Spatial-Temporal Modeling of Soil Organic Carbon Across a Subtropical Region

MISSING IMAGE

Material Information

Title:
Spatial-Temporal Modeling of Soil Organic Carbon Across a Subtropical Region
Physical Description:
1 online resource (97 p.)
Language:
english
Creator:
Ross,Christopher
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Soil and Water Science
Committee Chair:
Grunwald, Sabine
Committee Members:
Martin, Timothy A
Silveira, Maria L
Myers, David

Subjects

Subjects / Keywords:
carbon -- climate -- florida -- landcover -- landuse -- modeling -- organic -- pedon -- pedosphere -- region -- sequestration -- soil -- sub -- tropical
Soil and Water Science -- Dissertations, Academic -- UF
Genre:
Soil and Water Science thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Rising levels of anthropogenic carbon-dioxide emissions have raised concern that human actions may be altering the world?s climate. Recent research suggests that landcover/landuse (LC/LU) change may have significant influence on soil carbon (C) cycling. Optimized landuse management may provide opportunity to offset anthropogenic C emissions by sequestering C into terrestrial systems. Accurately quantifying changes in soil organic carbon (SOC) stocks in response to LC/LU change is essential for determining C fluxes to and from soils and will aid policy makers in making better, informed decisions regarding current economic and environmental policy. The objectives of this project were to i) quantify and compare ?historic? and ?current? SOC stocks; ii) quantify SOC sequestration rates; and iii) assess the effect of LC/LU change on SOC. The study was conducted across a subtropical region comprised of four hydrologic basins in north east and east central Florida which include the Upper, Lower and Middle St. Johns River Basins as well as the Ocklawaha River Basin (22,266 km2). The study area has experienced substantial LC/LU change over the last 40 years; this change may have affected SOC stocks within the study area. Two SOC datasets were used for this study: i) Dataset 1 (DS1) n=304 representing historic conditions (1965 ? 1996) and ii) Dataset 2 (DS2) n=402 representing current conditions (2008/2009). Ordinary kriging and block kriging were used to derive SOC stock predictions from site-specific observations across the region and cross-validate and validate estimates. Geoprocessing techniques were used to calculate SOC stocks, SOC sequestration rates, and to assess the effects of LC/LU change on SOC. Significant differences were found to exist among various LC/LU?s in regards to mean SOC stocks (kg m-2, 0-20cm), with the highest amounts found in Cypress Swamp (9.7), Hardwood Swamp (9.6) and Mixed Wetland Forest (7.8). Additionally, significant SOC stock differences among various soil types exist as well, with the highest mean stocks (kg m-2, 0-20cm) belonging to Saprists (12), Aquolls (9.8) and Aquepts (9.4). Geostatistical (kriging) models developed for the study area show approximately 102 ? 108 Tg SOC (kg C m-2) are held within the upper 20cm of soils representing historical conditions (DS1) and 211 - 320 Tg SOC (kg C m-2) are held within the upper 20cm of soils representing current conditions (DS2), which suggests the soils in the study area have been a net sink for C during the last 40 years. Highest SOC stock sequestration rates were observed in Hardwood/Cypress Swamp (51 g C m-2 yr-1) and the lowest observed in Xeric Upland Forest (-129 g C m-2 yr-1). Additionally, site remaining in Row/field Crop lost SOC (-2g C m-2 yr-1) on average. Interestingly, three out of four classes switching to Urban resulted in net gains of SOC stocks. Geostatistical models improved the knowledge of the spatial distribution and variability of SOC in the study area with implications for SOC cycling, land management, environmental conservation and policy decisions.
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 Christopher Ross.
Thesis:
Thesis (M.S.)--University of Florida, 2011.
Local:
Adviser: Grunwald, Sabine.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0042917:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 SPATIAL TEMPORAL MODE LING OF SOIL ORGANIC CARBON ACROSS A SUBTROPICAL REGION By CHRISTOPHER WADE ROSS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Christopher Wade Ross

PAGE 3

3 To all soil scientists and environmental enthusiasts

PAGE 4

4 ACKNOWLEDGMENTS I would like to take this opportunity to thank my parents for their support and guidance. I also thank the University of Florida, the Soil and Water Science Department and the USDA for providing the resources and funding to conduct this study. I thank my m ajor advisor, Sabine Grunwald as well as my supervisory committee Brent Myers, Maria Silveira and Tim Martin for their guidance. I would also like to thank Brandon Hoover for Information Technology expertise, as well as my colleagues from the Geographic In formation Sy stems Research Laboratory, Pasicha Ch aikaew, Jongsung Kim, Jinseok Hong, Xiong Xiong, Baijing Ca o, Julius Adewopo and Nichola Knox for their friendship and support. I also thank Aja Stoppe and Elena Azuaje for their role in my study, particular ly the long hot days using the hammermill. Funding for this project was provided by Rapid Assessment and Trajectory Modeling of Changes in Soil Carbon across a Southeastern Landscape (USDA CSREES NRI grant award 2007 35107 18368, note: CSREES NRI has been renamed to AFRI NIFA).

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .........................................................................................................................9 ABSTRACT ...................................................................................................................................10 CHAPTER 1 INTRODUCTION ..................................................................................................................12 2 BACKGROUND ....................................................................................................................13 The Lithosphere ......................................................................................................................13 The Hydrosphere ....................................................................................................................13 The Pedosphere .......................................................................................................................14 The Terrestrial Biosphere .......................................................................................................14 The Atmosphere ......................................................................................................................14 3 STUDY AREA .......................................................................................................................16 Upper St. Johns River Basin ...................................................................................................16 Middle St. Johns River Basin .................................................................................................17 Lower St. Johns River Basin ..................................................................................................17 Ocklawaha River Basin ..........................................................................................................18 Landcover/ Landuse .................................................................................................................19 Soil Properties .........................................................................................................................19 Hydrology ...............................................................................................................................19 4 QUANTIFY REGIONAL SOIL ORGANIC CARBON STOCKS FOR HISTORICAL AND CURRENT CONDITIONS ...........................................................................................26 Materials and Methods ...........................................................................................................27 Historic Conditions Dataset 1 .......................................................................................27 Sampling Design .............................................................................................................28 Sample Collection Protocol .............................................................................................28 Labora tory Analysis ........................................................................................................29 Conversion of Soil Organic Carbon Concentration into Stocks ......................................30 Reconstruction of Horizon Based Soil Carbon Data into Fixed Depth Layer Intervals ........................................................................................................................30 Statistical and Geostatistical Analysis .............................................................................31 Modeling Soil Organic Carbon Stocks ............................................................................31 Current Conditions Dataset 2 ........................................................................................33

PAGE 6

6 Sampling Design .............................................................................................................34 Sample Collection Protocol .............................................................................................35 Sample Processing ...........................................................................................................36 Laboratory Analysis ........................................................................................................36 Statistical and Geostatistical Analysis .............................................................................36 Soil Organic Carbon Stock Estimates .............................................................................37 Relationship between Soil Organic Carbon and Environmental Landscape Factors ......38 Results and Discussion ...........................................................................................................38 Historic Conditions Dataset 1: Assessment of Soil Organic Carbon Stocks ................38 Estimated Soil Organic Carbon Stocks ...........................................................................39 Current Conditions Dataset 2 .......................................................................................41 Assessment of Soil Organic Carbon St ocks ....................................................................41 Estimated Soil Organic Carbon Derived from Kriging ...................................................42 Soil Organic Carbon Stock Assessment Derived from Class PedoTransfer Functions ......................................................................................................................43 Relationships Between Soil Organic Carbon And Environmental Landscape Factors ..........................................................................................................................44 Co nclusions .............................................................................................................................47 5 QUANTIFY SOIL ORGANIC CARBON SEQUESTRATION RATES ..............................62 Materials and Methods ...........................................................................................................63 Quantify S oil Organic Carbon Sequestration Rates for Collocated Sites .......................63 Quantify Soil Organic Carbon Change for Kriged Estimates .........................................64 Results and Discussion ...........................................................................................................65 Conclusions .............................................................................................................................68 6 ASSESSMENT OF THE EFFECT OF LANDCOVER/LANDUSE CHANGE ON REGIONAL SOIL ORGANIC CARBON STOCKS .............................................................75 Materials and Methods ...........................................................................................................76 Comparison of Historic and Current Landcover/Landuse Conditions ............................76 Classification Scheme .....................................................................................................77 Results an d Discussion ...........................................................................................................77 Conclusions .............................................................................................................................81

PAGE 7

7 APPENDIX A DATASET 1 VARIOGRAMS ...............................................................................................87 B DATASET 2 VARIOGRAMS ...............................................................................................89 C FLORIDA SOIL CARBON PROJECT LANDCOVER/LANDUSE RECLASSIFACTION SCHEME ...........................................................................................91 LIST OF REFERNCES .................................................................................................................92 BIOGRAPHICAL SKETCH .........................................................................................................97

PAGE 8

8 LIST OF TABLES Table page 31 Landcover/landuse (LC/LU) distribution within the study area ........................................25 32 Distribution of soil orders within the study area ................................................................25 41 Descriptive statistics of field observed soil organic carbon (SOC) stocks derived from historic Dataset 1. ......................................................................................................54 42 Variogram parameters for soil organic carbon (SOC) stocks derived from Dataset 1. .....54 43 Predictions for historic soil organic carbon (SOC) stocks derived from ordinary kriging (OK) and block kriging (BK) using historic Dataset 1. ........................................55 44 Descriptive statistics of field observed soil organic carbon (SOC) stocks derived from current Dataset 2 .......................................................................................................56 45 Variogram parameters for soil organic carbon (SOC) stocks derived from Dataset 2. .....56 46 Kriged estimates of soil organic carbon stocks derived from ordinary kriging (OK) and block kriging (BK) using Dataset 2. ...........................................................................57 47 Soil organic carbon stock observations using Regional Dataset 2 stratified by landcover/landuse (LC/LU) classes ...................................................................................58 48 Soil organic carbon stock observations using Regional Dataset 2 stratified by soilorder classes .......................................................................................................................59 49 Analysis of variance (ANOVA) for landcover/landuse classes derived from Dataset 2 using Dunnetts T3 test. .....................................................................................................60 410 Analysis of Variance (ANOVA) for soil suborders derived from Dataset 2 using Dunnetts T3 test. ...............................................................................................................61 51 Soil organic carbon (SOC) accumulation rates by soil suborder. ......................................74 61 Landcover/landuse (LC/LU) change classification scheme ..............................................84 62 Confusion matrix showing the soi l organic carbon change by landcover/landuse in the study area .....................................................................................................................85 63 Confusion matrix showing the average distance and number of si tes undergoing landcover/landuse change within the study area ................................................................86

PAGE 9

9 LIST OF FIGURES Figure page 31 Location and extent of study area. .....................................................................................21 41 Spatial distribution of Dataset 1 .........................................................................................50 42 Spatial distribution of Dataset 2 .........................................................................................51 43 Kriged estimates of historic soil organic carbon (SOC) stocks (kg C m2, 020 cm depth) derived from Dataset 1within the core study area. .................................................52 44 Kriged estimates of current soil organic carbon (SOC) stocks (kg C m2, 020 cm depth) derived from Dataset 2 within the core study area.. ...............................................53 51 Soil organic carbon gains and losses (g C m2 yr1) derived from Dataset 1 and Dataset 2 collocated sites (200 m) within the study area. ..................................................70 52 Variance maps derived from Dataset 1 soil organic carbon (log10) estimates classified by quantiles. .......................................................................................................................71 53 Variance maps derived from Dataset 2 soil organic carbon (Log10) estimates classified by quantiles. .......................................................................................................72 54 Series of prediction maps showing SOC stock gains and losses (kg C m2) within the study area. ..........................................................................................................................73 A 1 Variogram of log transformed soil organic carbon stocks ................................................87 A 2 Variogram of log transformed soil organic carbon stocks ................................................88 B 1 Variogram of log transformed soil or ganic carbon stocks ................................................89 B 2 Variogram of log transformed soil organic carbon stocks ................................................90

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science SPATIAL TEMPORA L MODE LING OF SOIL ORGANIC CARBON ACROSS A SUBTROPICAL REGION By Christopher Wade Ross August 2011 Chair: Sabine Grunwald Major: Soil and Water Science Ri sing levels of anthropogenic carbondioxide emissions have raised concern that human actions may be altering the worlds climate. Recent research suggests that landcover /landuse (LC/LU) change may have significant influence on soil carbon (C) cycling. Optimized landuse management may provide opportunity to offset anthropogenic C emissions by sequestering C into terrestrial systems. Accurately quantifying changes in soil organic carbon (SOC) stocks in response to LC/LU change is essential for determining C fluxes to and from soils and will aid policy makers in making better, informed decisions regarding current eco nomic and environmental policy. The objectiv es of this project were to i) quantify and compare historic and current SOC stocks; ii) quantify SOC sequestration rates ; and iii) assess the effect of LC/LU change on SOC. The study was conducted across a subtropical region comprised of four hydrologic basins in north east and east central Florida which include the Upper, Lower and Middle St. Johns River Basin s as well as the Ocklawaha River Basin ( 22,266 km2) The study area has experienced substantial LC/LU change over the last 40 years; this change may have affected SOC stocks within the study area Two SOC dat asets were used for this study : i) Dataset 1 (DS1) [n=304] re presenting historic conditions (1965 1996) and ii) Dataset 2 (DS2) [n=402] representing

PAGE 11

11 current conditions (2008/2009) Ordinary kriging and block kriging were used to derive SOC stock predictions from site specif ic observations across the region and cross validate and validate estimates. Geoprocessing techniques were used to calculat e SOC stocks SOC sequestration rates, and to assess the effects of LC/LU change on SOC. Significant differences were found to exist among various LC/LUs in regards to mean SOC stocks (kg m2, 020cm), with the highest amounts found in Cypress Swamp (9.7), Hardwood Swamp (9.6) and Mixed Wetland Forest (7.8). Additionally, significant SOC stock differences among various soil types exist as well, with the highest mean s tocks (kg m2, 020cm) belonging to Saprists (12), Aquolls (9.8) and Aquepts (9.4). Geostatistical (kriging) models developed for the study area show approximately 102 108 Tg SOC (kg C m2) are held within the upper 20cm of soils representing historical conditions (DS1) and 211 320 Tg SOC (kg C m2) are held within the upper 20cm of soils representing current conditions (DS2), which suggests the soils in the study area have been a net sink for C during the last 40 years. Highest SOC stock sequestration rates were observed in Hardwood/Cypress Swamp (51 g C m2 yr1) and the lowest observed in Xeric Upland Forest ( 129 g C m2 yr1). Additionally, site remaining in Row/field Crop lost SOC ( 2g C m2 yr1) on average. Interestingly, three out of four clas ses switching to Urban resulted in net gains of SOC stocks. G eostatistical models improved the knowledge of the spatial distribution and variability of SOC in the study area with implications for SOC cycling, land management, environmental conservation and policy decisions.

PAGE 12

12 CHAPTER 1 INTRODUCTION Climate change and its effects have been well documented (EPA, 2010; Hansen, et al., 2007, Mulkey et al., 2008) and there is growing concern that future global impacts of rising carbon dioxide (CO2) and other greenhouse gas em issions from human activity may be potentially devastating. Soils play an important role in global carbon ( C ) cycling and the global soil C reservoir is estimated to be approximately four times greater than the atmospheric C reservoir and five times greater than the biotic reservoir (Jacobson et al., 2004). As a result, relatively small changes in the amount of C stored in the soil C reservoir could have a significant impact on atmospheric C concentrations, and therefore have effect on clim ate. Currently, soils are estimated to capture, or sequester approximately 60 Pg C yr1 and contribute approximately 62 Pg C yr1 to the total annual CO2 emissions resulting in a net contribution to CO2 emissions (Brady and Weil, 2008). It has long been r ecognized that land management practices and landcover/landuse ( LC/LU ) change can affect soil organic carbon ( SOC ) storage (Blair and McLean, 1917; Greenland and Nye, 1959); allowing soils to act as both a sink and a source for C affecting soil quality an d atmospheric CO2 concentrations. It is estimated that LC/LU change currently contributes approximately 1.2 Pg (12 15%) of total anthropogenic atmospheric emissions (van der Werf et al., 2009). This st udy investigates the soil C LC/LU relationship and its effect on the C cycle in a subtropical region in northeastern Florida.

