A comparative study of fuzzy methods for improving satellite-derived precipitation datasets using novel indices for improvability and efficiency

Study Region Southwestern Iran, an arid to semi-arid region with complex topography and variable precipitation. Study Focus This research introduces a novel framework for improving the accuracy of satellite-derived precipitation datasets (SDPDs) using fuzzy inference systems (FIS). Two new indices are developed: the Improvable Data Ratio (IDR), which quantifies a dataset's potential for meaningful correction, and the Efficiency Factor (EF), which integrates IDR with residual error to evaluate the combined reliability of the satellite dataset and the correction model. Three FIS-based models—ANFIS, MamFIS, and W-MenFIS—are applied to enhance the GPM_3IMERGDL product, identified as the most improvable dataset among four candidates. Performance is assessed using Mean Error, RMSE, Relative Bias, and EF across diverse climatic zones. New Hydrological Insights for the Region The results demonstrate that MamFIS offers the most robust and consistent correction performance across varying precipitation regimes, achieving an EF above 0.79 in wet zones (MAP > 800 mm ). While ANFIS excels in arid areas, it tends to overcorrect in wetter climates. W-MenFIS achieves low RMSE in dry regions but introduces systematic underestimation elsewhere. This study underscores the importance of pre-correction assessment using IDR and highlights MamFIS as a reliable model for improving SDPDs in topographically and climatically diverse regions like southwestern Iran.