Opportunities and constraints for farmers of west Africa to use seasonal precipitation forecasts with Burkina Faso as a case study

Agricultural Systems 74 (2002) 331–349 www.elsevier.com/locate/agsy

Opportunities and constraints for farmers of west Africa to use seasonal precipitation forecasts with Burkina Faso as a case study K.T. Ingrama,*, M.C. Roncolia, P.H. Kirshenb a

Department of Crop and Soil Sciences, The University of Georgia, 1109 Experiment Street, Griffin, GA 30223-1797, USA b Civil and Environmental Engineering Department and Fletcher School of Law and Policy, Anderson Hall, Tufts University, Boston, MA 02155, USA

Abstract Skill of seasonal precipitation forecasts for west Africa has improved to the point that forecasts may be of value to agricultural users, especially farmers. We studied agricultural production systems in three agro-ecozones of Burkina Faso to establish: (1) farmer interest in and ability to use forecasts; (2) forecast information farmers request; (3) lead-time required for greatest forecast value; (4) needs for forecast dissemination, interpretation, and application; and (5) possible strategies for using climate forecasts to improve crop production and resource management. The three agro-ecozones studied were a cotton-based system in the relatively high rainfall Sudan area of southwest Burkina Faso; a sorghum and millet based system in the low rainfall central plateau; and a cattle-based system in the very low rainfall Sahel area in the north. Potential value of forecasts to farmers differed among the three zones, with greatest apparent value to farmers of the central plateau and least apparent value to cattle herders of the Sahel. While farmers in all three agro-ecozones expressed a strong interest in receiving seasonal precipitation forecasts, they were much more interested in receiving forecasts of when the rains would start and end, and whether there would be interruptions in rains. Our results suggest that if seasonal precipitation forecasts are disseminated, they should be a part of an extension package that includes discussion of the probabilistic nature of the forecasts, potential response strategies, and risk management. Furthermore, farmers may need greater access to basic agricultural technologies, such as plows, new crop varieties, and fertilizers, before they can benefit fully from precipitation forecasts. # 2002 Elsevier Science Ltd. All rights reserved. Keywords: Climate change; Climate variability; Drought; Flood; Knowledge systems; Livelihood; Rainfed agriculture; Resource management; Risk management; Sustainable agriculture

* Corresponding author. E-mail address: kingram@griffin.peachnet.edu (K.T. Ingram). 0308-521X/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved. PII: S0308-521X(02)00044-6

332

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

1. Introduction Seasonal rainfall predictions are based on the influences of sea surface temperatures (SSTs) on global atmospheric circulation and regional precipitation patterns. Boundary conditions for SSTs may persist for months, making climate forecasting in some tropical regions feasible on a similar time scale (Hulme et al., 1992a). This relationship has led to a search for regions of the world where such an influence might enable seasonal climate forecasting (Palmer and Anderson, 1994). In general, SSTs have a stronger influence on atmospheric conditions in the tropics than outside the tropics, which has allowed seasonal forecasts to be developed for tropical regions of northeast Brazil, South Africa, west Africa, India, and others (Hulme et al., 1992a; Mason et al., 1996). Applications of seasonal forecasts include prediction of human diseases (Linthicum et al., 1999), forecasting crop yields (Mjelde and Keplinger, 1998), and improved management of crops (Mozzocco et al., 1992) and fisheries (Costello et al., 1998). This paper addresses the application of seasonal precipitation forecasts to the management of rainfed agricultural systems in the Sahel–Sudan of west Africa (Fig. 1).

Fig. 1. Map of Africa showing the Sahel–Sudan.

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

333

1.1. Agroclimate of the Sahel–Sudan The Sahel–Sudan climatic zone is characterized by a strong gradient of decreasing annual rainfall from south to north. Rains fall during a single wet season consisting of short intense storms over a 3- to 5-month period, with about 90% of the rains falling during the months of July, August, and September. Total seasonal rainfall ranges from 100 to 650 mm in the semi-arid Sahel in the north of Burkina Faso, and from to 650 mm to over 1000 mm in the semi-humid Sudan climate of south, using the climate definitions of Sivakumar and Gnoumou (1987). Native vegetation ranges from dwarf shrub grasslands in the driest zones to woodlands in wetter zones. Several major rivers flow through the region, fed mostly by the wetter regions in the south. Few of these rivers have large enough reservoirs or sufficient flows to support large-scale irrigation. Within the region there are also ephemeral streams, which are too undependable for major irrigation. For example, in Burkina Faso the total cultivated land is 3.27 million ha, but only 15,000 ha are irrigated (Economist Intelligence Unit, 1999). Rainfall in the region is characterized by large variation among seasons. For example, annual rainfall in Boulsa, Burkina Faso had a range from 419 to 1142 mm over the past 30 years. During recent years, annual rainfall varied from 1010 mm in 1994, which is near the long-term maximum, to 491 mm in 1997, a drought year, to 617 mm in 1998, near to long-term average. Annual rainfall also differs significantly between neighboring villages or farms. During the 1985 wet season Flitcroft et al. (1989) observed nearly a 2-fold difference in total rainfall between sites 14 km apart in the Sahel. There is also large temporal and spatial variability within seasons (Nicholson and Palao, 1993). In general, the less the total rainfall, the greater the variability (Jackson, 1989). Adverse effects of climate variability on agriculture are further exacerbated by relatively wetter or drier periods which may persist for several years or decades (Nicholson, 1985). Such dry periods have led to prolonged droughts, which caused widespread famines in the 1910s, 1940s, mid-1970s, and mid-1980s (Nicholson, 1986). 1.2. Agriculture in Burkina Faso Burkina Faso is a poor country with a Gross Domestic Product of $210 per capita in 1998 (Economist Intelligence Unit, 1999). About 90% of the population depends on agricultural and livestock production for livelihood and 80% depends upon subsistence agriculture (Economist Intelligence Unit, 1999). Most of the cereal produced is for household consumption and is grown under rainfed conditions. The leading cash crops are cotton, peanut, and sesame. There are also small areas irrigated with water from wells, catchments, or reservoirs during the dry season that produce rice or vegetables for local and export markets. Use of fertilizers, mechanized cultivation, and other off-farm inputs is low. Livestock management is an important component of the agricultural system, particularly for the agro-pastoral groups in the drier northern areas where cattle and other livestock are the principal source of income.

