A ten-year meteorological simulation and optimization in China based on traditional data assimilation and machine learning methods

This dataset contains meteorological input fields derived from METCRO2D files, generated using the WRF model and processed through MCIP. It consists of two distinct data components: Long-term simulations over mainland China (2014–2023): These simulations were conducted at a spatial resolution of 27 km, centered at 35°N and 105°E, and represent the baseline meteorological fields without any further optimization. A high-resolution case study for the Yangtze River Delta (YRD) in 2023: This regional case features WRF simulations at resolutions of 9 km, 3 km, and 1 km. Additional optimization was applied using three-dimensional variational data assimilation (3DVAR) and machine learning models, including Random Forest (RF) and XGBoost. The dataset consists of the following files: 1. METCRO2D_DailyAvg.tar.gz This archive contains meteorological input fields from METCRO2D files, in which all available variables have been temporally aggregated into daily averages. These data are suitable for meteorological and air quality modeling applications, and provide a basis for evaluating the effectiveness of both data assimilation and machine learning approaches in enhancing regional meteorological simulations. 2. METCRO2D_TEMP2_WSPD10_WDIR10.tar.gz This archive includes three key surface meteorological variables extracted from the METCRO2D files: TEMP2: 2-meter air temperature WSPD10: 10-meter wind speed WDIR10: 10-meter wind direction These selected variables enable targeted analysis of near-surface meteorological conditions and facilitate comparative assessments of optimization techniques, particularly in regions with complex meteorological and topographic features such as the YRD.