Assessment of remotely sensed and statistical inventories of African agricultural fields

This paper critically examines different sources of remotely sensed and statistical inventories of African agricultural fields. Substantial discrepancies are found across alternative sources of information in both the extent and location of agricultural fields. In one-third of the countries, the difference between lowest and highest field extent estimate exceeds 25% of the total country area, and the maximum difference at the continental level is 2.6 million km(2). Much of the disagreement between land-cover maps arises from areas of low cropping density. These inconsistencies have important implications when using these data directly, e.g. for the assessment of land cover changes, or indirectly in economic or physical models, and indicate a need to explicitly quantify uncertainties arising from the limitations in land-cover data. They also highlight the need for development of regional land information systems for baseline development and informed policy decisions.