PAGE 13

13 CHAPTER 2 BACKGROUND Carbon (C) is one of the most important elements of life and is abundant as many different forms throughout the Earth. The C cycle is the biogeochemical cycle by which C is transferred among the Eart hs C reservoirs The C reservoirs of greatest concern for this project include the pedosphere, terrestrial biosphere and the atmosphere. Each reservoir may act as a s ource or sink for C and the C flux between these three reservoirs is relatively rapid. In short, C enters the terrestrial biosphere via photosynthesis (becoming plant biomass and eventually soil C via decomposition), and is released back to the atmosphere via plant and microbial respiration, decomposition and fires; all of which can be modified or accelerated by human activities. Carbon is removed from the atmosphere when photosynthesis exceeds the rate of plant and microbial respiration, fires and decomposition. There is a net emission of C into the atmosphere when the opposite is true. The Lithosphere The lithosphere comprises the Earths crust and upper mantle. Although the lithosphere is the l argest C reservoir, with approximately 90,000,000 Pg C in the Earths crust, fluxes between it and the other reservoirs are very small (Jacobson, 2004). Carbon flux from the lithosphere is considered inactive on short time scales, such as hundreds of ye ars, due to its residence time of about 108 years (Broecker, 1973). Because of this, the lithosphere has received much less attention from researchers studying the partitioning of fossil fuel C between the reservoirs. The Hydrosphere Of the active reser voirs, the hydrosphere is considered the largest but least reactive C reservoir. Oceanic C exists in four main forms: dissolved inorganic carbon (DIC), dissolved organic carbon (DOC), particulate organic carbon (POC), and marine biota (Jacobson, 2004).

PAGE 14

14 C arbon dating was used to estimate the average age of deep water marine sediments, in the form of DOC, to be approximately 3400 years (Williams, 1975). In total, the hydrosphere accounts for approximately 40,300 Pg C (Jacobson, 2004). The Pedosphere The pe dosphere, or soil, is the outermost layer of the Earth. The C stored within this reservoir is a major constituent of the global C cycle and is estimated to hold approximately 3,150 Pg C (Field, 2007). Soils are perhaps the most complex and least understood ecosystem due to overlapping ecosystem processes and effect the atmospheric composition of greenhouse gases (i.e. carbon dioxide ( CO2) methane ( CH4) nitrous oxide ( N2O), soil quality and net primary productivity, among others. Because soils are capabl e of removing C from the atmosphere as well as releasing it into the atmosphere, the soil C reservoir has received considerable attention from scientists, economists and politicians around the world. The Terrestrial Biosphere The terrestrial biosphere is defined as the biomass of living organisms and vegetation on the earths land surface. For the purpose of this study, we consider only vegetation for terrestrial biosphere C reservoir estimates. Like the pedosphere, the terrestrial biosphere is also capab le of removing C from the atmosphere through photosynthesis, as well as releasing C into the atmosphere through plant respiration. Field (2007) estimates the terrestrial biosphere contains approximately 650 Pg C globally. The Atmosphere The atmosphere co ntains approximately 750 Pg C and is present mainly as CO2, with small amounts of C present as CH4, carbon monoxide (CO), and other gases. Although the atmosphere is the smallest of the C reservoirs, its importance to the C cycle is critical because of its ability to drive climate change. CO2 levels in the atmosphere are perhaps one of the most studied and well

PAGE 15

15 known quantities of the C cycle. According to the U.S. EPA (2010), atmospheric CO2 concentrations have increased by approximately 38% from 1780 (just before the industrial revolution) to 2009.

PAGE 16

16 CHAPTER 3 STUDY AREA The study was conducted in the St. Johns River (SJR) drainage which consists of the Lower, Middle and Upper St. Johns River Basins as well as the O cklawaha River Basin and is managed by the St. Johns River Water Management District (SJRWMD) The study area is located in the northeast and east c entral location of Florida (Figure 31) and has an area of approximately 2 2,267 km2 (which corresponds to ap proximately 14% of Floridas area). Floridas longest river the S t. Johns, runs through the study area and drains som e of the states major interior wetlands. Like most rivers of Florida, the St. Johns has a slow rate of flow, 0.2 km/h with an average discharge of approximately 425 m3/s (Whitney, 2004). The low flow rate can be attributed to the basins flat topography along the St. Johns River The elevation of the river surface has a difference of just 9 meters occurring over the 500 km fr om the headwaters to the mouth, about 1.8 cm/ km Elevation of the topography within the entire study area ranges from 1.2 m below mean sea level to 95 m above mean sea level. The study area receives a high amount of rainfall (up to 1600 mm annually) and has an average annual tempe rature range of 20.4 C (National Oceanic and Atmospheric Administration, 2006) Major cities within the study area include Jacksonville, Sanford, Orange Park, Green Cove Springs, Deltona, De land and Palatka. Upper St. Johns Riv er Basin Bec ause the river flows north, the Upper St. Johns River Basin (USJRB) is located at the southernmost point of the study area The USJRB measures approximately 4,504 km2 and covers portions of Indian River Osceola, Orange and Seminole counties. Wetland marshes in the USJRB once spanned over 1,618 km2 of floodplain marsh, but have been reduced by 62% due to drainage for agriculture and flood control (SJRWMD, 2007). Today the dominant LC/LU classes

PAGE 17

17 include Improved Pasture (20%), Rangeland (15%) and Freshwater Marsh/Wet Prairie (14%). Major soil types include Aquods (29%), Aqualfs (17%) and Saprists (10%). Middle St. Johns River Basin The Middle St. Johns River Basin (MSJRB) is approximately 3,058 km2 and is dominated by Urban (22%), Pineland (12%) and Rangeland (11%) landcover/landuse ( LC/LU ) classes. The dominant soil orders include Aquods (29%), Psamments (23%) and Aquents (8%). The MSJRB is fed primarily by springs, connecting lakes and storm water runoff (SJRWMD, 2009). Here, the basin spreads throughout Orange, Lake, Volusia, and Seminole Counties which are home to about two million people and major tourist attractions. The topography in this basin is a bit more diverse, with clearly distinguishable river banks t o broad, shallow lakes. As the St. Johns meet s the Wekiva River, it broadens into an approximately 91 meter wide by 2.5 meter deep river. This basin encompasses five watersheds, the Wekiva and Econlockhatchee Rivers, Howell Creek through Lake Jessup, Deep Creek through Lake Harney, and Lake Monroe. Lower St. Johns River Basin The Lower St. Johns River Basin (LSJRB) drains an area of 9,263 km2 and extends from the confluence of the St. Johns and Oklawaha rivers Welaka, north to the mouth of the S t. Johns R iver in Jacksonville. The major LC/LU classes within the LSJRB include Pineland (22%), Urban (13%), and Xeric Upland Fores t (9%). Major soil orders include Aquods (29%), Psamments (18%) and Aqualfs (11%). The LSJRB drains 12 tributaries into the SJR and e ncompasses Lake George, the states second largest lake at 2,113 km2. Lake George is a wide but shallow lake measuring 10 km at its widest span and just 2.5 m deep. The Ocala National Forest encompasses the west side of the

PAGE 18

18 lake is dominated by sandy, well drained soils and Xeric Upland Forest. Lake George is receives an additional influx of water from Salt Springs, Silver Glen Springs, and Juniper Creek. Ocklawaha River Basin At 5,442 km2 the Ocklawaha River Basin (ORB) is the second largest tributary to t he SJR within the study area. The three most dominant LC/LU classes within the basin include Pineland (13%), Urban (11%) and Xeric Upland Forest (9.2%). All LC/LU classes are represented within the ORB except for Coastal Upland. The dominant soil orders include Psamments (31%), Aquods (14%) and Udults (12%). The ORB has had a variety of land uses over the last century. The soils surrounding the river within the upper ORB are particularly fertile with large amounts of soil organic matter ( SOM ) and much of t his land was drained for use as farmland. In general, SOM is comprised of approximately 50% soil carbon ( C ) but is a heterogeneous mixture of components with a range of C concentration (Magdoff and Weil, 2004). Because of this heterogeneity, organic components of soil are better measured and discussed in terms soil organic carbon ( SOC ) Soil organic carbon is more precisely defined and can be more accurately measured than SOM via techniques such as gas combustion analysis (Brady and Weil, 2008). The early 1900s marked the beginning of the rivers transformation when 23 km2 of sawgrass marsh was drained for farming arou nd the upper O c klawaha River and additional 26 km2 of sawgrass marsh was drained around Lake Griff in for farming in the 1950s. In the 1980s the SJRWMD purchased 60.7 km2 of former muck farmland and began numerous restoration projects, which include returning the muck farmland back to natural marshlands, establishing Total Maximum Daily Loads (TMDLs) to improve the water quality as well as h arvesting phosphorus producing shad from impaired lakes. While the overall goal of these restoration projects was aimed towards imp roving water quality, these efforts also provide

PAGE 19

19 opportunity for SOC sequestration by removing land from farming practices, w hich tend to promote mineralization of SOC. Landcover/Landuse The most dominant LC/LU classes across the entire stu dy area include P ineland (16% ), U rban (12 %) and R angeland (10%) (Table 3 1) (FFWCC, 2003) It should be noted however that although the Oth er LC/LU class was omitted from the delineation of sampling, it accounts for a substantial portion (12%) of the study area. The spati al distribution of LC/LU classes across the study area is provided in Figure 3 2. Soil Properties The soil properties data was derived from the Soil Survey Geographic Database (SSURGO) Soil Data Mart (Natural Resources Conservation Service, 1961 2004). The soils within the study area are predominantly sandy in texture with some loamy/clayey areas as well as organic soils Across Florida, the dominant soil orders include Spodosols (25%), Entisols (22%), Ultisols (18% ) and Alfisols (13%). Within the study area, dominant soil orders include Spodosols (31%), Entisols (24%) and Alfisols (11%) (Table 3 2) and the spatial distribution is provided in Figure 3 3. Areas excluded from this analysis are represented on the map as whit e spaces and therefore the total area represented differs from the LC/LU summary These white spaces represent areas where soi l su rveys are not yet available or do not exist, and large bodies of water. Hydrology Hydrology is a major environmental property with influence over soil properties, such as SOC and even LC/LU classes as well As such, it is important to give a brief overvie w of the basins hydrologic condition. The study area, as well as the entire state of Florida, sits above one of the most productive aquifers in the world, the Floridan Aquifer, with a range of over 258,999

PAGE 20

20 km2. The aquifer extends up to portions of Georgi a, Alabama, and South Carolina. Several large cities depend on the aquifer for their principle source of water, including Savanna, Brunswick, Jacksonville, Tallahassee, Orlando, and St. Petersburg. Additionally, the aquifer provides water to hundreds of thousands of people in rural areas and smaller communities. An average of 11.4 billion liters of water was withdrawn from the Floridan in 1985 (USGS, 1990). The confining layer is composed mainly of limestone and dolomite, ranging from approximately 60 meter s thick in Georgia to over 1000 meters thick in parts of south Florida (USGS, 1990). T he study area also contains numerous surficial and intermediate aquifers; both are located above the deeper Floridan aquifer. These aquifers are mainly used for domestic, commercial, or small municipal supplies. They are generally unconfined aquifers and can reach depths of 122 meters in Indian River and St. Lucie Counties, but typically less than 15 meters in most areas of the study area. (USGS, 1990) Figure 3 4 shows t he location of major aquifers within Florida.

PAGE 21

21 Figure 3 1. Location and extent of study area [Adapted from St. Johns River Water Management District. 2010. Water resources geodatabase quadbasin. Palatka, Fl. St. Johns River Water Management Dis trict.]

PAGE 22

22 Figure 3 2. Spatial distributi on of landcover/landuse classes. [Adapted from Florida Fish and Wildlife Commission. 2003. Florida vegetation and land cover data derived from 2003 Landsat ETM+ imagery by B Styes et al. Office of Environment al Services, Florida Fish and Wildlife Conservation Commission, Tallahassee, Fl.]

PAGE 23

23 Figure 3 3. Spatial distribution of soil orders [Adapted from Natural Resources Conservation Service, U.S. Department of Agriculture, 1999. Soil taxonomy: a basic system of soil classification for making and interpreting soil surveys, Agriculture Handbook vol. 436, 2nd Ed. U.S. Government Printing Office, Washington, DC.]

PAGE 24

24 Figure 3 4: Aquifers of Florida [Adapted from U.S. Geological Survey (USGS). 1990. Groundwater atlas of the United States: Alabama, Florida, Georgia, South Carolina. Floridan aquifer. HA 730G. Accessible through http://pubs.usgs.gov/ha/ha730/ch_g/G text6.html.]

PAGE 25

25 Table 3 1. Landcover/landuse (LC/LU) distribution within the study area LC/LU Area (km2) Coverage (%) Citrus 413.5 1.86 Coastal Upland 1.6 0.01 Crop 751.1 3.37 Cypress Swamp 843.0 3.79 Freshwater Marsh/Wet Prairie 1,292.5 5.80 Hardwood Swamp 1,504.4 6.76 Improved Pasture 1,607.6 7.22 Mesic Upland Forest 1,135.9 5.10 Mixed Wetland Forest 1,141.9 5.13 Other 1 2,706.1 12.15 Other Ag. 188.7 0.85 Pineland 3,488.2 15.67 Rangeland 2,287.9 10.28 Shrub Swamp 775.5 3.48 Urban 2,671.9 12.00 Xeric Upland Forest 1,456.7 6.54 Total 22,266.5 100.0% 1LC/LU class designated as Other represents water, mangrove swamp, scrub mang rove, tidal flat, and bare soil. [Adapted from Florida Fish and Wildlife Commission. 2003. Florida vegetation and land cover data derived from 2003 Landsat ETM+ imagery by B Styes et al. Office of Environmental Services, Florida Fish and Wildlife Conservation Commission, Tallahassee, Fl.] Table 3 2. Distribution of soil orders within the study area Soil type Area (km 2 ) Coverage (%) Alfisols 2,540 10.7 Entisols 5,715 24.2 Histosols 2,319 9.8 Inceptisols 692 2.9 Mollisols 1,016 4.3 Spodosols 7,275 30.8 Ultisols 1,592 6.7 Vertisols 11 0.05 Urban Land 479 2.0 Water 2,019 8.5 Total 23,658 100% [Adapted from Natural Resources Conservation Service, U.S. Department of Agriculture, 1999. Soil taxonomy: a basic system of soil classification for making and interpreting soil surveys, Agriculture Handbook vol. 436, 2nd Ed. U.S. Government Printing Office, Washington, DC.]

