Document Type : Research Paper
Authors
1 Ph.D., Department of Water Engineering, Sari Agricultural Sciences and Natural Resources University, sari, Iran. ORCID: https://orcid.org/0000-0002-7941-4948 , Email: alirezau3fikrbriya@gmail.com
2 Associate Professor, Department of Agricultural Meteorology, Department of Water Engineering, Faculty of Agricultural Engineering, Sari University of Agricultural Sciences and Natural Resources, Sari, Iran. (Email: Mehdi.nadi@gmail.com)
Abstract
High-resolution precipitation data in both temporal and spatial domains are considered fundamental inputs in a wide range of environmental and hydrological studies. In this study, precipitation data from three satellite-based products (CHIRPS, GPM, and PERSIANN) were extracted using Google Earth Engine and evaluated against ground-based station observations in Khuzestan Province for the years 2010, 2018, and 2022 at a monthly time scale. A set of statistical indicators, including RMSE, MAE, MBE, Pearson correlation coefficient, R², and Kendall’s tau, was employed to assess the accuracy of the products.
The results indicated that CHIRPS and PERSIANN exhibited higher absolute accuracy compared to GPM. The RMSE values ranged from 17.3 to 22.8 mm for CHIRPS and from 15.3 to 25.9 mm for PERSIANN, whereas GPM showed a substantially higher RMSE of approximately 97.1 mm in the wet year of 2018. Bias analysis also revealed a significant overestimation in GPM, with MBE values ranging between 30.8 and 57.6 mm, while CHIRPS showed a near-zero bias and PERSIANN demonstrated a moderate positive bias.
In terms of temporal correlation, GPM achieved the highest Pearson correlation coefficients (up to 0.92) and R² values (up to 0.85), although this strong agreement in pattern was accompanied by large absolute errors. Kendall’s tau values above 0.65 for all three products indicate their general ability to capture precipitation temporal dynamics. Multivariate analyses further revealed that the performance of satellite products is highly dependent on rainfall intensity; CHIRPS and PERSIANN showed more stable performance under low to moderate rainfall conditions, whereas GPM consistently exhibited overestimation across all rainfall regimes.
Accordingly, CHIRPS and PERSIANN can be considered more reliable options for quantitative applications in Khuzestan Province, while the use of GPM requires bias correction tailored to local climatic conditions.
Keywords
)