Ecological sustainability in Amazonian agroforests


Material Information

Ecological sustainability in Amazonian agroforests
Physical Description:
McGrath, Deborah Anne, 1963-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 30032783
oclc - 41869669
System ID:

Full Text








I am extremely fortunate to have had an excellent advisory committee. Each member

had a unique and indispensable role in my academic training, and for their generous assistance

I am very grateful. Both Drs. Nicholas Comerford and Marianne Schmink provided the

inspiration, encouragement, and support that ultimately allowed me to carry out research in

areas that I would not have otherwise felt free to explore. I also greatly benefitted from the

input of Drs. Wendell Cropper and Kimberlyn Williams, who introduced me to ecological

concepts that broadened my view of agricultural systems as functioning ecosystems. Dr. P.K.

Nair was very gracious to accept the task of reading my dissertation so late in the process.

My research interests were certainly stimulated by the dynamic agroforestry program at

Florida initiated by Dr. Nair and his students. My major advisor, Dr. Mary Duryea, was, quite

simply, a mentor in every way, and her friendship and guidance have been invaluable.

My field research was funded by fellowships from the Inter-American Foundation,

the National Security Education Program and the University of Florida Tropical Conservation

and Development Program. I am grateful for their financial support, as I am to the University

of Florida's Center for Latin American Studies, not only for the two-year Title VI Foreign

Languages and Area Studies (FLAS) fellowship in Brazilian Portuguese, but for encouraging

collaborative research among the social and natural sciences with our southern neighbors.

From the University of Florida I would like to thank Cherie Arias for generous

administrative help, James Bartos, Christina Bliss, Jeff English, Wayne Hogan, Dave Noletti,

Larry Schwandes, and Beverly Welch for laboratory assistance; Jay Harrison for statistical

counsel; Ken Clark for critical input to the study design, and Gretchen Greene and Bea

Covington for being girlfriends. Mary McLeod in the Forest Soils Lab was especially

generous with her time and patience. My interaction throughout the years with Dr. Peter

Hildebrand has been very important, because he is, in spirit, a "farmer's farmer".

Special thanks are extended to John Haydu, Peter Cronkleton, Coral Wayland, and

Richard Wallace for priceless laughs and a place to get away in Rio Branco when the

grittiness of the field became too much. Karen Kainer, Jon Dain, and Connie Campbell are

also acknowledged for their valuable insight into conducting participatory research in the

Amazon, as well as for their hospitality in Rio Branco. I also thank Sonia Alfaia for her

hospitality in Manaus. I greatly appreciated being able to count on Charles Clement, from the

National Institute of Amazonian Research (INPA) in Manaus, Brazil, and Pauline Grierson,

of the Ecosystems Research Group at the University of Western Australia, for thoughtful

advice and dialogue via e-mail.

This research would not have been possible without the generous and conscientious

collaboration of my Brazilian colleagues and friends. I am extremely grateful to the non-

governmental organization, Grupo PESACRE (Pesquisa e Extensao do Sistemas

Agroflorestais no Acre), both for institutional collaboration and tremendous logistical support

during the field research. Both the assistance and friendship provided by members of

PESACRE, especially that ofNilton Cossan Mota, greatly enriched my research experience.


I thank SOS Amaz6nia and Projeto Tapiri for their helpful feedback and for allowing me to

participate in their environmental education courses. I also appreciate the use of laboratory

facilities at the Universidade Federal do Acre (UFAC). My most heart-felt thanks are

extended to Maria Lucia Hall de Souza for her superb field assistance. Her conscientious

work enabled me to entrust her with field data collection when I needed to leave the research

site. Her friendship, as well as the hospitality of both her and her husband, will not be


I must express my deep gratitude to the farmers of Projeto RECA (Reflorestatmento

Econ6mico Consorciado e Adensado) for their insight and enthusiastic collaboration in this

research. I am very grateful to the families of Sr. and Sra. Nelson Barbosa, Arnaldo da Costa,

Sr. Joio and Francisca Craveiro, Sr. Aluizio and Sra. Carmelita Goncalves, Sra. Linda

Hendricks, Semildo and Zali Kaefir, Bernadete and Sergio Lopes, Sr. and Sra. Raimundo

Ant6nio Roderigues, and Marcio and Menesilda Sorde, all of whom participated directly in

the study, offering their farms, friendship, generosity, hard work, and unwavering

commitment to the investigative process.

Finally, my gratitude to my parents, Tom and Patty McGrath and Janet and Garry

Bostwick, for their continual support and enthusiasm is endless. I know that to a large extent

my research interests were by the curiosity and love of my mother, Janet Bostwick, for all

things growing in soil, and by the commitment of my father, Thomas McGrath, to protecting

our environmental heritage. Finally, I thank my husband, C. Ken Smith, for his patience,

support, and collaboration as a colleague, as well as for being my hero and best friend.


ACKNOWLEDGMENTS ............................................... ii

AB STR A C T ........................................................ vii

1 INTRODUCTION ................................................... 1

The Problem: Ecological Instability in Amazonian Land-Use ............... 1
Research Questions and Objectives .................................. 6

2 THE RECA PROJECT ............................................... 9

The Settlement of Nova Calif6rnia, Acre, Brazil ................ ........ 9
Projeto RECA ................................................. 12
Agroforest Establishment ......................................... 15
Agroforest Tree Species .......................................... 17
Challenges Facing RECA ......................................... 21

RESEARCH ...................................................... 26

Introduction ................................................... 26
M ethods ...................................................... 29
The Participatory Process: Lessons Learned ........................... 42
Information Gained Using a Participatory Approach .................... 47
Conclusions ................................................... 56

CONVERSION .................................................... 58

Introduction ................................................... 58
M ethods ...................................................... 62
R results ....................................................... 71
Discussion and Conclusions ....................................... 76
Conclusions: Implications for Amnazonian Agroforest Sustainability ......... 87


Introduction ................................................... 90
M methods ...................................................... 94
R results .. .................................................... 102
D discussion .................. .... ............................ 111
Conclusions: Implications for Agroforest Management................. 121

CON V ER SION ................................................... 124

Introduction .................................................. 124
Methods ................................................... 127
Results ................................................... 139
D discussion ................................................. 150
Conclusions: Agroforest Sustainability ............................. 166

MANAGEMENT AND ECONOMIC VIABILITY ........................ 169

Accelerated P Cycling and Agroforest Management .................... 169
Nitrogen Removal and Agroforest Management ....................... 173
Conclusions: Prospects for Ecological and Economic Sustainability ........ 176

LIST OF REFERENCES ............................................. 181

BIOGRAPHICAL SKETCH ........................................... 201

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



Deborah Anne McGrath

December 1998

Chairperson: Dr. Mary L. Duryea
Major Department: School of Forest Resources and Conservation

Raising land productivity with perennial cash crops may allow Amazonian farmers to

meet food demands and increase livelihoods with less forest clearing. Despite more efficient

nutrient cycling in tree-based agroecosystems, maintaining phosphorus (P) availability to

plants growing in weathered tropical soils challenges the sustainability of commercial

plantation agroforests. The primary objective of this research project was to examine

phosphorus and nitrogen (N) dynamics in a widely-adopted peach palm (Bactris gasipaes

Kunth)-cupuassu (Theobroma grandiflorum)-Brazil nut (Bertholetia excelsa) agroforestry

system to evaluate the potential of commercial agroforests to offer a more sustainable

alternative to other Amazonian land-uses. The research was conducted in Acre, Brazil, using

a participatory approach so that farmers would benefit from both the investigative process and


study results, perhaps enabling them to maximize the agroecosystem's potential for sustained

production. A comparison of soils from eight agroforests and adjacent native forests

demonstrated that despite greater cation exchange capacity and pH in agroforest soils,

extractable P was significantly lower, suggesting a decline in P availability since conversion

of forest to agroforest. Phosphorus limitations to productivity, assessed using a root

ingrowth bioassay, were not apparent, although greater root growth by peach palm suggested

a competitive advantage by this species. Monthly measurements of resin-exchangeable P

demonstrated greater P availability in soil beneath peach palm litter than under cupuassu trees.

Nitrogen and phosphorus were mineralized rapidly from decomposing palm litter but

immobilized in P-poor cupuassu leaves. Soil P availability was greatest early in the rainy

season, decreasing during the mid-rainy season when fruit production was highest. An annual

budget for an eight-year-old agroforest revealed that P removal with harvest was half that

expected for pasture and shifting cultivation, and that system-level P cycling was more rapid

than in Amazonian forests growing in similarly P-poor soils. High N removal with harvest

suggests that this nutrient may eventually limit agroforest productivity. Peach palm and

cupuassu phosphorus-use-efficiencies were similarly low, while that of Brazil nut was

comparable to Amazonian forest species. Leguminous cover crops and directed application

of soil amendments, including plant residues, beneath cupuassu and Brazil nut canopies are

recommended to increase soil nutrient availability and sustain productivity.



The Problem: Ecological Instability in Amazonian Land-use

Since the late 1970s, Amazonian deforestation has proceeded at alarming rates,

raising world-wide concern because of its potentially negative consequences for global climate

change, biodiversity, hydrology and biogeochemical cycles (Skole and Tucker 1993).

Approximately one third of forest clearing in the Brazilian Amazon is undertaken by the

region's growing population of colonist farmers for the shifting cultivation of annual crops,

while roughly 60% occurs at the hands of large-scale cattle ranchers for pasture creation

(Fearnside 1993, Skole et al. 1994, Serrao et al. 1996). Geologically, the Guyana and

Brazilian shields that dominate the northern and southern ends of the Amazon Basin are the

oldest and most highly weathered soils found on the South American continent (Toledo and

Navas 1986). Consequently, annual crop production in the Basin's acidic nutrient-poor soils

(predominately Oxisols and Ultisols) is generally limited to two years because nutrient pulses

released by burning native forest vegetation decrease rapidly with crop removal and leaching,

after which agricultural fields are abandoned to fallow and additional forest land is cleared for

continued cultivation (Uhl and Jordan 1984, Ewel 1986, H61olscher et al. 1997). In areas

where low population densities and high land availability permit long fallow periods for soil

restoration (i.e., 10 to 25 years), shifting cultivation can be productive, however, when the

fallow is shortened, the practice results in rapid soil degradation (Nicholaides et al. 1985, Juo

and Manu 1996). Similarly, pasture productivity and longevity in the Amazon are limited by

soil fertility and disruptions in nutrient cycling processes. Generally, three to five years

following forest conversion, a rapid decline in the productivity of planted grasses, associated

with decreases in soil nutrient availability, permits the invasion of herbaceous and woody

"weeds" that characterize degraded and subsequently abandoned pastures (Toledo and Navas

1986, Dias-Filho et al. In press). Regrowth in both abandoned shifting cultivation plots and

degraded pastures occurs as succession proceeds, but often the species composition of the

secondary vegetation differs from that in primary forests and soil C and N stocks, as well as

other properties favorable for agricultural production, decline (Nepstad et al. 1991, Trumbore

et al. 1995, Moraes et al. 1996, H1olscher et al. 1997). Managed extensively, without the use

of soil amendments or germplasm suited to the region's physiography, these two principal

Amazonian land-uses are largely unsustainable. This lack of ecological stability combined

with the small economic return per unit area of land yielded by these land-uses results in

accelerated deforestation, habitat fragmentation, lowered agricultural production, failure of

small-scale farms, and greater rural poverty (Hecht and Cockburn 1990, Fearnside 1993,

Skole et al. 1994).

More recently, perennial crop-based commercial plantation agroforestry systems have

emerged as a promising Amazonian land-use alternative with the potential to reduce soil

degradation, improve living standards, and decrease pressures on remaining forested areas

(Smith et al. 1997). Whereas annual and perennial crops have traditionally been grown

together in multistory tree gardens, the production of high value perennial cash crops in

plantation agroforests represents a relatively new practice in Amaz6nia (Nair and Muschler

1993, Smith et al. 1997). Both the potential economic and ecological advantages of tree-

based agroecosystems arise in part from their longevity which promotes a more closed cycling

of nutrients that may extend the productivity of land already cleared (Ewel 1986, Smith

1990). In principle, deep-rooted perennials intercept cations and nitrate otherwise leached

from the soil surface, storing and cycling these nutrients in living biomass, fallen litter and

decaying fine roots, while reducing erosion losses by physically protecting the soil (Nair 1989,

Young 1989). Moreover, soil degradation and nutrient depletion resulting from crop harvest

is potentially less if the products harvested represent only a small proportion of the system's

total organic matter and nutrient stocks (Jordan 1988). Perennial crop-based agroforestry

systems comprising cashew (Anacardium occidentale), coconut (cocos nucifera), babassu

(Orbignya prunifera) and cacao (Theobroma cacao) have long provided an economically

important and ecologically stable land-use in the more arid Northeastern region of Brazil

(Johnson and Nair 1989).

In the late 1980s, an independent producer organization of colonist farmers initiated

Project RECA (Reflorestamento Econ6mico Consorciado e Adensado) with the establishment

of a perennial crop-based commercial plantation agroforestry system in the western

Amazonian community of Nova Califrrnia. Project RECA's objectives were two-fold: to

improve the economic security of farmers and decrease farm-level deforestation by providing

a more sustainable alternative to other land-uses (RECA 1989). Essentially, the farmers

believed that a multi-species system comprised of native forest trees would be more

productive than their failing annual crops and monospecific plantations of coffee and cacao.

The original system comprised three perennial components, cupuassu ("cupuaqu" in

Portuguese, Theobroma grandiflorum), peach palm ("pupunha" in Portuguese, Bactris

gasipaes), and Brazil nut ("castanha"in Portuguese, Bertholetia excelsa), and was planted on

over 400 ha on nearly 200 farms throughout the region. This particular configuration of

agroforest species has proven highly productive during the initial years following

establishment, and consequently, the RECA project became known throughout Brazil as a

model of sustainable agriculture and grass roots initiative. As a result, farmers in Acre and

Rond6nia continue to convert both primary and secondary forest into perennial crop

plantations that now include coffee, citrus, palm heart, and other native fruit and timber


Although the RECA agroforestry system initially exhibited high productivity,

sustaining yields in the future is a constant and justifiable concern for these farmers because

information about the behavior of these species as components of intensively harvested

agroforestry systems is limited (Clement 1991 & 1993, Venturieri 1993). Many studies have

demonstrated that mixed perennial crop-based systems offer greater ecological stability than

annual monocultures by improving soil properties (Nair 1989, Ewel et al. 1991, Chander et

al. 1998), however, little on-farm research has been conducted to determine if these systems

are sustainable in Amazonian Ultisols and Oxisols without the use of soil amendments (Szott

et al. 1991). For the most part, the RECA agroforests are low- to no-input systems because

most farmers have limited access to chemical fertilizers and little experience using the large

inputs of organic residues recommended to maintain soil fertility (e.g., Nicholaides et al.

1985, Szott et al. 1991). Throughout the Amazon Basin, it is estimated that nitrogen (N) and

phosphorus (P) deficiencies limit crop production in 90% of the region's upland soils

(Nicholaides et al. 1985). Maintaining P availability to crops plants may present a larger

challenge to sustained agroecosystem productivity because much of the total soil P stock is

geochemically bound to iron and aluminum oxides in forms largely unavailable for plant

uptake (Dias-Filho et al. In press). In agroecosystems where N requirements are met with

organic residues from leguminous plants, organic matter decomposition, mineralization and

fixation of N2 may be limited by soil fauna and bacteria sensitivity to P deficiency (Ewei 1986,

Crews 1993). From long-term studies of continuous cropping in Amaz6nia, Sanchez et al.

(1985) concluded that attempts to produce food crops in acid Oxisols and Ultisols without

the use of soil amendments are likely to fail. The question is, then, how sustainable are these

low- to no-input tree-based cropping systems planted by farmers throughout the Amazon


At the time this study was initiated in 1995, RECA farmers were beginning to realize

the economic benefits of the original cupuassu-peach palm-Brazil nut agroforestry systems.

The continued adoption of commercial plantation agroforests throughout western Amaz6nia

underscores the importance of determining if these tree-based systems do offer greater

ecological stability than other land-uses, especially because conversion of native terrafirme

forest into perennial crops is likely to increase as federally-sponsored colonization projects

planned for Acre proceed (Brown pers. comm., Slinger 1996). The RECA project provides

a timely and much-needed case study for evaluating the potential for both economic and

ecological sustainability of these commercially-harvested tree-based agroecosystems.

Research Question and Objectives

The overall objective of this research was to evaluate the potential for ecological

sustainability of a widely-adopted commercial plantation agroforestry system comprised of

native Amazonian forest tree species. The central question guiding the studies was

1. How sustainable are low-input, commercial plantation agroforests in Amaz6nia, and

what are some of the factors that control sustained productivity in these systems?

Specific study objectives intended to address this question were to

1. Analyze soils from eight RECA agroforestry systems and adjacent native forests to

determine how soil chemistry has changed since conversion of primary forest to


2. Assess potential P limitations to agroforest and native forest plant productivity using

a root ingrowth bioassay.

3. Quantify the stocks and fluxes of P and N in an eight-year-old RECA agroforest, and

construct an annual budget to determine how much of the system's P and N

requirements are (a) met through internal cycling, (b) taken up from soil stocks, and

(c) removed with harvest.

4. Identify socio-economic challenges to the sustainability of RECA agroforestry

systems, particularly those that constrain modifications in agroforest management

practices, through interviews, focus groups, and discussions with RECA families and

other local NGO's and research institutions.

