2025/07/25 by Takuya Inoue, Takuya Kawabata, Inoue, Takuya +1
Earth and Planetary Sciences · Environmental Science · Computer Science · #Meteorological Phenomena and Simulations #Hydrological Forecasting Using AI #Solar Radiation and Photovoltaics
paper · pdf · doi:10.1175/mwr-d-26-0003.1
Abstract Due to limited computational resources, medium-range temperature forecasts typically rely on low-resolution numerical weather prediction (NWP) models, which are prone to systematic and random errors. We propose a method that integrates a convolutional neural network (CNN) with an ensemble of low-resolution NWP models (40-km horizontal resolution) to produce high-resolution (5 km) surface temperature forecasts with lead times extending up to 5.5 days (132 h). The trained CNN is applied to all ensemble members to produce a CNN-corrected ensemble. The CNN-based postprocessing performs both bias correction and spatial downscaling, thereby reducing systematic errors and enhancing the horizontal resolution for each ensemble member, which ultimately improves the deterministic forecast accuracy. The resulting CNN-corrected ensemble constitutes a new high-resolution ensemble forecasting system with enhanced probabilistic reliability and an improved spread–skill ratio that differs from the simple error reduction mechanism of ensemble averaging. Whereas averaging reduces forecast errors by smoothing spatial fields, our memberwise CNN correction reduces error from noise while maintaining forecast information at a level comparable to that of other high-resolution forecasts. Experimental results indicate that the proposed method provides a practical and scalable solution for improving medium-range temperature forecasts, which is particularly valuable for use in operational centers with limited computational resources. Significance Statement Reliable temperature forecasts lasting more than 5 days are vital for planning and public safety; however, current forecasting methods rely on low-resolution numerical weather prediction models that frequently misestimate temperatures. We developed a method that employs artificial intelligence to correct each member of a set of forecasts (an “ensemble”), thereby reducing systematic and random errors. This method provides more accurate, reliable, and spatially detailed ensemble forecasts than the model’s original output while remaining feasible under limited computing resources. This method can help communities better prepare for temperature-related risks by making accurate forecasts accessible to operational centers with limited resources.