The lake is now being asked to file its own layering report
the lake is now being asked to file its own layering report: a foundation model trained on irregular multi depth time series spontaneously obeys the thermal stratification law it was never taught, a three stage transfer network learns lake surface temperature from satellite data before local fine tuning, and a study found a long short term memory network beats a physics model on accuracy yet blows up its bias on cloud covered winters.
the lake now files its own depth ledger.
Context
LakeFM (arXiv 2606.11268) is a foundation model for aquatic systems pre-trained on simulated and observed lakes with irregular multivariate multi-depth time series, and the paper compares its physical consistency against Chronos 2 across 100 simulated lakes using an inversion rate against the thermal stratification law and a Beer-Lambert check. Transfer-PIDL, in a Cambridge-hosted paper, is a transfer learning framework for lake surface temperature prediction with a three-stage training strategy of pre-training, re-training and fine-tuning, integrated into physics-informed deep learning. A separate paper builds a physics-informed network on Koopman embeddings and LSTM for multi-depth lake water temperature.
LakeFM and the three-stage transfer framework are supported. The note's claim that the foundation model obeys the stratification law it was never taught was not checked beyond the paper's inversion-rate metric, and the Transfer-PIDL text read does not mention learning from satellite data, so that link is unverified. The study in which an LSTM beats a physics model on accuracy with a much larger bias in cloud-covered winters was not found. The lake now files its own depth ledger is the author's framing.
Related work
- Earlier note in this series: the river ↗Same pattern: a natural system forecast by learned models.
Watch next
- The LakeFM results and the LSTM versus physics model study.
Sources
- LakeFM: Toward a Foundation Model for Aquatic Ecosystems (arXiv 2606.11268)arxiv.org
- Predicting Lake Surface Water Temperature With Transfer-Based Physics-Informed Deep Learningapi.repository.cam.ac.uk
- Lake Water Temperature Modeling Using Physics-Informed Neural Networks (ICLR 2025 workshop)climatechange.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 19:34 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 →