Search

Quantitative precipitation forecasting in numerical weather prediction (NWP) models is contingent upon physicals parameterization schemes. However, uncertainties abound due to limited knowledge of the precipitating processes, leading to degraded forecasting skills. In light of this, our study explores the application of a Swin‐Transformer based deep learning (DL) model as a supplementary tool for enhancing the mapping trajectory between the NWP fundamental variables and the most downstream variable precipitation. Constrained by the observational satellite precipitation product from NOAA CPC Morphing Technique (CMORPH), the DL model serves as the post‐processing tool that can better resolve the precipitation patterns compared to solely based on NWP estimation. Compared to the baseline Weather Research and Forecasting simulation, the DL post‐processing effectively extracts features over meteorological variables, leading to improved precipitation skill scores of 21.7%, 60.5%, and 45.5% for light rain, moderate rain, and heavy rain, respectively, on an hourly basis. We also evaluate two case studies under different driven synoptic conditions and show promising results in estimating heavy precipitation during strong convective precipitation events. Overall, the proposed DL model can provide a vital reference for capturing precipitation‐triggering mechanisms and enhancing precipitation forecasting skills. Additionally, we discuss the sensitivities of the fundamental meteorological variables used in this study, training strategies, and performance limitations. Plain Language Summary: Numerical weather prediction models depend on certain empirical formulations known as parameterizations to estimate precipitation. However, these methods often fall short due to the intricate dynamics of rainfall, which involves numerous small‐scale interactions that these models are unable to fully capture. To counteract these limitations, our study deploys a form of machine learning known as deep learning (DL) to predict precipitation. This DL model utilizes fundamental weather variables derived from NWP models to make its estimations, serving as a remedy for the inherent weaknesses of traditional models caused by the uncertainties in their parameterization schemes. The implementation of our DL model resulted in a significant enhancement in rainfall prediction accuracy, particularly in the case of extreme precipitation events. This suggests that the application of machine learning strategies could be a promising approach to improve the reliability of rainfall forecasts, a crucial element for effective weather prediction and water resource management. Key Points: We employ a transformer based deep learning model to improve the accuracy of precipitation estimation in numerical weather prediction modelsVarious training strategies were implemented to manage the highly skewed precipitation data leading to improvements in heavy rainfall eventsEvaluation conducted with multiple metrics including skill score, quantile and spatial distribution as well as two case studies [ABSTRACT FROM AUTHOR]

Most state-of-the-art climate models substantially underestimate the near-surface air temperature (SAT) over the Tibetan Plateau (TP), especially for the cold season. While previous studies have attributed this cold bias to local factors such as the elevation difference, coarse resolution, and excessive snow cover, this investigation addresses the potential contributions of the systematic bias of tropical sea surface temperature (SST) to the TP cold bias. Experiments with the NCAR Community Atmosphere Model demonstrate that tropical SST bias results in an apparent cold bias over the TP, especially in boreal winter, and explains about 40% of the deviation in multi-model mean SAT over the TP relative to station observations. Forced by the tropical SST bias, heat flux exhibits an anomalous divergence over the plateau, causing a cooling center in the mid- and lower-troposphere over the TP. This atmospheric cooling in turn leads to a reduction of the downward longwave radiative fluxes reaching the surface, less energy supply, and thus a cold bias over the TP. [ABSTRACT FROM AUTHOR]