5. Conduct the research using a participatory approach that (a) encourages farmers'

involvement in the formation of research objectives, data collection and interpretation


of results, and (b) fosters a multilateral exchange of knowledge, information, and

experience among researchers and land managers.

This dissertation is divided into eight chapters, including the introduction. The

second chapter is a synthesis of the history of the RECA project largely based upon

unpublished reports, local newspaper articles, and interviews and discussions with RECA

families and other Brazilian organizations conducted as part of this research. It summarizes

the past and present challenges faced by the producers' organization, as well as the socio-

economic benefits enjoyed by RECA farmers as a result of community-level agroforestry

adoption. Although RECA is somewhat atypical from many producers' groups (e.g., it has

received large amounts of outside financial assistance), the decade-old project demonstrates

many of the complex socio-economic issues associated with agroforestry system adoption,

underscoring the fact that the search for ecological stability addresses only half (or maybe

less) of the sustainability question.

Farmer participation was an important element of this project. It is hoped that

participation by RECA farmers improved the likelihood that the scope of this research

ultimately renders results that are useful to land managers in Acre. However, given the time,

monetary, and academic constraints of doctoral research projects, perhaps the most

immediately useful research output to RECA farmers was the participatory process itself,

described in Chapter Three. This chapter briefly outlines reasons for adopting a participatory

approach to on-farm agroecological research, describes the methods used to encourage

farmer participation in this research, evaluates the positive outcomes of the process, and

identifies areas in need of improvement.

different aspects of ecological sustainability from a nutrient cycling perspective, since

nutrients, such as P and N, are cited as the resource most limiting to productivity in natural

and managed ecosystems throughout Amaz6nia. The fourth chapter examines soil changes

six years following conversion of native forest to agroforests, especially those in extractable

P, and attempts to identify P limitations to plant productivity. The fifth chapter analyzes

seasonal fluctuations in soil P availability in relation to leaf litter decomposition and the

agroforest's production cycle. In chapter six, system-level P and N dynamics in the

agroforest are compared with those of other Amazonian land-uses, including native forests,

pasture and shifting cultivation. The concluding chapter attempts to synthesize information

on agroforest P and N dynamics within the context of the socio-economic constraints and

opportunities faced by rural households (identified by researcher and RECA farmers) to

develop recommendations for management that enhance the cycling of organic matter and

nutrients to sustain productivity in this and other tree-based Amazonian agroecosystems.


The Settlement of Nova Calif6rnia

The state of Acre, located in the western Amazon Basin on the borders of Bolivia and

Peru, is one of the last frontiers in the Brazilian Amazon (Grupo PESACRE 1989). An

important rubber-producing region previously considered part of Bolivia, Acre was annexed

by Brazil in 1903 following a war with its South American neighbor (Hecht and Cockbumrn

1990). The economy in Acre was thus originally based in forest extraction, and its inhabitants

were primarily indigenous peoples and rubber tappers, Brazilians brought from other regions

in the country to extract latex from trees growing in native forests (Kandell 1984). In the

1970s, the governmental institution, INCRA (Instituto Nacional de Colonizaqao e Reforma

Agraria), launched a large colonization project, referred to as the Polonoreste, in the

neighboring state of Rond6nia. The project encouraged families from south and southestemrn

Brazil, where agricultural modernization was displacing small farms, to resettle in this

relatively undeveloped region of western Amaz6nia by giving them title to 100 ha lots of

largely undisturbed forestland to farm (Browder 1996). In the process, the national highway

BR364 was paved, linking Rond6nia and later, Acre, to the rest of Brazil. These events

initiated a wave of migration to Rond6nia and Acre that consequently led to accelerated

deforestation in the region as colonist farmers and large-scale ranchers cleared native forest

for pasture and shifting cultivation (Hecht and Cockburn 1990).

The colonist community of Nova Calif6rnia, which lies on the border of Acre and

Rond6nia (10S, 67W), was officially recognized as a town by INCRA in 1984. Previously

known as "Santa Clara", which was little more than a gas station, a restaurant and five

houses, Nova Calif6rnia's establishment represented INCRA's official claim to land

previously controlled by the former owners of the rubber estate (seringal) Calif6rnia (RECA

unpublished). The region's population increased considerably in the mid to late 1980s as

families migrated from the southern Brazilian states of Parana, Rio Grande do Sul, and Santa

Catarina, many of them stopping in Rond6nia for several years before finally settling in Acre.

Located 150 km east of Rio Branco, the capital of Acre, Nova Calif6rnia now provides a

political and economic base for over a thousand farm families living on unpaved "feeder"

roads connected to the BR 364.

At the time of this study, state ownership of the region in which Nova Calif6rnia was

located had been disputed since the early 1980s by the governments of Acre and Rond6nia,

both of whom claimed the region as their own (Moreira 1992). The lack of legal definition

of the border between the two states greatly aggravated the economic hardship of families

living in and around Nova Calif6rnia because neither government was willing to invest

resources to build and maintain infrastructure in a town that might ultimately become the

property of another state. As a result, road maintenance, schools and medical facilities were

poorly funded, and INCRA, the organization that had brought families to the region,

essentially abandoned the community (Leite unpublished). Most residents had to travel

several hours on an unpaved road to Rio Branco for even minor medical care. Malaria, in

particular, was a medical problem that plagued residents, although this is now treated at a

community health post (PESACRE and GENESYS unpublished). Schools on feeder roads

continued only to the fourth grade, and families had to send their children to Rio Branco to

attend high school (Campbell 1994). During the rainy season, many families were forced to

walk up to 50 km to reach the BR364 because the mud made the unpaved feeder roads

impassable by car, bike, and horses. Because Nova Calif6rnia was never linked to a utility

grid, at the time of this study, electricity was provided to residents in town by a small

unreliable generator that was operated only in the evening, from six to twelve PM. Families

were responsible for digging their own wells, and "running" water was acquired by pumping

well water (using a diesel or electric pump) into a cistern built above the house. Wells

frequently dried up during the months of July and August, leaving families and the RECA

organization to collect water from swamp areas located at the edge of town. Most families

on the feeder roads still did not have access to electricity or running water at the end of this

study period.

The difficulties faced by these colonist farmers were made worse by the fact that

many of the crops they initially planted grew poorly, and in some cases, failed entirely due to

poor soils, pests, and insufficient marketing infrastructure. For example, monocultural

plantations of cacao (Theobroma cacao L.) succumbed to witches' broom (Crinipis

perniciosa (Stahel) Singer), and poor access to highly competitive markets impeded farmers

from selling coffee (Coffea arabica) (RECA unpublished). Moreover, most families were

unaccustomed to farming in the nutrient-poor soils underlying native Amazonian forest.

Many faced extreme hardship when their shifting cultivation plots of rice, beans, and maize

failed to produce adequate harvests during the second or third year following forest clearing.

As a result, many families were forced to abandon their farms and resettle in other regions,

or return to southern Brazil. Although Brazilian law obliges farmers to maintain 50% of their

land in native forest, it is not difficult to encounter vast tracts of deforested land in the region

surrounding Nova Califrrnia. Many of these deforested areas were created by ranchers who

bought up dozens of 100-ha lots from desperate colonist farmers and cleared them entirely

to raise cattle. Unfortunately, these large cattle ranches often proved unsustainable

themselves, due to poor soils and the invasion of weeds that precluded the regeneration of

pasture grasses, and in some cases, to a more recent drop in cattle prices throughout Brazil

(Hecht and Cockburn 1990, Browder pers. comm.). Consequently, many of these vast

degraded pastures have also since been abandoned.

Proieto RECA

In response to the economic crisis suffered by the region's families, the RECA project

(Projeto de Reflorestamento Consorciado e Adensado) was initiated by a group of farmers

in 1988. Fatigued with the enormous labor and risk associated with the shifting cultivation

of annual crops, these farmers began to experiment with plants native to Amaz6nia, and in

particular, tree crops. One of the group's leaders, Sergio Lopes, a university-trained teacher

from Santa Catarina, was also concerned about the ecological impact of deforestation

associated with shifting cultivation and farm abandonment. With assistance from the Catholic

Diocese of Rio Branco, the Federal University of Acre (UFAC) and the Institute of

Amazonian Research (INPA), the producers' group submitted a project proposal to various

European philanthropic organizations. The objective of the RECA's proposed project was

to increase the economic well-being of colonist farmers through the production of high-value

perennial crops. Because the tree crops proposed were native to Amaz6nia, it was also

reasoned that they were better adapted to the region's weathered forest soils and natural

pests, and thus more likely to persist and sustain productivity in these conditions. Thus, by

offering a more ecologically sustainable alternative to shifting cultivation, these systems could

decrease land abandonment and deforestation associated with small-scale production.

In 1989, the project acquired funding ($ 2 million USD) from Catholic organizations

in the Netherlands (CEBEMO) and France (CCFD) to inititiate the establishment of a multi-

species perennial crop-based commercial plantation agroforest on over 200 farms (Martinello

1993, Smith et al. 1997). The first agroforestry system established by the RECA project in

1989 and 1990 (described fully below) was comprised of three perennial species native to

Amazonia, and participants were required to plant the components in a specific configuration

of species and spacing designed by the RECA organization.

Monetary incentive was an important factor attracting families to participate in the

RECA project. For every hectare of commercial plantation agroforest planted, participating

families received approximately $1,000 (USD) over a three year period from RECA to help

offset expenses incurred during plantation establishment and to help sustain households during

the initial years required before the perennial system began yielding fruit (Moreira 1992). In

return, the participating families were obliged to give a proportion of each harvest, beginning

with the fourth year, to the RECA organization for 10 years after plantation establishment.

The proportion of harvest increased, from 5% during the fouth and fifth years, to 30% by

year ten, and the proceeds from the sale of products were used to support administrative and

operating costs of the RECA project. Services provided by the organization included the

transportation of raw products from farms on feeder roads to a small factory located in Nova

California where the pulp of cupuassu fruit was processed and stored frozen, and later

transported to urban centers, such as Rio Branco, for sale. Farmers were responsible for

transporting and marketing peach palm fruit themselves, and because the fruit is so highly

perishable, many farmers simply sold the palm seed to buyers interested in establishing heart-

of-palm plantations. More recently, the organization received financial assistance from

another non-governmental organization (NGO) to build a canning factory for palm heart, and

an auditorium/dormitory in which regional meetings with other producer organizations are

held. In addition, RECA's role in the community has not been confined entirely to

agricultural production. Nuns of the Catholic church who run a homeopathic pharmacy in

Nova Calif6mrnia regularly train RECA health agents in basic first aid and medicinal plant use

(Campbell 1994). So while RECA started out with a community-based agroforestry project,

the organization has evolved to represent and address the social and economic needs of the

people in and around Nova California, and for better or worse, the organization has become

quite politicized.

Participating farmers entered RECA through regionally-based groups, usually defined

by the feeder road the producers lived on. At the time of this study, there were a total of 15

groups, each led by an individual who acted as a liaison between the regional group and the

RECA organization by representing the group at monthly RECA coordinators' meetings. In

each group there was also a "tecnico", a fannrmer provided with technical training, sponsored


by RECA, whose role was to assist members with production related problems, and

introduce new species to plant, such as leguminous cover crops or native timber trees for new

plantations. At least four of the fifteen groups were led by women, although Campbell

(1994) notes that despite the project's seemingly democratic organization, women's voices

are hard to hear. Twice annually, all members of RECA were assembled for a three to five

day period, during which they reported on the health and productivity of their farms, and

discussed issues regarding new plantations, production, transport, processing and marketing.

'These assemblies also provide opportunities for research and extension organizations, as well

as for other producer groups in the region to meet with RECA farmers and discuss problems

related to agroforest management and product marketing. RECA has also attempted to

provide farmers with greater access to information and technical assistance. During the

period this study took place, two Brazilian non-government organizations (Projeto SOS and

Projeto Tapiri) offered week long environmental education courses to RECA farmers, and

RECA has formed partnerships with extension and research organizations such as PESACRE

(Grupo de Pesquisa e Extensao em Sistemas Agroflorestais do Acre) and EMBRAPA

(Empresa Brasileira de Pesquisa Agropecualia).

Agroforest Establishment

The specific agroforestry configuration under study was planted on over 300 farms

on approximately 450 ha in 1989 and 1990 (Leite unpublished). The system is two-tiered,

dominated by an upper canopy of peach palm (Bactris gasipaes Kunth) and Brazil nut

(Bertholletia excelsa Humb. & Bonpl.) with a middle canopy formed by cupuassu

(Theobroma grandiflorum (Willdenow ex Sprengel) Schumann). The agroforest's principal

products include cupuassu pulp, peach palm fruit and seed, and heart-of-palm, all of which

are consumed domestically as well as marketed regionally.

Typically, these agroforests were established by cutting and burning native forest

vegetation and interplanting one-year-old peach palm and cupuassu seedlings in rows at a

spacing of 4 x 7 m. Some agroforestry systems were also planted on old fallow fields

(capoeira). In every third row, Brazil nut was planted alternately with cupuassu to complete

a stocking density of approximately 370 trees ha', the majority of which are cupuassu and

peach palm (190 and 150 trees ha', respectively). The seedlings were raised on farm from

seeds collected from marketed fruit and surrounding forest; thus there exists considerable

genetic heterogeneity within each of the agroforests' three perennial components. At the time

of establishment, enough farmyard manure to fill a "milk can" (approximately 250 ml, equal

to about 0.5 and 0.15 kg ha"' N and P, respectively) was added to each seedling's planting

hole. During the first year, annual crops (maize, beans, rice, or cassava) were cultivated

between the agroforest rows to offset the initial expense of establishing the perennial system

(Wallace 1994). Thereafter, understory regrowth of native vegetation was cut twice annually

and left to decompose on the agroforest floor. In some systems leguminous cover crops

(Macuna cochichinensis and Pueraria phaseoloides) were planted in agroforest rows for

nitrogen fixation and weed control. However, due to the increased risk of fire hazard and the

fact that the vines began growing over the top of the tree canopies, the legumes were

eradicated from the agroforests three years after establishment. Since establishment, grazing

livestock were excluded from the agroforest, and no other amendments of any kind were

applied to the system.

Agroforest Tree Species

Peach Palm

The origin of peach palm (Bactris gasipaes Kunth, Palmae) is somewhat

controversial. Historically, peach palm, or "pupunha" in Portuguese, was grown throughout

tropical America by many pre-Colombian Amerindian communities for food, fiber, and

medicine. Clement (1988) cites the great genetic diversity in peach palm populations of

western Amaz6nia as evidence that the species originated in this region, although Mora Urpi

(1992) suggests that multiple domestication events may have taken place.

At present, two products are harvested from peach palm grown in multi-species

commercial plantation agroforests; an energy- and nutrient-rich fruit that is consumed locally,

and the tender unexpanded leaves produced by the apical meristem of young offshoots,

referred to as palm heart. The latter product is potentially more lucrative as it can be sold

both throughout Brazil and abroad as a high-priced delicacy. Peach palm is multi-stemmed,

producing as many as 12 offshoots that arise from axillary buds encircling the main stem.

This feature, in particular, has made it very popular for "sustainable" palm heart production

because unlike single-stemmed palms from which palm heart is also harvested (Euterpe spp.),

peach palm offshoots can be removed without killing the whole plant. Stems of peach palm

may attain up to 24 meters in height, but the species' relatively small crown, typically

composed of 10 to 30 pinnate leaves (Arkcoll 1990, Mora-Urpi et al. 1997), minimizes

shading of other agroforest components. An undesirable characteristic of the palm is that

stem internodes are frequently covered with long spines which present a hazard to livestock

and complicate fruit harvest for farmers. Peach palm root growth is generally concentrated


in the top 20 cm of soil, although a superficial mat of adventitious roots often develops at the

stem base and may extend up to five meters from the trunk (Ferreira et al. 1980). As leaves

and fruit abscise from palm stems, decomposing organic matter from fallen litter accumulates

in the root mat. Vandermeer (1977) states that approximately 75% of peach palm roots are

located within the perimeter of the canopy, but Ferreira et al. (1995) observed absorptive

roots extending up to nine meters from the stem base. The palm is relatively productive in

well-drained Oxisols and Ultisols, tolerating up to 50% aluminum saturation, although studies

have shown that nutrient additions are necessary to sustain long-term productivity (Mora-

Urpi et al. 1997). Because P is generally limiting in tropical soils, some work has been

conducted to determine the importance of P availability and mycorrhizal associations to peach

palm development and productivity (St. John, 1988, Clement and Habte 1995). For example,

Habte and Clement (1994) demonstrated that P fertilization greatly increased seedling leaf

growth, biomass increment and overall vigor, and Ruiz (1991) found that peach palm

infection with vescibular-arbuscular mycorrhizae was negatively correlated with soil P


Flowering in peach palm begins between the ages of 3 and 5, and the palm may

produce annual crops for up to 50 years, although estimates for fruit production in nutrient-

poor Amazonian soils are much lower (i.e., 20 to 25 years, Clement pers.comm.). The oily

fruit produced by the palm is highly perishable, and thus difficult to transport fresh to markets.

A beta-carotene-rich flour is made from dried fruit, employing the same on-farm processing

technique used for the fabrication of cassava flour, a traditional staple in Amazonian

households (Dibari pers. comm.). The fruit is also used for animal feed, and RECA farmers

have found it lucrative to sell peach palm seed to buyers interested in establishing heart-of-

palm plantations. Clement (1989) notes that commercial production of fruit and heart-of-

palm in the same system is not practical because the latter requires high density (4,000 plants

ha"') monospecific plantations to be economically viable.


