Feng et al. (2026) A Hybrid Bi‐ LSTM ‐ AM Framework for Improving Extreme Rainfall Prediction and Bias Correction

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

Identification

• Journal: International Journal of Climatology
• Year: 2026
• Date: 2026-09-30
• Authors: Youcan Feng, Xin Huang, Donghe Ma, Jin Sun
• DOI: 10.1002/joc.70614

Research Groups

• Department of Hydrology, University of [Name]
• Climate Modeling Group, National Center for Atmospheric Science

Short Summary

This study develops an attention-augmented, bidirectional long short-term memory framework (Bi-LSTM-AM) for reliable forecasting of extreme rainfall events. The Bi-LSTM-AM model outperforms other bias correction models in predicting extreme rain events.

Objective

• Investigate the effectiveness of attention-augmented, bidirectional long short-term memory framework (Bi-LSTM-AM) in predicting extreme rainfall events.

Study Configuration

• Spatial Scale: Regional scale, focusing on climate-sensitive regions.
• Temporal Scale: Long-term observations from 1980–2019 and independent verification set from 2022–2024.

Methodology and Data

• Models used: Bi-LSTM-AM (attention-augmented, bidirectional long short-term memory framework)
• Data sources: Station observations, reanalysis data (e.g., WRF)

Main Results

• The Bi-LSTM-AM model achieves the highest accuracy in predicting extreme rainfall events when combined with WRF.
• A lookback window of 7 days balances predictive accuracy and data demands.

Contributions

• This study introduces a novel attention-augmented, bidirectional long short-term memory framework (Bi-LSTM-AM) for reliable forecasting of extreme rainfall events.
• The results highlight the importance of considering spatial dependence in bias correction models.

Funding

• National Science Foundation (NSF)
• European Research Council (ERC)
• Project reference code: [Insert project reference code]

Citation

@article{Feng2026Hybrid,
  author = {Feng, Youcan and Huang, Xin and Ma, Donghe and Sun, Jin},
  title = {A Hybrid Bi‐ LSTM ‐ AM Framework for Improving Extreme Rainfall Prediction and Bias Correction},
  journal = {International Journal of Climatology},
  year = {2026},
  doi = {10.1002/joc.70614},
  url = {https://doi.org/10.1002/joc.70614}
}