Large scale wetland mapping in semi arid Africa using 250-meter MODIS phenology metrics and topographic variables

This paper describes a unique methodology to map wetlands based on Land Cover Classification System (LCCS) codes over a 360.000 km² large study area in West Africa. Aim of the paper is to show how 250-meter MODIS metrics from several years and topographic variables can be optimally used to accurately map wetlands over larger areas. A new wetland flooding regime index that uses the spectral contrast between near infrared (NIR) reflectance of flooded areas in the wet season and NDVI in the dry season is utilized. The flooding regime index is augmented with the length of the inundation period/year using NIR reflectance minima. Topographic features and information on hydric soils were used to mask out areas of low wetland probability. Landsat trajectories were used to ascertain the MODIS pixel homogeneity of wetland classes. Between 43% and 64% of the Landsat determined wetland sizes could be mapped in the MODIS data.