Comparison of on-farm research and extension methods in small scale farm systems in lower Casamance, Senegal

Material Information

Comparison of on-farm research and extension methods in small scale farm systems in lower Casamance, Senegal
Lo, Mamadou, 1952-
Publication Date:
Physical Description:
xvi, 116 leaves : ill. ; 29 cm.


Subjects / Keywords:
Agriculture -- Research -- Senegal ( lcsh )
Agricultural extension work -- Research -- Senegal ( lcsh )
Farms, Small -- Senegal ( lcsh )
Agricultural Education and Communication thesis, M.S ( lcsh )
Dissertations, Academic -- Agricultural Education and Communication -- UF ( lcsh )
bibliography ( marcgt )
non-fiction ( marcgt )


Thesis (M.S.)--University of Florida, 1997.
Includes bibliographical references (leaves 111-114).
General Note:
General Note:
Electronic resources created as part of a prototype UF Institutional Repository and Faculty Papers project by the University of Florida.
Statement of Responsibility:
by Mamadou Lo.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact Digital Services ( with any additional information they can provide.
Resource Identifier:
002267621 ( ALEPH )
37230442 ( OCLC )
ALM0621 ( NOTIS )

Full Text

Copyright 1997
Mamadou Lo

To my parents and my wife for their prayers and support.

First, I would like to thank my scientific advisor Dr. Peter E. Hildebrand for his support and advice during all my studies at the University of Florida, for his help and advice defining the research proposal preparation and execution of this research.
I am thankful to my academic advisor Dr. Mathew T. Baker for his constructive instructions and advice.
I am also thankful to Dr. Tracy Hoover for her critical instructions and advice.
All three were excellent support as members of my
supervisory committee. I am also grateful to my Institution (ISPA), to USAID, to Thomas Jefferson Fellowship and to the University of Florida (UF) for their major support for my studies. In Senegal, I would like to thank farmers of Lower Casamance, my colleagues of Djibelor and ISRA for their support.
My thanks go to my parents and my wife for their
prayers and blessing. It is wonderful to have their magic wisdom, love and courage helping me in fighting for my dreams and ideals.

Finally, I would like also to thank all my friends and colleagues at the University of Florida, for their friendship, support, encouragement, and advice and for listening to me and making my life easy in Gainesville, especially Michael Dougherty, Jonathan Moscatello, Abib Araujo, Elena Bastidas, and Christopher Johnson. I am also grateful to all UF's professors and technicians who, near and far have contributed to the success of my studies.

ACKNOWLEDGMENTS ........................................ iv
LIST OF TABLES ......................................... ix
LIST OF FIGURES ........................................ xi
LIST OF ACRONYMS ....................................... xiii
ABSTRACT ............................................... xv
1 INTRODUCTION ................................... 1
1.1 History of Agricultural Research and
Extension .................................. 2
1.1.1 Early Model ............................ 2
1.1.2 Farming Systems Research and Extension. 5 1.1.3 Farm Management and Linear Programming. 19 1.1.4 Extension ..... ***** ... 20
1.2 History of Research Systems in Senegal ..... 23 1.3 Researchable Problem ....................... 26
1.4 Hypothesis ................................. 27
1.5 Objectives ................................. 27
1.5.1 Primary Objectives ..................... 27
1.5.2 Secondary Objectives ................... 28
1.6 Limitation of the Study .................... 28
1.7 Definition of Terms ........................ 29
AGRICULTURE .................................... 32
2.1 Overview of the Agricultural Sector ........ 32 2.2 Characteristics of the Studied Area ........ 36 2.3 Description of the Representative Household 42
3 METHOD AND MATERIALS ........................... 43

3.1 Population and Sample ..................... 43
3.2 Data Collection and Analysis .............. 44
3.3 Methods of Analysis ....................... 44
4.1 Characteristics of the Village of
Loudia Ouloff ........................... 47
4.2 Characteristics of the Studied Household.. 48
4.2.1 Schematic Modeling of the Household .... 48 4.2.2 Gender Analysis of the Household ....... 48
4.3 Linear Programming Analysis of the
Household ................................. 51
4.3.1 Resources and Minimum Survival
Constraints ..... 51
4.3.2 Maximizing the Family Income ........... 52
4.3.3 Influence of Gender in the Household ... 52 4.3.4 Maximizing Female Income ............... 53
4.3.5 Maximizing Male Income ................. 53
4.3.6 Proposition of Basket Making as one
Alternative Solution for the Household. 54
4.4 Comparison of Results Analysis of Linear
Programming with Farmap/Brads Results ..... 55
5 ADAPTABILITY ANALYSIS .......................... 58
5.1 Calculating the Environmental Index ....... 60
5.2 Relating Treatment Response to Environment
on EI ..................................... 62
5.3 Relationship Between EI and Environmental
Characteristics ........................... 66
5.4 Definition of Tentative Recommendation
Domains ................................... 67
5.5 Determining Risk Associated with the New
Technologies .............................. 68
5.6 Definition of Final Recommendation
Domains ................................... 71
5.7 Comparing Results Using Farmers' Evaluation
Criteria .................................. 71
5.8 Definition of Tentative Recommendation
Domains Using Farmers' Evaluation
Criteria .................................. 77
5.9 Determining Risk Associated with the
Different Varieties Using Farmers'
Criteria .................................. 77
5.10 Define Final Recommendation Domains
Using Farmers' Criteria ................... 81
5.11 Extension Recommendations for Each of the
Recommendation Domains .................... 81

6.1 Conclusions.................................... 83
6.2 Recommendations................................ 86
OULOFF............................................. 89
LIST OF REFERENCES......................................... 111
BIOGRAPHICAL SKETCH........................................ 115

2.1 Desciption of the representative household ........ 42
4.1 Farm budgets and income in the surveyed area for a Rainy Year (1984), comparing Farmap/Brads
and Linear Programming .............................. 57
5.1 Anova for yields of nappe rice variety trials (kg/ha) .............................................. 60
5.2 Response of four improved nappe rice varieties and the Farmers' Local Variety (kg/ha) from On
farm research results (1982-1985). Lower
Casamance, Senegal .................................. 61
5.3 Nappe rice variety trials regression estimates .... 63
5.4 Environmental characteristics data, sorted by EI, for nappe rice variety trials (1982-1985). Lower
Casamance, Senegal .................................. 67
5.5 Tentative recommendation domains and the technologies recommended for nappe rice variety
trials based on environmental characterisrics
(zones) and evaluation criterion kg/ha ............ 68
5.6 Calculation of lower confidence limits (kg/ha) for two yielding domains based on zones for five
nappe rice variety trials (1982-1985). Lower
Casamance, Senegal .................................. 68
5.7 Response of four improved nappe rice varieties (Height of Plants) and Farmers' Local variety
for On-farm nappe rice variety trials (1982-1985).
Lower Casamance, Senegal ............................ 73
5.8 Response of five improved nappe rice varieties (Plant Cycle) for On-farm nappe rice variety
trials (1982-1985). Lower Casamance, Senegal ...... 74

5.9 Response of four improved nappe rice varieties
(Tillers/m2) and Farmers' Local Variety
for On-farm nappe rice variety trials (1982-1985).
Lower Casamance, Senegal ............................ 75
5.10 Summary of the recommendation domains and the
technologies recommended for nappe rice variety
trials (1982-1985). Lower Casamance, Senegal ...... 81
5.11 Multiple recommendations and the recommended
technologies for nappe rice variety trials (19821985), based on environmental characteristics
and four evaluations criteria. Lower Casamance,
Senegal .............................................. 82
A.1 Schematic modeling of the household ............... 89
A.2 Activities analysis of the household .............. 90
A.3 Resource analysis of the household ................ 91
A.4 Benefits and incentives analysis ................... 92
A.5 Seasonal calendar of the household ................ 93
A.6 Basic matrix of the household ....................... 94
A.7 Maximizing family income ............................ 96
A.8 Gender analysis of the household .................. 99
A.9 Maximizing female income ........................... 102
A.10 Maximizing male income ............................. 105
A.11 Basket making as one alternative solution for the
household .......................................... 108

2.1 Administrative map of Senegal ..................... 33
2.2 Map of Senegal and farming systems zones in
Lower Casamance .................................... 38
5.1 Linear response in kg/ha of variety DJ12519 to environmental index (EI)............................ 64
5.2 Quadratic response in kg/ha of variety IRAT1l2 to environmental index (EI)......................... 64
5.3 Linear response in kg/ha of variety IRAT133 to environmental index (EI)............................ 64
5.4 Linear response in kg/ha of variety IKP to environmental index (EI)............................ 65
5.5 Quadratic response in kg/ha of variety Local to environmental index (EI)............................ 65
5.6 Estimated responses in kg/ha of five nappe rice varieties to environment index (EI) for nappe
rice variety trials (1982-1985). Lower Casamance,
Senegal .................. ........................ 65
5.7 Risk levels for five nappe rice varieties for zones IV and Station, (Evaluation criterion
kg/ha) .............................................. 70
5.8 Risk levels for five nappe rice varieties for zones III and V, (Evaluation criterion kg/ha) .... 71
5.9 Estimated responses (height of plants in cm) of five nappe rice varieties to environmental index
(EI) ................................................ 76
5.10 Estimated responses (plant cycle in days) of five nappe rice varieties to environmental index xi

(EI) ............................................. 76
5.11 Estimated responses (tillers/m2) of five nappe
rice varieties to environmental index (EI) ....... 77
5.12 Risk levels for five nappe rice varieties for
zone IV and Station, (Evaluation criterion
height of plants) .................................. 79
5.13 Risk levels for five nappe rice varieties for
zones III and V, (Evaluation criterion height of
plants) .......................................... 79
5.14 Risk levels for five nappe rice varieties for
zone IV and Station, (Evaluation criterion plant
cycle) ........................................... 79
5.15 Risk levels for five nappe rice varieties for
zones III and V, (Evaluation criterion plant
cycle) .............................................. 80
5.16 Risk levels for five nappe rice varieties for
zone IV and Station, (Evaluation criterion
(tillers/m2) ........................................ 80
5.17 Risk levels for five nappe rice varieties for
zones III and V, (Evaluation criterion tillers/
n2) ................................................. 80

AA: Adaptability Analysis
ARP: Agricultural Research Project CNRA: Centre National de Recherches Agricoles
National Center for Agricultural Research
CFA: Communaut6 Financibre Africaine francs EI: Environmental Index
FSRE: Farming Systems Research and Extension FSSP: Farming Systems Support Project GNP: Gross National Product GDP: Gross Domestic Product IRAT: Institut des Recherches Agronomiques Tropicales et
des Cultures Vivridres.
Institute for Tropical Agronomic Food and Crop
IEMVT: Institut d'Elevage et de Medecine V6t6rinaires des
pays Tropicaux.
Institute for Tropical Livestock and Veterinary
IRHO: Institut de Recherches sur les Huiles et Oleagineux.
Research Institute for Oil Plants
IRCT: Institut de Recherche sur le Coton.
Research Institute for Cotton and Textile Fibers
ISRA: Institut Sen6galais de Recherches Agricoles.
Senegalese Institute for Agricultural Research
LP: Linear Program

LNERV: Laboratoire National d' Elevage et de Recherches
National Laboratory for Livestock and Veterinary
MDR: Minist~re du Developpement Rural.
Ministery of Rural Development
MSU: Michigan State University. NGO: Non-Governmental Organization. ORSTOM: Office de la Recherche Scientifique et Technique d'
Overseas Scientific and Technical Research Office.
OMVS: Organisation pour la Mise en Valeur du Fleuve
Organization for Development of River Senegal.
PAPEM: Point d'Appui et d' Experimentation Multilocale.
Off-station Sites for Multilocational
PSR: Production Systems Research. RD: Research and Development.
SARPP: Senegalese Agricultural Research and Planning
USAID: United States Agency for International Development. UE: Unites Experimentales.
Experimental Units.

Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science
Mamadou LO
May, 1997
Chairperson: Dr. Mathew T. Baker Major Department: Agricultural Education and Communication
Rainfall deficits for more than fifteen years have
turned the Lower Casamance, southern region of Senegal, from a region of food self-sufficiency to one of food deficits. Farmers were not widely adopting research results, despite the contacts and opportunities provided through extension services and research. Agro-socioeconomic surveys of the representative small-scale farm and on-farm research trials, conducted in Lower Casamance, in 1984 and from 1982 to 1985, respectively, by the FSR of Djibelor were analyzed based upon the primary data that the researcher collected in Senegal, in Summer 1996. Research objectives were to (1) compare the performance of new methods of analysis designed for on-farm research and extension with existing methods in

the identification of agro-socioeconomic constraints and analysis of improved technologies of livelihood systems; (2) provide methods of analysis adapted to on-farm research and extension; (3) predict response of livelihood systems to improved technologies and give valuable feedback to researchers, policy makers and technology change agents; and
(4) propose adapted educational programs to facilitate the transfer and diffusion of improved technologies in different environments. Comparison of the results from the linear programming simulation with the Farmap/Brads data showed that the LP model resembles the data very well; so it will help to predict how proposed technologies fit farmers' socioeconomic environments. Comparison of the results from Adaptability Analysis (AA) with Analysis of Variance (ANOVA) showed that Anova, because of its complexity and sensitivity to violations of a number of statistical assumptions, is not easily adapted to data produced under farmers' conditions, which are characterized by a high degree of variability and heterogeneity. The results of this research suggest that LP and AA are more effective and efficient methods than the previous ones (FARMAP/BRADS and ANOVA) in the effort to predict responses of farmers to improved technologies on one hand and, on the other hand, to identify improved technologies and produce extension messages relevant to a range of farmer circumstances.

The West African country of Senegal is primarily an
agricultural country. In 1980, agriculture accounted for 28% of the GNP and provided employment for 80% of the economically active population.
In recent years, cereal production in Senegal has not been sufficient to meet consumption needs and most of the production was generated by the small holders which represent 60 to 70% of the farming population. The general crisis in Senegalese agriculture during the last two decades and in Lower Casamance in particular derives from an insufficient adaptation of traditional farming systems to new agroclimatic perturbations (decreased rainfall, shortening of seasons, degradation of the edaphic milieu) and evolution of the socioeconomic environment (population growth, government policies, trade conditions, development of social dynamics) This unfavorable agro-socioeconomic context negatively influenced the production strategies of small-scale farms and family livelihood systems.
The overriding concerns of these small-scale farms is to sustain the home and the family rather than to produce

for growth and profit. The important role they play in the economic development of the country has received ample attention from the Senegalese Research and Extension Institution. It has been recognized and accepted by the government and other development agencies that agricultural and social changes in the production systems of these farms, based on the utilization of new or modified technologies that meet their needs and aspirations, is an important strategy and a challenge for the agricultural, economic and social development of the country.
1.1 History of Aaricultural Research and Extension
1.1.1 Early Models
The organizational framework for agricultural research and development which has evolved over the past century, into the 1970s, has worked reasonably well for the industrialized nations. People working in agricultural development in increasing numbers, however, are coming to believe that this approach is not working in the developing countries. This conclusion has prompted a search for new agricultural R&D models specially designed to improve the productivity and well-being of the rural majority who have so far been by-passed. The plant breeding breakthroughs of the "Green Revolution" of the 1960s, which produced new high-yielding grain varieties, provided enormous advances in food production in a number of countries. Yet as many

critics have noted, the new technologies have tended to favor those rural producers already in relatively advantageous positions, doing much less to improve the lot of the rural majority. In some cases, these new technologies have negative effects by spurring labor displacement or land concentration. This uneven impact of new technologies has not been a consequence simply of different sizes of landholdings, yet on the average, larger farmers have benefited more than smaller farmers. Without the advance of the so-called "Green Revolution", worldwide production of cereal grains would be far less than it is today, and the brunt of shortfalls would surely fall on the poorest sectors of society.
Existing agricultural R&D strategies have not given
much direct support to those farmers who struggle to survive under poor climates, soil, and water conditions which are much less favorable than the conditions assumed for adoption of "Green Revolution" technologies. Moreover, the rural majority labor and produce within systems of agricultural production far more complex than the primitive conventional stereotype (Harwood, 1979; Scrimshaw and Taylor, 1980). Research and extension must go beyond dealing with one crop at a time, and consider the pattern of the farming system as a whole and relating that farming system to the total economic and social environment of the rural family.

Agricultural research and development models have
mostly been created in industrialized nations and then have been introduced into developing countries. Since the implications of these model are profound, it is instructive to consider two general types of models that have been transferred. The first type, the European colonial model, was already introduced before World War II in the African and Asian colonies. The second type was developed after 1945 through U.S. technical assistance in Latin America and some Middle Eastern and Asian nations.
The European model was based primarily upon large-scale plantations devoted to production of export crops. Until shortly before the end of the colonial period, when agricultural research began experimentation on locally consumed subsistence crops, the European model did not lend itself to effective work with small farmers. Thus, the Europeans and their African and Asian research counterparts were in need of a new agricultural research and development model. The structure of the European model, in its initial conception and supporting philosophy, was distinctly vertical. Research was carried out in the laboratories and sent "down". Little feedback went "upwards" to the scientists.
The U.S. model of agricultural research and extension,

with the passing of the colonial era, gained in popularity and influence. Indeed, in the late 1940s, many U.S. experts assumed that the transfer of their model to developing nations could result in the increase in productivity and farmer income that had occurred in the United States. The Point IV program, designed to bring technological and financial assistance to agriculture in developing nations, carried with it the model of American "land grant" universities linked to an extension service taking the results of university-based research "out" to farmers. The model was intended to be horizontal. Program planners focused primarily on one part of the model, agricultural extension. Until the failure of this transplant became evident, programmers did not undertake to build into developing countries the other components, particularly the university and experiment-station-based research programs, which were vital features of the US model.
1.1.2 Farming Systems Research and Extension
Still, by the early 1970s it was becoming obvious that even technology tailored for tropical climates was not trickling down to the small, limited resource farmers who had less than the best physical resources and little or no access to infrastructure such as markets and irrigation. The new approach called "Farming Systems Research and Extension" (FSRE), with heavy emphasis on the limited resource farmers,

was kindled. FSRE was based on a bottom-up approach rather than top down, (Norman, 1980) Heavy emphasis was on participation by the small-scale, limited resource farmers in diagnosis and evaluation of potential new technologies on their own farms. For a quarter century following World War II, the conventional technology generation and diffusion process was patterned on a progressive farmer strategy. This strategy (Roling, 1988), in turn, was based on several assumptions. First was the innovation bias (Rogers, 1983), under which it was assumed that any innovation resulting from the established research-extension process was good and therefore, should be adopted. Second, it was assumed that this kind of technology was broadly adaptable and scale neutral, meaning, anyone who was willing could adopt. Third, diffusion research had shown that innovations spread within a social system from one decision making unit to the next over time (Roling, 1988), so any introduced innovation should spread through out the community. Fourth, it was also assumed that early and late adopters, as well as nonadopters were all from the same social system simply because they lived in the same community late adopters or nonadopters were thought to be "laggards"1, and not interested in "improvement" (Rogers, 1995) It was also noticed from feedback messages (farmers to extension to research, as well

as from farmers directly to research) that it was the "progressive farmers" who were adopting the technology first, if not exclusively. However, this was not a concern because it was assumed that the good technology would trickle down from these progressive farmers to those who were less progressive or more conservative or risk averse.
Indeed, extension used contacts with progressive
farmers as a prime strategy. As it became obvious that these progressive farmers were becoming wealthier and larger relative to the other farmers in the community, this was of little concern. The emerging change in the nature of farms was supported by the concept that "bigger is better." Small farmers often were considered more of a social problem than an agricultural problem. It became obvious early in the 1970s that the "bigger is better" approach was disastrous for developing countries where the large majority of farmers were not being served by the progressive farmer, trickle down strategy. In the less advantaged countries, as opposed to the industrialized countries, little capacity was available to employ persons forced from agriculture into urban areas. The gains made by small farms in the Third World with Green Revolution technology were made only by those with the best resource base, a limited minority. The technology did not trickle down from them to other farmers

who did not have the advantages of a better resource base. Research on technology diffusion was able to show ex post what characteristics were related to slow or non-adopters (smaller farms, poorer, less education), but was unable to provide ex ante suggestions for effective intervention strategies (Roling, 1988). The progressive farmer strategy coupled with the trickle down concept fails when the farm population is not homogenous, but heterogeneous. "Laggards" and "innovators," originally considered to be members of the same social system simply because they lived in the same community, region or country, are very different farmers, with different production environments.
The strategy employed in agricultural extension also involved the term "transfer of technology" which is misleading because it implies that small farmers have such inadequate knowledge about agriculture that they must depend upon academic professionals to provide them with the information and ideas to improve production. In many developing countries there were many common deficiencies in knowledge and ability on the part of the extension agents (Chambers, 1974). The scope of the agent's responsibilities and the number of available agents also limited more intensive relations between change agents and farmers.
Developing countries lack the money and the

professional talent to provide a more intensive technical assistance relationship. They have a lack of integration among the various government agencies who have official responsibilities for serving small farmers. Researchers look down upon the extension agents, considering them incompetent and poorly trained. On the other hand, extension agents are often inclined to think that researchers are out of touch with the practical realities of farming and are simply pursuing esoteric projects designed to accumulate professional prestige.
Another major factor affecting equitable agricultural extension is the lack of agricultural credit and marketing. As various studies have shown (Blair, 1978), credit tends to go predominantly to the more affluent farmers. The cost of credit for small farms is also likely to be a major problem. In order to protect poor people from the exorbitant rates of moneylenders, many governments have established special credit programs for small farmers. But the cost, in terms of time to obtain the necessary certificates and the rates of interest, make the situation difficult for small farmers to obtain credit. To deal with marketing problems, some governments have established buying organizations that guarantee purchase of farmers produce at prices designed to provide a reasonable income. However, such government organizations are often so inefficient that they fail to

provide real help to small farmers.
Finally there was the gender bias that has been built into agricultural R&D organizations. In the past, all over the world, agricultural extension has been a job almost exclusively for men. As a result, extension services provided for women have traditionally centered only around the homemaking functions of women, such as cooking and sewing (Poats et al., 1988).
While all of these factors add up to a general
explanation of the ineffectiveness of the conventional agricultural extension system, one can gain a more systematic picture of the problem if it is placed in the context of the socioeconomic structure of developing countries. In developing countries there is a gap in social status, education, income and communication style between peasant farmers and extension agents. Farmers are hesitant to express their opinions and make demands upon the agent, and this tends to lead the agent to underestimate the intelligence and competence of peasant farmers. Without a means to modify the extension model to suit the resources and environments of "laggards", the conventional agricultural research and extension models will not provide improved technology for farmers whose resources and environments do not reflect the highly modified conditions found on research stations.

The 1970s produced a convergence of thinking, across a range of disciplines, toward integrated systems of rural development. Those who approach development problems from plant, animal and soil science perspectives have come to recognize the importance of farming systems research. Furthermore, natural scientists have increasingly come to recognize the importance of active participation by small farmers if a program of agricultural research and development is to be effective. During the same period, social scientists have abandoned the myth of the "passive peasant." Researchers have begun to recognize that small, resource poor farmers reject innovations offered by extension because they yield poor results in their particular situation. Typically, they simply do not have the resources that would allow them to follow the extension recommendations. This realization has led to an increasing interest in studies of what may be called the "social organization of agriculture"; a field in which social scientists must have a grasp of some of the key elements of plant, animal and soil sciences. This understanding needs to go far beyond that possessed by proponents of community development. Researchers and socioeconomic planners have come to recognize not only the need for small farmer participation in agricultural R&D programs, but also the importance of small farm organizations to give individual

farmers an effective voice in these programs. Both natural and social scientists have recognized the importance of an interdisciplinary approach to agricultural research.
In the early 1970s, the beginnings of FSRE research had begun in the developing countries of Asia, Africa and Latin America funded by several international institutions (CGIAR, Ford Foundation, Rockefeller Foundation, USAID, World Bank, etc.) These early FSRE efforts were prompted by the desire to improve farmer participation in the agricultural R&D process. The results of these early FSRE studies showed that any new R&D model should include the following characteristics (William, 1981):
1. Research must be carried out on the fields of small farmers as well as in the agricultural experiment stations. 2. Small farmers must actively participate in the research and extension activities carried out in their area helping to identify problems and set criteria as well as judge results.
3. The research program must include a major emphasis upon cropping and farming systems, field studies and experiments.
4. The research program should involve a strong emphasis upon interdisciplinary collaboration.
5. People with special responsibilities in extension and local economic development should not be isolated from the research process.

