The city is now being asked to file its own heat record
the city is now being asked to file its own heat record: a deep network downscales satellite land surface temperature to one kilometre squares refreshed every fifteen minutes across europe, foundation model embeddings map street level heat across 26 us cities at a mean error of 1.57 degrees, and a machine framework projects urban heat island intensity at 30 metre resolution for five cities under 2030 climate scenarios.
the pavement now files its own temperature log.
Context
The kilometre-scale network is from an arXiv paper (2605.13566), Spatiotemporal downscaling and nowcasting of urban land surface temperature. It combines a geostationary and a polar-orbiting satellite to give LST fields at 1 km every 15 minutes with a U-Net, applied to intraday forecasting; the excerpt reports a hold-out RMSE of 1.92 degrees C and near-zero bias on European cities with over one million inhabitants. Code is on GitHub as EnergyWeatherAI/LST_downscaling_and_nowcasting.
The foundation model embedding study downscales 30 m Landsat LST to 10 m using Google's AlphaEarth Foundations embeddings (64 dimensions, 10 m, global) and a neural network across 26 US cities, reporting a mean RMSE of 1.57 degrees C and a mean R2 of 0.733 on the test set.
The note says a mean error of 1.57 degrees. The paper gives it as a mean root-mean-squared error of 1.57 degrees C, which is a different statistic from a mean error, so the figure is right but the label is looser. It is also surface temperature from satellites, not air temperature people feel on the street. The 30 metre urban heat island projection for five cities under 2030 scenarios was not located: the searches returned a Kuala Lumpur study at 1 km and a Beijing study with 1 km intensity, neither matching, so that item is unsupported here, not refuted.
Related work
- the hillside is now being asked to file its own stability note ↗Another note in the same series on environmental AI.
Watch next
- The source of the 30 m, five-city, 2030 scenario projection. Cross-city error spread for the embedding approach.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 11:35 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 →