Theobroma grandiflonrm (Willdenow ex Sprengel) Schumann, Sterculiaceae, is one

of nine species in the same genus native to the Brazilian Amazon. Cupuassu ("cupuaqu" in

Portuguese) occurs naturally in forests of the eastern Brazilian states of Para and Maranhio,

but its distribution has spread across the Amazon Basin (Cabral Velho et al. 1990, Venturieri

1993). Like its relative, cacao (Theobroma cacao), it is a broad-leafmesic species that grows

naturally in the understory of terrafirme forests, tolerating both shade and nutrient-poor

soils. The species' growth habit is pseudo apical, resulting in a relatively small-statured (5

to 15 m height) tree, with a plagiotropic canopy projecting outward, up to eight meters from

the trunk (Ribeiro 1992, Ventureiri 1993). Cupuassu is generally pest-resistant, although like

cacao, it is susceptible to witches' broom (Crinipis perniciosa).

The large (12 to 25 cm length, 10 to 12 cm diameter) woody elliptical fruit pods

(loculicidal capsules) produced by cupuassu are harvested for the fragrant creamy pulp which

is used in desserts, candies, and drinks throughout Brazil (Cabral Velho et al. 1990). More

recently, methods have been developed to use fermented cupuassu seeds, much in the same

manner that cocoa beans are processed for chocolate, to make a confection known as

cupulatee" (Ribeiro de Nazare et al. 1990, Wallace 1994). Cupuassu typically flowers at the

end of the dry season, with maximum fruit production occurring during the mid to late rainy

season. The species is known for its low fecundity; Ventureiri (1993) found that 3,500

flowers per tree were necessary to produce 15 fruit. Trees begin bearing fruit as early as three

years of age, and by year six or seven, an average tree produces between 12 and 15 fruits per

season. Peak fruit production, reported to be as great as 100 fruits per tree per year, occurs

between ages 10 and 20, but trees can continue to bear fruit for up to 30 years (Ribeiro

1992). The author has seen 50-year-old productive cupuassu trees growing in mesic habitats

on homesteads in eastern Amazona. To maintain plantation productivity, Calzavara (1980)

recommends a yearly application of 100 g fertilizer (15% ammonium sulfate, 50%

superphosphate, 15% potassium chlorate) per tree, broadcast on the soil surfacejust beneath

the canopy's drip line.

Brazil Nut

A rare upper canopy emergent in Amazonian terra firme forests, Brazil nut, or

"castanha" in Portuguese, (Bertholletia excelsa Humb. & Bonpl.: Lecythidaceae) trees may

attain heights of 50 meters (Mori and Prance 1990). A mature tree has a straight, relatively

unbranched bole and small crown which makes it a favorable upper story agroforest

component. Taproot extension of Brazil nut trees growing in forests and pasture in eastern

Amazona has been observed 5 to 10 meters into the soil (Nepstadt et al. 1994). While the

tree grows naturally in the well-drained nutrient-poor soils underlying native forest

vegetation, Kainer et al. (1998) found that Brazil nut seedlings planted in shifting cultivation

plots, where light and nutrient availability were greater, grew more vigorously and had higher

foliar nutrient contents than those planted in forest gaps.


Most Brazil nuts are collected from wild trees growing in forests, however, the

species has more recently become a component ofmonospecific plantations and multi-species

agroforests in Amaz6nia. Although it may take 25 years for forest trees to reach maximum

production (estimated to be several hundred fruit per tree per year), trees grown in more

intensively-managed plantations may begin bearing fruit within eight years after establishment

(Mori and Prance 1990). In addition to being a food staple for forest dwelling communities,

Brazil nuts have become an important Amazonian cash crop, both sold for domestic

consumption and exported abroad (Kainer et al. 1998). Despite its economic potential,

Brazil nut was typically a minor component in the commercial plantation agroforests

examined in this study.

Challenges Facing RECA

Since its establishment, RECA's peach palm-cupuassu-Brazil nut agroforestry system

has been highly productive. The total harvest ofcupuassu fruit from RECA agroforests was

reported to be 75 tons in 1994 and 120 tons in 1995 (Leite unpublished). Evidence of

greater economic prosperity for many RECA farmers is demonstrated throughout Nova

Calif6rnia, with an increased building of new homes with all-weather roofs, as well as satellite

dishes and diesel generators that are found even in some homes located on the more secluded

feeder roads. Many farmers also claim that the profits made through the sale of cupuassu

pulp, peach palm seeds, and heart-of-palm have allowed them to invest in other productive

endeavors on their farm, including small-scale cattle ranching. Nova Calif6rnia itself has

grown considerably, with several more markets, bars, and furniture builders, providing further

evidence of increased prosperity in the community. Also, RECA's initial success has earned


attention, recognition and respect for the producers group from local research and extension

institutions who are eager to initiate on-farm studies with the organization to examine

everything from the use of leguminous cover crops and breeding of spineless peach palm

stems, to establishing dairy farms in the community. Visits from T.V. crews, journalists and

reporters from southern Brazil were common during the time this study took place. The

group's recognition has given them an advantage in applying for economic assistance from

other foreign NGO's, although many argue that this has created a dependency on outside aid

which prevents the organization from becoming a model of economic and ecological


Moreover, impressively high yields on relatively poor soils has not spared the

organization from other socio-economic problems, the largest of which have thus far been

associated with product processing and marketing (Smith et al. 1997). One of the first such

difficulties arose with the unexpectedly high yield of peach palm fruit. Because it is highly

perishable, the fresh fruit must be transported and marketed very shortly after harvest, and

the local market for fruit is somewhat limited in its demand for the starchy fruit. As a result,

much of the peach palm fruit harvested spoils while awaiting transportation or is fed to pigs

for lack of a buyer. As previously mentioned, some RECA farmers are now making a flour

from boiled fruit, and attempts are being made to market this product for commercial use in

cereals, cakes and pasta (Dibari pers. comm.).

In addition to creating daily hardship in rural households, the lack ofinfrastructural

development and maintenance in the region surrounding Nova California has presented

obstacles for the transport, processing and marketing of RECA products. During the period


this study took place, RECA had record high production of cupuassu fruit. However,

because the market in nearby Rio Branco became quickly saturated with the fruit pulp during

peak cupuassu production months, the fruit pulp was frozen for sale later in the year when

pulp stocks had decreased and fruit prices increased. Because electricity is so unreliable in

Nova Calif6rnia, RECA had a diesel-fueled generator that provided power to the small freezer

in the organization's local processing unit. However, extremely high cupuassu production

early in 1996 forced RECA to rent freezer space in Rio Branco to store 62 tons of pulp

(Smith et al. 1997). Renting the space increased the cost of processing and marketing the

product by RECA, so that the organization was short of funds to purchase unprocessed fruit

from local producers. Furthermore, the freezing and thawing of cupuassu pulp in RECA's

freezers during power lapses lowered the quality of this product and made it less marketable

in urban areas. By the end of the season, many producers were furious with the organization

who either owed them money or had quit purchasing their fruit. The problem was

compounded by the fact that impassable roads during the rainy season made it difficult for the

drivers of RECA's two vehicles to pick up fruit from more remotely-located farms. As a

result of transportation difficulties and RECA's inability to purchase fruit brought on by the

unanticipated high cost of frozen pulp storage, a large portion of the 1996 harvest was lost.

The inability of producers to sell their products to RECA, or transport them to other markets,

created economic hardship for many of the less well-off households, and provoked a lot of

bitterness towards the organization. Many of the farmers that originally received money from

RECA to establish the agroforest have since left the organization, and some now work for

a new privately-owned foreign company (Agro-Amaz6nia) that moved into Nova California


in 1996 to produce and process fruit such as pineapples, palm heart, watermelon, and

cupuassu. Nor is it clear that many of the farmers currently benefitting from the membership

in the organization are truly committed to repaying the loans they received to establish the

system. The default on payments by many producers has increased financial stress for the

organization (Lopes, pers. comm.).

Smith et al. (1997) concluded that RECA is not a viable model for sustainable

agroforestry in Amazonia because of the organization's inability to resolve processing and

marketing problems and chronic dependence on external financial aid. However, there is

evidence to suggest that RECA is learning from its early mistakes and slowly working out the

economic difficulties it faces, especially through collaboration with both non-governmental

and governmental institutions such as PESACRE and EMBRAPA. PESACRE, in particular,

has conducted short courses aimed at teaching farmers basic marketing principles and

choosing more marketable products to grow in the future (Haydu and Wallace 1997). The

Italian Aid agency, MLAL, has sent several volunteers to RECA to investigate ways to add

value to products through processing and decreased marketing costs. For example, RECA

is now working on producing jams and syrups from cupuassu pulp, which not only increases

the value of the pulp before it is marketed, but also decreases the costs of transporting and

storing frozen pulp. This will be an especially effective strategy if the cupuassu confections

can be marketed in sealed plastic bags to avoid the high processing and transporting costs

associated with glass containers. The toasted pupunha flour, as well as the chocolate

cupulatee) made from cupuassu beans also offer real possibilities to increase the return of

agroforest products to both farmers and the RECA organization. In addition, the MLAL


volunteer has worked with RECA to increase the standards of hygiene, quality and safety of

RECA products, and in particular, of canned heart-of-palm, which was one of the most

lucrative agroforest products produced by RECA because it is consumed throughout Brazil

and potentially exported abroad. Likewise, when Brazil nut begins to fruit, high marketing

potential also exists for the nuts, because they are already sold abroad from other areas in the

Amazon, and require relatively low-tech processing (Kainer 1997). Also, shortly after this

study's field research was completed, the border dispute between Acre and Rond6nia ended

with the legal incorporation of Nova Calif6rnia into the state of Rond6nia. State membership

will likely entitle Nova Calif6rnia to greater political and economic support, and

infrastructural improvements made by the government of Rond6nia should diminish some of

RECA's processing and marketing dilemmas. Finally, from the standpoint of biological

sustainability, RECA farmers have been very open to outside researchers attempting to

address soil fertility problems. Researchers from EMBRAPA and INPA continue to work

with individual farmers to determine which leguminous cover crops are most efficiently

managed by farmers while increasing soil nitrogen availability to crop plants. It is hoped that

future courses on product processing and marketing, as well as on agroecosystem

management, will be conducted in RECA's new auditorium in collaboration with producer

groups from all over Amaz6nia so that more of the region's rural families will benefit from

continued education and self-empowerment that will lead to more economically and

ecologically viable agricultural systems and healthier, more secure livelihoods. In this way,

RECA would indeed provide a model if it is able to confront challenges that face land

managers throughout Amaz6nia, and persevere to overcome these obstacles and assist others.



Over the past two decades, "on-farm research", "farmer participation", and "rural

people's knowledge" have become expected components of much agroforestry and

agroecological research, and a wide range of innovative and non-traditional research tools

have been developed to facilitate and strengthen the participatory process (e.g., Mascarenhas

1992, Feldstein and Jiggens 1994). In fact, the process of fannrmer participation has provided

a new paradigm for the development of more sustainable agricultural practices in resource-

poor risk-prone areas, as well as a tool of "empowerment" through which rural people may

achieve more secure livelihoods (Chambers et al. 1989, Rocheleau 1991).

Among the reasons for including the "recipients" or "target group" of technological

developments as informers in the investigative process is the ability to gain a greater

understanding of the interests, priorities and problems faced by user groups. These factors,

as well as household and community level socio-economic constraints (i.e., financial,

practical, educational, motivational, traditional, cultural, political), are critical considerations

in the development of appropriate agricultural technologies and their potential for future

adoption (Beer 1991). User participation thus has the potential to make research results more

practically applicable.


Moreover, the participatory process itself offers opportunities for both land managers

and researchers to interact collaboratively in the development and application of more

sustainable management techniques. The extent to which local people are encouraged to

participate in on-farm research varies considerably, from researcher-designed and -managed

trials that address problems identified, in part, by local land-users, to researcher study of trials

designed and managed by land-users (Rocheleau 1994). A mutually respectful and trusting

relationship fostered by the participatory process increases the likelihood that researchers will

benefit from the specific experience-based knowledge provided by farmers about the

landscape and land-use strategies that have succeeded or failed in the past (Scherr 1991).

Local participants may gain an enriched understanding of the processes controlling

agroecosystem productivity and health, as well as how such processes are maintained or

degraded through manipulation. This knowledge may enable them to manage their land more


Studies have shown that participation in agroecological research encourages farmers

to improve land management practices through continued experimentation on their own. For

example, Ruddell and Beingolea (1995) found that training farmers to conduct their own

research was a much more effective strategy for raising food security than attempting to

provide a "'technology package" that would not serve the diverse ecological micro-climatic

and socio-economic conditions faced by Andean potato farmers. Kainer (1997) found

participation in the research process helped rubber tapper families in the Brazilian Amazon

recognize their strategic position and power when negotiating with NGO's and other

conservation and research organizations.

There are critics who charge that "empowering" local people through participation

is "naive populism" because it implies that "powerful outsiders" must help "powerless

insiders" (Thompson and Scoones 1994). Several authors have pointed out that a "lack of

understanding" may not dictate how resource-poor farmers manage their land. Rather, a

complex set of personal socio-economic and political circumstances, often hidden to

outsiders, operates to constrain management choices (Chambers 1983, Thrupp 1989). Thus,

while useful, an increased understanding ofagroecological processes may not necessarily have

an immediate impact on management decisions. Nevertheless, both researcher and farmer

offer distinct sets of experience and tools, derived from different cultural backgrounds and

traditions of knowledge creation, none of which should be ignored when developing,

adapting, and applying more sustainable land-use practices.

A participatory approach was applied in this agroforestry research to (a) gain a greater

understanding about the role of perennial crops in Amazonian fannrming systems, and household

constraints to modifying agroforest management strategies, (b) stimulate household and

community-level discussions about the role of organic matter and nutrient cycling in

controlling agroecosystem productivity, (c ) elicit realistic management strategies from

farmers to enhance and sustain productivity, and in doing so, provide results useful to the

region's farmers. In this study, the participatory processes of dialogue and exchange among

researchers and farmers were facilitated through two principal channels; (a) focus group

discussions held during organized meetings and presentations, and (b) informal interviews

or "conversations" held with members of 10 individual households during field visits.


Research Initiation

RECA farmers participated during the formulation of research objectives, study site

selection, data collection, and presentation and interpretation of preliminary results. The

design of trials, analysis of data, and statistical interpretation of results were the responsibility

of the researcher. The presentation of final results to farmers is the last step awaiting

completion, as discussed later in this chapter. Another participatory dimension to the research

was close collaboration with local university, government, and non-governmental

professionals. Collaborative ties were made with local Brazilian agencies, such as PESACRE

(Pesquisa e Extensdo dos Sistemas Agroflorestais no Acre), a non-governmental organization

(NGO) working with colonist farmers on agroecological problems, EMBRAPA (Empresa

Brasileira de Pesquisa Agropecuaria), a national agronomic research institution, and SOS

Amaz6nia, a group of environmental educators. PESACRE, in particular, provided crucial

logistical assistance and a field assistant from the local university who was trained in

agroecological research methodologies during the research period.

As mentioned in the previous chapter, the original "RECA" agroforestry system was

both conceptualized and planted on over 200 farms in the late 1980's by the producers

themselves, although the group did receive some technical advice from local research

institutions and NGOs. By the time I conducted a six-week pilot study in Acre in 1994, the

RECA project was already well known throughout the western Amazon, and several extended

visits to RECA farms left me with two questions: how could these agroforestry systems be

so productive in seemingly nutrient-poor soils, and how long could such productivity last

without more intensive management? Conversations with host families and members of

RECA administration included discussions about their experiences with this agroforestry

system, as well as their concerns regarding the agroforest's health and productivity, and the

marketing of its products. From these conversations and additional feedback received from

PESACRE, I formulated preliminary study objectives to present to RECA during one of their

official assembly meetings before returning to the U.S. to write a research proposal.

Including RECA farmers in the formulation of research objectives as part of the

participatory process necessitated that I change my research plan entirely from that previously

delineated in the original pilot study proposal. The basic objective agreed upon between

myself and RECA farmers was that I would develop a project that would examine processes

controlling productivity in the system in relation to its management by households, so that

research results could be presented to the group in the form of recommendations for

sustaining and enhancing agroforest productivity. Moreover, after the proposal had been

developed, I attempted to meet the concerns of local farmers by adapting the root ingrowth

bioassay to include a study of root competition (Chapter 4). RECA farmers were concerned

about what they perceived to be "aggressive" root competition by peach palm, as evidenced

by the species' thick shallow network of roots that extended well beneath the canopies of

cupuassu. This was especially alarming for farmers because cupuassu was the most

economically important component of the agroforest at that time. Maintaining soil fertility

was also a concern for them, as farmers were committed to using only organic soil

amendments (the actual use of which appeared to be quite limited), primarily because

chemical inputs were expensive and not easily procured in this region. Other issues outside


my realm of training discussed by the organization and households included pest problems,

especially witches' broom, which had previously wiped out cacao plantations, and the

transport, processing and marketing of agroforest products.

