Machine Learning Applications to Warn-on-Forecast
This grant seeks machine learning/deep learning applications to improve the Warn-on-Forecast System, including hazard prediction, calibration, and model emulation.
This grant is no longer accepting proposals
NRC Research Associateship Programs has archived this opportunity.
Funder: NRC Research Associateship Programs
Due Dates: May 1, 2025 (Application Deadline)
Funding Amounts: Base stipend approximately $60,000/year plus $3,000 travel allowance; additional supplements available for experience and specific doctorates. Typical award duration 2-3 years.
Summary: Supports postdoctoral and senior researchers applying machine learning and deep learning to improve NOAA's Warn-on-Forecast System for severe weather prediction and model emulation.
Key Information: Open to U.S. citizens, permanent residents, and non-U.S. citizens; requires contacting a NOAA research adviser prior to applying.
Description
This fellowship opportunity supports research at NOAA's National Severe Storms Laboratory focused on advancing the Warn-on-Forecast System (WoFS) through machine learning (ML) and deep learning (DL) techniques. The WoFS is an ensemble-based convection-allowing weather prediction system designed to provide timely forecasts of severe weather hazards.
Key research objectives include:
- Developing ML methods to predict storm hazards not explicitly forecasted by the ensemble system.
- Calibrating ensemble forecast outputs based on historical system performance.
- Providing dynamic estimates of forecast accuracy to enhance forecast usability.
- Improving storm-scale data assimilation, such as replacing current cloud analysis schemes to accelerate thunderstorm spin-up.
- Emulating numerical weather prediction (NWP) models, particularly convection-allowing models like WoFS, using DL approaches inspired by recent successes such as Google DeepMind's Graphcast.
The program invites innovative proposals applying ML/DL to any aspect of the WoFS, aiming to improve severe weather forecasting capabilities.