PAGE 26

26 CHAPTER 4 QUANTIFY REGIONAL S OIL ORGANIC CARBON STOCKS FOR HISTORIC AL AND CURRENT CONDITIONS Environmental factors which exert influence on the vertical, or profile distribution of soil organic carbon ( SOC ) as well as other soil properties at the site specific/plot scale have received considerable attention from researchers globally (Rumpel et al., 2002; Jobbagy and Jackson, 2000; Stone, et al., 1993) However, research regarding environmental factors that g overn the spatial distribution of SOC at large and regional scales is less prominent. Soil carbon ( C ) can vary substantially in space as well as time and is dependent upon factors such as landuse, vegetation, geology, topography, hydrology, microbial act iv ity and climate, among others (Brady and Weil, 2008). Regional scale variability of most soil properties can be attributed primarily to differences in soil forming factors, such as elevation, vegetation and/or parent material (Brady and Weil, 2008) In Flo rida, the majority of soils are formed in sandy to loamy marine derived parent material with sand as the dominant particle size fraction. Although the range of elevation across the state (0 to 105 m) is relatively small compared to other U.S. states, the e ffect of slight changes in topography can have a profound effect on hydrology and subsequent accumulation of SOC In general, concentrations of SO C are greater near the surface (0 20 cm) of the soil profile, largely due to the decomposition of organic matt er deposited on the soil surface as plant litter (Brady and Weil, 2008; Jobbagy and Jackson, 2000) A large scale study that utilized three global databases of soil profiles sampled to 1m depth found the relative distribution for grasslands and forests to be 42% and 50% of SOC, respectively, was found in the upper 20 cm (Jobbagy and Jackson, 2000). A study based on State Soil Geographic (STAT S GO) data indicated that Florida ranks highest among all U.S. states for SOC stocks per unit area, with a

PAGE 27

27 mean of 35.3 kg C m2 (Guo et al., 2006). The study also found that Histosols and Spodosols, the two most prominent soil orders in Florida, have the highest potential to sequester C with a mean SOC of 97.6 and 9.9 kg m2, respectively. Another study, which was done i n a subtropical region of Florida by Vasques et al. (2010), also found the greatest amounts of soil C in the top soil (0 30 cm), with values ranging from 1.2 to 63.9 kg C m2 and a mean of 6.3 kg C m2. However, detailed assessment of SOC stocks in the topsoil which explain the underlying variability of soil C and environmental landscape factors within Florida are lacking, and served as motivation for this research. The goal of this study was to model the spatial and temporal variability of SOC across a subtropical region in Florida Specifically, objective 1 was to quantity historical and current SOC stocks for the upper 20 cm soil profile within the study area and identify linkages between current SO C stocks with landcover/landuse ( LC/LU ) and soil type. Materials and Methods To model and quantify historic and current SOC stocks (020 cm soil depth) within the study area, we utilized two datasets which cover a time scale of approximately 40 years (196 1 2009). Dataset 1 (DS1), also referred to as the historic dataset, represents historic conditions (1965 1996) for the respective time period. Dataset 2 (DS2), also referred to as the current dataset, represents current conditions (2008 2009) A thorough description of the datasets follows below. Historic Conditions Dataset 1 Dataset 1 is a subset of data from the Florida Soil Characterization Database (FSCD) (http:/flsoils.ifas.ufl.edu or http://TerraC.ifas.ufl.edu) and was projected in ArcGI S using the Albers Conical Equal Area map projection. Soil samples from the FSCD that fel l within the

PAGE 28

28 boundaries of the four St. Johns River ( SJR ) hydrologic basins (n=162) were clipped to the study area. The FSCD included over 1,300 site specific soil pr ofiles collected across Florida over an approximate 30 year time period (1965 1996). Over 8,300 soil horizons across the state were described and data was collected for 144 physical, chemical, biological, morphological, and taxonomic soil properties incl uding soil C and bulk density Data collection and lab analysis funding was provided by the Accelerated Soil Survey Program (United States Department of Agriculture (USDA) Natural Resources Conservation Service (NRCS) and Environmental Pedology Laboratory, Soil and Water Science Department, University of Florida. Sampling Design The original statewide FSCD s ample locations were determined based on tacit knowledge of soil surveyors at the time and the sampling locations that fell within the stud y (DS1) are provided in Figure 4 1. To our knowledge, no strategic sampling design was implemented. Sample Collection Protocol Standard profile descriptions of soil morphology were collected by horizon down to a depth of 2 m for the statewide FSCD and sam ples were collected over the course of decades (1961 2004) Documentation of site location data changed during this time span and a variety of techniques were adapted to georeference sample sites. These techniques relied on the following information: 1. Lat itude and longitude provided by sampling crews (when available) 2. Soil maps and aerial photographs on which sampling locations were marked and latitude and longitude data could be derived 3. Utilization of sampling crews notes and descriptions to project sampl e locations onto a township, section and range system (i.e., Public Land Survey System, PLSS) which then allowed latitude and longitude coordinates to be derived

PAGE 29

29 4. Soil Survey Geographic ( SSURGO) NRCS data which allowed correlation of site specific pedon data to soil map polygons Laboratory Analysis FSCD l aboratory analysis was conducted by the University of Florida Environmental Pedology Laboratory. S oil organic matter (SOM) was measured using Walkley Black dichromate extraction (WB) for mineral soils and loss on ignit ion (LOI) for organic soils. Florida specific pedo transfer functions (PTF) were developed to convert SOM ( WB ) and LOI from DS1 to SOC as measured by modern gas combustion analysis instrumentation (Myers et al., 2011). This approach harmonize d the SOC measurements and allowed for comparison of DS1 and DS2. To develop the PTF for conversion of WB and LOI to SOC, total carbon (TC), inorganic carbon (IC) and LOI were measured on a hor izonstratified sub set of archived DS1 samples (n=144) These samples included a mix of mineral and organic soils originally measured by the WB and LOI procedures and described above. Total carbon and IC were measured by gas analysis of combustion and acid reaction gases respectively on a Shimadzu TOC V/SSM 500 (carb ondioxide) CO2 analyzer. LOI was measured by heating samples to 600 Pedo transfer functions were developed for converting DS1 mineral and organic soils separately to SOC (representing the combustion method) using linear regression techniques. A robust model was fitted using iteratively re weighted least square (Venab les and Ripley, 2002) to produce a model ( R2 = 0.92; residual standard error = 0.11) between SOM ( WB ) and SOC (%) resulting in Equation 41. = 0 08 + ( 0 85 ) (4 1) with SOC: Soil organic carbon in % representing combustion method. SOMWB: Soil organic matter in % derived from Walkley Black dichromate extraction.

PAGE 30

30 Linear regression between LOI and SOC from the mineral soils of DS1 indicated a possible segme nted relationship. The Davies test (Davies, 1987) for nonconstant regression slope showed a significant indication of change in slope ar ound 75% LOI (p<0.001) in the regression graph plotting LOI and SOC (representing combustion method) A broken line r egression procedure was used to develop two equations for converting LOI to SOC (%), one for samples with LOI less than 75.8% (E quation 42) or greater than 75.8% (E quation 43) with R2 of 0.998 and residual standard error of 0.68. = 0 5 (4 2) = 26. 26 + ( 0 85 ) (4 3) with LOI: Soil organic matter derived from loss on ignition in %. Conversion of Soil Organic Carbon Concentration into Stocks Soil organic carbon (%) was converted into SOC stoc ks by using the bulk density (BD ) values measured in the field and multiplying with SOC (%) for the 20 cm profile depth (PD) (E quation 44). = ( 2000 ) /1000 (4 4) with SOC stocks: Soil organic carbon stocks in kg C m2 (0 20 cm soil profile) BD: Bulk density in g cm3 SOC: Soil organic carbon in %. PD: Profile depth ( 0.2 m) Reconstruction of Horizon Based Soil Carbon Data into Fixed Depth Layer Intervals Once DS1 was converted to SOC stocks, the samples were reconstructed into fixed d epth intervals (0 20 cm) to provide constant sample support for geostatistical analysis, which is not

PAGE 31

31 provided by soil properties measured by horizons. A depth weighted average of SOC horizonbased stocks was calculated for all observation sites where mult iple horizons occurred within the top 20 cm profile. Otherwise, SOC stock values from the top horizon were assigned to the 0 20 cm fixed top layer. Statistical and Geo statistical Analysis An exploratory data analysis (SPSS A n IBM Company. Armonk, New York), was performed and a log10 transformation was applied to DS1 to achieve an approximate Gaussian distribution to comply with geostatistical assumptions. Experimental variograms were derived to quantify the spatial dependence structure of historic SOC st ock observations and range (spatial autocorrelation), sill and nugget variances were delineated. V ariogram model s ( Appendix A and B ) were fitted using interactive fitting to match the experimental variogram. Modeling Soil Organic Carbon Stock s Developing geostatistical (kriging) models of SOC for the study area was a major objective. However, the prediction performance of kriging drops near the boundary of a study area due to the limited number of samples found near the boundaries. This is known as the edge effect. To increase modeling/kriging performance and reduce the edge effect, a 20 km buffer surrounding the study area was creat ed in ArcGIS The buffer included 240 additional sites to be included for modeling/kriging purposes. This additional set of data used for modeling is referred to as Regional Dataset 1(DS1R) and has a total of 402 samples. The prediction performance of kriged maps was assessed usi ng cross validation from DS1R (n=402) as well as validation (n=30%). To achieve this, DS1R was randomly split into a model calibration set (70%) and a model validation set (30%). A summary of the datasets representing historic conditions follows: 1. DS1 (n=162): Sites within the study area

PAGE 32

32 2. DS1R (n=402): Sites within the study area and 20 km buffer 3. Calibration Set (n=281): C hosen at random (70%) from DS1R 4. Validation Set (n=121): C hosen at random (30%) from DS1 Historic SOC stocks were quantified for the study area using ordinary point kriging (OK) and ordinary block kriging (BK). Kriging provides a solution to the problem of estimation based on a continuous model of stochastic spatial variation. It takes into account the way soil properties vary in space through the variogram model which quantifies spatial autocorrelation. In kriging, the estimates are weighted linear combinations of the data. The weights are allocated to the sample data within the neighborhood of the point (OK) or block (BK) to be estimated in such a way as to minimize the estimation of kriging variance (Webster and Olive r, 2001). In general, kriging allows us to predict, with some quantifiable certainty, the value of a variable of interest at an unsampled location by using samples of a known quantity from nearby locations. Variogram models were produced and OK and BK (We bster and Oliver, 2001) were used to interpolate SOC stocks and produce maps for the entire study area using the geostatistical software ISATIS (Geovariances Americas Inc., Avon Cedex, France ). Different SOC stock estimates were derived at pixel locati ons x using a spatial (grid) resolution of 30 m standardized to 020 cm. The block size was 30 m x 30 m with 5 x 5 estimates averaged within the block. The estimates derived from OK for soil depth increment 020 cm is denoted as SOC( x )DS1 OK20. Estimates derived from BK are denoted as SOC( x )DS1 BK --20 for depth 020 cm. All SOC stocks are expressed in units of kg C m2. The log10 SOC predictions and standard deviation maps (STD) produced in ISATIS at 30 m spatial resolution were exported i nto ArcMap 9.3 and converted to rasters. The raster calculator was used to back transform the prediction maps from log10 space into original units

PAGE 33

33 (kg C m2). To achieve the back transformation, variance maps ( ) first had to be derived from the standard deviation (STD) maps according to E quation 45 (Webster and Oliver, 2001). ( ) = (4 5) Once variance maps were produced, prediction map s could be back transformed with the raster calculator in ArcGIS using E quation 46 (Webster and Oliver, 2001). = exp { ( ) 10 + 0 5 ( ) ( ) ( 10 ) (4 6) where ( ) = the kriged estimate of the common logarithm at x and its variance is denoted as ( ) The back transformed maps were masked to the study area (excluding 20 km buffer) and the prediction performance of these SOC maps were evaluated using cross validation and validation approaches. Mean prediction error (ME) (Equation 47) and root mean square prediction error (RMSE) (Equation 48) from each data set were u sed to assess the prediction performance ( Webster and Oliver, 2001). = { y ( x y ( x) } (4 7) = { ( ) ( ) }/ (4 8) where N = number of observed values with i = 1, 2, n. = estimated values y = observed values Current Conditions Dataset 2 Dataset 2 (DS2) represents current conditions (2008/2009) and is a subset of data from a statewide project known as Rapid Assessment and Trajectory Modeling of Changes in Soil

PAGE 34

34 Carbon across a Southeastern Landscape (USDA CSREES NRI grant award 2007 35107 18368, note: CSREES NRI has been renamed to AFRI NIFA), more commonly known as the Florida Soil Carbon Project (FLSCP). The FLSCP consisted of 1014 site specific observations throughout the state of Florida DS2 was derived by projecting (Albers Conical Equal Area projection) and clipping FLSCP sites to the study area, as outlined above in DS1. To reduce edge effects for modeling of current SOC stocks the same procedures as outlined above in DS1 were used to create a regional D ataset 2 (DS2R), as well as calibration and validation datasets. A brief description of the datasets representing current conditions f ollows: 1. DS2 (n=143): Sites within the study area 2. DS2R (n=304): Sites within the study area and 20 km buf fer, used for cross validation 3. Calibration Set (n=213): C hosen at random (70%) from DS2R 4. Validation Set (n=91): Chosen at random (30%) from DS2 Sampling Design The statewide FLSCP sampling design was based on two criteria: i) A stratified random approach identified representative soillandscapes and ecosystems across Florida, and ii) half of the samples were chosen to coincide with those from the statewide FSCD representing historic conditions Two primary strata, soil suborder and LC/LU were used to capture the broad range of expected soil C variability across Florida. Both properties were selected due to their strong relationships to soil C documented in the literature (Powers et al., 2011; Brady and Weil, 2008) Soil suborder distinguishes between major soil characteristics, in particular hydrologic soil conditions (wetness/dryness) of sites. T he Florida Vegetation and Land Cover Data (2003) prepared by Florida Fish and Wildlife Conservation Commission (FFWCC), Tallahassee, FL was used to derive LC/LU cla sses. The original 43 classes as defined by the FFWCC were reclas sified into 16 broader classes ( A ppendix

PAGE 35

35 C ) for the sampling scheme to collect soil C Some classes, such as salt marshes and water, were excluded and are referred to as Other. The s tratification scheme used 13 classes of LC/LU (FFWCC, 2003) and 10 soil suborders (NRCS, 2006) Not all combinations of soil suborder and LC/LU classes exist ed in Floridas landscape and some combinations had minor (<1%) coverage, and thus were excluded fr om the sampling scheme, resulting in a total of 63 stratification classes. These classes represent approximately 69% of Floridas land area, reach nearly every county, and capture the most prominent combinations of LC/LU and soil suborders. A key goal of t he FLSCP (current) design was to examine potential change in SOC at a subset of FSCD (historic) locations. Sampling locations from FSCD that matched the stratification classes of FLSCP were randomly chosen in proportion to the surface area of each strata ( n=550). The remaining sites referred to as reconnaissance sites, were chosen randomly from the stratified land area of Florida also in proportion to the surface area of each strata. In all, 304 sites were sampled in the DS2R study area, 152 sites coincide d with the historic phase and 152 sites were reconnaissance, the sp atial distribution of sites is provided in Figure 4 2. Sample Collection Protocol The pre designed sample sites were located by differential global positioning system and f our 20 x 5.8 cm s oil cores were collected (0 20 cm depth) from each site within a 2 m diameter of the georeferenced site. The four core samples were combined into one composite sample (bulked) in the field and placed in a cooler until they could be processed. LC/LU was det ermined and/or confirmed in the field and profile morphology was observed by auguring (2 m) to determine soil sub order from observations of profile morphology.

PAGE 36

36 Sample Processing Upon return to the lab, a sub sample of the bulk sample was used to measure fresh weight and then oven dried to determine moisture content. Next the bulk sample s were air dried and a subsample was weighed, then oven dried to determine the air dried moisture content. Finally, the bulk air dried samples were sieved through 2 mm mesh sieve mixed thoroughly, and stored in plastic containers. Mineral and organic materials larger than 2 mm in diameter if present, were archived for later processing. Subsamples were collected from the air dried bulk sample and ball milled for TC, IC and LOI determination Determination for IC was performed for all samples. Lab oratory Analysis Lab measurements were performed for all samp les with 5 % replication. TC and IC were m easured by separate gas analysi s procedures and accomplished using a Shi madzu TOC V /SSM 5000 gas analyzer (Shimadzu Scientific Instrume nts, Kyoto, Japan). TC was measured by carbondioxide ( CO2) evolution using 50 500 mg of ball milled soil samples combusted at 900C. Inorganic carbon was also measured by CO2 evolution and derived by reacting 20 250 mg of ball milled soil with 42.5% phosphoric acid ( H3PO ) in the gas analyzer at 200 Soil organic carbon concentration was derived by subtracting IC from TC To derive SOC stocks (kg C m2), SOC concentration was multiplied with BD for each sample (standardized to a depth of 20 cm) according to Equation 44. Statistical and Geo statistical Analysis All versions of DS2 showed a positively skewed frequency distribution, and approximated a Gaussi an distribution after performing a log10 transformation. Experimental variograms were derived to quantify the spatial dependence structure of current SOC stock observations and range (spatial autocorrelation), sill and nugget variances were delineated. Var iogram analysis was

PAGE 37

37 performed as outlined above and variogram model s ( Appendix A ) were fit using interactive fitting to match the experimental variogram. Soil Organic Carbon Stock Estimates Log normal BK, as well as log normal OK (point), was performed on the SOC stock values from DS2 in ISATIS to characterize t he spatial variation of log10 SOC within the study area. Different SOC stock estimates were derived at pixel locations x using a spatial (grid) resolution of 30 m standardized to 020 cm. The bl ock size was 30m x 30m with 5 x 5 estimates averaged within the block. The estimates derived from OK for soil depth increment 020 cm is denoted as SOC( x )DS2 OK20. Estimates derived from BK are denoted as SOC( x )DS2 BK --20 for depth 020 cm. All SOC stocks are expressed in units of kg C m -2. The kri ged maps produced from DS2 were exported to ArcGIS converted to rasters and back transformed to original units (kg C m2) and masked to the study area, as described above for historic prediction maps. Both cross validation and validation were used to assess the predictive performance of the kriging estimates. Mean prediction error (ME) and root mean squared error ( RMSE ) from each data set were used to assess the prediction performance. In addition to log normal OK and BK, a class pedo transfer (PTF ) was also used to derive SOC stocks for current conditions in the study area using DS2. Quantification of SOC stocks by class PTF was a straightforward process. First the mean, median and standard deviations of SOC stocks were derived from the observa tions for LC/LU and soil suborder classes respectively Soil organic C stock observations were stratified by LC/LU classes and then mapped to the core study area by applying the mean SOC stock values to the grid cells in the LC/LU raster provided by Florida Fish and Wildlife Conservation Commission ( FFWCC 2003). Likewise, the observations of SOC stocks were stratified by soil suborder and then mapped to the Core Study

