Tong et al. (2026) Maximum Temperature Forecasts in the Yangtze River Valley Using a Predictor‐Corrector Lightweight Deep Learning

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Geophysical Research Letters
• Year: 2026
• Date: 2026-07-22
• Authors: Xuan Tong, Wen Zhou
• DOI: 10.1029/2025gl117592

Research Groups

• Yangtze River Valley Weather Forecasting Team
• Department of Atmospheric Sciences, University of Science and Technology of China
• European Centre for Medium-Range Weather Forecasts (ECMWF)

Short Summary

This study introduces a lightweight maximum temperature forecasting framework that combines a ResNet-based predictor with a multilayer-perceptron-based corrector, achieving improved accuracy compared to existing numerical weather prediction methods.

Objective

• Investigate the feasibility of using deep learning for accurate maximum temperature forecasting in the Yangtze River Valley

Study Configuration

• Spatial Scale: Regional (Yangtze River Valley)
• Temporal Scale: Short-term (up to 10 days)

Methodology and Data

• Models used: ResNet-based predictor, multilayer-perceptron-based corrector
• Data sources: Historical weather data from ECMWF reanalysis

Main Results

• The proposed framework achieved up to 7.8% accuracy improvement compared to individual predictors
• It demonstrated similar RMSE but better spatial correlation and tolerance-based accuracy compared to EC forecasts
• The framework outperformed EC forecasts by predicting heatwave onset 9 days earlier with a one-day lead

Contributions

• This study provides a reliable and cost-effective tool for extreme weather forecasting in the Yangtze River Valley
• It demonstrates the effectiveness of coupling a corrector with a predictor to enhance the forecasting capabilities of lightweight models

Funding

• National Natural Science Foundation of China (Grant No. 51979023)
• European Union's Horizon 2020 research and innovation program (Grant Agreement No. 776186)

Citation

@article{Tong2026Maximum,
  author = {Tong, Xuan and Zhou, Wen},
  title = {Maximum Temperature Forecasts in the Yangtze River Valley Using a Predictor‐Corrector Lightweight Deep Learning},
  journal = {Geophysical Research Letters},
  year = {2026},
  doi = {10.1029/2025gl117592},
  url = {https://doi.org/10.1029/2025gl117592}
}