptions and plant specific parameters can be found in the appendix.

PAGE 89

89 Soil erosion was estimated for each farm using the most recent ARS version o f the RUSLE2 model (Foster et al. 2008) and customizing it for site specific conditions. RUSLE2 models rain based erosion on overland flow paths. Not included in this model are wind based erosion and rain based erosion from ditches or other concentrated flow areas, which are less significant sources of erosion on Costa Rican pineapple farms. Climate data required for the model were interpolated with the FAO Locclim database from the nearest 12 weather stations, including temperature, monthly rainfall, and number of days with rain per month (FAO 2010) R values (rainfall intensity factors) were adopted f rom maps created in an implementation of the USLE model for the country of Costa Rica (Rubin and Hyman 2000) To parameterize the model, t he following measurements were taken in representative areas of each participating farm: the percent slope and effective length of the slope were measured for each unique slope in the farm segment using a clinometer and metric tape. A unique slope consist ed of a slope 2 3 % different from other slopes based on visual assessment or with unique drainage or contouring (e.g., bed direction) elements. In each area of the farm with a unique soil profile, the profile was described and samples were collected fo r soil texture analysis (Burt 2009) Slope and soil data collected in the field were used along with farm specific management data including production schedules and other general data on pineapple morphology. One model was run for each unique combination of soil, % slope, field geometry and production schedule within each farm. Results for each farm were then averaged based on the total farm area represented by those conditions. Erosion occurring during initial conversion of the land from previous

PAGE 90

90 land use was not estimated. All general a ssumptions and parameters selected for the RUSLE2 model are reported in the appendix ( Table D 7 ). Sensitivity analyses of the adaptations of the PestLCI, RUSLE2, FAO CROPWAT models were conducted by selecting environmental and ma nagement scenarios reported or assumed to exist based on expert knowledge of the sector. Analyses were performed using the production weighted average of sample data (described below) and the climate variables of the North region as the default condition. Percent changes from the default conditions were reported by sequentially varying model variables within ranges naturally present in climate, field conditions, pineapple physiology, or ranges reported in management and harvest schedule. Estimating the Sector Range of Environmental Performance In order to meet the goal of conducting an LCA representative of production in the sector and maintaining the anonymity of producers participating in the study, a single unit process was created from the inventorie s of the participating farms. This process was used to create a distribution of environmental impacts to characterize the sector, henceforth referred to as the sector range of environmental performance (RoEP). To create the unit process, production weigh ted average input data from the individual farms were used as means, and parameterized with confidence intervals based on ranges existing within and among farms, or moreover likely to exist within the sector. For pesticide inputs and related emissions, on ly inputs to conventional farms were used in the baseline because inventory data on biological control agents and their associated environmental impacts were not available. Each of these inventory inputs was parameterized with a standard deviation based on the variation among the sample farms, and assumed to have a normal distribution.

PAGE 91

91 A correction of uncertainty for each input had to be made to reflect the variation in yield within and between farms. A standard deviation of yield within each farm was est imated using the reported min, max, and mean production values. A production weighted combined uncertainty of the yield was estimated with a propagation of standard uncertainty formula (NIST 2010) of the form: CV y ield CV 2 a (P a /P total ) 2 + CV 2 b (P b /P total ) 2 CV 2 z (P z /P total ) 2 ) ( 14 ) where CV y ield is the coefficient of variation of the yield for the baseline scenario, CV 2 is the square of the coefficient of variation of the yield for a farm a and P a /P total is the percent of the total production of f arm a from the total production of participating farms. The uncertainty based on variation in production inputs per hectare and uncertainty based on yield were then combined to estimate total uncertainty for each input, using the simplified form of equat ion 14 : CV mod, input,i CV 2 y ield + CV 2 input,i, ) ( 15 ) where CV mod, input,i represents the yield modified coefficient of variation for input i The standard deviation used to parameterize a normal distribution for a given input, i was then estimated by multiplying CV input,i by the sample mean value. For the emissions inventory, log normal distributions were assumed and extremes from sensitivity analyses of the emissions models were assumed to represent the 2.5% and 97.5% values of these distributions. The geometric variance (GV emission ), or measure of spread of the lognormal distribution, of the modeled emission from the sensitivity analysis was estimated by taking the maximum positive % change from the tested parameter v alues, dividing by 100% and adding 1. 25 The variation based on the 25 For example, if they max pe rcent change from the default value from the sensitivity analysis was +60%, the estimated geometric variance = 1+60%/100% = 1.6.

PAGE 92

92 sensitivity analysis was combined with variation in farm yields and in the production input related to that emission (e.g. nitrogen fertilizers for nitrate). A variation of equation 4 for propagation of uncertainty for lognormal variables was used to combine uncertainty from sensitivity analyses with yield uncertainty using to the follow formulas: GV mod, emission i ln ( GV yield ) 2 + ln ( GV input, i ) 2 + ln(GV emission i, ) 2 ) ( 16 ) where GV mod, emission i is the yield modified GV of the emission, GV 2 yield i is again the GV of the yield, GV input, i is the GV of the respective input related to the emission, and GV mod, emission i is the GV of emission, i For emissions related to multiple inputs, the GV input, i used was the related input with the maximum coefficient of variation. GV for the inputs and emissions were calculated from the coefficient of variation with the formula (Slob 1994) : GV x x 2 )) ( 17 ) wher e GV x is either the GV of yield or input and CV x is the coefficient of variation of the input or emission. An exception to a production weighted average of emissions was made for modeling the emission of carbon dioxide potentially resulting from land use change. For estimation of carbon emissions, the PAS 2050 standard dictates that, for cases where an agricultural product is from an unknown location in a country, the land use transformation allocated to the product should be the carbon lost in conversi on of the most carbon rich ecosystem of the country divided by the lifetime of the crop (default = 20 years) (Sinden 2008) The max potential kg C/ha loss was estimated by overlaying the historical Holdridge life zones on current pineapple occupied areas (Holdridge 1967) selecting the life zone with the highest storage of above ground and below

PAGE 93

93 groun d carbon (Helmer and Brown 2000) adding in estimated soil carbon (IPCC 2007), and dividing this carbon loss over 20 years. The uncertainty r ange of carbon loss allocated to pineapples due to conversion from forest was then modeled with a uniform distribution with the min equal to 0 and the max equal to the max potential carbon loss, all in kg/ha. Monte Carlo simulations with 1000 runs were exe cuted in SimaPro for each impact (described below). The final RoEP was estimated by taking the ends of the 99% confidence intervals (0.5 th and 99.5 th percentiles) to represent the ends of the RoEP. LCIA Indicators The measures of environmental impact sele cted, or LCIA indicators, were chosen both because of their precedence in existing agricultural LCA and for their environmental relevance to both the geographically specific human health and environmental concerns of the regions as well as larger concerns associated with the farm stage in production of fruit products. Characterization was done for both farm stages and upstream processes for farm inputs (e.g., manufacture and transport of agrochemicals to the farm). Impact categories selected were cumulati ve energy demand, potential soil erosion, potential aquatic eutrophication, water footprint and stress weighted water footprint, human and freshwater toxicity, carbon footprint and land use. Soil erosion impact Soil erosion or loss is infrequently repor ted as an emission and lacks a suitable LCIA methodology to relate erosion to impacts to damage to ecosystems or human communities. Soil erosion was one impact category with particular concern to experts from non OECD countries and thus recommended for fu rther development in LCAs

PAGE 94

94 studies by members of the UNEP working group on LCIA in 2003 (Jolliet et al. 2003b) Soil loss or potential has been reported as an inventory indic ator in mass of soil lost or depleted per functional unit (Heuvelmans et al. 2005; Peters et al. 2010; Schenck 2007) and is done as such here. Cumulative e nergy d emand Energy use from non renewable resources is oft en considered an indicator appropriate for all product systems and has been shown to correlate well with other categories of environmental impact (Huijbregts et al. 2010) Total energy life cycle use in fuels and el ectricity is measured using the c umulative e nergy d emand (CED) indicator implemented in the Ecoinvent database (Frischknecht and Jungbluth 2007) Only characterization of non renewable energy from fossil sources is implemented here. A proposed indicator (Ingwersen Accepted) based on the emergy method is potentially a stronger indicator of resource use for agricultural systems, but, because characterization factors were not available for the majority of the Ecoinvent processes used in the inventory it was not applied here. Virtual water content and stress weighted water footprint Freshwater consumption and its resulting impacts on water availability and quality for ecosystems and human health is a signifi cant environmental concern, particularly in areas susceptible to drought or water scarcity from overuse. Food consumption is a strong driver of water use globally (Chapagain and Hoekstra 2004) Nevertheless, estimating freshwater consumption has only recently been developed in reference to the water required per unit of food output, a nd just in the last year been integrated into LCA as an LCIA method (Pfister et al. 2009) Here, water consumption is estimated both by the water footprinting method (Hoekstra et al. 2009) henceforth referred to as

PAGE 95

95 volumetric water footprint to reduce confusion of terms, and further extended as a midpoint LCIA method called stress weighted water footprint (SWWF), as des cribed by Ridoutt and Pfister (2010) The volumetr ic water content, also known as virtual water, represents the total consumptive water use of green water (rainwater), blue water (water stored in surface and groundwater), and grey water (equivalent water use required to dilute polluted water to background levels). Life cycle consumptive water use in background processes is not included in this study for lack of appropriate background data, which has been acknowledged as a shortcoming of existing LCI databases (Pfister et al. 2009). However, consumptive w ater use has thus far been shown to be heavily dominated by agricultural processes, and upstream process are assumed not to have a significant effects on the results. The green and blue water components in the farm stage were estimated with the FAO CROPWA T model as described above; grey water was estimated as the water required to dilute the nitrate emission from the farms to 10 mg/L (Hoekstra et al. 2009) Because the effects of water use for production are very different depending on the relationship of that use to regional water availability, the water stress index (WSI) i s applied as a characterization factor to relate use to its likelihood of depraving humans and ecosystems of water in the region. A WSI for Costa Rica of 0.0163 calculated by Pfister et al. (2009) as part of the creation of global characterization factors and was applied using an equ ation by Ridoutt and Pfister (2010) to calculate t he stress weighted water footprint: SWWF = WSI CR (WF proc,blue ) ( 18 )

PAGE 96

96 where WF proc,blue is blue water footprint in L/kg pineapple and WSI CR is the unitless water stress index for Costa Rica. Ridoutt and Pfister (2010) also propose calculating the SWWF by including the grey wat er. However, the water represented by grey water (the water necessary for dilution) is not depriving other users of water, so it is not included in the SWWF here. Aquatic eutrophication Macro nutrient excess is a threat to both terrestrial and aquatic eco systems, however it is perhaps more of a threat in aquatic ecosystems. The process of eutrophication in aquatic ecosystems (nutrient excess leading to sharp increase in primary production and subsequent increase in microbial oxygen consumption leading to a depletion of oxygen) is closely tied with runoff of N and P in agricultural fertilizers. The effects of N and P nutrient influx are system dependent, but freshwater systems are generally P limited and seawater, N limited. Studies in streams on the Car ibbean side of Costa Rica have shown that P addition can have cascading ecological effects on stream ecosystems (Rosemond et al., 2001). N escaping to the Pacific and Caribbean estuaries is assumed here to have the same effects documented in other estuari ne environments, such as the Gulf of Mexico (Miller et al. 2006) As a result, quantification of the effects of N and P in runoff from pineapple farms is performed here with regard to its potential to cause eutrophication. A variation of formula has been previously used (Gallego et al. 2010; Seppala et al. 2004) to create eutrophication characterization factors for aquatic ecosystems: cf e = tf e *af e *nf e ( 19 ) where the characterization factor for emission e is cf e (here in kg N/kg emission); tf e is the transport factor, the probability that emission e will be transported to an aquatic

PAGE 97

97 environment where it will have an effect; af e is the bioavailability factor for a emission e ; nf e is the nutritive factor for emission e whi ch is its ability to cause eutrophication relative to N. Because emissions to water from farms occur directly to freshwater environments, and because land in Costa Rica is 100% exorheic (rainfall terminates in ocean), so as for areas where this is the cas e in the US, as in Norris (2003) tf e is set to 1. Most of the air currents in Costa Rica move inward toward the mountains (Daly et al. 2007) with rainfall depositing airborne emissions back to the land so for emissions to air we also set tf e to 1. Availability factors are based on the relative proportion of readily available inorganic forms of nutrients to organic forms in this case only emissions of inorganic nutrients are characterized, so af e is set to 1 for all emissions. The nutritive factors for the emissions are all based on the Redfield ratio of 116:16:1 (C:N:P) as in Norris (2003) Because the ratio of N:P has been found to vary between 13 19 in aquatic systems, the CV applied to each nf and propagated the final cf e is 0.09. Each cf e is thus equivalent to the n f e since both the transport and availability factors are set to 1 here for all characterized emissions. The resulting values, especially for emissions to air, are notably higher those in the Ecoinvent implementation of TRACI (Frischknecht and Jungbluth 2007) which uses the average US characterization values, because they account for transport losses assumed not to occur here. Human and freshwater ecotoxicity Pesticides used in pineapple farming include herbicides, insecticides, nematicides and soil fumigants. Toxicity of these pesticides to humans and ecosystems is a function of fate in the environment, lifetime, transport, intake and effect. Models were reviewed that consider the fate, incidence of contact, and effect of pesticide emissions both on ecosystems and human health. Numerous models that have been used in LCA are

PAGE 98

98 ava ilable for this purpose, including USES LCA, IMPACT 2002+, CAL TOX, and others. Despite their similarities in purpose and orientation, results of these models have been shown to be widely divergent. Recognition of this divergence prompted the cooperative development of the USEtox model (Rosenbaum et al. 2008) USEtox was therefore selected to characterize toxicity here, in line with the intent of selecting models based on international consensus. USEtox is, however, base d on the European continent, and the characterization factors are based on the climate, population, land use, and other data geographically representative of Europe. Other authors have shown that characterization scores for pesticides in multimedia fate, transport and effect models are very sensitive to geographic variables (Huijbregts et al. 2003b) particularly soil erosion and fraction of surface water, which are very different in Costa Rica than in the Eu ropean continent. An evaluation of sources of uncertainty in the IMPACT model showed that the misrepresentation of geographic variables can potentially result in errors of three orders of magnitude (Pennington et al. 2005) Thus all geographic and demographic variables in the USEtox default model were tailored to the Costa Rican environment, which is henceforth referred to as USEtox CR. Results a re reported in number of disease cases for human toxicity, and potentially affected fraction of species/m3/day for freshwater ecotoxicity. Other indicators The IPCC global warming potential 100 year characterization factors (IPCC 2007) expressed in CO 2 equivalents, were used as characterization factors for emissions with a potential to cause global warming, which sum together to create the carbon footprint. Occupation of land is described in m2/yr without impact characterization.

PAGE 99

99 Results Pineapple Sector Inventory Pineapple field data on geographic location, topography, management and soils were collected for areas in total representing approximately 200 ha and producing approximately 18,000 tons pineapple/harvest or 10,000 tons/yr. Participating farms represented all three primary production districts (North, Atlantic, Pacific) and in cluded both conventional and organic, respectively represented by approximately 88% and 12% by total production of the sample. Complete data on production inputs in the questionnaires was provided for 93% of farms surveyed based on total production volume The production weight average yield among farms providing complete data was 95 36 tons/harvest with an average of 0.60 0.24 harvests/yr. The average yield reported for the sector is 67 tons/harvest (Gmez et al. 2007) Within farm yield v ariation between minimum and maximum yield/ha was up to 38 tons in one case, with an overall minimum of 48 tons/ha and a maximum of 129 tons/ha. Inputs per kg pineapple by category were 0.17 0.04 m 2 /yr of land, 0.0075 0.0030 kg fuels, 0.043 0.012 kg minerals in fertilizers, 7.8E 4 1.6E 4 kg pesticides and 3.3E 4 1.35E 4 kg machinery. The inputs and standard deviations for 1 kg of pineapple at the packing facility are presented in the Appendix. Soil Erosion The estimated average soil erosion for the sampled pineapple farms varied from approximately 2.5 to 5 tons/ha/yr, which was approximately 0.05 to 0.10 kg soil/kg pineapple. There was significant variation within individual farms with erosion estimates for slope profiles within farms varying fr om less than 1 to 40 tons/ha/yr in one case,

PAGE 100

100 which equated to a range of 0.05 to 0.82 kg eroded soil/kg pineapple; a maximum of 16 times the minimum that was diluted by the averaging of erosion within farms. For the sector range of environmental performan ce (RoEP), the median value was 0.02 kg eroded soil/kg pineapple with a lower confidence bound of 0.0005 and upper bound of 0.6 kg eroded soil/kg pineapple. The results of the sensitivity analysis show that % slope was the factor most strongly influencing the erosion results. An increase in % slope alone from 2.5% to 30% caused an increase in erosion in tons/ha/yr of 1680%. The sensitivity of soil texture, in reference to percent change in erosion from the baseline ( 38 to 92% of the baseline from low to highest erodibility), along with degree of contouring of the rows ( 53 to 0% of the baseline from standard to no contouring), use of plastic mulch ( 78%) and use of double harvesting systems ( 32% of the baseline) all had significant influences on the soi l erosion at the pineapple farms Summary tables of the sensitivity analyses for the soil erosion and other emissions inventory models can be found in the appendix. Cumulative Energy Demand (CED) of Pineapple The RoEP for life cycle cumulative non renewa ble energy demand of pineapple was 1.2 to 2.2 MJ/kg with a median value of 1.5 MJ/kg. Most of this energy is used to make production inputs (77%), particularly fertilizers (see Figure 4 2 ). Figure 4 3 shows a comparison with evaluations of apples (4 countries), oranges (2 countries), and strawberries (2 countries) using a serving of fruit 26 as the unit of comparison. This and 26 Servings/kg for fruits used for comparison in the results are: 1 kg pineapple = 3.09 servings; 1 kg apple = 8.26 servings; 1 kg orange = 4.06 servings; 1 kg mango = 4.18 servings; 1 kg cantaloupe = 2.88 servings (based on formula used for pineapple in methods, ((1 kg fruit)(edible fraction))/(weight of USDA kg/serving)). Comparisons to Pimentel and Coltro were made by calcul ating the CED of analogous inputs from Ecoinvent for reported inputs rather that using originally reported energy totals. See the Appendix for recalculations.

PAGE 101

101 forthcoming comparisons are only preliminary, a s the full ROeP of these other sectors, with the exception of orange (BR) in this case, is not fully characterized. Nevertheless, the median value of pineapple is higher than the values reported for apples and oranges, although there is likely cases in pro duction of these fruits (based on the RoEP of Brazilian oranges), where a better performing pineapple has a lower CED. This results differs from what is revealed in a comparison on a per kg basis, where the median of the RoEP for pineapple (1.5 MJ/kg) is in the middle of the RoEP of CED for the different apple sectors (1.2, 1.0, 1.67, and 2.4MJ/kg). The strawberries both show more than double the pineapple CED/serving Carbon F ootprint The carbon footprint RoEP for pineapple at the packing facility was between 0.16 and 1.42 kg CO 2 equivalent/kg, which is equivalent to a range of 52 to 469 g per serving. The majority of this carbon footprint could potentially come from carbon loss from land use change, which could add up to 1.24 kg CO 2 eq./kg pineapple in the case of conversion from tropical moist forest, which was estimated to contain 394 tons C/ha. Of the sample farms, no land conversion from primary forest was reported by the producers, with no resulting loss of carbon from land use change, and as th is is likely the case for many farms, RoEP is also reported without land use change. Not including land use change, approximately half of the carbon footprint occurred upstream of the farm (51%) and (49%) of the footprint occurred on the farm with 34% be ing contributed from N 2 O release from N fertilizer and 15% from CO 2 primarily from fuel combustion. Fertilizer production (44%), followed by pesticide production (4%), fuel production (2%), and machinery production (1%) dominated upstream carbon footprint ( Figure 4 4 )

PAGE 102

102 Figure 4 2 Contribution to CED of pineapple, at packing facility. Figure 4 3 Non renewable CED of one servin g pineapple in comparison with evaluations of the farming stage of other fruits. Sources: Apple DE and Apple ZA (Blanke and Burdick 2009) ; Apple NZ (Blanke and Burdick 2009; Canals 2003 ) ; Apple US and Orange US (Pimentel 2009) ; Strawberry ES (Blanke and Burdick 2009; Williams et al. 2008) ; Strawberry UK (Lillywhite et al. 2007; UoH 2005; Williams et al. 2008)

PAGE 103

103 The carbon footprint of pineapple, assuming no land use change, translates to approximately 0.03 to 0.08 kg CO 2 eq./serving. This is higher than reported for apples from New Zealand and the United Kingdom, close to that reported for strawberries from Spain but mostly lower than strawberries from the UK; noting that the full RoEP for these other fruits is not reported ( Figure 4 5 ). Figure 4 4 Contribution to carbon footprint of pineapple, at packing facility. Potential footprint from land use change is not included. Virtual W ater C ontent and S tress W eighted F ootprint Lower ET rates due to the physiological adaptati ons of the pineapple plants, along with infrequent to no use of irrigation due to high and consistent annual rainfall (with the exception of one farm) resulted in a lower evaporative portion of the virtual water content (green + blue water) for pineapple i n comparison with the farm stage for other fruits ( Figure 4 6 ). For pineapple, the non evaporative, grey water component is larger than the evaporative water, owing to the leaching of nitrate from use of N fertilizers in pineapp le cultivation. Most of the uncertainty in the virtual water content can be explained by the variation in the grey water footprint due to nitrate emissions; the sensitivity analysis of the CROPWAT model for pineapple showed little regional

PAGE 104

104 variation in es timated ET for pineapple fields; the most significant variable is the crop coefficient (relationship of crop ET to pan ET), which has variable estimates in the literature (Malzieux et al. 2003) The stress weighted water footprint (SWWF) of pineapple in the baseline scenario is negligible; the estimated confidence interval is 0.004 0.017 L/serving, because the water stress index for Costa Rica is very low (0 .02 on a scale of 0 to 1). In comparison with mango grown in AU, with a stress weighted water footprint on average of 74 L/serving, the effect on water deprivation caused by pineapple is negligible. Figure 4 5 Carbon footprint of one serving pineapple in comparison with evaluations of the farming stage of other fruits. Sources: Apple NZ (Canals 2003) ; Apple UK (Lillywhite et al. 2007) ; Strawberry ES (Williams et al. 2008) ; Strawberry UK (Lillywhite et al. 2007; UoH 2005; Williams et al. 2008)

PAGE 105

105 Figure 4 6 Virtual water content (VWC) for pineapple in comparison with other fruits. Evaporative and non evaporative water are included for pineapple and mango (green + blue + grey); only evaporative water is inc luded for apples and oranges (green + blue). Mango data is from Riddout et al. (2009) ; apple and orange data from Chapagain and Hoekstra (2004) Aquatic E utrophication The eutrophication RoEP was estimated to be between approximately 1 and 15 g N eq./kg pineapple or 0.3 to 4.8 g N eq/serving. More than 90% of potential eutrophication effects were related to NO 3 leached from fields (53%), phosphorus bound to eroded sediment, and leached phosphate (10%) ( Figure 4 7 ) P in eroded soil was a t he most variable of the contributors, with a cooefficient of variation of 173%, which relates to the high variability of erosion. The estimated percentage of P lost to erosion of all P applied varied between 0 and 20% among participating farms; percent of N estimated to leach from fields as NO 3 N varied between 10% and 34%. While direct comparison among evaluations of fruits using different methods of estimating eutrophication related field emissions is very difficult, preliminary

PAGE 106

106 comparisons can be made by multiplying emissions by the same TRACI characterization factors used in this study. The results are shown in Figure 4 8 Figure 4 7 Contribution to potential eutrophicatio n of pineapple by emission. Figure 4 8 Preliminary comparison of potential eutrophication effects of one serving pineapple in comparison with evaluations of the farming stage of other fruits. Sources: (Canals 2003) ; Apple UK (Lillywhite et al. 2007) ; Cantaloupe CR (Hartley B. and Daz P. 2008) ; Strawberry ES (Williams et al. 2008) ; Strawberry UK (Lillywhite et al. 2007)

PAGE 107

107 Human and E cological T oxicity The RoEP f or human toxicity was estimated to be 1.7E 10 to 1.1E 9 disease cases/kg pineapple, but could be as much as 1000 times greater or less, due to the uncertainties inherent in the USETox model. The RoEP for freshwater ecotoxicity was 0.2 to 1.4 PAF in m3/day /kg pineapple, but could be as much as 100 times up greater or less. The pesticides contributing the most to ecotoxicity are diuron, ametryne (herbicide), ethoprop, and paraquat (herbicide) ( Figure 4 9 ) Toxicity characterizati on does not necessarily correspond to quantity applied in the field; half as much ethoprop is applied as diuron and diazinon, and less of that applied is emitted from the field (5% for ethoprop vs. 26% and 27% of diuron and paraquat), but its toxicity effe cts when being transported and coming into contact with humans and freshwater ecosystems is much stronger on a unit basis. Not all pesticides have demonstrated human toxicity effects although they do cause damage to freshwater ecosystems, including ametry ne and bromacil. In contrast to the temperate environment (Denmark) in which PESTLCI was originally calibrated, the Costa Rican environment has higher average annual rainfall and solar insolation which increases the estimated runoff and abiotic degradation of pesticides, respectively. The PestLCI CR model shows a greater fraction being delivered to water, but a smaller fraction being delivered to air than in the default PestLCI model. Total emissions of pesticides are greater overall in the default model. The USETox CR characterization model for the toxicity effects of these pesticides also shows differences from the default European parameterization. The USETox CR

PAGE 108

108 characterization factors for ecotoxicity for emissions range from 1.5 to 6 times less than in USETox EU; characterization factors for human toxicity for emissions are equal for emissions to air but 1.5 to 3 times less for emissions to water. Despite these absolute difference, relative toxicities among these pesticides are modeled similarly. Figure 4 9 Relative contribution of active ingredients of pesticides used in pineapple production to (a) human toxicity and (b) freshwater ecotoxicity.