PAGE 38

38 Area by applying the mean SOC stock values f rom site specific observations to the map units (polygons) in the soil order layer provided by SSURGO Soil Data Mart (NRCS). Relationship between Soil Organic Carbon and E nvironmental L andscape F actors The spatially explicit relationships between field observed LC/LU and soil type respectively, and SOC stocks were investigated Post hoc pair wise group comparisons were used to test for significant differences between the environmental factors to help explain the variation of logSOC using Dunnetts T3 test (Dunnett, 1980). This test is recommended (Myers and Well, 2003) when groups of less than 10 samples are present; it does not assume equal variance between classes. Dunnetts T3 test was used for post hoc pair wise group comparisons. Results and Discussion Historic Conditions Dataset 1 : Assessment of Soil Organic Carbon Stocks Overall th e descriptive statistics for DS1 and DS1R were similar with SOC stock mean s of 5.4 and 4.7 kg C m2, respectively. The minimums and maximums for DS1 and DS1R wer e the same and ranged from 0.4 to 60 kg C m2 ( Table 4 1) 43 samples were identified as high with values greater than 9.1 kg C m2. These high SOC stock values represent ecosystems such as wetlands and swamps, which store large amounts of C The greatest SOC stock values were as follows: Hemists > Saprists > Aquepts, with mean SOC stocks of 24.4, 24.3 and 22.1 kg C m2, respectively. The relatively high amounts of SOC accumulation in these soils may be explained by the aquic soil moisture regimes in which they are found. Soils of aquic moisture regimes experience seasonal flooding in which O2 levels become reduced. The flooded environment promotes accumulation of SOC by inhibiting the metabolic processes of aerobic microorganisms that utilize soil C to fulfill basic energy requirements. Vasques et al. (2010) had similar findings from a study in another subtropical

PAGE 39

39 region of Florida, in which they found the highest SOC stocks (kg C m2) to be associated with Histosols. The smallest SOC stock value s were as follows: Psamments < Orthods < Udults with mean SOC stocks of 1.9, 2.0 and 2.1 kg C m2. Again, these findings were similar with the findings from Vasques et al. (2010), in which they concluded the lowest SOC stocks (kg C m2) were associated wit h Entisols. Psamments typically consist of quartz sands and are formed on well drained uplands, which promote the mineralization of SOC in the upper soil profile (Natural Resources Conservation Service, 2009) and lack the ability to form larger aggregates incorporating SOM Somewhat surprising was the low SOC stocks found in Orthods and Udults, which typically have higher SOC stocks. But considering the sampling depth and moisture regime these soils were sampled in, it makes sense. Spodoso ls as well as Ultisols, typically accumulate soil C in subsurface (>20cm) horizons. The presence of iron and aluminum in spodic ( Bh ) horizons of Spodosols complexes with organic material, promoting SOC accumulation (Zhengxi et al ., 1999). Similarly, Ultisols typically accum ulate SOC in subsurface horizons where the presence of clay colloids binds with SOC (Brady and Weil, 2008). T he lowest SOC stocks were associated with Entisols. Estimated Soil Organic Carbon Stocks Geostatistical analysis of DS1 was impacted by a data gap within the approxi mate west central portion of the study area, and is visible in Figure 4 1. This d ata gap, or hole, forced model ing of historic SOC stocks over a much smaller lag spacing and range compared to DS2 which had a more even spatial distribution ( Figure 42). Properties for the fi t t ed variograms are provided in Table 42. The ranges for DS1 were large, with 7,689 m (cross validation set) and 13,627 m (calibration set), respectively, indicating that regional s patial autocorrelations prevail The sill/nugget ratios were about 62% suggesting

PAGE 40

40 that short range spatial structure in SOC is complex relative to the long range trends present. It is critical to note that the standard deviations of SOC estimates were quite small with 0.8 kg C m2 across both kri ging methods (OK and BK) and evaluation modes (cross validation and calibration/validation) (Table 4 3). This pinpoints that the dispersion of SOC stocks representing historic conditions has been low across the region which was for med under historic LC/LU and management, climate and other environmental landscape factors. Table 4 3 provides descriptive statistics for historic SOC stock predictions (020 cm depth) as derived by the various kriging methods with similar results achieve d for each method. According to Journel and Huijbregts (1978) backtransformations (Eq. 4.6) are sensitive to deviations from lognormality and as a result, the estimates of SOC stocks may be biased. In general, the back transformation underestimates the ma ximums and overestimates the minimum values To test for bias, Journel and Huijbregts (1978) suggest ed comparing the means of the estimates to the means of data. Interestingly, the mean SOC kriged estimates from DS1 ( 5.2 to 5.5 kg C m2) were similar to th e field observed means from DS1 (4.7 to 5.4 kg C m2), whic h would suggest low bias for kriged estimates. However, the kriged SOC estimates were substantially overestimated for the minimums (3.4 kg C m2) and underestimated for the maximums (9.4 kg C m2) in comparison to DS1 field observed minimums ( 0.4 kg C m2) and maximums (59.7 kg C m2) Soil organic carbon stocks (0 20 cm soil depth) were quantified for 19,557 km2 across the study area, and Other LC/LU class and bodies of water were excluded from SOC stock estimates Block kriging outperformed OK in validation mode but performed equally well as OK in cross validation mode Predicted SOC stocks for BK were 0.102 Pg C for both cross validation and calibration/validation modes Ordinary kriging produced SOC stocks from 0.107

PAGE 41

41 Pg C with the cross validation dataset and 0.108 Pg C in calibration/validation mode The spatial variability of predicted SOC stocks across the study area is displayed in Figure 43 for all kriging methods. The ME expresses the bias (over or under predictions) and ranged from 0.3 to 0.6 kg C m2 for validation datasets, and was 0.003 log10 SOC kg C m2 for cross validation datasets (Table 4 3). The RMSE values for validation datasets were both 7.6 kg C m2, respectively, and wer e relatively low considering the range of SOC observations of 0.4 to 59.7 kg C m2. Current Conditions Dataset 2 Assessment of Soil Organic Carbon Stocks Mean SOC stocks for all versions of DS2 were very similar and ranged from 4.4 to 4.6 kg C m2 and suggests calibration/validation and cross validation datasets are comparable. Dataset 2 had a minimum of 0.5 kg C m2 and a maximum of 24.6 kg C m2 (Table 4 4) and 27 samples were identified as high with values greater than 9.8 kg C m2. Th ese high values represent ecosystems that are capable of storing substantial amounts of SOC. M eans show ed consistent differences to median SOC stocks, suggesting that distribution functions of SOC stocks were skewed, and was confirmed by the skewness coefficients which ranged f rom 3.1 to 6.0 kg C m2. Samples wer e distr ibuted fairly evenly across the s tudy area (Figure 4 2) and buffer with no major data gaps ( holes ) and allowed model ing over a larger range, with larger lag spacings compared to DS1. All variograms were fitted with spherical models and showed longrange spatial autocorrelations with 37,037 m and 43,398 m, respectively (Table 45). The sill to nugget ratio was 66.7% (cross validation set) and 55.6% (calibration set) and suggests that long range spat ial autocorrelation prevails in the models. The ME ranged from 5.95 10.8 kg C m2 for BK and OK calibration sets indicating a slight positive bias and 0.04 log10 SOC (kg C m2) for cross validation suggesting a slight negative bias. The RMSE were 7.4 and 11.8 kg C m2 for BK and OK with the latter nearly half the maximum observed SOC stocks in the study area.

PAGE 42

42 Estimated Soil Organic Carbon Derived from Kriging Figure 4 4 s hows the spatial variability of SOC stocks across the study area as derived by the various kriging methods (OK and BK) and Table 46 provides a summary of the SOC stock estimates derived from the se prediction maps. A total of 24,740,510 pixels (30 m x 30 m) covered the study area and provided continuous SOC stock predictions with relati vely smooth variation. Total SOC stocks for the study area derived by kriging were ~ 102 Tg (0 20 cm) for BK and ~ 107 Tg (020cm) for OK. It should be noted that, for the calculation of SOC stocks across the study area, we removed pixels representing water and Other LC/LU classes which resulted in an area of 19,557 km2. As expected, standard deviations of SOC estimated stocks were lower for BK with 2.9 kg C m2 compared to 4.4 kg C m2 for OK. The dispersion coefficients for SOC stocks representing curre nt conditions (2008/2009) were more than three times higher than the ones representing historic conditions. Th e bias of the kriged estimates was checked by comparing the mean estimates of the kriged maps to the observed means of DS2. The m ean kriged estimates (9.5 to 14.4 kg C m2) were 2 to 3 times greater than the field observed means of DS2 ( 4.4 to 4.6 kg C m2), which suggests our total SOC stock kriged estimates may have been over estimated. However, t he m aximum kriged SOC estimates (33.4 kg C m2) was greater than the observed SOC maximums from DS2 (25 kg C m2), while SOC estimates of the minimums (3.9 to 6.5 kg C m2) were much higher than the observed SOC minimums ( 0.5 to 1.1 kg C m2). A comparison of the sill/nugget ratios (~66%) tells us that long range trends prevail, and that more samples would need to be collected to minimize the nugget effect and capture short range trends. However, this was not the main aim of this study.

PAGE 43

43 All four prediction maps derived through OK and BK had similar spa tial distributions for SOC stock estimates, and several areas with relatively high SOC stocks (> 15 kg C m2) were found in the center and the elongated portion of the south of study area Because the block size for BK was relatively small, the differences between point OK and BK were marginal. However, the differences in SOC stock predictions using the smaller calibration set (n=261) compared to the larger cross validation set (n=304) were more pronounced (Figure 44) with the latter set providing more rel iable SOC stock predictions. Soil Organic Carbon Stock Assessment Derived from Class Pedo Transfer Functions Total SOC stocks derived from the LC/LU class pedo transfer function we re 89.0 Mg C or 0.089 Pg C (Table 4 7) Pineland had a moderately high SOC stock with 3.8 kg C m2 and accounted for 13.3 million Mg C (or 14.9%) of C within the core study area due to the extensive coverage of Pineland across the region (15.7%). Interestingly, Hardwood Swamps had a substantially lower coverage (6.8%) within the Core Study Area compared to Pinelands but accounted for 14.4 million Mg C (or 16.2%) of all C stored in the topsoil of the Core Study Area which can be explained by its relatively high SOC stocks (9.6 kg C m2). In contrast, Crops (2.2 kg C m2) and Xeric Upland Forest (1.9 kg C m2) showed the lowest soil C values among LC/LU classes and contributed substantially less to the soil C stocks in the topsoil within the Core Study Area. Note that three LC/LU classes (Coastal Upland, Other, and Other Ag.) could not be populated by class PTF as there was no SOC stock data available from the study to represent these classes. Thus, the absolute SOC stocks derived from the LC/LU class pedotransfer function provides an approximate estimate for the study area, and is considered conservative because total SOC stocks were derived from mean class values and not all LC/LU classes could be associated with SOC values.

PAGE 44

44 Total SOC stocks derived by soil suborder class pedotransfer function were 93.3 Mg C (0.093 Pg C) (Table 4 8 ) Aqualfs had a moderately high SOC stock class average (5.0 kg C m2) and accounted for 11.6 million Mg C (12.4% of total ) within the study area and contributed to 10.4% of the overall soil coverage. In contrast, Saprists had a lower coverage (7.5%) wit hin the study area compared to Aqualfs, but contributed to nearly (21.3%) of the total SOC stocks. The relatively high total SOC stocks for this class can be explained by its high (12.0 kg C m2) class average. Aquods had a moderate class average at 4.3 kg C m2, but also had extensive coverage across the study area and contributed the most (25.9%) towards the total SOC stocks within the study area Note that six soil suborders classes could be not populated for lack of soil samples in these classes, and combined account for 3.5% of the total study area. Soil organic C stocks by soil suborder class pedotransfer function are again considered conservative as total SOC stocks were derived from class means, not accounting for some minority soil types. The class identified as N/A represents areas where soil surveys were not mapped, not yet available, W ater bodies and cover 13.9% of the study area. Relationship s Between Soil Organic Carbon And Environmental Landscape F actors Significant SOC differences between LC/LU classes and soil sub orders were identified at the 95% confidence level using Dunnets T3 test (Table 49). In regards to LC/LU, mean SOC stocks (2008/2009) decreased in the following order (Table 4 7) : Cypress Swam p > Hardwood Swamp > Mixed Wetland Forest > Freshwater Marsh/Wetland Prairie > Shrub/Swamp > Improved Pasture > Pineland > Urban > Mesic Upland Forest > Rangeland > Citrus > Crop > Xeric Upland Forests. The highest SOC stocks were observed in swamp and wet land classes (Table 44), with Cypress Swamp having the highest mean SOC at 9.7 kg C m2. This is explained by anaerobic c onditions present in swamps and wetlands. These anaerobic conditions arise when O2 diffusion

PAGE 45

45 into soil pores is limited as soil pores become inundated with water. When O2 levels are low, anaerobic bacteria dominate, and organic matter decomposition is slowed as these anaerobic bacteria must rely on lower energy level electron acceptors for metabolic processes (Brady and Weil, 2008). Relatively h igh SOC stocks were also observed in Improved Pasture, which may be explained by the addition of fertilizers and irrigation from agricultural uses Sites in this class were typically comprised of managed grasses such as Bahia grass ( Paspalum notat um ), which tend to accumulate substantial amounts of SOC in the top soil. According to a study by Conant et al. (2001), the greatest soil C sequestration rates for managed grasslands were found in the upper 10 cm soil profile and the largest soil C increas es were due to the introduction of irrigation (Conant, 2001). They concluded that grasslands may act as a significant C sink, particularly with improved management. A relatively high proportion of grassland plant residue is comprised of root matter, which decomposes more slowly and contributes more efficiently to the formation of SOM compared to forest leaf litter (Brady and Weil, 2008). Jobbagy and Jackson (2000) suggest the higher belowground alloc ations for grasses compared to trees may explain the relatively high amounts of SOC found in grasslands and improved pastures. For example, grasses have much higher average root to shoot ratios (34) in comparison to temperate forests (~0.26) (Jobbagy and Jackson, 2000). Additionally, the rate at whic h organic material decomposes is governed by the quality of C inputs, which is often characterized by lignin content (Jackson, et al., 1996). Carbon/Nitrogen (C:N) ratios greater than 25 often promote the immobilization of SOC in microbial biomass as well as mine ralization of SOC via respiration as the microbial population grows rapidly due to the added nutrients. However, as the microbial community depletes the additional nutrient inputs, the C immobilized in the microbes cell structure is released into the

PAGE 46

46 soil as the microbial population crashes due to competition for depleting resources (Brady and Weil, 2008) Surprisingly, Urban LC/LU classes also had high mean SO C stocks and were significantly greater than Xeric Upland Forest. These findings agree with a si milar study (Pouat, 2002) that compared SOC pools of forested mineral soils from urban and rural areas in the New York City metropolitan area, and concluded forest stands in urban areas had significantly (pvalue=0.02) higher SOC stocks (0 10 0 cm soil dept h) compared with both suburban and rural forest stands. Lugo and Sanchez (1986) also found that sites switching from agriculture to urban increased soil C over a 40 year period. It should be noted th at the majority of Urban sites for this study were sample d in lawns of St. Augustine grass with irrigation systems in developments belonging to homeowner associations which require homeowners to water and fertilizer their lawns on a regular basis. As mentioned above, this combination of planted grasses, fertiliz er additions and irrigation are favorable for SOC sequestration. Conant et al. (2001) found that the introduction of irrigation alone accounted for an average annual inc rease of 5.4% SOC in grasslands across a wide variety of climates. The lowest SOC stoc ks were observed in Xeric Upland Forest and may be explained by the soils this class was sampled in. Approximately 67% of sites sampled as Xeric Upland Forest in our study were sampled in Psamments, which tend to be well drained, well aerated soils where S OM is readily m ineralized by aerobic microorganisms. Detwiler (1986) reported similar findings for a study on tropical soils and reported soil C increased in the following order: Tropical Very Dry Forest < Tropical Dry Forest < Tropical Moist Forest < Trop ical Wet Forest, with soil C values of 6.9, 10.2, 11.4 and 15.0 kg m2, respectively. These findings are not novel but illustrate the influence soil moisture has on SOC.

