Chongbo et al. (2025) Extreme rainfall south of the Yangtze River in China during June 2024: Observational diagnosis and dynamical downscaling prediction

Identification

• Journal: Atmospheric Research
• Year: 2025
• Date: 2025-11-04
• Authors: Zhao Chongbo, Lili Dong, Bing Xie, Wenzhong Shi, Anran Wang, Jie Wu, Qingquan Li
• DOI: 10.1016/j.atmosres.2025.108616

Research Groups

• China Meteorological Administration Key Laboratory for Climate Prediction Studies, National Climate Centre, Beijing, China
• Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science and Technology, Nanjing, China
• Department of Earth System Science, Tsinghua University, Beijing, China

Short Summary

This study investigates the physical mechanisms behind the extreme rainfall south of the Yangtze River in June 2024 and evaluates the performance of dynamical downscaling for its prediction. It finds that combined external sea surface temperature forcing and internal atmospheric variability drove the record rainfall, and a regional climate model significantly improved spatial prediction by correcting global model biases, primarily through better representation of lower-tropospheric meridional wind.

Objective

• To investigate the physical forcing associated with the unprecedented extreme precipitation event in regions south of the Yangtze River during June 2024.
• To demonstrate the clear advantage of Climate-Weather Research and Forecasting (CWRF) dynamical downscaling over using a global climate model alone in predicting this event.
• To determine the extent to which the CWRF model skillfully predicted this extreme rainfall, identify the sources of its prediction skill, and explain its remaining limitations.

Study Configuration

• Spatial Scale: Regions south of the Yangtze River (SYR) in China, eastern China, Hunan–Hubei–Jiangxi border area, Dongting Lake (111°E–114°E, 28°N–30°N), Bay of Bengal and South China Sea (10°N–25°N, 80°E–130°E). Global climate model (BCC_CPSv3) resolution of approximately 45 km (T266). Regional climate model (CWRF) resolution of 15 km horizontally (462x342 grid points) with 36 vertical layers up to 50 hPa. U-Net model input fields of 128x128 grid cells.
• Temporal Scale: Extreme rainfall event in June 2024. Historical precipitation data since 1961. Model hindcasts and SST index analyses from 2008 to 2024 (or 2009 to 2023 for anomalies). Forecasts for June 2024 were initialized from May 8 to May 22, 2024, providing 10–25 days lead time. Intraseasonal oscillations (BSISO1: 30–60 days, BSISO2: 10–30 days).

Methodology and Data

• Models used:
  • Global Climate Model: Beijing Climate Center–climate prediction system version 3 (BCCCPSv3).
  • Regional Climate Model: Climate-Weather Research and Forecasting (CWRF) model, nested with BCCCPSv3.
  • Deep Learning Model: U-Net model, designed for circulation–precipitation nonlinear relationships and occlusion sensitivity analysis.
• Data sources:
  • Observational: Basic Chinese surface meteorological data (from >2400 national stations since 1951), Global Precipitation Climatology Project (GPCP) precipitation dataset (2.5° x 2.5°), NOAA Optimum Interpolation (OI) version 2 SST dataset (0.25° x 0.25°).
  • Reanalysis: China’s first-generation global atmospheric and land reanalysis (CRA-40; 0.25° x 0.25°), European Centre for Medium-Range Weather Forecasts (ECMWF) analysis (for BCC_CPSv3 initial conditions).
  • Satellite: Daily outgoing longwave radiation (OLR) Climate Data Record (CDR) from NOAA TIROS-N series and MetOp satellites (2.5° x 2.5°), OLR data from Fengyun-3D (FY-3D) satellite.

