COMPARISON OF BIAS CORRECTION METHODS FOR APPLICATION OF SATELLITE RAINFALL DATA AND EFFECT OF STREAMFLOW IN ZIWAY-SHALLA LAKE SYSTEM, ETHIOPIA

Abstract

Rainfall data availability is a major constraint to the performance of rainfall-runoff modeling. However, satellite rainfall estimates have the potential to fill data gaps. Several previous studies have applied different bias correction methods to remove the systematic error of satellite rainfall products. However, most studies arbitrarily selected single bias correction algorithms without comparing the performance of multiple methods. In this study, the bias correction and the performance of Climate Hazards Group InfraRed Precipitation (CHIRP) satellite estimate were applied in Ziway-Shalla and rainfall-runoff simulation at Meki catchments were evaluated using the Hydrologiska Byråns Vattenbalasavdelning (HBV) hydrological model. The objective of this work is to increase the usability of the CHIRP rainfall product by reducing its systematic error using the best performing bias correction method. Five bias correction methods were compared to correct CHIRP bias using rain gauge data as a reference with four temporal scales. The performance of the methods was evaluated at daily,14-days, monthly, and seasonal scales. Correlation (R2), Mean Error (ME), Percent of bias (Pbias), and Root Mean Squared Error (RMSE) were used as performance measures. The Gamma distribution (QMG) and Power transformation (PT) methods performed relatively better with high correlation (0.99) and small Pbias (0.007) whereas the Distribution transform (DT) bias-correction method performed poorly in reducing the bias of the CHIRP data. The accuracy of the bias corrected rainfall data showed spatial variation across the sub-basin. Most of the methods did not perform well in capturing the spatial pattern of annual average rainfall and its temporal variation. The raw CHIRP data led to a large volume error (10.59%) of the streamflow which was simulated by the HBV model. However, the model performance improved in capturing the volume (RVE<5%) and pattern (NSE>0.7) of the streamflow using bias-corrected CHIRP rainfall estimates which were corrected using five bias correction methods. However, parameter values changed during calibration when the source of the rainfall data was changed. Overall, all bias correction methods reduced the bias of CHIRP data but there is a large difference between the methods in terms of their performance over the study area.