PAGE 47

47 A comparison of the means for soil types indicates SOC stocks (020 cm soil depth) de crease in the following order (Table 4 8) : Saprists > Aquolls > Aquepts > Aquents > Arents > Aquults > Aqualfs > Aquods > Udults > Orthods > Udalfs > Psamments. The comparison indicates SOC stocks are highest among soil types with aquic moisture regimes. Aq uic soil conditions are associated with seasonal and/or permanent inundation, in which during this time soils are virtually free of dissolved O2 because soil pore spaces are filled with water. These conditions favor accumulation of SOM as anaerobic microbial activity becomes limited Significant differences between suborders at the 95% confidence interval were observed and the results are shown in Table 410. Conclusion s Overall, the models captured the general trends of the spatial distribution of SOC at the regional scale. These trends can be identified in both DS1 and DS2 prediction maps (Figures 43 and 44) representing both historic and curre nt landscape conditions respectively Generally speaking, some of the largest stocks of SOC were predicted in the southern most portion of the study area. These relatively larger SOC stocks run northward in a somewhat consistent manner to the east central portion of the study area, near Lake George and cont inue to the mouth of the river Of the various geostatistical methods used to compare regional SOC stocks, BK proved to be the most accurate for both DS1 and DS2. The predicted SOC stock estimates diffe red between historic and current kriged maps. Historical prediction maps ranged from 3.4 to 9.4 kg C m2, whereas current prediction maps ranged from less than 1 to 30 kg C m2. On an absolute basis, total SOC stock estimates differed among sets, ranging from 102 108 Tg for DS1 and 211 320 Tg for DS2. These differences may be explained by differences in the spatial distribution of site observations, different modeling parameters between DS1 and DS2 or may even suggest the

PAGE 48

48 study area has sequestered C o ver the time period. A comparison of nugget to sill ratios (~6266%) shows that long range spatial autocorrelation prevailed in the models and suggests that major portions of the short range variability in SOC could not be explained and quantified. The inc lusion of other environmental properties, such as topography and nitrogen levels may help capture the short range variability of SOC in the models. According to the ANOVA, environmental factors such as soil type and LC/LU had significant effects on the distribution of SOC across the region. Although significant differences were found between SOC stocks of wetlands and many other LC/LU classes we must reject the 1st hypoth esis because significant differences between wetlands and all other classes were not found. The highest absolute total SOC stocks in the top soil were found in Hardwood Swamp > Pineland > Urban > Mixed Wetland Forest, contributing 14.4, 13.3, 9.9 and 8.5 Tg C towards total SOC stocks, respectively. These 4 classes accounted for over 50% of the total stocks found within the study area, which is attributed to the high mean stocks present in some classes and extensive coverage present in others. For example, although Hardwood Swamp had moderate coverage (7%) throughout the study area, it accounted for the highest total stocks due its high mean stock class average. In contrast, Pineland and Urban classes had relatively low mean SOC stocks, but extensive coverag e (16 and 12%, respectively) throughout the study area. Interestingly Hardwood Swamp had significantly higher SOC stocks compared to Urban and Pineland, but were not significantly different compared with Mixed Wetland Forest. In comparison, the lowe st total SOC stocks within the study area were found in Citrus, Crop and Xeric Upland Forest, and accounted only for 1.0, 1.7 and 2.8 Tg C, respectively. These classes had low mean stock values as well as low coverage throughout the study area and contributed approximately 6% to the total SOC stock values in the study area. Soil organic

PAGE 49

49 carbon accumulation differs by LC/LU, however, and is confounded by many other soil and environmental factors and processes which modulate C dynamics. Findings from this re search suggest that future land use and management change may lead to disparate changes in SOC across the basin.

PAGE 50

50 Figure 4 1. Spatial distribution of Dataset 1 (n=402). Circles symbolize relative soil organic c arbon (SOC) stocks (kg C m2, 020 cm depth).

PAGE 51

51 Figure 4 2. Spatial distribution of Dataset 2 (n=304) Circes symbolize relative soil o rganic carbon (SOC) stocks (kg C m2, 020cm depth).

PAGE 52

52 A B C D Figure 4 3. Kriged estimates of historic soil organi c carbon (SOC) stocks (kg C m2, 020 cm depth) derived from Dataset 1within the core study area. Maps were derived by A) Block kriging using the calibration dataset (n=402). B) Block kriging using the cross validation dataset (n=281). C) Ordinary kriging using the calibration dataset ( n=402). D) Ordinary kriging using the cross validation set (n=281).

PAGE 53

53 A B C D Figure 4 4. Kriged estimates of current soil organic carbon (SOC) stocks (kg C m2, 020 cm depth) derived from Dataset 2 within the core study area. Maps were derived by A) Block kriging for calibration set (n=261), B) Block kriging for cross validation set (n=304), C) Point kriging with calibration set (n=261), D) Point kriging with c ross validation set (n=304).

PAGE 54

54 Table 4 1. Descriptive statistics of field observed soil organic carbon (SOC) stocks derived from historic Dataset 1. DS1 DS1 R Calibration Set Validation Set Observations 162 402 281 121 Minimum 0.4 0.4 0.4 0.9 Mean 5.4 4.7 4.7 4.8 Maximum 59.7 59.7 59.1 59.7 Median 2.6 2.6 2.6 2.6 Standard Deviation 8.8 6.9 6.5 7.8 Skewness 4.0 4.5 4.4 4.6 Kurtosis 19.1 25.6 25.4 25.1 DS1: Dataset 1. Sites sampled within the 4 hydrologic basins of t he study area (core study area). DS1R: Regional Dataset 1. Sites sampled within the 4 hydrologic basins as well as a 20km buffer, used for geostatistical (kriging) models. Calibration Set: Represents a 70% split of Regional Datase t 1, used for model calibration. Validation se t: Represents a 30% split of Regional Datas et 1, used for model validation. Values represented in kg C m2, 0 20 cm profile. Table 4 2. Variogram parameters for soil organic carbon (SOC) stocks derived from Dataset 1. Datasets Kriging Type1 # of Lags Lag Spacing Model Range Sill Nugget (m) (m) (kg C m 2 ) 2 Calibration OK 1 12 4,000 Spherical 13,627 0.13 0.08 BK 2 12 4,000 Spherical 13,627 0.13 0.08 Cross Validation OK 12 3,000 Spherical 7,689 0.13 0.07 BK 12 3,000 Spherical 7,689 0.13 0.07 1OK Ordinary Kriging, 2BK Block Kriging. Sill and nugget are reported on log10 transformed SOC stocks (kg C m2, 020 cm depth).

PAGE 55

55 Table 4 3. Predictions for historic soil organic carbon (SOC) stocks derived from ordinary kriging (OK) and block krig ing (BK) using historic Dataset 1. Min Mean Max STDEV Area ME3 RMSE4 SOC Stocks Kriging Type Prediction Map (kg C m2) (km2) (kg C m2) (T g C) Block Kriging A 1 3.9 5.2 8.3 0.8 19,557 0.3 7.6 102.5 B 2 3.4 5.2 9.1 0.8 19,557 0.003 5 102.3 Ordinary Kriging C 1 4.3 5.5 8.6 0.8 19,557 0.6 7.6 108.1 D 2 3.8 5.5 9.4 0.8 19,557 0.003 5 106.9 1Estimates of soil organic carbon derived in calibration mode (n=281) the ME and RMSE are reported for validation mode. 2 Cross validation mode (n=402) 3 Mean Prediction Error 4 Root Mean Square Error 5Reported on log10 transformed soil organic carbon stocks Soil organic carbon stocks reported as Teragrams of carbon, 0 20 cm soil profile.

PAGE 56

56 Table 4 4. Descriptive statistics of field observed soil organic c arbon (SOC) stocks derived from current Dataset 2 DS2 DS2R Calibration Set Validation Set Observations 143 304 212 91 Minimum 0.5 0.5 0.5 1.1 Mean 4.6 4.4 4.4 4.5 Maximum 19.8 24.6 24.6 19.8 Median 3.2 3.3 3.3 2.9 Standard Deviation 3.6 3.4 3.3 4.1 Skewness 1.7 2.1 2.0 2.3 Kurtosis 3.1 6.0 6.0 5.9 DS2: Dataset 2. Sites sampled within the 4 hydrologic basins of the study area (core study area). DS2R: Regional Dataset 2. Sites sampled within the 4 hydrologic basins as well as a 20km buffer, used for geostatistical (kriging) models. Calibration Set: Represents a 70% split of Regional Dataset 2, used for model calibration. Validation set: Represent s a 30% split of Regional Dataset 2, used for model validation. Data reported in kg C m2, 0 20 cm soil profile. Table 4 5. Variogram parameters for soil organic carbon (SOC) stocks derived from Dataset 2 Dataset Kriging Method1 # of Lags Lag Spacing (m) Model Range (m) Sill (kg C m2)2 Nugget (kg C m2)2 Cross Validation BK 15 8,000 Spherical 37,037 0.09 0.06 OK 15 8,000 Spherical 37,037 0.09 0.06 Calibration BK 12 7,500 Spherical 43,398 0.09 0.05 OK 12 7,500 Spherical 43,398 0.09 0.05 1BK: Block kriging; OK Ordinary kriging Sill and nugget reported on log10 transformed SOC stocks (kg C m2, 020 cm)

PAGE 57

57 Table 4 6. Kriged estimates of soil organic carbon stocks derived from ordinary kriging (OK) and block kriging (BK) using Dataset 2. Min Mean Max STDEV Area ME RMSE SOC Stocks Kriging Type Map Type (kg C m2) (km2) ( Tg C ) Block Kriging A1 3.9 10.0 20.4 2.9 19,557 6.05 7.4 223.4 B 2 4.3 9.5 22.1 3.0 19,557 0.04 6 210.9 Ordinary Kriging C1 5.9 14.4 28.6 4.0 19,557 10.85 11.8 320.3 D 2 6.5 13.9 33.4 4.4 19,557 0.04 6 310.1 1 Estimates of soil organic carbon derived in calibration mode (n=261); the ME and RMSE are reported in validation mode 2 Cross validation mode (n=304) 3 Mean Prediction Error 4 Root Mean Square Error 5Unit s reported in kg C m2 6Units reported in log10 space Soil organic carbon stocks reported as T eragrams carbon 0 20 cm soil profile

PAGE 58

58 Table 4 7. Soil organic carbon stock observations using Regional Dataset 2 stratified by landcover/landuse (LC/LU) classes LC/LU Class Pixel 1 Count Area2 Mean3 Median3 STDEV3 Total Stocks4 (km 2 ) (Mg C) Citrus 459,494 413.5 2.5 2.6 0.8 1,033,861 Coastal Upland 1,828 1.6 Crop 834,589 751.1 2.2 2.1 0.7 1,652,486 Cypress Swamp 936,674 843.0 9.7 9.5 4.9 8,177,164 Freshwater Marsh/Wet Prairie 1,436,097 1,292.5 6.0 5.3 3.6 7,754,923 Hardwood Swamp 1,671,568 1,504.4 9.6 9.8 5.1 14,442,347 Improved Pasture 1,786,240 1,607.6 4.1 3.3 2.8 6,591,225 Mesic Upland Forest 1,262,100 1,135.9 3.5 2.9 1.9 3,975,615 Mixed Wetland Forest 1,268,745 1,141.9 7.4 7.8 2.7 8,449,841 Other (excluded) 3,006,723 2,706.1 Other Ag. 209,710 188.7 Pineland 3,875,771 3,488.2 3.8 2.9 2.4 13,255,136 Rangeland 2,542,135 2,287.9 3.1 2.7 1.5 7,092,556 Shrub Swamp 861,686 775.5 5.1 5.1 5.1 3,955,138 Urban 2,968,759 2,671.9 3.7 3.5 1.9 9,885,967 Xeric Upland Forest 1,618,597 1,456.7 1.9 1.5 1.0 2,767,800 Total 24,740,716 22,266.5 89,034,059 1Number of pixels in data layer (30 m x 30 m spatial resolution). 2Total area (square kilometers) of LC/LU classes derived from the 2003 FFWCC LC/LU layer. 3Mean, median and standard deviation SOC (kg C m2, 020 cm) derived from Regional Dataset 2. 4To tal stocks (Megagrams of carbon) for the Core Study Area derived by applying the SOC LC /LU class pedo transfer function. [LC/LU coverage adapted from Florida Fish and Wildlife Commission. 2003. Florida vegetation and land cover data derived from 2003 Landsat ETM+ imagery by B Styes et al. Office of Environmental Services, Florida Fish and Wildlife Conservation Commission, Ta llahassee, Fl.]

PAGE 59

59 Table 4 8. Soil organic carbon stock observations using Regional Dataset 2 stratified by soilorder classes Soil Suborder Pixel 1 Count Area2 Mean3 Median3 STDEV3 Total 4 Stock s (km 2 ) (kg C m 2 ) (kg C m 2 ) (kg C m 2 ) (Mg C) Aqualfs 2,572,303 2,315 5.0 4.0 3.1 11,575,364 Aquents 799,399 719 6.4 4.9 4.0 4,604,538 Aquepts 695,758 626 9.4 9.4 4.3 5,886,113 Aquerts 12,604 11 Aquods 6,234,019 5,611 4.3 6.4 2.5 24,125,654 Aquolls 1,104,227 994 9.8 10.1 2.7 9,739,282 Aquults 533,087 480 5.1 3.1 5.6 2,446,869 Arents 122,871 111 5.3 5.3 0.3 586,095 Hemists 594,012 535 Humods 212,810 192 Orthents 10,879 10 Orthods 925,926 833 3.1 2.5 1.9 2,583,334 Psamments 4,493,480 4,044 2.2 2.1 1.0 8,897,090 Saprists 1,838,994 1,655 12.0 11.3 5.4 19,861,135 Udalfs 34,895 31 2.4 2.1 1.2 75,373 Udepts 29,327 26 Udults 1,001,905 902 3.2 2.6 1.5 2,885,486 Umbrepts 11,523 10 N/A 3,441,146 3,097 Total 24,669,165 22,202 93,266,333 1Number of pixels in the data layer (30 m x 30 m spatial resolution) 2Total area (square kilometers) of soil suborders derived from the NRCS data layer. 3 Mean, median and standard deviation SOC (kg C m2, 0 20 cm ) values were derived f rom Regional Dataset 2. 4 Total stocks (Megagrams of carbon) for the Core Study Area derived by applying the SOC soil suborder class pedotransfer function [Soils coverage adapted from Natural Resources Conservation Service, U.S. Department of Agriculture, 1999. Soil taxonomy: a basic system of soil classification for making and interpreting soil surveys, Agriculture Handbook vol. 436, 2nd Ed. U.S. Government Printing Office, Washington, DC.]

PAGE 60

60 Table 4 9. Analysis of variance (ANOVA ) for landcover/landuse classes derived from Dataset 2 using Dunnetts T3 test LC/LU Citrus Crop Cypress Swamp Freshwater Marsh/Wet Prairie Hardwood Swamp Improved Pasture Mesic Upland Forest Mixed Wetland Forest Pineland Rangeland Urban Xeric Upland Forest Citrus * Crop * Cypress Swamp * (.066) Freshwater Marsh/ Wet Prairie Hardwood Swamp * * * * Improved Pasture * Mesic Upland Forest * (.055) Mixed Wetland Forest * * * * Pineland * Rangeland (.066) * (.053) Urban * Xeric Upland Forest * * (.055) * (.053) * indicates significance at the 95% confidence interval. P values for comparisons that were close to significant at the 95% confidence interval are shown in parenthesis.

PAGE 61

61 Table 4 10. Analysis of Variance (ANOVA) for soil suborders derived from Dataset 2 using Dunnetts T3 test Suborder Aqualfs Aquents Aquepts Aquods Aquolls Aquults Arents Orthods Psamments Saprists Udalfs Udults Aqualfs * Aquents Aquepts Aquods * * Aquolls * * * Aquults Arents (.066) * * Orthods (.066) Psamments * * * Saprists * * * Udalfs * Udults * * * indicates significance at the 95% confidence interval. P values for comparisons that were close to significant at the 95% confidence interval are shown in parenthesis.

