Evaluation of four meteorological reanalysis datasets for satellite-based PM2.5 retrieval over China

Abstract

Meteorological reanalysis data is widely used for satellite-based retrieval of fine particulate matter (PM2.5); however, selecting appropriate data for specific regional applications remains a challenge. In China, numerous studies have used different meteorological reanalysis datasets for PM2.5 modeling, but few studies have provided their standards for data selection. In this study, we first evaluated the performance of four reanalysis datasets in China: the fifth-generation ECMWF Atmospheric Reanalysis (ERA5), the National Aeronautics and Space Administration Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA2), the GEOS-FP Atmospheric Data Assimilation System, and the Final Operational Model Global Tropospheric Analyses (FNL). The validation results show that ERA5 was the most accurate in terms of temperature, relative humidity, wind speed, and boundary layer height, whereas FNL had the highest uncertainty. The spatial accuracy of all four reanalysis datasets in the eastern region of China was notably higher than that in the western region (Qinghai-Tibet Plateau), owing to complex terrain and few ground-based observations for assimilation. We found that the use of ERA5 exhibited the best agreement with in situ measurements for PM2.5 retrieval, followed by MERRA2 and GEOS-FP, while FNL was the most inaccurate. Compared with FNL, ERA5 exhibited improved accuracy in retrieving PM2.5 during summer seasons by 12%, and reduced underestimation on heavy pollution days. Our results demonstrated that the ERA5 meteorological reanalysis is preferable for better PM2.5 retrieval in China. This study provides an actional guideline for selecting meteorological reanalysis data in subsequent satellite-based PM2.5 retrieval studies in China.