Within a year of my initial visit, a simplified version of the research proposal approved

by my advisory committee was translated into Portuguese and sent to PESACRE and RECA

leaders for review. PESACRE determined that the research would fit within their

professional priorities and contacted the Federal University of Acre (UFAC) to find a an

agronomy student in need of a supervised field research project to meet graduation

requirements. The student was paid to work as a field assistant with the understanding that

I would train her to conduct agroforestry research and guide her through her senior thesis

project, with the prospect of her future employment with PESACRE.

Presentations and Group Discussions

Nutrient budget study. Prior to initiating the biological studies, as well during the field

data collection period, the research plan was presented to RECA farmers through a series of

meetings arranged by leaders of the producers' group. A summary of group presentations

made to RECA and other local organizations is provided in Table 3-1.

The phosphorus (P) and nitrogen (N) budget study was first introduced during an

ecology course given by SOS Amaz6nia, an interdisciplinary group ofBrazilian environmental

educators. Following an SOS-led session on soil fertility, I used a "bank account" analogy

to explain how the system, or "agroforestry account" is comprised of different reserves (i.e.,

soil, above- and below-ground biomass, litterfall, microorganisms, etc), and how this"capital"

is transferred in and out of the account with fertilization, mulching, harvesting, leaching and

other processes that add or remove nutrients from the system. We then briefly discussed how

management practices may contribute to conserving and building nutrient capital, increasing

the likelihood that the agroforestry system would be productive in the future.

After the basic study methods and participants' responsibilities were outlined,

recommendations by farmers were solicited and recorded on a flip chart. Their requests

included that (a) the research involve several farms, (b) I attend official RECA group

meetings to become acquainted with farmers and update them on the study's progress, and

(c) a Brazilian be trained to continue this type of research after my departure. Following the

discussion, several farmers volunteered their agroforestry plots for the study, and it was

agreed that the final site selection would be made after visiting the farms.

Root ingrowth and soil studies. In addition to identifying P limitations to agroforest

productivity, the root ingrowth bioassay was also designed to address the farmers' concerns

about peach palm root competition by comparing the growth of roots by this and other

agroforest components into ingrowth cores buried in the soil for a specified period of time

(Chapter 4). Soil samples from agroforestry systems and adjacent native forest on eight

farms were to be analyzed to determine how the agroforest soils had changed chemically since

conversion from native forest. These studies were introduced at RECA's semi-annual

assembly (August 1995) in which all members ordinarily meet to discuss problems they are

having on-farm and issues facing RECA as an organization. A presentation on the basic

Table 3-1. Summary of presentations and discussion groups conducted as part of participatory research process to study nutrient and fine
root dynamics in the RECA agroforestry system (AFS). Number of participants in parentheses. Translated from Portuguese.
Date Topic and Objectives Participants Methodologies Outputs

8/19/95 Nutrient budget proposal
-Discuss nutrient cycling
-Present proposed study
-Solicit feedback & volunteers

9/16/95 Root study proposal
-Discuss root competition
-Present proposed study
-Solicit feedback & volunteers

2/24/96 Root study update
-Update farmers on progress
-Discuss hypotheses
-Discuss participants' field observations

8/23/96 Root & soil study update
-Present preliminary data
-Discuss possible interpretations of
preliminary results

-RECA leaders
attending SOS
course ( 35)

-RECA members
attending semi-
annual assembly

RECA members
attending semi-
annual assembly

RECA members
attending semi-
annual assembly

-Flip chart drawing of
AFS nutrient cycle
-List of study objectives
-Bank analogy

-Flip chart drawings of
root competition
-Ingrowth core used
-Group discussion

-Demonstration of actual
samples (ingrowth cores
with roots)
-Group discussion

-Simple bar graphs on
flip chart paper
-Sample soil analysis
sheet on flip chart paper

-Introduced concepts
of nutrient cycling
-Farmers' criteria
-Volunteer study sites

-Group discussion of
root competition,
-Farmers' criteria
-Volunteer study sites,

understanding of root
study by farmers,
demonstrated by
-Updated and received
feedback from farmers
on root study results

Table 3-1--continued.

Date Topic and Objectives



9/18/96 AFS nutrient cycling
-Present preliminary data on P&N
harvest removal, soil & plant stocks
-Discuss current management
-Discuss strategies to minimize soil
degradation & enhance nutrient cycling
-Generate list of management options
9/20/96 Nutrient cycling in RECA AFS
-Present methods & preliminary results
to local research institution
-Solicit feedback on results
-Encourage future research
collaborations with RECA

9/22/96 AFS Nutrient Cycling (SOS course)
and -Present nutrient cycling & removal
9/27/96 with harvest using preliminary data
-Discuss current land management &
effect on nutrient cycles & soil fertility
-Determine viable management options
to maintain productivity

11/4/96 Nutrient cycling in RECA AFS
-Present methods & preliminary results
to PESACRE & other local NGO's
-Discuss implications for AFS
management and productivity

reforestation team
(Equipe de
plantagAo, 17)

Rio Branco

Farmers in RECA
attending SOS
Education course
(25 each session)
UFAC students
EMATER agents

-Colored drawings on
flipcharts of AFS with
approximate nutrient
stocks in soil, biomass
and harvest
-Flip chart list

-Slides of methodology
-Overhead transparencies
with results

-Game: Banco do Brasil
(felt board and cut outs)
-Enlarged photo series
(forest, burn, planting,
mature AFS)
-Small group discussions
-Flip chart lists

-Reintroduce nutrient
cycling in RECA AFS
-Farmers list practices
affecting nutrient cycle
-Listed options for
reducing soil

-Greater awareness by
research institutions of
the potential for on-
farm research with

-Farmers discuss
nutrient cycling in
diverse land-uses
-Small groups generate
list of management

-Slides & transparencies
with methods & results
-Banco do Brasil game
-List of farmer options
for AFS management




concepts of root competition generated an animated discussion among farmers who had

observed the "aggressive roots" of peach palm in their own plantations, and were concerned

that it might eventually dominate, or even "kill" other components of the system. Farmers

were also interested in having their soil analyzed, as long as they also received the results.

After discussing criteria for farm site selection and participation in the study, the producers

themselves selected eight farm sites from among those volunteered by individuals.

Implementation of field studies. Once the field studies were underway, farmers were

updated on the progress of the research during the next two official assembly meetings (Table

3-1). These brief presentations simply served to keep RECA members not directly

participating in the field studies informed on the status of the research and foster interest in

the investigative process. Attending the two to three day meetings also allowed me to

become better acquained with RECA farmers and learn a great deal more about the

organization, and the challenging issues facing it and individual households. During this

period, however, most of the contact I had with farmers occurred through conversations with

household members held during my extended stays with families participating in the field

studies (discussed below).

Preliminary data generated from the concluded root and soil field studies, as well as

the ongoing P and N budget research, provided the basis for the nutrient cycling presentations

and discussion groups held in September and November, 1996. As indicated in Table 3-1,

a variety of techniques and tools were used to present the data and stimulate discussion. An

example is shown here for the nutrient cycling modules of the SOS Ecology courses (Table

3-2). These two sessions, along with an earlier session with RECA's reforestation team,

culminated with a list of management options for maintaining soil fertility generated by


Alternative (non-lecture) training techniques, such as games and small group

discussions, were employed to create a non-threatening and engaging environment, especially

as education levels varied considerably among RECA farmers. The basic objective of the

Banco do Brasil (official state bank) game used during the SOS courses was to demonstrate

how nutrients were transferred from different "accounts" in the agroforestry system, or

removed entirely with harvest. For example, when the cut-out of a peach palm crown was

added to the system depicted on the felt board, six Brazilian dollars (reais) were moved from

the soil account in the bank to the plant biomass account (the monetary value being based on

preliminary data, e.g., one Brazilian dollar equals one kg P/ha). When peach palm fruit was

added to the tree, its value was moved from the soil to the plant biomass account. When the

fruit was harvested, this amount was withdrawn from the bank. After a demonstration,

participants were asked in which account to place nutrients (or remove nutrients from) as

each component was added to or moved around in the agroforestry system. The felt board

and bank were also used to demonstrate changes in N cycling dynamics when understory

weeds were cut down and left to decompose, or when leguminous shrubs were included in

the system to add nitrogen through fixation of atmospheric N2. During both sessions, the

game encouraged a discussion about the importance of organic matter as a source of plant

nutrients, and management practices that favor organic matter build-up in the agroforestry


Table 3-2. Lesson plan for participatory research activities, nutrient cycling module for SOS
environmental education course given to RECA farmers on September 22 and 27, 1996.
Total session time: 3 hours. Translated from Portuguese.
Session Objectives
1 Demonstrate the concept of nutrient cycling and its role in maintaining agro-
ecosystem productivity to RECA producers using data from RECA
agroforestry system
2 Identify and discuss current land management practices of RECA farmers that
potentially benefit or degrade nutrient cycling processes
3 Develop a list of practical land management options that may help maintain soil
fertility and future agroforestry system productivity

Time Method


20 min Interactive lecture (emphasis on
questions to farmers) to define
concepts of nutrient cycling
40 min Game: Banco do Brasil (uses
bank analogy to describe
nutrient inputs, outputs &

15 min Break

40 min Group Discussion: Impact of
Agricultural Practices on
Nutrient Cycling

20 min Small Groups: discuss AFS
management & adaptations to
enhance nutrient cycling
30 min Group summary & evaluation
of proposed practices & effects
on nutrient cycling & AFS

-Flip chart with questions and room to list
answers (what is a nutrient cycle, why
important to land managers?, etc.)

-2 lxI m pieces of felt
cardboard cut-outs of AFS components
(trunks, leaves, fruit, roots, soil, legumes)
-Paper cut-out of bank with 3 accounts
(soil, fallen litter, live biomass in plants)
-Felt cut-outs of money (different values)
-Chart with "monetary (kg/ha) values for
AFS components based upon data

-Series of enlarged color (Xerox) photo
depicting a) native forest, b) cleared &
burned land, c) newly planted seedlings on
recently burned land d) pupunha
-5 groups, each given one photo to
stimulate discussion and flip chart & pen
to list possible adaptations
-Flip chart table to be filled out as by
entire group with four columns:
Management practice; Objective; Benefit
or degrade nutrient cycle; Why?


Another important objective of the participatory process was to familiarize local

research and extension organizations with the on-farm studies underway in Nova California,

and facilitate future investigative collaborations with RECA farmers. For this reason,

research methodology, as well as preliminary study results, were presented to PESACRE and

EMBRAPA in Rio Branco (Table 3-1), as well as to the Soils Department at the University

of Vigosa in the south central Brazilian state of Minas Gerais. Included in the results

presented to these groups were the flip chart lists of management options generated by RECA


Household Interviews

Root ingrowth and soil studies. Informal interviews, or conversations, were held with

both men and women of the eight families participating in the root ingrowth and soil studies

regarding the role ofagroforestry in a household's production strategy. Current agroforest

management practices, as well as constraints to and opportunities for modifying management

were also discussed. These conversations took place during three two-day visits with each

family. Previous visits to these farms prior to initiating the study had fostered a good working

relationship with each family. The actual installation of the root ingrowth study on each farm

required about a day with the help of family members. To encourage a "learning"

environment, the study objectives were reviewed and participants were asked to make

predictions about the results based upon their previous experience with the agroforestry


The root ingrowth core bioassay methodology is relatively straightforward (Chapter

4), andparticipants appeared to understand it conceptually, as demonstrated by their questions

and predictions. Conversations with other family members occurred during meals, farm

walks, and other "leisure periods". Observation of farm and household activities provided

additional insight into agroforest and overall farm management. The families were visited a

second time when ingrowth cores were removed from the soil, and the results from soil

analyses were returned to each family on the third visit. The latter visit was used to discuss

soil fertility and agroforest management. The soil analysis "sheet" (Table 3-3) was modeled

somewhat after the form given to farmers by EMBRAPA. Although the language used was

rather technical, the sheet was quite useful in initiating discussions about soil acidity,

nutrients, and the potential use of leguminous cover crops, fertilizer and organic residues to

improve soil quality.

P and N budget study. Frequent (often daily) contact with the families participating

in the nutrient budget study provided an opportunity for in-depth conversations regarding

constraints to agroforest management and production, as well as continual observation of

farm and household activities. While the nutrient cycling study was installed on only one site,

the farm site itself was owned by one family who hired another family to live on the property

and manage it, and a close relationship with both households provided considerable insight

into differing perceptions of agroforest productivity and sustainability. Because it was

impossible for me to be present every time fruit or palm heart was harvested from the system,

family members recorded the weight of fruit removed from the study plots, and collected

rainfall samples immediately after storms.

Table 3-3. Sample soil analysis sheet given to farmers participating in root ingrowth bioassay and studies
after completion of analysis collected from their farms. Analyses performed included pH, % organic
matter, extractable cations and P, and total N and P. Translated from Portuguese.
University of Florida The RECA Project
Soil Analyses 0-20 cm depth (November 1995)
Property of Sr. Aluizio e Sra. Carmelita Gongalves, Group BR

Your soil mean Your soil mean
agroforest native forest
pH 4.9 4.9 4.4 4.3

Organic matter (%) 1.4 2.0 1.7 2.0

Ca 189 350 31 102

Mg 41 61 35 42
K 26 35 26 36
A] 179 267 196 310
Fe 39 32 32 49
Na 3.5 4.7 2.2 3.0
P 1.4 1.1 1.6 1.5
P total 383 410 352 360

N total 1,217 1,690 943 1,599
mg/kg = ppm

Return of Final Results

As mentioned earlier, the final step of returning the final results to RECA and

PESACRE is awaiting completion. That it will have taken two years from the time I left Acre

to return the results is not entirely satisfactory, and this may represent one drawback of

participatory research, at least at the doctoral level. The lag time between process and

product underscores, once again, the need to make the process count. Currently, I am

planning, in collaboration with PESACRE and RECA, a course to be conducted in Acre in

on "Nutrient Cycling and Agroecosystem Sustainability". One of the objectives of this course

will be to present the final research results to RECA, as well as review and evaluate the

research process. Included in the final results will be the lists of management options

generated by the participants themselves, as well as actual data and interpretation of the field

studies. Hopefully this will provide a forum in which to discuss which practices are actually

applied in the field and by whom. Field visits would help identify how both management

practices and the agroforestry system itself have evolved since the research was conducted.

The final research results will be discussed in relation to past and present agroforest

management. Management options recognized by farmers as immediately feasible will be

reviewed. Because of their importance to sustained production (Chapters 4-6), practices that

may require more resources and training, such as widespread planting and regular pruning of

leguminous cover crops, and seasonal directed application of soil amendments, will also be

discussed. A review of the research process will provide an opportunity to discuss basic

research methods and their application by farmers as outlined earlier in this chapter.

We will present data generated from the field research in an extension pamphlet that

will be translated into Portuguese and disseminated by PESACRE. The focus of the

extension brochure will be how nutrient cycles of tree-based agroecosystems can be managed

to maximize their potential for sustained productivity. In addition to offering information and

management recommendations to farmers, the pamphlet may also provide a framework for

NGO's, such as PESACRE and SOS Amazonia, to use when conducting environmental

education courses for other rural producers' groups in Acre. PESACRE has also indicated

that additional collaborative efforts to develop an environmental education program could

provide an important last step in the participatory process.

The Participatory Process: Lessons Learned

The Benefits of Participation

Undoubtedly, the research benefitted directly from the participatory process. The

open discussions held with families and focus groups, engendered, in part, by the trust built

as a result of encouraging genuine multi-lateral exchange, revealed a lot of information that

that might not have otherwise been apparent to outside researchers. The role of perennial

crops in colonist households was clarified, and constraints to and opportunities for improving

agroforest management were identified. Extremely important to this process was the fact

that RECA was well organized and held regular meetings in which my participatory activities

could be integrated. As a result, RECA farmers remained aware of and interested in the

investigative process and the results it would potentially render, as demonstrated by their

thoughtful questions, observations, recommendations and continued willingness to participate

throughout the field research period.

The management recommendation lists generated by different discussion groups

(discussed below) represent the most tangible output of the process itself. The process might

have been much more limited had I not had a formal venue in which to conduct the

presentations. The real test of the participatory process lies in the extent to which

management recommendations generated by farmers are actually applied, both now and in the

future. Hopefully, the official fora provided by organized group discussions helped stimulate

on-going dialogue among RECA farmers and other research and extension institutions about

sustained production, ecological processes, such as nutrient cycling, and farmer management.

Further evidence to suggest that farmers benefitted from the process was the fact that

several individuals approached me to help them design their own on-farm research.

Experimentation among RECA farmers is nothing new. For example, many farmers had

already conducted "informal" on-farm research with different legumes species, and I

encouraged these producers to share their results with other families, as well as with

EMBRAPA in Rio Branco, whose researchers were in the process of initiating "new" studies

of legumes in agroforest understories on farms surrounding Nova Calif6rnia.

Although it is important that producers believe in the validity of their own research,

such farmer-initiated research could also benefit from training by professional researchers.

For this reason, a short course designed for farmers on "basic field research methods",

conducted in collaboration with organizations such as PESACRE and EMBRAPA, may be

an extremely effective way to improve agroecosystem management. Such a course would

provide an opportunity to discuss (a) the valuable experience farmers have gained through

experimentation on their own, as well as the strategies they employ, (b) differences farmers

may have noted between their research and that conducted by trained scientists, and (c) the

pros and cons to different investigative approaches. We could also discuss why controls and

replication are used in scientific research, and how they might use these "tools" to enhance

their own experimentation, if they are not already doing so. Grassroots developmental

organizations, such as World Neighbors, have successfully taught indigenous rural people to

use mathematics and statistical analyses in the design and interpretation of on-farm research

(Ruddell and Beingolea 1995). How the results of on-farm research can be shared (both

formally and informally) with fellow farmers, researchers, and extensionists would also be a

useful topic for discussion. Such a course could be included in an environmental education

program for rural producers conducted by NGO's such as PESACRE.