PAGE 109

109 Results S ummary Table 4 1 presents a summary of the life cycle environmental performance of pineapple production through transport to the packing facility. On farm processes are responsible for the majority of impacts (given since some impacts were only modeled at the farm s tage due to assumption it contributes the majority of this type of impact) with the exception of the cumulative energy demand and to carbon footprint; about half of the carbon footprint occurs upstream and half on the farm. The uncertainty of each modeled impact, as measured by the coefficient of variation, varies markedly from less than 10% for land use, for which yield variation is the sole contributor to uncertainty, to human toxicity, which has a high level of uncertainty due to the large uncertainty i n the toxicity characterization factors. Discussion The data underlying the inventory represent medium to large size farms in the three primary geographic zones in Costa Rica. Sufficient input data from the smallest producers (<10 ha) was solicited but not acquired, likely due to less stringent bookkeeping practices and also heavier reliance upon larger producer associations for tasks, managements, and equipment. The other end of the spectrum of producers, the largest national and multi national companies with farms >250 ha, is neither directly represented. A lthough solicited, none of the four largest companies agreed to provide primary data for this study. All emissions and inventory results reveal the importance of yield in impact estimations, confirm ing recent findings in agricultural LCA (Roos et al. 2010) With higher yields and an equal amount of impact/area, impacts are diluted across mo re product, representing higher environmental efficiency. The average yield reported for

PAGE 110

110 the sector (67 tons/ha) falls at the 9th percentile of the yield distribution of the sample farms that contributing production data, indicating a bias toward more prod uctive farms in the sample used to create the baseline scenario. However, because the reported average sector yield falls within the confidence intervals for yield varied here, this national average pineapple falls within the distribution modeled. It is n ecessary to reiterate here that the objective was to model the expected range of environmental performance in the sector, and that the range rather than the median or mean values should be the focus of the results. The wide ranges of performance evident fo r all impacts categories indicate the importance of farm level assessment to differentiate environmental performance of pineapple production among farms. In the initial comparisons of environmental performance between farm stage production of pineapple an d other fruits, where such comparisons were possible, pineapples perform within a similar range, seemingly better in some categories and worse in others, but the full RoEP for the other fruits was not published nor calculable in most cases, limiting the ab ility of comparison. The estimated RoEP for energy demand for pineapple showed it to be higher in energy demand than apples and oranges on a per serving basis, but lower than Spanish and British strawberries. The carbon footprint reflected a similar patte rns with less of a relative difference between pineapples and other fruits. Pineapple was lower in consumptive water use than apples, oranges and mangos, but higher than mangos in its gray water requirement. Without the need for irrigation in most areas and because of its physiological adaptations to water stress, water use impacts were minimal in comparison with other fruits. The broad RoEP of eutrophication for pineapple indicates

PAGE 111

111 the relatively higher degree of uncertainty for this category, and consid erable potential overlap in this respect with other fruits. Because production inputs dominate energy demand and carbon footprint, the relatively high agrochemical input intensity of pineapple cultivation (FAO 2006; Su 1968) may explain in part why these indicators are higher for pineapple in relation to other fruit. Additional expl anation is provided by the fact that there are less servings of pineapple per kg than the fruits compared here, largely because of the higher non edible potion of pineapple (about 50%). The Significance of Regionalized Emissions and Impact Models The sig nificance that climatic, geographic, crop, and field specific factors have in emissions and impact models is supported by the differences in outcomes of the regionalized and the original versions of models used here. Water loss estimates from CROPWAT are dependent on water balance calculations based on climatic, soil, and plant conditions, and estimated will differ greatly among different climate zones and by crop. The PESTLCI model showed great variation in emissions between the default conditions (Denm ark) and Costa Rica. Characterization factors for pesticides differed by up to 70 times for toxicity factors between the default USETox and the USETox CR model. Using regionalized models will likely have significant effects on LCA outcomes, and should be applied with careful attention to the capacity to accurately describe conditions, but is essential for more accurate characterization of local and regional impacts. Although regional data was incorporated into these models, all those adapted here operate independently and use a unique set of field parameters. Attempt was made to use consistent parameterization of these models, but there is no guarantee of

PAGE 112

112 consistency of model calculations of common parameters (e.g. runoff is estimated in PestLCI, CROPWAT, and RUSLE2). Some models achieve a higher degree of specificity (RUSLE2) than others (CROPWAT) and thus some do not utilize all data that could theoretically influence results. However, the use of freely, publically available models adaptable to a wide range of conditions is of high utility for likelihood of use and for comparability. The N and P fertilizers emissions model was adapted based on average pineapple nutrient uptake rates, but otherwise did not account for regional climatic conditions or soi l properties. The model presented here is an improvement upon solely arbitrary designation of emissions fractions of all forms of N and P (e.g. 35% of N leaches to soil), some of which, including N leaching, has been estimated to vary between 10 and 80% o f applied N (Miller et al. 2006) and may be sufficient for relative comparison among farms, but could be replaced with a more detailed process based model as is used here for soil erosion, water use and pesticide emissions. These models could all be improved with better parameterization based on data collection on pineapple farms in Costa Rica for variables including pineapple biomass, nutrient uptake, water use, and leaf permeability to pesticides. Estimated Environmental Impacts All estimates of environmental impacts need to be considered in light of the accuracy of their characterization and of the inputs data underlying this characterization. Experimental quantification of soil erosion is typically marked by high variability, usually because erosion is strongly event based and the difficulty of capturing a representative sample of eroded sediment. Data from experimental measurement of soil loss in CR are no exception to this (see Table 15 1, Rubin and Hyman 2000) In consequences models based on long term climatic and management data may be

PAGE 113

113 preferable and yield more comparable results for quantification of soil erosion in LCA. However they sho uld still be validated with existing data. The RoEP of 0.02 to 32 tons/ha soil erosion tons/ha/yr found here does confer with e xisting estimates of erosion of mineral soils under pineapple cultivation in Hawaii and Australia. Land use, energy use and car bon footprint were estimated with the lowest uncertainty, however the latter two are both heavily dependent upon the quality of the input data for upstream processes. Carbon loss through land transformation has been calculated to be a dominant factor in the carbon footprint of crops occupying former tropical forest (Fargione et al. 2008) and that could possibly occur for pineapple cultivation, if it replaces tropical forest. There is, however, little evidence to suggest that pineapple expansion in Costa Rica has been a direct cause of deforestation since 1990 (Joyce 2006) Nevertheless conversion from other types of land use, including secondary forest and pasture, could also result in carbon loss but is not quantified here. As far as eutrophication and toxicity impacts are concern, which are impacts based on potentially long range transport, persistence and availability in environmental media, the effects on ecosys tems (freshwater ecotoxicity) and humans (human toxicity) should be read with appropriate skepticism of the capacity of generic models to make accurate estimations without explicit spatial data; nevertheless because these aspects (fate, transport, toxicity effects) are all relevant to their ultimate effect, they should be considered superior to just reporting quantities of pesticides released. Potential Impacts Not Measured The scope of this LCA was strictly limited to environmental impacts, and did not inc lude any evaluation of social or economic impacts. Both of these impacts can

PAGE 114

114 potentially be accounted for in LCA, with the related tools of Life Cycle Costing (LCC) and the newly developed Social Life Cycle Assessment (SLCA). Aside from loss of stored car bon, land use conversion and occupation can have ecosystem consequences on biodiversity across multiple scales (ME Assessment 2005) and this should be accounted for in the LCA, and has been recommended for consideration and methods are under development, but none were judged to be sufficient to capture effects on biodiversity of pineapple production in the st udied environment. Handling and application of pesticides in the field could have direct impacts on worker health, but no suitable methodology exists for measuring this in LCA. However all farms sampled reported use of protective equipment among workers in the field to reduce this risk. Residual organic waste on pineapple fields has been blamed for ecological consequences such as providing the substrate for the larval stage development of biting flies (Sandoval 2009) which have potential consequences for local livestock. Such consequences have not been addressed here. Conclusions and Recom mendations for Farm Level LCA of Fruit Products The development of inventories of agricultural processes and the characterization of their impacts are two separate but interdependent stages of the LCA. Since fruit products depend on further downstream proc esses before reaching the final consumer, inventories should include sufficient information that impacts can be characterized for their entire farm to disposal life cycle stages. Yet particular attention should be paid to those inventory items that need t o be recorded in the farm stage because of their

PAGE 115

115 likel ihood to d ominant full life cycle impacts: these include water use, eutrophication, toxicity, and soil erosion. Evidence here shows that it is essential to include upstream processes to fully characteri ze energy use for farm LCA, because energy use in agricultural inputs such as fertilizers may dominate cumulative energy use through the life cycle stage. Acknowledging this importance, life cycle data on farm input production adapted from LCI databases w ith a EU focus such as Ecoinvent used here needs to be validated for its application in other world regions. Because actual farm level energy use is dominated by liquid fuels for farm equipment such as tractors, energy use is likely to be strongly correla ted with other impacts during the farm stage dominated by fuel combustion, including greenhouse gas production, acidification, and photochemical oxidant production. Emissions to air causing these impacts should be included in agricultural inventories for use in full life cycle studies, but for sake of brevity and increased interpretability of LCA users, characterization of these impacts at the farm level is likely to be unnecessary because of its redundancy. This may not be the case if other energy source s (e.g. biofuels or electricity) comprise a substantial proportion of farm stage energy use. Use of LCIA indicators should be based both on environmental relevancy and sufficient characterization models and uncertainty estimation. In this case we recommend use of a measure of cumulative energy consumption, such as CED. Use of other broader measures of energy use, such as emergy, would present a richer picture of energy use that is more informative for measurement of long term sustainability, but should onl y be used if accurately integrated into the life cycle inventory and for which

PAGE 116

116 model uncertainty is described. Energy use also is characterized by relatively low model uncertainty, which increases comparability of different products. Local and regional environmental impacts related to soil erosion, water stress, eutrophication, and ecological and human toxicity are particularly relevant for farm level process and require characterization adapted to the region of production. Soil erosion is a particularl y localized indicator requiring a large amount of field specific information to accurately model. It is highly relevant for areas with sloped terrain and high rainfall. The direct downstream impact of soil erosion on water quality though sedimentation, wa s not quantified here but is a relevant environmental impact that deserves future investigation for LCA characterization. And as demonstrated here, accurate quantification of soil erosion can be particularly relevant for other impacts, including eutrophi cation, due to loss of nutrients bound to soil in erosion, and potentially for toxicity impacts, although the contribution of eroded sediments to those impacts was not quantified here. Farm level emissions are marked by high levels of variability, especia lly related to yields, and uncertainty due to complex and site specific fate, transport, and effect processes of agricultural emissions. We recommended that farm stage LCAs reported data along with sufficient range parameters to quantify uncertainty in in put data related to those emissions, uncertainty in the emissions themselves, and if characterized, uncertainty in the characterization factors. Finally, farm stage assessment data must be coupled with data on downstream life cycle stages before being ful ly evaluated by the end consumer.

PAGE 117

117 Table 4 1 Summary table for impacts of 1 kg pineapple delivered to packing facility. RoEP Contribution to Impact Variance of Impact Indicator Unit Min Max % contributi on of farm stage Most significant contributor CV Factor most responsible for variance a Land occupation m2/yr 0.14 0.21 100% yield 9% yield Soil erosion kg eroded soil 0.0005 0.6 100% farm slope 165% farm slope NR cumulative energy demand MJ 1.2 2.2 23% fertilizer production 25% yield Carbon footprint (with LUC) kg CO2 eq. 0.16 1.4 89% land use change 48% carbon loss from land use change Carbon footprint (no LUC) kg CO2 eq. 0.10 0.3 49% fertilizer production 19% yield Virtual water content L 124 1450 1 00% water for dilution of pollution 21% nitrate emission Stress weighted water footprint L 0.0044 0.017 100% water for application of fert./pest. 21% yield Aquatic eutrophication kg N eq. 0.0008 6 0.015 96% nitrate emission to water 62% P in soil eroded Human toxicity disease cases 1.7E 10 1.1E 09 100% Ethoprop (nematicide) 46% amount of ethoprop applied Freshwater ecotoxicity PAF/m3/da y 0.2 1.4 100% Diuron (herbicide) 44% fraction of diuron emitted to water Notes a Based on the largest CV for r elated inventory item among yield, associated input, or emission model. If this was the emissions model, the most sensitive variable in the sensitivity analysis was used.

PAGE 118

118 5 C HAPTER 5 SUMMARY AND SYNTHESI S Summary The primary objectives of this disserta tion were to better equip life cycle assessment to relate the production of goods and services to their associated environmental impacts by means of the following tasks: to provide a new process based life cycle assessment (LCA) impact method for quantifyi ng impacts of resource use with emergy; to provide this method with an accompanying method of uncertainty analysis; and to create a method for establishing the range of environmental performance for agricultural products with a set of indicators adapted to a tropical environment. To do so, this dissertation included two original LCA studies, one of gold silver bullion from the Yanacocha mine and one of pineapple production in Costa Rica, and an original uncertainty model for use with emergy results. The m ajor conclusions that can be drawn from these studies are first listed by chapter and followed by a general synthesis of the dissertation along with the ramifications of the findings. Chapter 2 Summary Emergy is an ideal measure of total resource use becau se it traces energy directly and indirectly used in creation of products back to the driving energies of the biosphere (sunlight, tides, and deep heat) and can be used to measure environmental contribution to raw and processed resources and materials as w ell as direct environmental flows (e.g. sunlight, wind, rain). All indirect and direct energy can then be aggregated as emergy in sunlight energy equivalents for a single numeric value of resource use. Emergy can be integrated into conventional process b ased LCA databases to track direct and indirect energy flows associated with complex process chains and in this manner is compatible with process based LCA. In order to characterize resource use with emergy for a mining product, an LCA of the gold silver m ining operation at the Yanacocha mine in Peru was conducted using an boundary that extended from the environmental contribution to the

PAGE 119

119 inputs to mining (permitted by emergy) to the creation of gold silver bullion. A gram of gold silver bullion was used as the functional unit. Total emergy in 1 gram of gold silver bullion is in the range of 4.4E+11 to 1.3E+13 sunlight equivalent joules (sejs), which is orders of magnitude higher than most common resources, including other minerals, fuel sources, foods, and ecosystem products. 95% of the emergy in gold silver bullion comes from inputs to mining processes rather than gold formation (and thus is based on environmental contribution that occurs off site), despite the millennia of environmental work used to form gold deposits. The contribution of emergy to chemicals and fuels used in the mining and refining processes dominate the emergy contributing to the bullion. The breakdown of emergy used to make gold silver bullion does not reflect the same pattern as cumul ative energy demand, indicating the failure of the latter to characterize all indirect environmental flows to processes, and reinforcing the role of emergy in LCA to quantify these flows for a more complete measure of resource use. Use of allocation rules from LCA for allocating impacts among by products and those traditionally used in emergy result in drastically divergent outcomes; allocation rules from LCA are more consistent with LCA data and should be used if results are to be adopted in future downstr eam LCAs ( e.g., for a product that uses gold silver bullion as an input). Tracking labor and information inputs into processes is not typically done in LCA and thus integrating emergy into life cycle assessment databases will not permit the quantification of emergy in labor or information which is a shortcoming to using emergy in LCA because it arguably omits important environmental contributions to final products that should be accounted for. Chapter 3 Summary The range of accuracy, or uncertainty of emer gy values should be quantified so that the model uncertainty of using emergy is quantified in an LCA study, as this could be the dominate form of uncertainty present in the LCA results that use emergy as an indicator. Two options are demonstrated for estim ating uncertainty of unit emergy values including an analytical model based on mathematical rules for propagation of uncertainty and a stochastic model using Monte Carlo analysis. Results of either approach show that unit emergy values have confidence int ervals that resemble lognormal distributions and that these confidence intervals can be represented mathematically with the median value times or divided by the geometric variance.

PAGE 120

120 Three forms of uncertainty are present in emergy calculations, including parameter, scenario, and model uncertainty. All three components can be combined using the propagation of uncertainty approach to result in the broadest estimation of potential uncertainty, but depend upon the estimation of the uncertainty in parameters and existing models. Unit emergy value confidence intervals f or table form unit emergy value calculations, the most common calculation approach, are only renderable with a Monte Carlo model approach because there is no simplified mathematical form for esti mating them analytically; thus the stochastic approach is suggested to be the most valuable of the approaches introduced. The estimated factor of uncertainty for emergy values does not always correspond to the presumed range of an order of magnitude. The uncertainty is variable but will be smaller than the uncertainty factor of the largest contributing input, demonstrating that uncertainty is not infinitely compounded in more highly transformed products. Issues remain with using uncertainty factors for com parison of unit emergy values that share common parameters. The method requires further adaptation for handling the issue of covariance. Chapter 4 Summary LCA based environmental performance of tropical fruit production in non OECD countries is largely un characterized in comparison with agricultural activities in temperate countries, yet the production of fruit has growing importance in the diet of North Americans and Europeans, and occupies increasing area in the tropics Environmental product declaration s provide one means of providing both LCA based information and a market based mechanism for reduction of impacts associated with production activities. A host of LCA methods need to be developed or adapted to account for the potential environmental impac ts that are very relevant especially in humid tropical environments. Fresh pineapple from Costa Rica is a crop of both growing export importance and increasing environmental concerns with production. A farm to gate LCA was designed to sample representativ e production systems and conditions present in the Costa Rican pineapple sector. A statistical method was used to combine variability in yield, production inputs, and emissions models to estimate a range of inputs and emissions relevant to energy use, wa ter consumption, soil erosion, land use, carbon footprint, eutrophication, and toxicity. Combined with impact characterization methods, this variability in inputs and emissions was used to create ranges of environmental performance for the sector.

PAGE 121

121 In addit ion to a functional unit of mass (1 kg), the functional unit of 1 USDA serving was used in order to compare LCA results with those generated for products that serve the same function providing 1 serving of fruit. Soil erosion is a primary environmental c oncern associated with pineapple production in Costa Rica because of exposure of topsoils, sometimes on steep slopes, to high rainfall, but no commonly used LCA method incorporates soil to the climate conditions and observed field parameters in Costa Rican pineapple plantations for estimating soil erosion. Methods developed for characterization of pesticide emissions (PestLCI), toxicity assessment (USETox), and crop water consumption (wa CROPWAT), and were each adapted to the extent possible to account for the local conditions. The result of these adaptation s were significant differences in characterization of impacts occurring in Costa Rica from the same charact erization in the default models (developing mainly in Europe), suggesting the importance of adaptation of emissions and impacts models to the environments in which the emissions occur. The ranges of environmental performance for pineapple, described by the coefficients of variation, ranged from 9% for land use to 165% for soil erosion, demonstrating significant variation within the sector, with range of performance for impacts where models incorporated local conditions being the most variable. The largest c ontributor to farm to gate energy use and carbon footprint was fertilizer production, thus stemming from upstream processes. On the farm level, greenhouse gas emission were dominated by N 2 O. Water consumption was low because of the low water requirement of pineapple and sufficient precipitation. Soil erosion was highest (close to 0.5 kg soil/kg pineapple) in areas with steep slopes, no contouring, and erodible soils, but is potentially as low as 0.005 kg soil/kg pineapple in flat sites with good drainage erosion resistant soils, and management practices that involve contouring, use of plastic mulch, and minimal exposure time. Eutrophication was dominated by nitrate emissions but was highly variable due to variable emission of P in eroded sediment. The pesticides contributing the most toward human toxicity and freshwater ecotoxicity were complex factors related to transport, degradation as well as potential health effect s and could not be predicted simply by mass applied or their specific toxicity, sup porting the importance of using emissions and impact models. Because of the variability within the sector, as well as potential model differences for some indicators, comparisons with other fruits were inconclusive. However, pineapple largely had a highe r energy use and carbon footprint than apples and oranges, but lower values than greenhouse grown strawberries. Pineapple had a lower virtual water content than apples, oranges, and mango although it had a large grey water footprint (pollution dilution wat er requirement) because of significant nitrate losses.

