Corey K Potvin

In EOR at @AMS_AIES: "Using machine learning to predict convection-allowing ensemble forecast skill: Evaluation with the NSSL Warn-on-Forecast System" doi.org/10.1175/AIES-D…

Can we use AI to predict thunderstorm evolution? Recently published in @theAGU GRL, we refactored @GoogleDeepMind's GraphCast for regional modeling and trained it on @CIWRO_ / @NOAANSSL's archived Warn-on-Forecast System (WoFS) data.

Meteorologists are turning to AI as a quick, efficient alternative to traditional weather prediction algorithms. Now researchers including @MontePhD & @corey_potvin have extended this tech to cover localized extreme weather as well as global forecasts. eos.org/research-spotl…

🤖 How can AI help scientists predict thunderstorms faster? 🤔 A new paper from NSSL and @CIWRO_ discusses how WoFSCast, a new AI-powered model, can accurately predict how storms will evolve up to two hours in advance. ⛈️ 📝: inside.nssl.noaa.gov/nsslnews/2025/…

🚨 New paper! 🚨 Available in EOR is our paper exploring training AI models on MRMS MESH vs. reported hail sizes: journals.ametsoc.org/view/journals/…

🌪️🌩️ Studying tornadoes in the field has always been at the heart of the NSSL mission. With roots in the 90s, the VORTEX-USA project continues to blaze a trail in tornado science with the ultimate goal to protect lives and property. 🌪️ UP CLOSE: inside.nssl.noaa.gov/nsslnews/2024/…

WoFSCast is the first AI system trained and developed to emulate a sub-hourly convection-allowing weather model (CAM). WoFSCast speeds up forecast generation versus WoFS by at least a factor of ten, from minutes to seconds. #AMS2025

Monte Flora: Future of storm-scale AI: generating large-member (100+) hi-res (~1-km) ensemble forecasts in seconds. AI system in the WoFS framework already can produce “closely matching” forecasts with far less computational burden than #WoFS (far faster). #31sls

Huge kudos to @MontePhD for leading this awesome work! AI will undoubtedly play a major role in NWP at all scales.

@MontePhD Plus, we can fit ~10 WoFSCast members on one GPU, whereas all 18 WoFS members use 1080 CPUs total.

Best of all? WoFSCast achieves this in 30-40 seconds on 1 GPU, compared to the 8-12 minutes and 60+ CPUs WoFS needs.

🚨 Exciting Announcement! 🚨 Introducing WoFSCast: our AI-powered system for predicting thunderstorm evolution at watch-to-warning scales. 🌩️⚡ essopenarchive.org/users/829074/a…

Our paper combining AI/ML, storm-scale ensemble prediction, and "forecasting forecast skill" is now in final form in @AMS_AIES: doi.org/10.1175/AIES-D… Co-authors: @MontePhD, Patrick Skinner, @tonyreinhart, @BrianWxFLA

🤔How can science provide more warning when 🌪️tornadoes threaten? The Warn-on-Forecast System, equips forecasters with critical information between watches and warnings, enabling them to give longer lead times in the face of severe weather and tornadoes. inside.nssl.noaa.gov/nsslnews/2024/…

Our paper on the usefulness of the "WoFS-ML-Severe" products examined during the 2022 HWT-SFE is available in EOR. This paper explores subjective verification, objective verification, and qualitative analysis of the participant's feedback. journals.ametsoc.org/view/journals/…

Thank you to my co-authors: @MontePhD, Pat Skinner, @tonyreinhart, @BrianWxFLA.

I believe this is the first published application of machine learning to predicting storm-scale forecast skill. This work motivates additional applications of ML to predicting forecast skill, including at larger scales / longer lead times.

In EOR at @AMS_AIES: "Using machine learning to predict convection-allowing ensemble forecast skill: Evaluation with the NSSL Warn-on-Forecast System" doi.org/10.1175/AIES-D…

**Experimental (Not Public Yet)** The new AI hail size algorithm says 2 in is possible!

Great presentations from #NSSL scientists continue here at #AMS2024, including researchers Kim Hoogewind and Adam Clark at the @NOAA booth talking about trends in severe thunderstorms and tornadoes in a changing climate.

Hot off the press: The OU School of Meteorology (@ousom) & School of Civil Engineering and Environmental Science have a joint tenure-track faculty position focused on the broad areas of Hydrometeorology and Hydrology that is now open. Apply here: apply.interfolio.com/137372

Please share!

Possible NRC postdoc topics: (1) WoFS forecast emulation via deep learning (2) Other ML applications (3) Thunderstorm predictability (4) Multi-resolution ensemble design (5) Novel verification techniques (6) ??? Feel free to reach out here or at corey.potvin@noaa.gov.

The Feb 1 deadline for NRC postdoc apps is coming up! I am looking for someone to work with me on a topic related to our Warn-on-Forecast System. Search "Potvin" at ra.nas.edu/RAPLab10/Oppor…. These project descriptions are flexible. Possible topics in next post....