The Challenges of Participation

There is a lot to be learned from attempting to combine community participation and

doctoral field research. Although definitely rewarding, it added responsibilities, as well as

risks, to the research process. Initially I was determined to present "scientific" research to

farmers as simply as a series of steps, much like planting and harvesting a crop. In the end,

there were many "steps" that I found difficult to explain to participants, or for which it was

difficult to create a situation that allowed participants to arrive at a better understanding of

the process on their own. For example, while I believed the root ingrowth bioassay would

provide a relatively easy-to-understand, yet scientifically-sound, means to assess root

competition among species and phosphorus deficiencies to plant growth in the field, the

method presented a few problems. For one thing, root growth in natural conditions is

tremendously variable, and although the study was designed to accommodate variability


across farms, it appeared that farmers made up their mind about study results based upon

what they observed in their own field, and from conversations with other farmers. For

example, when farmers saw more peach palm roots protruding from unwashed, unseparated

ingrowth cores, they concluded that peach palm was, as they predicted, an aggressive

competitor. At this point it was difficult, and even questionable, to encourage participants

to withhold judgement on the preliminary results until a log transformation and analysis of

variance had been performed on the total length of roots found in cores (Chapter 4). This is

part of the process they did not participate in, and all that could be done was encourage

discussion among participants about what they were seeing as we removed the ingrowth cores

from the soil. Potentially, this presents a dilemma, because it is the researcher's responsibility

that study results are not erroneously interpreted, but one cannot resort to the approach:

"take my word for it, this is the way it is" if one is to maintain a participatory process. In this

case, peach palm root growth was greater than that ofcupuassu in cores buried in agroforest

alleys, supporting the farmers' hypothesis. However, in cores bured in agroforest rows,

beneath the dripline of the cupuassu canopy, there was no statistically significant difference

in root growth between the two tree species. Moreover, when it came to assessing

phosphorus limitations using the ingrowth cores, the data were not easily interpreted, even

after statistical analyses were performed, and the results pointed out some methodological

weaknesses of the root ingrowth bioassay (Chapter 4). Similar situations are not uncommon

in many fields of research; thus, it is an issue that must be addressed both prior to research

initiation and throughout the entire participatory process. Perhaps it was more problematic

in this study because of the type of agroecological research conducted, which concentrated

on biological processes, and not technology creation or evaluation. For example, in many

"on-farm trials", farmers may test fertilizer applications, genetic varieties (e.g. Hildebrand and

Poey 1985) or even planting locations (Kainer et al. 1998). From these types of studies,

farmers can "pick" the technology which performs the best under the specific environmental

and socio-economic conditions they face. In the present study, there was no technology

tested, rather, farmers were asked to evaluate their own practices in relation to its effect on

a process, so it was important that they understand the process. A frank discussion at the

outset about the scientific method, and and how it is met through the research objectives, may

facilitate a better understanding among participants about the limits of particular studies in

addressing specific questions.

This also points out the need for careful selection and execution of research methods;

however this is not always possible in more remote resource-limited areas, and could

therefore preclude research in regions that need it the most. Rocheleau (1991) notes that we

can improve our capabilities for participatory research if we "abandon fixed packages of

research methodology and broaden our horizons to include a wide variety of principles,

methods and other peoples' field experience". Such an approach does not necessarily allow

for controlled conditions, nor the use of tools that produce predictable outcomes. This

demonstrates the delicate balance between remaining faithful to the scientific method and

open to new constructs of knowledge creation.

Finally, participatory research takes a lot of energy and concentration to ensure that

the process is continually beneficial, and not exploitative, for all parties involved. I had to

be on guard constantly so as not to let my personal research anxieties prevent me from

hearing what the farmers had to say. I remember one day in particular when I was appalled

to realize that I had been so concerned with the difficult logistics involved with field work that

I had not concentrated enough on my interaction with the farmers. One must continually ask

oneself "are participants really gaining from the process, or just supplying labor, land or


Information Gained Using a Participatory Approach

The Role of Agroforests in RECA Households

From interviews with only 10 out of 300 RECA households, it is not possible to

generalize about the role of agroforestry in RECA farms. The households interviewed had

very diverse cultural and socio-economic backgrounds, and their reasons for adopting and

maintaining agroforestry as part of their production strategy also differed. However, three

common themes emerged from these interviews about the role of the cupuassu-peach palm-

Brazil nut agroforestry system in household production strategies. In general, an increase

in farm household income from the sale of agroforest products

a. allowed poor farmers, who might otherwise have abandoned their land or sold it to

ranchers, to continue farming profitably on the same land,

b. provided farm families with the means to purchase (i) household durable goods (such

as furniture, diesel generators and satellite dishes), (ii) labor to help with farm

activities, and (iii) livestock, especially cattle,

c. motivated farmers to open new areas of forest each year for perennial crop

monocultures (such as coffee and palm heart), because they anticipated a future drop

in agroforest productivity and/or change in the marketability of crops such as

cupuassu fruit.

These points serve as hypotheses to be tested with rigorous surveys. They also offer

some insight into the role of agroforestry system adoption in decreasing farm-level forest

clearing. Although several families claimed that agroforestry system adoption allowed them

to clear less native forest because they were no longer obliged to produce annual crops for

sale, the third point suggests that many RECA farmers are practicing a "shifting cultivation"

of perennial crops, that is, continual clearance of forestland for the establishment of perennial

crops in anticipation that the older systems will lose productivity in the near future. In fact,

when questioned about the period of time they anticipated the first cupuassu-peach palm-

Brazil nut systems to remain productive, most believed that cupuassu production would cease

within eight to ten years of planting, and many households had already established younger

monocultural plantations to avoid root competition. This same attitude was demonstrated by

the recommendations made by discussion group participants from the RECA reforestation

team to "intensify production of one species by planting monocultures" (Table 3-4).

This transitory approach to perennial cropping reveals farmers' anxieties about soil

fertility, and perhaps a basic disbelief in the potential for sustained production by tree-based

agroecosystems. Coffee was cultivated in a similar manner from the 1800s to mid 1900s in

regions of southeastern Brazil previously covered by the Atlantic forest (Laakkonen 1996).

In states such as Minas Gerais, monocultural plantations of coffee were planted on slopes

cleared of native forest vegetation. Without soil amendments, the plantations were

productive for an average of seven years before they were abandoned, during which time

additional forest was cleared for new coffee plantations that would come into production

about the time the others failed (Dean 1995). One RECA farmer frankly admitted that it was

a shortage of labor, and not the potential longevity of perennial crops, that kept him from

clearing additional forest.

Constraints to Agroforest Management

Labor was cited by all households as one of the largest constraints to modifying

existing agroforest management practices. For example, the labor burden incurred when

cutting climbing legume (Macuna spp.) vines from the canopies of cupuassu trees was

mentioned as a reason for eliminating the thriving N-fixing species from the system. Nearly

every farmer had experimented with other "shrub" forms of legumes (e.g. Pueraria spp);

however, in most plantations, they were left to grow, unpruned, in the understory throughout

the season because it required a lot of work to cut them down. During the dry season the

dead legumes were viewed as a fire hazard and for this reason many farmers eradicated the

shrubs. Interestingly, some families found it lucrative to harvest the seeds of some legumes

and sell them to buyers interested in establishing leguminous cover crops. However, by the

time the field studies took place, legumes had been eliminated from many agroforest

plantations because replanting the legumes every year was not feasible for families.

Unfortunately, the legume species most successful at reestablishment through natural

reseeding wasMacuna, a species viewed by farmers as most impractical from a management

standpoint. Furthermore, other legumes did not compete well with the native understory

herbaceous vegetation, which often exceeded 2 meters in height before it was cut down and

left to decompose. Families understood that the "weeds" in the agroforest understory

competed for nutrients and water with the system's tree components, but generally

households had only enough labor to cut down the herbaceous understory once or twice a

year, at best. Other labor-intensive farm production-related activities included annual crop

production (for household consumption); seed germination and seedling propagation in on-

farm nurseries; establishment of new perennial systems that included crops such as pineapples,

coffee and native timber trees; vegetable gardening; small and large livestock care, including

pasture creation and maintenance; well maintenance and water transport; processing and

transport of harvested products, medicinal plant propagation and collection; and forest

extraction (medicinal herbs and clay, Brazil nuts, game, fruits, seeds and seedlings for

planting, etc).

Another constraint to agroforest management was a lack of access to chemical and

organic fertilizers, and technical information about their use. Because little station research

had been done in the region on fertilizer use in alternative cropping systems (i.e. non-annual

crops), even extension agents had little idea about the most effective and efficient use of the

prohibitively-expensive soil amendments available, such as triple superphosphate and lime.

Organic fertilizers, such as cow manure, were often applied to home gardens in which

vegetables for household consumption were grown, or applied to enrich the soil used for

seedling germination and propagation. Plant residues were fed to small livestock, or were

viewed as too burdensome to transport material from the point of origin to the agroforest.

Most families, however, did leave peach palm residue (leaves and stem) originating from the

harvest of small basal offshoots for heart-of-palm in the agroforest. Operationally, these

residues were not strategically placed, but left to decompose where they fell. As mentioned

earlier, while leguminous cover crops were not pruned for maximum residue production, they

were initially left to decompose when they died during the dry season, and some families

found the residues to be a good livestock food supplement.

Finally, as discussed in Chapter Two, reliable transport, on- and off-farm processing,

and marketing of agroforest products were cited as some of the largest obstacles facing

households and the RECA organization. Particularly disturbing was the fact that large

harvests of peach palm fruit frequently spoiled while awaiting transport from farm to market.

High processing costs ofcupuassu pulp, due to Nova Calif6rnia's unreliable electricity supply,

raised the cost of marketing this product, and thus lowered the price farmers received from

the RECA organization. Financial losses such as these made farmers reluctant to invest

precious resources in more intensive agroforest management practices, regardless of the

system's potential for sustained production.

Management Options Generated by Farmers

Reforestation team. A very detailed list of agroforest management options was

generated by members of RECA's reforestation team (Table 3-4), who were "tecnicos" or

farmers trained technically to help other farmers. Most of the recommendations on this table

are presented as stated by farmers (translated from Portuguese), although there are a few

(identified by italics) that I suggested myself From previous farm visits it was apparent that

many households had previously employed, or were currently practicing, some of the options

listed, especially under the "leguminous cover crops" and "organic matter" categories.

Moreover, these practices had been discussed repeatedly in sessions led by myself, SOS

educators, and extension agents, and the lists demonstrated that producers were aware of

these practices. However, comments made by individuals during the session, as well as farm

visits and household interviews also revealed that many of the recommendations listed were

not necessarily being implemented. To a great extent this may have been due to the labor

constraints already discussed. For example, many farmers did not have the labor resources

to cut down/cultivate legumes and weeds to maximize understory nutrient cycling and

minimize competition, or to ferment cupuassu pods and other plant residues for compost.

However, at the end of the discussion session, participants did agree that most of the options

listed were desirable in order for their system to sustain productivity for a longer period of

time. Labor, time and monetary constraints made it difficult for farmers to adopt some

practices (indicated by the letter c or d in Table 3-4), despite their beneficial role in maintain

productivity. Some options, such as the use of lime and phosphate rocks, were not viewed

as feasible, because of their expense and inavailability.

Another recommendation viewed as impractical was felling the larger (> 8 m in height)

peach palm offshoots, and cutting up the stems and leaves for use as mulch beneath the

cupuassu and Brazil nut trees. I suggested this in response to comments made by producers

about the fact that during agroforest establishment most farmers had allowed the

offshoots to grow very tall (up to 16 m), not realizing that fruit and heart harvest from these

stems would become difficult, if not impossible. Although felling some of the larger

offshoots would theoretically (a) liberate and "recycle" nutrients stored in "underutilized"

Table 3-4. Management recommendations for maintaining agroforest soil fertility and decreasing root competition among agroforest
components generated by members of the RECA reforestation team during a participatory session entitled: Nutrient Removal from
Agroforestry Systems (held on September 19, 1996). Translated from Portuguese, italics indicate recommendations suggested by researcher.
Goal of Management Practice
Maintaining Soil Fertility Reducing Root Competition

Leguminous cover crops in agroforest understory
-Plant legumes (with bacteria) to promote N2-fixation ab
-Plant legumes in fallow fields b
-Plant legumes without burning fallow vegetation
-Cut down/cultivate legumes to recycle organic matter 0,b
-Cut down/cultivate weeds to recycle organic matter'
Organic matter in agroforestrv system
-Maintain an efficient nutrient cycle with green cover
crops and tree crops b
-Diversify plantations with legumes, native
timber tree species, shrubs, coffee, medicinal &
other native herbaceous plants 'b
-Bring/incorporate organic matter from forest'
(branches & leaves, etc.) into agroforestry system
-Apply cow manure to agroforest soil'
-Apply plant residues, especially palm heart harvest'
-Compost plant residues'
-Ferment cupuassu pods at factory for compost
-Maintain organic matter layer with weeds & legumes ab

Reduce peach palm offshoots (maintain only three),'
-Roots die with elimination of stems?
-Transfer nutrients stored in biomass to soil
-Increase harvest of peach palm heart
-Use residues from cut peach palm offshoots as green
manure beneath cupuassui and Brazil nut
Intensify production of one species by planting monocultures
-Plant peach palm for heart production in monocultures'
-Eliminate peach palm in future mixed cropping systems"'b
-Larger spacing between trees in future plantings
-Plant monocultural plantations of cupuassu (discussed
potential problems of disease and pests)"

Inorganic inputs Apply residues from palm heart harvest beneath cupuassu'
-Phosphate rocks in organic matter layer' -Transfer nutrients stored in peach palm to soil beneath
-Apply lime to add calcium and lower soil acidity4 cupuassu
-Directed and sparing application of chemical fertilizerd -Orient root growth toward decomposing organic matter?
'practiced operationally by many farmers; practiced experimentally by a few farmers; "rarely practiced, dnot practiced

Table 3-5. Management recommendations for maintaining agroforestry system (AFS) soil fertility generated RECA farmers attending
nutrient cvcling module of SOS Ecolovgy class (held on Seotember 27. 1996) during small uroup exercise. Translated from Portuuese.

Management Recommendations
Group one Group two Group three Group four
-Directed application of cow -Manage AFS adequately to -Organic soil amendments' -Leave wood residues to
manure maintain efficient cycling of decompose on soil'
-Green (leguminous) cover -Increase the AFS diversity by -Leguminous cover crops ",b -Maintain "dead" cover with
crop ofMacuna spp.5 b planting native timber tree palm residues to enrich soil'
-Fertilization with legumes in -Fertilization with organic -Native timber tree species' -Plant native timber tree
most practical manner' matter' species to add organic matter'
-Reduce root competition by -Nutrient export (?) -Control nutrient removald -Plant legumes to furnish
eliminating weeds and older nitrogen *b
peach palm stems'
-Leave organic matter when -Reduce competition by' -Apply animal manure to"
harvesting fruits to minimize controlling understory weeds improve plant production
impoverishment of soil'

'practiced operationally by many farmers; practiced experimentally by a few farmers; 'rarely practiced, dnot practiced

above-ground biomass, and (b) potentially decrease intraspecific competition by decreasing

peach palm overstory biomass, participants concluded that the practice would be both labor

intensive and dangerous to farmers, possibly damaging to agroforest components (if other trees

were struck by fallen palm stems), and would not result in higher palm heart yields since the

larger stems no longer produced "tender" apical buds. Participants did agree that an effort

should be made to maintain only three stems per palm in younger plantations; two young

offshoots for heart harvest, and the larger "mother" stem for fruit production. They also

concluded that it would be feasible to place as mulch the residues of young palm offshoots

harvested for heart beneath cupuassu and Brazil nut canopies.

RECA farmers. Despite the fact that the producers comprising the reforestation team

had access to greater technical training, the management option lists generated by "untrained"

RECA farmers in small groups during the SOS nutrient cycling module were very similar to

those listed by the "tecnicos" (Table 3-5). Moreover, recommendations were similar among

the four groups, i.e., use of leguminous cover crops and organic residues, agroforest

diversification with native timber species, and cow manure application. The strength in these

lists is that they were made by the producers themselves, organized in small groups, so that

their independent responses indicate that these practices were commonly recognized by

farmers as beneficial to sustaining agroecosystem productivity. The "agroforest

diversification" option also demonstrated the producers' knowledge of improved organic

matter and nutrient cycling in structurally diverse tree-based systems, and their role in

maintaining soil fertility.

Compared to the "tecnico" group, however, it was more difficult to elicit from these

producers which practices were truly feasible and which were impractical. When the small

groups were reunited to "report out" their lists, we discussed management options in greater

depth; for example, how legumes could be pruned to provide mulch and decrease competition.

In retrospect, it might have been very useful to introduce the "tecnico's" management options

list to see if this group would comment on the feasibility of the recommendations made by the

reforestation team. Furthermore, the "untrained" producers may have been less hesitant to

comment on the practical application of the management options they had cited had they seen

the similarity between their lists and that made by the "tecnicos".

Most participants were aware of management options that could help sustain

agroforest productivity in the future. Yet field visits, interviews, and comments made by

participants during the focused discussion, indicated that, for whatever reason, they were not

currently practicing some of these techniques (i.e., application of manure and plant residues).

