Evaluation and Comparison of Satellite Rainfall Products in Mountainous Areas with Lack of Meteorological Data of Lorestan Province

Introduction The large temporal and spatial changes of precipitation, especially in mountainous areas, have turned it into a controversial variable in climate models. Measuring precipitation (rain and snow) along with its distribution and changes is very important to improve our understanding of global water cycle and energy, water resources monitoring, hydrological modeling. Lack of reliable data is one of the most important challenges in rainfall analysis. Due to the significant temporal and spatial variability of precipitation in mountainous areas, accurate spatially distributed data is crucial for effective water resource assessment and management. However, many mountainous regions have limited rain gauge stations. Today, satellite products are commonly used to measure precipitation in these areas, but the variability among these products raises concerns about their accuracy in mountainous regions. Additionally, the quality of satellite products differs between various products and across different climatic regions, making it essential to thoroughly evaluate them before use. The purpose of this research was to evaluate the precipitation data of two satellite products (GPM, PERSIAN) and reanalysis data (ECMWF) in the estimation of precipitation in mountainous areas without stations in Lorestan province. Method This study utilized rainfall data from 24 synoptic and rain gauge stations across Lorestan province. Emphasis was placed on stations situated in or near mountainous regions. The selected stations were chosen based on their suitable spatial distribution and record length. The rainfall data spanned the period from 2015 to 2021 and included daily, monthly, and annual measurements. To evaluate satellite rainfall algorithms and estimate rainfall in regions with limited data, data from the GPM and PERSIAN satellites were employed, along with ECMWF reanalysis data. The PERSIAN rainfall algorithm is a remote sensing-based method that utilizes artificial neural networks. It calibrates infrared data with passive microwave estimates and converts longwave infrared images into rainfall estimates using a three-step process. The spatial resolution of this product is 0.25° x 0.25°, and it offers hourly, daily, and monthly temporal resolution. The PERSIAN rainfall algorithm data can be accessed from https://chrsdata.eng.uci.edu. The GPM mission aims to provide continuous observations of Earth's precipitation. It employs the GPM Microwave Imager (GMI) and Dual-frequency Precipitation Radar (DPR) to observe both snow and rain. The final product, called IMERG, is generated through multiple runs of the algorithm for each observation time. Initial estimates are quickly provided, and subsequent estimates improve as more information becomes available. The spatial resolution of the GPM product is 1° x 1°, and it offers hourly, daily, and monthly temporal resolution. IMERG data can be obtained from https://gpm.nasa.gov/data. CMWF reanalysis data is derived from the combination of short-term simulations of numerical weather prediction models with ground-based observational data. These simulations are controlled with observational data, and the resulting reanalysis database provides global coverage from 1979 with a spatial resolution ranging from 0.125° x 0.125° to 3°. The temporal resolution of ECMWF reanalysis data is hourly, daily, and monthly. More information about ECMWF data can be found at https://www.ecmwf.int/ (Azizi, 2019). To evaluate the accuracy of the products, R-squared correlation (R2), root mean square error (RMSE), standard deviation (MAD), correlation coefficient (R), error deviation (MBE) and Nash-Sutcliffe coefficient (NS) were used. Also, the probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) indices were used to validate the data. Results The results showed that none of the three products are suitable for estimating daily precipitation in mountainous areas. However, on a monthly scale, these products provide reasonable estimates. Among the three, the GPM satellite product demonstrated better accuracy on a monthly scale, based on error levels and the spatial distribution of estimated precipitation. On an annual scale, GPM also performed best, as indicated by both statistical errors and the spatial patterns of average annual precipitation. According to the MBE index, on daily and monthly scales, the ECMWF product tended to overestimate precipitation, while the PERSIANN and GPM products underestimated it. On an annual scale, GPM and ECMWF products overestimated precipitation, whereas PERSIANN underestimated it.