Document Type : Research Paper

Authors

• Mahdi Naseri ¹
• Mitra Noorizehab ²

¹ Assistant Professor, Faculty of Engineering, Department of Civil Engineering, University of Birjand, Birjand, Iran

² Ph.D. Candidate of Water Resources Engineering and Management, Department of Civil Engineering, Faculty of Engineering, University of Birjand, Birjand, Iran

Abstract

Accurate precipitation estimation in hyper-arid regions is fundamentally challenged by sparse observational networks and complex atmospheric dynamics. This study evaluates and corrects the errors of IMERG (Final V07) and GSMaP satellite precipitation products using 18 years (2005–2023) of daily synoptic data from Birjand, representing the hyper-arid climate of eastern Iran. Baseline evaluations indicated that GSMaP outperformed IMERG in continuous metrics (RMSE = 3.25 vs. 4.40 mm/day); however, both exhibited systematic underestimation, primarily driven by sub-cloud evaporation (the Virga effect). Categorically, IMERG demonstrated higher detection sensitivity (POD = 0.763), whereas GSMaP more effectively minimized false alarms. Bivariate density analysis revealed a notable finding: absolute estimation errors are significantly driven by surface thermodynamic conditions (maximum temperature and relative humidity, P-value < 0.01), while the dynamic impact of wind speed was statistically insignificant. Finally, applying a multiple linear regression (MLR) bias correction framework incorporating these meteorological covariates successfully reduced IMERG's RMSE by 14.1%. The findings demonstrate that integrating surface thermodynamic data with satellite retrieval algorithms via regression models substantially mitigates precipitation uncertainties in data-scarce hyper-arid basins.

Keywords

• Satellite precipitation products
• Error evaluation
• Multiple linear regression
• Hyper-arid climate