AI Weather Quest Python Package

To support participants in the AI Weather Quest, a dedicated Python package has been developed. This package enables users to submit forecasts, access potential training datasets, and perform local forecast self-evaluation. 

Comprehensive documentation for the package is available here: AI Weather Quest Python Package Documentation

As the AI Weather Quest is a cutting-edge and evolving competition, regular updates to the package are anticipated. This page provides an up-to-date overview of all changes, improvements, and enhancements to ensure participants have the latest tools at their disposal.

Latest updates

To upgrade the python package, please use the following command: 

python3 -m pip install --upgrade AI-WQ-package

Package VersionRelease DateSummary of updates
2.4.428th July 2025
  • New function: forecast_submission.AI_WQ_check_submission to verify forecast submission success. 
  • Excluded grid points from RPS calculation where all quintile climatologies are zero (i.e., dry regions).
  • Applied land-sea mask for grid points with less than 50% land coverage. Originally mask was applied for grid points with less than 80% land coverage. 
  • Separated latitude weighting and global mean computation in RPS calculation.
  • Extended ERA5-based training data availability to up to 4 months prior to the current date. Originally only up to 2024. 
  • Enabled download of historical quintile climatologies at a daily resolution. Originally only for every Monday start date. 
  • Added functionality to evaluate forecasts for selected geographic regions to enable regional evaluation. 
  • Implemented a check to exclude submissions from withdrawn teams.
2.0.06th May 2025 Initial public release prior to the JJA (June to August) testing period.
  • No labels