Dynamic mapping of cropland areas in Sub-Saharan Africa using MODIS time series

Abstract

Mapping cropland areas in a dynamic way is of great interest to successfully monitor agricultural areas and food security. Existing cropland masks are either too coarse or inaccurate or are limited in spatial coverage. This study aims at developing a method for dynamic mapping of cropland areas in Sub-Saharan Africa and at producing a multi-annual map of cropland extent at 250m using MODIS time series. The originality of the approach consists of including a dynamic and automatic stratification that allows tuning the classification parameters according to the inter-annual variability, and exploiting the local differences of spectral signatures between natural vegetation and crops during the green-up season. The accuracy of the product is assessed using a large sample of points interpreted on high resolution images and is compared to the accuracy of two existing cropland maps.