The aquifer is now being asked to keep its own ledger
the aquifer is now being asked to keep its own ledger: a decision tree ensemble trained with shap attribution reconstructs two decades of human and climate driven groundwater storage shifts across brazil, a framework that imputes missing records and searches for network architectures forecasts 1,493 wells in six basins, and a hierarchical graph network coupled with an lstm projects terrestrial water storage six months ahead.
the underground now files its own balance.
Context
The Brazil reconstruction is the paper Two decades of human- and climate-induced groundwater storage shifts in Brazil. The text read says the model ensemble is tree-based and includes extreme gradient boosting, and it analyses groundwater storage trends across 2002 to 2023 including before and after the 2015/2016 El Nino, with a consistent total water storage decline seen between 2009 and 2019.
The well forecasting framework is GWL-IMFO, a Groundwater Level Imputation and Forecasting framework that uses neural architecture search to choose ANN, LSTM or GRU forecasters under imputed data. It was evaluated across 1,493 monitoring wells in six basins, with controlled missing-data scenarios of 20%, 30% and more.
The terrestrial water storage model is HiGNN-LSTM, a hierarchical graph neural network on two spatial scales coupled with an LSTM, forecasting global TWS anomalies up to six months ahead (NeurIPS ML4PS 2025 workshop paper and an EGUsphere 2026 preprint).
The note's three items are separate studies. The SHAP attribution in the note was not seen in the excerpt read for the Brazil paper, which names XGBoost, so SHAP is unsupported here, not refuted. The 1,493 wells and six basins match the abstract. The six month horizon matches both HiGNN-LSTM sources, and the workshop abstract describes the gridded seasonal forecast as a short-term improvement over a seasonal long-term mean, so it is a modest gain and the EGUsphere paper is a preprint.
Related work
- the field is now being asked to file its own harvest forecast ↗Another note in the same series on environmental AI.
Watch next
- SHAP results from the Brazil paper. HiGNN-LSTM skill by lead month.
Sources
- Two decades of human- and climate-induced groundwater storage shifts in Brazillib.icimod.org
- A unified framework for groundwater level imputation and forecasting in data-limited catchmentsresearchnow.flinders.edu.au
- Hierarchical Graph Networks for Forecasting Terrestrial Water Storage Anomalies (NeurIPS ML4PS 2025)ml4physicalsciences.github.io
- Hierarchical Graph Networks for Seasonal Forecasts of Terrestrial Water Storage Anomalies (EGUsphere)egusphere.copernicus.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 11:18 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 →