6. Government program designs should enhance the quality of human resources among small farmers and build material resources into the organizational base of the community and be maintained under the control of community.
There are four distinct stages in adaptive technology development (Hildebrand, 1988): 1-The Descriptive or Diagnostic Stage: The diagnostic phase in FSRE provides researchers with the information required for identifying farmer problems and constraints and determining appropriate solutions. There are a number of methods for collecting data (sondeos, Rapid Rural Appraisal, Participative Diagnostic, etc.) Secondary data are also very important to consider before the diagnosis. Data can be collected and summarized from unpublished as well as published sources. The main objective of secondary data is to have an overview of the entire community being studied (Hildebrand, 1979) Commonly used methods for collecting data in FSRE are informal surveys, formal surveys, observation, on-farm experiments, and case studies. However, the most widely used method is the formal and informal survey.
2-The Design or Planning Stage: In this stage, a number of alternative intervention strategies (solutions to problems) are identified which may be appropriate in dealing with the

constraints delineated in the descriptive or diagnostic stage. Then, the multidisciplinary team draws the community into the planning process first by discussing the diagnostic report and its initial recommendations. In the discussion with the community, suggestions are made regarding needed modifications to the team findings and the discussion provides the basis for planning the first field activities. 3-The Testing Stage: The purpose of FSP.E is to develop new agricultural technologies that address priority problems of farm households. The ultimate measure of the success of a new technology is the acceptance, adoption, and sustained use by farmers (Hildebrand, 1988) FSRE teams use on-farm experimentation involving active participation of farm household members to test and adapt alternative technologies. On-farm research is characterized by farmers' participation on their own land. This participation varies according to the nature of the experiments. During the testing stage, potential recommendations derived from the design stage are examined under actual farm conditions. This is done to evaluate the suitability and acceptability of the new practices in the existing farming systems. 4-The Recommendation and Dissemination Stage: Extension is very important in this stage because successfully tested technologies or practices developed by researchers and a

group of farmers are made available to other farmers with similar circumstances. Farm households are divided into different recommendation domains based on factors such as the farmers' knowledge base, their level of resources, etc., which combine to determine the particular farming system of the household. In this stage, *standard extension techniques are used to spread the news about new technologies to other farmers in similar recommendation domains (Wotowiec et al., 1988).
The grassroots origins of FSRE began to coalesce into a school of thought during the late 1970s and early 1980s. Innovative international agricultural scientists, working independently in dispersed locations around the globe, sensed the need for whole systems analysis for dealing with the small scale farm populations. Peter Hildebrand, working in Guatemala, Collinson in Tanzania, and Norman in Nigeria, were a few of these lead innovators. Their initial successes and alternative paradigm for dealing with this special population was recognized by international donors like USAID, the Ford Foundation, the Rockefeller Foundation, and others. As a result of the initial success by innovators, USAID sponsored a meeting in Fort Collins, Colorado, to create a forum for discussion and to begin the commercialization of FSRE concepts. From these meetings one

can see the first steps of the commercialization of FSRE by the production of conference proceedings. These papers were intended by their authors to serve as guidelines for application of the newly named Farming Systems concepts and methods (Shaner et al., 1982) In concert with the publication of these methods, USAID began to fund projects purporting to use the FSRE research approach. USAID, in this case the centralized authority, initiated diffusion by offering the incentive of funding for projects which applied the FSRE approach. However, it was soon realized that these projects did not integrate natural science research with social science research, and that the farmer was not being included in the study. Systematic findings were central to these early studies, but these examples did not capture the idea behind FSRE, which is the establishment of technologies for the needs of, as perceived by, and with the cooperation of the farmer.
As a result of the shortcomings of some early farming systems projects, donors realized that further commercialization was needed for the effective diffusion of the FSRE concept and methods. Donors, in cooperation with innovators, realized the need for people trained to conduct research using FSRE methods. To meet this need, USAID funded the Farming Systems Support Project (FSSP) at the University of Florida. This massive initiative involved 21 institutions

and over 600 professionals worldwide, with the goal of creating instructional materials and training practitioners in the methodologies and concepts used to conduct farming system research exercises (Hildebrand .P.E, 1996, personal communication). The FSSP constitutes an active commercialization and diffusion program by international donors to transform it into new methodologies which could then be diffused to researchers working on USAID (and other) funded projects. The experience gained in these projects showed the need for other methods and concepts. As a result, livestock system analysis, gender analysis, and later community development components were included (Hildebrand.P.E, 1996, personal communication). The FSSP helped to formulate a cluster of research methodologies to apply when conducting farming systems research, development, and extension.
By the late 1980s two very important things occurred simultaneously. A new buzzword had entered the discourse: sustainability. The concept of sustainable farm production in the face of global warming, rising population, and limited energy reserves, seemed logical and necessary. As a result, the concept of sustainability supplanted FSRE and was quickly adopted by funding agencies. USAID and others switched from funding FSRE projects to those explicitly working towards sustainable production. A sudden

discontinuance of explicit FSRE projects was mandated by the donor organizations who held the purse strings. However, the actual level of discontinuance of FSRE methodologies was small. Since the methodologies proved useful for dealing with the problems underlying sustainable production, FSRE methods were highly compatible with the new buzz word.
It seems that the adoption of FSRE had reached a
critical mass near the time that sustainability became the credo of the donor organizations. FSRE methodologies continue to diffuse without donor incentive. FSRE methodologies are now widely used and are understood as important tools in many research projects which do not explicitly claim to be FSRE projects (Okali, Sumberg & Farrington, 1994). When lower than expected adoption rates began to be recorded, FSRE methodologies were developed to respond to the lack of farmer participation of the past. FSRE efforts began to move more in the direction of consultative participation in an effort to identify why more widespread adoption was not taking place. Farming systems research has begun the process of normalizing collaborative work with farmers, from the problem identification stage through technology evaluation with on-farm trials.
Individual farmers must repeatedly make decisions about what commodities to produce, by what method, in which seasonal time periods, and in what quantities. Decisions are

made subject to the prevailing farm physical and financial constraints, and often in the face of considerable uncertainty about the planning period. Uncertainty may arise in forecasted yields, costs and prices for individual farm enterprises, in enterprise requirements for fixed resources, and in the total supplies of the fixed resources available.
1.1.3 Farm Management-and Linear Programming
Traditionally, farmers have relied on experience, intuition and comparisons with their neighbors to make decisions. Formal techniques of budgeting and comparative analysis were developed by farm management specialists, and these can be useful aids for making decisions in less complex situations or for analyzing selected decisions when all the other farm decisions are taken as given. It is only with the more recent advances in computers and in mathematical programming software that satisfactory procedures have been developed for whole-farm planning in more complex situations. Whole-farm planning can assist farmers in efficiently adapting to a changing economic and technological environment. Optimization models such as Linear Programming (LP) which adequately articulates the goals and constraints of representative farmers can also predict quite accurately what these farmers do. This is particulary true in more stationary situations where farmers have time to adapt to the economic and technological

environment. It is this predictive possibility of representative farm models that make them useful for inclusion in agricultural sector models intented for agregated policy anlysis. Delicate technologies developed solely under the conditions common on experiment stations are, however, rarely transferable directly to limitedresource farmers because they require a certain specific growth factors in order to perform well. Robust technologies, those capable of good, or at least adequate results under the broader, more variable, and often unpredictable range of conditions faced by farmers, tend to be systematically disfavored under usual experiment station conditions (Hildebrand & Russell, 1996).
1.1.4 Extension
On-farm research-extension, as carried out in the FSRE approach can convey knowledge of important problems to be addressed by on-station research. It can also attempt to adapt the results of this necessary discipline and commodity-based experiment station research through joint research-extension farmer evaluation of promising technologies under a broad range of environmental conditions. To incorporate farmers' perspectives and evaluations into the development of agricultural technology, two traditional methods have been used. The first is through

farmer feedback about the technology after viewing "demonstrations," or after some initial adoption during the extension/diffusion process. The second is through field-day visits by farmers to experiment station trials. Although these two methods are useful, it must be recognized that farmers are not likely to give pertinent opinions and evaluations of technologies they see "demonstrated" to them under foreign conditions, conditions almost always substantially more favorable than their own (Hildebrand & Russell, 1996). By conducting a considerable portion of the development of new technologies in close collaboration with farmers, through on-farm research-extension activities, incorporation of farmers' own evaluation criteria of those technologies will help to ensure relevance of the final recommendations eventually made. To be most effective, farmers' perspectives and judgments should be elicited continually during their participation in all phases of the technology development process. Such participation is essential in the identification of priority problems and potential solutions to those problems through both formal and informal diagnostic efforts, in design and implementation of on-farm trials, and in analysis and interpretation of trial results. Collaborating farmers also learn, through participation in trials, how best to implement new technological options and how to modify and

adapt them to specific local conditions in order to get the most out of them. On-farm researchers also learn from these adaptation efforts how better to design future technologies aimed at similar farmers, and also how better to interpret and extend the results of on-farm research.
In recent years, there has arisen an increased
awareness that the ecological and economic costs of creating and maintaining the high controlled, and homogeneous agricultural environments will not be sustainable in the long term. Greater attention is currently devoted to the development of "alternative" or "sustainable" agricultural systems. As agricultural policy makers begin to change the incentives which have encouraged the use of technologies broadly adapted to these superior environments, new technologies will have to be developed to conform with the the environments where they will used, not dominate them (Hildebrand, 1990). This change in emphasis and the concurrent imperative to target specific environmental conditions will make on-farm research-extension activities central to the development of technologies conducive to more sustainable agricultural systems. Socioeconomic analysis of limited-resource farms using Linear Programming (LP) and Adaptability Analysis (AA) using environmental characterization and evaluation criteria appropriate to specific farmers' production circumstances, can give

valuable feedback to policy makers, researchers, extension agents and farmers on one hand, and on the other hand, can help in ensuring the identification of technologies specifically adapted to the more variable environmental conditions of these systems.
1.2 History of Research Systems in Senegal
Senegal has a rich tradition of agricultural research, spanning over sixty years. During this period, the research strategy has evolved in response to both political changes-dating back to the colonial period--and the introduction of new methodologies from abroad. From the turn of the century until the mid-1970s, agricultural research in Senegal and through out Francophone West Africa was implemented under the auspices of the French overseas research institutes-most notably IRAT (crop research), IEMVT (livestock), IRHO (oil plants), IRCT (cotton) and ORSTOM (overseas scientific and technical research).
In 1975, the Government of Senegal established the
Senegalese Agricultural Research Institute (ISRA) as part of its policy to nationalize agricultural research. Under ISRA, crop research was concentrated at the Centre National de Recherches Agricoles (CNRA) at Bambey and carried out through a network of regional stations. This research focused on variety improvement and plant protection. Animal

production and veterinary research were carried out the Laboratoire National d'Elevage et de Recherches Veterinaires (LNERV) in Dakar and at two substations (Dahra and Kolda). Because Senegal served as the headquarters for agricultural research in Francophone West Africa until the 1960s, today the country has one of the most extensive research infrastructures in Sahelian Africa (Bingen & Faye, 1987). Since 1975, ISRA has conducted extensive research on groundnuts, cereals, cotton, cowpeas and soybean at CNRA/Bambey and has generated many research results for improving agriculture in the in the Sudano-Sahelian zone. In spite of these successes, in the 1960s many observers pointed out several weaknesses in the research system. For example, crop research had few links with extension and was concentrated on stations, with little research conducted at the farm level. Similarly, livestock research, which focused on veterinary research and breed improvement, neglected the production constraints faced by pastoralists (World Bank, 1981).
In response to these criticisms, in the early 1960s, ISRA established outreach (off-station) sites called Point d'Appui et d'Experimentation Multilocale (PAPEM) where researchers implemented multi-locational experimentation on soil fertility, crop rotations and variety evaluation