PAGE 122

122 Synthesis LCA serves the purpose of providing a measureable link between production of goods and services and pressure on environmental resources but requires further methodological expansion and refin ement to provide more relevant and accurate information for environmental decision making regarding production and consumption. Expanded methods described here include the use of emergy as an indicator of resource use accompanied with an uncertainty model for emergy and a unique combination of LCIA indicators applied with an original method of describing the range of environmental performance of a tropical agricultural product across the product sector in a country. The use of LCA as a tool to inform and direct sustainable production and consumption depends both on its methodological capacity to describe impacts accurately and a means of conveying complex environmental information in a form that is useful when making product design, management, or purchasi ng decisions as well as for informing policy making. In reference to its current methodological capacity, the incorporation of a measure of resource use that characterizes all process inputs in a common form using emergy provides a way of measuring the env ironmental contribution to products in the context of the availability of resources. Integration of emergy in LCA as a measure of resource use is not limited to mining products but is applicable to all products and services for which process data is avai lable to characterize them. The incorporation of additional or modified methodologies for impact assessment in LCA is also needed for cases where relevant indicators do not exist or are not accurate in the environments in which production processes take pl ace. Soil erosion is

PAGE 123

123 a relevant environmental concern with agricultural production in Costa Rica, but no adequate methodology for characterizing it existed in LCA. To this end the RUSLE2 model was incorporated into LCA as a measure of rain based soil eros ion. Furthermore other impacts models were adapted for use in the Costa Rican environment, including PestLCI and USETox, that had been developed for the Danish and European environments, respectively. Customization of impacts models for local conditions are not commonly performed in LCA, however, they are essential for characterizing impacts of emissions that have local or regional effects, such as eutrophication and toxicity associated with nutrient and pesticide emissions from farms. Through adapting m odel parameters to the local environment, strong comparability can still be achieved between products in different environments because the same model structure and its underlying physical assumptions are used. This is particularly necessary when environm ental conditions are sufficiently different than those in which a baseline model was developed, to the extent that they create meaningful differences in model outcomes. Despite the increased accuracy of LCA emissions and impact models that may be conveyed with greater environmental customization, uncertainty in the results will always be an issue because of lack of full primary data availability and because of model uncertainty. The uncertainty associated with model results can be large, as shown in some o f the impact assessment results for pineapple and those for gold, which can complicate the task of selecting environmentally preferable products. Incorporation of uncertainty information should be associated with any LCA impact method. Emergy was adapted as a new impact method, but there was no associated method of measuring model uncertainty in emergy. The new methodology developed now

PAGE 124

124 provides a means of characterizing the uncertainty of the emergy of a product that can accompany the use of emergy in L CA, as was demonstrated with the LCA of gold. One of the most valuable uses of LCA is for direct comparison of the environmental performance of products that serve the same function. By incorporating data and model uncertainty into LCA results, comparison s can be more realistic and subject to statistical tests that are not possible by comparing averages. Because incorporation of uncertainty may hinder interpretability and the purposes of comparison (because comparing ranges is less intuitive than comparing points), further methodological detail and guidance can assist in the consistency of uncertainty application and its practical usage. This is the very purpose of a current UNEP working group on uncertainty management in LCA. Another challenge related to the use of LCA for promoting sustainable production and consumption, other than the challenge of improving the accuracy of LCA data and models, is the meaningful presentation of LCA results in a form that both producers and consumers can interpret to make informed production and consumption decisions. One way to improve interpretability is to permit direct comparisons between a product and others in the same product category. Environmental performance for products within a category may include a high degr ee of variability that comes from differences in production practices and production site characteristics. Describing this variability can be potentially used to situate the environmental performance of individual product supply chains within the product category as an improved means of interpreting its environmental performance in addition to describing the performance range present across the sector for comparison with other products that serve the same function (e.g.

PAGE 125

125 a serving of fruit). The constructi on of such a range is often hindered by lack of sufficient production data to describe an entire product sector. However this range can be approximated by sampling representative production processes and for accounting for variability of environmental con ditions that could occur based on differences in production location, which are particularly relevant for agricultural production activities. A method for combining variation in production practice and environmental conditions to describe the range of env ironmental performance was developed here for pineapple production in Costa Rica, with the outcome being ranges of environmental performance for farm level pineapple production for Costa Rica. The production systems modeled in this dissertation were two pr imary sector products, gold silver bullion and fresh pineapple, from two different non OECD countries, Peru and Costa Rica. Product supply chains in non OECD countries, particularly those which are largely located in the tropics, have been poorly characte rized thus far through LCA. Implementation of LCA in non OECD countries requires adaptation of data and impact assessment (LCIA) methodologies for measuring the environmental impact associated with production chains in these countries for exports that are consumed in OECD countries. LCA is uniquely appropriate for quantifying the environmental burden of this production consumption pattern because it is only by accounting for impacts over the full life cycle that the responsibility of OECD consumers for en vironmental burdens in non OECD countries is quantifiable and thus can be addressed by associated market based or policy measures to reduce these burdens. This was demonstrated through two unique adaptations of LCA for one Peruvian and one Costa Rican exp ort product, with implications in each

PAGE 126

126 case for improved environmental management from both producer and consumer perspectives. The availability of process data and technical capacity to use LCA in non OECD countries is likely to be less than in OECD cou ntries. Where the data or capacity does not exist, incentives and new mechanisms for use of LCA in non OECD countries, particularly for products exported to OECD countries where there is a demand for environmental information, are required. The use of LC A based labels called environmental product declarations (EPDs) could be a market based mechanism for improvement of export production by providing information for buyers and consumers in importing nations that could be used to select the products tha t hav e the lesser impact. EPD p rograms for these product labels exist in a number of EU and Asian countries and are emerging in the US. Products from non OECD countries could be registered in these programs and used to inform purchasing decisions of buyers. Alternatively or in concert with OECD programs, EPD programs could be developed in non OECD countries as a way to gauge, publish, and promote environmental performance of export production. Thus EPDs are an application of life cycle assessment that could promote trade of more environmentally benign products by influencing both the production and consumption aspects of the supply chain. This particular application of life cycle assessment should improve LCA interpretability and function to broaden use into international markets.

PAGE 127

127 A APPENDIX A SUPPLEMENT TO CHAPTE R 2: PROCESS TREE AND UNC ERTAINTY ESTIMATES Figure A 1 SimaPro process tree of environmental contribution (sej) to 1 g dor. Inputs contributing 5 % or more of the total emergy visible.

PAGE 128

128 Table A 1 U ncertainty estimates for UEVs for inputs into gold silver bullion production. Item for which uncertainty estimated Uncertainty estimate used for 2 geo Refe rence Electricity, from oil Electricity from all sources in mix 2.8 1 Gold, in ground Gold, in ground 10.7 T able A 2 Groundwater, global All process water 2.0 1 Iron, in ground Pig iron, steel 7.5 1 Lead, in ground Pb in lead acetate and Zn in zinc powder 11.1 1 Oil, crude Crude and refined oil, natural gas 3.6 1 Silver, in ground Silver, in ground 10.6 Table A 3 Sulfuric acid sulfuric acid, HCl, general acids 3.3 1 Sources 1 (Ingwersen 2010) T able A 2 Estimation of total uncertainty in gold in the ground. No. Parameters 2 geo 1 crustal concentration (ppm) 4.00E 03 0.001 1.96 2 ore grade (ppm) 0.87 0.04 1.10 3 crustal turnover (cm/yr) 2.88E 03 6.77E 04 1.58 4 density of crust (g/cm3) 2.72 0.04 1.03 5 crustal area (cm2) 1.48E+18 2.1E+16 1.03 Models 6 Alternate Model UEVs 5.68E+14 9.22E+14 9.28 Summary 3.65E+11 Parame ter Uncertainty Range (No. 1 5) geo (sej/g) ( x ) 2 geo 3.35E+11 (x ) 2.27 Total Uncertainty Range (No. 1 6), geo (sej/g) ( x ) 2 geo 1.75E+11 (x ) 10.74 Sources 1 Butterman and Amey (2005) 2 Newmont (2006c) 3 Odum (1996) ; Scholl and von Huene (2004) 4 Australian Museum (2007) ; Odum (1996) 5 UNSTAT (2006) ; Taylor and McLennan (1985) ; Odum (1996) 6 ER method and Abundance Price Methods (Cohen et al. 2008) Odum (1991)

PAGE 129

129 Table A 3 Estimation of total uncertainty of silver in the gr ound. No. Parameters 2 geo 1 crustal concentration (ppm) 7.50E 02 0.007 1.20 2 ore grade (ppm) 1.13 0.06 1.10 3 crustal turnover (cm/yr) 2.88E 03 6.77E 04 1.58 4 density of crust (g/cm3) 2.72 0.04 1.03 5 crustal area (cm2) 1.48E+18 2.1E+16 1.03 Models 6 Alternate Model UEVs 4.97E+14 8.60E+14 10.03 Summary 2.54E+10 Parame ter Uncertainty Range (No. 1 5) geo (sej/g) ( x ) 2 geo 2.46E+10 (x ) 1.65 Total Uncertainty Range (No. 1 6), geo (sej/g) ( x ) 2 ge o 1.23E+10 (x ) 10.59 Sources 1 Butterman and Hillard (2004) 2 6 See Table 1 sources

PAGE 130

130 B APPENDIX B SUPPLEMENT TO CHAPTER 2: LIFE CYCLE INVENTORY OF GOLD MINED AT YANACOCHA Background The gold mine at Yanacocha, Peru operated by Minera Yanacocha, S.R.L is the largest gold mine in South America, and th e second largest in the world in terms of production volume. Ya nacocha is co owned by Newmont Mining Company(US), Buenaventura (Peru), and the International Finance Corporation. The Yanacocha mine is one of the largest gold mines (in terms of production) in the world. The mine produced 3.3275 milli on ounces of gold in 2005 (Buenaventura Mining Company Inc. 2006) This represented more t han 40% of Peruvian production (Peruvian Ministry of Energy and Mines 2006) 100% recovery of gold from dor and using the total of 2467 tonnes reported by the World Gold Council (World Gold Council 2006) Yanacocha is an open pit mine. Ore is obtained through surface extraction. Gold and silver are extracted from ore through cyanide he ap leaching and further refined through a series chemo and pyrometallurgical processes. The output of the Yanacocha mine is a gold silver bullion called dor, with a mercury by product. The dor is shipped overseas for further refining. Methodology Scope The scope of the life cycle inventory (LCI) included gold mining and processing from the stage of the deposit formation to the overseas export of a semi refined gold product (dor). The purpose was to include every critical link in the mining process, inc luding background and auxiliary processes, with the exception of administrative, community, and information and other mine support services. The choice to include all mine operations, described later, is based on the supposition that are all these operati ons are necessary for gold mining to occur within the current regulatory and business contexts. The scope is consistent with a cradle to gate LCI but extends further upstream to encompass both pre mining activity of the company and geologic work of the e nvironment. The downstream life cycle of gold production was not included. The inventory is based on total reported production in year 2005 This a source side LCI accounting for all the inputs to the process but not the emissions and wastes. Therefore t his inventory would not be sufficient for characterizing pollution impacts such as air, water, or soil contamination. Purpose This LCI was constructed to provide a measure of total environmental contribution to mining. Total environmental contribution was measured as the total energy used to supply all inputs tracing back to the energies that drive the biosphere (e.g. solar, tidal, deep heat). This energy, a form of embodied energy which includes environmental inputs, was estimated following the emergy meth odology (Brown and Ulgiati 2004; Odum 1996)

PAGE 131

131 The aim of this LCI is generally descriptiv e, rather than decision oriented (Frischknecht 1997) Neither was it completed for specific comparison. As a consequence, no inputs or processes were omitted because of redundancy with similar pr oducts or systems. Furthermore, the purpose was to complete a detailed LCI, rather than a screening LCA. Therefore rather than relying on existing LCI data, primary data from Yanacocha was used or original calculations specific to processes at Yanacocha w ere performed in all main unit processes and significant 27 indirect processes. Inventory Contents and Organization As is customary in LCI, the inventory was grouped into a series of unit processes (National Renewable Energy Laboratory 2008) Nine primary unit processes were identified and grouped into three unit process types These unit processes and types are identified in Figure B 1 Background and auxiliary processes are not always included in mining LCIs, but are both essential to the mining process. A generic mining LCI model called LICYMIN i ncludes auxiliary processes (Durucan et al. 2006) This inventory is unique among mining LCIs, in that background processes, including natural processes, are included. Data for the mining a ctivities are grouped by nine units processes except in cases where data was available only at the mine level, which was the case for labor. T his item is only tracked at the system level. Water included in the inventory was water used and evaporated in the process. Other water used that is recycled or released downstream was not included, as it was not considered to be consumed. Both raw materials inputs and core capital goods are included in the inventory. Core capital goods are defined as installations and heavy equipment critical to processes at Yanacocha. These include heavy vehicles, processing units such as ovens and reaction tanks, primary pipes, and large storage tanks. Auxiliary equipment such as connector pipes, structural skeletons, monitoring equipment are not included. Capital goods included elements of process infrastructure such as pad and pool geomembranes, pipes conveying process material and waste bet ween units, and earthen materials supporting pads and used in restoration. Earthwork was not included. Elements of non process mine infrastructure included in the inventory are roads, steel buildings, water supply, electricity transmission line, and dams. Equipment used in mine administration and maintenance such as small trucks, computers, protective clothing, were omitted. Employee support services such as food, medical, and housing services were not included due to lack of data. Infrastructure and mana gement of the San Jose reservoir, a reservoir for mine and community water storage within the mine boundary, was not included. 27

PAGE 132

132 Figure B 1 Process overview Nine unit processes (boxes) are grouped by th ree process types: background, production and auxiliary. Geologic processes led to deposit formation. Deposit discovery occurs during exploration. B efore a deposit can be mined the necessary infrastructure such as roads, electricity and water supply, and o ffice facilities are put in place.Mining itself begins with extraction which requires drilling and blasting away surface rock, and loading and hauling ore to leach pad. Leach pads and pools are prepared to contain and extracted ore and capture gold in solu tion in the leaching process. The leached solution is fur ther refined in multiple stages, including a retort process in which the mercury is separated. Pouring into dor bars completes the processing steps that occur at the mine. Excess water from process ing and acid runoff from pit is treated before release at water treatment plants. To prevent degradation of stream function sediment control structures are used to capture sediments. Once an area becomes inactive it is filled with waste rock, covered with top soil and in cases other protective layers, and replanted during reclamation. Data Collection The mining process was modeled based on written and graphic descriptions in corporate literature from primary sources. The model was corrected and/or confirme d through visits to the mine in July 2007 and in conversations with mine employees. Primary, public data from Newmont and partners were used as the source whenever based o n stoichometric formulas (for chemical reactions), equations in reference books

PAGE 133

133 (for mine equipment, operations and infrastructure), or calculated using, when necessary, generic industry data. Areas and distances utilized in calculations, when not publishe d in primary data, were estimated by delineating polygons of pertinent process footprints from satellite imagery in Google Earth software, saving them as KML files, and using a freely available web based KML polygon area calculator (GeoNews 2008) Inventory Cutoffs Rather than choosing a strict material, energetic or economic cutoff for data collection, inventory cutoff was based on contribution to final measure of resource impact from mining measured in emergy. Inputs estimated to contribute to 99% of all emergy were included. In many cases items with less than 1% of contribution to impact were included, because lack of significance could not be assumed prior to calculation. Many of these in puts were left in the inventory both to demonstrate their lack of significance and to make the inventory more complete for use with other measures of impact, for which relative impact would vary. Data Management The inventory data was managed in SimaPro 7. 1 software (PR Consultants 2008) Original processes and product stages were created for the primary unit processes identified ( Figure B 1 ) as well as for direct and indirect inputs to those processes. For some input data was replicated from processes available in the Ecoinvent database version 2.0 (Ecoinvent Centre 2007) The Ecoi nvent database was the only third party data used to avoid boundary issues that would result from incorporation of processes from other LCI databases available in SimaPro. Data underlying Ecoinvent processes were altered in some cases, such as for heavy ve hicles, where the most analogous Ecoinvent process (e.g. lorry, 40 ton) was modified with manufacturer data on weight to make it applicable to the mining process at Nature scope of this LCI. Transport and excavation inputs were omitted for infrastructure items adapted from Ecoinvent Processes were stored either as unit processes or system proces ses. Unit processes were used in all cases except for those indirect processes (e.g. fabrication of infrastructure) for which emergy values already existed, in which cases system processes were used. The process were named according to the following schem e: processes based on general estimates or calculation from the mining literature or other mines, no additional ending was attached to the name. When inputs were prepared of f site but added to the end and if this unit emergy value did or did not include labor a nd services

PAGE 134

134 Results The LCI consists of 164 SimaPro processes ( Table B 16 ) is the process for the final product ( Table B 1 ) and Mercury for the by product. All results are presented relative to the total production at the mine in 2005 of 2.17E+08 g dor which comprised 9.43E+07 g of gold, 1.23E+08 g of silver, and had by product of 5.99E+07 g of mercury. Mercury is represented by an Processing, Yanacocha Processing, without smelting, Yanacocha Gold at Yanaco cha, geologic emergy at Yanacocha, geologic emergy at Yanacocha, geologic emergy mining inputs are allocated to both the dor and mercury by products. Table B 1 No Process Amount Unit 28 1 Processing, Yanacocha 1 yr 2 Water Treatment, Yanacocha 1 yr 3 Gold at Yanacocha, geologic emergy 9.43E+07 g 4 Silver at Yanacoch a, geologic emergy 1.23E+08 g 5 Exploration, Yanacocha 1 year 6 Mine infrastructure, Yanacocha 1/mine_lifetime p 7 Extraction, Yanacocha 1.33E+11 kg 8 Leaching, Yanacocha 1.20E+14 g 9 Sediment and dust control, Yanacocha 1 year 10 Reclamation, Yanaco cha (6.56E+10*waste_to_reclam)+8.3E+07 kg 11 Labor, total, Yanacocha 1 p Notes All variables with their default values are listed in Table B 24 D escription s of the nine primary unit processes depicted in Figure B 1 and procedures for collection of data associated with these process are presented by process below. Deposit Formation The gold deposits at Yanacocha were formed by the flux of hydrothermal fluids containing Au and other minerals from deeper within the crust. These fluids pushed up and crystallized on near surface rock that had been previously altered by flows of magma. At Yanacocha, periods of volcanic activity producing magmatic flows alternated with hydrothermal flows over approximately 5. 4 million years created the deposits. Greater depth and detail on the formation of gold deposits at Yanacocha is provided by 28 All symbols for units are the same as those used in SimaPro 7.1.

PAGE 135

135 Longo (2005) The inventory for this process only contains the estimated mass of gold, silver, and mercury in the final products. Exploration The exploration model consists of land based exploration with a drill rig. I nventory data is presented in Table B 2 Drill rig use is based on Newmont worldwide ratio of oz reserve added to meters drilled, and reported reserve oz added at Yanacocha (Newmont 2006b) This results in 0.8 m drilled/oz reserve added. Drilling includes a diamond drill rig, diamond drills bits, and and water and diesel use for operation. Drilling calculations are based on Hankce (1991) Water us e is reported by the company (Minera Yanacocha S.R.L. 2005) Initial exploration is done though aerial surveys and remote sensing techniques, but this phase was not accounted for due to lack of data. Support for exploration teams and sample processing was also omitted. Table B 2 Inputs to process 'Exploration, at Yanacocha'. Output is 1 yr of exploration. No. Process Amount Unit 2 geo 1 Process water, at Yanacocha 1.37E+11 g 1.2 2 Di amond exploration drill, Yanacocha 50665 hr 1.3 3 Diamond drill bit 2.00E+02 p 1.3 4 Oil, refined, at Yanacocha 5.67E+13 J 1.3 Infrastructure Inputs to mine infrastructure are presented in Table B 3 Land use prior to mining was predominately pasture (Montgomery Watson 2004) Loss of aboveground biomass due to clearing for mining is included. Mine roads, water and ele ctricity supply, and buildings were included in the inventory. Total length and width of mine roads was estimated using satellite imagery. Models for road materials and constructions were created for three roads types: (1) hauling roads for use by heavy mine vehicles (approx 25m in width), (2) service roads (approx. 10 m in width), and a provincial highway connecting Cajamarca and the mine which was improved by the mining company for support of increased traffic and weight (Minera Yanacocha S.R.L. 2007) Road models were based on standards in accordance for support of vehicle weight and material type, based on California Bearing Ratios obtained from Hartman (1992) Table B 17 provides assumed road layer depths. Road materials and diesel used in transport of mat erials in road construction was included. Materials were assumed to be gathered on site, at an average distance of 2.5 km based on visual estimate Equations for transport of mine dump trucks (CAT 777C) were used to estimated trips and fuel use (see next section). Material and fuel use for (Spielmann et al. 2004) (Dones et al. 2003) (Althaus et al 2004) Distance for electricity and

PAGE 136

136 water supply networks were assumed equal to major mine road length (hauling road), and total water supply was reported by the company (Newmont 2006a) Total mine building area was estimated f rom satellite photos to the nearest 10000 m 2 (Althaus et al. 2004) Table B 3 Inputs t o process 'Mine infrastructure, Yanacocha'. Output is 1p. No. Process Amount Unit 2 geo 1 Hauling Road, Yanacocha 44 km 1.5 2 Service Road, Yanacocha 110 km 1.5 3 Highway, provincial 3.60E+06 my 1.5 4 Building, hall, steel 3.00E+04 m2 1.5 5 Pump station 6.21 p 1.2 6 Water supply network 44 km 1.2 7 Transmission network, electrici ty, medium voltage 44 km 1.5 8 Standing biomass before mining, Yanacocha 7895 acre 1.5 SimaPro. Extraction The extraction phase model is based on a process descriptions reported by the mining company (Minera Yanacocha S.R.L. 2005, 2006, 2007) and third parties (Infomine 2005; International Mining News 2005; Mining Technology 2007) The extraction phase commences with the remo val and onsite storage of topsoil. Drill rigs drill bore holes for placement of ANFO explosives for loosening overburden. Explosives are assumed to be ANFO type (Newmont 2006 a) Large mining shovels scrape overburden and ore into large dump trucks. Overburden is transferred into waste rock storage piles. Gold bearing ore is transported and stacked on heap leach pads. The total amount of ore mined, explosives used, percentag e waste rock, and water used are reported by Newmont (Minera Yanacocha S.R.L. 2005; Newmont 2006a) Inputs are presented in Table B 4 Table B 4 Inputs to process 'Ext raction, Yanacocha'. Output is 1.99E+11 kg extracted material. No. Process Amount Unit 2 geo 1 Scraper, Yanacocha' 596 hr 1.3 2 Drill rig, Yanacocha 2273 hr 1.3 3 Explosives (ANFO), at Yanacocha 7.71E+03 tn.sh 1.0 4 Mining shovel, Yanacocha 4.60E+04 hr 1.3 5 Rear dump truck, at Yanacocha 2.1+E+05 hr 1.3 6 Oil, refined, at Yanacocha 2.83E+15 J 1.3 7 Process water, at Yanacocha 3E+11 g 1.2

