Modelling cropping plan strategies: What decision margin for farmers in Burkina Faso?

Agricultural Systems 167 (2018) 17–33

Contents lists available at ScienceDirect

Agricultural Systems journal homepage: www.elsevier.com/locate/agsy

Modelling cropping plan strategies: What decision margin for farmers in Burkina Faso?

T

⁎

Jahel C. , Augusseau X., Lo Seen D. UMR TETIS, CIRAD, TA C-91 / MTD, 500 Rue J.F. Breton, Montpellier Cedex 5 34093, France

A R T I C LE I N FO

A B S T R A C T

Keywords: Cropping plan strategies Decision influencing factors Spatial modelling Burkina Faso

In a context of strong climatic and economic uncertainty in West Africa, agricultural statistics reveal interannual variations in the proportions of crops. The underlying causes of these variations remain, however, poorly documented although their understanding is essential for crop production monitoring and agricultural policyrelated decision-support. Regional scale cropping plan fluctuations arise directly from multiple individual farmer decisions. The purpose of this study is to understand the decision-making processes involved when farmers choose their cropping plans, in order to assess the respective weight of the different factors underlying the observed fluctuations in cropped areas. The study zone is the Tuy Province, occupying around 6000 km2 in central-western Burkina Faso. An initial field work showed how farmers' decision-making processes depend on external factors. This led to a separation and prioritization of the decisions taken in response to the physiological needs of the family (primary objectives) from those taken in response to other needs (secondary objectives). Four decision-influencing external factors were identified: i) climate, ii) price of cash crops, iii) incentive measures and dissemination processes and iv) credits for inputs. Decision-making rules were then determined by combining the objectives with the external factors. A decision model built on these rules was applied to Tuy Province between 2002 and 2014 to simulate every year the decision-making process of each farmer depending on several influencing factors. The model was verified by annually comparing the proportions of each crop grown in each cultivated area with those recorded in the agricultural statistics. The annual weight of each of these factors was then assessed: over the study period, 55% of the cropping plans satisfied unavoidable primary needs (the factors involved being internal determinants and credit), and 45% concerned secondary objectives (influenced by prices and promotion drives). With this approach, we evaluated the weight of the price factor to be only 6%. This result did not tally with the literature where the price factor is seen as a major element influencing cropping plan decisions. It is then discussed and considered in the specific context of this study. Sixty percent of the areas planted in cotton were linked to the access to credit granted by the cotton company in the zone, and tallied with the primary objectives of the farmers. The farms were therefore fully dependent on the cotton company. This study also illustrates the merits of modelling to assess how the respective weights of factors change over time, and to provide some major methodological perspectives for using spatial models to strengthen and validate typologies and processes arising from field analysis.

1. Introduction Farmers in West Africa have to cope with climatic, economic and political contingencies, limited access to credit and markets, and little backing from public policies (Gafsi et al., 2007). While, at first glance, this lack of means and opportunities combined with all these constraints might suggest that cropping plans will hardly vary from one year to the next, agricultural statistics seem to show the opposite (FAOSTAT, 2017). Understanding the reasons of these fluctuations is essential for the monitoring and control of the agricultural production

⁎

of a region and for agricultural policy-related decision-support (Edwards-Jones, 2006). These cropping plan dynamics, which can be seen on a regional scale, arise directly from the decisions taken individually by farm managers. The purpose of this article is to explore cropping plan dynamics on a regional scale, through an analysis of farmers' decision-making processes and the decisional factors involved, in order to understand what decisional factors lay behind the observed fluctuations in cropped areas. Although family farms in Burkina are characterized by their great diversity, they have one point in common in producing goods and

Corresponding author. E-mail addresses: [email protected] (C. Jahel), [email protected] (X. Augusseau), [email protected] (D. Lo Seen).

https://doi.org/10.1016/j.agsy.2018.08.004 Received 25 January 2018; Received in revised form 12 July 2018; Accepted 5 August 2018 0308-521X/ © 2018 Elsevier Ltd. All rights reserved.

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

decision-making rules depending on the influencing factors. The influencing factors are the means or information at the farmers' disposal for making their choice. They may be internal to the farm, such as the amount of land available or the labour that can be called upon; or external such as prices or the climate (Wood et al., 2014). Decisionmaking rules (Sebillotte and Soler, 1990; Aubry et al., 1998b; Mérot et al., 2008; Schaller, 2011) are rules drawn up by the farmer when making a decision. For example, one rule might be: “if the season is late, cotton areas are reduced and sesame areas are increased”. Our study uses this “influencing factors/decision-making rules” approach to assess the weight of each of the factors influencing the ultimate cropping plans. The study zone is the Tuy Province in western Burkina Faso. Over the last 15 years, the crops in that province have seen some major fluctuations, with maize for example increasing from 25% to 40% of the total areas cropped. Such variations are particularly visible from one year to the next, the most striking example being the reduction in cotton proportions by a half between 2006 and 2007. The first section of the article presents the data used and the method developed. The preliminary results are then described, namely identification of the objectives, influencing factors and decision-making rules for the different types of farming systems in the study zone. Based on those initial results, a model was constructed, calibrated and validated and was used to explore the impact of the different influencing factors on the cropping plans. A final section puts into perspective this impact with the degree to which farmers are free to make their decisions, given their integrated contractual relationship with the cotton company in the zone. This final section comments on the original approach taken, prioritizing farmers' objectives, along with the importance of modelling for strengthening field results.

services in the aim of creating wealth and satisfying the family's needs (Byerlee and Collinson, 1980; Gafsi et al., 2007). These two main objectives are accompanied by a raft of other more precise objectives, varying from one farm to another depending on the motivations, abilities and experiences of the farm manager, on the stage in the life cycle of the farm, or the particular needs of the family. For example, it may be a matter of developing the farm, or preparing for it to be taken over by an heir, etc. (Gafsi et al., 2007). Achieving these objectives, by producing goods, means combining the internal resources of the farm, which are often limited (labour, capital, land), with the information farmers have about external opportunities and constraints (Dillon, 1980; Jean-Pierre and Bernard, 1993). Over the length of a farming year, a farm is managed in three stages: planning, implementing and monitoring (Gafsi et al., 2007; Kay and Edwards, 1999). Planning consists in knowing what to grow, in what amount and in which way, in relation to the specific objectives of the farm. The achievement of that planning means bringing into play the means at the farm's disposal and making operational decisions (Duru et al., 1988). Work progress is monitored regularly in order to take decisions to redress any drift. At any given moment, these three stages require decisions to be taken to address waves of opportunities and constraints that occur (Brossier et al., 1997; Gafsi et al., 2007). Kay and Edwards (1999) described decision-making processes as a permanent mechanism that consists in making resource allocation choices. Decisions can be of three types: strategic decisions, tactical decisions and routine decisions (Gafsi et al., 2007). Strategic decisions concern long-term decisions focusing on the major orientations of the farm, such as investments in equipment to be used over a number of years. Tactical decisions establish the major lines of agricultural operations over a season and are taken in the planning phase. They include the cropping plan choices, technical decisions, choice of product use, etc. Lastly, routine decisions are taken on a day-to-day basis and consist in implementing and adapting the chosen techniques to everyday events and occurrences. In this article, we primarily focus on tactical decisions concerning the choice of cropping plan, which is largely made during the planning phase and may be readjusted during the monitoring phase. Kay and Edwards (1999) divided this process into three stages, which we have simplified for the choice of cropping plan: i) identify the objectives to be fulfilled, ii) recap the available sources, i.e. labour, capital and land, along with the external influencing factors and iii) allocate resources to the different crops. The study of these three stages in the choice of cropping plan is often complicated since it means taking into account a complex combination of factors which evolve over time and to which not all farmers will react in the same way. Resorting to modelling is a recognized way of working on these complex issues, since it makes it possible to represent a complicated reality in a simplified manner in the aim of understanding it better. For instance, a great deal of work has contributed to the modelling of these decision-making processes and their impacts on farm efficiency. Whilst most of that work has concerned technical strategies and how farming operations are conducted (Dury et al., 2011; Aubry et al., 1998a; Cerf and Sébillotte, 1997; Aubry, 2007; Dounias et al., 2002; Schaller, 2011; Jain et al., 2015), only a small part of it has concerned the processes involved in the choice of cropping plan and how external factors affect those choices (Houet, 2006; Deressa et al., 2009; Seo and Mendelsohn, 2008; Robert et al., 2016; Robert et al., 2017). Few studies have scrutinized the combinations of factors leading to cropping plan choices, using models retrospectively to address this issue. Nevertheless, we adopted the approach commonly taken when modelling decision-making processes, which consists in constructing