334

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

1.3. Production and dissemination of seasonal precipitation forecasts in west Africa The meteorological forecast research community has developed methods to produce regional and national forecasts of total rainfall for the three months of July, August, and September, during which 90% of total annual precipitation occurs in most countries of the Sahel–Sudan. The US Center for Climate Prediction, the United Kingdom Hadley Meteorological Center, the European Center for Medium Range Weather Forecasting, and the International Research Institute for Climate Prediction have produced seasonal forecasts and transmitted them to national meteorological services (NMS). These forecasts, however, have not been widely disseminated to potential users within the region, with little effort to disseminate forecasts to farmers. To promote further development and application of seasonal forecasts, the African Center for Meteorological Applications for Development (ACMAD) has organized annual fora to develop probabilistic consensus forecasts for the region since 1998. These consensus forecasts draw from the individual forecasts produced by organizations both inside and outside of west Africa. Seasonal precipitation forecasts are presented as the probabilities of expected total rainfall being in the upper third, normal third, or lower third of total seasonal rainfall in the region over the historic period for which weather records were available to develop the forecast model. For example, the 2000 consensus forecast predicted rainfall in Burkina Faso to have the following probabilities: 40% above normal; 40% normal; and 20% below normal. Such forecasts may have sufficient precision, which is directly related the number of forecast categories, that governments, businesses, and farmers can use them to improve decision making in agricultural production (Kirshen and Flitcroft, 2000). In June 1999 regional and national forecasts were distributed at the regional climate forum and were posted openly on the internet [http://www.acmad.ne]]. After the 1999 regional forum most west African NMSs used their own models to update the seasonal forecasts in July. An important unanswered research question is how best to incorporate probabilistic forecasts into farmers’ decision-making processes. Hulme et al. (1992b) caution that the skill, which is the frequency that a forecast is correct based on historic data, of these types of forecasts may be insufficient for use at the village level to make direct farming decisions. They warn further that an ‘‘incorrect’’ forecast could result in ‘‘a waste of resources. . .at worst destitution and even threaten survival.’’ Forecast skill has improved greatly in the past decade, but it is still imperative that we use caution in the application of precipitation forecasts to agriculture. 1.4. Climate forecasting for agricultural resources (C FAR) In 1997 Tufts University and the University of Georgia started C FAR, a multidisciplinary project with the ultimate goal of helping farmers in the Sahel–Sudan use climate monitoring and forecasts to enhance their agricultural sustainability and food security. To date, C FAR has emphasized observation and assessment, including a survey of institutional capacity to produce and disseminate climate

335

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

forecasts (Kirshen and Flitcroft, 2000) and qualitative interviews, household surveys, and focus groups to establish farmers’ interest in and ability to use climate forecasts. Project research combined expertise in agricultural systems, meteorology, water resources, and anthropology with an objective to better understand how farm families, institutions, and organizations in the Sahel–Sudan region could use seasonal precipitation forecasts to improve agricultural production and sustainability. Toward this end we undertook field research to ascertain: (1) farmer interest in and ability to use forecasts; (2) forecast information farmers request; (3) lead-time required for greatest forecast value; (4) needs for forecast dissemination, interpretation, and application; and (5) possible strategies for using climate forecasts to improve crop production and resource management. The project aims to identify and recommend points of entry and appropriate modalities for integrating meteorology-based forecasts with local predictions and farming practices in ways that are culturally acceptable and positively affect quality of life, food security, and environmental sustainability. Burkina Faso was used as a case study because it has a range of distinct agro-ecological zones representative of the Sahel–Sudan and the research team had previous institutional and field experience there.

2. Methods We studied three village sites that represent the three main agro-ecological zones of the country: (1) Bwahoun village, near Hounde´, in the cotton production area of the wetter Sudan of the Southwest; (2) Bonam village, near Boulsa, which represents a subsistence cereal producing area of the Central Plateau; and (3) Koria and Table 1 Salient climatic, agricultural, and economic characteristics of the three study sites Site, longitude, latitude

Average rainfall, mm

Principal food crops

Principal cash source

Tillage methods

Inputs

Annual

Jul–Aug–Sep

Dori, 14
020 N, 0
020 W

500

435

Millet, sorghum

Livestock

Hand hoe

Manure

Bonam, 12
400 N, 0
300 W

745

695

Sorghum, millet

Labor migration, peanut, sesame, fruit trees

Mostly hand hoe with a few bullock plows

Manure, local water and soil conservation systems

Bwahoun, 11
390 N, 3
450 W

965

900

Maize, rice

Cotton

Mostly bullock plows with some tractors

Fertilizers, herbicides, insecticides, improved varieties

336

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

Sambonaye villages, near Dori, in the Sahel of the North where agro-pastoralists herd cattle, principally for milk (Fig. 2). 2.1. Site descriptions Major features of the three village sites are listed in Table 1. Sites differ considerably from one another in terms of livelihood configurations, degrees of integration into a market economy, levels of capital investment, and technological development in agriculture. Expansion of cotton production has led to a relatively high level of market integration in the southwest. Labor migration to Coˆte d’Ivoire plays a key role in the livelihood of farming households in the Central Plateau, and livestock provide the main source of cash in the Sahel. Still, agricultural production of all sites depends largely on seasonal rainfall and is therefore a high-risk endeavor. Each of the three communities has about 3000 residents and is located within 30 km of a provincial administrative center. Ethnic composition differs among sites, including an indigenous Bwa community increasingly outnumbered by an influx of Mossi immigrants in the southwest, a predominantly Mossi farming population in the Central Plateau, and a majority of agro-pastoral Fulbe in the Sahel, subdivided between ‘free’ Fulbe and caste groups, including formerly enslaved Rimaibe´.