PAGE 137

137 Transport of Ore and Waste Rock Models and makes of mine vehicles were confirmed from the primary and secondary sources listed in the previous paragraph. Weight and capacity specifications for these vehicles were acquired from vehicle manufacturers. Fuel economy was estimated from data for another Newmont mine (Newmont Waihi Gold 2007) These specifications were used as parameters for vehicle production equations from the SME Mining Engineering Handbook (Lowrie 2002) for estimating total hours of use for scrapers, mechanical shovels, dump trucks, and stackers (see Table B 19 ). The estimated number of hours of use of each vehicle was then used to estimate fuel consumption. Mine Vehicle Model Fabrication and transport of mine vehicles was included in the inventory. Material composition electricity and gas used in fabrication of mine vehicles were scaled up from a simplified version o v1.3 ( (Spielmann et al. 2004) based upon the difference in weight. Only mass inputs into included, with the addition of copper, lead, electricity, and natural gas. M aterials were aggregated together in the case of iron (e.g. weights of wrought iron and pig iron were manufacturer of larger vehicles was attributed to steel for all vehicles ( 40% of weight) and rubber for vehicles (7% of the weight) with larger tires including the rear dump truck and scraper. Remaining additional weight was assumed to have the same composition as the 40 ton lorry. Vehicle models including weights and lifetime s and equations for scaling weights of materials and energy in vehicle fabrication are given in Table B 20 Leaching The leaching process at Yanaococha is a hydrometallurgical process whereby a dissolved cyanide so lution is dripped through gold and silver bearing ore to strip these metals and collect them in lined pool before being pumped out for further processing. Total leached solution processed in 2005 was 1.21E+14 g (Condo ri et al. 2007) The leaching process is a circular process whereby barren solution (from CIC plant) is recycled after replenishment with cyanide. A stacker is used to stack the extracted and delivered ore on the leach pads. Estimated use is based on or e quanity and S ME Reference Handbook equations (see Table B 19 ). A total of 4845.5 tons as of sodium cyanide as CN were consumed in this process in 2005 (Newmont 2006a) This was multiplied by molecular weight ratio of Na CN: CN to get estimated NaCN used. Calcium hydroxide, or lime, is added to raise the pH for optimal leaching. The estimated quantity of lime is based on an addition of .38 g CaOH:kg or e, which matches the total use reported by Newmont (Newmont 2006a) and is consistent with the range of 0.15 0.5 gCaOH:kg ore reported in Marsden and House (2006) Use of the leachpads and pool were based on a ratio of ore capacity to total pad area (Buenaventura Mining Company Inc. 2006) Details on leach pad and pool facilities were obtained from a mine tour and primary sources (Minera Yanacocha S.R.L. 2007; Montgomery Watson 1998) Leach pads consists of a clay layer, two layers of geomembranes, a gravel layer and collection

PAGE 138

138 and conveyance pipes. These inputs were estimated based on area and specifications. Total leach pad and pool areas in 2005 were reported by Buenaventura Mining Company Inc. (2006) The leach pad process is based on the largest pad at La Quinua. Fuel used in transport of the gravel from China Linda lime plant (12 km) and of the clay from borrow pits wit hin the mine (2.5 km) was estimate d assuming dump truck equations ( Table B 19 ), assuming use of a CAT 777C with a fuel economy of 129L/hr. Pipe network for leachate irrigation was not included. Leach pools for collecting leachat e prior to processing consist of three layers of geomembranes, a geotextile, pipes for collection and pumping to treatment, and storage tanks for NaCN and mixing. Table B 5 Inputs to process 'Leaching, Ya nacocha'. Output is 1.21E+14 g leachate. No. Process Amount Unit 2 geo 1 Stacker, Yanacocha 1.54E+05 hr 1.3 2 Sodium cyanide, at Yanacocha 6.74E+09 g 1 3 Lime, loose, hydrated, at Yanacocha 4.6E+10 g 1.2 4 Process water, at Yanacocha 4.23E+12 g 1.2 5 Leach Pad, Yanacocha 6.69E+05 m2 6 Leach Pool, Yanacocha 3.28E+04 m2 7 Recycled leach solution 1.25E+14 g Table B 6 Inputs to process 'Leach Pad Yanacocha'. Output is a 2.1E+6 m 2 leachpad. No. Process Amount Unit 1 Geomembrane, HPDE, 2mm thickness 2.10E+06 m2 2 Scraper, Yanacocha' 1.86E+03 hr 3 Geomembrane, LLPDE, 2mm thickness 2.10E+06 m2 4 HDPE Pipe, 40" dia. 6.67E+04 m 5 Fill material, Yanacocha 8.00E+08 kg 6 Gravel, crushed and washed, Peru 1.12E+09 kg 7 Oil, refined, at Yanacocha 1.63E+15 J Table B 7 Inputs to process 'Leach Pool, Yanacocha'. Output is a 1.03E+05 m 2 leachpool. No. Process Amount Unit 1 Geomembrane, HPDE, 2mm thickne ss 4.81E+04 m2 2 Geomembrane, LLPDE, 1mm thickness 1.03E+05 m2 3 Geomembrane, HPDE, 1.5mm thickness 2.06E+05 m2 4 Steel Pipe, 36" dia., at Yanacocha 2.74E+04 m 5 Geotextile, 8 oz. 3.09E+05 m2 6 Steel Pipe, 36" dia., at Yanacocha 1.70E+04 m 7 Storage tank, steel 1.50E+04 kg

PAGE 139

139 Processing G old bearing leachate is fu rther processed and refined on site into dor The process train includes carbon in column adsorption and stripping, Merrill Crowe precipitation r etorting, and smelting (Mimbela 2007) Wastes from these various stages go into process water treatment. These stages are aggregated together in an inventory Table B 8 ). Processing is assumed to be the major consumer of electricity. Electricity is purchased by the mine from the national grid. Provision of electricity was modeled after the national feedstock mix for Peru (Energy Information Administration 2007) Table B 8 Inputs to process 'Processing Yanacocha '. Ouput is 1 yr of processing. No. Process Amount Unit 1 CIC process solution, Yanacocha 1.06E+13 g 2 Merrill Crowe process, Yanacocha 1.16E+13 g 3 Smelting, Yanacocha 2.17E+08 g 3 Retort process, Yanacocha 1.16E+13 g 4 Electricity, at powerplant, Pe ru 1.07E+06 GJ The inputs included for the CIC process was activated carbon and the CIC plant infrastructure. A ration of 4 g Au: 1000g activated carbon with a reuse rate of 90% of Carbon in pulp process, 1.89E+08 g of zinc powder and 4.45E+08 g of lead acetate are assumed to be included. Estimates are based on ratios from Lowrie (2002) The retort process is m erely an empty place holder. The smelting process includes two smelters in addition to 1.68E+03 GJ natural gas, an amount based on a calculation of the energy necessary to heat gold to its melting point of 1337K, assuming a heat capacity of 25.4 J mol 1 K 1 and the operational parameters of the smelter (see below). Mass Balance Model A dynamic mass balance model was used to track the fate of core species through the process train (see Table B 21 ). Company reported concentrations of elements in the feedstock at various stages and concentrations of reagents used were set as constants in the model (e.g. Water used in process; cyanide used; ppm CN in the leachate; gold and silver in final product). Other ranges of con centrations not reported were gathered from the literature and upper and lower limits were used as constraints. Recycle loops back to the leaching process exists at each stage, as the solution is reused in the process. Values for unknown quantities were manipulated within upper and lower limits until all mass balance conditions were satisfied, within an error of 2% for water flows, and up to 5% for constituents. The following species were tracked through the processing stages: H 2 O (including pumped water and precipitation), CN, Au, Ag, Hg, and Cu, primarily to account for the various reagents used in the treatment chain, including activated carbon, zinc and lead acetate (for precipitation in the presence of lead acetate), and to account for the quantities of reagents used in treatment of the process water.

PAGE 140

140 Process Infrastructure Significant components of processing and water treatment infrastructure were included based on estimates during a site visit and through measurements of geo referenced aerial photog raphs (Google 2008) Infrastructure includes storage and processing tanks and steel buildings. Tanks were assumed to be steel and weights were estimate d from formulas from The Tank Shop (2007) Other process capital components included in the inventory were 2 tilt ing electric arc furnaces fo r smelting and a r everse o smosis membrane treatment system for process water. The tilting furnace was based on the Lindberg 61 MNP 1000 model 29 For simplicity the furnace was assume d to be 100% steel. Water Treatment Water treat ment at Yanacocha consists of treatment of process water and treatment of acid water from previously mined open pits and reclaimed pits. Treatment type, plus includes reported additional acid use in excess of the modeled requirements from the mass balance model ( Table B 9 ). Table B 9 treatment. No. P rocess Amount Unit 1 Acid Water Treatment, Yanacocha 1.42E+13 g 2 Conventional Process Water Treatment, Yanacocha 7.02E+12 g 3 Reverse Osmosis Process Water Treatment, Yanacocha 4.68E+12 g 3 Acid,Yanacocha, unaccounting for 1.08E+09 g Table B 10 Inputs to process Conventional Process Water Treatment, Yanacocha '. Output is 3.1E+12g treated water. No. Process Amount Unit 2 geo 1 Chlorine, at Yanacocha 1.17E+10 g 1.2 2 Iron(III) Chloride 3.02E+08 g 1.2 3 Sodium hydrosulfide, 100% 3.62E+07 g 1.2 4 Polyacrylamide (PAM) 3.00E+08 g 1.2 5 Sulfuric acid, 98%, emergy w/out L&S 4.91E+04 g 1.2 6 Electricity, at powerplant, Peru 1.16E+06 kWh 1.31 7 Conventional Process Water Treatment Plant, Yanacocha 0.05 p 29 Approx. weight 8000 lbs empty. Uses maximum of 3,100 cf per h r of natural gas based on 1,000 Btu/ cf natural gas. Max load 2,800 lbs. Melt t ime for this load about 3 hrs (Hosier 2008)

PAGE 141

141 Table B 11 Inputs to process 'Reverse Osmosis Proces s Water Treatment, Yanacocha'. Output is 5.55E+12 g treated water. No. Process Amount Unit 2 geo 1 Chlorine, at Yanacocha 2.09E+10 g 1.2 2 Sulfuric acid, 98%, emergy w/out L&S 5.40E+04 g 1.2 3 Electricity, at powerplant, Peru 1.20E+14 J 1.31 4 RO System 1.71 p Table B 12 Inputs to process Acid Water Treatment Yanacocha '. Ouput is 1.42 E+13g treated water. No. Process Amount Unit 2 geo 1 Lime, loose, at Yanacocha 7.96E+09 g 1.2 2 Iron(III) Chloride 7.10E+08 g 1.2 3 Polyacrylamide (PAM) 9.22E+08 g 1.2 4 Sulfuric acid, 98%, emergy w/out L &S 2.24E+04 g 1.2 5 Electricity, at powerplant, Peru 2.74E+06 kWh 1.31 6 Acid Water Treatment Plant, Yanacocha 0.05 p Water treatment process models are based on site visits and personal communication with engineers at Yanacocha. Process water treat ment included both conventional and reverse osmosis systems. Allocation between these systems is based on installed capacity in 2005. Chemical reagents used in these processes are included. Reagents quantities are based on reported quantities used when ava ilable or calculated based on total water treated and requirements specified in water treatment literature. Sludge waste from treatment is slurried and pumped back to the leach pads no additional long term management for sludge is included other than lea ch pad reclamation, as none is planned. Conventional process water treatment inputs were based on the following. Chlorine calculations were based on the stochiometric calculation of 4 mol Cl per mol CN with an e xcess ratio of 1.1 mol Cl (National Metal Finishing Resource Center 2007) NaSH is added to release cyanide bound to copper. Inputs is based o n the stochiometric equation from Coderre and Dixon (Coderre and Dixon 1999) PAM added is based on an optimal concentration of 65 ppm (Wong et al. 2006) The sulfuric acid add ition is based on a stochiometric requirement to adjust the pH of the water. Electricity of 0.193 kWh/ m 3 (Abou Elela et al. 2008) The reverse osmosis process only requires the addition of CN to destroy cyanide and sulfuric acid to adjust the pH after treatment. It does require additional electricity. The assumed electricity requirement was 6 kWh/m 3 treated water. Acid water treatment is assumed similar to process water treatment, without the addition of chlorine for cyanide destruction, and w ith the addition of additional lime for pH treatment. Lime added is based on the lime needed to adjust the pH of the influent from 2 11.

PAGE 142

142 Reclamation Reclamation models are based on primary data on restoration methods and long term mine closure plans (Montgomery Wat son 2004; Montoya and Quispe 2007) Total reclamation amount is based on the total amount of waste rock (material extracted), which is the difference between total extraction and total ore to leachpads. Inputs are all estimated relative to the mass of ov erburden returned to mining pits. All waste rock was assumed to be loaded from waste rock piles, transported and backfilled in pits, and limed at a ratio of 1gCaOH:1 kg fill. Fuel consumption for mining shovels and dump trucks is included and based on mini ng equations ( Table B 19 ). Protective layering, capping, seeding/planting and reclamation maintenance activities were not included due to assumption of insignificance to entire process (< 1%). Inputs to reclamation are shown in Table B 13 Table B 13 Inputs to process 'Reclamation, Yanacocha'. Output is 1 kg of returned overburden. No. Process Amount Unit 2 geo 1 Lime, loose, at Yanacocha 1 g 1.2 2 Rear dump truck, at Yanacocha 1.32E 06 hr 1.3 3 Mining shovel, Yanacocha 2.33E 07 hr 1.3 4 Oil, refined, at Yanacocha 9.79E+03 J 1.3 Sediment and Dust Control The primary measures taken at Yanacocha to redu ce sediment in runoff are serpentine structures immediately adjacent to mine facilities and three large sediment dams. Sediment runoff is based on sediment storage capacity in dams and dam lifetime. Thirteen serpentines are reported (Campos 2007) Dimensions of a representative serpentine were estimated from satellite imagery (Google 2008) Serpentines were assumed to be constructed of 1540 m 3 reinforced concrete. Flocculants to cause sediments to drop out of the water column were not included. Reinforced concrete w as also the only input included in sediment dams. Total concrete volume was reported as 7000 and 3000 m 3 for the Grande and Rejo dams respectively (Newmont 2004) Concrete for the Azufre dam, not reported, was estimated as the average of the aforementioned dams. The contribution of these structures is annualized over the assumed mine lifetime of 25 years. Mine roads are regularly watered to reduce particulates i n the air. The amount of water used by the mine in dust control was reported (Minera Yanacocha S.R.L. 2005) An evaporation rate of 50% was as sumed for water spayed on roads, and only this water, a total of 1.34 E+11 g, was included.

PAGE 143

143 Table B 14 Inputs for process 'Sediment and dust control, Yanacocha'. Output is 1 yr. No. Process Amount Unit 1 Sediment control structures, Yanacocha 0.04 p 2 Dust control, Yanacocha 1 year System Level Inputs Becau se labor was not reported by unit process, it was included as a system level input, Table B 1 Inputs to process Table B 1 ). Labor Energy in l abor was included based on the total hours worked and average human energetic consumption. Total hours worked by employees and contractors is reported by the company (Newmont 2 006a) Total J of energy in human labor at Yanacocha was calculated as: (3.82 E+09 J/yr avg human consumption)/(365*8 working hrs/yr)(2.3 E+07 hrs worked at Yanacocha) = 3.01 E+1 3 J/yr (1) work daily for 365 days a year. Transport Transport of materials and capital goods making up 99% of the mass of all inputs was considered. Sea, land, and air transport were all included. Inputs to transport included transport infrastructure constructi on and operation. Transport distance was based on origin of the item if known. If unknown, origin was first determined to be domestic or foreign by consultation of the Peru statistical companion for domestic production data and United Nation trade data for import export data ( Instituto Nacional Estadistica y Informacion 2006; United Nations 2008) If the item was produced or exported in quantities sufficient to supply the usage at Yanacocha, origin was assumed domestic and assumed to originate in Lima. If item was assumed to be of foreign origin, a sea distance of 5900 km was assumed (Los Angeles to Lima) in addition to road transport from Lima. Top ten items, mass inputs, and transport distances are given in Table B 23 Inputs for sea and air transport were based on the Ecoinvent processes 'Transport, transoceanic freight ship/OCE U', 'Transport, transoceanic tanker/OCE U', (Spielmann et al. 2004) An inventory of US truck transport from Buranakarn (1998) was adapted with data from Spielman and data on the Peruvian truck fleet (Instituto Peruano de Economia 2003) Data and notes are given in Table B 22 Due to complex g eography, an older fleet, and significantly less transport, ton km efficiency was assuming to be 50% of that of the United States.

PAGE 144

144 Life Cycle Model Parameters Various life cycle parameters c an be switched to include or ex clude inpu t of geologic emergy of ore, to clay and gravel construction material. By default these inputs are switched to '0', indicating they are not included. Lifetime of all mine infrastructure and long term activities such as reclamation are based on the 'mine_lifetime' variable, which is set to 25 years, representing the time the mine area is occupied and run by the represents the time of active mi ning and processing at the mine, and is set by default to 20 yrs. leach pad and carrying capacity and are used for leach pad capital estima tions; (2) related to the mine vehicle models; (3) the ore grade at Yanacocha (Au_ore_grade); (4) the percent of process water treated with reverse osmosis (per_RO_treat); and (5) the way that emergy of labor is included. Parameters are given in Table B 24 Uncertainty The inventory estimates were complemented with uncertainty ranges for direct inputs to the nine primary unit processes. For these inputs, uncertainty range was estimated using the same model specifi ed for the Ecoinvent v2.0 database (Frischknecht et al., 2007). This model assumes inventory data fit a log normal distribution, and that uncertainty can be estimated according to six factors: reliability, completeness, temporal correlation, geographic cor relation, technological correlation, and sample size. The uncertainty is reported as the square of the geometric standard 2 Uncertainty estimates are presented in Table B 25 Model parameters relat ed to lifetime of operations were also assigned ranges. Parameters for mine infrastructure, transport distances, and mine vehicle models were estimated with the Ecoinvent method. For processes based on Ecoinvent data, uncertainty data was perpetuated from Ecoinvent processes. Emergy Conversions All system processes consi sted solely of an These processes served as conversion factors b etween inventory units and emergy values (e.g. 1.1 E+05 sej per J of refined oil), commonly called unit emergy values (UEVs). The UEVs were applied in order to calculate total environmental contribution as energy in sunlight equivalents. Sources for emergy values per unit input were based on previous emergy evaluations of an identical or similar product. Like inventory values, UEVs were assigned an error range, due to uncertainty in the equivalence of the product, uncertainty in processes in nature, or due to methodological differences in emergy calculations. A log normal distri bution is assumed for the UEVs. Discussion This inventory may be directly compared with an existing process in the silver production, a (henceforth ) and its accompanying description (Classen et al. 2007) which is also based on production at the Yanacocha mine.

PAGE 145

145 This study reports a total production of 9.43E+07 g of gold in dor while the process assumes 1.03E+08 g gold in dor. In the process, the inventory data has already been allocated between gold and silver in the dor. This process assumes an additional inputs for separating the gold from the silver in the dor. In this study, the inventory data has not been pre allocated between gold and silver. The structure of this inventory is much more elaborate than that of the process in Ecoinvent. The Ecoinvent process is essenti ally a system process, where inputs to dor production are all grouped under the aforementioned process. This inventory is based on nine unit processes, each of which have additional unit processes contributing to them. The process does n ot consider any inputs into deposit formation, or exploration. Mine infrastructure in the Ecoinvent process is based on a generic Swedish mine. In this study major infrastructure, such as mine building, roads, and processing structures, are based on origi nal analysis of the mine site. The remaining infrastructural components, included power delivery and water supply, are based on generic Ecoinvent processes. For extraction, the process does not estimate the contribution of mine vehicles. F or leaching, the process does not include the leach pad and pool architecture or its construction. For processing, the process does not include the leach pad and pool architecture or its construction. In this inventory, r eagents added during processing and water treatment are based on mass balance calculations of the process. This inventory explicitly includes some of the major components of the process, water treatment, and sediment control infrastructure at Yanacocha, w hich are missing from the process. There are other notable differences in the inventories. Land use and transformation are not included as inputs in this study, but are included in the process. Standing biomass from land transformation, however, is included in this inventory. This is only a source side LCI, but the process includes estimates of emissions to air and water. The electricity mix in the process is based on the Brazilian elec tricity mix. In this study a new electricity mix process specific to Peru was created. The assumed mine lifetime presents a significant difference between the inventories, which effects the contribution of all capital goods and infrastructure. The process assumes a mine lifetime of 50 years; this study only 25 years. A comparison of the outputs and direct non durable inputs to mining in reference to output of 1 g of dor is presented in Table B 15

PAGE 146

146 Table B 15 Comparison of this inventory with the equivalent Ecoinvent process No Item this inventory 'Dore, at Yanacocha' Ecoinvent v2.0 Unit Total production 1 Gold 9.43E+04 1.03E+05 kg 2 Silver 1 .23E+05 3.67E+04 kg 3 Dore 2.17E+05 1.40E+05 kg Rel. to dore production 100% 64% Direct non durable inputs to 1 g of dore 4 Electricity 6.77 12.3 MJ 5 Diesel 18.4 47.7 MJ 6 Sodium Cyanide 30.8 42.9 MJ 7 Lime 0.55 1.17 g 8 Sodium hydroxide 0 52.6 g 9 Activated carbon 6.73 17.1 g 10 Zinc 0.873 3.33 g 11 Sulfuric acid 6.74 7.67 g 12 Hydrochloric acid 6.75 0 g 13 Transport, truck 0.352 1.92 tkm 14 Explosives 0.032 0.416 kg 15 Water 0.022 0.016 m3 16 Lead acetate 2.05 0 g 17 Chlorine 0 .203 0 kg 18 Sodium hydrosulfide 0.378 0 g 19 Iron chloride 6.430 0 g 20 Polyacrylamide 7.38 0 g Notes 4.61E (=1/99.4 % allocation to gold) were compared here as each represent 1 g of dor. Post dor Item references (format: this inventory; Ecoinvent) 4 'Electricity, at powerplant, Peru'; 'Electricity Mix /BR' from Ecoinvent 5 Oil, refined, at Yanacocha'; 'Diesel, burned in building machine /GLO U' 6 'Sodium cyanide, at Yanacocha'; 'Sodium cyanide, at plant/RER U' 7 'Limestone, loose and hydrated, at Yanacocha'; 'Lime, milled, packed, at plant' 8 NA; 'Sodium hydroxide, 50% i n H2O, production mix, at plant/RER U' 9 'Activated carbon'; 'Charcoal, at plant/GLO U' 10 'Zinc, geologic emergy'; 'Zinc, primary, at regional storage/RER U' 11 'Sulfuric acid, 98%, emergy w/out L&S', 'Sulphuric acid, liquid, at plant/RER U' 12 NA, Hydrochloric acid, liquid, at plant/RER U' 13 'Transport, truck, Peru';'Transport, lorry >16t, fleet average/RER U' 14 'Explosives (ANFO), at Yanacocha'; 'Blasting/RER U' 15 'Process Water, Yanacocha'; 'Water, river' + 'Water, well, in ground' 16 'L ead Acetate'; NA 17 'Chlorine, at Yanacocha'; NA