These findings suggests that a lack of resources (time, labor, money), rather than a lack of

"understanding," may have prohibited households from employing more sustainable agroforest

management practices.


It has been assumed that offering economically and ecologically viable land

management strategies to Amazonian farmers is crucial to decreasing tropical deforestation.

Amazonian farmers are rapidly adopting perennial-crop based agroforestry systems as an

alternative to shifting cultivation. This land-use may offer a greater degree of ecological

stability if the biological processes that control sustained productivity in tree-based ecosystems

are maintained and/or enhanced through management practices. Encouraging farmer

participation in agroecological research allowed us to gain a better understanding of the

constraints to more intensive agroforest management faced by rural Amazonian households.

Farmer input through focus group discussions also revealed the existing opportunities for

modifying agroforest management to increase its potential for sustained production. As such,

the most valuable output of the participatory process for RECA farmers may be the list of

management options they themselves generated. These lists were very instrumental in

formulating management recommendations that address biological constraints to optimal

agroforest nutrient cycling identified by this research (Chapter 7). Although not an "ideal" list

of management practices, working within the framework provided by the farmer-generated

lists does offer the most promising approach to maximizing the agroforests' potential for

sustained production, given the constraints faced by rural Amazonian households. Stimulating

household and community-level discussions about the role of nutrient dynamics in sustained

agroecosystem production may also encourage farmers to continue experimenting on their own

to develop innovative practices that enhance agroforest productivity and minimize soil

degradation. Finally, the information gained by encouraging local participation underscores

the fact that the longevity of this system as a viable alternative to other Amazonian land-uses

depends on the extent to which it improves the livelihoods of rural households as much as its

potential for ecological sustainability.



Since the late 1970's, the rate of deforestation in the Brazilian Amazon is among the

highest in the world, raising concern because of its potentially negative consequences for

global climate, hydrology, biogeochemical cycles and biodiversity (Skole and Tucker 1993).

While a great share of the destruction is attributed to large-scale cattle ranching, nearly one

third of forest clearing is undertaken by the region's growing population of small farmers,

primarily for the shifting cultivation of annual crops (Fearnside 1993, Skole et al. 1994,

Serrio et al. 1996). As one of many strategies to decrease deforestation rates, it has been

proposed that adding perennial crops to agricultural systems may raise land productivity, and

subsequently allow small farmers to meet food demands with less forest clearing (Sanchez et

al. 1982, Anderson 1990, Smith 1990). Increasingly over the past decade, as the practice of

shifting cultivation has proven economically unviable in the region's nutrient-poor soils,

Amazonian farmers have begun adopting perennial crop-based agroforestry systems, largely

because many agroforest products are high value cash crops that often require less labor to

produce (Smith et al. 1997). Some studies point out that these agroforests can be more

ecologically sustainable than annual cropping systems because the longevity of tree-based

ecosystems promotes a more closed cycling of organic matter and nutrients, a key factor for

the growth of native forests in weathered Amazonian soils (Sanchez et al. 1982, Ewel 1986).

Despite more efficient nutrient cycling offered by tree-based agroecosystems,

maintaining phosphorous (P) availability to plants growing in tropical Ultisols and Oxisols is

problematic for a number of reasons. While nitrogen fixation and rainfall deposition may

serve as significant external sources of N, there is no comparable atmospheric input that can

dramatically increase P availability in P-deficient habitats (Schlesinger 1995). Thus, the

amount of P cycling through natural and low input agricultural systems is determined by the

initial state of the various pools comprising the soil P stock (Stevenson 1986). Phosphorus

uptake occurs from the most vulnerable soil pool, free phosphate ions desorbed or dissolved

from the soil solid phase, often referred to as "labile" P (Fardeau 1996). While this "labile"

pool is difficult to actually quantify because it is continually affected by a myriad of biological

and geochemical factors, a number of procedures are used to quantify readily-extractable P,

and the P measured in such extracts is presumed to be correlated with plant uptake. Much

of the total P stock in tropical Oxisols and Ultisols has been precipitated as insoluble Al and

Fe phosphates, or occluded in hydrous oxides of Al and Fe, as a result of intense weathering,

which render it largely unavailable for short-term plant and microbial uptake. Solution

phosphate concentrations are maintained at low levels because any plant available P remaining

in, or added to, the soil system is sorbed by Al and Fe oxides on the surfaces of clay minerals

(van Wambeke 1992, Fontes and Weed 1996). In these conditions, mineralization of organic

P becomes increasingly important to P nutrition (Stewart and Tiessen 1987, Cross and

Schlesinger 1995), as do mycorrhizal associations and Al- and Fe-solubilizing root exudates

that increase P availability to plants (Chapin 1980, Fox et al. 1990, Bolan 1991).

Nonetheless, numerous studies provide evidence that tropical forest productivity is P-limited

(Vitousek and Sanford 1986, Attiwill and Adams 1995). Thus, maintaining P availability in

biologically and structurally less diverse agroforestry systems undergoing repeated nutrient

removal with crop harvests is a dilemma certain to face Amazonian land managers. The

problem is further aggravated by the fact that many farmers have limited access to chemical

fertilizers and little experience using the large inputs of organic residues recommended to

maintain soil fertility (e.g., Nicholaides et al. 1985, Szott et al. 1991). Studies of low input

annual cropping have shown that with continued harvest, cation leaching and soil

acidification, P availability may decrease to the extent that organic matter decomposition, N

mineralization, and N-fixation is limited because of soil fauna and bacteria sensitivity to P

deficiency (Ewel 1986, Crews 1993).

Critical to addressing the problem of P maintenance in Amazonian tree-based

agroecosystems is a knowledge of (1) how P dynamics are altered when native terrafirme

forest is converted to agroforest, and (2) how long readily-extractable soil P pools can sustain

agroforest productivity without the use of amendments. These questions are particularly

important if commercial agroforestry systems are to be considered both economically viable

and ecologically sustainable alternatives to other more destructive land uses in Amazonia.

Establishing nutrient limitations to plant productivity often requires fertilization

experiments that test the relationship between nutrient supply and plant growth and

development (Marschner 1995). Cuevas and Medina (1988) used root proliferation in

nutrient enriched-ingrowth cores as a bioassay to infer nutrient limitations to fine root growth

in Amazonian forests. Raich et al. (1994) demonstrated that this method could be used to

identify specific nutrient limitations to aboveground forest productivity by comparing root

proliferation response in nutrient-enriched cores to previous forest fertilization studies. It is

generally accepted that many crop plants do proliferate in nutrient-rich patches, but studies

have shown that nutrient-deficient plants exhibit a greater proliferation response than nutrient-

sufficient plants (Caldwell 1994). For example, Ostertag (1998) found that root growth of

tropical forest trees established in P-poor soils was greater in response to P fertilization than

the same forest types growing in less P-limited soils. In P-deficient habitats, roots and

associated mycorrhizae must grow to P sources as phosphate concentrations become depleted

around the rhizosphere because P diffusion through the soil solution is slow (Nye and Tinker

1977). Thus, root proliferation in response to P microsite enrichment could be an effective

tool for assessing P limitations to ecosystem productivity, as well among species within a

system growing in P-poor Amazonian soils.

The objectives of our study were threefold. First, readily-extractable inorganic and

organic P pools, as well as other chemical properties, were compared between agroforest and

adjacent native forest soils to determine how short-term (< 10 years) P dynamics change when

primary forest is converted to perennial crop-based agroforestry systems. Secondly, P

limitations to productivity in eight six-year-old agroforestry systems were studied using a root

ingrowth bioassay to examine fine root response to phosphate microsite enrichment by

agroforest and native forest plants. Finally, a differential response in root proliferation

among agroforest species was examined as a component of inter-species competition. This

third objective was added at the request of the Brazilian farmers collaborating in this study

who were concerned about what they perceived to be aggressive root competition by the

agroforest's palm component.


The Study Area

The study was conducted on eight farms within a 30 km radius from the town ofNova

California, a rural community which lies on the border of the Brazilian states of Acre and

Rond6nia in the western Amazon Basin (67W, 10 S). The life zone in this region is humid

moist tropical forest (Holdridge 1967) and the native non-flooded terrafirme vegetation

comprises both deciduous and evergreen broadleaftree species. Average air temperature is

22 C and mean annual rainfall over the last 10 years is approximately 2,000 mm with a three-

month dry period occurring from June through August (UFAC unpublished). The region's

topography is slightly undulating and soils are predominately Ultisols and Oxisols (Sombroek

1966, Souza 1991). Soils from the study sites are acidic (pH < 5), with an effective cation

exchange capacity less than 12 cmol+ kg clay', high levels of exchangeable aluminum (>40%

Al saturation), and 40 percent or more clay in the top 20 cm (Table 4-1). These properties

are consistent with Oxisols of the Ustox suborder (van Wambeke 1992). Colonist farmland

holdings in the region are typically 100 hectares, half of which are maintained in primary

forest, as dictated by Brazilian law (IBGE 1990). Land use includes livestock pasture, annual

and perennial crops, homegardens, and forest extraction.

Table 4-1. Soil properties ( one SE) in eight agroforests and adjacent native forests at 0-20 and
20-40 cm depth (n=8).
0-20 cm 20- 40 cm Paired t-test
Soil Property p values"
Agroforest Forest Agroforest Forest by soil depth

Sand(%) 27.6 5.6 24.2 5.2 26.2 5.7 25.3 3.9
Silt(%) 25.92.0 35.35.4 22.92.0 27.74.4 0.139 0-20 cm
Clay(%) 46.2 6.0 40.6 5.4 50.9 5.9 47.0 4.7
pH 4.9 0.2 4.3 0.1 4.7 0.2 4.3 0.1 0.006 0-40 cm
Organic matter (%) 2.0 0.1 2.0 0.2 1.3 0.1 1.3 01
M-l Ca(mgkg-')b 349.9152 102.436 105.1 43 45.5 18 0.035 0-40 cm
M-I Mg(mgkg-') 60.813 42.2 7.2 23.45.4 24.25.0 0.115 0-20 cm
M-I K(mgkg') 35.1 17 36.0+2.8 19.0 1.8 21.22.2
M-l Pi(mgkg') 1.080.11 1.540.22 0.190.22 0.430.11 0.051 0-40cm
Ca (cmol+kg") 1.96 0.50 0.50 0.17 0.009
Mg (cmol+kg') 0.50 0.07 0.32 0.05 0.045
K (cmol+kg-') 0.14 0.02 0.13 0.07
Al (cmol+kg") 1.88 0.53 2.17 0.35
ECEC(cmol+kg-')c 4.42 0.34 3.11 0.27 0.004
A] sat (%) 41.0 10.9 68.2 7.7 0.033

Table 4-1-continued.

0-20 cm

Soil Property



20- 40 c


;m Paired t-test
p values'
Forest by soil depth

Total C (gkg-) 16.2 1.5 15.3 1.3
Total N (g kg-') 1.69 0.02 1.60 0.13
Total P (g kg-') 0.41 0.06 0.36 0.04
7 P values ; 0.15 not reported.
bMehlich-1 extractable elements.
SEffective cation exchange capacity (sum of base cations + exchangeable Al).

Table 4-2. Readily-extractable inorganic (Pi) and organic (Po) phosphorus ( one SE) in 0-20 and 20-40 cm depth (n=8).
0-20 cm 20- 40 cm Paired t-test
Soil Property P values'
Agroforest Forest Agroforest Forest by soil depth

M-l Pi(mgkg-') 1.080.11 1.540.22 0.19 0.22 0.430.11 0.051 0-40cm

Bray Pi (mg kg-') 2.64 0.29 3.56 0.63 0.86 0.20 1.08 0.23 0.045 0-20 cm

Resin Pi (mg kg-') 1.31 0.10 2.00 0.29 0.07 0.01 0.28 0.12 0.087 0-40 cm

Bicarb Pi (mg kg-') 0.75 0.19 1.32 0.37
Bicarb Po (mg kg-') 6.19 0.71 6.98 0.50 0.139 0-20 cm
' P values 0.15 not reported.

The eight farms included in this study were volunteered by members of the producers'

organization, Projeto RECA (Economic Partnership for Reforestation). In the late 1980's the

group established a perennial crop-based commercial plantation agroforestry system, one to

two hectares in size, on more than 200 farms. The system is two-tiered, dominated by an

upper canopy of peach palm (Bactris Gaesipaes Kunth), a multi-stemmed monocot cultivated

for centuries by Amerindians throughout Amaz6nia (Clement 1986). The middle canopy is

formed by cupuassu (Theobroma grandiflorum (Willdenow ex Sprengel) Schumann), a

shade-tolerant broad leaf tree native to non-flooded forests of the central Amazon Basin

(Venturieri 1993). A third component of the system is Brazil nut (Bertholletia excelsa Humb.

& Bonp.), a broad leaf upper canopy dominant, also a native to the region's forests (Mori and

Prance 1990). The agroforest's principal products include cupuassu pulp, peach palm fruit

and seed, and heart-of-palm, all of which are harvested as early as three years after system

establishment. Brazil nuts are also an important cash crop throughout Amaz6nia (Kainer et

al. in press), however, at the time of the study, this species had not yet begun to produce


Typically, the agroforest was established by cutting and burning native forest

vegetation and interplanting one-year-old cupuassu, peach palm and Brazil nut seedlings at

a spacing of 7 x 4 meters to complete stocking densities of 190, 150 and 30 trees ha'1,

respectively. During the first year of establishment, leguminous cover crops were planted

in agroforest rows. However, legumes were largely eradicated from the agroforests in the

years following establishment, and native understory herbaceous vegetation was cut down and

left to decompose twice annually. Since agroforest establishment, grazing livestock


were excluded from the system, nor were chemical fertilizers applied on the eight farms under


Farmer Participation

The research was carried out using a participatory (Feldstein and Jiggins 1994)

approach that encourages farmer involvement throughout the investigative process. By

involving farmers it was hoped that the results might ultimately be more useful to them. Prior

to initiating the study, the research objectives and overall plan were presented to RECA

producers. As a result of feedback received by the farmers, a third objective was added,

which was to compare root proliferation response to phosphate among agroforest species,

in an attempt to address their concerns about aggressive root competition by the peach palm.

Personal observation and interviews during repeated stays (3 visits each x 2 days) with nine

families, five focus group discussions with RECA farmers (2 30 participants each), and

project reports all provided information about agroforestry system establishment, farm

management practices, and crop harvests. A more detailed description of the participatory

process is provided in Chapter 3.

Plot Establishment

On each farm a 20 x 20 m plot was located in both the agroforest and adjacent native

forest. Adjacent native forest refers to the primary terra firme vegetation that had been

growing contiguous to the forest that was cut and burned to establish the agroforestry system.

As fire often enters standing forest during agricultural site preparation, a 50-m border was

maintained between agroforest and adjacent native forest plots to avoid sampling under

previously-burned vegetation. There were no observed topographic differences between

paired forest and agroforest plots on any of the farms.

Soil Sampling and Analyses

One well-mixed composite soil sample consisting of 10 randomly located cores was

taken at two depths (0-20 cm and 20-40 cm) from both the agroforest and adjacent native

forest plots on all eight farms. The samples were air dried, passed through a 2-mm mesh

sieve, and hand-picked free of fine roots prior to chemical analyses.

Mehlich-1I (M-1) cations and P at 0-20 and 20-40 cm soil depth were extracted by

shaking 5 g mineral soil in 20 ml of dilute double acid (0.05 N HCL in 0.025 N H2SO4) for

five minutes. Percent organic matter (OM) was quantified using the Walkley-Black

dichromate procedure (Nelson and Sommers 1982), and pH was measured using a Beckman

pH meter and electrode in a 2:1 water to soil ratio. A particle size analysis was performed

using the pipet method (Kilmer and Alexander 1949).

In the top 20 cm of soil, exchangeable base cations (K", Ca2 and Mg") and aluminum

(Al3") were measured after extracting 10 g soil in 100 ml 1.0 M NH4OAc and 1.0 M KCI,

respectively, for 16 hours. Total P in 200 mg finely ground soil was extracted using a

concentrated H2S04/H202 digest at 360C for two hours. Ion concentrations in the filtered

extracts were measured using inductively coupled argon plasma (ICAP) spectroscopy. Total

soil nitrogen (N) and carbon (C ) were analyzed after Dumas (flash) combustion (Nitrogen

Analyzer 2500; Carlo Erba Strumentazione, Milan, Italy).

