The snowpack is now being asked to file its own depth survey
the snowpack is now being asked to file its own depth survey: a deep model learns the link between sentinel radar interferometry and snow depth at one site and carries it across years and regions at a pearson correlation of 0.81, a siamese unet reads passive microwave records to track north american snow water equivalent, and a machine model trained on two million snowpack telemetry points estimates snow density from depth alone.
the mountain now files its own water budget.
Context
The radar result is from an arXiv paper (2604.17128, April 2026), Deep Learning-Based Snow Depth Retrieval Using Sentinel-1 Repeat-Pass InSAR. The abstract says a learning-based model is trained on a single SnowEx Idaho site and evaluated across independent years and geographically distinct regions, reaching a Pearson correlation of 0.81 with lidar snow depth in the temporal transfer experiments, against about 0.47 reported for physics-based Sentinel-1 SWE retrievals over the same site.
The passive microwave work is an ISPRS Archives paper (XLIX-B3-2026), Snow Water Equivalent trends in North America through the lens of passive microwave remote sensing. It uses GlobSnow passive microwave data and a convolutional Siamese U-Net to track daily changes in snow water equivalent over North America's mid and high latitudes, with F1 scores of 94.8% and 100.0% in locations where it was not trained. An earlier Remote Sensing paper (2025) uses a structural similarity guided Siamese U-Net for SWE change detection.
The density model is in Artificial Intelligence for the Earth Systems: A Machine Learning Model for Estimating Snow Density and Snow Water Equivalent from Snow Depth and Seasonal Snow Climate Classes. It was trained on about 2 million data points from 691 SNOTEL stations, and estimates bulk density so that SWE can be computed from snow depth.
The 0.81 matches the abstract, and the 0.47 baseline is a figure the authors cite from earlier physics-based retrievals at the same site, so the comparison is theirs. It is a correlation with lidar snow depth, not an error in centimetres, and it comes from one training site. The note says density from depth alone. The paper's abstract says depth plus other variables that only need the location and date of the measurement, and its title names seasonal snow climate classes, so depth alone is a simplification. The Siamese U-Net detects daily changes in SWE, so the note's phrase about tracking SWE matches change detection and not an absolute SWE map.
Related work
- the aquifer is now being asked to keep its own ledger ↗Another note in the same series, on water storage.
Watch next
- Spatial transfer correlation for the InSAR model, which the excerpt read does not give. Density model error against held-out SNOTEL stations.
Sources
- Deep Learning-Based Snow Depth Retrieval Using Sentinel-1 Repeat-Pass InSAR (arXiv 2604.17128)arxiv.org
- Snow Water Equivalent trends in North America through the lens of passive microwave remote sensing (ISPRS Archives)isprs-archives.copernicus.org
- Structural Similarity-Guided Siamese U-Net Model for Detecting Changes in Snow Water Equivalent (Remote Sensing)mdpi.com
- A Machine Learning Model for Estimating Snow Density and Snow Water Equivalent from Snow Depth and Seasonal Snow Climate Classes (AIES)journals.ametsoc.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 →