(Bingen & Faye, 1987). A major objective of PAPEM was to develop simple extension themes--mainly on fertilization and improved varieties (millet, sorghum, groundnuts)--that collaborating farmers could verify in their own fields; and which would serve both demonstrations and training purposes. In the late 1970s, ISRA realized that (1) farmers were not widely adopting research results, despite the contacts and opportunities provided through PAPEM, (2) it was possible to develop coherent extension packages and improved cropping systems models from many available research results, and (3) it was necessary to conduct on-farm adaptative research on a wide enough scale to take into account differences in farmers' socio-economic constraints. In responses to these realizations, ISRA initiated the Unites Experimentales (UE) project in 1979 to assess the relevancy to farmers of technologies developed on research stations. At the time, this project represented a unique attempt in West Africa to apply a systems approach, with a farm-level focus, to agricultural research (Gilbert et al., 1980; Norman et al., 1981). The UE project generated considerable knowledge about the technical and economic feasibility of intensifying agriculture and the process of transferring technology from research stations to farmers (Faye, 1978; Benoit-Cattin, 1982). Researchers and extension services gained a better

understanding of the structure and functioning of the farm family.
In 1981, the World Bank-funded Agricultural Research Project (ARP) was initiated to decentralize agricultural research at ISRA, develop interdisciplinary programs to address commodity improvement constraints, and implement production systems research (PSR) in the country's major agro-ecological zones. Under the umbrella of the Agricultural Research Project, USAID/Dakar subcontracted Michigan State University (MSU) to assist ISRA in implementing the Senegal Agricultural Research and Planning Project (SARPP)--designed to organize and carry out production systems research, as well as macroeconomic research (Kamuanga et al., 1992). As part ot the reorganization, ISRA established the Production Systems Department in 1982, responsible for establishing five PSR teams and managing several research support programs (bioclimatology, weed control, post-harvest technology and soil fertility). By 1986 the department had three operating production systems teams, stationed at Djibelor (Lower Casamance), Kaolack (Peanut Basin) and St Louis (Fleuve region).
1.3 Researchable Problem
The low adoption of improved technology by small-scale farm systems in Lower Casamance contributes to low average

productivity, deficient agricultural policy and marketing systems that do not meet farmers'needs and aspirations. Traditional methods used for the analysis of these small scale farm systems have not performed well for technology producing research. Methods and tools more adapted for onfarm research and extension to meet farmers' needs and situations are needed by the national research and extension organizations.
1.4 Hypothesis
There are methods of analysis of on-farm research and extension developed for small scale farm systems, that if used by researchers, extension agents and NGOs, could be of assistance in formulating recommendations that meet farmers' needs and situations, and provide valuable feedback to policy makers and policy change agents.
1.5 Objectives
1.5.1 Primary Objectives:
1- Compare the performance of new methods of analysis designed for on-farm research and extension with the traditional methods in the identification of agrosocioeconomic constraints and analysis of improved technologies of small-scale farm systems.
2- Propose adapted educational programs to facilitate the creation, transfer, and diffusion of improved technologies in the different agro-socioeconomic situations.

1.5.2 Secondary Oblectives:
1- Provide methods of analysis and tools adapted to onfarm research and extension to better understand, conceptualize, and model the nature and the complexity of family livelihood systems.
2- Provide appropriate methods of analysis and tools to predict responses of small-scale farm systems to improved technologies and give valuable feedback to research, policy makers and technology change agents.
1.6 Limitation of the Study
The study focuses mainly on the small-scale limited resource family farms in Lower Casamance, using two approaches: 1) Linear Progamming analysis (LP) of small farm livelihood systems and 2) Adaptability Analysis (AA) of on-farm Research-Extension data. The study intends to compare previous results of socio-economic analysis realized in 1984, and on-farm trials analyses realized in 1982-1985 by the Farming Systems Team of Djibelor (Lower Casamance). It also intends to bring to the attention of a larger audience the most up-to-date thinking in the use of comprehensive methods for design, analysis and interpretation of on-farm research and agronomic survey results of limited resource family farms, in different diffusion domains and different environments by incorporating farmers' perspectives and evaluation criteria

to make recommendation domains socially acceptable, technically and economically feasible.
1.7 Definition of Terms
Linear Programming Analysis (LP) is a model for determinating a profit maximizing combination of farm enterprises (activities) that is feasible with respect to a set of fixed farm constraints.
Adaptability Analysis (AA) is a simple and accurate method of identifying and comparing the performance of agricultural technologies under a wide range of specific biophysical conditions and socioeconomic circumstances (research domain).
Family farms are those which are first a home rather than a business. Decisions are made from the point of view of the home and the farm includes a wide diversity of enterprises or activities.
Farm constraints include land, labor, capital, management of many different types of activities, information services, and fixed resources.
Diffusion domains are informal and naturally occurring interpersonal communication networks for diffusion of information on agricultural technology. Often specific to the commodity or product involved.
Environment is the natural biophysical and

socioeconomically modified or created conditions existing for plant or animal growth in the trial location. It includes influences of any differences in management practices not part of treatments.
Environmental Index (E-l) is a convenient measure of the environment at the location of the trial. For a specific environment, it is the average response of all the treatments for that environment, usually based on physical yield per hectare.
Evaluation criterion is the measure or (measures) used to compare the treatments in a trial. It can reflect a researcher's concern (t/ha) for example, or a farmer's concerns and/or needs (kg/kg seed, kg/CFA cash cost, tillers/m2, among many others).
Recommendation domains are the situations for which specific treatments or technologies will be recommended. Extension messages can be designed differently for use in specifically defined diffusion domains within a single recommendation domain.
Risk is the probability (or percentage of time) that the selected evaluation criterion, such as t/ha, will fall below a certain level.
Trial usually refers to a set of treatments being
evaluated over a range of environments. It can also refer to

the set of treatments at each environment.
Diffusion is the process by which an innovation is
communicated through certain channels over time among the members of a social system.
Technology transfer is the exchange of technical
information between R&D workers who create a technological innovation and the users of the new idea.

2.1 overview of the Agricultural Sector
The West African Country of Senegal covers 196,722 square kilometers with maximum dimensions of 700 km east/west and 500 km north/south (Figure 1). Senegal has a total of 2,640 km of land boundaries with the Gambia, Guinea, Guinea-Bissau, Mali and Mauritania. The national territory is divided into ten administrative regions, each with three departments- which are in turn divided into districts (arrondissements). The population is further organized into urban communes and rural communities; there are 314 rural communities and 33 urban communes. The climate is tropical with two main seasons: a dry season from May to November and a rainy season from June to October. The country has six primary surface water systems, each with its own potentials and problems. Potentially irrigated lands are estimated at 240,000 hectares in the River Senegal Valley and at 20, 000-30, 000 ha in Eastern Senegal and in the Upper Casamance.
The 1994 population census counted eight million

z Low-$ I 4EGIO tv LIBERIA
Louga f malAm
REGION DU T ao ... e I to'JG-4
CAP VERT % _- 0
sea of
04ka, % %
mso., Golus 0
!?. 0 I
,0 AtCk
F dm % ft KAOIAC fKafftm MALI
< AMBIE t&-bfCDu.d. IS6
0 ol
0 o Too
Limille d' Etat Limited de R6gion Krn
Limited de Diparlement

people, 39% of whom live in urban areas. The population density, estimated at 41 persons/km2, is unequally distributed throughout the country. The population is composed of diverse groups, the most important of which are Wolofs, Sereres, Toucouleurs, Peuls, Diolas and Mandingues (MDR, 1986).
Senegal, as a Sahelian country, has few resources. Agriculture constitues the main source of revenue of two thirds of the Senegalese population. Because of the Sahelian drought, cultivated areas from 1961 to 1990 did not change and the average during this period was 2,352,000 hectares with 52% used for cereal production. Senegalese agricultural production is dominated by rainfed crops which represent 90% of the the total agricultural production and 90% of cereals produced locally (World Bank, 1991). Rice represents 3 percent; maize 4 percent; cotton 2 percent; cowpea 2 percent; groundnut 50%, and millet/sorghum 40%. Vegetable production, fruit production and cassava occupy very limited areas. Groundnuts and cotton are the main cash crops and their average yields are respectively 760 kg/ha and 1201 kg/ha (World Bank, 1991). Livestock is constituted mainly of cattle, sheep and goats. Livestock production fluctuates between 65,000 and 75,000 tons of carcass weight (Faye & Bingen, 1989).

Agricultural development today is the central point of Senegalese policy. The 60% of Senegalese living in rural areas have seen a decline of four to six percent per year in real per capita cash income over the last fifteen years (World Bank, 1991). Senegal has lost control of two of its most basic and fragile resources: land and forest. Pasture lands and protected forests represent 20% of the total surface area of the country. Available arable lands are limited and according to the "Septieme Plan" of Senegalese development policy, they are estimated at 3.7 million hectares that represents 19% of Senegalese surface area. As the result of stagnant total production and a shift from cash crops to food crops, the real market value of agricultural output has declined by 2.5% per year since 1960. Cash income to purchase food, to fill the production consumption gap, to purchase health and education services, farm inputs and implements and to invest in improved technology has declined. This means that a resource baseddevelopment strategy is essential to obtain improvement in the quality of life for Senegalese people. The primary sector (crops, livestock, forestry and fishing) represents about 20% of GDP in Senegal (World Bank, 1991), which is extremely low for an economy with a 60% rural population. In the Senegalese agricultural sector, it is clear that the resources, even though marginal, can be managed more

effectively and that productivity can be increased. This management for productivity requires that investments be redirected to support it, that policies be modified to encourage it, that technologies be proven to accomplish it, and that all these be brought to farmers, local organizations and NGOs (World Bank, 1991).
2.2 Characteristics of the Studied Area
The Lower Casamance covers an area of 7,300 km2 in Southern Senegal from the Soungrougrou Valley to the Atlantic Coast (Figure 2). It became the Ziguinchor administrative region in July, 1984 and comprises the Ziguinchor, Oussouye and Bignona Departements (MDR, 1984).
The region's subguinean climate receives a strong
maritime wind and is characterized by two seasons: a dry season from November to May and a rainy season from June to October, with August receiving the heaviest rainfall. The Atlantic Ocean has a dominant influence on the hydrology of the region because of its very low relief and the current rainfall deficit. Salt water frequently flows as far as 220 km upstream from the mouth of the Casamance River. The region has also an extensive network of lowland swamps which facilitate even further penetration of the sea water.
The nature of the soils depends on their position in the topographical sequence. On the upland, the soil is a sandy loam with a sandy surface. Two types dominate: (1)

red, low base status iron soils (ferralitique) with a higher clay content in the B-horizon and (2) ferruginous, leached tropical beige soils found in the central well-drained uplands (Beye, 1972; Bertrand, 1973). Along the talwegs (inland valleys) and the river itself can be found a sandy zone (sol gris de nappe) that is temporarily flooded during the rainy season and is favorable to palm groves. In the low-lying areas of the talwegs, rice is grown during the rainy season and commercial production of vegetables is carried out in the off-season. Saline soils are found in the river bottomland, the lowest position on the topographical sequence--alluvial flats (para-sulphate and acids or sulphate soils) and the mangrove swamps (potentially acid sulphate soils). "Mangrove" rice can be cultivated in these soils if there is sufficient rainfall to leach salt.
The rural population in the Lower Casamance is
estimated at 262,000 (ISRA, 1985) living in 330 villages. The population is evenly distributed and ranges from a density of around 10 per km2 in the Northeast (Bignona District) to 35 per km2 in the Southeast. The population is young and fairly mobile. There are two main ethnic groups:
(1) the Diola, who constitue the majority (85%) of the totalpopulation and who form a group which is composed of several distinct sub-groups (Kassa, Blouf, Fogny-Kalounayes,

1: Soilogny flbr-Dos oaia
-. MAURITANIE traction; transplanted rice dominant
ll: Social organize. of labor.- Dicta; no animal traction: upland crops impL.. direct seeding of rice SENEGAL
- Social orgartiz. of labor Maridinka ES dominant With Diolas & others. limited animal
traction; direct seeding of rice %ATBAC ~IV: Social 0rzanjL of labor.- Mandinka;, ex.panding use of animals, expanding upland
BISJV: Social organiL. of labor Didla: widespread U1 Sanmal traction: transplanted rice remains impt.
................ ............. ....
. . . . . . . . . . . . . . .
:~~.1ANV~l ....... .. .*:ODAA
............. ................ ............
......... ZONE IV

Fogny-Combo) and (2) the Mandingues, a minority group (5%), but whose cultural influence is great in the North, the Northeast and around Ziguinchor (ISRA, 1984).
The Lower Casamance is essentially an agricultural
region and because of its relatively abundant rainfall it plays an important role in Senegal' s agricultural development policy. Groundnuts, rice, millet/sorghum and maize are the main crops. Different types of livestock are found in the Lower Casamance, such as N'dama cattle and Guinean species of sheep, goats and pigs; however, very few donkeys and horses exist. The cattle are not equally distributed throughtout the region: 84% of the herd is in the Bignona District, 9% in Ziguinchor and 7% in Oussouye. Herd organization and management varies from North to South. Like other regions of Senegal, the Lower Casamance has experienced and continues to experience a considerable decline in rainfall. During the last ten years the mean has fallen to 1,000mm-l,100mm in Ziguinchor (compared to the normal 1,500mm). This average, however does not reflect the significant inter-annual variations. Each of the three districts has suffered several years of drought over the past ten-year period--so severe as to put the agricultural system in serious situations, decreasing the total area cultivated and the production of all cereals.
In spite of its relatively small area and the

predominance of one ethnic group, the social organization of agricultural production in Lower Casamance is quite heterogeneous. In order to control for this heterogeneity, the Djibelor Farming Systems divided the region into five zones which cover relatively homogenous agricultural situations (geographical area in which farmers confront similar constraints, have comparable potential to produce, and thereby constitue a group for which a common development strategy can be devised). The three criteria used by the Djibelor Farming Systems Team (ISRA, 1983) and derived from the peasant farmers' method of exploiting land, to divide the region are:
1) Division of Labor: Among the Diola in the South and the Northwest (Figure 2), men clear and plow the fields while the women transplant and harvest rice. The plowing and transplanting of rice fields, the most labor intensive activities, do not begin until mid-August, allowing farmers sufficient time to plant their rice nurseries, household gardens and upland fields in July. Among the Mandingues and the Diola "mandinguise" (influenced culturally by the Mandingues), the division of labor is more simply defined by the topographical location of the crops: men cultivate upland rainfed crops while women cultivate rice.

