The river is now being asked to file its own rise forecast
the river is now being asked to file its own rise forecast: a global lstm called aifl trained on 18,588 basins pretrains on four decades of reanalysis and fine tunes on operational weather forecasts, google's upgraded flood system uses a mean embedding lstm to push reliable forecasts six days further, and a convlstm attention network averaged a nash sutcliffe efficiency of 0.84 across fourteen flood events.
the waterway now files its own stage.
Context
AIFL is described in arXiv 2602.16579 (February 2026), a global daily streamflow forecasting model using a deterministic LSTM pre-trained on ERA5-Land and fine-tuned on IFS forecasts. The abstract says it is trained on 18,588 basins curated from the CARAVAN dataset with a two-stage training strategy to bridge the gap between reanalysis and operational forecast products.
The Google paper is Extending Medium-Range Global Flood Forecasts: The Google Global Flood Forecasting Model V2. It moves from an Encoder-Decoder LSTM to a Mean Embedding LSTM, with an expanded Caravan training set and GraphCast forcings, and says v2 extends the reliable predictive horizon by 6 days in gauged basins relative to the v1 nowcast. The ConvLSTM attention result is from a Multi-Source Environmental Data Flood Forecasting System paper (2026), reporting an average Nash-Sutcliffe Efficiency of 0.84 and an average RMSE of 0.18 meters.
The 18,588 basins and the 0.84 match their sources. The six days is for gauged basins: the Zenodo record says 2 days in ungauged basins and the EGUsphere abstract says 1 day, so the two Google sources disagree on the ungauged figure and the note's wording covers neither. The note says fourteen flood events for the ConvLSTM system. The excerpt read gives the 0.84 and 0.18 m RMSE but not the event count, so that count is unsupported here, not refuted.
Watch next
- The ungauged-basin horizon in the final Google paper. The number of events behind the ConvLSTM score.
Sources
- AIFL: A Global Daily Streamflow Forecasting Model Using Deterministic LSTM Pre-trained on ERA5-Land and Fine-tuned on IFS (arXiv 2602.16579)arxiv.org
- Extending Medium-Range Global Flood Forecasts: The Google Global Flood Forecasting Model V2 (Zenodo)zenodo.org
- The Next-Generation Google Flood Forecasting Model and Community Resources (EGU26)meetingorganizer.copernicus.org
- Multi-Source Environmental Data Flood Forecasting System Based on ConvLSTM Networksexa.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 10:47 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 →