2. Material and method 2.1. Description of the study site The study zone is the Tuy Province, occupying around 6000 km2, located in central-western Burkina Faso. The climate is of the Sudanian type. The zone is crossed by a line of hills not exceeding 400 m in elevation, separating two broad plains of ferruginous soils (Fig. 1). Twenty percent of the territory has been protected forest since 1990, the remainder being primarily devoted to crops, of which the main ones are cotton, maize and sorghum. Table 1 shows the main crops grown in the zone and their use. A typology of farming systems in the zone was taken from literature (Vall et al., 2017; Marre-Cast and Vall, 2013). It distinguishes between crop farmers (less than 10 cattle), crop-livestock farmers (more than 10 cattle and more than 6.5 ha) and livestock farmers (less than 6.5 ha). Crop farmers make up 80% of the farms in the region, crop-livestock farmers 15%, and livestock farmers, 5% (Jahel et al., 2017).These three types of farming system will then have different strategies when choosing cropping plans. 2.2. Input data Two types of data were used: those used as input data for the model, and those used to calibrate and validate the model. The model input data were price trends for the main crops, rainfall and soil texture. The rainfall data came from a TAMSAT time series from 2001 to 2015, with a spatial resolution of 4 km and a daily time step. Soil texture came from soil maps drawn up by the Institut Géographique du Burkina (IGB). Prices were taken from the paper archives of the Agriculture Services and concerned the annual mean of prices paid to farmers in Tuy Province. Appendix 2 presents all the data used in this study.

18

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Fig. 1. Location map of the study area in Burkina Faso.

more or less constant price from one year to the next. It is interesting to note that changes in the proportions of crops did not mirror the changes in prices, even though some tendencies could occasionally be comparable (e.g. the increase in cotton price, in 2011, with that of the cotton volumes). Whilst it is accepted that prices have a substantial influence on cropping plan management (Pierye et al., 2009), it seems that considering that factor alone is not enough to explain the fluctuations in crop proportions.

Table 1 Recap of the characteristics of the main crops. Crops

Uses (listed in order of importance)

Maize Cotton Sorghum Millet Groundnut Cowpea Sesame

Food, sale Sale Food, drink, sale Food, sale Food, sale Food, sale Sale, food

2.3. The modelling platform used: Ocelet The study used an existing model (Jahel et al., 2017), constructed on the Ocelet modelling platform (Degenne and Lo Seen, 2016). It only used, and further developed, part of the model that concerns the processes of cropping plan creation. Ocelet is a modelling language and environment for simulating spatial dynamics. The modelling is based on interaction graphs. The graphs are made up of entities characterized by properties, and linked to each other by different relations. Once the entities and their relations have been described, scenarios are executed and cause the entities to evolve in line with their relations. We chose the Ocelet platform as it is adapted to spatial simulations over large areas. The model used in this study involved two types of entities: the “plot” entity and the “farm” entity. The plots were derived from a map of the plot structure in 2000 based on Landsat7 images (April and November 2000) and Spot5 images (January 2014). During the initialisation stage of the model, plots are linked to the farms they belong. The farms were of different types, the proportions between each of the types were based on the results of Marre-Cast and Vall (2013). The links between farm and plot were made according to farm types (Jahel et al., 2017). Each farm type has several plot characteristics (range of surface areas, distance between plots and distance of plots to the farm) that conform to the farm types defined in MarreCast and Vall (2013). At the end of 2001, the fields were linked to the ~7000 farms in the zone. The scenario consisted in allocating some crop species each year to the plots by simulating the decision-making process of each farmer depending on several influencing factors.

The data used to calibrate and validate the model were changes in the areas harvested for each crop obtained from AGRHYMET statistical surveys from 2001 to 2011 (Fig.2). The cotton data came from the cotton company in the zone. Two crop maps were drawn up for 2012 and 2013 in the Municipality of Koumbia from Pleiades satellite images. While the crops remained the same, their proportions underwent some major fluctuations over the study period (Fig.2). The variations were most marked for cotton, maize and sorghum, with the other crops remaining more or less stable – between 2 and 5% of the total areas. Whereas the proportion curves for maize displayed a relatively gradual increase, cotton and sorghum crops underwent some major fluctuations from one year to the next, the most striking example being the 2006–2007 period. The prices of the main crops also displayed major variations throughout this period: the prices of the three cereals (maize, sorghum, millet) evolved in a similar and relatively stable way (except for 2005 when some dry spells resulted in poor harvests and a price hike). The price of cotton saw limited variations, compared to the other two cash crops, soybean and sesame. Unlike for the latter two crops, cotton prices did not directly reflect international market prices. Indeed, they were fixed each year in advance by the three cotton companies in the country and depended on the trends of the previous three years and market trend forecasts. This mechanism guaranteed a 19

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Surfaces proportions of cultivated species

40% 35% 30% 25%

20% 15% 10% 5% 0%

2001

2002

Groundnut

2003 Maize

2004

2005

2006

Millet

2007

Sorghum

2008

2009

Cotton

2010

2011

Cowpea

Price (current CFA F/kg)

300 250

200 150 100 50 0 2000 Groundnut

2001

2002 Maize

2003

2004

2005

Millet

2006

2007

Sorghum

2008

2009

Coon

2010

2011 Cowpea

Fig. 2. Changes in proportions (top graph) (AGRHYMET) and prices (bottom graph) (Agriculture Services) for the main crops.

before. At the end of an interview, we had, for each field of the farm, the current cultivated crop and the reason of its choice for the current and previous years.

2.4. Method: from the field to the model The method was divided into two stages. Some interviews with farmers were first of all carried out in an attempt to understand their cropping plan strategies and identify influencing factors. Three types of surveys were carried out:

• Interviews with the agricultural supervisory services, farmers groups (one per village composed of village chiefs and some producers) and the cotton company of the zone to cross-reference the information gathered with some more global visions of the dynamics in the zone.

• Collective interviews took place in 18 villages (Appendix 0) to meet •

village chiefs, elders and members of the village committee. The purpose of these semi-directive interviews was to i) identify crop changes over the last 15 years and, ii) identify the factors responsible for those changes. Individual interviews were carried out with 50 farm managers (between 2 and 10 per village) in an attempt to understand the choice of cropping plans, in order to pinpoint key decision-making moments within the year and identify the nature of the different factors influencing the decision. The farmers interviewed were selected from those proposed by the village chiefs so as to cover the different types of farmers.

These interviews enabled us to establish some decision-making rules depending on the types of farmers. In fact, the rules followed the crop choice processes described by the farmers, and the decision-making thresholds (i.e. values used in the model to delimit decision rules) were fixed by statistical analysis of the data from the farms (the mean, maximum, minimum and variance of the surfaces of each crop). For example, if a farmer claimed to choose maize areas according to the people to be fed, that choice process was translated into a rule, and the “maize area/mouths to be fed” ratio stemmed from the survey data. The approach taken in this article was to prioritize the objectives to be fulfilled in order to establish the decision-making rules and give them an order of priority. The rules were then translated into the model, via the “field-farm” relation. That relation was divided into three interactions, occurring before each cropping season, and following the stages described by Kay and Edwards (1999):

During the interview, quantitative data about the farm were collected (number of animals, material, surface of each crop, evolution of cropping plan, etc). More qualitative data were also collected, as we asked farmers to explain their choices of crop allocation for each cultivated field. This question was asked for the current year and the years 20

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

i) Recap for each farm of the information available to it before choosing its crops: external factors (prices, links with the cotton company, climate, etc.) and internal factors (location of plots, plot history, number of labourers, etc.). In the model, the external factors were renewed each year and came from data external to the model, while the internal factors were derived from properties specific to each farm entity. ii) Choice of an “ideal” cropping plan depending on the factors. In this study we developed an approach based on a prioritization of farm objectives. For instance, the farmer chose as a priority the crops intended to fulfil the most important objectives, and gradually completed the cropping plan with crops intended to meet objectives of lesser priority. iii) Allocation of crops to plots so as to be as close as possible to this “ideal” cropping plan, but respecting the constraints linked to soil properties and plot location.

Table 2 Recap of the different influencing factors. Internal factors

External factors

1) Provide the quantities needed to feed the family. 2) Provide a varied diet for the family. 3) Have income staggered over time to satisfy secondary requirements for different periods of the year, in particular those occurring at the end of the rainy season, when the money from the main crops harvested has yet to be received. This is the case, for example, for schooling expenses incurred in September, which can only be covered with the income linked to early crops. 4) Have cash assets in order to share certain benefits between members of the family to maintain family cohesion. This objective will especially be achieved through cash crops. Cash assets also make it possible to acquire cattle or invest in infrastructure.

In the case of family farms in West Africa, which are still mostly managed by large families, it is often complicated to identify decisionmaking centres due to the existence of collective management, alongside individual management (e.g. collective fields, women's fields, children's field; Ancey, 1975), although the elders often hold a predominant position. For our study, we simplified this complexity of relations by imagining a single decision-making centre. We used collected data to fix thresholds in the decision rules. For example, the maximum proportion of maize in the cropping plan of crop farmers was 50% (in 2016). In the decision rules, we choose thresholds in order not to go beyond 50%. The decision-making model was calibrated using data gathered between 2002 and 2006 and validated with data from 2007 to 2013. Intermediate validation, called conceptual validation (Rykiel, 1996), was carried out to ensure the coherence of the decision-making rules incorporated into the model: the results of the field surveys were reported back to farmers who validated the different strategies proposed for each type of farmer. This was carried out in two villages where all the producers were invited. Around thirty farmers came, half of whom had not taken part in the surveys. The decision-making model validated in that way can be used to explore the different factors and decision-making processes guiding cropping plans. For each field to which a given crop has been allocated, the factor underlying that allocation was stored, thereby making it possible to estimate each year the surfaces linked to each factors for the Tuy province by summing all the surfaces of fields linked to the same factor (SurfF). The division of this sum by the total surface of cultivated area of Tuy province (SurfTot) gives the weight of the factors (WF) in play in the cropping plans.

WFn =

Number of labourers (influencing factor 1) Equipment (influencing factor 2) Capital (influencing factor 3) Land (influencing factor 4) Climate (influencing factor 5) Cash crop prices (influencing factor 6) Incentive measures and dissemination processes (influencing factor 7) Credits for inputs (influencing factor 8)

These four objectives can be divided into two categories: those meeting the physiological needs of the family (objectives 1 and 2) and those satisfying the other needs (objectives 3 and 4). We find similar classifications in the literature (Ancey, 1975; Gafsi et al., 2007; Thornton et al., 2007). In this study, we refer to them as primary objectives and secondary objectives. 3.2. Identification of the factors influencing the choice of cropping plan At the end of the interviews, 8 main influencing factors were identified by the farmers, broken down from factors internal to the farm and factors external to the farm (Table 2). The internal factors were the number of labourers that could be called upon, equipment, capital and land (plot areas and soil texture). According to the typology pre-established by Marre-Cast and Vall (2013), the internal factors are already known for each type of farm. They concern the structure of the farm (land, capital, equipment and number of labourers). The four external factors identified were specific to the study zone and required a more detailed description. These were the climatic conditions, cash crop prices, incentive measures amplified by the dissemination process and arrangements for access to inputs on credit.

∑ SurfFn SurfTot

• Climate: During the collective interviews, one of the first factors of

3. Preliminary results: understanding cropping plan strategies and constructing the model This section presents the identification of farm objectives in the zone, the different decision-making factors and the different decisionmaking rules.

3.1. Identification of objectives and requirements

•

The objective that was common to all the farms was to produce wealth and satisfy the needs of the family. Starting from these two main objectives, other more detailed objectives could be derived. The objectives were common to all the farms, but their importance differed depending on the farm types. Four main objectives were identified following the field surveys: 21

change mentioned by the farmers was “rainfall”, particularly late starts to the season, drought events or floods. The date of the first rainfall was decisive in the choice of cropping pattern, with a late season leading to a preference for short-cycle or late-sown crops (such as sesame). The most widely grown cotton variety is longcycle, which the farmers tended to replace with short-cycle maize for a late season (even though the cotton company supplied a late cotton variety, but lower yielding). Cash crop prices: Most of the farmers interviewed stated they only took into account the price of cash crops, namely cotton and sesame. For cotton, prices were announced before the season, so that the farmer could choose the areas to be sown without too much uncertainty. The sesame supply chain was less structured in the zone, and the prices were directly those of the international market. Whilst some stakeholders, running “projects”, might sometimes

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Fig. 3. Decision-making process depending on objectives.

In 2008, the introduction of a GM variety, Bt-cotton, gave new impetus to the company, promising a reduction in the associated input doses. In 2011, the company launched some new attractive strategies, such as a new, faster payment policy, access to credit to buy tractors via the “Union des Producteurs de Coton” and the introduction of a rebate. The success of the promotional campaigns was often enhanced by a dissemination phenomenon. Although the promotion was first carried out vertically (from the company to some of the producers), the information then circulated horizontally between the producers, with successful innovations thus being passed on via social links and cotton producers groups. In order to represent that dissemination process, we introduced a dissemination index from 0 to 1, being at 0 when there was no promotion in the previous year, and increasing by 0.2 if there was a promotion campaign in the previous year and it was still under way. For example, for the 4 years running of a promotional campaign taking place in years 2, 3 and 4, the indices for the respective years would be 0.0–0.0–0.20–0.40.

commit to purchasing harvests, prices were not regulated as for cotton, and the only information available to the farmer prior to the season was in reference to the prices of previous years. In order to simplify the choice of areas attributed to cash crops depending on their respective prices, we normalized prices to obtain a ratio between 0 and 1, 1 meaning the price was the highest of the period, 0.5 meaning that the price reached its average value over the period, and 0 meaning that it was the lowest. It was thus possible not to consider a price in terms of absolute value, but to place it on a scale of values. For cotton, farmers based themselves on the price declared for the current season, while they looked at the price of the previous year for sesame, in the absence of declared prices.

• Cotton company or “project “loyalty-building strategies: This section is divided into two sub-categories: i) incentive measures, and ii) credit for inputs. Credit for inputs could be included in the “incentive measures” categories, but we have preferred to address it separately given its importance in the choice of cropping plan.

ii) Credit for inputs, a contractual relationship between the cotton company and the cotton producer groups: each year, the cotton company in the zone concluded a contract with the farmer groups to encourage them to grow cotton. It provided fertilizers on credit to grow the cotton, and for some of the maize (for three hectares of cotton, the farmer could acquire fertilizers on credit for one hectare of maize). In order to have access to that contract, the farmer had to grow more than 3 ha of cotton and belong to a cotton producer group. The group enabled a system of joint surety between farmers when dealing with the cotton company in the event of outstanding payments.

i) Incentive measures: These concerned the strategies of the cotton company for increasing the areas sown by farmers with the targeted cash crop. For instance, several strategies were developed, such as purchase commitments, easier access to seeds, supervision and promotion in the field. The purchase commitments and easier access to seeds often came from “projects” run by different types of stakeholders, which consisted in distributing seeds at the start of the season, with a commitment to come back and buy the harvests at the end of the season. In recent years, there has been an increase in the number of such projects for sesame and soybean. The cotton company developed the same strategy by selling seeds on credit and committing to buying harvests. The promotional campaigns involved the presence of the company in the field to promote cotton, sending staff for supervision purposes, communication measures, etc. For cotton, the promotional campaigns for the study period were linked to how the cotton company was doing. For instance, the noteworthy dates indicated a strong promotional campaign between 2003 and 2006, after the “Union des Producteurs de Coton” purchased shares in the company in 2000 and the “Société des fibres et textiles” was split into three firms, leading to liberalization of the supply chain. Starting in 2006, the company entered an internal crisis linked to the drop in cotton prices on the international markets since 2004 (Kaminski et al., 2011). There then ensued cash-flow problems, late payments and a reduction in agricultural supervision.

