Progress
Tested in Latin America and Sub-Saharan Africa, SPAM currently generates a global distribution of crop area and production for 20 major crops: wheat, rice, maize, barley, millet, sorghum, potato, sweetpotato, cassava and yams, plantain and banana, soybean, dry beans, other pulse, sugar cane, sugar beets, coffee, cotton, other fibres, groundnuts, and other oil crops. The detailed spatial datasets represent a unique and rich platform for exploring the social, economic, and environmental consequences of agricultural production in a strategic policy context.
While new technologies are improving data collection, working at a spatial scale of individual pixels creates many data management and computational challenges. Some of these challenges need to be met through improved numerical methods and mathematical optimization software. Though the current model provides what appear to be reasonable results, more work is underway to improve its performance. Potential enhancements include better approximations of the agricultural extent, more realistic crop suitability surfaces, and more research on the association between crop production and population density.
On the other hand, more information can also be added into the model. For example, household or agricultural survey information on the location and quantity of crop production would provide a direct, sampled calibration of the entire crop distribution surface. Reasonable behavioral assumptions, such as that farmers would opt to plant higher-revenue crops in any given location, all other things being equal, could also be useful, although risk minimization might also play a role. Thus several options are available for further exploration of alternative drivers of crop choice, both individually and in crop combinations, in each location. Furthermore, changes in cropping patterns over time are as important as the cropping patterns over space. Therefore, HarvestChoice will estimate similar crop maps continuously in the future, most likely every five years. Solving the model’s problems is central to achieving national and regional development objectives in developing countries and will certainly be boosted by better and more disaggregated production data.
Collaboration
HarvestChoice validates its results through collaboration with CGIAR scientists and continues to compile newer and higher resolution data and to regenerate successive versions of SPAM data. This feedback from crop scientists and local experts in the field increases the accuracy of the crop distribution maps. Updates to the SPAM database are undertaken on a country-by-country basis as and when any significant new data input source is obtained. While HarvestChoice maintains its own collection of sub-national production data, a major resource for such data is the Agro-Maps website, hosted by FAO. HarvestChoice scientists helped establish Agro-Maps and continue to collaborate closely.
