The lightning is now being asked to file its own strike calendar
the lightning is now being asked to file its own strike calendar: a boltcast network turns gfs convective fields into one to four day lightning probabilities trained on goes glm flashes, a two step cnn predicts lightning and the fires it ignites across northern california with 0.82 accuracy, and a two dimensional scheme runs global lightning on a single cape field at 0.89 correlation.
the storm now files its own strike map.
Context
BoltCast is a medium-range lightning prediction system described in AI for the Earth Systems (AMS), with code at ai2es/boltcast on GitHub for CONUS lightning forecasts at a multi-day scale. The repository says it downloads and matches GFS forecast fields to GOES GLM lightning observations, and the paper evaluates predictions for days 1 to 4 with thresholds on probability.
The wildfire paper is Predicting and Explaining Lightning and Lightning-Ignited Wildfires in Northern California With a Two-Step Convolutional Neural Network Framework (2026). It predicts lightning from meteorological conditions and then wildfire ignition conditional on lightning, using meteorological and fuel predictors, with testing accuracy of 0.82 for predicting lightning-ignited wildfires. The global scheme is A Two-Dimensional Deep Learning Scheme With One Predictor Only to Parametrize Global Lightning (Geophysical Research Letters, 2026), using CAPE only, with a determination coefficient of 0.89, a 24% increase over an existing machine learning global lightning scheme.
The note says the global scheme reaches 0.89 correlation. The abstract says a determination coefficient of 0.89, which is R squared and not a correlation coefficient, so the note's word is loose. The 0.82 is accuracy for lightning-ignited wildfires, a classification measure that depends on how common ignitions are in the test data. BoltCast covers CONUS, and the excerpt read lists days 1 to 4 and does not give a probability skill score here, so none is claimed.
Watch next
- BoltCast skill by lead day. Class balance behind the 0.82 accuracy.
Sources
- BoltCast: Medium-Range Lightning Prediction with Neural and Long Short-Term Memory Networks (AI for the Earth Systems)journals.ametsoc.org
- ai2es/boltcast (GitHub)github.com
- Predicting and Explaining Lightning and Lightning-Ignited Wildfires in Northern California With a Two-Step Convolutional Neural Network Frameworkexa.ai
- A Two-Dimensional Deep Learning Scheme With One Predictor Only to Parametrize Global Lightning (GRL)exa.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 08:47 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 →