3.3. Decision-making rules We divided decision-making rules into two stages: the first consisted in choosing the proportions of each crop in the cropping plan depending on the objectives and influencing factors identified above. The second consisted in spatially distributing these cropping plans by allocating the crops to the plots and taking into account the biological rules applying for each crop (soil texture needed, previous crop cover, fallow resting period, etc.). a. Choice of cropping plan depending on objectives and influencing factors During the interviews, we found that the decision-making rules declared by the farmers when choosing crop proportions were prioritized according to the importance of the objectives they satisfied,

22

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Table 3 Decision-making rules to fulfil objectives. Name

Decision-making rules

Influencing factors involved

Decision-making rules to fulfil objectives 1, 2 and 3: Objective 1: providing the quantities of food needed for the family Objective 2: providing a varied diet for the family and securing food production Objective 3: stagger income R1 Choice of maize areas: between 5% and 15% of the total area. If the farming system is of the crop-livestock farmer type 2, the proportions of cotton are double that of maize. For the other types of farms, the cotton area is between 0 and 2 times the maize area. R2 The farm has a 50% probability of allocating between 0 and 30% of its areas to groundnut. When this is not the case, it devotes between 0 and 25% of its areas to cowpea. R3 The farm has a 50% probability of allocating between 20 and 50% of its areas to sorghum or 20 and 50% to millet. Decision-making rules to fulfil objective 4: Objective 4: obtaining cash assets R4 If the cotton company promotes cotton and a project ensures the promotion of sesame, then the remaining areas will have a 50% probability of being allocated to one or other of the crops. R5 If the cotton company promotes cotton and there is no project promoting sesame, then:

R6

R7

- crop-livestock farmers will allocate all the remaining areas to cotton. - The other types of farming systems will choose their areas according to the threshold derived from the price and dissemination. The threshold determines the likelihood that they will sow half of the remaining areas with cotton and share the rest between sorghum and maize (probability = T); or not grow cotton whatever happens (probability = 1 T). If the cotton company does not promote cotton and there are projects promoting sesame, then: - if the normalized cotton price is over 0.5, the farmers will grow sesame on the remaining areas in 15% of cases, cotton in 45% of cases and maize in 30%. - if the normalized cotton price is under 0.5, the normalized sesame price determines the likelihood that the farmer will grow maize, sorghum or sesame on the remaining areas (seeAppendix 1). If the cotton company does not promote cotton and there are no projects promoting sesame, then the normalized cotton price determines the likelihood that the farmer will grow cotton, maize or sorghum on the remaining areas (seeAppendix 1).

Decision-making rules for crop allocation to plots R8 Proportions: between 0 and 20% of the area. Fallow as a priority in plots under fallow for less than 3 years, then in plots cropped for the longest period. R9 Priority rotations: - cotton-maize - cotton-legumes or maize-legumes - no rotation R10 No groundnut on sandy soils Decision-making rule to readjust cropping plans R11 If the season is late, for each of their cotton plots farmers will have a 33% probability of replacing cotton with maize, 33% probability of keeping cotton and 33% probability of replacing cotton with sesame.

• Decision-making

namely those concerning the primary needs of the family, and those satisfying secondary needs (see 3.1). This choice was therefore made in two chronological stages. The first i) concerned the proportions of compulsory crops to meet the requirements of the family in terms of quantity and quality (varied diet). Once the areas allocated to those basic needs had been achieved, the second stage ii) consisted in provisioning the remaining areas (if any) by choosing crops according to external opportunities, with a view to fulfilling objectives 3 and 4, namely staggering income and obtaining cash assets (Fig. 3).

Internal factors, credit (IF8)

Internal factors Internal factors Incentive measures for sesame, incentive measures for cotton (IF7). Incentive measures for cotton (IF7), price of cotton (IF6).

Cotton price (IF6), sesame price (IF6), incentive measures for sesame (IF7).

Cotton price (IF6)

Internal influencing factors

Climate (IF5).

rules for fulfilling objective 1, providing the quantities of food needed for the family: Between one to five crops were grown with a view to satisfying the family's food requirements. In general, this was particularly maize, whose proportions largely exceeded those of the other cereals. A small area was nonetheless often set aside for sorghum and sometimes millet, which is less demanding in inputs. Legumes such as groundnut and cowpea were minority crops, often in plots managed by women, and also providing a food supplement. The areas planted to cereals were usually directly linked to the number of people to feed (Youl et al., 2008). Most of the farms in the zone grew cotton. The main reason put forward by the farmers was linked to the contract concluded with the cotton company which guaranteed access to credit for inputs used on maize (influencing factor 8). Although cotton did not directly satisfy the family's food requirements, it was nonetheless considered as a crop able to fulfil the primary objectives as it was necessary for growing maize in order to provide the quantities of food needed for the family. This first objective was to fulfil needs in term of quantity; the next

i) Basic cropping plan to fulfil the objectives essential for the food security of the family The following rules took external influencing factors into account very little. They concerned the crops that were essential for the survival of the family and were therefore applied whatever the external context. The influencing factors taken into account, which were internal to the farm, determined the cropping capacities; they were therefore linked to the type of farming system.

23

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

based on the rainfall in 2015 and was evaluated empirically at less than 70 mm of cumulated rainfall on 13 June. The decision-making rule was as follows (Table 3).

one concerns the importance of having diversified food.

• Decision-making rules for fulfilling objective 2, providing a varied

diet for the family and securing food production: Two thirds of the farms interviewed had 4 or more crops in their cropping plan, of which at least two cereals. These diversification processes have been widely studied as a strategy for risk reduction, variations in food intake and farming calendar organization (Milleville, 1989). This diversification involves sowing groundnut, cowpea, millet and sorghum.

3.4. Different rules depending on farm type The decision-making process presented above is the general process for all the farms. However, as the different farming systems did not possess the same capital, or even the same equipment, or the same objectives, some variations in the proportions between crops were found between the farm types. Appendix 1 presents these variations. These differences primarily involved the supplementary part of the cropping plans, with the “primary needs” being generally common to all. The only visible difference in the “primary needs” part of the cropping plans concerned livestock farmers who, possessing a lot of manure, were less drawn than the others towards cotton in exchange for fertilizer credits. The other two largest differences between the farm types were with regard to the supplementary part of the cropping plan and the proportion of fallow land. Indeed, the large crop-livestock farmers had a tendency to allocate all the remaining areas to cotton if prices were good, while the other types of farms nonetheless made sure to diversify with maize, sesame or sorghum. As regards the fallow time, it varied depending on the types of farmers: the crop-livestock farmers and livestock farmers were generally able to compensate for the loss of fertility by applying mineral and organic fertilizers, while the crop farmers often lacked inputs and sometimes were unable to sow all their areas due to limited labour.

The decision-making rules (called R1, R2, etc.) for objectives 1 and 2 were therefore as follows (Table 3); the proportions for each of the crops were fixed in accordance with the cropping plan seen during the surveys: At the end of this first cropping plan, there generally remained some areas yet to be allocated. The farmer chose the additional crops in the next stage. ii) Additional crops to fulfil secondary objectives: satisfy the family's secondary needs and ensure the perpetuation of the farm

• Decision-making rules to fulfil objective 3, stagger income: This objective was fulfilled at the same time as objective 2, a varied diet. In fact, monetary income was staggered by harvesting and selling at different periods. The crops enabling money to be obtained rapidly, notably to cover schooling costs, were groundnut, cowpea and certain varieties of sorghum. The rules were therefore the same as for objective 2.

