Science et changements planétaires / Sécheresse


Rainfall estimation in the Sahel: Using an error model to evaluate rain gauge and satellite products Volume 15, issue 3, JUILLET-AOÛT-SEPTEMBRE 2004


See all figures

Centre régional de formation et d’application en agrométéorologie, hydrologie et météorologie, (AGRHYMET). BP 11011, Niamey Niger, Institut de recherche pour le développement (IRD), Laboratoire d’étude des transferts en hydrologie et en environnement (LTHE), UMR 5564, BP 53 38041 Grenoble cedex 09 France

Drought is continuing in the Sahel. Lower rainfall and the decrease of the number of raingauges are two factors that lead to larger errors in the estimation of rainfall from ground observations. This paper consequently investigates the quality of rainfall estimation from the raingauge operational network and the differences existing between ground-based rainfall estimates and various satellite-based rainfall estimates. To provide an objective base for this evaluation, an error model approach is presented. The error model is calibrated on a densely instrumented test area. A first application of the model is to study the concept of standard optimal networks, optimal referring here to a given level of average expected error for a given space-time resolution. Taking 1° x 1° monthly resolutions as the lowest acceptable values for many water resources and agronomical applications, it is shown that the AGRHYMET Regional Center (CRA) operational network is far from being optimal, since it guarantees in rare places only an average estimation error lower than 10%. The proposed error model also makes it possible to carry out an intercomparison of the most commonly used satellite products, that is CMAP, GPI, GPCC, and GPCP. All these products have a distribution of their estimated rainfall values which is significantly less dispersed than the distribution of the CRA ground-based reference values (computed at the same resolution as the satellite rainfall products). The differences between the reference values and the satellite products and the differences between the various satellite products are strongly dependent on the latitude. The performance of the GPI product is the worst, whereas the CMAP product, which in fact includes a large quantity of rain gauge data, is the best on average.