A comprehensive evaluation of the accuracy of satellite-based precipitation estimates over Thailand

Thailand This study assessed the performance of fifteen SPE products from the TRMM, GPM IMERG, CHIRP, CMORPH, and PERSIANN families, using 1779 gauged rainfall measurements, covering the period from 2001 to 2015, after undergoing fourteen quality control steps. Various statistical indicators ranked fifteen candidates with normalized scores, revealing performance differences across regions and products. This analysis helps to identify the most suitable products for specific objectives. It also evaluates the performance of SPEs across changes in elevation, rainfall probability, and data-scarce conditions by progressively reducing the amount of gauged rainfall data used. Finally, an ensemble product was developed. Firstly, CMORPH-BLD, CHIRPS-Preliminary, TRMM-3B42, and GPM IMERG-Final are the top four performers. Secondly, rainfall estimates from the SPE products were generally underestimated at rare probabilities and overestimated at more frequent ones. Their accuracy was slightly lower at elevations below 200 m, improved up to 800 m with consistent accuracy, and then decreased above 800 m, with minimal gauged rainfall stations increasing uncertainties in assessing SPE product performance at higher elevations. Thirdly, bias correction improved the performance of all SPEs, and accuracy gains were still achievable even with reduced gauged rainfall data. The ensemble of multiple SPEs reduced individual errors, demonstrating their reliability in representing precipitation in Thailand. These findings confirm the robustness of fixed simulation weights for future applications. • CHIRPS-PL is suggested due to its high accuracy, fine resolution and short latency. • Satellite Precipitation Estimates are underestimated, especially at low probabilities. • The accuracy of SPE stabilizes up to 800 m and decreases with increasing elevation. • Bias correction improves SPE accuracy, even in areas with limited rainfall gauges. • Fixed weights for each SPE product can be used to create future ensemble products.