Hui et al. (2025) Refine Extreme Hot Day Predictions With the Sea Surface Temperature Tendency

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

Identification

• Journal: Geophysical Research Letters
• Year: 2025
• Date: 2025-09-17
• Authors: Tan Hui, Zhiwei Zhu, Fenghua Ling, Bin Wang
• DOI: 10.1029/2025gl116339

Research Groups

[Not specified in abstract]

Short Summary

This study investigates the underlying mechanisms influencing extreme hot days over Western North America (WEHDs) and aims to improve their seasonal prediction, revealing that two independent sea surface temperature (SST) precursor signals enable robust and enhanced prediction using a physics-informed convolutional neural network.

Objective

• To explore the underlying mechanisms influencing extreme hot days over Western North America (WEHDs).
• To improve the seasonal prediction of WEHDs.

Study Configuration

• Spatial Scale: Western North America, tropical eastern Pacific, tropical North Atlantic.
• Temporal Scale: Seasonal prediction, focusing on springtime precursor signals and extreme events (e.g., 2021).

Methodology and Data

• Models used: Physics-based empirical model, Convolutional Neural Network (CNN), physically informed CNN.
• Data sources: Sea surface temperature (SST) anomalies, SST tendency fields.

Main Results

• Two independent precursor signals, persistent negative sea surface temperature (SST) anomalies in the tropical eastern Pacific and a cooling tendency in tropical North Atlantic SST during springtime, significantly influence WEHDs.
• A physics-based empirical model constructed with these two predictors demonstrates robust independent prediction skills.
• A physically informed Convolutional Neural Network (CNN), integrating SST tendency fields, achieves significantly improved prediction performance.
• The physically informed CNN successfully predicted the extreme WEHD events of 2021.

Contributions

• Identification of two independent precursor sea surface temperature (SST) signals (tropical eastern Pacific negative SST anomalies and tropical North Atlantic SST cooling tendency) for extreme hot days over Western North America (WEHDs).
• Development of a physics-based empirical model for WEHD prediction using these precursors.
• Introduction and successful application of a physically informed Convolutional Neural Network (CNN) that integrates SST tendency fields to significantly enhance WEHD prediction accuracy.
• Emphasizing the critical role of physical understanding in advancing deep learning-based climate prediction.

Funding

[Not specified in abstract]

Citation

@article{Hui2025Refine,
  author = {Hui, Tan and Zhu, Zhiwei and Ling, Fenghua and Wang, Bin},
  title = {Refine Extreme Hot Day Predictions With the Sea Surface Temperature Tendency},
  journal = {Geophysical Research Letters},
  year = {2025},
  doi = {10.1029/2025gl116339},
  url = {https://doi.org/10.1029/2025gl116339}
}