The funnel is now being asked to give a week of warning
the funnel is now being asked to give a week of warning: ncar runs daily ai forecasts for tornadoes, hail and damaging winds three to eight days out, a kalman filtered conv network lifted tornado recall from 72 percent past 99, and radar models now read dual polarization fields with dual attention.
the sky now files its own spin forecast.
Context
NCAR's page describes real-time medium-range convective hazard forecasts for days 1 to 8, built from two 52-member ensembles run with two AI weather model emulators, PanguWeather and FengWu, on NCAR supercomputers and initialized daily at 00 UTC. NCAR and UCAR News on May 4, 2026 describes experimental forecasts that use AI weather models to extend forecasts of tornadoes, hail and damaging winds, giving a percent chance at any location.
The arXiv paper presents a Kalman-Convolutional BiLSTM with Multi-Head Attention for tornado prediction in the United States using MRMS radar data, and says it outperforms methods such as K-Nearest Neighbors and LightGBM. A 2026 paper presents TDA-DARKNet for tornado detection from dual-polarization radar data, with a dual attention mechanism, dense residual connections and a Kolmogorov-Arnold network.
The note joins three separate things. NCAR's product covers hazards at days 1 to 8, and its project page says physics models lose skill beyond about day 4. The note's three to eight days is not the wording of the pages read, so that range is unsupported here, not refuted.
The recall rise from 72 percent to past 99 percent was not found in the abstract of the Kalman paper, which reports better precision, recall, F1 and accuracy against baselines. That figure is unsupported here, not refuted. The dual attention work is a different paper on detection from dual-polarization radar, not forecasting.
'The sky now files its own spin forecast' is the author's line. The NCAR forecasts are for hazard chances over areas, not for individual tornadoes, as NCAR's own news piece says.
Related work
- NCAR Medium-range AI NWP Convective Hazard Forecasts (NSF NCAR) ↗Source for the day 1 to 8 forecasts and the two AI emulators.
- Identifying severe weather hazards further in the future with AI (NCAR & UCAR News, May 4, 2026) ↗NCAR news piece on the experimental AI forecasts.
- A Novel Hybrid Approach for Tornado Prediction in the United States: Kalman-Convolutional BiLSTM with Multi-Head Attention (arXiv 2408.02751) ↗Source for the Kalman-Convolutional BiLSTM tornado prediction paper.
- TDA-DARKNet: A Deep Learning Model Based on Dual-Polarization Radar Data for Tornado Detection (2026) ↗Source for the dual attention dual-polarization detection model.
Watch next
- Read the Kalman paper's results table for recall values. Check NCAR's verification page for skill at each lead.
Sources
- NCAR Medium-range AI NWP Convective Hazard Forecasts (NSF NCAR)www2.mmm.ucar.edu
- Identifying severe weather hazards further in the future with AI (NCAR & UCAR News, May 4, 2026)news.ucar.edu
- A Novel Hybrid Approach for Tornado Prediction in the United States: Kalman-Convolutional BiLSTM with Multi-Head Attention (arXiv 2408.02751)arxiv.org
- TDA-DARKNet: A Deep Learning Model Based on Dual-Polarization Radar Data for Tornado Detection (2026)exa.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 3 October 2026 at 19:18 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →