Evaluation of Bias correction Approaches for Daily Satellite-based Rainfall Estimates in the Upper Zambezi, Africa

The Zambezi Basin is located in the semi-arid region of southern Africa and it is one of the largest basins in Africa. The Upper Zambezi River Basin (UZRB) above Katima Mulilo is sparsely gaged (only 11 rain gauges are currently accessible) and real-time rainfall estimates are not readily available. On the other hand, Satellite Precipitation Products (SPPs) offer the opportunity to complement that information, thereby allowing for a real-time forecasting of streamflows. In this study we evaluate, correct, and validate three SPPs for the UZRB: (1) CMORPH (Climate Prediction Center’s morphing technique), (2) PERSIANN (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks), and (3) TRMM-3B42RT (Tropical Rainfall Measuring Mission). The bias is initially assessed by using raingauges in a point-to-pixel comparison for daily, monthly and yearly estimates of precipitation. Then all SPPs are bias-corrected using the CHIRPS rainfall estimates as reference data. For these purposes two Bias Correction (BC) methods are applied and their performance is evaluated for the period 2001-2016. The methods include the well-known quantile-mapping and a proposed method using a Principal Components-based correction. The ultimate goal of this bias-correction is to use the corrected SPPs as input for real-time, short-time (8 days ahead) and seasonal (180 days) forecasts in the UZRB upstream of Katima Mulilo.