In addition to M-1 phosphorus, readily-extractable P was measured using anion

exchange resins to exhaustion, the Bray P1 (Bray and Kurtz 1945), and sodium bicarbonate

procedures (Olsen and Dean 1954). Although these four extracts are all used to quantify

readily-extractable P, they solubilize varying quantities of the "labile" pool as a result of

different chemical reactions. The dilute mixed acid in the Mehlich- I extract dissolves Al and

Fe phosphates (Olsen and Sommers 1982) and is used as an index for P availability in Oxisols

throughout Brazil. Anion exchange resins desorb exchangeable Pi without drastic changes

in pH or other soil chemistry. High correlations of resin-extractable P with plant uptake

suggests that resin extracts more closely simulate the physical action of plant roots (McKean

and Warren 1996). The Bray P 1 extract removes easily-acid-soluble Al- and Fe- phosphates

through the formation of fluoride complexes with Al and Fe. The sodium bicarbonate

(bicarb) extract solubilizes a small portion of what is presumed to be readily-mineralizable

organic P (Po), which includes some microbial biomass, in both alkaline and acid soils

(Bowman and Cole 1978, Stevens 1986). Bicarb Pi, used only to calculate Po in this study,

is traditionally used to measure extractable Pi in neutral to alkaline soils (Olsen and Sommers


For resin-extractable P, 2 g of soil were placed with a 2.5 x 5 cm2-' anion exchange

membrane sorptionn capacity = 272 mg P per membrane) in a 50 ml centrifuge tube filled with

30 ml deionized (DI) H20. The tubes were shaken for 16 hours, after which the membrane

was removed and rinsed with deionized water to remove soil. Phosphorus on the membrane

was desorbed by shaking it in 20 ml 0.5 M NH4OAc for two hours. Using the Bray P1

extract, 1 g soil was shaken for one minute with 7 ml of 1.0 N NH4F and 0.5 N HCL in DI

H20 and filtered. For bicarb-extractable Pi, 10 g soil were shaken in 30 ml 0.5 M NaCO3

(pH 8.5) for 30 minutes. Concentrated HCI (1.5 ml) was added to the filtered and centrifuged


extracts to precipitate organic matter. Aliquots of the bicarb Pi extracts were ashed in a

muffle furnace at 560 C and wet-digested with concentrated HCI to extract total P (Ptot).

Bicarb Po was calculated as the difference between bicarb-extracted Ptot and Pi (Olsen and

Sommers 1982). All procedures were conducted in triplicate, and extract P concentrations

were determined colorimetrically using the molybdate blue method (Murphy and Riley 1962)

on a Milton Roy Spectronic 1201 spectrophotometer.

Root Ingrowth Bioassay

A root ingrowth bioassay (Cuevas and Medina 1988) was used to study root

proliferation response to phosphate microsite enrichment. Two hypotheses were tested: (1)

fine root length in phosphate-treated cores would be greater relative to the paired control,

indicating a P limitation to plant productivity, and (2) peach palm fine root length would be

greater than that of cupuassu in both core treatments and agroforest locations, signaling root

competition by the palm that could be detrimental to cupuassu nutrition. Root biomass

between P-treated and control cores was also compared. Root proliferation by Brazil nut was

not studied because the species is a minor component of the system, contributing to less than

8% of the agroforest's trees (30 trees ha').

The ingrowth cores were constructed from high density polyethylene mesh tubes (10

cm tall, 6.5 cm diameter, 4x2 mm mesh size), and filled with 40 g medium-sized vermiculite,

treated with either 100 mL 0.08 M Na2HPO4 (phosphate treatment) or deionized water

(control). The cores were placed individually in 7.5 cm diameter holes dug in the top 15 cm

of soil using a bucket auger, and buried in pairs consisting of both a P-treated and control

core spaced 30 cm apart. Five pairs were buried in each of three locations per farm: (a)

between trees in agroforest rows; (b) in agroforest alleys (between tree rows), and (c )

randomly in adjacent native forest. In the agroforest rows, where both cupuassu and peach

palm roots grew, the ingrowth core pairs were buried midway between the two trees, which

usually fell beneath the dripline of the cupuassu canopy. Core placement in rows was located

at random, but the pairs were buried so that each treatment was equidistant from both the

cupuassu and peach palm. Cores were placed randomly in agroforest alleys, an area densely

populated by peach palm roots, but where cupuassu roots were rarely found, to determine if

the latter would proliferate outside of its "rooting zone" in response to P microsite

enrichment. Ingrowth pairs were placed randomly in adjacent native forest plots to examine

native forest plant response to P microsite enrichment. The cores were buried at the

beginning of the rainy season (mid November) 1995, concurrent with soil sampling, and left

for 100 days.

Upon removal, roots were cut flush with the outside of the ingrowth cores, and the

roots inside the tubes were washed and separated according to Volt and Person (1990). Total

root length was calculated using the line intersect method (Tenant 1975). Root lengths for

native forest species were estimated together, while agroforest root lengths were calculated

separately for peach palm, cupuassu, and the remaining "other" roots. Total root length per

ingrowth core is reported as m'2 to facilitate comparisons with root mass (g m'2). Oven-dried

and ground root tissue was wet-digested with H2SO/H2'02 (Thomas et al. 1967) for analysis

of P concentrations using ICAP spectroscopy.

Statistical Analyses

A paired-comparison t-test was used to identify differences in soil properties between

agroforest and adjacent native forest (n= 8 farms). Root length data were log transformed

to meet the equal variance assumption of analysis of variance (ANOVA) after a normal

probability plot of the residuals revealed heteroscedasticity; untransformed mean values are

presented. Paired differences between control and phosphate-treated cores were analyzed as

a function of location, species, and land-use system treatments by using ANOVA models with

these effects and their interactions. All analyses were performed using SAS (SAS Institute,

Inc., Cary, NC).


Agroforest Soil Six Years Following Forest Clearing

Physical and chemical properties of agroforest and native forest soils are presented

in Table 4-1. Across the eight farms, particle size distribution did not differ between

agroforest and native forest soils. In particular, the clay fraction at both depths did not differ

between agroforest and native forest, providing evidence that soil properties were initially the

same in the paired agroforest and forest plots, because particle size distribution varies little

over time or as a result of management (Sanchez 1987).

Exchangeable Ca and Mg were significantly greater in the agroforest soil, as was pH,

resulting in a higher effective cation exchange capacity (ECEC) and lower aluminum

saturation (Al sat) in this system. Exchangeable Ca, in particular, was nearly four times

higher in the agroforests than in adjacent forests (P < 0.009). Overall, total C, N, P, and soil

organic matter were low when compared to other Amazonian forest soils, and did not differ

between the two systems. Although the concentration ofagroforest M-I extractable bases

was either higher or unchanged from that of forest soils six years after clearing, M-l Pi

decreased nearly 30% in the top 20 cm, and over 55% at the 20-40 cm depth (P < 0.049).

Readily-extractable Pi concentrations in the Bray P 1 and resin extracts of native forest soils

(0-20 cm) were also higher than those of agroforest soil (Table 4-2). Overall, the Bray P1

extract produced the highest extractable Pi concentrations in agroforest and forest soils,

presumably due to the dissolution of Al-phosphates, however, in both systems these

concentrations would be considered inadequate (< 7 mg kg"') for agricultural production

(Olsen and Sommers 1982). Bicarb Po was the largest extractable pool measured, however

neither it nor Bicarb Pi differed between agroforest and native forest soils.

Fine Root Response to Phosphate-Enriched Microsites

In all three locations (agroforest rows, alleys, and adjacent native forest) there was

a trend towards greater root length in phosphate-treated cores (Fig. 4- 1 A). However, using

a t-test, the difference in root length between paired P-treated and control cores was

statistically significant only in the agroforest alleys (P < 0.015). Overall, when data from both

rows and alleys were pooled, mean root length in P-treated cores (74.92 m mf2) was still

greater than in the control (46.96 m m"2) (P < 0.039). Root weight did not differ between the

control and P-treated cores in either agroforest location, but was significantly greater in P-

treated cores buried in native forest (Fig. 4-1B).

A significant root proliferation response to P-treatment was exhibited by cupuassu

only in agroforest alleys (Fig.4-2). Both cupuassu and "other" root length in P-treated cores






rows alleys



I 'll, I I I y
rows alleys

Fig.4-1. In all ingrowth core locations, there was a trend towards greater root length in P-treated cores; this
significant only in agroforest alleys. Root biomass was significantly greater in P-treated cores buried in forests.

effect was statistically



Rows p<0.18



40- '
20- p<0.13 T
20- T 1

0 I 1 --Il I I
cupuassu peach palm other

""I Phosphate



p<0.10 T
.' ;;

cupuassu peach

Fig. 4-2. In agroforest rows, root length did not differ between cupuassu and peach palm
in either core treatment. Root length in phosphate-treated ingrowth cores was significantly
greater than the control for cupuassu and "other" roots only in agroforest alleys.


palm other



(6.38 and 44.61 m m"2, respectively) were significantly greater than in control cores (0.52 and

13.58 m m"2) (P < 0.012 and P 0.007) Cupuassu root weight in P-treated cores (2.41 g

m-2) was also higher than that found in the paired control (0.90 g m-2) (P !< 0.100). Peach

palm root length did not differ between phosphate and control cores, but exceeded that of

cupuassu for both the P- and control treatments in the alley location (P < 0.0001) (Fig. 4-2).

In agroforest rows, there were no differences in root length or weight between the P-treated

and control cores, nor were there differences between peach palm and cupuassu root lengths

in either treatment (Fig 4-2).

Table 4-3. Phosphorus content ( one SE) in fine root tissue growing in phosphate-treated

and control ingrowth cores (n=8).

Tissue P Content (mg g'1) T-test Increase in

Control Phosphate P values P content (%)
cupuassu 0.62 0.02 0.86 0.05 0.041 38.7
peach palm 0.88 0.06 1.01 0.05 0.031 14.8
other 1.10 0.15 1.22 0.10 10.9
forest spp. 0.83 0.07 0.93 0.10 12.1
a P values a 0.15 not reported.

Across all four root groups (cupuassu, peach palm, "other" and forest), tissue P

contents were greater for roots growing in P-treated cores than those in the control (P

,0.010). Analyzed separately by group, this effect was significant only for cupuassu and

peach palm, with cupuassu exhibiting over twice the increase in root P content (38.7%) than

the palm (14.8%) (Table4-3).


Soil Chemistry Following Conversion of Forest to Agroforest

The greater exchangeable Ca and Mg in the top 20 cm ofagroforest soil relative to

adjacent native forest indicates a common effect of the slash-and-bum conversion of primary

forest to agricultural land-use. Ash deposited on the future agroforest site from forest

biomass burning undoubtedly produced a pulse of base cations in the mineral soil,

precipitating an increase in pH and a decrease in exchangeable Al. Many studies report the

favorable effects of burning on soil chemical properties initially following forest clearing

(Ewel et al. 1981, Sanchez et al. 1983, Andriesse and Kioopmans 1984), and this study

suggests that such changes may persist six years after agroforest establishment.

Although the nutritional quality of forest biomass growing in tropical Oxisols is

relatively low (Vitousek and Sanford 1986), the total quantity of nutrients released after

burning mature forest usually supports two to three years of no-input annual cropping before

fields are abandoned to fallow (Serrao et al. 1995, Juo and Manu). In perennial crop-based

agroforests, nutrients otherwise removed from the system during the first few years following

the burn through annual crop harvest or leaching, were stored in growing agroforest biomass

and cycled in fallen litter and decaying roots. When nutrient export commenced with

agroforest harvest three to four years after planting, it would be less than that expected for

annual crops because the first few years of crop production are usually low in a maturing

perennial system comprised of these Amazonian species (Ventereiri 1993, Mora-Urpi et al.

1997). In addition, a change in species composition when forest was converted to agroforest

may have further modified soil properties by altering the quantity and quality (i.e., nutrient

content) of above- and below-ground litter (Binkley 1995, Smith et al. 1998). While it might

be expected that nitrogen-fixing legumes growing in the agroforest understory during the first

three years after establishment would add nitrogen to the soil, total N did not differ between

the two systems. This may be due to the fact that volatilisation of N during forest biomass

burning can be up to 68% of total N content in vegetation (Kaufman et al. 1995) Moreover,

N export from this particular configuarion of agroforest species can be relatively high, often

comparable to that of annual cropping systems as the system approaches six to eight years

(Chapter 6).

Regardless of the origin, an increase in exchangeable bases and pH can also stimulate

decomposition and mineralization of organic matter by creating a more favorable environment

for microbial populations (Nye and Greenland 1960), as well as decrease the soil's P fixation

capacity by reducing APl and Fe3 solubility. Combined with the transfer of P from biomass

to soil following a slash and burn, these factors could initially increase P availability to plants

(Sanchez 1976). Kainer et al. (inpress) found that M-1 Pi concentrations in recently burned

shifting cultivation plots (pH = 5.9, 8.1 mg kg') were markedly greater than those in native

forest (pH = 4.7, 2.8 mg kg') located in the same western Amazonian extractive reserves.

Similarly, Lessa et al. (1996) attributed an increase in bicarb-extractable Po in the top 20 cm

of an Oxisol to accelerated organic matter decomposition and mineralization one year

following savanna clearing in northeastern Brazil.

Readily-Extractable Pi

Six years after forest clearing, there was no evidence that Pi concentrations in

agroforest soil increased as a result of forest biomass burning. In fact, while agroforest M-1

extractable bases had increased or remained unchanged from native forest levels at the time

of sampling, M-1 Pi was considerably lower, as were Pi in the Bray P1 and resin extracts.

Such decreases in agroforest readily-extractable Pi are evidence that phosphate is being taken

up by the aggrading agroforest faster than it can be restored into these pools from less readily-

extractable forms.

Without a measure of total soil microbial biomass in the two systems, it is unknown

if temporary differences in Pi immobilization contributed to lower agroforest extractable Pi

concentrations. However, as discussed below, similar total C-to- P ratios and organic matter

content in agroforest and forest soils suggest that this would not be the principal cause of

lower agroforest readily-extractable Pi. It is also improbable that extractable Pi decreased as

a result of downward movement or leaching through the soil profile, given the inherent

adsorptive characteristics of tropical Oxisols (van Wambeke 1992). However, it is possible

that soil Pi plus that released from plant biomass during the bum was "fixed" into more stable

P fractions not measurable in the extracts used in this study. Linquist et al. (1997) found that

despite large P fertilizer applications exceeding crop removal in a Hawaiian Ultisol, M- 1 Pi

decreased from 35 to 30.5 mg kg' during a four year period. In our study, however, the

relatively short period occurring between forest burning and sampling makes it unlikely that

agroforest P moved into the most recalcitrant occluded soil pools (Cross and Schlesinger

1995). Thus, one might expect agroforest solution Pi to be restored, either through

desorption from secondary minerals or Po mineralization. Using sequential fractionation

procedures, studies have shown that readily-extractable Pi in many agricultural systems is

maintained in equilibrium with less labile pools, such as NaOH-extractable Pi (Hedley et al.

1982, Tiessen et al. 1983, Crews 1996). Despite an 86% decrease in resin Pi (from 2.76 to

0.38 mg kg1) after 13 years of no-input cropping in a Peruvian Ultisol, Beck and Sanchez

(1994) surmised that resin Pi had been sustained by Po mineralization and more stable Pi

fractions because this Pi pool was not large enough to support the total removal of 38 kg P

ha' that resulted from grain harvests. The role of these less readily-available P fractions in

maintaining solution Pi, and ultimately, the productivity of perennial crop-based agroforests,

is relatively unstudied and certainly merits further investigation.

Extractable Po

How soil bicarb-extractable Po concentrations were affected initially following forest

burning is unknown, but at the time of sampling six years later this pool did not differ between

the two systems. This suggests that despite soil conditions more favorable to organic matter

decomposition and mineralization, other factors, such as lower turnover in above- and below-

ground biomass at this stage of agroforest development, may have stabilized or even limited

bicarb Po accumulation in agroforest soils. Linquist et al (1997) suggested that readily-

extractable Po is coupled with C mineralization after observing that bicarb Po declined at the

same rate as soil organic carbon and total N during four years of continual cropping in an

Ultisol. In our study, soil organic matter (OM) was equally low (2.0 %) in both forest and

agroforest soils, most likely because decomposition and mineralization are so rapid in the

tropical environment that only the most recalcitrant OM fractions remain in either system.

In these conditions one would expect the labile Po pool to be maintained at a constant level,

unless modified temporarily by seasonal pulses or immobilization. He et al. (1997)

demonstrated that total soil C-to-P ratios were closely related to soil microbial biomass and

P availability. In our study, neither C-to-P ratios nor the total soil C content differed between

the two systems, suggesting that varying rates of mineralization and immobilization are not

the primary causes for a difference in P availability between forest and agroforest.

Other research has shown that while bicarb-extractable Po may act as a sink or source

during periods of fertilization or P deficiency, absolute changes in this pool after years of

cultivation are negligible (Sharply and Smith 1985, Crews 1996, Schmidt et al. 1996). These

and other studies indicate that increasingly stable P pools, such as NaOH-extractable and

"residual" Po, contribute significantly to plant P uptake over the long term by buffering more

readily-extractable Pi pools, and therefore may better represent the soil's potential for P

maintenance (Beck and Sanchez 1996). Tiessen et al. (1982) suggested that as the most

labile Po fractions mineralize and become depleted through crop uptake, P dissolution from

primary and secondary minerals becomes increasingly important to plant nutrition.

Total P

Whereas significant decreases in agroforest readily-extractable Pi might be explained

by its redistribution among various P fractions not measured in this study, total P provides

an index of the absolute amount of P in the soil. As in the case of base cations, one might

expect greater total P concentrations in agroforest soils as a result of the net transfer of P

from forest biomass to soil following burning. Kauffman et al. (1995) found that total soil

P in slashed primary terrafirme forests in Rond6nia, Brazil increased 40% from preburn

concentrations immediately following burning. Any such increase in agroforest soil P

following forest biomass burning was no longer apparent six years later, as demonstrated by

similar total P concentrations in agroforest (902 kg ha-') and adjacent forest (792 kg ha')

soils. Likely sinks for agroforest P include accumulation in above and below-ground biomass

and loss through three years of crop harvests, which, combined, could account for

approximately 35 to 40 kg ha"' (Chapter 6). Peach palm biomass, in particular, was found

to store over twice as much P in above-ground biomass, than in cupuassu and Brazil nut

combined in an eight-year-old agroforest (Chapter 6). The lack of difference between forest

and agroforest total soil P stocks implies that the latter has already depleted any post-burn P

additions. As a result, the readily-extractable Pi fraction in agroforests will likely continue

to decrease with time, eventually precipitating a decline in productivity under current

management practices, unless Pi is made available through other mechanisms, such as root

exudates, or buffered in the soil system by other less readily-available fractions not measured

in this study.

