Generating multi-period crop distribution maps for Southern Africa using a data fusion approach

Abstract Detailed data on the location of crops is essential to inform national food and agricultural policies. A key source of information on the spatial distribution of crops are the global datasets produced with the Spatial Production Allocation Model (SPAM). SPAM uses an optimization approach to allocate national and subnational crop statistics for four production systems, informed by spatial information on both biophysical (e.g. suitability and potential yield) and socio-economic (e.g. market access and population density) drivers of crop location. The SPAM crop distribution maps are produced at a resolution of 5 arc minutes, which is often too coarse for detailed country and subnational assessments, which require higher resolution products and the flexibility to subsume more detailed information from national sources. The aim of this paper is to demonstrate an extended and improved version of SPAM, which is able to (a) produce maps at a higher resolution than the current existing global maps; (b) incorporate additional and more detailed information on the location of crops (i.e. from OpenStreetMap); and (c) create linked maps that can be compared over time. The model is applied to seven countries in Southern African region. Detailed results are presented for maize and irrigated wheat production in Southern Africa. Maize is widespread in Mozambique, Malawi and Zimbabwe, Angola and Zambia, while very little maize is grown in Botswana and Namibia. Wheat is predominantly located in the Central and Southern provinces of Zambia, which overlaps with the location of commercial farm blocks, and is widespread in Zimbabwe. A validation using information on the location of a large number of crops from household surveys for Malawi and Zambia showed 75%-100% true positives for most crops. The model is performing less well for a few marginal crops that are only grown in specific regions and several crops that are observed throughout the country. The crop distribution maps can be used to support regional crop monitoring systems, guide investment decisions and inform national assessments of food security under socio-economic and climate change.