4. Modelling results: weight of factors in the cropping plans

• Decision-making rules for fulfilling objective 4, obtaining cash as-

4.1. Model structure: choice of cropping plans and plot allocation

sets:

The initialisation stage of the model was presented in paragraph 2.2. We present here the part of the model that was built for this study, that is, the annual loop describing the choice of cropping plans and of plot allocation. The rules revealed by the surveys were translated into the model according to the previously described algorithms, via four interactions, applied each year to each of the farms in the zone: the recap of information, the choice of cropping plans, the allocation to plots, and any readjustment depending on the climate data. Fig. 4 shows the structure of the model and the rules applied. The majority of the rules contain ranges of values for a given parameter, instead of having a fixed value. For example, the rule R2 “the farm has a 50% probability of allocating between 0 and 30% of its areas to groundnut” will be translated into the model in two steps: i) for each farm, a temporary variable will be created and its value will be randomly picked between 0 and 1. If the value is higher than 0.5, the farm will not grow groundnut, otherwise ii) the surface area of groundnut is randomly chosen between 0 and 30% of the total area of the farm following a uniform distribution.

Obtaining cash assets was mainly enabled by cash crops, the main two of which were sesame and cotton. The choice of areas to be sown for each of the crops was linked to prices, promotion drives, purchase commitments, and access to seeds for sesame (influencing factors 6, 7, 8). A threshold, T, was calculated to establish the rules, cumulating a normalized cotton price and the dissemination index for a given year, n. This was used to adjust the areas according to price and the dissemination process. An increase in cotton often went hand in hand with an increase in maize so as to be able to respect rotation requirements. The decision-making rules depending on these influencing factors were as follows for a given year (Table 3). b. Allocation of the different crops to plots and resting the land The interviews brought out three major biological constraints: the need for rotations, the need for a fallow period and taking into account soil texture (Table 5). As regards fallow, the proportions were between 0 and 20% and varied depending on the type of farming system (R8). Rotations were carried out as a priority between cotton and maize, then between all the other crops. While some plots could not be in rotation, it was nevertheless possible to grow a crop in the same plots several times in a row (R9). Lastly, the texture parameter particularly concerned the growing of groundnut, which can only be done on sandy or gravelly soils – a clay or loam soil prevents easy extraction of pods during harvesting (R10).

4.2. Model calibration and validation The calibration consisted in readjusting the thresholds in accordance with the proportions data for the main crops between 2002 and 2006. Very minor calibration was required, as the tendencies were very similar to the AGRHYMET data from the outset. Appendix 1 presents thresholds of the model with their sources of information and distinguishes the parameters that have been calibrated from those that were fixed. Once the model calibrated, we simulated a map of crops for each year between 2002 and 2014. We run the model several times to ensure

c. Readjustment of cropping plans according to the climate The threshold fixed to assess the late start to the rainy season was

24

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Fig. 4. Model structure (elements written in blue are the data used and “R1, R2, etc” are the decision rules (Table 3)). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Output for the year 2013: each plot of the Tuy Province has a crop. The white areas belong to the non-crop domain.

25

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Surfaces proporons of culvated species

40% 35% 30% 25%

20% 15% 10% 5%

0%

2001

2002

2003

2004

2005

2006

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Simulated proporons of culvated species

a) 40% 35% 30% 25% 20% 15% 10% 5% 0%

Maize

Coon

Sorghum

2007

Millet

2008

2009

2010

Groundnut

2011

Cowpea

Naonal stascs values (2006-2011)

b) 0.5 0.4 0.3 0.2 0.1 0 0

0.1

0.2

0.3

0.4

0.5

Simulated values (2006-2011)

c) Fig. 6. Comparison of the simulated proportions with those from agricultural statistics (AGRHYMET).

Table 5 Weight of the influencing factors in the choice of cropping plans for the whole study period (2002–2014).

Table 4 Comparison of the crop proportions from the satellite images and the model.

Images Model

Year

Groundnut/sesame/ cowpea

Cotton

maize_sorghum_millet

Fallow

Primary needs – Internal influencing factors (cereals and legumes) (IF 1,IF2, IF3, IF4)

43,9%

2012 2013 2012 2013

4% 19% 19% 21%

56% 18% 36% 16%

35% 58% 40% 58%

5% 6% 5% 5%

Primary needs – Credit (IF8) Secondary needs – Cotton promotion (IF 7) Secondary needs – Cotton promotion and cotton dissemination (IF 7) Secondary needs – Cotton price (IF 6) Secondary needs – Sesame promotion and sesame price (IF 6) Secondary needs – Sesame promotion (IF 7) Secondary needs – Internal influencing factors (cereals) (IF1, IF2, IF3, IF4)

9,3% 3,4% 4,5% 2% 5,3% 1,1% 30,4%

26

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2002

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Secondary needs - Sesame promoon and sesame price (IF 6, IF 7) Secondary needs – Internal influencing factors (cereals) (IF1, IF2, IF3, IF4) Secondary needs - Coon price (IF 6) Secondary needs - Coon promoon and coon disseminaon (IF 7) Secondary needs - Coon promoon (IF7) Secondary needs - Sesame promoon (IF 7) Primary needs - Credit (IF 8) Primary needs - Internal influencing factors (cereals and legumes) (IF1, IF2, IF3, IF4) Land resng Fig. 7. Changes in weight of the different influencing factors.

Unlike the AGRHYMET data, they also contained the values for sesame and fallow, so it could be seen (without having been calibrated beforehand for those crops) that the model was able to reproduce orders of magnitude close to those observed on the images for 2013. For 2012, the model underestimated the cotton proportions by 20% and overestimated the groundnut, sesame and cowpea proportions by 15%. That error may have been due to the fact that, while the surveys highlighted the arrival of several sesame projects in 2011 and 2012, the promotion drive undoubtedly had weaker effects than those modelled, with 2012 being a trial year for the farmers.

that the simulations were not significantly different from each other. For instance, the crop percentage difference between simulations was always below 1%. Fig. 5 shows the map simulated for the year 2013 and illustrates the type of output we obtain. For the rest of this article, we study the proportions of crops aggregated from these outputs to the whole province. The comparison with the data from 2006 to 2011 was used for validation (Fig. 6). Over that period, the simulated crop proportions were similar to those of the AGRHYMET data, with a mean error of 2% and a maximum error of 10%. The linear regression between simulated and crop statistics data between 2006 and 2011 gives an r2 of 0.93 (Fig. 6.c). Further validation was carried out using satellite data. Only crop groupings by class were visible on the images (Table 4) but they nonetheless made it possible to ensure that the proportions were right.

4.3. Analysis of the weight of the influencing factors and of the objectives a. Evaluation of the weight of the objectives and of the influencing factors, all crops combined

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

Secondary needs - Coon price Secondary needs - Coon promoon and coon disseminaon Secondary needs - Coon promoon Primary needs - Credit Fig. 8. Variation in the weight of objectives and influencing factors affecting the cotton area in the cropping plans.

27

Agricultural Systems 167 (2018) 17–33

maize surfaces (ha)

C. Jahel et al.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2002

2003

2004 2005 2006 Secondary needs

2007

2008 2009 2010 Primary needs

2011

2012

2013

2014

Fig. 9. Variation in the weight of objectives affecting the cereal area in the cropping plans.

“satisfying primary needs” corresponded to a constant area throughout the period. On the other hand, its weight as regards the total area of each of the crops varied (this weight is the area proportion of the cotton (or maize) whose choice has been influenced by an influencing factor compared to the total cotton (or maize) area of the Tuy province). On average, the proportion of cotton areas intended to meet the family's primary needs was greater than that of maize (Figs. 8 and 9). For instance, on average, 58% of the cotton areas satisfied primary needs, and that could range from 30 to 100% depending on the years. The rest was linked to supervision efforts and the dissemination process on the one hand, and on price attractiveness on the other hand. The years when prices were not attractive, and where the cotton company was not very present, the cotton areas were restricted to the primary needs threshold. On the other hand, for maize, only an average 39% of the areas were grown to meet primary needs, with that percentage varying between 20 and 60% (apart from 2012 when all the maize was grown for that purpose). The rest corresponded to the part of the cropping plan fulfilling the objectives of secondary needs. This score for maize was easily explained as it is central to the diet. However, the presence of cotton, a cash crop, in the cropping patterns fulfilling primary objectives was less obvious as regards its contribution to the primary needs of the family. Its presence was actually due to the fact that cotton growing was a prerequisite for access to credit for maize inputs.

In this section, we analyse the weight of the different influencing factors and objectives in the ultimate cropping plans. The weight of an influencing factor is the area proportion of the crops whose choice has been influenced by this influencing factor compared to the total cropped area of the Tuy province. To facilitate the analysis, each influencing factor was considered in relation to the objective in which it participated (but the objective is not taken into account in the calculation of the weight). We saw earlier that cereals and legumes were chosen according to influencing factors internal to the farm, not appearing in the analysis. The table giving the weights of the objectives and influencing factors (Table 5) indicates that between 2002 and 2014, 53% of the cropping plans satisfied unavoidable primary needs and 47% concerned secondary objectives. In the part of the cropping plans intended to meet secondary needs, only a third arose from attractive influencing factors linked to cash crops (price, promotion and dissemination). They were mainly linked to cotton growing. The rest comprised cereals (maize or sorghum) not reflecting any particular external influencing factor. According to the model, late rainfall played virtually no role in the choice of cropping plans. As regards the evolution of these factors and objectives over time (Fig. 7), the influencing factors correlated to the primary objectives (internal influencing factors and credit for inputs) had a relatively constant weight throughout the period, unlike the influencing factors involved in the secondary objectives. This explains the stability seen in the proportions for the minority crops such as cowpea, groundnut and rice. Another salient feature was how the influencing factors evolved: while the promotion and dissemination of cotton had a large impact between 2002 and 2006, some other influencing factors gained in weight from 2007 onwards. This was the case for cotton and sesame prices – independently from the promotion drives – which did not occur previously (or they were not high enough to cause a change in cropping plans). Two years, 2007 and 2012, stood out: 2007 was marked by an absence of influencing factors other than credit for inputs. The secondary objectives were therefore satisfied by cereals (only depending on internal factors). Conversely, in 2012, all the secondary objectives brought into play some influencing factors linked to cash crops. The comparison of these two years confirmed the fact that cereals were then used to complete the cropping plans when there was no external opportunity.

5. Discussion 5.1. Limitations of the field study One of the limitations of our study lay in the difficulty of identifying from surveys all the drivers of the complex dynamics observed in the field. While the study sought to be exhaustive in terms of the factors influencing the choice of cropping plan, it was difficult – during the surveys – to pinpoint and represent certain highly specific factors, notably regarding social influencing factors. One example was the social pressure surrounding cotton growing, in addition to the other factors influencing the choice of cotton. In fact, society seemed to be veritably constructed around cotton growing, with the practice becoming an identity factor, totally rooted in the organization of the villages. The cotton supply chain very quickly became organized around village groups then later around cotton producer groups (Schwartz, 1997; Tallet, 2007). These groups enabled a system of joint surety between farmers when dealing with the cotton company and rapidly became structuring within society.

b. Particular case of cotton and maize For cotton and maize, the main two crops grown, the aim of

28

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

influencing cropping plan decisions (Alene et al., 2008). Several elements can explain this apparent limited importance of the price factor in our study. Firstly, what we call “price factor” in this study is in fact a “worthwhile price” factor, as it only intervened when prices were high enough to influence the farmer's choice. When crop price was too low, the farmers did not increase their areas for that crop, so the price affected the decision-making but did not appear in the factors having played a role. However, some influencing factors with weight in the decisionmaking were indirectly linked to prices. For example, such was the case of cotton incentive factors, which reflected farmer supervision and promotion drives. In 2007, the cotton company had to halt that promotion as it was in financial difficulty due to the slump in international cotton prices. As domestic cotton prices were stabilized by national price regulating mechanisms, that drop in international prices was barely passed on in the purchasing price paid to producers, but lay behind the end of the promotion. Here again, the price factor did not appear directly in the influencing factors considered. Lastly, price stability over the study period was responsible for the fact that the farmers declared that they did not take prices into account when choosing their cropping plans. For example, cereal prices have varied very little over the last 15 years, so the farmers took those prices for granted and did not modify their strategy in line with them. But, if maize prices were to have trebled from one year to the next, it is likely that corresponding changes in the areas sown with maize would have ensued. This analysis of factors therefore needs to be placed in perspective with the context in which it was carried out, bearing in mind that the stability of certain factors made them invisible, while a break in that stability would certainly have brought them out. Although the price factor seemed to have limited weight in decisionmaking, this study was nonetheless unable to conclude that prices played a minor role in the areas sown. However, this study did enable some conclusions to be drawn about the prices directly accessible to the farmers, in a context where those prices remained stable around a range of values.

Social pressure was mentioned three times during the surveys, with one cotton grower telling us that if all his neighbours grew cotton, he had to do the same thing otherwise he would be marginalized. Two other farmers having recently decided to stop growing cotton told how they had been ostracized by society, not being invited to meetings. It was impossible to show these factors in the model, due to the difficulty of positioning them within the decision-making processes. 5.2. Modelling as back-up for field analyses The semi-open interviews were valuable tools for identifying the decision-making processes of the stakeholders in the field, but they had the drawback of sometimes being subjective and their limited number raised the question of their representativeness to cover the whole of Tuy Province. Translation into the model of the field dynamics observed and obtaining simulations close to reality after very limited calibration strengthened the credibility accorded to the analyses from the field surveys. The approach developed, based on prioritizing the decisionmaking rules by “objectives”, was also partially validated by the modelling results. This link between typologies derived from field studies and modelling opens up some important prospects, notably for the spatialization of typologies and their validation on a regional scale. The study undertaken showed that the typologies established during earlier work (Marre-Cast and Vall, 2013) were effectively representative of the field dynamics and were applicable throughout the province. In addition, while the field studies made it possible to identify the processes and factors underlying the dynamics observed, the quantification of their respective weights was more arduous. The study undertaken showed that once the model had been calibrated and validated, it was able to unravel the combinations between factors and achieve such a quantification. Nonetheless, one of the criticisms that could be made of the model lies in the simplification of the reality observed. Indeed, several factors influencing decision-making that were identified in the field were simplified, such as the cultural techniques adopted, the equipment available, the yields obtained, farmers' incomes, or off-farm activities and incomes, even though they were factors influencing the directions taken by the farm. The simplification was made by using farm types, with each type being characterized by a range of structuring influencing factors. For example, basing ourselves on the surveys and the literature, we considered that crop-livestock type farms had efficient equipment and enough capital to pay for additional labour and inputs. They were therefore not considered as being limited in their ability to sow all their areas. On the other hand, some farmers did not have the means required to cultivate all their land, leading them to leave some of it fallow. Indeed, the causal relationship was thus simplified by associating the choices of cropping plan with types of farmers and not directly with the factors involved (income, equipment, etc.). This type of generalization can rapidly lead to approximations that are likely to bias the model results. One of the actions taken to compensate for this effect was the use of probabilities. Indeed, translating the rules observed into probabilities made it possible to simulate all the diversity of the existing situations. This also made it possible to represent the diversity inherent to the same type of farming system. In fact, all choices were not made with a single mechanical logic of economic optimization but also depended on the personality of the person making the decision, their own constraints and opportunities, and individual external events (wedding, funeral, social network, etc.) (Singh et al., 2016).

5.4. What choices do farmers have when composing their cropping plans? The other conclusion that can be drawn from this ease of reproducing the dynamics of crop choices in the model, without a great deal of calibration work, came from the fact that the reality to be modelled was quite simple on the whole: the farmers in Tuy Province have little adjustability and few opportunities. Cropping plan choices are made under constraint, and the large number of those constraints is responsible for the fact that few possible crop combinations exist. The decision to prioritize decision-making rules by objective made it possible to estimate that more than half of the cropping plan was established with a view to satisfying the primary needs and food security of the family, and only 45% satisfied secondary needs, i.e. education, health or perpetuation of the farm. Taking an average per capita income of 69,0001 CFA F (FAO, 2013), i.e. an income of 897,000 CFA F generated by the farming systems in the zone (taking an average of 13 adults per farm), and considering – broadly speaking – that half of the secondary needs was reinjected into the farm, there remained, on average, 448,500 CFA F (i.e. 684 euros) to run the farm the following year (purchases of inputs, equipment, infrastructure, etc.). Under such conditions, where most of the harvests of the year are intended for the family's survival, it is very difficult for farmers to succeed in capitalizing, and the poverty spiral continues.

5.3. Minor influence of the price factor in decisions? At first glance, it seems that price played little role in the farmers' decision-making process, with an average weight of 6% over the study period. This result did not tally with the literature, which seemed to show that the price factor was, on the contrary, a major element

1 This value is an estimation made by FAO of the income per capita in rural areas in 2010, derived from the agricultural added value per capita (agricultural added value divided by the entire rural population). It does not take into account the monetary value of the on-farm consumption of products.

29

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

cotton company ensured minimum cotton production for itself, stable from one year to the next, in the absence of a rival creditor who could also offer credit for inputs.

To qualify this reasoning, it should be noted nonetheless that only the cropping systems were taken into account in the study. For an increasing number of farms with a livestock activity, the panel of choices is slightly larger once primary needs have been met, as it includes some strategic herd management choices. In addition, the study did not take into account multiple job holding, or off-farm earnings, which are often major elements in the functioning of farms in the zone. Most of the cropping plans seemed to be governed by constraints rather than opportunities. However, the revealed limited adjustability obviously does not mean that a single path was followed and the farmer did not have a strategy. On the contrary, the farmers needed to develop a raft of strategies to successfully cope in an unsuitable environment with few means at their disposal.

6. Conclusion This study showed that farmers in Tuy, who are subjected to numerous constraints and have few opportunities, have little adjustability when choosing their cropping plans. More than half of their areas were thus earmarked for meeting the primary needs of the family. Over the entire study period, external influencing factors linked to prices and the incentive strategies of the cotton company were responsible for less than a quarter of the cropping plans. It was this part of the cropping plans, linked to external factors, that gave rise to fluctuations in the proportions of the crops grown in the province from one year to the next. The cotton company in the zone nonetheless succeeded in maintaining a minimum cotton area, by making itself essential in meeting the primary needs of the families. Another aspect to be highlighted as a conclusion of this study concerns the novel use of modelling to address cropping plan issues. While there has been a great deal of work using modelling to analyse the complex reality of decision-making, the use made here of a model to assess the evolution over time of the respective weights of the factors of change is relatively innovative. This new approach enables analyses to be undertaken that it would have been difficult to carry out with field surveys alone, as interviews cannot be used to reliably obtain a quantification and prioritization of factors influencing decisions. In addition, the study provided some important methodological perspectives regarding the use of spatial models to strengthen and validate typologies and processes arising from field analyses. With spatial modelling it is possible to go beyond the limits of field typologies, namely a risk of low representativeness and subjectivity, by spatializing them and assessing their coherence on a regional scale. The ability of simulating crop surface dynamics on a regional scale thus offers a valuable tool for public policy decision-making, such as when advising on food security related issues.

5.5. The cotton company present in the subsistence objectives of the farmer The contractual relationship between the cotton company and the farms in the zone has been the subject of numerous debates. Indeed, the technical progress made since the arrival of the company, the new organization of the villages by cotton producer groups and the inarguable increase in living standards, at least in the first 10 years following the cotton boom around the 1980s, are an image factor favourable to this model developed around the cotton supply chain. However, the recent difficulties encountered by the supply chain, namely falling cotton prices on the international market, increase in input doses and costs (Bainville and Dufumier, 2009), gradual decrease of cotton yields (SOFITEX) and the difficulties recently encountered in introducing genetically modified cotton (Dowd-Uribe, 2014), are all valid arguments against this system. This study does not hold the keys for entering this debate, as no economic analysis was conducted that might help to estimate the economic importance of the cotton crop on the farms. Nevertheless, it provided a new angle of analysis, regarding these contractual relationships, by positioning them within the farmers' decision-making processes. The cotton company strategy for increasing the cotton areas grown by farmers was based on: providing inputs on credit in exchange for the areas planted to cotton, a commitment to buy harvests, the stabilization of prices and announcement of prices in advance, along with on-site supervision and communication. These influencing factors linked to the cotton company covered around 22% of all the other factors entering into the choice of cropping plans. Of all these conditions, the proposal of inputs on credit was alone responsible for almost 60% of the areas planted to cotton. The same decision-making factor was also found in the primary decision-making rules for food security, while all the company's other types of strategies were considered by the farmer in second place, to fulfil secondary objectives. This result had a strong implication for understanding the contractual relationships between the cotton company and the farmer: it means that at least half of the areas planted to cotton on the farms became essential for ensuring the family's food security, as they were the prerequisite for obtaining credit that gives access to maize fertilizers. These are key for producing maize in sufficient quantities. The farms were therefore wholly dependent on the cotton company. By making itself essential for satisfying the families' primary needs, the

Acknowledgment This study was supported by the SIGMA European Collaborative Project (603719) (FP7-ENV-2013 SIGMA -Stimulating Innovation for Global Monitoring of Agriculture and its Impact on the Environment in support of the GEOGLAM- project). The authors wish to thank the Meteorology Regional Direction of Burkina Faso and the AGRHYMET Centre in Niamey, the Centre International de Recherche-Devloppement sur l'Elevage en Zone Subhumide (CIRDES) in Bobo-Dioulasso, Stephane Dupuy, Jacques Imbernon and Raffaele Gaetano for providing information used in this study. The authors are very grateful to Elodie Maître d'Hotel (CIRAD) and Tristan Lecotty (CIRAD) for having given good advices on the question of prices in this article. The authors also thank Thomas Houet (CNRS) and Delphine Leenhardt (INRA) for their careful rereading and relevant remarks.

30

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Appendix 1. Parameters used in the model and their sources. The parameters orange and bold are those who were calibrated. The others are fixed.

NSP: normalized sesame price. NCP: normalized cotton price.

31

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Appendix 2. Data The next table presents the soil, rainfall and meteorological data used in the model.

Data type

Spatial resolution

Temporal resolution

Soil Rainfall

1 km2 30*30 km2

day

Meteorological data Price Plot structure

The province Province

day year

Source BDOT, 2007 Meteorology Regional Direction of Burkina Faso 18 rain gauges for 2014 and 2015 Experimental meteorological station of Farako Ba (INERA-CIRAD) Agriculture Services Spot5 image of January 2014

Appendix 3 The investigation work took place in 18 villages of the Tuy province: Banere, Boho-Bereba, Boho-Kari, Boni, Dabere, Dankari, Dimikuy, Dossi, Dougoumato, Founzan, Gombeledougou, Kongolekan, Koumbia, Mambo, Nahi, Pe, Saho and Waly (Fig. 1).

Fig. 1. Localisation of investigated villages (taken from Jahel et al., 2018).

l'exploitation agricole familiale. Eléments théoriques et méthodologiques, Nouvelle édition, 2002. Enesad- Cnerta, Dijon, France 215 p. Byerlee, D., Collinson, M.P., 1980. Planning technologies appropriate to farmers: Concepts and procedures. Cerf, M., Sébillotte, M., 1997. Approche cognitive des décisions de production dans l'exploitation agricole [Confrontation aux théories de la décision]. Économie rurale 239, 11–18. Degenne, P., Seen, D.L., 2016. Ocelet: Simulating processes of landscape changes using interaction graphs. SoftwareX 5, 89–95. Deressa, T.T., Hassan, R.M., Ringler, C., Alemu, T., Yesuf, M., 2009. Determinants of farmers' choice of adaptation methods to climate change in the Nile Basin of Ethiopia. Glob. Environ. Chang. 19 (2), 248–255. Dillon, J.L., 1980. The definition of farm management. J. Agric. Econ. 31 (2), 257–258. Dounias, I., Aubry, C., Capillon, A., 2002. Decision-making processes for crop management on African farms. Modelling from a case study of cotton crops in northern Cameroon. Agric. Syst. 73 (3), 233–260. Dowd-Uribe, B., 2014. Engineering yields and inequality? How institutions and agroecology shape Bt cotton outcomes in Burkina Faso. Geoforum 53, 161–171. Duru, M., Papy, F., Soler, L.G., 1988. Le concept de modèle général et l'analyse du fonctionnement de l'exploitation agricole. CR Acad. Agric. Fr. 74 (4), 81–93.