Fine Root Response to Phosphate Microsite Enrichment

While soil extracts provide evidence that labile Pi has decreased since agroforest

establishment, the question remains whether agroforest productivity is currently limited by the

size of this pool. The greater proliferation of cupuassu and "other" plants roots in P-treated

cores buried in agroforest alleys compared to the control cores suggests that these species

may invest more resources into roots that find P-enriched microsites. However, definite P

limitations to plant productivity could not be inferred using the root ingrowth core bioassay

because roots of the same species did not respond to P microsite enrichment in agroforest

rows. There was a similar trend towards greater mean root length in P-treated cores buried

in native forest, but this effect was not statistically significant. Cuevas and Medina (1988)

used proliferation of fine root biomass to infer both a P and Ca limitation to native terrafirnme

forest in the Venezuelan Amazon, and an analysis of native forest root mass in this study

revealed a significantly higher root mass in P-treated cores. As earlier noted, extractable Pi

concentrations measured in any of the extracts used in this study are considered "low" for

both forest and agroforest soils. However, a P-limitation to native forest productivity cannot

be inferred based on the root ingrowth bioassay, despite the fact that many studies cite P as

the nutrient most limiting to tropical forest productivity (Vitousek and Sanford 1996).

Length Versus Mass as an Indicator of Proliferation

Overall, root mass in P-treated cores did not differ from that of the control in

agroforest alleys, and this is perhaps due to differing patterns of root biomass allocation

among the system's components. Cupuassu root length appeared to be a more sensitive

measure of root proliferation than mass, and a plot of root length versus mass reveals a

significant linear relationship in which a small increment in mass produced a large increase in

length. Similarly, the majority of the various unidentified herbaceous species that comprised

the "other" group had very fine roots (diameter < 0.5 mm), which produced over twice as

much length than peach palm roots (in P-treated cores) with less than a third of the mass. The

difference in "other" root mass between to P-treated and control cores was not significant,

perhaps because the greater mass of coarser (diameter 2 mm) roots (of different species)

that occasionally grew into cores of both treatments overwhelmed the fine root mass,

obscuring any proliferation response based upon this variable. Ultimately, we placed greater

emphasis on root length in our evaluation of root proliferation response because the surface

area of contact between root and soil is more indicative of a root system's capacity to take

up nutrients than is mass (Newman 1966).

Root Proliferation in Agroforest Alleys and Rows

Studies have shown that root proliferation in patches where nutrients are more

abundant is a foraging strategy that may be more effective for fine- rooted species growing

in environments where nutrients are heterogeneously distributed or "patchy" (Caldwell 1994).

The proliferation response to P-enrichment exhibited by cupuassu in agroforest alleys may

reflect the heterogeneity of nutrient availability and the density of root distribution in this

location. With extractable Pi concentrations so low in agroforest soils, root growth would

likely be concentrated in areas of accumulating organic matter to take advantage of Po

mineralization. Decomposition and mineralization of more abundant and homogeneously

distributed litterfall in agroforest rows probably contributed to a more constant supply of P

to cupuassu roots growing in this location. Overall, cupuassu root length was greater in cores

placed in rows than in alleys, regardless of treatment. This, and the fact that the row location

often fell beneath the cupuassu canopy, suggest that roots were more densely distributed in

agroforest rows. Higher root density would increase the likelihood that roots would

"randomly" grow into both P-enriched and control cores, perhaps explaining why cupuassu

root lengths did not differ significantly between the P-treated and control cores in rows.

In contrast, litterfall in alleys was patchy, often exposing bare mineral soil. In such

an environment, the cost of root growth is relatively high if nutrient acquisition is not

increased as a result of the investment. Thus, we see very little cupuassu root length or mass

in the control cores buried in agroforest alleys compared to rows. Cupuassu roots growing

into alleys proliferated only when they encountered the P-enriched cores. Greater light and

water availability, and perhaps lower root densities in agroforest alleys may have also

contributed to a more favorable environment for the root growth of "other" weedy species

which are often better adapted to efficient exploitation of patchy nutrient availability.

Root Competition Among Agroforest Components

Differences in root response to P-microsite enrichment among agroforest components

may also reflect differing ecological strategies for nutrient acquisition that determine

competitive interactions among agroforest components. While the roots of cupuassu and

"other" plants proliferated in P-enriched cores in agroforest alleys, peach palm roots did not

exhibit similar foraging behavior. In fact, overall, neither palm root length, nor mass, differed

among core treatments or locations. This suggests that this species may acquire P in P-

deficient habitat through mechanisms other than proliferation of fine roots.

Fitter (1994) suggested that coarse-rooted species are less adapted to proliferate in

nutrient-enriched microsites because the investment necessary for the growth of longer-lived

high diameter roots may not be offset by the resources gained in short-lived nutrient-rich

patches. Potential differences among plant species in their capacity for root proliferation

seriously weakens the root ingrowth bioassay as a means to detect nutrient deficiencies in

crops, because a lack of proliferation might be misinterpreted to mean that the species is not

nutrient-limited, when in fact, a lack of response may be due to species-specific differences

in carbon allocation. The specific root length of peach palm is approximately half that of

cupuassu (Haag 1997), so that the palm allocates twice as much biomass to an equivalent

length of root as cupuassu. Furthermore, such differences in carbon allocation may indicate

different ecological strategies for resource capture. For example, it may be that instead of

proliferating short-lived fine rootlets in ephemeral nutrient patches, the palm may invest in

longer-lived higher diameter roots to locate and exploit soil resources that may be spatially

beyond the reach of competitors. Ferreira et al. (1995) estimated that when growing in

heavy-textured clay Oxisols, absorptive roots of peach palm may extend up to 9 meters from

the stem. In addition, the palm roots form thick superficial mats at the base of stems and

offshoots that "catch" fallen litter. In another study (Chapter 6) resin-Pi concentrations in the

organic matter trapped in the peach palm's root mat were 10 to 100 times greater than Pi

concentrations in the top 5 cm of surrounding soil. Finally, peach palm roots are known to

form vesicular-arbuscular mycorrhizal symbioses in Amazonian Oxisols, and other studies

suggest that this species may be able to solubilize less readily-extractable forms ofP (Clement

and Habte 1994, Femrnandes and Sanford 1995). Despite these potential mechanisms for P

acquisition, it cannot be concluded that peach palm is P-sufficient based upon a lack of root

proliferation response to P microsite enrichment, due the inherent weaknesses of the root

ingrowth bioassay mentioned above.

Mora-Urpi et al. (1997) note that P deficiencies are rarely observed in peach palm

growing in tropical Ultisols and Oxisols, and the various strategies for P acquisition described

above likely increase the palm's competitive ability in P-deficient soils. The RECA farmers'

concerns of root competition by the peach palm were based upon observations that palm

roots regularly grew in the soil beneath cupuassu canopies. In an eight-year old agroforest

of the same configuration planted by RECA farmers peach palm comprised 72.3% of total

above ground biomass, providing further evidence of the palm's dominance in this

agroecosystem (Chapter 6). The results of this study demonstrate that readily-extractable Pi

concentrations in agroforest soil decreased relative to adjacent forest, perhaps to the

detriment of other less-competitive species in the system. In another study, resin-extractable

P measured monthly over one year was higher underneath peach palm canopies than those of

cupuassu, presumably due to faster leaching, decomposition and mineralization of the

relatively P-rich peach palm leaf litter (Chapter 5), and the palm appears more efficient in

procuring and storing P in rapidly growing above- and below-ground biomass than the other

two agroforest components (Chapter 6). Thus, if competition is defined as the reduction in

plant fitness resulting from resource exploitation by neighboring plants (Grime 1977), it might

be concluded that peach palm competition threatens productivity in cupuassu and Brazil nut

under current no-input management practices. However, this study also demonstrates that

cupuassu roots do proliferate when they encounter P-enriched patches in close proximity to

its canopy, resulting in nearly a 40% increase in root tissue P content. While peach palm root

length and mass was greater than that of cupuassu in alley ingrowth cores, root length did not

differ between the two agroforest components in cores buried in rows, near the dripline of the

cupuassu canopy. Thus, despite the presence of neighboring peach palm roots, cupuassu P

nutrition might benefit from directed application of organic residues and/or fertilizer beneath

and around the canopy dripline, although further on-farm study is required before this can be


Conclusions: Implications for Amazonian Agroforest Sustainabilitv

The results of this study demonstrate that six to eight years following clearing, labile

inorganic phosphorus decreases when native forest is converted to agroforest in western

Amaz.nia, which could result in early P-limitations to agroforest productivity in P removal

is not offset with additions. Clearly, the role of less labile Pi and Po pools in maintaining P

availability in perennial cropping systems deserves further research. However, the decrease

in agroforest labile Pi six years after forest conversion represents not only an irreplaceable loss

from the system under current management practices, but more importantly, a significant

difference in P cycling between tree-based agroecosystems and native forest. Like secondary

forest ecosystems, storage in live tissue represents a significant P sink during early stages of

vegetational succession, and presumably, much of the agroforest's future standing biomass

production requirements would be met by nutrient fluxes in the intrasystem cycle as the tree-

based agroecosystem reaches steady state (Attiwill and Leeper 1987). For example,

Polglase et al. (1992) found that M- 1 Pi in Eucalyptus reglans forests decreased from 34 mg

kg' at time zero to 2.3 mg kg' at age 16, and remained constant thereafter in stands aged 40

to 80 years old.

In contrast, P removal with the harvest of agroforest products, estimated during the

sixth year following establishment to be between 3.2 and 4.0 kg P ha"1 yr"1, represents a

permanent loss from the total soil P stock. While Po mineralization may sustain production

requirements of native forest at steady-state, the decrease in agroforest labile Pi indicates that

Po pools cannot adequately restore solution Pi as it is taken up by an aggrading

agroecosystem undergoing P removal with successive crop harvests. Unless replenished

through external inputs, this drain on soil P will only increase as the system matures and

harvest ofagroforest products continues. Arguably, the 25 to 50% reduction (depending on

the extract) in readily-extractable Pi six years after forest conversion, represents less than one

percent of the agroforest's total soil P stock (410 mg kg'), and the large difference between

total P and extractable Pi may include pools that are plant-available over the long term.

However, the rate at which P is supplied to agroforest plants determines the system's

productivity on a short term basis, and hence, its potential for economic sustainability. The

decrease in agroforest readily-extractable Pi relative to that in native forest soils demonstrates

that it is being taken up more rapidly than it can be restored by other P pools in the soil

system, and this decrease in "labile" P may affect components of the system differentially is

one species, for example, is has "access" to less readily-soluble P forms while another is not.

Therefore, it cannot be assumed that the processes sustaining mature native forest ecosystems

will maintain productivity in all agroforest components without management intervention.

While it is unlikely that production in the agroforest will cease entirely in the short

term, continually low or reduced productivity may exclude commercial agroforestry from

consideration as an economically viable alternative to other more destructive land uses in

Amaz6nia. In the absence of perceived economic sustainability, farmers will clear more forest

to establish new agricultural systems, because forest land is not a scarce resource in this

region. In all five focus group discussions held with RECA farmers, producers admitted that

they continued to clear forest every year following agroforest establishment to plant more

perennial crops. Their objective was to maintain household income when productivity of the

first agroforestry systems planted ultimately fell. Coupled with their fear of aggressive

competition by the peach palm, these farmers did not believe that the initially "high"

productivity ofagroforests could be maintained, and thus, they chose to minimize economic

risk by planting new systems every year. Although the producer's concerns are

understandable, perceived and managed in this way, agroforests do not offer a means of

decreasing deforestation rates on Amazonian small farms.

Undoubtedly, the sustained production in Amazonian agroforests will require practices

that both offset nutrient export with crop harvest using soil amendments, as well as enhance

organic matter cycling to maintain soil solution P concentrations and protect the system from

further nutrient losses. Aside from cost, a major constraint to the use of most organic and

inorganic amendments in these soils could be that phosphate ions are adsorbed almost as fast

as they are released into solution, either through dissolution or mineralization. Hands et al.

(1995) recommend that inorganic P be added to the mulch layer of alley cropping systems to

avoid fixation by the mineral soil. In this case, directed fertilizer application in fallen litter

beneath the cupuassu canopy, where root growth was shown to be comparable to that of

peach palm, would add P, stimulate organic matter decomposition and mineralization, and

perhaps encourage greater fine root growth towards sources of mineralizing P on the

agroforest floor. Obviously this and other potential management practices need to be tested

before they can be recommended, and further participatory on-farm research is necessary to

design management strategies that are both economically feasible and practical so that

commercial agroforestry systems do indeed offer a sustainable alternative to more destructive

land uses driving Amazonian deforestation.



Maintaining phosphorus (P) availability to crop plants growing in highly weathered

soils is one of the largest challenges facing the development of sustainable agroecosystems

throughout much of the humid tropics (Sanchez 1976). Previous studies have shown that

less than 1% of total P in Oxisols and Ultisols of South America's Amazon Basin is

extractable using procedures for the most common indices of P availability (Tiessen et al.

1993, Dias-Filho et al. in press, Chapter 4), and it is estimated that P deficiencies limit

crop production in 90% of the region's upland soils (Nicholaides et al. 1985, Smyth and

Cravo 1990). Much of the soil P stock is geochemically bound to iron and aluminum

oxides in forms that are largely unavailable for uptake, rendering plant P nutrition highly

dependent upon biologically-mediated transformations of organic P (Cross and

Schlesinger 1995, Hedley et al. 1995). Thus, in non-fertilized agroecosystems,

fluctuations in soil P availability over a growing season are often associated with factors

controlling litter decomposition and Pi mineralization from soil organic matter, such as

temperature, moisture and resource quality (i.e. the biodegradability of organic material),

as well as with seasonal variations in P demand by plants and competing microbial

populations (Tate 1984, Stewart and Tiessen 1987, Lajtha and Harrison 1995). In tree-

based ecosystems, such as perennial crop-based agroforests, Pi mineralized from

decomposing litterfall and dead roots contributes to the long-term productivity of these

systems, although the highest rate at which Pi is released from various organic sources

may not necessarily coincide with periods of greatest demand by the system's crop

components. It is the rate of Pi release by mineralization, rather than the amount of

organic P (Po) present, that frequently controls Po availability to plants (Tate 1984).

Consequently, the most efficient use of soil amendments, including inorganic fertilizers,

green manures and organic residues, often requires synchronized and directed application

during periods of high demand by crop plants and low soil availability (Young 1989,

Fernandes et al. 1997). For this reason, monitoring spatial and temporal fluctuations in

soil P availability in relation to the production cycle of an agroecosystem is important to

developing management practices that sustain productivity in low/no input systems.

Assessing short term changes in soil P availability is, however, difficult using the most

common soil extracts because they often solubilize portions of solid-phase P not available

to plants (Bolan 1991). Moreover, soil extractions carried out in laboratories do not

necessarily reflect the ambient conditions that control P mineralization, or the

biogeochemical processes that cause short term fluctuations in soil solution Pi, such as

microbial immobilization, adsorption to soil solids, and plant uptake (Lajtha 1988).

In general, resin extracts more closely simulate the physical action of plant roots

because exchangeable ions, such as H2PO04, are desorbed from soil solids without drastic

changes in soil chemistry (McKean and Warren 1996). Conceptually, the resin acts as a

sink for phosphate ions desorbing from soil solids, continually removing them from

solution so that an equilibrium between the solid and solution phases is not established

(Vaidyanathan and Talibudeen 1970). In a laboratory study, Parfitt and Tate (1994) used

resin-impregnated membranes to measure P mineralization by extracting the soil to

exhaustion before and after an incubation period. Under field conditions, especially in

soils with high sorption capacities, resins behave more like dynamic exchangers, so that P

measured in resin extracts represents a composite index of the soil's retention capacity,

microbial P demand, and the status of plant available P (Cooperband and Logan 1994).

As a composite index, resin-filled bags or impregnated membranes are useful for making

in situ comparisons of temporal and spatial variations in P availability within or among

systems (Huang and Schoenau 1996, Femrnandes and Coutinho 1997). Fluctuations in P

availability under field conditions have been monitored in a number of different ecosystems

using resin bags placed in or on top of the soil for varying lengths of time (Gibson 1986,

Lajtha 1988, Giblin et al. 1994, Yavitt and Wright 1996). Krause and Ramlal (1987) used

resin bags to show that P availability over a four month period remained 2.5 times greater

in soil under a clear-cut area than in adjacent forest, presumably due to increased

decomposition stimulated by higher temperatures and forest floor mixing resulting from

the timber harvest activities. The relatively new use of resin-impregnated membranes as

an index of P availability in field conditions is especially attractive because the two-

dimensional rigid structures can be placed in soil or litter to achieve maximum surface area

contact with minimal disturbance (Cooperband and Logan 1994, Huang and Schoenau

1996), and unlike resin-filled bags, resin membranes do not trap fine roots and soil

particles that interfere with analyses (Fernandes and Warren 1996). Numerous studies