Evaluation of High‐Resolution Downscaling Predictions for the July 2023 Extreme Rainstorm in the Beijing‐Tianjin‐Hebei Region Based on <scp>CMA</scp>‐<scp>CPSv3</scp>

ABSTRACT The integration of weather and climate prediction represents the current frontier in the development of numerical modeling in China. Dynamic downscaling serves as a pivotal approach, improving the performance and resolution of global climate models to the weather scale. Focusing on the ‘23.7’ extreme rainstorm (July 29, 00:00–August 2, 00:00 UTC) in the Beijing‐Tianjin‐Hebei region, this study assesses predictions from the China Meteorological Administration Climate Prediction System version 3 (CMA‐CPSv3, 45 km resolution) and 9‐km dynamic downscaling hindcasts from the Weather Research and Forecasting model (WRF‐9 km). In contrast to the conventional climate anomaly approaches, direct outputs are used for evaluation, similar to weather forecasting tests. By examining, both the CMA‐CPSv3 predictions and the WRF‐9 km hindcasts provide a 5‐day prediction window for this rainstorm. They successfully predict the rainstorms and related atmospheric circulations from July 24th onward, aligning with observed and reanalyzed data. WRF‐9 km, with the higher resolution and optimised physical processes, outperforms CMA‐CPSv3, especially in precipitation spatial distribution and center intensity. The WRF‐9 km 7/24 hindcast demonstrates the most significant enhancement compared to the corresponding CMA‐CPSv3 prediction. This improvement is notably reflected in the substantial increase in spatial correlation, from 0.68 to 0.79, as well as a reduction in the difference of center values, decreasing from −51% to −20%. Furthermore, the WRF‐9 km 7/24 hindcast improves the Critical Success Index by 0.08, the Success Rate by 0.08, and the Probability of Detection by 0.29 for heavy rainfall (over 25.0 mm/d). However, improvements in large‐scale circulations with WRF‐9 km are limited, which may restrict advancements in predictability. In conclusion, the WRF‐9 km enhances the performance and resolution of CMA‐CPSv3 predictions, which can be regarded as a viable pathway for CMA‐CPSv3 to achieve weather‐climate integration.