2) The Relative Proportion of Upland versus Lowland Crops: As the rainfall levels decline from the Southeast to the Northeast, the type of cropland gradually changes from aquatic rice land (transplanted rice) to rainfed, upland crops. Upland crops represent less than 50% of the total land under production in the Southeast, but exceeds 80% of cropland in the North.
3) The Use of Animal Traction: This criterion differentiates one production system from another in the Lower Casamance. More than 90% of the upland areas in the Sindian-Kalounayes (Northeast) and 50% in the villages of Fogny-Combo (North) are plowed with animal-drawn equipment. Animal traction is almost unknown in the Southeast. The five existing agricultural zones are: (1) Zone I: characterized by a Diola-type organization without animal traction. Transplanted rice is dominant; (2) Zone II: is a Diola-type organization with a dominance of rainfed crops and direct seeding of rice; (3) Zone III: is a cosmopolitan zone with some animal traction and direct seeded rice; (4) Zone IV: is characterized by a Mandingue-type organization with an importance of upland crops and animal traction. Zone V has a Diola-type organization of labor. Animal traction is important with a significance of transplanted rice.

2.3 Description of the Representative Household
The main characteristics of the household in Loudia Ouloff
(Zone I), are summarized in Table 2.1.
Table 2.1: Description of the representative household
Farm resources
- Family farm size (ha) 2
- Community land (ha) 3
- Family labor (person-day/ha) 700
- Cash available (francs CFA) 6000
- Seeds groundnut (kg) 40
- Seeds maize (kg) 20
- Seeds aquatic rice (kg) 100
- Seeds cowpea (kg) 12
- Seeds sweet potatoes (cuttings) 100
Labor for the different activities (persdays/ha)
- Groundnuts 96
- Maize 94
- Aquatic rice 144
- Cowpea 30
- Sweet potatoes 8
- Palm oil 158
- Fruit production 15
Family consumption needs (kg/year)
- Groundnuts 83
- Maize 500
- Aquatic rice 602
- Cowpea 30
- Sweet potatoes 46
- Palm oil (liters) 50
- Fruit production 84
Seeds for different crops (kg/ha)
- Groundnuts 60
- Maize 15
- Aquatic rice 120
- Cowpea 10
- Sweet potatoes (cuttings) 10000
Yields of crops (kg/ha)
- Groundnuts 608
- Maize 554
- Aquatic rice 120
- Cowpea 122
- Sweet potatoes 4600
- Palm oil (liters) 120
- Fruit production 200
(Source: FSR/E, ISRA. Djibelor, 1984)

3.1 Population and Sample
In 1982, a map delimiting farming zones in Lower
Casamance (Senegal) was constructed by the Farming Systems Team of Djibelor on the basis of three criteria: the sexual division of labor, the extent to which animal traction had been adopted, and the importance of transplanted rice versus other cereals in the cropping systems. These criteria resulted in five agricultural zones (Figure 2). The Team then chose two villages in each zone in order to take in account the variability within each zone. A random selection of between 10 and 15 farms per village resulted in a sample of 125 farms across ten villages. These farms were studied from 1982 through 1985. For the purpose of this present work, a representative farm based on the data in Table 2.1 was choosen in zone I (Loudia Ouloff). These data were used for the linear programmining analysis. For the Adaptability Analysis (AA) of on-farm trials, the sample includes nappe (phreatic) rice variety trials conducted from 1982 to 1985. Data from these on-farm trials were used for the adaptability analysis.

3.2 Data Collection and Analysis
Data to augment those from earlier stidies were
collected in Lower Casamance (Senegal) during the summer of 1996. Variables included existing data from an agronomic survey and on-farm trials conducted by the Farming Systems Team of Djibelor from 1982 to 1985. For the socio-economic analysis of small limited-resource farms using the linear progamming model, data were collected from the 1984 agronomic survey realized by the Farming Systems Team of Djibelor in Lower Casamance, Senegal. Data included available resources of the representative farm choosen in Loudia Ouloff, zone I (Table 2.1), different characteristics of the farm systems, systems components and interaction, gender analysis including activities, seasonal calendars, responsiblities and benefits, resource use and control, farmers' criteria, etc. For the Adaptability Analysis, data were collected from the agronomic on-farm trials conducted by the Farming Systems Team of Djibelor from 1982 to 1985. The data set included nappe rice variety on-farm trials. The data set also included characteristics of the different environments (biophysical and socio-economic factors) where the trials took place, and farmers' own evaluation criteria.
3.3 Methods of Analysis
The methodology of analysis used includes the two
methods of analysis (Linear Programming and Adaptability

Analysis) developed for on-farm Research and Extension, compared to the traditional methods (FARMAP/BRADS and ANOVA).
1) Method of Economic Analysis of technology and policy changes in small livelihood systems using Linea-r Progarmming
(LP): This method is used to conceptualize, and to model the nature and complexity of small farm systems to predict the response of family livelihood systems to improved technologies and give valuable feedback to research, policy makers, local organizations (public and private) and technology change agents. The method includes the following:
(1) definition of basic linear programming matrix to adequately reflect the resources, activities and constraints, (2) definition of the family consumption constraints and intermediate products and interactions, (3) gender analysis and resource flow, (4) simulaton of the representative farm systems and (5) assessing alternative technology and policy analysis. Linear programming uses sets of linear equations in an optimization procedure that allocates scarce resources among competing alternatives to maximize specified objectives. The standard form of a linear programming model is composed of three sections: (1) the objective function, (2) resource constraints, (3) activities or competing alternatives. The model used in this thesis is

a maximization procedure whose objective function maximizes cash income.
2) Method of Adaptability Analysis (AA) of on-farm
research and extension: Adaptability analysis is a method used to analyze and interpret on-farm research trials, to provide technological recommendations for specific biophysical and socioeconomic environments and for individual farmers' evaluation criteria. It includes the following: (1) calculation of the environmental index (El),
(2) identification of treatment response to environment, (3) assessement of treatment interaction with environment, (4) characterization of the environment, (5) interpretation of results and definition of recommendation domains, (6) comparison of results using alternative evaluation criteria, and (7) creation of extension recommendations for each recommendation domain and formulation of messages appropriate to each of the diffusion domains. Data were analyzed with the Quatro-pro 6.0 statistical software program.

4.1 Characteristics of the Village of Loudia Ouloff
Loudia Ouloff, a village located in Zone I, south of the River Casamance, represents the traditional Diola system. The traditional Diola live in households (butong) composed of conjugal units with autonomy in economic matters. Villages are organized in groups of individual residential units (eluf). Households follow an intensive aquatic rice production system with a marked division of labor by task. The heavy work of dike building and ridging is done by men, while women transplant and harvest rice. Land is nominally owned by patrifilial groups (Linares, 1981; Diouf, 1984), but usufruct rights to land lie with the conjugal unit under the direct responsibility of the head of the compound. Women, as a rule do not own land. The village of Loudia Ouloff has a population of 306 people with 43 households (Farming Systems Team Djibelor, 1983). The average size of the work force by household/farm is five. The village also has 172 hectares of community land. The average farm size is 2 to 2.5 ha.

4.2 Characteristics of the Studied Household
4.2.1 Schematic Modelina of the Household
The representative household is composed of nine people (3 men, 4 women and 2 children) It has a total of two hectares of family land for crop production and three hectares from the community land used for different purposes (fruit production, firewood, palm oil collection, etc.). The active labor force is five (Table A.1) Palm oil and fruit production are the main sources of family revenues. The farm also has six sheep, six goats, a family herd of 15 cattle and some chickens. The family constitues the main source of labor for all activities. During peak periods (plowing, transplanting and harvesting), the farm can hire a limited amount of labor if cash is available. Rice production which is never sold according to Diola tradition, remains the staple food. Livestock is used for ceremonies and rituals and constitues the main source of manure for crop production. Fuel, medicine and other family needs are purchased from the market.
4.2.2 Gender Analysis of the Household
The household activities (Table A.2) include crop production (rice, groundnuts, maize, sweet potatoes and cowpea), livestock management, non-agricultural activities (fruit picking, palm oil) and domestic activities. Both men and women are involved in these different enterprises.

Both men and women have access to and control of the resources (Table A.3) but land management is the responsibility of men. Farm revenues come from selling palm oil and fruit. The farm does not have access to chemical fertilizers and other inputs, nor credit. Most of the varieties used for crop production are local varieties characterized by their long cycles (late maturing). Both men and women have access to the farm benefits and incentives (Table A.4). All the upland crop production is controled mainly by men while women control the production of lowland crops (rice). For livestock, women control only the small ruminants (goats, sheep). Women also control the domestic household activities such as cooking, fetching water, firewood collection, etc. Construction material and building material are in the hands of men. Palm oil collection is realized by men while women do the extraction and the benefits of this activity are controlled by women. Benefits and incentives gained from all these activities are shared commonly, part of them are reinvested in agricultural production or used for family needs.
For crop production, the seasonal calendar (Table A.5) begins in May and ends in October-November, depending on the rainfall pattern. For groundnuts, land preparation ranges from May until the first good rain (June). It is done by men. Land preparation takes 34 person-days and seeding 12

person-days per ha. Weeding of groundnuts is done generally one time at the end of August or the beginning of September and takes around 30 days of family labor. Harvesting, done both by men and women takes 13 person-days per ha, while threshing and conditioning total 7 person-days. For maize, land preparation takes 34 person-days per ha of the family labor; seeding, two person-days per ha; weeding and harvesting take 33 and 25 person-days respectively. For aquatic rice (transplanted rice), the activities begin after the seeding of upland crops (groundnuts and maize) when the lowlands are wet and easy to plow. Clearing/nurseries and plowing take about 50 person-days per ha of planted rice. Transporting and transplanting of seedlings total an average of 46 person-days of female labor. Weeding when it is done takes around 2 person-days per ha. Harvesting rice is a slow and difficult task due to the traditional method used (sheaves). This activity, done only by women, takes 44 person-days while conditioning takes around two person-days per ha. Cowpea and sweet potatoes are secondary crops in this area and take 30 person-days and 8 person-days respectively of the total family labor. Palm oil production and fruit harvesting are done throughout the year but more intensively during the dry season. They take respectively 158 person-days and 15 person-days of the family labor. Domestic activities such as firewood collection, cooking,

fetching water, and child care are daily activities. The farm total wet season labor is estimated at 700 person-days (66% for the wet season and 34% for the dry season). The male wet season labor and the female wet season labor represent respectively 48% and 52% of the total family wet season labor. Male dry season labor represents 46% of the total family dry season labor and female dry season labor is 54%.
4.3 Linear Proaramming Analysis of the Household
4.3.1 Resources and Minimum Survival Constraints
The farm possesses two hectares of family land for crop production to satisfy family consumption needs (Table A.6) and can also use three hectares of the community land for palm oil and fruit picking production, firewood, building material, etc. Family labor remains the main source of farm labor. The farm does not use chemical fertilizers nor improved varieties due to the lack of credit. Aquatic rice (transplanted rice) constitues the main crop for the household. Its average production is 854 kg/ha. For groundnuts and maize the yields are 608 kg/ha and 554 kg/ha respectively. Cowpea produces 122 kg/ha. Palm oil production and fruit production count 120 liters and 200 kg/ha, respectively. The farm's annual consumption needs are 602 kg of rice, 500 kg of maize, 83 kg of groundnuts, 50 liters of palm oil and 84 kg of fruit.

