The Combined Power of Double Mass Curves and Bias Correction for the Maximisation of the Accuracy of an Ensemble Satellite-Based Precipitation Estimate Product

Abstract

Precise estimation of the spatial and temporal characteristics of rainfall is essential for producing the reliable catchment response needed for proper management of water resources. However, in most parts of the world, gauged rainfall stations are sparsely distributed and fail to properly capture the spatial variability of rainfall. Furthermore, the gauged rainfall data can sometimes be of short length or require validation. Following this, we present a procedure that enhances the trustworthiness of gauged rainfall data and the accuracy of the rainfall estimations of five satellite-based precipitation estimate (SPE) products by validating them using the 1779 gauged rainfall stations across Thailand. The five SPE products considered include CMORPH-BLD; TRMM-3B42; CHIRPS; CHIRPS-PL; and TRMM-3B42RT. Prior to validation, the gauged rainfall dataset was verified using double mass curve (DMC) analysis to eliminate questionable and inconsistent readings. This led to the improvement of the Nash–Sutcliffe Efficiency (NSE) between the station of interest and its surroundings by 13.9% (0.758–0.863), together with an average 11.8% increase with SPE products, whilst dropping only 7% of questionable dataset. Three different bias correction (BC) procedures were applied to correct SPE products using gauge-based gridded rainfall (GGR). Once DMC and BC procedures were implemented together, the performance of the SPE products was found to increase significantly. Finally, the application of the ensemble weighted average of the three best-performing bias-corrected SPE products (Bias-CMORPH-BLD, Bias-TRMM-3B42, and Bias-CHIRPS) further enhanced the NSE to 0.907 and 0.880 in calibration and validation time periods, respectively. The proposed DMC-based correction SPE and the weighting procedure of multiple SPE products allows for an easy means of obtaining daily rainfall in remote locations with sufficient accuracy.