Comparison of satellite remote sensing derived precipitation estimates and observed data in Kenya

Abstract

This study evaluated the accuracy of four satellite remote sensing (SRS) based products in predicting rainfall (amounts and spatial distribution) over Kenya between 1998 and 2013. The four SRS products used include; two satellite products (Climate Hazards Group InfraRed Precipitation with Station (CHIRPS 2.0) and Tropical Rainfall Measuring Mission - Multi-satellite Precipitation Analysis (TMPA) 3B42 version 7 (TRMM)), one gauge-interpolated product (Global Precipitation Climatology Centre (GPCC)) and one re-analysis product (Modern-Era Retrospective Analysis for Research and Application (MERRA)). The monthly precipitation data were evaluated for completeness, converted to individual raster files, projected to the World Geodetic System (WGS) 1984 - Universal Transverse Mercator (UTM) zone 37 N ensuring a similar processing extent, rescaled to a common resolution and reclassified following defined uniform intervals for ease of comparison. Thereafter, they were subjected to five different metrics based on eight agro-ecological zones (AEZs) of Kenya, in reference to observed rainfall data obtained from Kenya meteorological department (KMD). Results show that all SRS products both overestimated or underestimated rainfall amounts on a pixel to pixel comparison. Based on point to point proportion of variance evaluation (r2), TRMM best-estimated rainfall in the tropical cool humid (r2 = 0.64), tropical warm humid (r2 = 0.58) and tropical cool subhumid (r2 = 0.39) zones and can be used for agricultural advisory services. The GPCC product best-estimated rainfall in the tropical warm semiarid (r2 = 0.46) and warm tropical sub-humid (r2 = 0.21), while CHIRPS 2.0 best-estimated rainfall in the tropical warm arid (r2 = 0.33) and therefore the two products could be best used to predict rainfall in the ASALs and drought-related studies, with potential for irrigation. The MERRA product best-estimated rainfall in tropical cool arid (r2 = 0.97) and tropical cool semiarid (r2 = 0.53) and could, therefore, be best used for high elevation and drought-related studies. These results demonstrate the promising potential of the satellite remote sensed data in complementing the existing meteorological observed data which are often marred by inconsistency and scarcity, and hence unreliable in the existing agricultural advisory and other climate-based applications in Kenya, and sub-Saharan Africa at large. However, given the observed AEZ dependant variations in the satellite estimates, it is advisable to choose the most suitable SRS product for specific activities per AEZ and calibrate before utilisation.