The impact of biophysical conditions on food security in three Sahelian countries (Mali, Niger, Burkina Faso)

Extensive efforts have been made to find operational metrics of food security and its geographical variation in the West African Sahel since the catastrophic famines of the 1970s and 1980s. Food insecurity has been estimated using many different types of data, including satellite-derived vegetation indices and food prices in markets. By developing a series of multi-linear models, the relationship between the price of the primary subsistence crop, coarse grain millet, and seasonal fluctuations in satellite-derived vegetation indices has been established for Mali, Niger and Burkina Faso. It was found that, although the use of Normalized Difference Vegetation Index (NDVI) time series improved the price estimation in the linear models, the improvement was small. Using a new time series decomposition technique known as Empirical Mode Decomposition (EMD), a seasonal component and an interannual trend were extracted from the price and NDVI time series. Although Sahelian NDVI has 80% of its variance in the seasonal component, only 7% of the variance of the price was due to seasonal oscillations. By using the EMD components instead of the raw time series in the model, the errors in the price per kilogram were reduced from an average RMSE of 13.2 CFA/kg to 7.9 CFA/kg.; Building on these models, historical and future price movements were estimated. Continuous images of vegetation health were integrated with known historical behavior of prices in Mali, Niger and Burkina Faso in order to derive maps of annual anomalies from mean millet prices. A one-year forecast of millet prices was made in the form of a map that combines information on the previous growing season and a three-year cycle found in the interannual millet prices between 1981 and 2000. The combination of forecasting and imaging capabilities provides a powerful new indicator of access to food in West Africa.