PAGE 62

62 CHAPTER 5 QUANTIFY SOIL ORGANIC CARBON SEQUESTRATION RATES As one of the largest carbon ( C ) reservoirs, soils are capable of sequestering and storing significant amounts of C Because of this, soil C sequestration has been suggested as a means of offsetting rising anthropogenic carbondioxide ( CO2) emissions and considered an environmentally fr iendly, cost effective solution. Soil C sequestration is defined as the transfer of atmospheric C, in the form of CO2, to the soil C reservoir via humification of biomass and/or the formation of secondary ca rbonates (Lal et al., 2009). Soil organic carbon sequestration begins with utilization of CO2 by photosynthetic vegetation to create biomass. Eventually this biomass enters the soil as soil organic matter ( SOM ) which includes living biomass such as microorganisms and plant roots, dead biomass, and amorphous mixtures of complex colloidal organic substances known as soil humus. Increasing SOM, and therefore soil organic carbon ( SOC ) has a positive impact on soil properties and processes. Benefits of increasing SOM include improving soil structure, increas ing water and nutrient retention capacities, enhancing plant growth by improving soil fertility, and improving the overall soil quality. Management strategies that aim to increase SOC via C sequestration do so by increasing the density or stocks of SOC and stabilizing it in soil aggregates so that SOC is protected from microbial processes. Rates of SOC sequestration in the worlds soils depends largely on climate, soil texture and structure, temperature, precipitation and soil management. Lal (2004) estimat es that some cultivated soils have lost one half to two thirds of the original SOC pool, translating to a loss of 3040 Mg C ha1 and overall soils have contributed approximately 136 15 Pg of C to the atmosphere through depletion of SOC pools since the i ndustrial revolution. Generally speaking, soils in temperate regions have lost as much as 60% of SOC and as much as 75% in tropical regions due to conversion from natural ecosystems to agricultural uses (Lal, 2004). However, the

PAGE 63

63 potential to recover losses from degraded soils approximately equals historical losses and may be realized by proper soil management. In short, climate change and SOC sequestration on a global scale has received considerable attention from researchers around the world ( Batjes, 1996; Post and Kwon, 2000; Lal, 2004), but knowledge of environmental properties and processes that govern soil C cycling at regional scales is still somewhat limited. The objective of this chapter was to quantify SOC sequestration rates within the study area Materials and Methods Quantify Soil Organic Carbon Sequestration Rates for Collocated Sites To establish SOC sequestration rates within the study area three different modes of comparisons between historic Dataset 1 ( DS1 ) and current Dataset 2 ( DS2 ) were made. The first method (Equation 51) compared locations from DS1 that were collo cated (200 m) with DS2 on a per year basis. All historic DS1 site observations located within 200 m of current DS2 site observations were used for the comparison. The original method was to compare DS1 sites collocated within 30m of DS2 sites, but the number of sites meeting this requirement was inadequate. In all, 115 sites met the 200 m collocation criteria. ( 2009 ) ( ) = ( ) (5 1) ( ) = ( ) (5 2) YM: Year of measurement (DS1) NY: Number of years between historic and current observations (NY): 2009 YM DS1: Historic soil organic carbon dataset DS2: Current (2008/2009) soil organic car bon dataset

PAGE 64

64 SOC: Soil organic carbon stocks in g C m2 SOCsa: Soil orga nic carbon sequestration rate ( g C m2 yr1) according to method a constrained to collocated historic and current sites (xo) xo: Original geographic coordinate (x and y coordinates) of measurements Positive SOCsa values represent soil carbon gains (sequestration) and negative SOCsa values represent soil carbon losses over the considered time period (2009 YM). Quantify Soil Organic Carbon Change for Kriged Estimates A simplified comparison was made to assess spatially explicit gains and losses in SOC stocks across the entire study area. This method assumes that historic sites (DS1) pooled together covering the time period (1965 1996) provide a generic signature to represent histor ic carbon status. Essentially, historic SOC stock predictions (DS1) derived in Chapter 4 were subtracted from c urrent SOC stock prediction (DS2) maps (Figure 52 and Equations 53 and 54) ( 2009 ) ( ) = ( ) (5 3) ( 2009 ) ( ) = ( ) (5 4) with SOCsc: Soil organic carbon sequestration rate (g C m2) according to method C derived from OK and BK Note that Equations 53 and 54 only derive change in SOC stocks (kg C m2) because a single year for historic SOC stock estimates could not be assigned to pixel values. Before the simplified SOC subtraction maps could be derived, areas of high variance or uncertainty first had to be identified and removed from the analysis. This was achieved by classifying variance maps ( Figure 5 2 and Figure 53) by quantiles in ArcGIS Essentially, the variance maps show the uncertainty of the kriged estimate at each estimate (i.e., 30 m x 30 m pixel) for the respective SOC p rediction map. Classifying the variance maps by quantiles divided

PAGE 65

65 the data into equal sized subsets of data, in which the quantiles represent the boundaries between the subsets. The second quantile is often referred to as the median and because the variance maps represented lognormal data, the median approximated the mean as well. Areas of high uncertainty may arise due to the limited sampling support in certain areas, such as the data hole in DS1 that was discussed in Chapter 4, and can affect the krigin g process, causing kriged estimates in the area to be less accurate. Areas from the kriged estimates with values greater than the third quantile of the variance maps were removed from the analysis to remove areas of high variance, or uncertainty. This was achieved by implementing a conditional statement using the rast er calculator in ArcGIS (Equation 55) which classified the SOC stock variance maps for both historic and current conditions by quantiles. If the conditions are true: { ( ) ( ) } (5 5) return the values derived through subtraction of historic SOC stock estimates from current SOC stock estimates: ( ) ( ) where: = the variance of the kriged estimate q = variance value at the third quantile. Result s and Discussion Globally, soils have great potential to sequester SOC, especially in degraded soils and ecosystems where SOC has been mineralized due to LC/LU change and land use management practices. Lal (2004) concluded that the potential for SOC sequestration decreased in the following order: degraded soils > cropland > grazing lands > forestlands and permanent crops. Soil C sequestration has the potential to offset CO2 emissions, as well as improve soil quality and fertility.

PAGE 66

66 T he comparison of collocated sites (n=115) yielded a net gain in SOC stocks of 92 g C m2 yr1 across the study area with a mean of 0.8 g C m2 yr1. Site specific gains and losses were symb olized to show the relative magnitude and direction in relation to one another and are provided in Figure 5 1. The majority of sites experienced gains in S OC between sampling dates with the largest gains realized in Hardwood/Cypress Swamp and other wetland classes Overall, wetland classes accounted for ju st over 25% of the LC/LU classes in the study area. S ome of the site specific losses were substa ntial and the largest loss ( 799.5 g C m2 yr1) was documented in Saprists. Of the six soil types with aquic soil moisture regimes, three experienced net loss es in SOC stocks. Surprisingly, Saprists experienced a net loss in SOC stocks and this finding disagrees with the findings of another study in Florida by Guo et al. (2006) who reported Histosols have the highest potential to sequester C. Howeve r, it shoul d be noted that these three soil classes ( Aquents, Aquults and Saprists ) had small sample sizes (n

PAGE 67

67 The variance of the kriged estimates ranged from .38 to .66 log10 SOC for DS1 (Figure 5 2) and .30 to .55 log10 SOC for DS2 (Figure 5 3) Dataset 1 had consistently higher variance across the study area compared to DS2, which likely resulted from the spatial di stribution of historic samples. Modeling SOC was not an objective at the time historic sites were sampled and may explain why a strategic sampling method was not implemented. Additionally, OK resulted in higher variances across the study area for historic and current conditions when compared to BK. Once areas of high uncertainty were identified and removed, the kriged estimates from DS2 were subtracted from the kriged estimates of DS1 and the results are displayed in Figure 5 2. The spatial distribution and magnitude of gains and losses by the various kriging methods were similar and resulted in a range of approximately 1 to 20 kg C m2 (0 20 cm) across the study area. Net losses of SOC stocks across the study area were small and trend s could be identified. Ordinary kriging resulted in higher SOC gains compared to BK, especially near the central portion of the study area where many of the sites were sampled in wetland ecosystems with high SOC stocks (Figure 52). Findings from the comparisons of kriged estima tes of SOC change agree with and support results from the site specific comparisons Additionally, these findings suggest that soils in the study have been sequestering C over the last several decades. The relatively large increase in SOC stocks derived b y this method, compared to the collocated sites comparison, may be explained by the spatial distribution of historic sites. As mentioned earlier, the presence of the data hole in DS1 forced modeling over a much smaller range compared to DS2 and because o f this DS2 had a much longer spatial autocorrelation and may have resulted in higher averages for the kriged estimates. When compared to the field observed values, the means of the kriged estimates were very similar to their respective field observed mean s for DS1. However, the means of the kriged estimates for DS2 were 2 to 3 times

PAGE 68

68 higher than the means of their respective field observed values. Accordingly, the total stocks for the kriged estimates of DS2 were approximately 2 to 3 times higher than the total stocks from the kriged estimates of DS1. However, the field observed means for DS1 were approximately equal in comparison to the field observed means of DS2. This indicates that the kriged estimates for DS2 were likely overestimated. Conclusions Over all, the findings from this study captured general trends of SOC change across the basin. The comparison of collocated, site specific collocated resulted in a mean of 0.8 kg C m2 yr1 for all soils. This suggests soils in the study area have acted as a si nk for C over the last 50 year s. Additionally, the positive direction of SOC sequestration documented in this study agrees with findings from literature, and one such study concluded that North America has acted as a net C sink with a mean uptake of 1.7 0.5 Pg C yr1 (Fan et al., 1998). There has been some disagreement on whether this net sink was predominantly oc eanic or terrestrial, but studies (Ciais et al., 1995; Keeling et al., 1996) have concluded the sink is primarily terrestrial uptake. Although the direction of SOC sequestration agreed with findings from the literature, the magnitude of SOC sequestration was often substantially greater in some comparisons from this study when compared to findings from literature. Schlesinger (1990) documented acc retion rates as high as 10.0 g C m2 yr1 in forest soils and an average increase of 2.4 g C m2 across all ecosystems. In comparison, the highest accretion rates from this study were 93.5 g C m2 yr1 with an overall average increase of 0.8 g C m2yr1 across all soil types. Findings from the simplified co mparison of kriged estimates also agre ed with the previous findings from the collocated site comparison. The highest SOC sequestration rates (25.4 kg C m2) from the simplified comparison were derived by OK and occurred in the area just southeast of

PAGE 69

69 Lake George. Small losses ( 0.6 kg C m2) were estimated by BK for some small portions of the study area. No SOC losses were estimated with OK. Among the methods used to compare SOC change across the stud y area, the collocated comparisons were more accurate and both gains and losses of SOC were identified (Figure 5 1). Conclusions from the simplified comparison of kriged estimates are limited because the minimums are typically overestimated and the maximums underestimated by the back transformation process Additionally, the different spatial dependence structures of historic and current datasets may explain why the SOC stock losses derived from the simplified subtraction method were substantially smaller w hen compared to the losses from the collocated site comparison.

PAGE 70

70 Figure 5 1. Soil organic carbon gains and losses (g C m2 yr1) derived from Dataset 1 and Dataset 2 collocated sites (200 m) within the study area

PAGE 71

71 A B C D Figure 5 2. Variance maps derived from Dataset 1 soil organic carbon (log10) estimates classified by quantiles. A) Calibr ation set (n=281) using block kriging. B) Cross valid ation set (n=402) using block kriging. C) Calibration set (n=281) using ordinary kriging. D) Cross validation set (n=402) using ordinary kriging.

PAGE 72

72 A B C D Figure 5 3. Variance maps derived from Dataset 2 soil organic carbon (Log10) estimates classified by quantiles. A) Calibr ation set (n=261) using block kriging. B) Cross valid ation set (n=304) using block kriging. C) Calibration set (n=261) using ordinary kriging. D) Cross validation set (n=304) using ordinary kriging.

PAGE 73

73 A B C D Figure 5 4. Series of prediction maps showing SOC stock gains and losses (kg C m2) within the study area. Areas with high uncertainty were ignored. SOC change derived by subtracting historic prediction maps from current prediction maps using: A) Block kriging maps derived with calibration datasets. B) Block kriging maps derived with cr oss validation sets. C) Ordinary kriging maps derived with calibration dataset. D) Ordinary kriging derived with cross validation dataset.

PAGE 74

74 Table 5 1. Soil organic carbon (SOC) accumulation rates by soil suborder for 200 m collocated sites N Mean Median S TD 1 Min Max Aqualf s 17 23.1 49.0 138.8 468.4 216.7 Aquent s 3 97.3 106.9 168.6 260.9 75.8 Aquods 46 6.6 24.6 122.1 499.1 253.9 Aquoll s 2 93.5 93.5 61.8 49.8 137.2 Aquult s 2 52.0 52.0 27.9 71.7 32.3 Orthods 6 16.8 32.6 61.7 73.3 93.1 Psamment s 23 17.3 10.6 35.5 66.1 91.7 Saprist s 3 364.4 180.1 378.2 799.5 113.7 Udult s 13 31.0 17.0 69.6 61.1 242.4 Total 115 0.8 17.0 129.6 799.5 253.9 1Standard deviation Units reported in g C m2 yr1, 0 20 cm profile Derived by comparing Dataset 1 sites that were collocated with Dataset 2

PAGE 75

75 CHAPTER 6 ASSESSMENT OF THE EFFECT OF LANDCOVER/LANDUSE CHANGE ON REGIONAL SOIL ORGANIC CARBON STOCKS Accurately quantifying changes in soil organic carbon ( SOC ) stocks from landcover/landuse ( L C/LU ) change is important for determining the impact of anthropogenic decisions regarding landuse and management, such as urban developm ent and agricultural practices. There is growing concern that the effect of landuse change on soils may substantially i ncrease carbondioxide ( CO2) levels in the atmosphere. For instance, the conversion of native grasslands and forest lands to cultivated agricultural soils in various states across the U.S. was reported to have resulted in a loss of approximately 3075% of the SOC pool, particularly on soils where conventional tillage is used (Follet et al., 2009). The sink capacity of soils for CO2 can be greatly enhanced by restoring degraded lands and ecosystems and converting marginal agricultural lands to more sustaina ble land uses through adoption of site specific best management practices (BMPs). Best management practices aim to improve soil quality and fertility by increasing SOC through conservation tillage, minimizing erosion, use of cover crops, crop rotations and restoration of degraded lands through landuse change (Lal, 2004). For example, Guo and Gifford (2002) found in a meta analysis that carbon ( C ) losses from soils can be countered by converting agricultural lands back to native land uses, resulting in increased soil C stocks. The meta analysis showed that significant increases in soil C stocks can occur by converting forests to pasture (8%), crop to pasture (19%), and 53% for crop to secondary forest (Gu o and Gifford, 2002). Lal and Follett (2009) estimate d the adoption of management practices such as reduced till age and notillage could sequester 45 to 98 Tg C yr1 in the United States alone. In a ddition to improving the overall health of soils, increasing SOC levels will help mitigate CO2 emissions.

