Validation of TRMM and Other Rainfall Estimates with a High-Density Gauge Dataset for West Africa. Part I: Validation of GPCC Rainfall Product and Pre-TRMM Satellite and Blended Products

Gauge data over North Africa are used to provide an assessment of satellite and blended rainfall products for 1988–94 and for 1998. A comparison is also made with the Global Precipitation Climatology Center (GPCC) gauge dataset. For the 1988–94 period, mean fields and latitudinal transects for the June–July–August season are presented, based on a 515-station gauge dataset, the GPCC gauge data, the Global Precipitation Climatology Project (GPCP) blended data, the infrared-based Geostationary Operational Environmental Satellite precipitation index (GPI), and the Special Sensor Microwave Imager (SSM/I) microwave estimates. Error calculations are also presented. The mean fields derived from the dense gauge network, the GPCC gauge-only analysis, and the GPCP are remarkably similar. The bias, with reference to the seasonal rainfall field based on the denser network, is about 3%–4% for either GPCC or GPCP. Agreement is relatively good, even in individual years. The rms error associated with these datasets is 12% for seasonal rainfall totals; thus, the error is largely random. In contrast, there are large systematic errors in the satellite-only analyses of GPI and SSM/I, with biases of 20% and 40% for the mean rain field as a whole and much larger biases in individual years. The rms errors are nearly 2 times as great. For 1998, a 920-station gauge dataset was available for a smaller section of West Africa. The comparison confirmed the superior performance of GPCP and demonstrated the lower level of performance of both GPCP and GPCC at the monthly scale as compared with the seasonal scale. Overall, the results of this study underscore the continued need for extensive gauge networks to describe adequately the large-scale precipitation field over Africa.