PAGE 147

147 18 'Sodium hydrosulfide, 100%'; NA 19 'Iron chloride'; NA 20 'Polyacrylamide'; NA Due to the difference in output one would expect the values in the process to be 1.58 tim es greater than those in this inventory, but there are still discrepancies bey ond this difference. Electricity, diesel, lime, activated carbon, zinc, truck transport and explosives are all greater in the Ecoinvent inventory than expected. Sodium cyanide, sulfuric acid, and water use are less than the expected difference. Appendix Table B 16 No. Process Unit No. Process Unit 1 Acid Water Treatment P lant, Yanacocha p 83 Mercury, in ground, geologic emergy g 2 Acid Water Treatment, Yanacocha g 84 Merrill Crowe plants, Yanacocha p 3 Acid,Yanacocha, unaccounting for g 85 Merrill Crowe process, Yanacocha g 4 Activated carbon kg 86 Mine infrastructure, Yanacocha p 5 Aircraft, long haul p 87 Mining shovel, Yanacocha hr 6 Airport p 88 Natural gas, emergy w/out labor & services J 7 Aluminum ingot, emergy w/out labor & services g 89 Oil, crude, emergy w/out labor & services J 8 Ammonium nitrate, emergy w /out labor & services g 90 Oil, refined, at Yanacocha J 9 Ammonium, emergy w/out labor and services g 91 Oil, refined, emergy wout/labor & services J 10 Antifreeze g 92 Operation, aircraft, freight, intercontinental tkm 11 Azufre Dam, Yanacocha p 93 Ope ration, maintenance, airport p 12 Bitumen, emergy w/out labor and services g 94 Operation, maintenance, port p 13 Brass, emergy w/out labor & services g 95 Operation, transoceanic freight ship tkm 14 Brick, emergy w/out labor and services g 96 Operation transoceanic tanker tkm 15 Bronze, emergy w/out labor & services g 97 Paint, emergy w/out labor and services g 16 Building, hall, steel m2 98 Pesticide, orthophosphate, emergy w/out labor and services g 17 Cement, emergy w/out labor and services g 99 Pig iron, emergy w/out labor and services g 18 Chlorine, at Yanacocha kg 100 Polyacrylamide g 19 Chlorine, emergy w/out labor and services kg 101 Polybutadeine rubber, emergy w/out labor & services g 20 CIC plant, Yanacocha p 102 Polystyrene, emergy w/o ut labor and services g 21 CIC process solution, Yanacocha g 103 Polyurethane g 22 Clay, in ground, geologic emergy g 104 Port Facilities p 23 Concrete, at Yanacocha g 105 Primary steel, emergy wout/labor & services g 24 Concrete, emergy w/out labor an d services g 106 Process water, at Yanacocha g 25 Conventional Process Water Treatment Plant, Yanacocha p 107 Processing without smelting, Yanacocha year 26 Conventional Process Water Treatment, Yanacocha g 108 Processing, Yanacocha year 27 Copper, emer gy w/out labor & services g 109 Pump station p

PAGE 148

148 28 Diamond drill bit p 110 PVC, emergy w/out labor and services g 29 Diamond exploration drill, Yanacocha hr 111 Quicklime, emergy w/out labor and services g 30 Diamond, in ground, geologic emergy g 112 Rea r dump truck, at Yanacocha hr 31 Dor from Yanacocha PE, at CH g 113 Reclamation, Yanacocha kg 32 Dor, at Yanacocha g 114 Recycled leach solution g 33 Drill rig, Yanacocha hr 115 Reinforced concrete, at Yanacocha m3 34 Dust control, Yanacocha year 116 Rejo Dam, Yanacocha p 35 Electricity from coal, emergy w/out labor and services J 117 Retort process, Yanacocha g 36 Electricity from hydro, emergy w/out labor and services J 118 Reverse Osmosis Process Water Treatment, Yanacocha g 37 Electricity from natural gas, emergy w/out labor & services J 119 RO membrane p 38 Electricity from nuclear, emergy w/out labor and services J 120 RO System p 39 Electricity from oil, emergy w/out labor and services J 121 Road construction, Peru kmy 40 Electricity, at p owerplant, Peru J 122 Road operation, Peru kmy 41 Electricity, at powerplant, USA J 123 Rock wool, emergy w/out labor and services g 42 Emergy in dollar, Peru, 2004 USD 124 Salt, NaCl 100%, emergy w/labor and services g 43 Ethylene propylene rubber (EBR ), emergy w/out labor and services g 125 Sand, in ground, geologic emergy g 44 Exploration, Yanacocha year 126 Scraper, Yanacocha' hr 45 Explosives (ANFO), at Yanacocha kg 127 Sediment and dust control, Yanacocha year 46 Extraction, Yanacocha kg 128 Sed iment control structures, Yanacocha p 47 Fill material, Yanacocha g 129 Serpentine, Yanacocha p 48 Generic inorganic acid, 100%, emergy w/out labor and services g 130 Service Road, Yanacocha km 49 Generic organic chemical, emergy w/out labor and service s g 131 Silt, in ground, geologic emergy g 50 Geomembrane, HPDE, 1.5mm thickness m2 132 Silver in dor, at Yanacocha g 51 Geomembrane, HPDE, 2mm thickness m2 133 Silver, in ground, at Yanacocha, geologic emergy g 52 Geomembrane, LLPDE, 1mm thickness m2 134 Smelters, Yanacocha p 53 Geomembrane, LLPDE, 2mm thickness m2 135 Smelting, Yanacocha g 54 Geotextile, 8 oz. sq.y d 136 Sodium cyanide, at Yanacocha kg 55 Glass, emergy w/out labor and services g 137 Sodium hydrosulfide, 100% kg 56 Gold in dor, at Yanacocha g 138 Sodium hydroxide, 100%, at Yanacocha g 57 Gold, in ground, at Yanacocha, geologic emergy p 139 Sodium hydroxide, 100%, emergy wout/labor and services g 58 Grande Dam, Yanacocha g 140 Stacker, Yanacocha hr 59 Gravel, crushed and washed, P eru g 141 Standing biomass before mining, Yanacocha m2 60 Ground water, emergy km 142 Standing biomass, tropical savannah, emergy g 61 Hauling Road, Yanacocha m 143 Steel Pipe, 36" dia., at Yanacocha ft 62 HDPE Pipe, 40" dia. g 144 Storage tank, steel g 63 HDPE, emergy w/out labor & services kg 145 Sulfuric acid, 98%, emergy w/out labor g

PAGE 149

149 and services 64 Heavy Vehicle my 146 Sulphur hexaflouride g 65 Highway, provincial g 147 Surface water, emergy g 66 Hydrochloric acid, 100%, emergy w/out labor and services g 148 Tetrafluoroethylene g 67 Hydrogen cyanide g 149 Tilting Furnace p 68 Hydrogen sulfide, emergy w/out L&S g 150 Transmission network, electricity, medium voltage km 69 Iron ore, emergy w/out labor and services g 151 Transoceanic freight shi p p 70 Iron(III) Chloride J 152 Transoceanic tanker p 71 Labor, Peru, emergy p 153 Transport of Dore, Yanacocha to Switzerland g 72 Labor, total, Yanacocha m2 154 Transport truck, operation, Peru km 73 Leach Pad, Yanacocha m2 155 Transport, aircraft, f reight, intercontinental tkm 74 Leach Pool, Yanacocha g 156 Transport, aircraft, freight, Peru tkm 75 Leaching, Yanacocha g 157 Transport, transoceanic freight ship tkm 76 Lead acetate g 158 Transport, transoceanic tanker tkm 77 Lead, in ground, geolog ic emergy kg 159 Transport, truck, Peru tkm 78 Lime, loose and hydrated, at Yanacocha g 160 Transport, truck, USA, emergy w/out labor and services tkm 79 Limestone, in ground, geologic emergy g 161 Water supply network km 80 Lumber, emergy w/out labor a nd services g 162 Water Treatment, Yanacocha year 81 Mercury, at Yanacocha g 163 Wood preservative g 82 Mercury, in ground, at Yanacocha, geologic emergy g 164 Zinc, in ground, geologic emergy g Table B 17 Mine hauling road parameters based on Hartman (1992) Course Thickness (m) Material Cross sectional area (m2) Surface 0.1 Gravel 2.5 Base 0.1 Clay sand silt 2.5 Subbase 0.5 Clay sand silt 12.5 Table B 18 Mine service road parameters, based on Hartman (1992) Course Thickness (m) Material Cross sectional area (m2) Surface 0.1 Gravel 2.5 Base 0.1 Clay sand silt 2.5 Table B 19 Mining equations Equat ion Reference 1 Shovel and stacker loading production, loose m 3 /hr = 3600(Bucket capacity, loose m 3 )(efficiency)(fill factor)(propel time factor)/(load cycle time, seconds) SME, Equation 12.21 Total shovel and stacker use, hrs = (m 3 /mine/yr/ loose m 3 /hr) NA Scraper load, m 3 = (capacity, m 3 )(swell factor, ratio of bank m 3 to loose m 3 ) SME, Equation 12.9

PAGE 150

150 Scraper travel time, min = (distance to soil storage, m)/(speed, km/hr)(16.7 m h/km min) SME, Equation 12.18 Scraper cycle time, min = (load time,min)+( travel time,min*2)+ (spread time,min) SME, Equation 12.19 Scraper production, m 3 /hr= (60)(bucket capacity, m 3 )(operating efficiency)/cycle time (hrs) SME, Equation 12.21 Scraper use, hrs (Topsoil to be moved, annualized)/(scraper production) NA Dump tr uck spot and load time, min = (spot time, min)+(passes 1)(loading cycle time) SME, Equation 12.15 Travel time to dump point, min = (Distance,m)/(speed, km/h)(16.7 m h/km min) SME, Equation 12.18 Dump truck cycle time, min= (load time) + (travel time) + ( travel time) + (dump time) SME, Equation 12.19 Dump truck production, m3/hr =(60)(haulage units)(load, bank m 3 )(efficiency)/(cycle time,min) SME, Equation 12.21 Dump truck use, hrs = (ore mined, m 3 /yr/ haulage production, m 3 /hr) NA Drill rig use, hrs/y r = (holes/layer)(layers/year)(digging, hrs/hole+travel time, hrs/hole) NA 1 All references with SME refer to the SME Handbook (Lowrie 2002) Table B 20 Mine vehicle data Type Manufacturer/Model Weight (kg) 1 Lifetime (hrs) 2 Rear Dump Truck CAT 793D 166866 30000 Stacker CAT 325D w/boom 29240 14000 Scraper CAT 651E 62000 14000 Mining shovel Hitachi EX5500 518000 90000 Drill rig Atlas Co pco Simba 1250 11830 14000 1 From manufacturer specifications 2 Estimated from (Lowrie 2002)

PAGE 151

151 Table B 21 Mass b alance of leaching, processing, and water treatment.

PAGE 152

152

PAGE 153

153

PAGE 154

154 Table B 22 Inventory of P eruvian r oad t ransport No. Item Flow Unit 1 Trucks 4.44E+10 g Road Construction 2 Concrete 6.00E+09 g 3 Bit umen 1.75E+10 g 4 Gravel 2.42E+11 g 5 Electricity 4.92E+11 J 6 Diesel 1.18E+12 J Road operation 7 Electricity 7.31E+09 J 8 Paint 6.04E+03 g 9 Herbicide 3.37E+02 g Transport 10 Diesel consumption 8.90E+15 J Product 11 Annual yield of tru cks 1.50E+09 ton km NOTES Input references from Spielman et al. (2004) Trucks 1 (Class 8 weight lb)(class 8 trucks)*(Class 6 weight lb)(class 6 trucks)*( 454 g/lb) / (10 yr lifetime) 4.44E+10 g Truck weights from Buranakarn (1998) UEV from heavy mine vehicle model Highway construction Demand by trucks of infrastructure creation Good transport percent road wear 0.424 Based on Swiss situation. Table 5 117. road length=(length of road network, km)(14.4% paved) Highway km 11351 (Economic Commission of Latin American and the Carribbean 2006) Improved unpaved km 18634 Concrete kg/ (m*yr) 37 Bitumen kg/ (m*yr) 15.4 Gravel for highway subbase kg/ (m*yr) 470 Gravel for unpaved road surface kg/ (m*yr) 101.25 Lifetime Concrete yr 70 Bitumen yr 10 Gravel for highway subbase yr 100 Gravel for unpaved road surface yr 10 Standard Equation for road materials (Good transport percent road wear)(material kg/m*yr)(road length km) (1000m/km) (1000g/kg) / (material lifetime yr) 2 Concrete g 6.00E+09 3 Bitumen g 1.75E+10 4 Gravel g 2.42E+11

PAGE 155

155 Electricity for highway constr. MJ/m*yr 98.7 Motorway. Table 5 94. Electricity for unpaved road constr. MJ/m*yr 2.18 2nd class road. Table 5 94. (Good transport percent road wear)(energy MJ/m*yr)(road length km) (1000m/km) (1 E+6 J/MJ) 5 Electricity for construction J 4.92E+11 Diesel for highway construction MJ/m*yr 192 Motorway. Table 5 94. Diesel for unpaved road construction MJ/m*yr 33 2nd class road. Table 5 94. (Good transport percent roa d wear)(energy MJ/m*yr)(road length km) (1000m/km) (1 E+6 J/MJ) 6 Diesel J 1.18E+12 Operation Demand by trucks of infrastructure operation Good transport percent road use 0.103 Based on Swiss situation. Table 5 117. Electri city for highway operation KWH/m*yr 0.67 Motorway. Table 5 101. Electricity for unpaved road operation KWH/m*yr 3.4 2nd class road. Table 5 101. (Good transport percent road use)(electricity use KWH/m*yr)(road length km) (3600000 J/KWH) 7 Electrici ty for operation J 7.31E+09 Paint for highway operation kg/m*yr 0.00517 (Good transport percent road use)(paint usekg/m*yr)(road length km) (1000 kg/g) 8 Paint g 6.04E+03 Herbicide for highway operation kg/m*yr 2.88E 04 (Good tr ansport percent road use)(herbicide usekg/m*yr)(road length km) (1000 kg/g) 9 Herbicide g 3.37E+02 UEV for orthophosphate from Nepal (2008) Transport Mid size truck fuel economy diesel kg/vkm 0.25 (Kodjak 2004) Tractor trailer truck fuel economy diesel kg/vkm 0.37 (Kodjak 2004) Mid size truck vkm/ton km vkm/ton km 0.62 Lorry 3.5 16t. Table 5 119. Tractor trailer vkm/ton km vkm/ton km 0.12 Lorry >16t. Table 5 119. Tractor trailer ton km percentage 0.88 Table 5 119. Mid size truck ton km ton km 1.75E+08 Lorry >16t. Table 5 119. Tractor trailer ton km ton km 1.32E+09 Lorry 3.5 16t. Table 5 119. Truck fuel u se = (Truck ton km)(ton km/vkm)(diesel kg/vkm) (4.36 E+07 J/kg) Mid size truck fuel use J 1.20E+15 1.08E+08 Tractor trailer fuel use J 2.53E+15 1.56E+08 10 Total diesel fuel use J 3.73E+15 2.64E+08 11 No. trucks= total vehicles* portion o f trucks in import data (Economic Commission of Latin American and the Carribbean 2006; United Nations 2008) (5.04 E+04 Ton km/truck/yr USA)(.5 Peru/US productivity)(142872 trucks in Peru fleet) Annu al truck transport ton km 1.50E+09

PAGE 156

156 Table B 23 Assumed origins and transport distances for inputs to mining. Input Mass (kg) Assumed Origin Data Source Sea Distance (km) Road Distance (km) Refined Oil 9.75E+07 Imported 2.34E+07 Balao, Ecuador 1 1148 250 Domestic 7.41E+07 Chimbote 1 0 250 Lime 7.36E+07 China Linda 2 0 12 Chlorine 4.41E+07 Lima 3 0 850 Caustic soda 2.52E+07 Lima 1 0 850 Explosives (ANFO) 7.00E+06 Lima 3 0 850 Sodium cyanide 6.6 9E+06 US 3 5900 850 Concrete 4.68E+06 China Linda 2 0 12 Steel pipe 2.97E+06 US 3 5900 850 Other 1.27E+07 Local NA 0 0 TOTAL 2.74E+08 Notes Only inputs comprising 1% of total mass input are listed. Data Sources 1. (Instituto Nacional Estadistica y Informacion 2006) ) 2. (Buenaventura Mining Company Inc. 2006) 3. (United Nations 2008)

PAGE 157

157 Table B 24 System level par ameters Parameter Default Value 2 geo Units and Comments include_geo 1 NA 1=Include geologic emergy of gold ore; 0=do not include include_clay_em 0 NA 1=Include geologic emergy of clay for roads and leach pads; 0=do not include include_grav_em 0 NA 1=Include geologic emergy of g ravel for roads and leach pads; 0=do not include mine_lifetime 25 1.3 yrs. 1993 2018. End date estimate from http://www.newmont.com/csr05/protest_yanacocha/1.html process_lifetim 20 1.3 yrs. Avg process lifetime for all processing facilities. Less than m ine_lifetime waste_to_reclam 1 NA Fraction of waste rock used to refill pits. 1=All waste rock used for backfilling lima_yan_distan 850 1.1 km. ( 1.05,1,1,1.01,1,NA ) Au_output 3327500 1 oz/yr, Buenaventura 2006 Hg_output 5.5 1 short tons/month, Ne wmont 2006a veh_add_steel 0.4 1.2 Additional frac tion steel for heavy vehicles. (1.2,1,1.03,1,1,NA) veh_add_rubber 0.07 1.2 Additional fraction rubber for heavy vehicles. This is substituted with steel for track vehicles. ( 1.2,1,1.03,1,1,NA ) veh_wei ght 15500 1.2 kg. Based on 40ton Lorry (Ecoinvent). (1.2,1,1.03,1,1,NA) kgore_topadarea 198891 1.5 kg/m2. Based on avg of 5 leach pad areas and capacities. Actual SD*2 kgoretopoolarea 4057275 1.5 kg/m2. Based on avg of 5 leach pad areas and capacities. Actual SD*2 per_RO_treat 0.4 1 Fraction of excess water treatment using reverse osmosis tot_excess_wat 1.2E+13 1 G Au_ore_grade 0.028 1 oz/ton labor_use_J 0 NA 1 = include labor by using sej/J emergy in labor. See emergy in labor process. 0= Do not use labor_use_dol 0 NA 1 = include labor using emergy/$ ratio. 0=do not include. sea_transport 5900 1.1 km. Los Angeles to Lima sea distance. Used for generic sea transport distance. ( 1.05,1,1,1.01,1,NA )

PAGE 158

158 Table B 25 Uncertainty estimates for inventory data using Ecoinvent method (Frischknecht and Jungbluth 2007) Unit Process(es) Input or Variable reliability completene ss temporal correlation geographic correlation other tech correlation sample size Uncertainty score Exploration, Extraction, Reclamation Oil, refined 1.2 1 1 1.1 1. 2 NA 1.3 Exploration, Extraction, Sed. & Dust control Water for process 1.2 1 1 1 1 NA 1.2 Extraction, Reclamation, Mine Infrastructure Heavy Vehicle Use 1.2 1 1.1 1.1 1 NA 1.3 Mine infrastructure Infrastructure based on visual es timates 1.0 5 1 1 1 1. 5 1.2 1.5 Extraction Explosives 1 1 1 1 1 NA 1.0 Leaching CN 1 1 1 1 1 NA 1.0 Processing Natural gas 1.2 1 1 1.1 1. 2 NA 1.3 Water treament, Reclamation Chemicals for water treatment (CaOH, Cl, FeCl3, PAM, H2SO4); and reclamation (C aOH) 1.2 1 1 1 1 NA 1.2 Variables Distance variables 1.0 5 1 1 1.0 1 1 NA 1.1 Variables Mine vehicle model variables 1.2 1 1.03 1 1 NA 1.2

PAGE 159

159 C APPENDIX C SUPPLEMENT TO CHAPTE R 3 : R CODE FOR STOCHASTIC UNCERTAIN TY MODELS The following sections contains code for stochastic uncertainty models for both the formula and table form uncertainty models, as described in chapter 2. This code can be run in R statistical software. Code for Formula UEV Uncertainty Estimation #A script for a Monte C arlo simulations of formula type unit emergy values to estimate uncertainty #Author: Wes Ingwersen, wwi@ufl.edu ##Do a Monte Carlo simulation for a formula UEV calculation, with uncertainty expressed for all variables #################Instructions#################### #Prepar e a tab separated table of items in your emergy table in the form of: #variable_name average standard deviation # the following is a sample for the lead UEV this can be copied and pasted into a new .txt file crust_conc_ppm 15 1.41 ore_grad_frac 0.06 0. 03 crust_turn_cm_yr 1 2.88E 03 6.77E 04 den_crust_g_cc 1 2.72 0.04 crustal area_sqcm 1.48E+18 2.1E+16 # This file has to be saved at C: \ RData \ UEV \ directory unless the path name is changed in the script for the script to function. ##################Import Data##################### #Input data in the form of a tab delimited txt file with var name, mean, sd, on 1 line #Uncomment lines related to UEV of interest #To see the table that translates into this format, see Table 3 in Ingwersen (2009) #UEV for lead #fname < "C: \ \ RData \ \ UEV \ \ lead.txt" #item < "lead" #fractions < c(1,2) #den_unit < "g" #mag < 12 #Order of mag of deterministic mean UEV #UEV of iron #fname < "C: \ \ RData \ \ UEV \ \ iron.txt" #item < "iron" #fractions < c(1,2) #den_unit < "g" #mag < 10 #Order of mag of deterministic mean UEV #UEV of oil #fname < "C: \ \ RData \ \ UEV \ \ oil.txt" #item < "oil" #fractions < c(2,3,4,5) #den_unit < "J" #mag < 5 #Order of mag of deterministic mean UEV

PAGE 160

160 #UEV of groundwater #fname < "C: \ \ RData \ \ UEV \ \ gw.txt" #item < "gw" #Groundwater #fractions < c(2) #den_unit < "g" #mag < 5 #Order of mag of deterministic mean UEV #UEV of labor #fname < "C: \ \ RData \ \ UEV \ \ labor.txt" #item < "labor" #fractions < c() #den_unit < "J" #mag < 6 #Order of mag o f deterministic mean UEV #Loads the text file, stores it in a data frame cols < c("var","mu","sig") df < read.delim(fname,header=FALSE,strip.white=TRUE,row.names=1, col.names=cols) df #Verify that the data loaded properly #########Set Initial Paramete rs#################### #Run the following code ##Number of MC results n < 100 #Number of MC's to run from which to calculate the uncertainty j < 100 ##Case 1: Assume variables are normally distributed ##Case 2: Assume varibales are log normally distri buted case < 2 #Note Model only stable using case 2 ###########Functions for MC just load on first use##################### ##Function to return logforms of mean and standard dev returnlogforms < function(mu,sig) { lamda < 1+(sig/mu)^2 logformsi g < sqrt(log(lamda)) logformmu < log(mu) 0.5*logformsig return(c(logformmu,logformsig)) } #n will also be the number of replicates of each variable in the model chosen #Make a matrix to hold n of each model parameter) make_params < function() { mc_ vars < matrix(nrow=nrow(df),ncol=n) for (x in 1:nrow(df)) { #Put the mean and sd in a matrix m < df[[1]][x] s < df[[2]][x] if(case==2)

PAGE 161

161 { logforms < returnlogforms(m,s) mc_vars[x,] < rlnorm(n,meanlog=logforms[1],sdlog=logforms[2]) } el se { mc_vars[x,] < rnorm(n,mean=m,sd=s) } } return(mc_vars) } clean < function(parameters) { a < 0 b < 0 for (a in 1:length(fractions)) { ind < fractions[a] for (b in 1:n) { if ((parameters[ind,b]<=0 || parameters[ind,b]>=1) & & !is.na(parameters[1,b])) { parameters[,b] < NA } } } } ######################Unit emergy value model#################################### #Run the desired model, or enter your own model #Model for land cycle is #ER < 2.ore_grad_frac/( 1.crust_conc_ppm/1E6) #ER #Land_UEV < 15.83E24/(3.crust_turn_cm_yr 1)*(4.den_crust_g_cc 1)*(5.crust_area_sqcm) #Mineral_UEV < ER*Land_UEV #Model for water = UEVwater, sej/g = (global emergy base, 15.83E24 sej/yr)/Annual Flux, g/yr) #turnover time = (Glo bal groundwater resevoir)/ #(Global precip on land, mm/day)(365days/yr)/(1E6 mm/km)*(global land area (km2)*(infiltration rate) #Function to do the model calculation mod < function (mat) { res_vec < c() #Result vector for (i in 1:n) { UEV < NA i f ((item=="lead" || item=="iron")&& !is.na(mat[1,i])) { pred < 2.64 # Predicted sq_sig_geo for lead_UEV pred < 2.03 # Predicted for iron_UEV #Formula for Mineral UEV calc if (item=="lead") { er < mat[2,i]/(mat[1,i]/1E6)#when conc is i n ppm } else { er < mat[2,i]/(mat[1,i]) #when conc is a frac }