References Alene, A.D., Manyong, V.M., Omanya, G., Mignouna, H.D., Bokanga, M., Odhiambo, G., 2008. Smallholder market participation under transactions costs: Maize supply and fertilizer demand in Kenya. Food Policy 33 (4), 318–328. Ancey, G., 1975. Niveaux de décision et fonctions objectifs en milieu rural africain. Ed. Amira, Paris, France. Aubry, C., 2007. La gestion technique des exploitations agricoles. Composante de la théorie agronomique. Mémoire d'habilitation à diriger les recherches. Institut National Polytechnique De Toulouse, pp. 101. Aubry, C., Papy, F., Capillon, A., 1998a. Modelling decision-making processes for annual crop management. Agric. Syst. 56 (1), 45–65. Aubry, C., Biarnes, A., Maxime, F., Papy, F., 1998b. Modélisation de l'organization technique de la production dans l'exploitation agricole: la constitution de système de culture. Études et Recherches sur les Systèmes Agraires et le Développement 31, 25–43. Bainville, S., Dufumier, M., 2009. Diversité des Exploitations Agricoles en Zone Cotonnière du Burkina Faso. AFD. Brossier, J., Marshall, E., Chia, E., Petit, M., 1997. In: Éducagri (Ed.), Gestion de

32

Agricultural Systems 167 (2018) 17–33

C. Jahel et al.

Robert, M., Dury, J., Thomas, A., Therond, O., Sekhar, M., Badiger, S., Bergez, J.E., 2016. CMFDM: a methodology to guide the design of a conceptual model of farmers' decision-making processes. Agric. Syst. 148, 86–94. Robert, M., Thomas, A., Sekhar, M., Badiger, S., Ruiz, L., Raynal, H., Bergez, J.E., 2017. Adaptive and dynamic decision-making processes: a conceptual model of production systems on Indian farms. Agric. Syst. 157, 279–291. Rykiel, E.J., 1996. Testing ecological models: the meaning of validation. Ecol. Model. 90 (3), 229–244. Schaller, N., 2011. Modélisation des décisions d'assolement des agriculteurs et de l'organization spatiale des cultures dans les territoires de polyculture-élevage. Thèse de Doctorat. AgroParisTech. Schwartz, A., 1997. Des temps anciens à la edévaluation du franc CFA, les tribulations de la culture du coton au Burkina Faso. In: Annales de géographie. Armand Colin, pp. 288–312. Sebillotte, M., Soler, L.G., 1990. Les processus de décision des agriculteurs: acquis et questions vives. In: Brossier, J., Vissac, B., Lemoigne, J.L. (Eds.), Modélisation systémique et systèmes agraires. INRA, Paris, pp. 93–102. Seo, S.N., Mendelsohn, R., 2008. An analysis of crop choice: Adapting to climate change in South American farms. Ecol. Econ. 67 (1), 109–116. Singh, C., Dorward, P., Osbahr, H., 2016. Developing a holistic approach to the analysis of farmer decision-making: implications for adaptation policy and practice in developing countries. Land Use Policy 59, 329–343. Tallet, B., 2007. A l'arrière des fronts pionniers, recompositions territoriales dans l'Ouest du Burkina Faso et le Sud du Veracruz (Mexique). volume 3/3 HDR, Universite de Paris 1 Pantheon-Sorbonne. Thornton, P.K., Boone, R.B., Galvin, K.A., BurnSilver, S.B., Waithaka, M.M., Kuyiah, J., Herrero, M., 2007. Coping strategies in livestock-dependent households in east and southern Africa: a synthesis of four case studies. Hum. Ecol. 35 (4), 461–476. Vall, E., Marre-Cast, L., Joél Kamgang, H., 2017. Chemins d'intensification et durabilité des exploitations de polyculture élevage en Afrique subsaharienne: contribution de l'association agriculture-élevage. Cahier Agric. 26 (2). Wood, S.A., Jina, A.S., Jain, M., Kristjanson, P., DeFries, R.S., 2014. Smallholder farmer cropping decisions related to climate variability across multiple regions. Glob. Environ. Chang. 25, 163–172. Youl, S., Barbier, B., Moulin, C.H., Manlay, R., Botoni, E., Masse, D., Hien, V., Feller, C., 2008. Modélisation empirique des principaux déterminants socioéconomiques de la gestion des exploitations agricoles au Sud-Ouest du Burkina Faso. Biotechnol. Agron. Soc. Environ. 12 (1), 9–21.

Dury, J., Schaller, N., Garcia, F., Reynaud, A., Bergez, J.E., 2011. Models to support cropping plan and crop rotation decisions. A review. Agron. Sustain. Dev. 32 (2), 567–580. Edwards-Jones, G., 2006. Modelling farmer decision-making: concepts, progress and challenges. Anim. Sci. 82 (6), 783–790. FAO, 2013. Suivi des politiques agricoles et alimentaires en Afrique. Revue des politiques agricoles et alimentaires au Burkina Faso. Rome (Italie), pp. 225. FAOSTAT, 2017. Agricultural data, Food and Agricultural Organization of the United Nations [WWW Document]. URL. http://faostat.fao.org/site/291/default.aspx. Gafsi M., Dugué P., Jamin J.-Y., Brossier J., 2007. Exploitations agricoles familiales en Afrique de l'Ouest et du Centre. Editions Quae. Houet, T., 2006. Modélisation prospective de l'occupation du sol en zone agricole intensive dans la France de l'Ouest. Norois 1, 35–47. Jahel, C., Baron, C., Vall, E., Karambini, M., Castets, M., Coulibaly, K., Bégué, A., Lo, Seen D., 2017. Spatial modelling of agro-ecosystem dynamics across scales: a case in the cotton region of West-Burkina Faso. Agric. Syst. 157, 303–315. Jahel, C., Vall, E., Zermeno, A., Bégué, A., Baron, C., Augusseau, X., Lo Seen, D.L., 2018. Analysing plausible futures from past patterns of land change in West Burkina Faso. Land Use Policy 71, 60–74. Jain, M., Naeem, S., Orlove, B., Modi, V., DeFries, R.S., 2015. Understanding the causes and consequences of differential decision-making in adaptation research: adapting to a delayed monsoon onset in Gujarat, India. Glob. Environ. Chang. 31, 98–109. Jean-Pierre, Darré, Bernard, Hubert, 1993. I. Les raisons d'un éleveur sont notre raison de coopérer. In: Études rurales, n°131–132, pp. 109–115 Droit, politique, espace agraire au Brésil, sous la direction de Afrânio Garcia Jr. Kaminski, J., Headey, D., Bernard, T., 2011. The Burkinabé cotton story 1992–2007: sustainable success or sub-Saharan mirage? World Dev. 39 (8), 1460–1475. Kay, R., Edwards, W., 1999. Farm management, 4e édition. McGraw-Hill, Boston, ÉtatsUnis, pp. 464. Marre-Cast, L., Vall, E., 2013. Stratégies et trajectoires des exploitations de polycultureélevage de l'Ouest du Burkina-Faso. In: Séminaire DP Asap, Plateformes d'innovation et intensification écologique, Bobo-Dioulasso, Burkina Faso, 15 p. Mérot, A., Bergez, J.E., Capillon, A., Wéry, J., 2008. Analysing farming practices to develop a numerical, operational model of farmers' decision-making processes: an irrigated hay cropping system in France. Agric. Syst. 98, 108–118. Milleville, P., 1989. Activités agropastorales et aléa climatique en région sahélienne. In: ORSTOM. Le risque en agriculture, Paris, pp. 233–241. Pierye, W., Dabire, I., Barbier, B., Savadogo, K., 2009. Evaluation ex-ante de la prévision saisonnière en petit paysannat burkinabè. Atelier régional CIRAD-CIRDES, pp. 17.

33