4.3.2 Maximizing the Family Income
By maximizing the family cash income after satisfying consumption requirements (cash for discretionary spending), selling palm oil and fruit bring to the family a net income of 92,945 francs CFA ($265). The farm produces 114 liters of palm oil, 242 kg of fruit and satisfies its consumption needs using the two hectares of the family land and the three hectares of community land (Table A.7). The farm uses 207 person-days of the family wet season labor and 240 person-days (the limit) of the family dry season labor. The solution is not sensitive to the palm oil price of 700 francs CFA and will not change over a price range from 214 to more than 1,200 francs. For the resources, the family land and the community land are not limiting although all is essentially used. Family wet season labor is not binding but family dry season labor is binding and an additional day of dry season labor would add 435 francs CFA ($1.24) to family income.
4.3.3 Influence of Gender in the Househol
In the previous solution, labor is not disaggrated by gender. The particularity of Zone I, however, is the sexual division of labor. Men and women work together but the labor time remains different for each sex. The division of labor depends on the task that each sex plays in the household and this is incorporated in the next model, Table A.8. Wet

season labor still is not binding. By disaggrated dry season labor, less total labor is used (233 days versus 240) because the woman's labor is a constraint. Men use 93% of their labor during the dry season, while women use 100%. An additional day of female dry season labor would add 729 francs CFA ($2.08) to family income.
Cultures vary as to who has the most decision making
authority between the man and the woman in the household. A way to analyze this effect is to maximize income to each separately and compare these solutions with the solution that maximizes income to the family.
4.3.4 Maximizing Female Income
By maximizing female income, after the farm satisfies all its consumption needs, using 2 hectares of the family land and 1.99 hectare of the community land (Table A.9), the farm also produces 111.77 liters of palm oil for selling which brings to the female an income of 78,238 francs CFA ($224) No fruit is sold, so family income is reduced.
Female dry season labor is binding, and an additional day of female dry season labor would add 884.21 francs CFA ($2.53) to her income.
4.3.5 Maximizinar Male Income
By maximizing male income, after the family also
satisfies all its consumption needs it produces 426.94 kg of fruit for sale (Table A.10) Labor is not a restriction but

all available community land is used. Selling fruit brings to the male an income of 34,154 francs CFA ($98).
4.3.6 Pronosition of Basket Making as one Alternative Solution for the Household
Based upon linear programming, we have seen that the farm has limited resources but earns substantial revenues from non-agricultural activities such as fruit production and palm oil collection. Increased production is limited due to the lack of land; labor also constitues a constraint to increased production of palm oil during the dry season. As an alternative solution to increase family income, basket making can be an alternative for the farm because of the availability of the raw material and the prices offered in the market (500 francs CFA) Basket making is a female activity and is done during the dry season. The LP model that best simulates the farm situation, (maximizing family cash income) shows that female dry season labor is binding. To achieve this activity, the farm needs to hire labor to collect raw material and make baskets. The acceptability of this innovation depends mainly on the availability of cash which can be obtained easily from the revenues of palm oil and fruit production. The results of Table A.11 show that by hiring labor (50 person-days) at the rate of 250 francs CFA ($0.71) per person, the farm can produce 133 baskets, 252.96 kg of fruit, 107.83 liters of palm oil and satisfy easily

easily the family consumption needs. When sold, basket making brings to the family an income of 66,500 francs CFA ($190) which increases the total family income to 143,915 francs CFA ($412).
With this scenario, community land is binding. Male wet season labor, male dry season labor and female wet season labor are not binding but female dry season labor still is binding. An additional day of female dry season labor will add 729 francs CFA ($2.08) to female income. But if more than 3 extra days were available, female dry season labor would no longer be limiting. A decrease by 71 days would also bring change to the results. Availability of hired labor is also a constraint and an additional day would add 583 francs CFA ($1.66) to family income.
4.4 Comparison of Results Analysis of Linear Programming
(LP) with Farmap/Brads Method
The FARMAP data were used to construct the Linear Programming (LP). This comparison is a form to show how closely we were able to simulate the representative farm with the linear programming model.
Comparison of the linear programming model of the
representative farm in Lower Casamance with FARMAP/BRADS data shows that the linear programming model closely resembled the representative farm (Table 4.1). But one disavantage of the FARMAP/BRADS method is the fact that it

does not allow to predict how the improved technologies, taking in account existing farmers' resources and constraints, fit farmers' situations. Linear Programming provides this opportunity and shows how well the introduced technologies fit farmers' situations and how the resources should be allocated to the different farm entreprises to better meet farmers' needs and aspirations, allowing researchers, extensionists, NGOs and other partners of the development process to have valuable feedback prior to the introduction of the technololgies in the farm level. The example used with Linear programming analysis in zone I is a perfect illustration of the usefulness and the performance of this method.
Introduction of basket making as an alternative
innovation in the household in Loudia Ouloff (Zone I), fits well in the farms' environments and increases farm income to 143,915 francs CFA ($411). Linear progamming facilitates (1) a better understanding of the objectives and constraints facing the farm family, as well as the dynamics of the farming system, (2) a study of the factors likely to affect wide-scale adoption of proposed improved technologies and
(3) valuable feedback to researchers, extention agents, NGOs, local communities, decision-makers and other partnerships of the technological development process.

Table 4.1: Farm budgets and income (francs CFA)a in the surveyed areab for a rainy year (1984), Lower Casamance,
Senegal, comparing FARMAP/BRADSc and Linear Programming (LP) analyses.
Indicators Representative Farm LINEAR PROGRAMMING RESULTS
(FARMAP/BRADS) (Gender Analysis)
Zone I(Loudia Ouloff) Zone I(Loudia Ouloff)
Family farm area 1.89 2
Community land 3 3
Total labor used 141 89.6
Value of
production 98,422 89,748
(francs CFA)
Production cost 5,907 5,966
(francs CFA)
Net farm income 92,515 83,782
(francs CFA)
Income per ha 18,919 16,756
(francs CFA)
Income per 134 187
a = US $1.00 equals 350 francs CFA in 1984 b = Average budget for representative farm c = FARMAP/BRADS is the programm used for setting and analyzing data collected from households. It is based on the system of codes developed by FAO and adapted to computer IBM5120. This programm was used by the Farming Systems Team (Djibelor/Lower Casamance) to analize agrosocioeconomic survey of small households in Lower Casamance.

In the late 1970s, ISRA realized that (1) farmers were not adopting research results despite the contacts and opportunities provided through PAPEM and the "Unit6s Experimentales" (UE), (2) it was possible to develop coherent extension packages and improved cropping system models from many available research results, and (3) it was necessary to conduct on-farm adaptative research on a wide enough scale to take in account differences in farmers' socioeconomic constraints. During the 1970s, most farmers acquired farm equipment, fertilizer and improved varieties through the governmental credit program. When it was discontinued in 1979, many farmers, particulary in Lower Casamance stopped using fertilizers due to the lack of credit for purchasing inputs.
In response to this situation, the Farming Systems Team of Djibelor, created in 1982, began introducing some new technologies for main crops in Lower Casamance, to improve farmers' agricultural productivity. The example of on-farm trials used for this following analysis is taken from nappe rice variety on-farm trials conducted by the Farming Systems 58

Team of Djibelor from 1982 to 1985.
The object of the trials was to study and evaluate
several improved nappe rice varieties. The varieties used had already been tested on the station and in multilocational trials. The field-size comparisons consisted of five plots. Four were planted with the improved varieties and one plot with the farmers' local varieties. The plots were managed under local agronomic practices. There was only one plot for each variety on each farm each year with a total of 13 environments over four years. Four improved varieties were studied: DJ12619, IRATII2, IRAT133, and IKP (Table 5.1). All of these are short cycle varieties (90-105 days) and have good tillering capacity (average of 235 tillers/m2 on-station) Data regarding the environmental characteristics for each site (farmer) and year were recorded. They include plowing type (flat vs ridge), seeding dates, rainfall pattern and soil samplings (pH and percent of organic matter). Trials were implemented on 13 sites (12 on farm and one on station).
The results analysis of ANOVA (Table 5.1) over the four years (1982-1985) in 13 sites (environments) shows that variety IKP gave the best average yield (2666 kg/ha), followed by variety DJ12519 with 2564 kg/ha and farmers' local varieties with 2119 kg/ha. The analysis also shows significant differences that can be attributable to the

separate effects of year, site and variety. Based on this analysis of ANOVA, two improved varieties would be recommended: IKP and DJ12519 and the indigenous local variety. But the recommendations to be more efficient should take in account the environmental characteristics of each site.
Table 5.1: ANOVA for yields of nappe rice variety trials (kg/ha). Lower casamance. Senegal (Source: FSR/E. ISRA/Djibelor, Annual report 1985/1986).
Villages DJ1251 IRAT112IRAT133 IKP LOCAL MEAN Fcal. CV% /year
Maoua/82 1383 0 0 1633 1766 956 NS 22
Maoua/83 1750 616 1067 1408 1350 1238 NS 26 Maoua/84 767 317 908 1017 433 688 14.3 (b) Boul./82 3466 1700 2233 2933 2200 2506 9.3 (a) Boul./83 3366 433 3000 2633 2600 2406 24.5 (b) Boul./84 4433 3766 3366 4516 2516 3719 21.7 (b) Medeg/82 2366 0 2171 2900 2200 1927 NS 26
Medeg/83 3883 2055 2883 3533 3083 3087 24.4 (b) Medeg/84 4017 2183 3100 4833 2916 3409 6.9 (a) Bandj/84 1500 1126 1750 2034 1750 1632 NS 30
Bandj/85 1250 1750 1250 2250 2416 1783 5.1 (a) Suel/84 1111 740 842 1379 1425 1099 NS 28
Stat./84 4039 2357 2363 3595 2892 3049 12.6 (a) Mean 2564 1311 1918 2666 2119 2116
aSignificant at 5%; bSignificant at 1%
5.1 Calculatina the Environmental Index (EI)
The range of environments (Els) calculated for each site/year is presented in Table 5.2.

Table 5.2: Response of four improved nappe rice varieties (kg/ha) and the farmers' local varieties from on-farm research results (1982-1985). Lower Casamance, Senegal. Farmer/year Yields (kg/ha) of varieties sorted by EI
6/84 4433 3766 3366 4516 2516 3719
9/84 4017 2183 3100 4833 2916 3410
8/83 3883 2055 2883 3533 3083 3087
13/84 4039 2357 2363 3595 2892 3049
4/82 3466 1700 2233 2933 2200 2506
5/83 3366 433 3000 2633 2600 2406
7/82 2366 0 2171 2900 2200 1927
11/85 1250 1750 1250 2250 2416 1783
10/84 1500 1126 1750 2034 1750 1632
2/83 1750 616 1067 1408 1350 1238
12/84 i111 740 842 1379 1425 1099
1/82 1383 0 0 1633 1766 956
3/84 767 317 908 1017 433 688
Col. AVG 2563.92 1311.00 1917.92 2666.46 2119.0 2116
The overall mean is 2116 and the range of EIs is 3031 so the ratio of range to mean is 1.4. This is an acceptable ratio that indicates a wide range of environments has been sampled. The other two conditions to indicate high quality data (Hildebrand & Russell, 1996) generally are also met. It appears that the yield of the local varieties are generally quite high, but the minimum is 433 kg/ha. The distribution of EIs is acceptable. The best varieties, based on yield

(kg/ha), are IKP with an average of 2666kg/ha, followed by DJ12519 with 2564 kg/ha and farmers' local varieties with 2119 kg/ha.
5.2 Relating Treatment Response to Environments on EI
Treatment (variety) response to environment, by regression on environmental index (EI), is presented in Table 5.3. From this table, it was concluded that data for varieties DJ12519, IRAT133 and IKP can be represented by linear regressions because R2 of the linear regression of each of these varieties is similar to their respective R2 quadratic regressions. For varieties IRAT112 and local varieties, their respective quadratic regressions better fit the data because their respective R2 (0.75 and 0.85) are greater than R2 of their linear regressions. The yields and regressions, plotted on the environmental index (EI), are presented respectively in Figures 5.1 to 5.5. Figure 5.6 presents the estimated responses of all varieties on environmental index
(EI), using kg/ha as the selection criterion. In good environments (EIs > 2000 kg/ha), the best varieties are IKP, and DJ1519. In poorer environments, the farmers' local varieties join the other two as superior.
Based on the criterion kg/ha, we can tentatively
recommend varieties IKP and DJ12519 for good environments (EIs > 2000 kg) and varieties IKP, DJ12519 and farmers' varieties for poor environments (EIs < 2000 kg).