PAGE 162

162 land_UEV < 15.83E24/(mat[3,i]*mat[4,i]*mat[5,i]) UEV < er*land_UEV #For mineral calcs } if (item=="oil" && !is.na(mat[1,i])) { #Formula for oil #Deterministic solution #mat < df #i< 1 ep_c < (mat[1,i]*1.78E4)/mat[2,i] ek < ep_c/mat[3,i] UEV < (1.68*ek*mat[5,i])/(mat[4,i]*4.19E4) #UEV #If UEV is negative take absolute value } if (item=="gw" && !is.na(mat[1,i])) { #Formula for g roundwater #Deterministic solution #mat < df #i< 1 global_land_area < mat[3,i] #km2 precip < mat[1,i] #mm/yr infiltration < mat[2,i] annual_flux < ((precip/1E6)*global_land_area*infiltration*1E12*1000) UEV < 15.83E24/annu al_flux } if (item=="labor") { #Formula for labor #(Global emergy use per yr/global population)/(Daily per capita calorie intake*365 days* 4184J/kcal) #mat < df #i< 1 UEV< (1.61E26/mat[1,i])/(mat[2,i]*365*4184) } if (UEV<0) { UEV < NA } res_vec[i] < UEV } return(res_vec) } ##################RUN SIMULATION#################### #Hightlight and run the following code #Run the Monte Carlo, j times mc < c() #Store the results of one Monte Carlo here Quot_upper_by_med < c() #St ore the results of the upper limimit divided by the median for each MC upperlims < c() lowerlims < c() medians < c() means < c() sds < c() all_mc < matrix(nrow=j,ncol=n)#Store each mc result in a row for graphing later for (a in 1:j) { params < ma ke_params()

PAGE 163

163 if (length(fractions)) ( clean(params) ) #Removes values <0 or >1 for fractions mc < mod(params) all_mc[a,] < mc med < median(mc,na.rm=TRUE) std < sd(mc,na.rm=TRUE) CIs < format(quantile(mc, probs = c(0.025,0.975),na.rm=TRUE, digits=3 scientific=TRUE)) upperlim < as.double(as.vector(CIs["97.5%"])) lowerlim < as.double(as.vector(CIs["2.5%"])) up < upperlim/med #low < med/lowerlim upperlims[a] < upperlim lowerlims[a] < lowerlim medians[a] < med medians[a] < med sds[a] < std Quot_upper_by_med[a] < up } #Take averages of medians of distributions and geometric variances med < mean(medians) geo_var < mean(Quot_upper_by_med) lower_bound < mean(lowerlims) upper_bound < mean(upperlims) #Print the results c('Median=',me d) c('Geometric variance=',geo_var) c('Lower bound',lower_bound) c('Upper bound', upper_bound) Code for Table form UEV Uncertainty Estimation #A script for a Monte carlo simulations of table form unit emergy values to estimate uncertainty #Author: Wes Ingw ersen, wwi@ufl.edu ##Do a Monte Carlo simulation for a table form UEV calculation, with uncertainty expressed for all variables #################Instructions#################### #Input data in the form of a tab delimited txt file with var name flow_quani ty_mean flow_quanity_geo_var UEV_mean UEV_geo_var # the following is a sample for the sulfuric acid UEV this can be copied and pasted into a new .txt file Secondary_sulfur 214 1.32 5200000000 3.59 Diesel 3410 1.34 121000 3.59 Electricity 63000 1.34 371 000 2.77 Water 241000 1.23 189572.5914 1.95 # This file has to be saved at C: \ RData \ UEV \ directory unless the path name is changed in the script for the script to function. ##################Import Data##################### #UEV for electricity #fname < "C: \ \ RData \ \ UEV \ \ electricity.txt" #item < "electricity" #den < 3.6E6 #Joules of electricity This is the denominator for the UEV calculation #den_unit < "J" #mag < 5 #Order of mag of deterministic mean UEV

PAGE 164

164 #UEV for sulfuric acid fname < "C: \ \ RData \ \ UEV \ \ sulfuric_acid.txt" item < "sulfuric acid" den < 1000 #g of H2SO4 This is the denominator for the UEV calculation den_unit < "g" mag < 7 #Order of mag of deterministic mean UEV cols < c("param","value","k_value","UEV","k_UEV") df < read.deli m(fname,header=FALSE,strip.white=TRUE,row.names=1, col.names=cols) df #########Set Initial Parameters#################### ##Number of MC results n < 100 #Number of MC's to run from which to calculate the uncertainty j < 100 ##Case 2: Assume varibale s are log normally distributed #Now it only works for log normally distributed variables case < 2 ###########Functions for MC just load on first use##################### ##Function to return logforms of mean and standard dev #Only used for formula UEV s copied here for reference returnlogforms < function(mu,sig) { lamda < 1+(sig/mu)^2 logformsig < sqrt(log(lamda)) #Source: Wikipedia, "Lognormal distribution" logformmu < log(mu) 0.5*logformsig #Wikipedia return(c(logformmu,logformsig)) } ##Fu nction to return logforms of with determininstic mean and k value (ref: Slob (1994)) returnlogforms_withKvalue < function(mu,k) { logformsig < sqrt((log(k)/1.96)^2) logformmu < log(mu) 0.5*logformsig logformsig logformmu return(c(logformmu,logform sig)) } #Make a matrix to hold n of each model parameter) make_params < function() { #Create a matrix to store n random values(3rd dimension) of both the value and UEV (2nd dimension) of each variable (1st dim) mc_vars < mc_vars < array(NA,dim=c(nrow(d f),2,n)) for (x in 1:nrow(df)) { #Gets the values from the input matrix val < df[[1]][x] k_val < df[[2]][x] uev < df[[3]][x]

PAGE 165

165 k_uev < df[[4]][x] #Call the script to get the logforms of mu and sigma val_logforms < returnlogforms_withKvalue(va l,k_val) uev_logforms < returnlogforms_withKvalue(uev,k_uev) #Use the log forms in a lognormal distribution random generator function mc_vars[x,1,] < rlnorm(n,meanlog=val_logforms[1],sdlog=val_logforms[2]) mc_vars[x,2,] < rlnorm(n,meanlog=uev_logfo rms[1],sdlog=uev_logforms[2]) } return(mc_vars) } #Function to do the model calculation mod < function (mat) { res_vec < c() #Result vector for (i in 1:n) { UEV < NA #Calculate the UEV for that random set of params em < 0 for (r in 1 :nrow(df)) { #Multiply the value and UEV var_em < mat[r,1,i]*mat[r,2,i] #Add the emergy to the sum em < em + var_em } UEV < em/den #UEV is sum of emergy divided by denominator (usu. J or g) res_vec[i] < UEV } return(res_vec) } # #################RUN SIMULATION#################### #Run the Monte Carlo, j times mc < c() #Store the results of one Monte Carlo here Quot_upper_by_med < c() #Store the results of the upper limimit divided by the median for each MC upperlims < c() low erlims < c() medians < c() means < c() sds < c() all_mc < matrix(nrow=j,ncol=n)#Store each mc result in a row for graphing later for (a in 1:j) { params < make_params() mc < mod(params) all_mc[a,] < mc med < median(mc,na.rm=TRUE) m < mean( mc,na.rm=TRUE) std < sd(mc,na.rm=TRUE) CIs < format(quantile(mc, probs = c(0.025,0.975),na.rm=TRUE, digits=3, scientific=TRUE))

PAGE 166

166 upperlim < as.double(as.vector(CIs["97.5%"])) lowerlim < as.double(as.vector(CIs["2.5%"])) up_by_med < upperlim/med u pperlims[a] < upperlim lowerlims[a] < lowerlim medians[a] < med means[a] < m sds[a] < std Quot_upper_by_med[a] < up_by_med } #Take averages of medians of distributions and geometric variances med < mean(medians) geo_var < mean(Quot_upper_by_ med) lower_bound < mean(lowerlims) upper_bound < mean(upperlims) #Print the results c('Median=',med) c('Geometric variance=',geo_var) c('Lower bound',lower_bound) c('Upper bound', upper_bound)

PAGE 167

167 D APPENDIX D SUPPLEMENT TO CHAPTE R 4 : ADDITIONAL TABLES AND F IGURES Table D 1 Inputs to one kg pineapple at the packing facility. Category Input name Country Sr c Unit Amount SD Active Ing. Energy Diesel, at regional storage RER e kg 7.29E 03 2.97E 03 n/a Petrol, unleaded, at regional storage RER e kg 2.40E 04 2.20E 04 n/a Fertilizer Ammonium nitrate, as N, at regional storehouse RER e kg 1.92E 03 1.08E 03 n/a Boric acid, anhydrous, powder, at plant RER e kg 1.73E 04 1.89E 04 n/a Calcium nitrate, as N, at reg ional storehouse RER e kg 1.72E 04 4.66E 05 n/a Compost, at plant CH e kg 4.33E 03 2.43E 03 n/a Dolomite, at plant RER e kg 2.03E 04 4.58E 05 n/a Fosfomax (0,30,0) fertilizer CR o kg 4.51E 04 3.67E 04 n/a Iron sulphate, at plant RER e kg 2.97E 04 2 .45E 04 n/a Kaolin, at plant RER e kg 8.20E 04 6.74E 04 n/a Lime, hydrated, packed, at plant CH e kg 1.63E 03 3.68E 04 n/a Magnesium ammonium nitrate, (22,0,0,0,7) RER o kg 2.11E 03 1.19E 03 n/a Magnesium sulphate, at plant RER e kg 2.03E 03 2.09E 03 n/a NPK (12,24,12) fertilizer RER e kg 1.18E 02 9.63E 03 n/a NPK (18,5,15) fertilizer RER o kg 2.11E 03 1.72E 03 n/a NPK (2,10,10) fertilizer RER o kg 7.93E 05 6.46E 05 n/a Potassium chloride, as K2O, at regional storehouse RER e kg 5.82E 03 4.7 4E 03 n/a Potassium sulphate, as K2O, at regional storehouse RER e kg 4.33E 03 3.52E 03 n/a Single superphosphate, as P2O5, at regional storehouse RER e kg 5.54E 05 4.51E 05 n/a Sugar, from sugarcane, at sugar refinery BR e kg 2.51E 04 5.67E 05 n/a Urea, as N, at regional storehouse RER e kg 3.62E 03 2.04E 03 n/a Zinc monosulphate, ZnSO4.H2O, at plant RER e kg 2.74E 04 7.58E 05 n/a fungicide benzoic compounds, at regional storehouse RER e kg 5.63E 05 3.55E 05 Metalaxil pesticide unspecified, at regional storehouse RER e kg 1.49E 04 9.40E 05 Fosetyl aluminium triazine compounds, at regional storehouse RER e kg 1.20E 06 7.54E 07 Thiazole, 2 (thiocyanatemethylth io)benzo triazine compounds, at regional storehouse RER e kg 6.58E 06 4.15E 06 Tria dimefon growth organophosphorus compounds, at regional storehouse RER e kg 2.58E 05 3.69E 05 Ethephon herbicide diphenylether compounds, at regional storehouse RER e kg 6.58E 06 3.43E 06 Fluazifop p butyl diuron, at regional storehouse RER e kg 1.12E 0 4 5.83E 05 Diuron glyphosate, at regional storehouse RER e kg 3.76E 05 1.96E 05 Glyphosate pesticide unspecified, at regional RER e kg 6.60E 05 3.44E 05 Bromacil

PAGE 168

168 storehouse phenoxy compounds, at regional storehouse RER e kg 1.38E 06 7.21E 07 Quizalof op P triazine compounds, at regional storehouse RER e kg 7.96E 05 4.14E 05 Ametryn insecticid e [thio]carbamate compounds, at regional storehouse RER e kg 3.08E 05 1.60E 05 Carbaryl organophosphorus compounds, at regional storehouse RER e kg 1.24E 04 7 .84E 05 Diazinon nematicid e organophosphorus compounds, at regional storehouse RER e kg 6.80E 05 5.47E 05 Ethoprop Machiner y tractor, production CH e kg 3.13E 04 1.35E 04 Table D 2 Emissions from one k g pineapple at the packing facility. Substance To Amount GV Note Ametryn air 4.90E 06 3.0 from pesticide application. Includes yield and pesticide input uncertainty. Ametryn water 9.87E 06 5.9 Ammonia air 1.55E 07 2.3 from fuel combustion Ammonia a ir 1.10E 04 2.9 volatilized from N fertilizers Benzene air 2.33E 06 2.3 from fuel combustion Benzo(a)pyrene air 2.28E 10 2.3 from fuel combustion Bromacil air 9.62E 06 2.1 from pesticide application. Includes yield and pesticide input uncertainty. Brom acil water 5.42E 06 4.8 Cadmium air 7.53E 11 5.9 from fuel combustion Carbaryl air 5.16E 06 4.3 from pesticide application. Includes yield and pesticide input uncertainty. Carbaryl water 1.78E 07 7.5 Carbon dioxide, fossil air 2.35E 02 2.1 from fue l combustion. Combines uncertainty of diesel input, diesel emission factor, and yield Carbon dioxide, fossil air 6.45E 04 2.8 from urea application Carbon dioxide, land transformation air 1.00E 10 from land use change Carbon monoxide, fossil air 2.05E 04 5.9 from fuel combustion Chromium air 3.77E 10 5.9 from fuel combustion Copper air 1.28E 08 5.9 from fuel combustion Diazinon air 6.01E 06 3.0 from pesticide application. Includes yield and pesticide input uncertainty. Diazinon water 3.60E 07 5.9 Dinitrogen monoxide air 9.06E 07 2.3 from fuel combustion Dinitrogen monoxide air 1.78E 04 2.9 from N fertilizers Diuron air 6.55E 06 1.8 from pesticide application. Includes yield and pesticide input uncertainty. Diuron water 2.20E 05 4.5 Ethephon air 1.51E 05 8.4 from pesticide application. Includes yield and pesticide input uncertainty. Ethephon water 2.27E 07 12.8 Ethoprop air 2.25E 06 6.6 from pesticide application. Includes yield and pesticide input uncertainty. Ethoprop water 1.13E 06 10 .4

PAGE 169

169 Fluazifop p butyl air 1.73E 06 8.4 from pesticide application. Includes yield and pesticide input uncertainty. Fosetyl aluminium water 1.59E 05 2.8 Glyphosate air 2.17E 05 8.4 from pesticide application. Includes yield and pesticide input uncerta inty. Glyphosate water 2.95E 06 12.8 Lead air 3.51E 08 5.9 from fuel combustion Metalaxil water 1.82E 06 2.4 from pesticide application. Includes yield and pesticide input uncertainty. Metalaxil air 4.92E 07 5.1 Methane, fossil air 1.64E 06 2.3 fr om fuel combustion Nickel water 5.27E 10 5.9 from fuel combustion Nitrate air 6.84E 03 3.0 leached from N fertilizers Nitrogen oxides air 2.79E 04 2.3 from fuel combustion Nitrogen oxides water 8.57E 08 2.8 from N fertilizers NMVOC, non methane volati le organic compounds, unspecified origin air 1.87E 05 2.3 from fuel combustion PAH, polycyclic aromatic hydrocarbons air 2.31E 08 3.8 from fuel combustion Paraquat air 4.44E 07 8.4 from pesticide application. Includes yield and pesticide input uncertaint y. Paraquat air 2.17E 06 12.8 Particulates, < 2.5 um air 1.27E 05 3.8 from fuel combustion Phosphate water 1.15E 04 4.3 runoff of P fertilizers Phosphorus air 1.17E 04 18.7 P in eroded soil. Uncertainty includes soil erosion, P content in soil, and y ield uncertainty Quizalofop P water 6.88E 08 8.4 from pesticide application. Includes yield and pesticide input uncertainty. Quizalofop P water 1.48E 07 12.8 Sediment, eroded air 6.28E 02 18.2 estimated with RUSLE2 model. Includes yield and emission u ncertainty Selenium water 7.53E 11 5.9 from fuel combustion Sulfur dioxide water 7.38E 06 2.1 from fuel combustion Triadimefon air 1.16E 06 8.4 from pesticide application. Includes yield and pesticide input uncertainty. Triadimefon air 3.38E 08 12.8 Water air 1.62E+00 1.5 evaporated blue water. Includes yield and emission uncertainty Zinc air 7.53E 09 5.9 from fuel combustion

PAGE 170

170 Table D 3 Emissions estimation s for mineral N in applied fertilizers No Pathway Equation Source 1 Uptake 0.018 dry biomass Su (1968) 2 NH 3 N to air 1 15 % N applied Brentrup and Kusters (2000) 3 N 2 O N to air 1.25 % N applied IPCC 2007, for estimating direct N 2 O emissions 4 NO 2 N to air 0.001 % N 2 O N Nemecek and Kagi (2007) 5 NO 3 N to water 1 residual N in soil Brentrup and Kusters (2000) Item notes 1 Based on percent concentration of N in dry pineapple biomass 5 Assuming exchange ratio (rainfall/field capacity) = 1, all residual N leaches Table D 4 Emissions estimations for miner al P in applied fertilizers Item Pathway Equation Source 1 Uptake 0.18% *biomass Su (1968) 2 P 2 O 5 P to water 2.5% of applied Powers (2007) 3 P in erodible sediment 0.00186 kg P/kg soil Nemecek and Kagi (2007) Table D 5 General assumptions used in the FAO CROPWAT model. CROPWAT Component A ssumption Climate Penman ET based on geographically specific data from LocClim Rain Rainfall from LocClim; calculation with USDA S.C. Method Soil Medium (loam) from CROPWAT database Crop water requirement See Table X Schedule Irrigate at user defined intervals; 70% (default) efficiency Table D 6 Crop water requirement variables for CROPWAT. Parameter Value Source K c init 0.9 Bartholomew (2003) p. 95, for non mulched system K c mid and end 0.4 Bartholo mew (2003) p. 95, for non mulched system Kc init1 0.5 Allen et al. (1998), refers to plastic mulched system Kc mid 0.3 Allen et al. (1998), refers to plastic mulched system Stage initial, days 90 Based on average reported harvest schedule Stage dev elopment, days 180

PAGE 171

171 Stage mid season, days 120 Stage late season, days 180 Yf (all stages) 1.0 Estimated based on crops with similar critical depletion Rooting depth, m 0.45 Smith (1992), p. 61, Critical depletion, p 0.5 Smith (1992), referre d to as fraction of available soil water p. 61 Crop height, m 0.9 Bartholomew (2003), average height Table D 7 RUSLE2 parameters for Pineapple in Costa Rica RUSLE2 Componen t Parameter Value/Setting Notes Introductio n Template ARS Basic Uniform Slope Profile Horiz. overland flow path length, m 16.1 production weighted average Avg. slope steepness, % 2.1 production weighted average Contouring Up and down slope Strips/barriers None Diversion/ter race, sediment basin None Subsurface drainage none Adjust res. burial level Normal Climate How to get erosivity Enter R & choose EI zone R factor, US units 450 for North zone Standard EI Enter half montly EI based on relative intensity of stor m events during the month How determine runoff? based on 10 yr 24 hr ppt 10 yr, 24 hr rain (mm) from FAO Clim Annual precip from FAO Clim Soils Erodibility get from standard nomograph Erodibility, SI 0.036 production weighted average of sampl es Hydrologic class C mod. high runoff Hydrologic class w/subsurface drainage B mod. low runoff drainage decreases runoff Rock cover% 0 Cal. Consolidation from precip Yes Normal consolidation time, yrs 7 Default Manageme nt Rel. row grade % 100

PAGE 172

172 Long term natural rough, mm 6 Normally used as a rotation? No Duration, yr NA Operations Dates for 1st complete cycle Cropland \ disks \ disk, tandem heavy primary op. 1/1/2000 Cropland \ bedders/hippers \ hipper 2/1/2000 Add mulch NA basic/general \ begin growth 3/1/2000 basic/general \ harvest pineapple 5/1/2001 basic/general \ harvest pineapple NA basic/general \ kill vegetation 11/1/2002 Operation: harvest pineapple Portion of total biomass effected 0.4975 Assume all pineapples harvested at once, with fruit and 25% of plant being effected. Assuming 1.5 green biomass:fruit weight, 33% is the removed fruit. Of the remaining 66%, assumed 25% is chopped. 33%*22% Portion of effected left on surface 0.17 Portion of biomass effected removed fruit Portion of effected left as standing residue 0 0 Vegetation First yield for biomass conversion (kg/ha) 67000 1st above ground biomass at max canopy (kg/ha) 16000 Biomass yield ratio 0.097 Develop growth chart for a production ( yield) level other than base level Yes Adjust fall height based on canopy shape? NA Adjust biomass yield relationship NA Adjust senescence relationship see Senescence relationship Adjust yield/flow retardance relationship see Vegetation_retarda nce Setup long term veg NA Residue Responds to tillage like non fragile med (corn) default Decomp. half life, days 130 Use exponential decay equation with average lifetime from Bartholomew (2003) mean life ln(2) Weight required for area covered, 60%, kg/hec 4000 calculation (above ground biomass) (percent chopped)

PAGE 173

173 0 0 Assume of plant mass 25% is chopped and used to cover 60% of area. The mass can be related to the harvest Decomposition half life 0 Time until decay, weeks 26 from Bartholome w (2003) Halflife, weeks 18.02182669 mean life*ln(2) Senescenc e Relationshi p Above ground biomass subject to senesence, % 0 Plant continues growing until killed Vegetation retardance Type of row spacing Veg. on ridges Max. expected retardance High Avg. yield for this expect. Retardance 67000 Does 'no retardance' apply for yields >0 No Retard class at zero yield Low Table D 8 Parameters modified for USETox CR model. Item Name USETox CR USE Tox Default Source 1 Continent, Area land, km2 5.11E+04 9.01E+06 INEC 2009 2 Continent, Area sea, km2 5.00E+04 9.87E+05 Humbert et al. 2006 3 Continent, Areafrac, freshwater 8.61E 03 3.00E 02 Humbert et al. 2006 4 Continent, Areafrac, natural soil 4.60E 01 4.85E 01 INEC 2009 5 Continent, Areafrac, ag soil 5.29E 01 4.85E 01 6 Continent, Areafrac, other soil 9.78E 03 1.00E 20 7 Continent, Temperature, C 2.50E+01 1.20E+01 Humbert et al. 2006 8 Continent, Rain rate, mm/yr 3.24E+03 7.00E+02 Hum bert et al. 2006 9 Continent, Soil erosion, mm/yr 4.20E 01 3.00E 02 Rubin and Hyman 2000 10 Human pop., Continent 4.45E+06 9.98E+08 INEC 2009 11 Human pop., Urban 2.80E+06 2.00E+06 INEC 2009 12 Exposed produce, continent, kg/day/capita 2.38E+00 7.5 3E 01 Humbert et al. 2006 13 Unexposed produce, continent, kg/day/capita 8.62E 01 2.35E 01 Humbert et al. 2006 14 Meat, continent, kg/day/capita 1.11E 01 8.39E 02 Humbert et al. 2006 15 Dairy products continent, kg/day/capita 4.25E 01 2.50E 01 Humber t et al. 2006 16 Fish freshwater, kg/day/capita 5.43E 03 1.26E 02 Humbert et al. 2006 17 Fish marine, kg/day/capita 1.76E 03 3.57E 02 Humbert et al. 2006 Item Notes 4 Based on total protected area

