The river is now being asked to file its own breakup forecast
the river is now being asked to file its own breakup forecast: a machine learning model reads sar radar returns to predict ice breakup with 92 percent accuracy, a long short term memory network forecasts breakup timing across 33 locations in alaska with an average error of 5.4 days, and a random forest on the upper yellow river found a freezing degree hour threshold near 4000 that separates slow growth from rapid thickening.
the river now files its own ice calendar.
Context
An ISPRS Annals paper of 2026 on near-real-time river ice breakup prediction from SAR and meteorological data reports leave-one-out cross-validation with an overall accuracy of 92 percent, an F1-score of 0.91, a Kappa of 0.84 and a mean absolute error under 6 days. A Water Resources Research paper (Limber and co-authors) builds an LSTM on 33 locations along eight major rivers across Alaska and Western Canada, 23 of them used for tuning, training and testing, and reports a mean absolute error of 5.40 days on the annual breakup date. A 2026 study of the Shisifenzi Reach of the Upper Yellow River uses a random forest with SHAP and ALE on six winters of monitoring data and 26 features, and finds that 30-day cumulative freezing degree-hours dominate, with a threshold near 4000 degree-hours separating slow growth from rapid thickening.
The three results come from three different papers, rivers and tasks, so the figures do not combine. The 92 percent is classification accuracy in the SAR paper, which names a mean absolute error of under 6 days, and it is not a breakup date accuracy. The 5.4 days is a mean absolute error across the Alaska sites, with a 4.03 day standard deviation, on 23 locations used for model work and not all 33. The 4000 figure is from one reach and six winters, and describes ice thickening, not breakup. The river now files its own ice calendar 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
- The ISPRS paper's operational status, and whether the LSTM has been run on live forecasts in a later season.
Sources
- Machine Learning for Near-Real-Time River Ice Breakup Prediction Using SAR and Meteorological Data (ISPRS Annals, 2026)isprs-annals.copernicus.org
- Long Short-Term Memory Model to Forecast River Ice Breakup Throughout Alaska USA (Water Resources Research)impact.ornl.gov
- A Data-Driven Framework for Characterizing Nonlinear Responses of River Ice Growth and Decay, Shisifenzi Reachouci.dntb.gov.ua
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 18:52 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 →