Fig. 2. Map of Burkina Faso showing mean average annual rainfall (after Sivakumar and Gnoumou, 1987) and location of study sites.

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

337

Sahelian Fulbe rigorously observe Islam precepts and condemn anything that deviates from them. The Mossi of the Central Plateau integrate elements of Islam and Christianity with an enduring allegiance to ancestor-based spirituality. Among Bwaba, who have resisted the penetration of monotheistic religions, local beliefs are being eroded by the socioeconomic changes triggered by cotton production. Export agriculture has enabled Southwest farmers to have far greater access to technology, inputs, and credit, which are channeled through SOFITEX, a semi-private partnership of the Burkina government and the French textile industry that purchases, processes, and exports cotton. In the southwest, a few wealthy farmers have tractors and most farming households own at least one plow and pair of oxen. Less than one-fourth of households in the Central Plateau own plows. Although none of the farm households in the Sahel owns a plow, this may be because the top soil layer is too shallow and the subsoil too rocky to allow plowing rather than because of an economic constraint. Farmers in the Central Plateau and in the Sahel apply only manure and organic matter to their fields, but southwest farmers apply chemical fertilizers, herbicides, and insecticides. Inputs used by farmers in the southwest are bought on credit from SOFITEX through farmer groups. Costs are subtracted from remittances from cotton sales after harvest. Because of the importance of cotton to international trade, farmers of the southwest have been better served by agricultural extension than have other areas. In the other two sites of this study the outreach capacity of government extension is limited, but foreign-funded projects or international NGOs provide resources for some activities, many of which emphasize household food security or natural resource management. These programs along with recent national policies of decentralized natural resource management have led to the formation of viable networks within villages that can be tapped into to disseminate forecasts. Different levels of access to formal education, agricultural extension, and development programs, and differences in adherence to monotheistic religions affect the extent to which local knowledge, including local climate forecasts, remains a viable basis for farmer decisions. Farmers of the Central Plateau described a rich repertoire of local climate indicators including tree fruit and flower production, duration and intensity of cold and hot periods, bird and insect behavior, and movements of stars and moon. Despite this wealth of local indicators, farmers report that climate is changing with rains becoming increasingly erratic and harder to predict. The influences of Islam in the Sahel and of cotton-focused extension services in the southwest have eroded confidence in local systems of precipitation forecasting (Roncoli et al., submitted for publication). In all sites, producers concur that drawing a livelihood from the natural environment, whether by farming or by herding, has become increasingly arduous and risky during the last two–three decades. Climate variation is one of the factors farmers mentioned most often as increasing risk. Farmers in Burkina Faso have a broad perception of climate variability, including more frequent water-deficit years, late onset of the rainy season, premature end of rains, and anomalous rainfall distribution. Meteorological data do not always support farmers’ perception that rainfall has declined over time or that climate variation is increasing (Ovuka and Lindqvist,

338

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

2000). In Burkina Faso, however, there does appear to be a 40- to 50-year cycle of total annual rainfall (Sivakumar and Gnoumou, 1987), and it is possible that recent years reflect another low rainfall phase of the cycle. Hulme (1992), on the other hand, suggested that changes in land cover and greenhouse gases could explain why seasonal rainfall was lower in the Sahel from 1961 through 1990 than during the preceding three decades. In any case, global circulation models predict further increase in drought frequency and temperatures for the Sahel over the next century (Dirmeyer and Shukla, 1996), so farmer interest in forecasts is likely to increase further. 2.2. Field methods Research was conducted between January 1998 and October 1999 in the course of five fieldwork periods, each 4–6 weeks long. In each village, we used a combination of interview and survey techniques to collect data. We began with large discussion groups to introduce our project and to establish the parameters of study. We then conducted open-ended interviews of three to six key informants in each village to study broad issues related to agricultural practices and local methods for predicting climate as described in detail by Roncoli et al. (submitted for publication). Then we conducted in-depth interviews of at least 20 randomly selected adults from different households in each village to collect specific data on crop production and economics as described in detail by Roncoli et al. (in press), and use of weather and climate data. Finally, we used focus groups of 4–10 participants to study the different perceptions and needs of men and women with respect to access to and use of forecasts and other resources. Research periods were staggered throughout the year to include wet and dry seasons. Observations also encompassed a range of significant climate events, such as the aftermath of a major drought in the Central Plateau in 1997 and equally devastating heavy rains in the southwest in 1999. The time span enabled us to observe how these events influenced production strategies and outcomes, how local households coped with the resulting shortages, and how the events shaped farmers’ expectations and adaptations relative to the subsequent farming season.