PAGE 174

174 5 Remainder of other land area fractions 6 Assume 500 km2 of urban area + semi urban 11 63% of population lives in urban areas 14 Pork+beef+chicken+goat

PAGE 175

175 Table D 9 Sensitivity analysis of the RUSLE2 model customized for pineapple in C R Category Variable changed New Parameter Set/Value Erosion (MT/ha/yr) % Change Note Baseline NA NA 7.3 NA Climate Geographic location Pacific climate 7.5 3% Atlantic climate 7.1 3% Profile % slope 1 3.3 55% Smallest slope among sites % slope 5 20 174% % slope 10 41 462% % slope* 20 89 1119% % slope* 30 130 1681% Approximately largest slope among visited sites Soil Soil erodibility, SI 0.071 9.4 29% Silt loam with 80% silt. Estimate of most highly erodible soils present in pineapple zone Soil erodibility, SI 0.022 3.0 59% Loamy sand with 10% silt. Estimate of least erodible soils present in pineapple zone Contouring Cross slope moderate 4.3 41% Contouring Standard contouring 4.0 45% Management Management schedule Double ha rvest 4.9 33% Double harvest Management schedule Initial preparation during rainy season rainfall 9.3 27% Mulch Add plastic mulch 1.6 78% Typical in organic practice Vegetation Residue half life, days 260 5.2 29% Residue half life, days 65 8.9 22% Yield, tons/ha 33.5 14 92% Half of average yield, assume limits of competitve production Yield, tons/ha 110 4.1 44% Max yield reported, Gomez et al. 2007 Above ground dry biomass: harvest weight ratio 0.0647 7.4 1% Lowest plant biomass; based o n highest fruit:biomass fresh weight ratio of 1 (Bartholomew 2003) Above ground dry biomass: harvest weight ratio 0.144 6.8 7% Highest plant biomass; based on lowest fruit:biomass fresh weight ratio of 0.45 (Bartholomew 2003) Max for farm of unknown origin 1681% Geometric variance 17.8

PAGE 176

176 Table D 10 Sensitivity analysis of the FAO CROPWAT model to variables found in pineapple cultivation. Category Variable changed New Parameter Set/Value ET (mm/ cro p cycle) % Change Baseline NA NA 767.6 NA Climate Geographic location Pacific climate 811.7 6% Atlantic climate 712.8 7% Field Soil texture Clay 763.0 1% Sand 723.0 6% Add plastic mulch kc init= 0.6, kc mature=0.3 565.0 26% Vegetation Highe r relative crop transpiration kc init= 0.9, kc mature=0.74 1335.0 74% Root depth depth = 1m 770 0.3% Critical depletion, p High (p=0.75) 768 0.1% Yield in response to water High (Yf = 1.25) 765 0.3% High (Yf = .75) 767 0.1% Crop height, m Tal l (height = 1.25 m) 767.7 0.0% Max 74% Geo var 1.74 Table D 11 Sensitivity analysis of PestLCI model for pineapple conditions. Category Variable changed New Parameter Set/Value Sensitivity ratio % Change, fair Sensitivity ratio % Change, fsw Climate Solar radiation, MJ/m2/yr 6595 1.32 5.8% n/a n/a Solar radiation, MJ/m2/yr 6271 1.32 1.0% n/a n/a Field farm average % slope 1 n/a 1.1 66.0% farm average % slope 5 n/a 1.1 110.0% farm ave rage % slope 10 n/a 1.1 330.0% Sand content (top layer) % 82 n/a 2.0 177.8% Sand content (top layer) % 10 n/a 2.0 150.4% % canopy cover when applied 20% 0.75 55.3% 3.01 220.5% % canopy cover when applied 97% 0.8 22.1% 3.0 88.2% MAX 5 5% 330% Geo var 1.55 4.30

PAGE 177

177 Table D 12 Recalculation of Pimentel (2009) energy demand for US oranges. Input Quantity Unit CED (1E3 kcal) Machinery 50 kgc 1206.172 Diesel 337 La 3739.454 Nitrogen 196 kga 2687.112 Phosphorus 98 kga 374.5104 Potassium 196 ka 337.0593 Lime 1,120 kga 1 051.304 Herbicides 0.8 kga 34.39381 Insecticides 0.3 kga 12.89768 Fungicides 1.5 kga 64.48839 Electricity 40 kWha 86.36668 Transport 228 kga 0 Yield 48,000 kg Total without labor kcal 9593758 MJ 40167.15 MJ/kg 0.84 Table D 13 Recalculation of Pimentel (2009) energy demand for US apples. Input Quantity Unit CED (1E3 kcal) Machinery 88 kga 2123 Diesel 2,000 Ld 22193 Nitrogen 50 kge 685.5 Phosphorus 114 kga 435.7 Potassium 114 kga 196 Lime 682 kga 640.2 Herbicides 6 kgi 258 Insecticides 47 kgi 2021 Fungicides 49 kga 2107 Electricity 40 kWh 86.37 Transport 3,000 kgk 0 Yield 54,000 kg Total without labor kcal 3E+07 MJ 1E+05 MJ/kg 2.4

PAGE 178

178 Table D 14 Recalculation of Coltro (2009) energy demand for BR oranges. Input Quantity Unit CED (MJ/kg) Diesel 4.1 9 kg 53.4 Fertilizers (NPK) 11.75 kg 48.4 Bactericide 0.017 kg 180 Acaricide 1.12 kg 180 Fungicide 0.049 kg 180 Herbicide 0.149 kg 180 Insecticide 0.0093 kg 180 Lime 17.75 kg 3.93 Yield 1000 kg MJ 1005.73 MJ/kg 1.0 Table D 15 CED values for inputs used in recalculations of Orange BR, Orange US and Apples US. Process Amount Unit NR fossil CED (MJ) pesticide, unspecified 1 kg 180 ammonium nitrate, as N 1 kg 57.4 diesel, at regional storage 1 kg 53.4 single superphoshate, as P2O5 0.436 kg 16.0 tractor, production 1 kg 101 lime, hydrated, packed, at plant 1 kg 3.9 electricity, US 1 kWH 9.0 potassium chloride, as K2O 0.837 kg 7.2

PAGE 179

179 Figure D 1 Emission fractions of applied pesticides in PestLCI CR vs. the PestLCI default.

PAGE 180

180 Figure D 2 Freshwater ecotoxicity characterization factors for pesticides in USETox CR vs USETox Default

PAGE 181

181 Figure D 3 Human toxicity characterization factors for pesticides in USETox CR vs USETox Default

PAGE 182

182 Figure D 4 Human toxicity and freshwater ecotoxicity f or pesticide emissions from pineapple production in the baseline scenario.

PAGE 183

183 LIST OF REFERENCES Abou Elela, S., H. Ibrahim, and E. Abou Taleb. 2008. Heavy metal removal and cyanide destruction in the metal plating industry: an integrated approach from Egypt. The Environmentalist 28(3): 223 229. Althaus, H. J., M. Chudacoff, R. Hischier, N. Jungbluth, M. Osses, and A. Primas. 2004. Life cycle inventories of chemicals Final report ecoinvent 2000. No. 10. Dbendorf, CH: Swiss Centre for LCI EMPA DU. Australian Museum. 2007. Structure and composition of the Earth. www.amonline.net.au/geoscience/earth/structure.htm Accessed 9 September 2008. Ayres, R. U., L. W. Ayres, and K. Martinas. 1998. Exergy, waste accounting, and life cycle analysis. Energy 23(5): 355 363. Bach, O. 2008. Agricultura e implicaciones ambientales con nfasis en algunas cuencas hidrogrficas principales [Agriculture and environmental implications wit h emphasis on selected principal watersheds]. Decimotercer Informe de Estado de la Nacion en Desarollo Sostenible. San Jose: Consejo de Rectores. Baral, A. and B. R. Bakshi. 2010. Thermodynamic methods for aggregation of natural resources in life cycle a nalysis: Insight via application to some transportation fuels. Environmental Science & Technology 44: 800 807. Bare, J., T. Gloria, and G. Norris. 2006. Development of the method and U.S. normalization database for life cycle impact assessment and sustaina bility metrics. Environmental Science & Technology 40: 5108 5115. Bare, J., G. A. Norris, D. W. Pennington, and T. McKone. 2003. TRACI The tool for the reduction and assessment of chemical and other environmental impacts. Journal of Industrial Ecology 6: 49 78. Bartelmus, P. 2003. Dematerialization and capital maintenance: two sides of the sustainability coin. Ecological Economics 46: 61 81. Bastianoni, S., A. Facchini, L. Susani, and E. Tiezzi. 2007. Emergy as a function of exergy. Energy 32: 1158 1162. Bastianoni, S., D. E. Campbell, R. Ridolfi, and F. M. Pulselli. 2009. The solar transformity of petroleum fuels. Ecological Modelling 220(1): 40 50. Birkveda, M. and M. Z. Hauschild. 2006. PestLCI A model for estimating field emissions of pesticides in agr icultural LCA. Ecological Modelling 198: 433 451.

PAGE 184

184 Blanke, M. M. and B. Burdick. 2009. An energy balance (as part of an LCA) for home grown (apple) fruit versus those imported from South Africa or New Zealand. Paper presented at Joint North American LCA Con ference, 2 October, Boston. Bsch, M. E., S. Hellweg, M. A. J. Huijbregts, and R. Frischknecht. 2007. Applying Cumulative Exergy Demand (CExD) indicators to the ecoinvent database. Int J LCA 12(3): 181 190. Boustead, I. and G. F. Hancock. 1978. Handbook of industrial energy analysis New York: Ellis Horwood Ltd. Brentrup, F. and J. Ksters. 2000. Methods to estimate potential N emissions related to crop production. In Agricultural data for Life Cycle Assessments edited by B. P. Weidema and M. J. G. Meeuse n. The Hague: Agricultural Economics Research Institute (LEI). Brown, M. T. 2009. Personal Communication with Brown, M. T., Professor of Environmental Engineering Sciences. Gainesville, FL, 10 September 2009. Brown, M. T. and S. Ulgiati. 1997. Emergy based indices and ratios to evaluate sustainability: monitoring economies and technology to toward environmentally sound innovation. Ecological Engineering 9: 51 69. Brown, M. T. and S. Ulgiati. 2002. Emergy evaluations and environmental loading of electricity production systems. Journal of Cleaner Production 10(4): 321 334. Brown, M. T. and V. Buranakarn. 2003. Emergy indices and ratios for sustainable material cycles and recycle options. Resources, Conservation and Recycling 38: 1 22. Brown, M. T. and S. Ulgia ti. 2004. Emergy and environmental accounting. In Encyclopedia of Energy edited by C. Cleveland. New York: Elsevier. Brown, M. T., M. J. Cohen, and S. Sweeney. 2009. Predicting national sustainability: The convergence of energetic, economic and environmen tal realities. Ecological Modelling 220(23): 3424 3438. Brown, M. T., M. J. Cohen, E. Bardi, and W. W. Ingwersen. 2006. Species diversity in the Florida Everglades, USA: A systems approach to calculating biodiversity. Aquatic Sciences 68(3): 254 277. Brunn er, P. and H. Rechburger. 2003. Practical handbook of material flow analysis Vero Beach, FL: CRC Press. Buenaventura Mining Company Inc. 2006. Form 20 F for fiscal year 2005., edited by SEC.

PAGE 185

185 Buranakarn, V. 1998. Evaluation of Recycling and Reuse of Build ing Materials Using the Emergy Analysis Method. Ph.D. thesis, University of Florida, Gainesville. Burt, R. 2009. Soil survey field and laboratory methods manual Soil Survey Investigations Report No. 51. Lincoln, Nebraska: National Soil Survey Center, Natu ral Resources Conservation Service, U.S. Department of Agriculture. Butterman, W. C. and H. E. Hilliard. 2004. Silver Mineral Commodity Profiles. Reston, VA: U.S. Geological Survey. Butterman, W. C. and E. B. Amey. 2005. Gold Mineral Commodity Profil es. Reston, Virginia: U.S. Geological Survey. Campbell, D. 2001. A note on uncertainty in estimates of transformities based on global water budgets. In Proceedings of the Second Biennial Emergy Analysis Research Conference Gainesville, FL: Center for En vironmental Policy, University of Florida. Campos, L. 2007. Gestion de los recursos hidricos en las quencas con localizacion minera: Caso Yanacocha [Management of hydrologic resources in watersheds located in mining areas: The Case of Yanacocha] Cajamarca Peru: Minera Yanacocha, S.R.L. Canals, L. M. 2003. Contributions to LCA methodology for agricultural systems: Site dependency and soil impact assessment. Ph.D. thesis, Universidad Autonoma, Barcelona. Chapagain, A. K. and A. Y. Hoekstra. 2004. Water foot prints of nations Vol. 1, Value of Water Research Report Series No. 16 Delft, The Netherlands: UNESCO IHE Delft. Cherubini, F., M. Raugei, and S. Ulgiati. 2008. LCA of magnesium production Technological overview and worldwide estimation of environment al burdens. Resources Conservation and Recycling 52(8 9): 1093 1100. Christiansen, K., M. Wesns, and B. P. Weidema. 2006. Consumer demands on Type III environmental declarations Copenhagen: 2.0 LCA consultants. Classen, M., H. J. Althaus, S. Blaser, G. Doka, N. Jungbluth, and M. Tuchschmid. 2007. Life cycle inventories of metals Final report ecoinvent data v2.0. Dbendorf, CH: Swiss Centre for LCI, Empa TSL. Coderre, F. and D. G. Dixon. 1999. Modeling the cyanide heap leaching of cupriferous gold o res Part 1: Introduction and interpretation of laboratory column leaching data. Hydrometallurgy 52: 151 175.

PAGE 186

1 86 Cohen, M., S. Sweeney, and M. T. Brown. 2008. Computing the unit emergy value of crustal elements. In Proceedings of the 4th Biennial Emergy Confer ence edited by M. T. Brown. Gainesville, FL: Center for Environmental Policy, University of Florida. Cohen, M. J. 2001. Dynamic emergy simulation of soil genesis and techniques for estimating transformity confidence envelopes. In Proceedings of the Second Biennial Emergy Analysis Research Conference Gainesville, FL: Center for Environmental Policy, University of Florida. Coltro, L., A. Mourad, R. Kletecke, T. Mendona, and S. Germer. 2009. Assessing the environmental profile of orange production in Brazil The International Journal of Life Cycle Assessment 14(7): 656 664. Condori, P., S. Garcia, and C. Ramon. 2007. Administraction y optimizacion de operaciones de heap leaching haciendo uso de un simulador de procesos en Minera Yanacocha. In 28th Convencion Minera 10 14 September: Instituto de Ingenieros de Minas de Peru. Cuadra, M. and J. Bjrklund. 2007. Assessment of economic and ecological carrying capacity of agricultural crops in Nicaragua. Ecological Indicators 7: 133 149. Daly, G. L., Y. D. Lei, C. Teixeira, D. C. G. Muir, L. E. Castillo, and F. Wania. 2007. Accumulation of Current Use Pesticides in Neotropical Montane Forests. Environmental Science & Technology 41(4): 1118 1123. Dones, R., B. Bauer, R. Bolliger, B. Burger, M. Faist Emmenegger, R. Fr ischknecht, T. Heck, N. Jungbluth, and A. Rder. 2003. Sachbilanzen von Energiesystemen Final report ecoinvent 2000. Volume: 6. Dbendorf and Villigen, CH: Swiss Centre for LCI, PSI. Durucan, S., A. Korre, and G. Munoz Melendez. 2006. Mining life cycle modelling: a cradle to gate approach to environmental management in the minerals industry. Journal of Cleaner Production 14(12 13): 1057 1070. Ebeling, J. and M. Yasue. 2008. Generating carbon finance through avoided deforestation and its potential to crea te climatic, conservation and human development benefits. Philosphical Transactions of the Royal Society B 363: 1917 1924. Ecoinvent Centre. 2007. Ecoinvent data v2.0 Dbendorf, CH: Swiss Centre for Life Cycle Inventories. Economic Commission of Latin A merican and the Carribbean. 2006. Statistical yearbook for Latin America and the Caribbean, 2006 Santiago, Chile: Economic Commission of Latin American and the Carribbean.

PAGE 187

187 Ehrlich, H. and D. Newman. 2008. Geomicrobiology 5th ed. Boca Raton, FL: CRC Pre ss. Energy Information Administration. 2007. Peru energy data, statistics and analysis oil, gas, electricity, coal. Washington, DC.: Department of Energy. European Commission. 2003. Communication of the (European) Commission to the Council and the E uropean Parliament on Integrated Product Policy COM(2003) 302 final. FAO. 2006. Fertilizer use by crop FAO Fertilizer and Plant Nutrition Bulletin. Rome: Food and Agricultural Organization of the United Nations. FAO. 2009. FAOSTAT Trade Database. http://faostat.fao.org/site/342/default.aspx Accessed 8 July 2009. FAO. 2010. Web LocClim, local monthly climate estimator. http://www.fao. org/sd/locclim/srv/locclim.home Accessed 7 January 2010. Fargione, J., J. Hill, D. Tilman, S. Polasky, and P. Hawthorne. 2008. Land clearing and the biofuel carbon debt. Science 319(5867): 1235 1238. Fava, J. A. A. A. J., L. Lindfors, S. Pomper, B. d. Sm et, J. Warren, and B. Vigon. 1994. Lifecycle assessment data quality. A conceptual framework Pensacola, FL: SETAC. Federici, M., S. Ulgiati, and R. Basosi. 2008. A thermodynamic, environmental and material flow analysis of the Italian highway and railwa y transport systems. Energy 33(5): 760 775. Finnveden, G. 2005. The resource debate needs to continue. The International Journal of Life Cycle Assessment 10(5): 372 372. Foster, G. R., D. Yoder, and S. Dabney. 2008. Revised Universal Soil Loss Equation 2 ( RUSLE2) ARS Version May 20, 2008. USDA Agricultural Research Service, Oxford, MS. Franzese, P. P., T. Rydberg, G. F. Russo, and S. Ulgiati. 2009. Sustainable biomass production: A comparison between gross energy requirement and emergy synthesis methods. Ec ological Indicators 9(5): 959 970. Frischknecht, R. 1997. Goal and scope definition and inventory analysis. In Life Cycle Assessment: State of the Art and Research Priorites edited by H. U. d. Haes and N. Wrisberg. Bayreuth: Ecomed Publishers. Frischknech t, R. and N. Jungbluth. 2007. Implementation of Life Cycle Impact Methods. Data v2.0 (2007) Ecoinvent report No. 3. Dbendorf, CH: Swiss Centre for Life Cycle Inventories.

PAGE 188

188 Frischknecht, R., N. Jungbluth, H. J. Althaus, G. Doka, R. Dones, T. Heck, S. Hel lweg, R. Hischier, T. Nemecek, G. Rebitzer, M. Spielmann, and G. Wernet. 2007. Ecoinvent report No. 1: Overview and methodology Dbendorf: Swiss Centre for Life Cycle Inventories. Gabby, P. N. 2007. Lead. In U.S. Geological Survey Minerals Yearbook 20 05 edited by USGS. Washingon, DC: USGS. Gaillard, G. and T. Nemecek. 2009. Editorial. Paper presented at 6th International Conference on LCA in the Agri Food Sector, November 12 14, Zurich. Gallego, A., L. Rodriguez, A. Hospido, M. T. Moreira, and G. Feij oo. 2010. Development of regional characterization factors for aquatic eutrophication. International Journal of Life Cycle Assessment 15(1): 32 43. Gallopin, G. 2003. A systems approach to sustainability and sustainable development Santiago, Chile: Sustai nable Development and Human Settlements Division, United Nations Economic Commission for Latin America and the Caribbean. GeoNews. 2008. KML Polygon Area Tool. http://www.geo news.net/index_area _poligono.php Accessed August 12, 2008. Giljum, S. 2004. Trade, materials flows, and economic development in the South: The example of Chile. Journal of Industrial Ecology 8: 241 263. Gloria, T. 2009. Determination of empirical allocation measures for no n ferrous metals. Paper presented at Joint North American Life Cycle Conference, 1 October, Boston. Gobin, A., G. Govers, R. Jones, M. Kirkby, and C. Kosmas. 2003. Assessment and reporting on soil erosion Background and workshop report Copenhagen: Europe an Environment Agency. Goedkoop, M. and R. Spriensma. 2001. The Eco indicator 99: A damage oriented method for LCA Amersfoort, NL: PR consultants. Gmez, M. P., P. P. Quesada, and K. M. Bucheli. 2007. Implementation of good practices in the productio n of fresh pineapples for export: Case study of the Huetar Norte region, Costa Rica. In Implementing programmes to improve safety and quality in fruit and vegetable supply chains: benefits and drawbacks. Latin American case studies edited by L. B. D. R. M aya Pieiro. Rome: Food and Agriculture Organization of the United Nations. Google. 2008. Google Earth 4 software Palo Alto, CA: Google. Gossling Reisemann, S. 2008a. What is resource consumption and how can it be measured? Theoretical considerations. Jou rnal of Industrial Ecology 12(1): 10 25.

PAGE 189

189 Gossling Reisemann, S. 2008b. What is resource consumption and how can it be measured? Application of entropy analysis to copper production. Journal of Industrial Ecology 12(4): 570 582. Guine, J. B., ed. 2002. H andbook on life cycle assessment: operational guide to the ISO standards Vol. 7, Eco efficiency in industry and science Doldrecht, The Netherlands: Kluwer Academic. Hails, C., S. Humphrey, J. Loh, and S. Goldfinger, eds. 2008. Living Planet Report 2008 Gland, Switzerland: WWF International. Hankce, G. 1991. The effective control of a deep hole diamond drill. Paper presented at Industry Applications Society Annual Meeting, 28 Sep 4 Oct, Dearborn, MI. Hartley B., M. and R. Daz P. 2008. Mejoras ambiental es para el desarollo de la competitividad en tres cadenas agroalimentarias costarricenses [Better environments for competitive development of three Costa Rican agro food chains]. Heredia, Costa Rica: Centro Internacional de Politica Economica. Hartman, H L. 1992. SME mining engineering handbook 2nd ed. Vol. 2. Littleton, CO: Society for Mining, Metallurgy and Exploration. Hau, J. L. and B. R. Bakshi. 2004a. Expanding exergy analysis to account for ecosystem products and services. Environmental Science and Technology 38(13): 3768 3777. Hau, J. L. and B. R. Bakshi. 2004b. Promise and problems of emergy analysis. Ecological Modelling 178(1 2): 215 225. Helmer, E. H. and S. Brown. 2000. Gradient analysis of biomass in Costa Rica and a first estimate of coun trywide emissions of greenhouse gases from biomass burning. In Quantifying sustainable development the future of tropical economies edited by C. A. S. Hall, et al. San Diego: Academic Press. Heuvelmans, G., J. F. Garcia Qujano, B. Muys, J. Feyen, and P. C oppin. 2005. Modelling the water balance with SWAT as part of the land use impact evaluation in a life cycle study of CO2 emission reduction scenarios. Hydrological Processes 19(3): 729 748. Hill, A. R. and C. V. Holst. 2001. A comparison of simple statist ical methods for estimating analytical uncertainty, taking into account predicted frequency distributions. Analyst (126): 2044 2052. Hoekstra, A. Y., A. K. Chapagain, M. M. Aldaya, and M. M. Mekonnen. 2009. Water footprint manual State of the art 2009 En schede, The Netherlands: Water Footprint Network. Holdridge, L. R. 1967. Life zone ecology San Jose, CR: Tropical Science Center.

