Ensemble combining techniques to improve gridded satellite‑based products in precipitation estimation

Abstract

This study investigates the effectiveness of ensemble combining techniques in improving the accuracy of satellite-based gridded precipitation estimations in the Sirjan watershed, Iran. For this purpose, four satellite-based products (CHIRPS, MSWEP, PERSIANN-CDR, and PERSIANN-CCS-CDR) were analyzed over 25 years (1996–2020). Various ensemble combining techniques, including Simple Model Averaging (SMA), Weighted Averaging Model (WAM), Multi-Model Super Ensemble (MMSE), and Modified MMSE (M3SE), were implemented to generate new estimations of precipitation based on satellite-based products. The performance of the ensemble combining techniques was evaluated through a series of comparative tests. The findings indicate that the MMSE and M3SE significantly improve the accuracy of precipitation estimates compared to individual products, increasing the accuracy rate up to 48% at the annual scale. The results demonstrate that the ensemble combining techniques provide a more reliable classification of wet and dry days, resulting in a 1.1–2.6% increase in true detection. This study offers a robust framework for generating more reliable precipitation estimates in non-gauge regions.