3. Results 3.1. Forecast information farmers request and need Given that livelihood of most rural households depends on rainfed farming it is hardly surprising that most farmers expressed strong interest in receiving seasonal precipitation forecasts. At the same time, farmers said that by itself a forecast of total seasonal forecast is of limited usefulness. Farmers in all sites stressed that forecasts on the quantity of rainfall must include estimates of duration and distribution of rainfall over time and space to be most valuable. In order of declining priority, the most salient rainfall parameters farmers want in a forecast are:

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

339

(1) timing of the onset and end of the rainy season; (2) likelihood of water deficits, that is, rainfall distribution; and (3) total amount of rainfall. The relative importance of these forecast products varies according to the different agricultural strategies and crop requirements in various zones. For instance, cotton and maize, the two major crops in the southwest, are vulnerable to water stresses at critical growth stages. Thus, southwest farmers are particularly interested in knowing when water deficit periods may occur so they might stagger planting to avoid exposing crops to water deficit during vulnerable crop growth stages. Farmers in the Central Plateau, who grow mostly sorghum, are more interested in the duration of the rainy season, in order to select among varieties with different water requirements and growth cycles. In the Sahel, information on seasonal rainfall quantity can help farmers to know whether to plant millet in low or high water retention areas. Information priorities are also affected by recent experience. In 1998 southwest farmers stressed the need for forecasts of water deficit periods, which adversely affected cotton after planting and maize during tasseling and silking the preceding season. They were not concerned with rainfall quantity after the 1998 season, but reported strong interest in forecasts of rainfall quantity after the 1999 season because cotton and maize suffered from excessive rains during that season. Had farmers known in advance that rains would be excessive they would have planted rice instead of cotton or maize in low-lying fields. Farmers would prefer a combination of forecast parameters rather than a single parameter. For example, southwest farmers would have been able to prepare for flooding conditions in August had they known that the 1999 rains were expected to begin late but would be heavy with most seasonal rainfall occurring in July and August. 3.2. Forecast timing In developing forecasts, meteorologists must balance needs for forecast accuracy with timeliness. Farmers of Burkina Faso agreed with US farmers studied by Mjelde et al. (1988); a less accurate forecast with sufficient lead-time would be more valuable than a highly accurate forecast that arrives after farmers have made irrevocable decisions. Most farmers in Burkina Faso make decisions about planting based on what happened during the previous season. They revise decisions based on shortterm assessments of upcoming season. The planting period, which lasts 30–90 days according to climatic zone and date of onset of rains, is the most critical part of the farming season. Farmers consider the timing and nature of onset and the performance of crop establishment, especially whether or how many times they must replant, to be the most reliable indicators for the rest of the rainy season. Farmers associate late onset of rains with drought and early onset with good rainfall, assumptions that can prove erroneous as they did in 1999. Most farmers in all sites requested that a forecast of seasonal precipitation arrive 1–2 months before the expected onset of the rainy season, that is, by late April or early May. This lead-time would enable them to optimize labor and land allocation, to obtain seed of different varieties, and to prepare fields in different locations. Other

340

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

strategic responses to a forecast, such as clearing new fields, applying more or less manure, ordering inputs, and implementing local soil and water conservation systems require the information to be delivered early in the dry season, by January or February. For example, in the Southwest, SOFITEX would need a forecast by February at the latest if they are to use it to decide which varieties of cotton to distribute in different parts of the country and to ensure that seed is delivered to farmers in time for planting. Though of less value, forecasts of total seasonal rainfall provided at the beginning of the rainy season could still contribute to revisions of farm decisions, in combination with farmers’ own observations at the onset of the rainy season. 3.3. Forecast dissemination 3.3.1. Explaining the 2000 forecast to farmers In May of 2000, the C FAR team conducted a workshop to present the seasonal forecast to a group of three farmers from each of the three sites. We first explained in simple terms how the forecast is developed from SSTs and how the forecast may change unpredictably in the next few months. The forecast for Burkina Faso, predicting 40% probability of above normal rainfall, 40% normal rainfall, and 20% below normal rainfall, had been issued at a regional forum the night before, so that these farmers were the first agricultural audience that received it. This scenario confirmed some farmers’ local forecasts, which also predicted a relatively wet season, based on their observation that temperatures during the early dry season had been colder than normal. A practical demonstration emphasized the probabilistic nature of the forecast. We cut 5 cm5 cm squares from colored paper, with blue representing higher than normal rainfall, green representing normal, and red lower than normal. We arranged 20 squares each of blue and green and 10 squares of red to represent the probabilities forecast of each possible outcome. We emphasized that all outcomes were possible, only that there was a strong likelihood that total rainfall would be higher than normal or normal. We then placed all squares in a box, mixed them, and allowed farmers to draw squares randomly from the box to exemplify the relationship between forecasted probability and actual occurrence of seasonal rainfall. Squares were returned after each draw to assure that relative probabilities remained constant. This demonstration was similar to a lottery draw, with which farmers were already familiar. They appeared to understand both the probabilistic nature of the forecast and the fact that all outcomes were possible even if they had different probabilities of occurring. This was the first time that farmers from the study sites were exposed to a scientific forecast. In 1998 the Burkina Faso National Meteorological Service (NMS) did not disseminate the forecast because they felt it was too experimental (Fre´de´ric Ouattara, personal communication). In June 1999, NMS disseminated the forecast via the national radio and television to alleviate farmers’ anxieties due to the late onset of the rains. However, because broadcasts were in French and at inconvenient times, farmers attending the workshop had not received the information. In the