PAGE 190

190 Hopper, R. 2008. Emergy synthesis of sulfuric acid. In EES5306 Energy Analysis class, Spring 2008. Gainesville, FL: Univer sity of Florida. Hosier, B. 2008. Personal Communication with Hosier, B., Phone conversation with representative from Lindberg/MPH. November 3, 2007 2008. Huijbregts, M. A. J., W. Gilijamse, A. M. J. Ragas, and L. Reijnders. 2003a. Evaluating uncertainty i n environmental life cycle assessment. A case study comparing two insulation options for a Dutch one family dwelling. Environmental Science & Technology 37(11): 2600 2608. Huijbregts, M. A. J., S. Lundi, T. E. McKone, and D. v. d. Meent. 2003b. Geographica l scenario uncertainty in generic fate and exposure factors of toxic pollutants for life cycle impact assessment. Chemosphere 51: 501 508. and A. J. Hendriks. 2010. Cu mulative Energy Demand as a predictor for the environmental burden of commodity production. Environmental Science & Technology 44(6): 2189 2196. Infomine. 2005. Yanacocha Minesite. http://yanacocha.infomine.com Accessed Sept. 9, 2007. Ingwersen, W. W. 2010. Uncertainty characterization for emergy values. Ecological Modelling 221(3): 445 452. Ingwersen, W. W. Accepted. Emergy as a impact assessment method for life cycle assessment presented in a gold mining case s tudy. Journal of Industrial Ecology Ingwersen, W. W., S. A. Clare, D. Acua, M. J. Charles, C. Koshal, and A. Quiros. 2009. Environmental Product Declarations: An introduction and recommendations for their use in Costa Rica Gainesville, FL: University of Florida Levin College of Law Conservation Clinic. Instituto Nacional Estadistica y Informacion. 2006. Peru compendio estadistico 2006 Lima, Peru: Instituto Nacional Estadistica y Informacion. Instituto Peruano de Economia. 2003. La brecha en infraest ructura: Servicios publicos, productividad, y crecimiento en el Peru Lima: International Mining News. 2005. The Yanacocha Seven. International Mining News [Hertfordshire, UK] IPCC. 2007. Climate change 2007. IPCC fourth assessment report. The physical science basis. Geneva: International Panel on Climate Change.

PAGE 191

191 ISO. 2006a. 14044: Environmental management -Life cycle assessment -Requirements and guidelines Geneva: International Organization for Standardization. ISO. 2006b. 14025: Environmental labelling and declarations Type III environmental declarations Principles and procedures International Standard. Geneva: International Organization for Standardization. ISO. 2006c. 14040: Environmental management -Life cycle assessment -Principl es and framework Geneva: International Organization for Standardization. Jolliet, O., M. Margni, R. Charles, S. Humbert, J. Payet, G. Rebitzer, and R. Rosenbaum. 2003a. IMPACT 2002+: A new life cycle impact assessment methodology. International Journal of Life Cycle Assessment 8(6): 324 330. Jolliet, O., A. Brent, M. Goedkoop, N. Itsubo, R. Mueller Wenk, C. Pea, R. Schenk, M. Stewart, and B. Weidema. 2003b. Final report of the LCIA Definition study UNEP/SETAC Life Cycle Initiative. United National Envi ronmental Program. Joyce, A. 2006. Land use change in Costa Rica 1966 2006 as influenced by social, economic, political and environmental factors San Jose: Litografa e Imprenta LIL. Kodjak, D. 2004. Policy discussion Heavy duty truck fuel economy. P aper presented at 10th Diesel Engine Emissions Reduction (DEER) Conference, 29 August 2 September, Coronado, CA. La Rosa, A. D., G. Siracusa, and R. Cavallaro. 2008. Emergy evaluation of Sicilian red orange production. A comparison between organic and c onventional farming. Journal of Cleaner Production 16(17): 1907 1914. Lal, R. 1983. Soil erosion in the humid tropics with particular reference to agricultural land development and soil management. Paper presented at Hydrology of Humid Tropical Regions wit h Particular Reference to the Hydrological Effects of Agriculture and Forestry Practice, 15 October, Hamburg. Lenzen, M. and U. Wachsmann. 2004. Wind turbines in Brazil and Germany: an example of geographical variability in life cycle assessment. Applied E nergy 77: 119 130. Lillywhite, R., D. Chandler, W. Grant, K. F. Lewis, C., U. Schmutz, and D. Halpin. 2007. Environmental footprint and sustainability of horticulture (including potatoes) A comparison with other agricultural sectors UK: DEFRA. Limpert E., W. A. Stahel, and M. Abbt. 2001. Log normal distributions across the sciences: Keys and clues. Bioscience 51(5): 341 352.

PAGE 192

192 Lloyd, S. and R. Ries. 2007. Characterizing, propagating, and analyzing uncertainty in life cycle assessment. Journal of Industr ial Ecology 11(1): 161 179. Longo, A. 2005. Evolution of volcanism and hydrothermal activity in the Yanacocha Mining District, northern Per. Ph.D. thesis, Oregon State University. Lowrie, R. L. 2002. SME mining reference handbook Littleton, CO: Society f or Mining, Metallurgy and Exploration. Maia de Souza, D., R. Rosenbaum, L. Deschnes, and H. Lisboa. 2009. Crucial improvements needed for land use impact assessment modeling concerning biodiversity indicators. Paper presented at Life Cycle Assessment IX Joint North American Life Cycle Conference, 29 September 2 October, Boston. Malzieux, E., F. Cte, and D. P. Bartholomew. 2003. Crop environment, plant growth and physiology. In The pineapple: Botany, production, and uses edited by D. P. Bartholomew, et al. Oxon, UK: CABI Pub. Marsden, J. and I. House. 2006. The chemistry of gold extraction 2nd ed: SME. Matthews, E., C. Amman, S. Bringezu, M. Fischer Kowalski, W. Huttler, R. Kleijn, Y. Moriguichi, C. Ottke, E. Rodenburg, D. Rogich, H. Schandl, H. Sc hutz, E. V. d. Voet, and H. Weisz. 2000. The weight of nations: Material outflows from industrial economies 1st ed. Washington, DC: World Resources Institute. ME Assessment. 2005. Ecosystems & human well being: Biodiversity synthesis Millineum Ecosystem Assessment Washington, DC: World Resources Institute. Miller, S. A., A. E. Landis, and T. L. Theis. 2006. Use of monte carlo analysis to characterize nitrogen fluxes in agroecosystems. Environmental Science & Technology 40(7): 2324 2332. Mimbela, R. 200 7. Filosofia y gestion de agua [Philosophy and management of water]. Lima: Minera Yanacocha S.R.L. Minera Yanacocha S.R.L. 2005. Procidimiento: Plan Integral de Control de Polvo [Procedure: Integrated plan to control dust]. MA PA 026. Lima, Peru: Minera Ya nacocha S.R.L. Minera Yanacocha S.R.L. 2006. La produccion del oro en Yanacocha [Gold production at Yanacocha]. Informes de Centro de Informaccion. Cajamarca, Peru: Minera Yanacocha S.R.L. Minera Yanacocha S.R.L. 2007. Mine Tour. Cajamarca, Peru. Minin g Technology. 2007. Minera Yanacocha Gold Mine, Peru. http://www.mining technology.com Accessed October 1, 2007.

PAGE 193

193 Montgomery Watson. 1998. Estudio de impacto ambiental: Proyecto La Quinua [Environmental impact study: La Quinua project]. Santiago, Chile: Montgomery Watson. 2004. Plan de cierre conceptual: La Quinua [Conceptual mine closing plan: La Quinua]. Lima, Peru: Montoya, P. and J. Quispe. 2007. Maqui maqui: Ejemplo de cierre exitoso [Maqui maqui: Exam ple of a successful mine closing]. DDC Fabrica de Ideas. NAS. 1999. Nature's numbers Edited by W. N. a. E. Kokkelenburg. Washington, DC: National Academy of Sciences. National Metal Finishing Resource Center. 2007. Pollution prevention and control tech nologies for plating operations. http://www.nmfrc.org/ Accessed October 20, 2007. National Renewable Energy Laboratory. 2008. Notes regarding transparency, data publishing (unit processes) and data exchange Life Cycle Assessment Working Paper No. 7. Golden, Colorado: Nemecek, T. and T. Kagi. 2007. Life cycle inventory of agricultural production systems Dubendorf: Ecoinvent Centre. Ness, B., E. Urbel Piirsalu, S. Anderberg, and L. Olsson. 2007. Categorising tools fo r sustainability assessment. Ecological Economics 60(3): 498 508. Newmont. 2004. Social and environmental responsibility Denver, CO: Newmont External Affairs and Communication Department. Newmont. 2006a. Now and beyond 2005 sustainability report: Minera Yanacocha, Peru Denver, Colorado: Newmont. Newmont. 2006b. 2005 annual report Denver, Colorado: Newmont. 2006a. Now & Beyond 2005 Corporate Sustainability Report Denver, Colorado: Newmont. 2006c. Form 10 K for fiscal year 2005, edited by SEC. N ewmont Waihi Gold. 2007. Equipment at the Martha mine. http://www.newmont.com/en/operations/australianz/waihigold/mining/index.asp Accessed November 1, 2007. NIST. 2 010. The NIST reference on constants, units, and uncertainty. http://physics.nist.gov/cuu/Uncertainty/combination.html Accessed 26 January 2010.

PAGE 194

194 Norris, G. A. 2003. Impact characteriza tion in the tool for the reduction and assessment of chemical and other environmental impacts (TRACI) Methods for acidification, eutrophication, and ozone formation. Journal of Industrial Ecology 6(3 4): 79 101. O'Brien, E., B. Guy, and A. S. Lindner. 20 06. Life cycle analysis of the deconstruction of military barracks: Ft. McClellan, Anniston, AL. Journal of Green Building 1(4): 166 183. Odum, H. T. 1988. Self organization, transformity, and information. Science 242: 1132 1139. Odum, H. T. 1996. Environm ental Accounting New York: John Wiley & Sons. Odum, H. T. 2007. Environment, power and society for the twenty first century: The hierarchy of energy New York: Columbia University Press. Odum, H. T. 1991. Emergy of South African gold. In Ecological Phys ical Chemistry. Proceeding of a Conference edited by C. Rossi and E. Tiezzi. Siena, Italy: Elsevier. Odum, H. T., M. T. Brown, and S. Brandt Williams. 2000. Handbook of emergy evaluation folio #1: Introduction and global budget Gainesville: Center for En vironmental Policy, University of Florida. Pennington, D. W., M. Margni, C. Ammann, and O. Jolliet. 2005. Multimedia fate and human intake modeling: Spatial versus nonspatial insights for chemical emissions in Western Europe. Environmental Science & Tech nology 39(4): 1119 1128. Peruvian Ministry of Energy and Mines. 2006. Annual Production 2005: Gold Lima, Peru: Peters, G. M., H. V. Rowley, S. Wiedemann, R. Tucker, M. D. Short, and M. Schulz. 2010. Red meat production in Australia: Life cycle assessmen t and comparison with overseas studies. Environmental Science & Technology 44(4): 1327 1332. Pfister, S., A. Koehler, and S. Hellweg. 2009. Assessing the environmental impacts of freshwater consumption in LCA. Environmental Science & Technology 43(11): 409 8 4104. Pimentel, D. 2009. Energy inputs in food crop production in developing and developed nations. Energies 2: 1 24. Pizzigallo, A. C. I., C. Granai, and S. Borsa. 2008. The joint use of LCA and emergy evaluation for the analysis of two Italian wine far ms. Journal of Environmental Management 86(2): 396 406.

PAGE 195

195 Powers, S. E. 2007. Nutrient loads to surface water from row crop production. International Journal of Life Cycle Assessment 12(6): 399 407. PR Consultants. 2008. SimaPro 7.1. Ph.D. Version., Amsfoor t, NL. Rai, S. N. and D. Krewski. 1998. Uncertainty and variability analysis in multiplicative risk models. Risk Analysis 18(1): 37 45. Reap, J., F. Roman, S. Duncan, and B. Bras. 2008. A survey of unresolved problems in life cycle assessment: Part 2: impa ct assessment and interpretation. International Journal of Life Cycle Assessment 13(5): 374 388. Ridoutt, B. G. and S. Pfister. 2010. A revised approach to water footprinting to make transparent the impacts of consumption and production on global freshwate r scarcity. Global Environmental Change 20(1): 113 120. Ridoutt, B. G., P. Juliano, P. Sanguansri, and J. Sellahewa. 2009. Consumptive water use associated with food waste. Hydrology and Earth System Sciences Discussions 6: 5085 5114. Roos, E., C. Sunderbe rg, and P. A. Hansson. 2010. Uncertainties in the carbon footprint of food products: a case study on table potatoes. International Journal of Life Cycle Assessment 15(5): 478 488. Rosenbaum, R. K., T. M. Bachmann, L. S. Gold, M. A. J. Huijbregts, O. Jollie t, R. Juraske, A. Koehler, H. F. Larsen, M. MacLeod, M. Margni, T. E. McKone, J. Payet, M. Schuhmacher, D. van de Meent, and M. Z. Hauschild. 2008. USEtox the UNEP SETAC toxicity model: Recommended characterisation factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. International Journal of Life Cycle Assessment 13(7): 532 546. Rubin, B. D. and G. G. Hyman. 2000. The extent and economic impacts of soil erosion in Costa Rica. In Quantifying sustainable development the futur e of tropical economies edited by C. A. S. Hall, et al. San Diego: Academic Press. Rydburg, T. 2010. Personal Communication with Rydburg, T., Professor of Environmental Science. Gainesville, FL 2010. Sandoval, A. C. C. 2009. Insensatez piera [Foolish pin eapple production]. El Financiero [San Jose, CR], July 5, section En Portada. Schenck, R. 2007. Canning green beans Ecoprofile of Truitt Brothers process Vashon, WA: Institute for Environmental Research and Education. Schenck, R. 2009. The outlook and ppportunity for Type III environmental product declarations in the United States of America White Paper. Vashon, WA: Institute for Environmental Research and Education.

PAGE 196

196 Schenck, R. C. and S. Vickerman. 2001. Developing a land use/biodiversity indicator for agricultural product LCAs. In Proceedings of the First International Conference on LCA in Foods Gothenburg, Sweden. Schmidt Bleek, F. 1994. Wieviel Umwelt braucht der Mensch? MIPS, das Mass fur okologisches Wirtschaften [How much environment do we ne ed? MIPS, the measure for ecologically sound economic performance]. Berlin: Birkhauser. Scholl, D. and v. Huene. 2004. Crustal recycling at ocean margin and continental subduction zones and the net accumulation of continental crust. EOS Transactions 88(5 2). Sciubba, E. and S. Ulgiati. 2005. Emergy and exergy analyses: Complementary methods or irreducible ideological options? Energy 30(10): 1953 1988. Seager, T. P. and T. L. Theis. 2002. A uniform definition and quantitative basis for industrial ecology. J ournal of Cleaner Production 10: 225 235. Seppala, J., S. Knuuttila, and K. Silvo. 2004. Eutrophication of aquatic ecosystems A new method for calculating the potential contributions of nitrogen and phosphorus. International Journal of Life Cycle Assessm ent 9(2): 90 100. Sinden, G. 2008. PAS 2050:2008 Specification for the assessment of the life cycle greenhouse gas emissions of goods and services London: British Standards Institute. Slob, W. 1994. Uncertainty analysis in multiplicative models. Risk An alysis 14(4): 571 576. Sonnemann, G. and B. de Leeuw. 2006. Life cycle management in developing countries: State of the art and outlook. International Journal of Life Cycle Assessment 11(Special Issue 1): 123 126. Spielmann, M., T. Kgi, P. Stadler, and O. Tietje. 2004. Life cycle inventories of transport services Final report ecoinvent 2000. Volume: 14., UNS. Dbendorf, CH: Swiss Centre for LCI. Stewart, M. and B. P. Weidema. 2005. A consistent framework for assessing the impacts from resource use A f ocus on resource functionality. The International Journal of Life Cycle Assessment 10(4): 240 247. Stratus Consulting. 2003. Report on the independent assessment of water quantity and quality near the Yanacocha mining district, Cajamarca, Peru Washington, DC: IFC/MIGA Compliance Advisor. Su, N. R. 1968. Pineapple (Ananas comosus (L) Merr.) nutritional requirements Taipei: Taiwan Council of Agriculture.

PAGE 197

197 Sweeney, S., M. Cohen, D. King, and M. Brown. 2009. National Environmental Accounting Database. http://sahel.ees.ufl.edu/frame_database_resources_test.php Accessed 10 May 2009. Swennenhuis, J. 2009. CROPWAT version 8.0. Water Resources Development and Management Service, FAO, Rome. Taylor, S. R. and S. M. McLennan. 1985. The continental crust: its composition and evolution : an examination of the geochemical record preserved in sedimentary rocks Palo Alto, CA: Blackwell Scientific. The Tank Shop. 2007. Tank Weight Calculator Sprea dsheet Tool. http://www.thetankshop.ca/private/admin/upload/xls/Tank%20Weight%20Calcula tor.xls Accessed October 10, 2008. Thiesen, J., S. Valdivia, G. Sonneman n, J. Fava, T. Swarr, A. A. Jensen, and E. Price. 2007. Understanding challenges and needs: A stakeholder consultation on CICLA 2007 Sao Paolo, Brazil. Thornton, I. and S. Brush. 2001. Lead: The facts L ondon: IC Consultants Ltd. Tilley, D. R. 2003. Industrial ecology and ecological engineering: Opportunities for symbiosis. Journal of Industrial Ecology 7(2): 13 32. Ukidwe, N. and B. R. Bakshi. 2004. Thermodynamic accounting of ecosystem contribution to economic sectors with application to 1992 U.S. economy. Environmental Science & Technology 38: 4810 4827. Ulgiati, S., M. Raugei, and S. Bargigli. 2006. Overcoming the inadequacy of single criterion approaches to Life Cycle Assessment. Ecological Modellin g 190(3 4): 432 442. UN. 1992. Declaration on environment and development. Rio de Jainero, Brazil: United Nations. UN. 2005. Johannesburg Plan of Implementation Johannesburg, SA: United Nations. UN DESA. 2008. The Marrakech Process. http://esa.un.org/marrakechprocess/ Accessed 19 May 2010. UNEP. 2007. Life cycle management A business guide to sustainability Paris: United Nations Environment Programme. UNEP Life Cycle Initiative. 2007. Life Cycle Initiative Phase 2 2007 2010 UNEP DTIE Project Brief. Paris, France: United Nations Environment Programme, Division of Technology, Industry & Economics.

PAGE 198

198 United Nations. 2008. UN Comtrade Database. http://comtrade.un.o rg Accessed March 22, 2008. UNSTAT. 2006. Demographic Yearbook Table 3: Population by sex, rate of population increase, surface area and density. http://unstats.un.org/u nsd/demographic/products/dyb/dyb2006/Table03.pdf Accessed 13 August 2008. UoH. 2005. Sustainability of UK Strawberry Crop University of Hertfordshire. Urban, R. A. and B. R. Bakshi. 2009. 1,3 Propanediol from fossils versus biomass: A life cycle evalu ation of emissions and ecological resources. Industrial & Engineering Chemistry Research 48(17): 8068 8082. USDA. 2009. National Nutrient Database for Standard Reference, Release 22. http://www.nal.usda.gov Accessed November 2, 2009. Van Der Voet, E., L. Van Oers, and I. Nikolic. 2004. Dematerialization: Not just a matter of weight. Journal of Industrial Ecology 8(4): 121 137. Wackernagel, M., N. B. Schulz, D. Deumling, A. C. Linares, M. Jenkins, V. Kapos, C. Monfreda J. Loh, N. Myers, R. Norgaard, and J. Randers. 2002. Tracking the ecological overshoot of the human economy. Proceedings of the National Academy of Sciences 19(14): 9266 9271. Weidema, B. and G. Norris. 2002. Avoiding co product allocation in the metals sector. In Life cycle assessment of metals: Issues and research directions edited by A. Dubriel. Pensacola, FL: Society of Environmental Toxicology and Chemistry. Williams, A., E. Pell, J. Webb, E. Moorhouse, and E. Audsley. 2008. Strawberry and tomato pr oduction for the UK compared between the UK and Spain. Paper presented at International Conference on LCA in the Agri Food Sector, November 12 14, Zurich. Wong, S. S., T. T. Teng, A. L. Ahmada, A. Zuhairi, and G. Najafpour. 2006. Treatment of pulp and pape r mill wastewater by polyacrylamide (PAM) in polymer induced flocculation. Journal of Hazardous Materials B135: 378 388. World Gold Council. 2006. Mine Production. www.gold.o rg/value/markets/supply_demand/mine_production.html Accessed 12 October 2007. Yellishetty, M., P. G. Ranjith, A. Tharumarajah, and S. Bhosale. 2009. Life cycle assessment in the minerals and metals sector: A critical review of selected issues and challen ges. International Journal of Life Cycle Assessment 14(3): 257 267.

PAGE 199

199 Zhang, Y., Z. Yang, and X. Yu. 2009. Ecological network and emergy analysis of urban metabolic systems: Model development, and a case study of four Chinese cities. Ecological Modelling 220 (11): 1431 1442. Zhang, Y., S. Singh, and B. R. Bakshi. 2010. Accounting for ecosystem services in life cycle assessment, part I: A critical review. Environmental Science & Technology 44(7): 2232 2242.

PAGE 200

200 BIOGRAPHICAL SKETCH Wesley W. Ingwersen was born in Atlanta, GA in 1977 and grew up in the Stone Mountain area. He went to secondary school at Woodward Academy in College Park, GA, where he developed a keen interest in environmental science After a year at Wake Forest University he transferred to Geor getown University (Washington, DC) where he completed a B.A. in 1999. Wesley worked for an e commerce company, enews.com, and a software development company, Lokitech, as a web designer and I nternet applications developer until 2002. While in the DC area a nd volunteering with the National Park Service and the Casey Tree Foundation, he became determined to work toward greater scientific understanding of the dependence of human systems upon nature and the value it provides, an d returned to graduate school to pursue an M.S. in Environmental Engineering at the University of Florida. His M.S. thesis was an evaluation of long term term success of wetland reclamation efforts on phosphate mined lands. Following the completion of his M.S degree, Wesley joined Eco logic, and environmental policy think tank in Berlin as a Transatlantic Fellow, and at the end of 2006 returned to UF to pursue a Ph.D. under his M.S. adviser, Mark T. Brown. Wesley is a Life Cycle Assessment Certified Professional I n addition to the LC A work in this dissertation he contributed to a study of future transportation related GHG emissions for the state of Florida led a feasibility study of environmental product declarations (EPDs) in Costa Rica, and is involved in the development of nation al guidance standards for EPDs in the US He has published b ook chapters, peer reviewed journal articles, and presented papers for conferences on issue s of trade and the environment, environmental assessment, life cycle assessment, uncertainty modeling, a nd emergy analysis.