Geostatistical Tools for Validation of Satellite and NWP Model Rainfall Estimates

Traditionally, rainfall information has been in the form of raingauge records where the rainfall is observed at a particular point on the Earth’s surface. In recent years, increasing use has been made of satellite-based rainfall estimates, radar images and NWP (numerical weather prediction) model products. In all these cases, the rainfall information is presented as areal averages over gridboxes (NWP models) or pixels (radar, satellite). A question of comparison of these areal products with the point gauge data often arises in the process of calibration or evaluation of rainfall algorithms. The mismatch of spatial scale makes the problem non-trivial. The problems are exacerbated by sparse gauge networks and, as that is frequently the case in Africa, examples used in the course of this lecture will be taken from African rainfall data.