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

341

southwest, SOFITEX delivered the forecast to cotton farmers through its village agents. But because the latter lacked the training to understand and explain the forecast, most farmers ignored it: only one of the cotton farmers who participated in our workshop had taken notice and responded by expanding its cotton holdings. 3.3.2. Forecast interpretation and access Literacy levels remain very low in all sites, particularly in the Sahel where families still resist sending children to school because they are needed to tend to livestock. In most sites farmers identified local language radio programs to be the best way to deliver forecasts, with follow up meetings with agricultural extension agents. Radio broadcasting would ensure widespread and timely coverage, while extension work would enable farmers to ask questions and get advice. This approach also reflects local learning styles, which are based on experience and interaction rather than verbal instructions. Popular gathering places, such as Friday mosques in the Sahel, markets in the Plateau, and beer drinking places in the southwest, can also be points of information delivery and discussion. Burkina Faso and several other west African countries have relatively dense networks of rural radio stations. Farmers in all sites eagerly follow radio programs in local languages broadcast by the national radio or by neighboring countries. Farmers stressed that programs should be aired when farmers are home from the fields and herders and not engaged with their animals. Thus, timing of broadcasts must be optimized for each area. All households in the southwest and about half of them in the Central Plateau and Sahel have radios. Southwest farmers buy batteries in bulk to operate spraying equipment so they almost always have batteries to operate their radios. The batteries that are available on the local market are not very expensive, but they are generally of poor quality and last only a short time. Farmers of the Central Plateau and Sahel said that they occasionally do not have enough money to buy even the relatively inexpensive batteries available locally. Although not all farmers have radios, neighbors and friends often gather to listen to broadcastings. Television may also be medium for forecast dissemination. In all sites, a few teachers, extension workers, or wealthy farmers have televisions, and villagers often gather to watch soccer matches or other programs. Although farmers concluded that they would ultimately devise their own resource management strategy and that strategies may differ somewhat among neighbors, they did not feel that dissemination of a forecast alone would suffice. They preferred to have someone visit who could answer questions about the nature of the forecast and with whom they could discuss options for responding to a forecast. The ability of the government agricultural extension service is extremely limited, to the extent that agents may not have fuel to get to the field except when a foreign-funded project provides funds for specific areas and purposes. In the southwest, SOFITEX has deployed its own network of village-based agents, but most are young, poorly trained, and accountable to the company, hence farmers do not trust them deeply. For instance, in 1999, SOFITEX disseminated the 1999 census climate forecast from ACMAD to farmers in order to convince them to continue planting cotton even

342

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

though the rains were late in starting. Most field agents were unable to interpret the information for farmers so most farmers did not seriously heed the forecast. Nonetheless, SOFITEX reported that farmers planted an additional 50,000 ha because they disseminated the forecast (Georges Yame´ogo, personal communication). In order for forecasts to be understood by and useful to farmers, they need not only to provide relevant information, at the optimal time, and in the most appropriate form and language, but also be delivered by credible sources. Delivery of useful forecasts calls for a much closer inter-institutional collaboration between meteorological experts and agencies that intervene in rural areas, such as technical services, development projects, and local institutions. For instance, extension and development agents could be trained in the interpretation of forecasts and they could advise meteorologists about the information needs of rural producers. Unfortunately the ability of the NMS and other government agencies to provide training and technical assistance is severely limited as a result of structural readjustment policies enacted during the past decade (Roncoli et al., 2000). An issue that has greatly concerned C FAR researchers and others involved with disseminating climate forecasts to farmers is how we can avoid potential disaster that might arise if the forecast is ‘‘incorrect’’ (Hulme et al., 1992b). Strictly interpreted, because forecasts include the possibility of all outcomes, a forecast cannot be correct or incorrect. Nonetheless, there is great concern that farmers might invest resources in response to a forecast that predicts a greater probability of higher than normal rainfall, then lose their investments and more if rainfall is less than normal. Farmers have a long tradition of managing risk and generally develop strategies to accommodate a full range of potential rainfall outcomes. Furthermore, meteorologically-based forecasts are not farmers’ sole source of information. They will not rely heavily on these forecasts until the forecasts have proven themselves reliable. Finally, we believe it is imperative that forecast dissemination emphasizes that all outcomes are possible, that meteorologists issue regular forecast updates, and that forecasts are regional rather than location specific. 3.4. Strategies for using forecasts to improve crop production and resource management The main strategy whereby households attempt to reduce risk is by diversifying production and livelihood systems, although dominance of cotton production for export markets has decreased diversification in the southwest. In all areas, farmers diversify by becoming more involved in livestock production, whereas agro-pastoralists expand the area they cultivate. Most households engage in non-farm incomegenerating activities, such as making dolo (local beer), crafts, and trade. Agricultural diversification requires farmers to carefully orchestrate what, where, when, and how to plant their crops in response to both current and expected conditions. Diversification results in a household managing a constellation of fields with different soil types, in different locations, planted in staggered sequences with various combinations of crops and crop varieties having different growth duration and water requirements (Roncoli et al., 2001). Nonetheless, because of the similarities

343

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

among locations, there is a common set of strategies from which most farmers draw (Table 2). Although farmers draw strategies from a similar set or response options, each farmer must decide individually how to manage their own particular resources, considering such factors as household size, debt exposure, and credit available. Thus, each farmer may respond differently to a forecast as they design their particular response to a forecast, but farmers within the same village often follow one another to assure that their fields are not out of synchrony with other fields and thereby vulnerable to pests. In all sites, farmers interviewed strongly emphasized that their ability to respond adequately to forecasts is hindered by resource limitations (Table 3), especially availability of labor and productive land. Most farmers also mentioned that they are constrained by lack of access to credit, capital, or agricultural technologies. Because rapid crop establishment is a key factor in coping with a shortened rainy season, farmers stated that access to tractors, plows, and other technologies that could expedite crop establishment is critical. As was found by Blaikie et al. (1994), the household resource access profile largely defines which particular strategy is selected and, when and how it is enacted. Labor availability constrains many responses to a forecast, especially for poor households. At the same time that labor shortages constrain responses to a forecast, one strategy by which households cope with crises is for young men to migrate to the Table 2 Potential response strategies that farmers may implement in response to receipt of rainfall forecast of high probability for higher or lower than normal seasonal rainfall and lead-time needed to implement strategies Above normal 1. Agricultural responses Clear upland areas for planting Order less insecticide (Bwahoun) Orient furrows along slope Plant longer duration crops/varieties Plant flood tolerant crops Decide planting sequence based on location and toposequence position Increase area planted in uplands Plant more cash crops Decrease total area planted Apply more fertilizer or manure Sell grain stocks during rainy season 2. Non agricultural responses Acquire capital to purchase inputs

