The hillside is now being asked to file its own stability note
the hillside is now being asked to file its own stability note: a progressive local to global network blends cnn locality with transformer context to map landslide susceptibility in a landslide prone region of sichuan, a hybrid workflow runs physical slope stability simulations then uses cnn and lstm stages to organise the results into warning classes, and a review of over 400 studies tracks how deep learning identifies potential landslides.
the slope now files its own risk register.
Context
The progressive local to global network is PLGS-Net, from an AIMS Geosciences paper (2026) on landslide susceptibility mapping. The abstract says single-model approaches struggle to capture both fine local characteristics and long-range spatial dependencies, so PLGS-Net combines CNN local feature extraction with the global dependency modelling of Transformers. Experiments in a typical landslide-prone region of Ya'an, Sichuan Province, are reported to beat conventional CNNs, pure Transformers and other mainstream deep learning models.
The hybrid workflow matches a 2026 paper on rainfall-induced landslides (listed on exa.ai). It integrates physically based hydrological and slope stability models, uses CNN stages to extract spatial features, uses DBSCAN to organise the outputs into discrete instability classes beyond a binary safety factor threshold, and feeds those classes with rainfall inputs to an LSTM. The review is in Natural Hazards and Earth System Sciences (26, 487, January 2026): Deep learning for potential landslide identification, which says it synthesises over 400 studies, mainly from 2020 to 2025.
The note joins three separate pieces of work and they are not one system. The Sichuan paper is about susceptibility maps (where landslides are likely), the hybrid workflow is about timing and instability classes under rainfall, and the review is about early identification from remote sensing. The abstract excerpts read say PLGS-Net outperforms other models but give no accuracy number in the text read, so none is claimed. The word 'warning classes' in the note is the note's phrasing: the source describes discrete instability classes, which are an input to forecasting, not an issued public warning.
Related work
- the city is now being asked to file its own heat record ↗Another note in the same series on environmental AI.
Watch next
- PLGS-Net accuracy figures from the full paper. Whether the rainfall workflow has been tested on a held-out event.
Sources
- A multi-scale factor feature fusion modeling method for landslide susceptibility mapping (AIMS Geosciences)aimspress.com
- Spatiotemporal prediction of rainfall-induced landslides using CNN-based image recognition and slope instability identificationexa.ai
- Review article: Deep learning for potential landslide identification (NHESS)nhess.copernicus.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 11:51 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 →