Multi-Model Ensemble Forecasts of Surface Air Temperatures in Henan Province Based on Machine Learning

Based on the China Meteorological Administration Land Data Assimilation System (CLDAS) reanalysis data and 12–72 h forecasts of surface (2-m) air temperature (SAT) from European Centre for Medium-Range Weather Forecasts (ECMWF) three numerical weather prediction (NWP) models (CMA-GFS, CMA-SH, CMA-MESO), multi-model ensemble are conducted with a convolutional neural network (CNN) feed-forward (FNN) to improve SAT forecast in Henan Province, China. The results show that there large errors CMA, while ECMWF outperforms other raw NWP models, especially eastern southern Henan. CNN has best short-term forecasting skills. difference geographical distribution is small, without any apparent large-value areas. shows its advantages bias correction mountainous region (western Henan), indicating can capture spatial features atmospheric fields therefore more robust regions varied topography. In addition, extract through convolution kernel focus local features; it assimilate at higher level obtain global features. Therefore, takes advantage four further improves skill.