Below normal

Month required

Implement soil and water conservation Order less herbicide (Bwahoun) Sell livestock or go on transhumance Orient furrows across slope Plant shorter duration crops/varieties Plant drought tolerant crops/varieties

Jan Jan Feb May May May May

Apply less fertilizer or manure Store grain stocks

May May Jun Jun Jul

Ration food Increase income-generating enterprises Migrate Purchase or borrow food grain Send younger men abroad to work

Jan Jan Mar Apr Jun

Plant more cereal crops

344

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

Table 3 Obstacles that may prevent farmers from implementing responses to climate forecasts Labor shortage Competition for high quality fields Debt Lack of access to credit or inputs Lack of access to markets Disruptions to traditional lines of authority Insufficient lead-time of forecasts Lack of predicted date of onset and end of rains Inability to predict rainfall distribution, likelihood of drought or flood Regional rather than local nature of forecast

plantations of Coˆte d’Ivoire or gold mines of Niger to work and send cash back to the village. Such out-migration intensified in the aftermath of the 1997 drought in the Central Plateau. Thus, households balance needs for labor against opportunity for cash income accrued through out-migration. Each response tactic entails compromises and trade-offs. Though a particular tactic may mitigate some risks, it may also expose farmers to other risks, both foreseen and unforeseen. In making resource decisions, farmers balance forecasts against experience of the previous season and temper decisions with an expectation of high variability. Factors included in farmers’ decisions are summarized below for each of the three agro-ecozones. 3.4.1. Southwest The first strategy whereby southwest farmers reported that they would respond to rainfall forecasts is by changing furrow orientation during land preparation. They would orient ridges with regards to slopes to either facilitate flow in wet years or prevent water flow in dry years. Rainfall forecasts may also influence their decisions concerning the relative proportion of cotton and maize to be planted. Because cotton production has high labor requirements, other crops that farmers grew previously have largely given way to maize, which has higher productivity and timing of labor demand that is more compatible with labor demands for cotton. On the other hand, maize is vulnerable to waterlogging. If wet conditions are predicted and farmers have access to upland fields, they would plant maize on upland fields, possibly displacing peanut. If they do not have access to such land, they must plant maize as early as possible to ensure that plants are strong enough to withstand flooding when heavy rains begin in July. But the ability to plant early is predicated upon having enough household labor and either a tractor or several plowing teams. High rainfall favors productivity of rice in lowland fields and sesame in upland fields. Both crops can be highly profitable, but they also have high labor demands. Labor availability, especially during planting and harvesting, may constrain

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

345

cultivation of these cash crops because they compete with labor needs for maize and cotton. Thus, most farmers plant only small plots of rice or sesame. If low rainfall is predicted, farmers would emphasize food crops over cash crops, by reducing the cotton area and shifting to a short-duration, drought resistant maize variety. Even though farmers consider the short-duration maize to be less palatable, less marketable, and less durable in storage, farmers grow it for food security. Under low rainfall predictions, farmers would also plant maize and sorghum instead of rice in valley bottom fields. Availability of household labor and suitable land are key factors shaping productive decisions in response to forecasts. If the household is short of labor, high rainfall may require them to reduce acreage of cotton and maize in order to weed properly, because weeds proliferate under high moisture conditions. Under high rainfall conditions, farmers may plant more sorghum, which is less vulnerable to weeds. If a household has sufficient labor, it may increase cotton acreage if high rainfall is expected. There are costs and risks associated with this strategy. Households still need to produce enough maize to meet much of their consumption needs because uncertainty of grain prices makes it financially unwise to rely heavily on the market. Expanding the area of cotton cultivated also requires more inputs and therefore leads to a greater debt burden. Farmers reported that they would also adjust use of chemical inputs according to rainfall predictions. For instance, in drier situations, farmers would apply less herbicide, the most costly of all inputs. In contrast, under wetter conditions farmers said that they would apply fertilizer at planting and apply more herbicide. Farmers also said that ‘‘rains knock insects off of the leaves’’ so they would apply less insecticide and would adjust the timing of insecticide applications to minimize likelihood that rains would rinse insecticides from the leaves thus reducing their efficacy. 3.4.2. Central Plateau The principal mechanisms whereby Central Plateau farmers cope with climate variability is by selecting field locations, crop mixes, and varieties. Moisture retention capacity of soils and fields is a primary consideration in deciding what and when to plant. If below normal rainfall is expected, farmers will plant sorghum, the staple grain of choice, in valley bottoms or near watercourses. An increase in frequency of drought over the past several decades has led to increased competition for valley bottom fields. Households without access to valley bottom fields may plant drought resistant varieties of sorghum or millet in their regular fields. Upland areas, characterized by gravelly eroded soils and poor water infiltration, may be planted with millet or abandoned. In response to a forecast of above normal rainfall, farmers may plant flood-prone valley fields with rice, while sorghum may be planted in elevated areas instead of millet. Farmers rarely abandon valley fields to avoid losing their use rights over them. Millet is more drought tolerant than is sorghum (Payne et al., 1990), but millet also yields less than sorghum under normal rainfall conditions. Farmers with large

