Improving Satellite Derived Rainfall Products: Bayesian Approach

Abstract Advances in remote sensing have led to use of satellite derived rainfall products to compliment the sparse raingauge data. However, the indirect estimates of these products results into errors. The global and regional bias correction, have not been effective in poor raingauge network areas. This study focuses on improving seven satellite rainfall products using newly gridded raingauge data (0.05o) over East Africa for the period 1998-2012. They includes; The Precipitation Estimation From Remotely Sensed Information Using Artificial Neural Network-Climate Data Record (PERSSIAN-CDR-CDR),Tropical Applications Of Meteorology Using Satellite And Ground-Based Observations (Tamsat) African Rainfall Climatology And Time Series (TARCAT V2.0), Global Precipitation Climatology Project Version-2 (GPCP), The CPC Merged Analysis of Precipitation (CMAP), Tropical Rainfall Measuring Method (TRMM) 3B43 v.7, Climate Prediction Center (CPC) Morphing(CMORPH), and Climate Hazards Group Infrared Precipitation with Stations (CHIRPS). Each product was resampled to the raingauge spatial resolutions, and Bayesian approach was used to monthly correct each product using 10 years (1998-2007) average values. The relationship derived from training period was used to correct the products (2008-2012) in the absence of raingauge data. To quantify the performance of each product, categorical (e. FAR, POD) and continuous (e.g. correlation coefficients, mean error, Root mean squared error and standard deviation) statistics were used validation. Results indicated high performance improvements of rainfall detection (POD>0.9 and low FAR ~0) and retrieval after bias corrections, which is a proof, of the importance of local bias corrections before applications. In retrieval, before bias corrections, all products showed underestimations except CHIRPS which showed overestimation. Although the coarse resolution showed the lowest correspondence before corrections, this was associated with low gauge density within each grid. Over all, the CMORPH, CHIRPS and TRMM showed close agreements with raingauge data and may be used for climate studies over the region.