The slope is now being asked to give its own warning
the slope is now being asked to give its own warning: a multimodal fusion net predicts three gorges landslide displacement one to fifteen days out within 3.4 millimeters, a contrastive model learns rainfall fields that stay stable when forecasts drift across japan, and a deep classifier separates true alarms from sensor noise on monitored slopes in fujian.
the hill now files its own risk ledger.
Context
The Japan rainfall paper (arXiv, May 2026) trains a landslide warning model with Rainfall-Motion-Aware Contrastive Learning, which adds correlated perturbations to rainfall fields to mimic forecast displacement so the learned representation stays stable. It used two years of rainfall and terrain data across 19 regions in Japan and reports up to 37% higher precision than state-of-the-art baselines.
The Fujian slope paper (EGUsphere preprint, May 2026) treats crackmeter alerts as a classification problem: true early warning or false alarm. A patch-based dual-branch Transformer fuses displacement and rainfall recorded at different rates, using slopes in Fujian Province, China, with true warnings confirmed by field investigation. It reports precision 90.91%, recall 93.53% and F2 92.99%, and says its decisions follow local displacement trends and relative rainfall intensity.
The Three Gorges item, a multimodal fusion net predicting displacement 1 to 15 days out within 3.4 millimeters, could not be matched. Searches returned other Three Gorges displacement studies (for example Baishuihe and Shengjibao landslide models) but none with that figure or horizon, so it is unsupported here, not refuted. The Japan result is a precision gain against baselines over two years of data and is not an absolute accuracy. The Fujian result applies to small-scale slopes with crackmeters and to the alert-versus-false-alarm question, not to forecasting when a slope will fail. Both are preprints. The note's 'hill files its own risk ledger' fits neither study as a warning that has been used in operation.
Related work
- Learning Displacement-Robust Representations for Landslide Early Warning under Rainfall Forecast Uncertainty (arXiv) ↗Source for the rainfall displacement method and the 37% precision gain.
- Reducing False Alarms in Small-Scale Slope Early Warning Systems via Deep Learning-Driven Asynchronous Displacement and Rainfall Data Fusion (EGUsphere) ↗Source for the Fujian false alarm classifier and its precision and recall.
- Multi-Scale Response Analysis and Displacement Prediction of Landslides Using Deep Learning with JTFA: A Case Study in the Three Gorges Reservoir, China (Remote Sensing, MDPI) ↗A Three Gorges displacement prediction study. It is not the 3.4 mm paper.
- Mechanism guided forecasting model for landslide displacement prediction: case study for Baishuihe Landslide (Frontiers in Earth Science) ↗Another recent displacement forecasting paper on a Three Gorges landslide.
Watch next
- Find the multimodal Three Gorges paper that gives 3.4 mm and the 1 to 15 day horizon, and check which landslide and period it covers.
Sources
- Learning Displacement-Robust Representations for Landslide Early Warning under Rainfall Forecast Uncertainty (arXiv)arxiv.org
- Reducing False Alarms in Small-Scale Slope Early Warning Systems via Deep Learning-Driven Asynchronous Displacement and Rainfall Data Fusion (EGUsphere)egusphere.copernicus.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 01:03 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 →