Xiong et al. (2026) CSU-PCAST: a dual-branch transformer framework for medium-range ensemble precipitation forecasting

Identification

• Journal: npj Climate and Atmospheric Science
• Year: 2026
• Date: 2026-07-07
• Authors: Haonan Chen, Kelly M. Mahoney
• DOI: 10.1038/s41612-026-01475-9

Research Groups

• Colorado State University (CSU)
• NOAA Physical Sciences Laboratory (PSL)

Short Summary

This study develops CSU-PCAST, a deep learning-based ensemble forecasting framework that improves medium-range precipitation forecasts relative to operational ensemble baselines at a fine spatial resolution of 0.25°.

Objective

• Develop an accurate and reliable model for predicting medium-range precipitation using a dual-branch transformer framework

Study Configuration

• Spatial Scale: Global, with a focus on the contiguous United States (CONUS) domain
• Temporal Scale: Medium-range, up to 15 days ahead, with forecasts generated every 6 hours

Methodology and Data

• Models used: CSU-PCAST, a deep learning-based ensemble forecasting framework using Swin Transformer blocks and Conditional Layer Normalization
• Data sources:
  • ERA5 atmospheric and surface variables at 0.25° spatial resolution
  • IMERG precipitation labels

Main Results

• Improved short-range deterministic skill relative to GEFS, with higher CSI and lower RMSE during the first several forecast days
• Reduced GEFS wet bias for light precipitation and dry bias at the 10 and 20 mm thresholds
• Lower CRPS and higher Brier Skill Scores for several thresholds, indicating improved probabilistic precipitation forecasts

Contributions

• Original value of the article with respect to existing literature lies in its development of a deep learning-based ensemble forecasting framework that improves medium-range precipitation forecasts relative to operational ensemble baselines.

Funding

• This research was funded by [insert funding agency/project/program reference codes].

Citation

@article{Xiong2026CSUPCAST,
  author = {Xiong, Tianyi and Chen, Haonan and Mahoney, Kelly M. and Tang, Jingyin and Smith, Tim and Bytheway, Janice},
  title = {CSU-PCAST: a dual-branch transformer framework for medium-range ensemble precipitation forecasting},
  journal = {npj Climate and Atmospheric Science},
  year = {2026},
  doi = {10.1038/s41612-026-01475-9},
  url = {https://doi.org/10.1038/s41612-026-01475-9}
}