GENERATING PLAUSIBLE CROP DISTRIBUTION MAPS FOR SUB-SAHARA AFRICA USING SPATIAL ALLOCATION MODEL

Spatial data, which are data that include the coordinates (either by latitude/longitude or by other addressing methods) on the surface of the earth, are essential for agricultural development. As fundamental parameters for agriculture policy research agricultural production statistics by geopolitical units such as country or sub-national entities have been used in many econometric analyses. However, collecting sub-national data is quite difficult in particular for developing countries. Even with great effort and only on regional scales, enormous data gaps exist and are unlikely to be filled. On the other hand, the spatial scale of even the subnational unit is relatively large for detailed spatial analysis. To fill these spatial data gaps we proposed a spatial allocation model. Using a classic cross-entropy approach, our spatial allocation model makes plausible allocations of crop production in geopolitical units (country, or state) into individual pixels, through judicious interpretation of all accessible evidence such as production statistics, farming systems, satellite image, crop biophysical suitability, crop price, local market access and prior knowledge. The prior application of the model to Brazil shows that the spatial allocation has relative good or acceptable agreement with actual statistic data. The current paper attempts to generate crop distribution maps for Sub-Sahara Africa for the year 2000 using the spatial allocation model. We modified the original model in the following three aspects: (1) Handle partial subnational statistics; (2) Include the irrigation map as another layer of information in the model; (3) Add subsistence portion of crops in addition to the existing three input and management levels (irrigated, high-input rainfed and low-input rainfed). With the modified spatial allocation model we obtain 5 by 5 minutes resolution maps for the following 20 major crops in Sub-Sahara Africa: Barley, Beans, Cassava, Cocoa, Coffee, Cotton, Cow Peas, Groundnuts, Maize, Millet, Oil Palm, Plantain, Potato, Rice, Sorghum, Soybeans, Sugar Cane, Sweet Potato, Wheat, Yam. This approach demonstrates that remote sensing technology such as satellite imagery could be quite useful in improved understanding of the spatial variation of land cover, agricultural production, and natural resources.