Surface gust observations, station-interpolated WRF predictors, and model code for northwestern China, southeastern China, and the European Alps

This dataset was compiled for developing a deep-learning-based gust estimation scheme (DL-GUST). It provides surface gust observations and WRF-derived predictor datasets used as model input factors. The WRF simulations were driven by ERA5 reanalysis data. Model development is based on the northwestern inland region of China during 2020-2021, and external validation can be performed over the European Alps in February 2020 and the southeastern coastal region of China in August 2020 using stations unseen during training. The dataset provides 15 hourly WRF predictors representing dynamical, thermodynamic-moisture, and boundary-layer-process conditions. U, V, P, Z, q, T, Td, and RH are taken from the 10-m level of the WRF postprocessed height-level output (wrfpress); HFX, LH, UST, and PBLH are taken from the standard WRF history output (wrfout); TKE is derived from QKE at the lowest model level; and VWS is computed from wind components near 500 and 1500 m above ground level. All variables were bilinearly interpolated to station locations before use. The package includes three processed station observation files in CSV format and three WRF predictor files in NetCDF format. Original NOAA ISD observations and ERA5 reanalysis data are available from their official public archives and are not redistributed here.