346

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

households avoid planting millet on most of their fields because millet can be ground by hand and consumed uncooked. Thus, household heads have difficulty in controlling millet consumption. With a 70-day growth duration, maize is well-adapted to a shortened rainy season but needs regular rainfall, careful weeding, and good soil fertility so it is only planted in small manured plots around the compound. During the last 10–15 years farmers have shifted from long-duration (120–150 days) to short-duration (70–90 days) varieties of sorghum, most of which have been imported from the drier northern provinces. Short-duration varieties, however, are said to be less palatable and nutritious, less productive per unit of land, and more vulnerable to weeds, pests, and water stresses than local varieties, which are planted when the rains begin early, as we observed in 1999. A 50-day sorghum variety, also imported from the north, may be planted when the onset of rains is very late. This 50-day variety requires good soil fertility and good weed control so it cannot be planted in most fields. Moreover, unless they are planted late these short-duration varieties increase the risk of damage by birds or rains if they ripen before other nearby fields. Water conservation techniques, such as straw mulching, stone barriers, and zai (planting in small depressions filled with organic matter) are also implemented to cope with water deficits. Feasibility of these techniques depends of availability of labor, transportation, equipment, and materials such as grass and manure. Because these techniques are highly labor intensive, they must be started early in the dry season. 3.5.3. Sahel Seasonal rainfall affects Sahelian agro-pastoralists in a different manner than it affects sedentary farmers. Agro-pastoralists can move with their herds to find water or better pastures for at least part of the year. Still, changing ecological, demographic, and political conditions make it increasingly difficult for Sahelian famers to draw the majority of their livelihoods from herding. Farmers reported that decreasing availability of water and grass, human population growth and expansion of sedentary cultivation, political restrictions against border crossings, and legal prosecution against trespassing in neighboring countries all constrain herd management. Potential livestock management responses to forecasts elicited from herders were few. If below normal rainfall is predicted, pastoralists would collect and store fodder during the rainy season because pasture grasses would be expected to be scarce in the following dry season. If above normal rainfall is predicted, pastoralists would increase grazing time by taking animals to the bush at night in order to maximize weight gain and value. Though their control is diminishing, pastoralists have historically controlled land access, and pastoralists endeavor to prevent expansion of cultivation into pastures. Farmers and agro-pastoralists, on the other hand, reported several possible responses to forecasts, although they are limited by land scarcity. Choosing what to plant and where is the main way farmers can respond to rainfall forecasts. Farmers classify their lands in three categories: elevated clay soil areas (clairie`res), sandy soils (terroirs sablonneux), and valley bottoms (bas-fonds). Because farmers perceive that

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

347

the climate has become drier in the past several decades, they have moved away from clay soil areas to fields with sandy soils, which occupy a lower toposequence position and remain moist for longer periods. They have also attempted to cultivate more valley bottom fields, which were used previously only for grazing. Increased cultivation of valley bottom fields has led to conflict between farmers and pastoralists. Movement away from clay soil areas to valley bottoms is also associated with a shift from millet cultivation, which farmers prefer as a staple crop, to short-duration sorghum, which is better adapted to a short rainy season. Sorghum is less susceptible than millet to water logging, so farmers would increase the area planted to sorghum if higher than normal rain is forecast. Farmers in the Sahel cultivate only one variety of millet, which is well adapted to local conditions. On the other hand, they select from three main sorghum varieties according to growth duration and soil fertility requirements. Farmers would also respond to forecasts of high rainfall by planting more secondary crops, such as sesame, sorrel, maize, cowpea, and calabash. In a water deficit season, farmers would reduce manure application to prevent scorching, plant sorghum in valley bottoms if they have access to such land, plant millet in sandy soils, and abandon clay soil fields. Under high rainfall conditions, farmers may abandon valley bottom fields, plant sorghum and millet in sandy soils, and sorghum in clay soil areas. Under high rainfall conditions, farmers would expand area cultivated if they have adequate labor and land access, or they would manure heavily and weed intensively if labor or land access is limited.

4. Summary and conclusions Farmers of the Sahel–Sudan in Burkina Faso already use a variety of local forecast techniques in ways that may mirror their responses to meteorology based forecasts (Roncoli et al., submitted for publication). Farmers say that they are particularly interested in meteorologically based forecasts because they perceive their local indicators to have lost reliability. Among the three sites studied, forecasts would likely be of greatest value to farmers of the Central Plateau because they use the largest variety of local indicators to predict rains and adapt their resource management strategies accordingly (Roncoli et al., 2002). Ability to use forecasts is constrained in the southwest by debt incurred in cotton production. Because herds are movable, agro-pastoral systems in the Sahel are partly insulated from climate variation, whereas farming systems in the Sahel are constrained by competition with cattle herds. Farmers would benefit most from forecasts that are delivered at least 1 month and preferably 2 months before the start of the rainy season, but they can still benefit from later forecasts. Farmers need a combination of information, thus the full benefits of a forecast for the region will not be obtained until the timing onset of the rains and the likelihood of water deficit periods are forecast in addition to total rainfall amount. To obtain maximum value of forecasts, farmers also need access to resources such as credit and improved agricultural technology. Considering the extreme poverty of the area, particularly in the Central

348

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

Plateau, policy makers should consider linking dissemination of such key production technologies with dissemination of forecasts. Inter-institutional cooperation is needed to ensure that farmers receive the forecasts on time from credible sources and that farmers understand the forecasts. Fortunately, there appear to be adequate mechanisms and networks of governmental and private organizations to produce and disseminate forecasts in Burkina Faso (Kirshen and Flitcroft, 2000) and throughout the Sahel–Sudan. Forecasts should be disseminated by radio in the local language at times when farmers are available. Further research is needed on how best to support dissemination of forecasts to farmers and best strategies for farmers to use forecasts. Farmers need more than just a forecast. They also need explanations and interpretations of the forecast. Explanations should include forecast limitations as well as possible response strategies. Whether national extension services or NGOs, agents need training and resources before they will be able to support farmers in the use of forecasts to improve productivity and sustainability of agricultural production systems in the Sahel–Sudan. This research has established that farmers across a diverse transect of the Sahel– Sudan have strong interest in receiving climate forecasts and describes the complex of biological, physical, social, and economic factors that farmers consider in devising potential response strategies to forecasts. How farmers change their behavior in response to a forecast, and the benefits or costs of such responses are subjects for future research.

Acknowledgements This research has been supported by the Economics and Human Dimensions of Climate Fluctuations Program, Office of Global Programs, United States National and Atmospheric Administration under Grant Number NA76GP0327. We thank PLAN International, particularly Felipe´ Sanchez and Eric Mamboue´, for exceptional support in Burkina Faso.

References ACMAD (African Center for Meteorological Applications to Development), 1998. West African forum on climate variability and prediction. ACMAD, AGRHYMET, ICRISAT, MEDIAS/START, Abidjan, 4–8 May 1998. Blaikie, P., Cannon, T., Davis, I., Wisner, B., 1994. At Risk: Natural Hazards, People’s Vulnerability, and Disasters. Routledge, London. Costello, C.J., Adams, R.M., Polasky, S., 1998. The value of El Nino forecasts in the management of salmon: a stochastic dynamic assessment. American Journal of Agricultural Economics 80, 765–777. Dirmeyer, P.A., Shulka, J., 1996. The effect on regional and global climate of expansion of the worlds deserts. Quarterly Journal of the Royal Meteorological Society 122, 451–482. Economist Intelligent Unit, 1999. Country Reports, 1999–2000. Economist Intelligent Unit, London, UK. Flitcroft, I.D., Milford, J.R., Dugdale, G., 1989. Relating point to area average rainfall in semi-arid west Africa and the implications for rainfall estimates derived from satellite data. Journal of Applied Meteorology 28, 252–266.

K.T. Ingram et al. / Agricultural Systems 74 (2002) 331–349

349

Hulme, M., 1992. Rainfall changes in Africa: 1931–1960 to 1961–1990. International Journal of Climatology 12, 685–699. Hulme, M., Biot, Y., Borton, J., Buchanan-Smith, M., Davies, S., Folland, C., Nicholds, N., Seddon, D., Ward, N., 1992a. Seasonal rainfall forecasting for Africa. Part 1: Current status and future developments. International Journal of Environmental Studies 39, 245–256. Hulme, M., Biot, Y., Borton, J., Buchanan-Smith, M., Davies, S., Folland, C., Nicholds, N., Seddon, D., Ward, N., 1992b. Seasonal rainfall forecasting for Africa. Part II: Application and impact assessment. International Journal of Environmental Studies 40, 103–121. Jackson, I.J., 1989. Climate, Water and Agriculture in the Tropics. Longman, New York. Kirshen, P.H., Flitcroft, I., 2000. Use of seasonal precipitation forecsting to improve agricultural production in the Sudano-Sahel: an institutional analysis of Burkina Faso. Natural Resources Forum 24, 185–195. Linthicum, K.J., Anyamba, A., Tucker, C.J., Kelly, P.W., Myers, M.E., Peters, C.J., 1999. Climate and satellite indicators to forecast Rift Valley fever epidemics in Kenya. Science 285, 397–400. Mason, S.J., Joubert, A.M., Cosijn, C., Crimp, S.J., 1996. Review of seasonal forecasting techniques and their applicability to Southern Africa. Water South Africa 22, 203–209. Mjelde, J.W., Keplinger, K., 1998. Using the southern oscillation to forecast Texas winter wheat and sorghum crop yields. Journal of Climate 11, 65–70. Mjelde, J.W., Sonka, S.T., Dixon, B.L., Lamb, P.J., 1988. Valuing forecasts characteristics in a dynamic agricultural production system. American Journal of Agricultural Economics 70, 674–684. Nicholson, S.E., 1985. Sub-Saharan rainfall 1981–84. Journal of Climate and Applied Meteorology 24, 1388–1391. Nicholson, S.E., 1986. Climate, drought and famine in Africa. In: Hansen, A., McMillan, D.E. (Eds.), Food in Sub-Saharan Africa. Lynne Rienner, Boulder, CO, pp. 107–128. Nicholson, S.E., Palao, I.M., 1993. A re-evaluation of rainfall variability in the Sahel. 1. Characteristics of rainfall fluctuations. International Journal of Climatology 13, 371–389. Ovuka, M., Lindqvist, S., 2000. Rainfall variablity in Murang’a District, Kenya: meteorological data and farmers’ perception. Geografiska Annaler 82, 207–219. Palmer, T.N., Anderson, D.L.T., 1994. The prospects for seasonal forecasting: a review paper. Quarterly Journal of the Royal Meteorological Society 120, 755–793. Payne, W.A., Wendt, C.S., Lasiano, R.J., 1990. Root zone water balances of three low-input milled fields in Niger, West Africa. Agronomy Journal 82, 813–819. Roncoli, M.C., Ingram, K.T., Kirshen, P.H., 2000. Can farmers of Burkina Faso use rainfall forecasts? Practicing Anthropology 22, 24–28. Roncoli, M.C., Ingram, K.T., Kirshen, P.H., 2001. The costs and risks of coping: drought and diversified livelihoods in Burkino Faso. Climate Research 19, 119–132. Roncoli, M.C., Ingram, K.T., Kirshen, P.H., 2002. Reading the rains: local knowledge and rainfall forecasting among farmers of Burkina Faso. Society and Natural Resources 15, 411–430. Sivakumar, M.V.K., Gnoumou, F., 1987. Agroclimatology of West Africa: Burkina Faso. Information Bulletin No. 23. International Crops Research Institute for the Semi-Arid Tropics, Patancheru, India.