PAGE 76

76 Schulze and Freibauer (2005) suggested that land use shifts in conjunction with climate change have major effects on soil C stocks and their distribut ion. According to a review of 31 different studies by Post and Kwon (2000), the rate at which C may accumulate in soil and the length of time for soil C sequestration is highly variable among land use changes. Average rates of soil C accumulation for fores t and grassland establishment after agricultural use were 33.8 and 33.2 g C m2 y r1, respectively (P ost and Kwon, 2000). Landcover/landuse change may show different response in SOC sequestration due to site specific characteristics determined by soils, pa rent material, topography, hydrology, and geography. The objective of this chapter was to assess the effect of LC/LU change on SOC within the study area. Materials and Methods Comparison of Historic and Current Landcover/Landuse Conditions To assess the effect of LC/LU change on SOC stocks in the study area historic al conditions were compared to current conditions. It was necessary to harmonize LC/LU classes characterizing historic and current conditions. Only historic Dataset 1 ( DS1 ) sites that were collocated within 200 m of current Dataset 2 ( DS2 ) sites were used for the comparison. Unfortunately, no LC/LU classification scheme was implemented during the original statewide Florida Soil Characterization Database ( FSCD ) sampling efforts. To overcome this limitation, digital historic aerial photographs and images from the University of Florida (UF) Map and Digital Image Library were retrieved and analyzed for the collocated DS1 sites Outside of the National Achieves and Records Administration, the UF Map and Digital Image Library houses the largest and most complete collection of historic aerial photographs of Florida, over 160,000 archived photos that have been scanned and are available online at http://ufdc.ufl.edu/aerials. C omplete historic aerial coverage for the entire state was not available

PAGE 77

77 for every location and time period, and in some ca ses photographs from the closest available date were used to classify LC/LU at a particular location in time Digital images from Google Earth were also used to track LC/LU changes for collocated sites. Complete coverage for the entir e state dating back to the mid 90s is available using the time slider in Google Earth Classification Scheme To overcome the limitations of accurately interpreting h istoric imagery and differentiating between similar LC/LU classes, a classification scheme was developed and applied to DS1 and DS2 The new classification scheme (Table 6 1) combined similar LC/LU classes such as M ixed Wetland F orest and H ardwood S wamp, and reduced the number of classes in the study area from 16 to 10. This broader, more generalized classification scheme also helped improve the datasets robustness for underrepresented classes In all, 120 site observations met the 200 m c ollocation criteria. However, 5 were discar ded from the analysis either because LC/LU classes could no t be assigned to historic DS1 site due to lack of historic al aerial coverage, or the current DS2 site was resampled in a different LC/LU class compared to the origi nal LC/LU. In other words, the location of the collocated site (DS2) was never in the same LC/LU class as the historical site (DS1), and therefore not possible to assess the effects of LC/LU change between the two sites Once LC/LU classes for DS1 were identified, the LC/LU change analysis could be completed. A confusion matrix was created to document both LC/LU change as well as change in SOC stocks (Table 6 4). Results and Discussion A summary of the LC/LU conversions for collocated sites is provided in T able 6 2 and Table 63. In all, 6 LC/LU classes experienced LC/LU change. Rangeland/Dry Prairie

PAGE 78

78 experienced the most change among classes and changed from Rangeland/Dry Prairie to Pasture/Other Ag., Pineland, Row/Field Crop and Urban. The relatively high number of LC/LU changes experienced within this class may be explained by the nature of the class. Rangeland/Dry Prairie was a challenge and in a sense, a catch all class, combining several classes from the 2003 Florida Fish and Wildlife Conservation Com mission ( FFWCC ) classification scheme which include: Dry Prairie, Shrub and Brushland, Grassland and Unimproved Pasture and many of the sites sampled in this class could have possibly fit into other classes. Additionally Rangeland/Dry Prairie sites were s ampled in a wide variety of soil suborders. The remaining classes that experienced LC/LU change include d Pineland to Rangeland/Dry Prairie, Pasture/Other Ag. to Rangeland/Dry Prairie, Coastal Upland to Urban, Mesic Upland Forest to Urban and Xeric Upland Forest to Urban. Four LC/LU classes did not experienc e any LC/LU change and include d Freshwater Marsh/Wet Prairie, Hardwood/Cypress Swamp, Row/Field Crop and Urban. Of these four classes, Row/Field Crop was the only class that experienced net SOC losses, although minor at 1.5 g C m2 yr1 while remaining in the same LC/LU class. Other studies have also shown that continued use of land for agricultural purposes, particularly when conventional tillage is used, results in a net loss of soil C. For ex ample, one study concluded that cultivated soils on average, consistently had 20% less soil C compared to uncultivated soils (Mann, 1986). Davidson and Ackerman (1993) reviewed data from several studies regarding the impact of cultivation on soil C and co ncluded that cultivation of soils lead to a loss of 2040% of soil C on average They also concluded that the majority of soil C lost occurred within the first few years after cultivation. Both Hardwood/Cypress Swamp and Freshwater Marsh/Wetland Prairie L C/LU classes increased SOC sto cks on average, with gains of 50.9 and 30.7 g C m2 yr1, respectively. These

PAGE 79

79 findings are similar to the findings of Gosselink and P endleton (1984), who derived average accumulation rates of 0.48 t C ha1 yr1 for organic soils in Louisiana. This is not surprising considering both classes were sampled entirely (100%) on soils with aquic moisture regimes, where anaerobic conditions are found and tend to promote SOC accumulation. Pineland sites that remained in Pineland experienced a net i ncrease in mean SOC stocks (21 .1 g C m2 yr1) and had moderate variance among sites. Pineland also experienced small increases (5 .4 g C m2 yr1) in mean SOC stocks when converted to Rangeland/Dry Prairie. This finding is somew hat surprising and disagrees with findings from Post and Kwon (2000), who found that sites switching from forests to unimproved pasture lost 17.4 g C m2 yr1. This discrepancy from the literature may be explained by the wide variety of Pineland ecosystem s under different soil types, hydric regimes and management that were sampled For example, the classification of Pineland did not differentiate managed pines from sandhill pine or f latwoods pine. Additionally, Pineland class was sampled in seven different soils with Aquods being the most prominent (56%) and may also explain the high variance found in this class. Urban sites posed significant classification challenges as well, as many sites were considered high impact urban but other sites were borderline Rangeland/D ry P rairie as they occurr ed in somewhat rural areas, but ultimately fell under Urban classification Urban sites that remained Urban ex perienced a net increase in SOC ( 43.8 g C m2 yr1) I nterestingly all LC/LU classes that were converted to Urban, except Mesic Upland Forest, resulted in a net accumulation of SOC stocks. Classes that experienced net SOC gains after converting to Urban include d Coastal Upland, Rangeland/Dry Prairie and Xeric Upland Fores t. Anthropogenic factors, such as establishment of lawns, irrigation and fertilization may explain why most classes accumulated SOC after converting to Urban and most of the sites classified as Urban were sampled in lawns

PAGE 80

80 from communities belonging to home owner associations. Many homeowner associations in Florida require homeowners to maintain pristine lawns which generally require increased irrigation and fertilization. While the increased irrigation and nutrient inputs may have a negative impact on the st ates water resources, the increased soil moisture and nutrient additions promote SOC accumulation. Pouyat et al (2002) also found urban soils were consistently higher in SOC by about 30% (p = 0.03) compared to rural fores t stands and they concluded that urbanization can substantially affect soil C pools. Pouyat et al. (2002) also attributed the increase in soil C to human activities, such as irrigation, in addition to earth worms, which were prominent in their study area. Although earthworms have been found to accelerate organic matter decomposition, they also promote production of soil aggregates, which not only increases storage of soil C, but also protects it from microbial attack (Martin, 1991). The discrepancy in direction of SOC stocks upon conversi on to Urban between Xeric Upland Forest and Mesic Upland Forest is explained by the environmental conditions in which they are found, p articularly soil conditions. Fifty percent of sites sampled in Mesic Upland Forest were sampled in soils with aquic moisture regimes, such as Aquods and Aqualfs In contrast, 50% of sites sampled in Xeric Upland Forest were sampled in Psamments, which tend to form in uphill landscape positions well drained topographic settings and are typically comprised of oak scrub, sand pine scrub and sandhill communities. Additionally, s ites that remained in Upland Xeric Forest also resulted in net SOC losses on average ( 129.5 g C m2 yr1) Several constraints were imposed on the analysis. First, as mentioned in the description of DS1 historic site locations were reconstructed from field notes and site descriptions taken by the sampling crew at the time which may have impaired the positional accuracy of historic site locations. Lack of positional accuracy in addition to the absence of a LC/LU scheme for DS1

PAGE 81

81 may have confounded the ability to identify collocated sites. Additionally, some comparisons between collocated sites were made at distances up to 200 m, and may have resulted in a mismatch between comparisons, particularly soil typ es. For example, some historic sites may have been originally sampled as a Pineland Aquods, but the current site may have been re sampled as a Pineland Orthods. This type of mismatch could have resulted from errors in the Soil Survey Geographic Natur al Resources Conservation Commission ( SSURGO NRCS ) soil map units as well as from limitations encountered by the field crew. Soil map units defined in SSURGO NRCS may contain several components (Soil Series) where only the percentage of components is defined, but not their spatial distribution within a map unit. For example, often within a soil map unit, a particular area of interest may be delineated as Aquods, but in reality this area may contain both Aquods and Orthods, but dominated by Aquods. In su ch cases, it is not practical for field crews to delineate every soil type within an area, but rather the dominate soil. A particular pedon in this area could have been inadvertently sampled in Orthods as opposed to Aquods. Additionally, the field crew ma y have encountered pro perty ownership issues, and access to private property not granted in some cases In this case the field crew would have had to locate a representative pedon within the appropriate LC/LU class as near as possible to the historic site, which often proved to be challenging. Conclusion s The direction and magnitude of SOC change is governed by many environmental factors when vegetation and landuse management pr actices are altered. Increasing SOC inputs, the ease with which organic material is decomposed and the physical protection of SOC in aggregates are all important factors for increasing SOC storage. The small scale spatial variability in SOC can be substant ial, particularly in Florida. For example, the difference of just a few meters in elevation may have a significant effect on soil moisture and often can be the difference between

PAGE 82

82 aquic vs. udic soil moisture regimes. High variation was observed in the rate s of SOC change in some classes and may be explained small sample sizes in some classes, the collocation distance of sites (up to 200 m in some cases) as well as differences in the influence of environmental factors outlined above. Overall, the LC/LU chan ge analysis captured general trends in the direction and magnitude of SOC sequestration in the study area Most classes, with a few exceptions, experienced net positive SOC gains which suggest soils across the study area have acted as a C sink between hist oric and current sampling dates Some of the largest gains in SOC resulting from LC/LU conversion were found in classes that switched to Urban, with Xeric Upland Forest being the greatest ( 46.5 g C m2 yr1) and likely due to anthropogenic factors, such as irrigation and nutrient inputs Other studies have also observed significant SOC increases in urban lands which were also attributed to anthropogenic factors (Martin, 1991; Pouyat, 2002). As mentioned above, some classes in the LC/LU change analysis had relatively high variance ion regards to SOC change and may be explained by environmental factors, such as the type of soil these classes were sampled in For example, Pineland was sampled in 7 different soil types across the study area and the soil moistur e regimes ranged from aquic to udic. Aquic soil moisture regimes are typically free of dissolved oxygen for an extended duration, typically more than a few days, and tend to promote SOC accretion. In contrast, udic soil moisture regimes are much drier and aerobic conditions are present throughout the year Differences in soil moisture regimes have a profound influence on SOC and may explain the high variance in Pineland. Soil organic carbon cycling in agricultural lands, particularly where conventional til lage is used, has been a major concern of researchers globally for some time. Losses as great as 50% in the upper 20 cm soil profile have been observed over a time period of 3050 years due to

PAGE 83

83 cultivation (Post and Kwon, 2000). Although not as substantial, findings from this study agree with the literature and losses in SOC ( 1.5 g C m2 yr1) were in observed in soils where tillage is used (Row/Field Crop). V arious other studies have also reported that terrestrial systems in the northern hemisphere have been accumulating CO2 and storing it in biomass and soils at rates of about 1 2 Pg C yr1 over the last several decades ( Ciais et al., 1995; Fan et al., 1998). Some studies have hypothesized the C sink in the northern hemisphere is related to LC/LU change and findings from this study support this hypothesis. Changes in SOC across the study area may be attributed to both abiotic and biotic environmental factors, such as water table depth, water holding capacity, soil particle size, vegetation, LC/LU change and even microbial activity. Although the data is not robust enough to precisely quantify the amount of SOC being accumulated at the plot level or even the regional scale, the findings suggest that overall, soils in the study area have acted as a net sink f or soil C over the last several decades. Additional long term studies addressing SOC cycling, particularly where LC/LU change has occurred, would be valuable in fostering our understanding of C dynamics in Florida. These findings provide valuable insight r egarding the role of soils in the study area in regards to SOC sequestration.

PAGE 84

84 Table 6 1. Landcover/landuse (LC/LU) change classification scheme Original LC/LU Classification New LC/LU Classification N 1 Coastal Upland Coastal Upland 2 Citrus Row/Field Crop 2 Crop Row/Field Crop 2 Other Agriculture Pasture/Other Agriculture 12 Freshwater Marsh/Wet Prairie Freshwater Marsh/Wet Prairie 2 Hardwood Swamp Hardwood/Cypress Swamp 9 Mixed Wetland Forest Hardwood/Cypress Swamp 9 Shrub Swamp Hardwood/Cypress Swamp 9 Cypress Swamp Hardwood/Cypress Swamp 9 Improved Pasture Pasture/Other Agriculture 12 Rangeland Rangeland/Dry Prairie 27 Pineland Pineland 22 Urban Urban 2 Xeric Upland Forest Xeric Upland Forest 10 Mesic Upland Forest Mesic Upland Forest 10 1Number of sites from Dataset 2 [ LC/LU classification scheme adapted from the Florida Fish and Wildlife Conservation Commission. 2003. Florida vegetation and land cover data derived from 2003 Landsat ETM+ imagery by Styles et al. O ffice of Environmental Services, Florida Fish and Wildlife Conservation Commission, Tallahassee, Fl.]

PAGE 85

85 Table 6 2. Confusion matrix showing the soil organic carbon change by landcover/landuse in the study area over the study period DS22 Coastal Upland Freshw ater Marsh/Wet Prairie Hardwood /Cypress Swamp Mesic Upland Forest Pasture/ Other Ag. P ineland Rang e land /Dry Prairie Row/ Field Crop U rban Xeric Upland Forest DS11 Coastal Upland 37.1 /37 .1 / 0.7 Fresh w ater Marsh/Wet Prairie 30.7/30.7 /44 .0 Hardwood/ Cypress Swamp 50.9/54.9 /10 6.9 Mesic Upland Forest 34.2 /24 .1 /3 1.7 6.7/ 39.9/88 .0 Pasture/ Other Ag. 42.4/33.8 /140.2 8.5/3.5/5 2.5 Pineland 21.1/22.5 /8 3.8 5 .4 /12 .0 /3 2.5 Rangeland/ Dry Prairie 20.1/53.4 /7 4.5 2.7 /17 .2 /5 2.5 3.5 / 23.1 /90 .0 20.2/52.6 /6 7.8 8.5 /27 .0 /6 1.5 Row/Field Crop 1.5/ 1.5 /4 3.5 Urban 43.8/43.8/ 9 .4 Xeric Upland Forest 46.5/42 .2 / 35.0 129.5 / 36.3 /185.6 1DS1: Dataset 1 representing historic landcover/landuse conditions. 2DS2: Dataset 2 representing current conditions. Cell values represent: Mean/Median /Standard Deviation SOC sequestration rates (g C m2 yr1, 0 20 cm ).

PAGE 86

86 Table 6 3. Confusion matrix showing the average distance and number of sites undergoing landcover/landuse change within the study area over the study period DS22 Coastal Upland Freshwater Marsh/Wet Prairie Hardwood /Cypress Swamp Mesic Upland Forest Pasture/ Other Ag. Pineland Rang e land /Dry Prairie Row/ Field Crop Urban Xeric Upland Forest DS11 Coastal Upland 60/2 Freshw ater Marsh/Wet Prairie 50/2 Hardwood/ Cypress Swamp 77/9 Mesic Upland Forest 23/7 74/3 Pasture/ Other Ag. 38/9 6/3 Pineland 25/18 35/4 Rangeland/ Dry Prairie 17/3 11/6 97/9 4/3 49/6 Row/Field Crop 26/2 Urban 14/2 Xeric Upland Forest 71/4 44/6 1Dataset 1 representing historic conditions. 2Dataset 2 representing current (2008/2009) conditions. Cell values = average distance of sites in each class / number of sites in each class.

PAGE 87

87 APPENDIX A DATASET 1 VARIOGRAMS Figure A 1. Variogram of log transformed soil organic carbon stocks (kg C m2) used for ordinary kriging (OK) and block kriging (BK) using the calibration set from Dataset 1 (historic conditions).

PAGE 88

88 Figure A 2. Variogram of log transformed soil organic carbon stocks (kg C m2) used for ordinary kriging (OK) and block kri ging (BK) using the cross validation set from Dataset 1 (historic conditions).

PAGE 89

89 APPENDIX B DATASET 2 VARIOGRAMS Figure B 1. Variogram of log transformed soil organic carbon stocks (kg C m2) used for ordinary kriging (OK) and block kriging (BK) using the calibration set from Dataset 2.

PAGE 90

90 Figure B 2. Variogram of log transformed soil organic carbon stocks (kg C m2) used for ordinary kriging (OK) and block kriging (BK) using the cross validation set from Dataset 2.

PAGE 91

91 AP PENDIX C FLORIDA SOIL CARBON PROJECT LANDCOVER/LA NDUSE RECLASSIFACTION SCHEME Table C 1. Reclassification scheme for 2003 landcover/landuse layer Old Value Old Class New Value New Class 1 Coastal Strand 140 Coastal Upland 2 Sand/Beach 140 Coastal Upland 3 Xeric Oak Scrub 70 Xeric Upland forest 4 Sand Pine Scrub 70 Xeric Upland forest 5 Sandhill 70 Xeric Upland forest 6 Dry Prairie 50 Rangeland 7 Mixed Pine Hardwood Forest 80 Mesic Upland forest 8 Hardwood Hammocks and Forest 80 Mesic Upland forest 9 Pinelands 10 Pinelands 10 Cabbage Palm Live Oak Hammock 80 Mesic Upland forest 11 Tropical Hardwood Hammock 80 Mesic Upland forest 12 Freshwater Marsh and Wet Prairie 90 Freshwater Marsh and Wet Prairie 13 Sawgrass Marsh 90 Freshwater Marsh and Wet Prairie 14 Cattail Marsh 90 Freshwater Marsh and Wet Prairie 15 Shrub Swamp 100 Shrub Swamp 16 Bay Swamp 110 Hardwood Swamp 17 Cypress Swamp 130 Cypress Swamp 18 Cypress/Pine/Cabbage Palm 130 Cypress Swamp 19 Mixed Wetland Forest 120 Mixed Wetland Forest 20 Hardwood Swamp 110 Hardwood Swamp 21 Hydric Hammock 120 Mixed Wetland Forest 22 Bottomland Hardwood Forest 120 Mixed Wetland Forest 23 Salt Marsh 170 To be excluded 24 Mangrove Swamp 170 To be excluded 25 Scrub Mangrove 170 To be excluded 26 Tidal Flat 170 To be excluded 27 Open Water 170 To be excluded 28 Shrub and Brushland 50 Rangeland 29 Grassland 50 Rangeland 30 Bare Soil/Clearcut 170 To be excluded 31 Improved Pasture 40 Improved pasture 32 Unimproved Pasture 50 Rangeland 33 Sugarcane 20 Crop 34 Citrus 30 Citrus 35 Row/Field Crops 20 Crop 36 Other Agriculture 60 Other Agriculture 40 Brazilian Pepper 150 Exotic plants 41 High Impact Urban 160 Urban 42 Low Impact Urban 160 Urban 43 Extractive 170 To be excluded

PAGE 92

92 LIST OF REFERNCES Alavalapati, J. 2008. Role of forest s in r educing greenhouse gases In Opportunities for greenhouse gas reduction through forestry and agriculture in Florida Mulkey, S., J. Alavalapati, A. Hodges, A.C. Wilkie, and S. Grunwald.. University of Florida, School of Natural Resources. Retrieved January. 20: 2008. Blair, A.W. and H.C., McLean. 1917. Total nitrogen and carbon in cultivated land and land abandoned to grass and weeds. Soil Sci 4:283294. Brady, N.C., and R.R. Weil. 2008. Soil organic matter. In The nature and properties of soils. Prentice Hall, New Jersey. 503 531. Broecker, W.S. 1973. Factors controlling CO2 content in the oceans and atmosphere. p. 32. In Brookhaven symposia in biology. Dunnett, C.W. 1980. Pairwise Multiple Comparisons in the Unequal Variance Case. Journal of the American Statistical Association. 75(372): 796800. Ciais, P., P.P. Tans, M. Trolier, J.W.C. White and R.J. Francey. 1995. A l arge northern h emisphere t errestrial CO2 s ink i ndicated by the 13C/12C r atio of a tmospheric CO2. Science. 269(5227): 1098 1102. Conant, R.T., K. Paustian, and E.T. Elliott. 2001. Grassland management and conversion into grassland: Effects on soil carbon. Ecological Applications. 11(2): 343355. Davidson, E.A., and I.L. Ackerman. 1993. Changes in soil carbon inventories following cultivation of previously untilled soils. Biogeochemistry. 20(3): 161193. Davies, R.B. 1987. Hypothesis testing when a nuisance parameter is present only under the alternative. Biometrik a. 74(1): 33 43 Detwiler, R.P. 1986. Land use change and the global carbon cycle: the role of tropical soils. Biogeochemistry. 2(1): 6793. EcoAsset Solutions, Land and Timber Services Group, and University of South Florida. 2009. Carbon capture and se questration study of Florida board of trustees land. Environment Florida. 2007. The carbon boom. Environment Florida Research and Policy Center. Tallahassee FL. Fan, S., M. Gloor, J. Mahlman, S. Pacala, J. Sarmiento, T. Takahashi, and P. Tans. 1998. A Large Terrestrial Carbon Sink in North America Implied by Atmospheric and Oceanic Carbon Dioxide Data and Models. Science. 282(5388): 442 446. Field, C.B., D.B. Lobell, H.A. Peters, and N.R. Chiariello. 2007. Feedbacks of Terrestrial Ecosystems to Clima te Change *. Annu. Rev. Environ. Resourc. 32(1): 129.

PAGE 93

93 Florida Fish and Wildlife Commission. 2003. Florida vegetation and land cover data derived from 2003 Landsat ETM+ imagery by B Styes et al. Office of Environmental Services, Florida Fish and Wildlif e Conservation Commission, Tallahassee, Fl. Follet, R.F., J.M. Kimble E.G. Prussner, S. SamsonLiebig and S. Waltman. 2009. Soil c arbon s equestration and the greenhouse e ffect: co eds.: Rattan Lal and Ronald F. Follett. American Society of Agronomy. Ch. 3. Gosselink, J., and E.C. Pendleton. 1984. The Ecology of Delta Marshes of Coastal Louisiana: A Community Profile. U.S. Department of the Army. Corps of Engineers. Greenland, D.J., and P.H. Nye. 1959. Increases in the carbon and nitrogen contents of tropical soils under natural fallows. European Journal of Soil Science. 10(2): 284299. Grunwald S. 2008. Role of soils to sequester carbon. In Opportunities for greenhouse gas reduction through fore s try and agriculture in Florida. Mulkey, S., Alavalapati, J., Hodges, A., Wilkie, A.C., and Grunwald, S. 2008.University of Florida, School of Natural Resources and Environment Department of Environmental Defense, Washington D.C. Guo, L.B., and R.M. Gif ford. 2002. Soil carbon stocks and land use change: a meta analysis. Global Change Biology. 8(4): 345360. Guo, Y., R. Amundson, P. Gong, and Q. Yu. 2006. Quantity and spatial variability of soil carbon in the conterminous united states. Soil Science Soci ety of America 70: 590600. Online. soils under natural fallows. J Soil Sci 10:284299 Harris, W.G ., and K.A. Hollien. 2000. Changes across artificial E Bh boundaries formed under simulated fluctuating water tables. Soil Science Society of America. J 2 000 64: 967973 Hansen, et al. 2005. NASA Goddard Institute for Space Studies, GISS Surface Temperature Analysis: Global Temperature Trends: 2005 Summation, data.giss.nasa.gov/gistemp/2005/, accessed 11 June 2007. International Panel on Climate Change 2007. Climate change 2007: The physical science basis. Summary for policymakers. Available by International Panel on Climate Change, United Nations. http://www.ipcc.ch/SPM2feb07.pdf Jackson, R.B., J. Canadell, J.R. Ehleringer, H.A. Mooney, O.E. Sala, and E .D. Schulze. 1996. A global analysis of root distributions for terrestrial biomes. Oecologia. 108(3): 389411. Jacobson, M.C., R.J. Charlson, H. Rodhe, and G.H. Orians. 2004. Eath system science From biogeochemical cycles to global change. International Geophysics Series Vol. 72, Elsevier Academic Press, New York. Jobbgy, E.G., and R.B. Jackson. 2000. The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological Applications. 10(2): 423436.

PAGE 94

94 Journel, A. G., a nd C.J. Huijbregts 1978. Mining g eostatistics, New York: Academic Press. Keeling, R.F., S.C. Piper, and M. Heimann. 1996. Global and hemispheric CO2 sinks deduced from changes in atmospheric O2 concentration. Nature. 381(6579): 218221 Lal, R. 2004. Soil carbon sequestration impacts on global climate change and food security. Science. 304(5677): 1623 1627 Lal, R. 2004. Soil carbon sequestration to mitigate climate change. Geoderma. 123(12): 1 22. Lal, R., and R.F. Follett. 2009. Soil Carbon Seques tration and the Greenhouse Effect: coeds.: Rattan Lal and Ronald F. Follett. American Society of Agronomy. Lugo, A.E., M.J. Sanchez, and S. Brown. 1986. Land use and organic carbon content of some subtropical soils. Plant Soil. 96(2): 185196. Martin, A. 1991. Short and long term effects of the endogeic earthworm Millsonia anomala (Omodeo) (Megascolecidae, Oligochaeta) of tropical savannas, on soil organic matter. Biology and Fertility Soils. 11(3): 234238 Millennium Ecosystem Assessment (MEA). 2005. Ecosystems and human well being: synthesis. Mulkey, S. 2008. Introduction to greenhouse gas mitigation through forestry and agriculture in Florida. In Opportunities for greenhouse gas reduction through forestry and agriculture in Florida. Universi ty of Florida, School of Natural Resources. J. Alavalapati, A. Hodges, A.C. Wilkie, and S. Grunwald. 2008. Retrieved January. 20: 2008. Myers, D.B., S. Grunwald, G.M. Vasques, and W.G. Harris. 2011. Harmonizing legacy and emerging datasets for soil carbo n stock estimates. In press Myers, J.L., and A. Well. 2003. Research design and statistical analysis. Psychology Press. National Research Council. 2005. Valuing ecosystem services toward better environmental decision making. Washington, D.C. Natural R esources Conservation Service, U.S. Department of Agriculture, 1999. Soil taxonomy: a basic system of soil classification for making and interpreting soil surveys, Agriculture Handbook vol. 436, 2nd Ed. U.S. Government Printing Office, Washington, DC. Nowak, D.J., and D.E. Crane. 2002. Carbon storage and sequestration by urban trees in the USA. Environmental Pollution. 116(3): 381389 Post, W.M., and K.C. Kwon. 2000. Soil carbon sequestration and land use change: processes and potential. Global Change Biology. 6(3): 317327Available at (verified 20 July 2011).

PAGE 95

95 Post, W.M., and L.K. Mann. 1990. Changes in soil organic carbon and nitrogen as a result of cultivation. Soils and the greenhouse effect. 401406. Pouyat, R., P. Groffman, I. Yesilonis, and L. H ernandez. 2002. Soil carbon pools and fluxes in urban ecosystems. Environmental Pollution. 116(Supplement 1): S107S118 Powers, J.S., M.D. Corre, T.E. Twine, and E. Veldkamp. 2011. Geographic bias of field observations of soil carbon stocks with tropical land use changes precludes spatial extrapolation. Proceedings of the National Academy of Sciences. 108(15): 6318 6322 Rumpel, C., I. Kgel Knabner, and F. Bruhn. 2002. Vertical distribution, age, and chemical composition of organic carbon in two forest soils of different pedogenesis. Organic Geochemistry. 33(10): 11311142 Schlesinger, W.H. 1990. Evidence from chronosequence studies for a low carbonstorage potential of soils. Nature. 348(6298): 232234 Seaber, P.R., Kapinos, F.P., and Knapp, G.L., 198 7, Hydrologic Unit Maps: U.S. Geological Surve y Water Supply Paper 2294, 63 Skog, K.E., and Nicholson, G.A. 2000. Carbon cycling through wood products: The role of wood and paper products in carbon sequestration. Forest Products Journal 48: 75. St. Johns River Water Management District (SJRWMD). 2003. The upper Ocklawaha river basin. Accessible through http://sjr.state.fl.us/publications/pdfs/fs_upperockl.pdf. St. Johns River Water Management District (SJRWMD). 2007. The upper St. Johns river basin pr oject. Accessible through http://sjr.state.fl.us/publications/pdfs/fs_usjrbproject.pdf St. Johns River Water Management District (SJRWMD). 2009. The middle St. Johns River basin: St. Johns River Water Management District projects in central Florida. Accessible through http://sjr.state.fl.us/publications/pdfs/fs_msjrb.pdf St. Johns River Wa ter Management District (SJRWMD). 2010. Water resources geodatabase quadbasin. Accessible through http://floridaswater.com/gisdevelopment/docs/metadata/HydrologicBasins.htm Stone, E.L., W.G. Harris, R.B. Brown and R.J. Kuehl 1993. Carbon storage in fl orida spodosols. Soil Science Society of America. J 1993 57: 179182 U.S. Environmental Protection Agency (EPA). 2010. Climate change indicators in the United States. http://www.epa.gov/climatechange/indicators.html

PAGE 96

96 U.S. Environmental Protection Agency (EPA). 2009. Inventory of greenhouse gas emissions and sinks: 19902007 Washington, D.C. EPA 430R 09 004. U.S. Environmental Protection Agency (EPA). 2007. Inventory of U.S. greenhouse gas emissions and sinks 19902005. Washington, D.C. EPA 430R 07002. U.S. Environmental Protection Agency (EPA). 2005. Greenhouse gas mitigation potential in U.S. Forestry and Agriculture. Washington, DC. EPA 430R 05006. U.S. Geological Survey (USGS). 2009. Open file report 2009 1283. Rapid assessment of U.S. forest and soil organic carbon storage and forest biomass carbon sequestration capacity. U.S. Geological Survey (USGS). 1990. Groundwater atlas of the United States: Alabama, Florida, Georgia, South Carolina. Floridan aquifer. HA 730G. Accessible through http ://pubs.usgs.gov/ha/ha730/ch_g/G text6.html. van der Werf, G.R., D.C. Morton, R.S. DeFries, J.G.J. Olivier, P.S. Kasibhatla, R.B. Jackson, G.J. Collatz, and J.T. Randerson. 2009. CO2 emissions from forest loss. Nature Geosci. 2(11): 737738. Vasques, G.M ., S. Grunwald, N.B. Comerford, and J.O. Sickman. 2010. Regional modelling of soil carbon at multiple depths within a subtropical watershed. Geoderma. 156(3 4): 326336. Venables, W.N., and B.D. Ripley. 2002. Modern applied statistics with S. Birkhuser. Whitney, E.N., D.B. Means, and A. Rudloe. 2004. Priceless Florida: natural ecosystems and native species. Pineapple Press Inc. Williams, P., and K. Norris. 1987. Near infrared technology in the agricultural and food industries. American Association of Cereal Chemists, Inc. Williams, P.J. leB. 1975. Biological and chemical aspects of dissolved organic material in seawater. Chemical oceanography. 2: 301363. Zhengxi, T., W.G. Harris, and R.S. Mansell. 1999. Water table dynamics across an AquodUdult tra nsition in Florida flatwoods. Soil science. 164(1): 10.

PAGE 97

97 BIOGRAPHICAL SKETCH Christopher Wade Ross was born in Chiefland, Florida in 1981, the second of three children of Robert and Nancy Ross. He came to love and respect the outdoors as a child and some of his fondest memories are playing in the woods and springs along the Suwannee River as a child. His father was a huge influence in his life as well as his baseball coach, scout leader and life mentor. He continued playing baseball and scouting through hi gh school and was awarded the rank of Eagle Scout his senior year. He wa s introduced to soil science early in his undergraduate career through the introductory course Introduction to Soils in which he immediately fell in love with. During his junior year he decided he wanted to continue his education at the graduate level in the department that taught his favorite courses as an undergraduate, the So il and Water Science Department In 2004 he received his Bachelor of Science degree in Agricultural Operatio ns Management at the University of Florida. In 2011 he received his Masters of Science degree in Soil and Water Science from the University of Florida. He is a member of the Gamma Sigma Delta Honor Society as well as a Florida Bright Futures scholar. Durin g his graduate studies, he had the opportunity to gain professional work experience as an Environmental Scientist with the St. Johns River Water Management District through their college internship program. Wades scientific interests include soil science, biology, geography, hydrology and anthropology. He enjoys being outdoors and loves fishing, bow hunting, mountain biking, hiking, camping, geocaching, traveling and discussing research in general. Wade wants to continue working in the environmental scien ces in North Florida, the area he grew to love.