Li et al. (2025) Subseasonal prediction of early summer precipitation in the middle and lower reaches of the Yangtze River Basin based on circulation classification

Identification

• Journal: Atmospheric Research
• Year: 2025
• Date: 2025-10-27
• Authors: Mei Li, Yunyun Liu, Jinqing Zuo, Yinghan Sang, Jiaxi Yang
• DOI: 10.1016/j.atmosres.2025.108596

Research Groups

• State Key Laboratory of Climate System Prediction and Risk Management/China Meteorological Administration Climate Studies Key Laboratory, National Climate Centre, China Meteorological Administration, Beijing, China
• Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science &Technology, Nanjing, China
• State Key Laboratory of Severe Weather Meteorological Science and Technology, and Institute of Tibetan Plateau Meteorology, Chinese Academy of Meteorological Sciences, Beijing, China
• Institute of Urban Meteorology, China Meteorological Administration, Beijing, China

Short Summary

This study develops and evaluates Self-Organizing Map (SOM)-based statistical downscaling models to improve subseasonal precipitation predictions in the middle and lower reaches of the Yangtze River Basin (MLYRB) by establishing nonlinear relationships between circulation patterns and precipitation probability distributions. The models significantly enhance the accuracy and temporal predictability of early summer subseasonal precipitation, showing strong potential for operational regional climate prediction.

Objective

• To develop and evaluate Self-Organizing Map (SOM)-based statistical downscaling models for subseasonal prediction of early summer precipitation in the Middle and Lower Reaches of the Yangtze River Basin (MLYRB), by identifying key circulation patterns and establishing nonlinear relationships between these patterns and daily precipitation probability distributions.

Study Configuration

• Spatial Scale: Middle and Lower Reaches of the Yangtze River Basin (MLYRB) (109°–124°E, 27°–32.5°N), utilizing data from 56 meteorological stations. Large-scale circulation data (500 hPa geopotential height) covered 40°–160°E and 0°–70°N. Data resolutions were unified to 1.25° × 1.25°.
• Temporal Scale: Early summer (June).
  • Training period: 1961–1990.
  • Independent testing period: 1991–2002.
  • Evaluation period: 2015–2022.
  • Prediction scales: Monthly and pentad (5-day) mean precipitation.
  • Forecast lead times: Evaluated up to two weeks for monthly predictions and up to one month for pentad predictions.

Methodology and Data

• Models used:
  • Self-Organizing Map (SOM) neural network for circulation classification and statistical downscaling.
  • Monte Carlo random simulation method for generating precipitation statistics.
  • NCEP Climate Forecast System Version 2 (CFSv2) dynamic climate model outputs were used to drive the downscaling models.
• Data sources:
  • Observational: Daily precipitation records from more than 2400 meteorological stations in China (56 stations selected in MLYRB) for 1961–2002 and 2015–2022, provided by the China Meteorological Administration.
  • Reanalysis: ERA5 daily reanalysis data (500 hPa geopotential height) from the European Centre for Medium-Range Weather Forecasts (ECMWF) for 1961–2002, with an initial spatial resolution of 0.25° × 0.25°, unified to 1.25° × 1.25°.
  • Model outputs: Ensemble means of hindcasts and real-time predictions from the NCEP Climate Forecast System Version 2 (CFSv2) for January 1982–December 2022, with an initial spatial resolution of 1.5° × 1.5°, unified to 1.25° × 1.25°.

Main Results

• The SOM method successfully identified key circulation patterns, primarily reflecting east-west shifts of the Western Pacific Subtropical High (WPSH), which influence early summer monthly and pentad rainfall variability.
• A clear nonlinear relationship was established: a westward WPSH shift increases the probability of heavy precipitation, while an eastward retreat is associated with light rainfall.
• Significant improvements in spatial prediction accuracy after downscaling:
  • For monthly predictions, the proportion of years with statistically significant spatial correlation coefficients (PCCs) increased from 60% (before downscaling) to 95% (after downscaling). Multi-year averaged PCCs nearly doubled, increasing by approximately 0.3 (to around 0.6).
  • For pentad predictions, the proportion of years with statistically significant PCCs increased from 53% to 92%. Multi-year averaged PCCs nearly doubled, increasing by approximately 0.2.
  • Substantial improvements were also observed in normalized standard deviations (approaching 1) and centered root mean square errors (falling below 0.5 for monthly precipitation).
• Extended predictable lead times: The downscaling approach promoted consistency across lead times within two weeks for monthly precipitation predictions and within one month for pentad predictions.
• Limited improvement in temporal variability: While spatial patterns were significantly enhanced, the temporal correlation coefficients (TCCs) for precipitation anomalies, particularly extreme events, largely failed to reach statistical significance, indicating modest gains in predicting intensity and timing of extreme precipitation.

Contributions

• This study provides the first systematic evaluation of the applicability and effectiveness of a Self-Organizing Map (SOM)-based downscaling approach at the subseasonal scale for early summer precipitation in the MLYRB.
• It demonstrates a significant enhancement in both the accuracy of spatial precipitation patterns and the predictable lead times for subseasonal forecasts, addressing limitations of dynamic models.
• The research establishes robust, nonlinear, and physically realistic empirical relationships between daily circulation patterns (specifically WPSH shifts) and observed precipitation probability distributions.
• The developed SOM-based statistical downscaling models offer a valuable framework with strong potential for operational regional climate prediction and early warning capabilities in the MLYRB.

Funding

• National Key Research and Development Program of China (2023YFC3007503)
• National Natural Science Foundation of China (42175056)
• Jianghuai Meteorological Joint Project of Anhui Natural Science Foundation (2208085UQ10)
• National Climate Center Innovation Team (NCCCXTD003)
• Youth Innovation Team of China Meteorological Administration (CMA2023QN15, CMA2024QN06)

Citation

@article{Li2025Subseasonal,
  author = {Li, Mei and Liu, Yunyun and Zuo, Jinqing and Sang, Yinghan and Yang, Jiaxi},
  title = {Subseasonal prediction of early summer precipitation in the middle and lower reaches of the Yangtze River Basin based on circulation classification},
  journal = {Atmospheric Research},
  year = {2025},
  doi = {10.1016/j.atmosres.2025.108596},
  url = {https://doi.org/10.1016/j.atmosres.2025.108596}
}