Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study introduces a hybrid LSTM-Vision Transformer (LSTM-ViT) framework designed to enhance the prediction of forecast errors in the High-Resolution Rapid Refresh (HRRR) model by integrating temporal surface observations with atmospheric profiles from the New York State Mesonet. The LSTM-ViT significantly improves predictive skill for hourly precipitation, wind speed, and temperature errors, particularly during periods of complex vertical atmospheric evolution, achieving nearly double the skill for precipitation forecasts compared to the baseline LSTM. These advancements highlight the importance of incorporating vertically structured atmospheric data to mitigate performance degradation in numerical weather prediction systems.
Forecast error prediction in high-resolution weather models can be dramatically improved by integrating atmospheric structure with a hybrid LSTM-ViT approach, yielding up to 2x better performance for precipitation forecasts.
Forecast errors in high-resolution numerical weather prediction (NWP) systems are often linked to unresolved planetary boundary layer (PBL) processes, convection, terrain-induced circulations, and other vertically structured atmospheric phenomena. Previous work demonstrated that Long Short-Term Memory (LSTM) networks can successfully predict forecast errors in the High-Resolution Rapid Refresh (HRRR) model using mesonet observations, but we believe performance degradation is linked to periods of complex vertical atmospheric evolution. To address this limitation, we develop a hybrid LSTM-Vision Transformer (LSTM-ViT) framework that combines temporal sequence learning from surface observations with atmospheric profiles from the New York State Mesonet profiler network. The LSTM-ViT framework is trained to predict HRRR hourly precipitation, 10 m wind speed, and 2 m temperature forecast errors at individual mesonet stations. Across all three predictors, incorporation of profiler-derived atmospheric structure improves forecast error prediction skill relative to the baseline LSTM architecture, with the largest gains occurring at shorter forecast lead times and during periods of enhanced PBL activity. Improvements are particularly pronounced for precipitation forecast error, where the LSTM-ViT framework achieves approximately a twofold increase in predictive skill relative to the baseline LSTM while better capturing convectively driven error evolution and reducing degradation associated with PBL processes. These results demonstrate that combining temporal sequence learning with vertically informed attention mechanisms provides a physically meaningful pathway for improving forecast error prediction in operational NWP systems. Our research offers forecasters enhanced guidance regarding model bias and forecast confidence.