Table 5.3: Nappe rice variety trials regression estimates.
DJ12519 lin Regression Output DJ12519 q Regression Output
Constant -153.132 Constant -183.0222
Std Err of Y Est 393.7941 Std Err of Y Est 412.9594
R Squared 0.91981 R Squared 0.9198319
No. of observations 13 No. of Observations 13
Degrees of Freedom 11 Degrees of Freedom 2.0
X Coefficient(s) 1.284258 X Coefficient(s) 1.31802 -7.709E-05
Std Err of Coef. 0.114331 Std Err of Coef. 0.662824 0.0001488
IRAT112 lin Regression Output IRAT112 q Regression Output
Constant -647.2731 Constant 758.9227
Std Err of Y Est 650.5132 Std Err of Y Est 603.534
R Squared 0.685882 R Squared 0.754203
No. of observations 13 No. of observations 13
Degrees of Freedom 11 Degrees of Freedom 10
X Coefficient(s) 0.9256079 X Coefficient(s) -0.66275 0.000363
Std Err of Coef .0.188655 Std Err of Coef. 0.968691 0.000218
IRAT133 lin Regression Output IRAT133 q Regression Output
Constant -115.755 Constant -646.985
Std Err of Y Est 431.7577 Std Err of Y Est 436.5858
R Squared 0.842413 R Squared 0.853518
No. of Observations 13 No. of Observations 13
Degrees of Freedom 11 Degrees of Freedom 10
X Coefficient(s) 0.961249 X Coefficient(s) 1.561298 -0.00014
Std Err of Coef 0.125354 Std Err of Coef. 0.700746 0.000157
IKP lin Regression Output IKP q Regression Output
Constant 180.7105 Constant 558.04262
Std Err of Y Est 313.50131 Std Err of Y Est 317.52598
R Squared 0.938073 R Squared 0.9422482
No. of Observations 13 No. of observations 13
Degrees of Freedom 11 Degrees of Freedom 10
X Coefficient(s) 1.174929 X Coefficient(s) 0.748715 0.9732E-05
Std Err of Coef 0.09102 Std Err of Coef. 0.509648 0.0001144
Local lin Regression Output Local q Regression output
Constant 735.4496 Constant -486.958
Std Err of Y Est 399.4582 Std Err of Y Est 314.6677
R Squared 0.742962 R Squared 0.855001
No. of Observations 13 No. of observations 13
Degrees of Freedom 11 Degrees of Freedom 10
X Coefficient(s) 0.6539564 X Coefficient(s) 2.034722 -0.00032
Std Err of Coef 0.1159759 Std Err of Coef. 0.50506 0.000113

7 o DJ12519 in
500 1000 1500 2000 2500 3000 3500 4000 ENVIRONMENTAL INDEX (El)
Figure 5.1: Linear response in kg/ha of variety DJ12519 to environmental index (EI).
3000 obs
SIRAT112q 1000.
500 1000 1500 2000 2500 3000 3500 4000 ENVIRONMENTAL INDEX (El)
Figure 5.2: Quadratic response in kg/ha of variety IRAT112 to environmentalal index (EI).
200 0 o IRATi33 1in
S00 1000 1500 2000 0500 3000 3500 4000 ENVIRONMENTAL INDEX (El)
Figure 5.3: Linear response in kg/ha of variety IRAT133 to environmental index (EI).

5 IKP lin
So00 1000 1500 2500 30o 00 4000 ENVIRONMENTAL INDEX (El)
Figure 5.4: Linear response in kg/ha of variety IKP to environmental index (EI).
10 IO I s 000 20no 3 m 3500 4o
Figure 5.5: Quadratic response in kg/ha of variety Local to environment index (EI).
"oo /LOCAL El
500 1000 1500 2000 2000 3000 3500 4000 ENVIRONMENTAL INDEX (El)
Figure 5.6: Estimated responses in kg/ha of five rice varieties to environment (EI) for nappe rice variety trials (1982-1985). Lower Casamance, Senegal.

5.3 Relationship between El and Environmental Characteristics
Characterization of the different environments usually can help better understand the existing differences in yields of the different varieties and to see how the specific environmental characteristics are related to Els. The environmental characteristics available include the different existing plowing type (flat vs ridge), soil samplings for pH and organic matter, number of days between seeding and weeding, the rainfall pattern and the agricultural zones. Data characterizing the environments are presented in Table 5.4. The data are sorted by El.
On this case, it is evident that none of the measured
data nor plowing practices are related to El. However, it is clear that the best environments are in zone IV and on the experimental station, and the poorest environments are in zones III and V. Zones III and V generally correspond to an El below 1800. Zone IV is located in the manding area characterized by a large use of animal traction and flat plowing. Zones III and V are located in the cosmopolitan and in the Fogny-Combo zones respectively. Ridge plowing is dominant in zone V, while in zone III, the two types of plowing (flat and ridge) are used.

Table 5.4: Enviromental characteristic data, sorted by EI, for nappe rice variety trials (1982-1985). Lower Casamance. Senegal.
EI Zones Plowing Rain Interval pH %
type (mm) seeding/weeding organic
(days) matter
3719 IV Flat 1015 47 4.30 6.10
3410 IV Flat 766 23 3.30 4.40
3087 IV Flat 638 19 3.90 4.10
3049 STI Flat 1259 19 4.40 3.00
2506 IV Flat 843 28 4.10 5.60
2406 IV Flat 580 15 4.50 5.10
1927 IV Flat 786 18 3.30 3.38
1783 V Ridge 1159 32 4.00 3.31
1632 V Ridge 687 30 4.10 3.58
1238 III Flat 806 17 3.30 4.00
1099 V Ridge 698 28 5.10 0.13
956 III Flat 898 17 3.70 4.30
688 III Flat 1199 29 3.60 4.10
1 = station
5.4 Definition of Tentative Recommendation Domains
For the researcher' s criterion (kg/ha), the following tentative recommendation domains presented in Table 5.7 can delineated.

Table 5.5: Tentative recommendation domains and the technologies recommended based on environmental characteristics (zones) and the evaluation criterion kg/ha. Environmental characteristics Recommendations
(Zones) Evaluation criterion (kg/ha)
IV and Station IKP and DJ12519
III and V IKP Local and DJ12519
5,5 Determining Risk Associated with the New Technology
The formula Y [ (t(df=nl,p)) (Sd) /Vnl provides
information on the probability that values lie below the confidence interval and is a measure of the level of risk associated with the technology in these tentative recommendation domains. Table 5.6 presents summaries of the risk calculations with environmental characteristic based on Zones. Figures 5.7 and 5.8 present graphs for each variety comprised the different selected environment. Table 5.6: Calculation of lower confidence limits (kg/ha) for two yielding domains based on Zones for five nappe rice variety trials (1982-1985). Lower Casamance, Senegal.
EI Zones DJ12519 IRAT112 IRAT133 IKP LOCAL
3719 IV 4433 3766 3366 4516 2516
3410 IV 4017 2183 3100 4833 2916
3087 IV 3883 2055 2883 3533 3083
3049 STATION 4039 2357 2363 3595 2892
2506 IV 3466 1700 2233 2933 2200
2406 IV 3366 433 3000 2633 2600
1927 IV 2366 0 2171 2900 2200

Table 5.6 (continued)
EI Zones DJ12519 IRAT112 IRAT133 IKP Local
1783 V 1250 1750 1250 2250 2416
1632 V 1500 1126 1750 2034 1750
1228 III 1750 616 1067 1408 1350
1099 V i111 740 842 1379 1425
956 III 1383 0 0 1633 1766
688 III 767 317 908 1017 433
Avg of IV-Station 3653 1785 2731 3563 2630
STDS of IV-Station 673 1259 471 839 351
Square Root of n=7 2.65 2.65 2.65 2.65 2.65
Avg of III-V 1294 758 970 1620 1523
STDS of III-V 338 618 576 455 654
Square Root of n=6 2.45 2.45 2.45 2.45 2.45
Risk analysis for Zone IV and Station (n = 7) ALPHA PROB. t,df=6 DJ12519 IRAT112 IRAT133 IKP LOCAL
0.25 25.00 0.71 3472.29 1447.11 2604.42 3338.11 2535.37
0.20 20.00 0.91 3421.43 1351.97 2568.81 3274.68 25.8.84
0.15 15.00 1.13 3365.48 1247.32 2529.63 3204.91 2479.65
0.10 10.00 1.44 3286.64 1099.85 2474.42 3106.60 2438.52
0.05 5.00 1.94 3159.48 861.99 2385.39 2948.02 2372.18
0.03 2.50 2.45 3029.77 619.39 2294.57 2786.28 2304.51
0.01 1.00 3.14 2854.29 291.16 2171.70 2567.45 2212.96
0.01 0.50 3.71 2709.33 20.00 2070.19 2386.68 2137.34
0.00 0.05 5.96 2137.11 -1050.3 1669.52 1673.10 1838.81
Risk analysis for zones III and V (n = 6) ALPHA PROB. t,df=5 DJ12519 IRAT112 IRAT133 IKP LOCAL
0.25 25.00 0.73 1192.75 574.01 797.92 1484.57 1328.56
0.20 20.00 0.92 1166.53 526.09 753.26 1449.28 1277.87
0.15 15.00 1.16 1133.40 465.55 696.85 1404.70 1213.83
0.10 10.00 1.48 1089.24 384.83 621.64 1345.26 1128.45
0.05 5.00 2.01 1016.09 251.13 497.07 1246.81 987.04
0.03 2.50 2.57 938.80 109.86 365.44 1142.79 837.63

Table 5.6 (continued)
ALPHA PROB t,df=5 DJ12519 IRAT112 IRAT133 IKP LOCAL 0.01 1.00 3.36 829.77 -89.42 179.76 996.04 626.85 0.01 0.50 4.03 737.30 -258.43 22.28 871.59 448.08
0.00 0.05 6.86 346.72 -972.32 -642.88 345.92 -306.99
Results of risk analysis for the two domains: highyielding environments (Zone IV and Station) and low-yielding environments (Zones III and V) are presented in Figures 5.7 and 5.8 respectively. In Zone IV and on Station, varieties DJ12519 and IKP would be recommended. In Zones III and V, IKP would be prefered but also Local and DJ12519 would be recommended.
4000 L
- - - - - - - -3000 DJ12519
-IRAT112 IRAT133
0 5 10 15 20 25 RISK (% time below y-a ids vakue)
Figure 5.7: Risk levels for five nappe rice varieties for Zone IV and Station (Evaluation criterion kg/ha).