Main Results

• June 2024 experienced record-breaking precipitation south of the Yangtze River (SYR), with rainfall near Dongting Lake exceeding 450 mm, the highest since 1961.
• The extreme rainfall was driven by a synergistic combination of external sea surface temperature (SST) forcing and internal atmospheric variability.
• SST Forcing: Elevated SSTs in the tropical North Atlantic (1.19 °C anomaly) and Indian Ocean (0.57 °C anomaly), coupled with cold SSTs in the eastern equatorial Pacific (Niño3 index -0.35 °C), contributed to the event. The Indian Ocean played the most dominant role. A triple-linear regression using these three SST indices explained approximately 86% of the June 2024 SYR precipitation anomaly.
• Atmospheric Variability: The Boreal Summer Intraseasonal Oscillation (BSISO), particularly BSISO1 in late June, triggered and intensified subseasonal precipitation anomalies. When superimposed on SST-forced components, combined BSISO indices explained approximately 97% of the total observed precipitation anomaly.
• Model Performance:
  • The global climate model (BCCCPSv3) failed to accurately capture the spatial distribution of the heavy rainfall belt, showing a northward deviation and negative anomaly correlation coefficients (ACCs) over eastern China. Its biases were linked to flawed SST-precipitation relationships and suppressed BSISO intensity due to cooler-than-observed Indian Ocean SSTs.
  • Dynamical downscaling with the regional climate model (CWRF) significantly improved the spatial distribution of precipitation prediction, correcting the global model's northward-displaced rain belt. The ACC over eastern China improved from -0.32 (BCCCPSv3) to 0.02 (CWRF) for June 2024.
  • CWRF's improved performance was attributed to a more realistic representation of warmer SSTs in the Bay of Bengal and the South China Sea, and their relationship with precipitation.
• U-Net Diagnosis:
  • U-Net-based occlusion sensitivity analysis revealed that CWRF's prediction skill for SYR precipitation primarily stemmed from its accurate representation of lower-tropospheric meridional wind (at 850 hPa, 700 hPa, and 500 hPa).
  • However, the simulation of lower-tropospheric humidity (especially 850 hPa relative humidity) remains a major source of prediction errors. Replacing CWRF's 850-hPa relative humidity field with reanalysis data boosted the ACC by 0.32.
  • Sliding-window experiments highlighted that precipitation prediction accuracy was most sensitive to humidity errors in the SYR, meridional wind anomaly shifts near the Yangtze River, and cold SST biases in the South China Sea.

Contributions

• Provides a comprehensive mechanistic analysis of the record-breaking June 2024 extreme precipitation in China, elucidating the synergistic effects of remote SST forcing from the Tropical North Atlantic, Indian Ocean, and Pacific, and subseasonal internal variability from BSISO.
• Highlights the critical importance of accurately representing multi-scale ocean-atmosphere interactions in models for improved predictability of extreme events.
• Demonstrates the significant value of high-resolution dynamical downscaling using regional models (CWRF) in correcting precipitation biases from global driving models.
• Introduces an innovative application of deep learning (U-Net) for rigorously attributing sources of prediction skill and error in subseasonal precipitation forecasts.
• Identifies key atmospheric variables (low-level humidity and meridional wind) that are crucial for targeted model development, emphasizing the need to enhance the representation of ocean-atmosphere coupling and moisture transport processes to advance extreme rainfall forecasting.

Funding

• National Natural Science Foundation of China (U2242207, 42575020, 41790471)
• National Key Research and Development Program of China (2022YFE0136000, 2023YFF0805104)
• Natural Science Foundation of Anhui Province of China (2208085UQ08)
• Innovative Development Special Project of the China Meteorological Administration (CXFZ2023J003)
• Joint Research Project for Meteorological Capacity Improvement (24NLTSQ016)
• China Meteorological Administration (CMA) Youth Innovation Group (CMA2024QN06)

Citation

@article{Chongbo2025Extreme,
  author = {Chongbo, Zhao and Dong, Lili and Xie, Bing and Shi, Wenzhong and Wang, Anran and Wu, Jie and Li, Qingquan},
  title = {Extreme rainfall south of the Yangtze River in China during June 2024: Observational diagnosis and dynamical downscaling prediction},
  journal = {Atmospheric Research},
  year = {2025},
  doi = {10.1016/j.atmosres.2025.108616},
  url = {https://doi.org/10.1016/j.atmosres.2025.108616}
}