The aurora is now being asked to file its own visibility forecast
the aurora is now being asked to file its own visibility forecast: a two stage cascade learns the probability of aurora occurring from space weather inputs and then multiplies in cloud and moonlight to get a clean chance of seeing it, a model trained on 28 years of solar wind and magnetometer records predicts it one and a half to two hours ahead, and a resnext network picks auroras out of satellite images using motion across frames.
the sky now files its own viewing log.
Context
An arXiv paper (2605.24038), Aurora Hunter, describes a two-stage cascade: stage one uses gradient-boosted trees on 51 space weather features, trained on about 16,600 hours of Tromso and Kiruna data from 2015 to 2023, to predict the probability that aurora occurs; stage two uses logistic regression on 21 cloud-cover and lunar-illumination features to predict the chance of a clear observation given aurora. A CEDAR 2026 poster applies a ResNeXt-50 network to near-infrared sequential images from the e-POP Fast Auroral Imager to separate aurora from static features such as clouds, mountains and city lights. The GeoDGP paper, using SuperMAG data from about 350 stations over 28 years from 1995 to 2022, forecasts ground magnetic perturbations from solar wind.
The 28 year record and the lead time belong to a geomagnetic perturbation forecast and not an aurora forecast, and the one and a half to two hour range was not confirmed in the text read. The two-stage model combines occurrence with visibility as the note says, and multiplying in cloud and moonlight was inferred from the stage design. ResNeXt-50 picking auroras by motion across frames fits the poster's sequential imagery, and its accuracy was not read. These are research systems, and no operational comparison was read. The sky now files its own viewing log is the author's framing.
Related work
- Earlier note in this series: the monsoon ↗Same pattern: a natural system forecast by learned models.
Watch next
- Aurora Hunter's skill scores, and the exact source for the 1.5 to 2 hour lead time.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 18:35 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 →