Food Production Variability and Modeling in East Africa

Dr. Nathan Moore, Michigan State University

FARMD (March 2012) | Food security depends on a host of political, socioeconomic, and biophysical factors. These factors are especially critical for subsistence farmers, for whom changes in any of these factors can easily lead to bumper harvests or destitution.  Subsistence farmers do not have political leverage or the economic resources to manage many aspects of food security, but one aspect—food production – can be influenced by farmers’ management decisions. Food production thus relies on weather, nutrient inputs, pests, and other biophysical components that can be measured.

"...high-sensitivity areas can serve as “canaries in the coal mine”—places to monitor the vanguard of climate shifts and their impacts on maize cultivation."

For farmers in most of the world, all of these elements of food production are a source of risk. Weather data and forecasts are often sparse or unavailable; fertilizer can be expensive with irregular availability; cultivated and non-cultivated land can be managed in complex strategies that are difficult to measure remotely; and strategies for dealing with pests are only beginning to be applied. Because of these limitations, food production in these areas is very difficult to estimate or measure even with detailed geospatial data.

When data on food production are unavailable, models can be applied to estimate rough levels of food production based on a combination of remotely sensed geospatial data, on-ground surveys where available, and simple models. These estimates (e.g. FEWSNet) can be applied in near real-time to evaluate food production risk, with an eye towards assessing food security. Combined with weather forecasts, near-term prognoses of crop yield and food production can be estimated in order to help with seasonal planning. For longer-term outlooks, however, a more complicated approach is needed.

Historical trends in modeling and forecasting food production risk started in the 1960s as part of Department of Defense programs aimed at estimating wheat yields for the Soviet Union. These initial models developed over time into modern incarnations of “processed-based” crop models like those being compared to one another as part of a massive model-improvement effort (see http://www.agmip.org/ for more details) and interests in developing ensembles of models to improve crop forecasts. The growth in crop models over the last decade has been driven in part by a need to provide guidance to institutions, farmers and crop breeders for the potential conditions they may face in a greenhouse-gas-warmed world. Current efforts to bring together institutions, modelers and stakeholders are helping all participants understand how farmers apply adaptive strategies (bottom up) instead of investment-heavy top-down approaches.

While local-scale adaptation strategies and global scale climate perturbations can be clearly delineated, significant obstacles remain in linking these disparate scales. One such effort at bridging the global and local scale influences is the Climate-Land Interactions Project at Michigan State University. Here, a coupled system of regional climate models, crop models, and local decision making have been combined to understand what processes are significant for maize production in East Africa, and which land-use decisions can significantly affect atmospheric processes in return. The crop simulations were designed to identify and project sensitivity of crop yields where nitrogen and water stress were projected to increase due to climate activity. Our recent research has identified both weather-sensitive regions and land-use-sensitive regions (i.e., where land use change has a significant effect on local weather. Some areas even appear to exhibit a dynamic feedback between land use and regional atmospheric convection. These high-sensitivity areas can serve as “canaries in the coal mine”—places to monitor the vanguard of climate shifts and their impacts on maize cultivation.
 
There are several limitations in making these kinds of crop yield estimates. In Kenya and Tanzania, for example, no reliable surveys exist of what lands are actually cultivating which crops in a given year; farm plots are too small (and irregular) for satellite remote sensing, and heterogeneous uses make it difficult to identify individual crops. Thus, any estimate of the area under maize cultivation in East Africa is highly uncertain. In addition, the food production strategies shift quickly from year to year, so maps of maize production lack fidelity to on the ground changes. Estimates of nitrogen inputs are similarly limited, as is maize variety, planting time, and soil characteristics. Significant field work in the form of soil surveys, cultivation methods, nitrogen use, and cultivated area are all needed to better improve crop projections and future crop breeding needs.

 

Given these limitations of data, agricultural yields in rural parts of East Africa have proven difficult to estimate but possible to examine from the standpoint of sensitivity. Data from FEWSNet show significant year-to-year variability in yields. Much of this variability is due to weather – droughts and floods—in regions where maize cultivation is an adaptation strategy by subsistence farmers against unreliable shifts in rainfall and temperature. Climate change due to elevated greenhouse gases is expected to lead to increased nighttime temperatures that accelerate maturation and shorten the growing season. These shortened growing seasons in already warm equatorial regions threaten to seriously diminish yields. Cooler highland areas may exhibit projections of better yields, but East Africa’s highland areas are currently used for high-value cash crops like coffee, tea, and vegetables. A switch to maize would mean a loss. As for rainfall changes, most IPCC AR4 models indicate wetter conditions for East Africa for the growing seasons. This would help limit the lowered yields due to warmer temperatures. However, work by Chris Funk and colleagues in PNAS shows that the projected warming shown by global climate models is not consistent with the observed data: a warmer Indian Ocean shows a significant association with drier conditions over East Africa.

With uncertain climate changes ahead, adaptation will require diverse strategies with multiple-scale integrated research teams. Our methods need broader involvement. Better short-term forecasts, better communication between stakeholders and scientists, and consistent, reliable on-the-ground data will be needed to develop more resilient, food-secure systems in East Africa and elsewhere.