--IRAT112 IRAT133
...... IKP
==.. LOCAL
0 5 0 I5 20 25
RISK (% Ume below Y- &ods value)
Figure 5.8: Risk levels for five nappe rice varieties for Zones III and V (Evaluation criterion kg/ha).
5.6 Definition of Final Recommendation Domains
Based on the risk analysis and the environmental characteristics, the tentative recommendation domains presented in Table 5.5 become the final recommendation domains for the researcher' s criterion kg/ha.
5.7 Comparina Results Using Farmers' Evaluation Criteria
Most improved nappe rice varieties developed by
research are often rejected by farmers even though these varieties can produce better than the local varieties. The technology to be accepted by farmers should be technically feasible, simple and socially conform to the norms and beliefs of the social system. Thus, the technology should respond to farmers' own criteria which depend generally on resource scarcity and the product that farmers want to

maximize. In Lower Casamance, seeds are scare, manual harvesting is the only method used by farmers.
Farmers in Lower Casamance harvest rice fields manually with knives. Rice is cut panicle by panicles and then panicles are tied together to form sheaves of 3-5 kg, before storage in the granaries. This process is a hard physical process particulary when farmers have to bend all day to harvest short stature rice varieties. Also because of the deficit of rain, farmers' local varieties characterized by their long cycle (more than 120 days) became less and less adapted to the new agroclimatic conditions.
In this case, we will use three farmers' evaluation criteria to better meet their expectations. The three criteria are: (1) height of plants (cm), (2) plant cycle (days), and (3) number of tillers/m. Data for the different varieties in response to the farmers' three criteria are presented in Tables 5.7, 5.8 and 5.9. Figures 5.9, 5.10 and
5.11 present relationships of all five varieties with the environmental index (EI) using, respectively the criteria: height of plants, plant cycle and tillers/m2.
From figure 5.9, we can see that farmers' local
variety, which grows to about one meter in height, would be prefered across all environments. Among the improved varieties IRAT133 approaches the height of the locals in the best environments. If the farmers' selection criterion is

plant height, all four improved varieties would be rejected. This is one example of the reasons why farmers do not adopt improved varieties even if they produce better than the local varieties.
Table 5.7: Response to environment of four improved nappe rice varieties (height of plants in cm) and farmers' local variety for on-farm nappe rice variety trials (1982-1985). Lower Casamance. Senegal.
Farmer/year Height of plants (cm) sorted by El
DJ12519 IRAT112 IRAT133 IKP 10CAL EI 6/84 90 89 98 94 138 3719
9/84 91 88 95 85 83 3410
8/83 76 89 84 78 79 3087
13/84 70 78 70 78 99 3049
4/82 90 108 99 94 94 2506
5/83 79 76 96 77 72 2406
7/82 74 68 69 71 71 1927
11/85 62 72 64 72 109 1783
10/84 65 74 68 70 110 1632
2/83 80 89 78 74 103 1238
12/84 68 55 61 62 103 1099
1/82 79 85 78 79 118 956
3/84 82 87 77 76 110 688
Col. AVG 77 81 80 78 99
For the second farmers evaluation criterion (plant
cycle), estimated responses presented in Figure 5.10 shows that farmers' local varieties and variety IKP have the

largest plant cycles. Due to the drought, farmers are looking for shorter cycle varieties better adapted to the new agroclimatic conditions. Table 5.8: Response of five nappe rice varieties (plant cycle ) for on-farm nappe rice variety trials (1982-1985). Lower Casamance, Senegal.
Farmer/year Plant cycle (days) sorted by EI
DJ12519 IRAT112 IRAT133 IKP lOCAL EI 6/84 119 101 119 117 108 3719
9/84 106 91 100 116 119 3410
8/83 108 97 102 115 118 3087
13/84 105 91 95 100 110 3049
4/82 108 100 100 108 110 2506
5/83 106 88 101 ill ill 2406
7/82 106 98 105 107 105 1927
11/85 105 94 100 110 105 1783
10/84 105 93 110 114 110 1632
2/83 105 92 105 101 112 1238
12/84 110 110 109 107 115 1099
1/82 105 97 108 115 118 956
3/84 112 107 109 105 112 688
Col. AVG 108 97 105 10 112
Farmers' local varieties is more stable in poor and good environment due to their genetic stability and their rusticity. Improved varieties in general need a minimum input to express their minimum yielding in poor environments. Variety IRAT112 has the shortest cycle but

because of its yielding capacity in poor environments and its precocity (it is too early and would require labor for bird scaring), it would not be recommended in farmers' conditions. The best variety to recommend would be variety DJ12519 with an average cycle between 105 and 110 days with good yielding capacity in both poor and good environments. Table 5.9: Response of four improved nappe rice varieties (Tillers/M2) and farmers' local variety for on-farm nappe rice variety trials (1982-1985). Lower Casamance, Senegal. Farmer/year Tillers/2 sorted by EI
DJ12519 IRAT112 IRAT133 IKP lOCAL EI 6/84 254 231 221 267 217 3719
9/84 270 183 202 352 248 3410
8/83 238 199 169 307 311 3087
13/84 289 207 219 248 337 3049
4/82 267 230 248 289 273 2506
5/83 193 132 144 198 212 2406
7/82 200 177 156 235 232 1927
11/85 62 122 100 162 109 1783
10/84 70 135 74 165 120 1632
2/83 263 120 153 285 148 1238
12/84 133 110 106 198 153 1099
1/82 172 166 131 229 134 956
3/84 167 147 154 177 157 688
Col. AVG 198 166 160 239 203
For the third farmers' evaluation criterion
(tillers/m2), responses of the different varieties to EI presented in Figure 5.11, shows variety IKP would be

recommended because of its highest tillering capacity (187300 tillers/m2 ), followed by variety DJ12519 (138-290 tillers/m2 ). Farmers in Lower Casamance like improved varieties with good tillering capacity because they plant fewer seeds than recommended and depend on tillers to increase harvest. Thus they prefer rice varieties having good tillering capacity.
1W. -DJ-12-519
- g IRAT-112
0 IRAT-133
500 000 I500 2000 2500 3000 3500 4000 ENVIRONMENTAL INDEX (El)
Figure 5.9: Estimated responses (height of plants) of five nappe rice varieties to environmental index (EI).
-110 DJ-12-519
105 IRAT-112
o IRAT-133
so 1 i 5 2000 a 250oo 30o 35W 4000o ENVIRONMENTAL INDEX (El)
Figure 5.10: Estimated responses (plant cycle) of the five nappe rice varieties to environmental index (EI).

MP RCEVAMTRW9L. 198271"5
290- - - - - -12-519
- -- IRAT-112
F -
500 10o0 15 2999 2500 30 39D9 4090
Figure 5.11: Estimated responses (tillers/M2) of the five nappe rice varieties to environment index (EI).
5.8 Definition of Tentative Recommendation Domains Using
Farmers' Evaluation Criteria
For the first farmers' criterion: height of plants
(cm), farmers' local varieties would be recommended. For the second criterion: plant cycle (days), variety DJ12519 would be recommended because of its intermediate cycle and its yielding capacity in poor and good environments. For the third farmers' evaluation criterion: number of tillers/M2, variety IKP would be recommended, followed by farmers' local variety and possibly DJ12519. Variety IKP, because of its cycle (longer than cycle of variety DJ12519), and its height, (< 87 cm), may be rejected by farmers.
5.9 Determining Risk Associated with the Different Varieties Using Farmers' Criteria
Results of risk analysis, comprised environmental
characteristics are presented in Figures 5.12, 5.13, for the first farmers' criterion (height of plants). From Figure

5.12, which represents the high-yielding environments (Zone IV and Station), variety IRAT133 would be recommended if we consider a probability less than 15%. Over 15%, farmers' local variety and IRAT112 would be recommended. In lowyielding environments (Zones III and V), farmers' local would be the best to recommend.
For the second criterion (plant cycle), results
analysis are presented in Figures 5.14, and 5.15 for the high-yielding and low-yielding environments respectively. In high-yielding environments (Zone IV and Station), varieties IRAT133 and DJ12519 would be recommended. Variety IRAT112 might not be recommended because of its early maturity and its needs of labor for bird scaring. In low-yielding environments (Zones III and V), varieties IR{AT133 and DJ12519 would be the best to recommend. For the farmers'
third criterion (tillers/n2), results of risk analysis are presented in Figure 5.16 for the high-yielding environments and Figure in 5.17 for the low-yielding environments. Variety IKP would be the best to recommend, followed by farmers' local variety and DJ12519 for Zones IV and Station. If one considers farmers' local as a selected criterion, only variety IKP would be recommended. In low-yielding environments (zones III and V), variety IKP still remains the best to recommend over time, followed by farmers' local varieties.

' DJ12519
P. 7----0 IRAT133
S 5 10 15I 20 25 RISK (% time below Y- axis value)
Figure 5.12: Risk levels for five nappe rice varieties for Zone IV and Station,(Evaluation criterion height of plants).
em = = : : .0DJ12519
75 IRAT112
0 IRAT133
5 t0 15 20 25
RISK (% time below Y- aids value)
Figure 5.13: Risk levels for five nappe rice varieties for Zones III and V,(Evaluation criterion height of plants).
- -- DJ12519
0 5 10 15 20 25 RISK (% below Y- axis value)
Figure 5.14: Risk levels for five nappe rice varieties for Zone IV and Station, (Evaluation criterion plant cycle).

u IRAT112
e IRAT133
0 s 10 15 20 2 RISK (% time below Y- adxis value)
Figure 5.15: Risk levels for five nappe rice varieties for Zone III and V, (Evaluation criterion plant cycle).
2oo- rDJ12519 IRAT112
0 *
0 5 10 15 20 25 RISK (% time below Y. axis value)
Figure 5.16: Risk levels for five nappe rice varieties for Zone IV and Station, (Evaluation criterion tillers/m2)
200- - - --- ----- 9
. ..... IRAT112
-- . n IRAT133
S0 is 20 25
RISK (% time below Y- as value)
Figure 5.17: Risk levels for five nappe rice varieties for Zones III and V, (Evaluation criterion tillers/m2)

5.10 Define Final Recommendation Domains Using Farmers' Evaluation Criteria
The different recommendations for each recommendation domain based on farmers' s three criteria are summarized in Table 5.10.
Table 5.10: Summary of the recommendation domains and the technologies recommended for nappe rice variety trials (1982-1985). Lower Casamance, Senegal.
Environmental Recommendations
Evaluation criteria
(Zones) Height of Plant cycle Tillers/2
plants (cm) (days)
IV and Station IRATI33, IRAT133 and IKP, Local, DJ12519
Local, IRATI12 DJ12519
III and V Local IRAT133 and IKP, Local, DJ12519
5.11 Extension Recommendation for each of the Recommendation Domains
The recommendations and the resulting analysis based on researcher' s criterion (kg/ha) and the farmers' three evaluation criteria are summarized in Table 5.11.

Table 5.11: Multiple recommendations and the recommended technologies for nappe rice variety trials (1982-1985), based on environmental characteristics and four evaluation criteria. Lower Casamance, Senegal.
characteristics Evaluation criteria
(Zones) Researcher Height Plant Number of
(kg/ha) of plants cycle tillers/n2
(cm) (days)
IV and Station DJ12519, IRAT133, IRAT133 IKP,Local,Dj12519 IKP Local, and
IRAT112 DJ12519
III and V IKP, Local, Local IRAT133, IKP,Local,DJ12519
DJ12519 DJ12519

6.1 Conclusions
Agricultural research and extension services in Senegal very often lack of methodological tools adapted to the analysis of agro-socioeconomic constraints of small livelihood farms, and on-farm research. The traditional methods used have not performed well for technology producing research. Farming Systems Research and Extension (FSR/E) requires the collection of a wide range of data in order to understand the many factors that, taken together as a system, underlie and determine the production of crops and animals. More traditional research concentrates on individual commodities or aspects of production, and experiments and trials are conducted primarily or completely on research stations under highly controlled conditions. As development programs in Senegal, and particulary in Lower Casamance, have shifted from the traditional audiences ("Big producers") to smaller family farmers, often people who produce for food rather than for marketing, the traditional type of research analysis has not proven very successful. Farming systems research combines aspects of both research 83

and extension and serves to orient research toward the immediate concerns of farmers while it also bridges the gap between the pool of knowledge from traditional research and the extension of new information and practices onto farmers, fields.
The comparison of socioeconomic results analysis previously made in 1984, of a representative household chosen in zone I, using the FARPJ/BRADS method with a linear programming analysis (LP), shows the following similarities and differences:
1) Linear programming simulated very well the limitedresource farm in Lower Casamance. The traditional Diola system does not use animal traction, and has limited land compared to zones IV and V, where the use of animal traction is intensive.
The FARMAP/BRADS method does not allow prediction of how improved technologies fit into the households prior their introduction at the farm level. On the other hand, Linear programming helps to predict responses of family livelihood systems to improved technologies and to give valuable feedback to researchers, extentionists and other partners of the rural development. The example of alternative technology used (basket making) in the household in Zone I, showed how well the proposed technology, based on the optimization procedures that allocates scarce resources