The forest is now being asked to keep its own death registry
the forest is now being asked to keep its own death registry: an open dataset of 385,000 aerial patches teaches models to find dead trees at 2.5 to 20 centimeter resolution, a unet adapted across four countries via knowledge distillation, and a unet beat ndvi thresholds and random forests for early bark beetle mortality in the bohemian forest.
the canopy now files its own mortality report.
Context
The deadtrees.earth-aerial preprint describes DTE-aerial-train with 385,000 overlapping, multi-resolution patches at 2.5 to 20 cm, using expert annotations and audited pseudo-labels. Its separate benchmark comprises 25 globally distributed orthoimages and 525 patches with expert labels. Patch count is not a count of independent trees or sites.
Cross-Domain Dead Tree Detection via Knowledge Distillation trains on Finland, adapts to Poland and evaluates Germany and Estonia without further training. The preprint reports fewer false positives for feature-level distillation than fine-tuning, with a precision-versus-recall trade-off rather than one uniformly best model.
The May 18, 2026 Bohemian Forest publisher article reports U-Net overall accuracy of 91.4% and green-attack F1 of 0.72 versus Random Forest’s 0.51. These are attributed study results, not independently validated operational detection rates.
The aerial dataset maps tree cover and visible mortality; the Bohemian study addresses earlier infestation from satellite time series. These tasks and resolutions cannot be treated as the same detector. The distillation preprint contains differing annotation totals between prose and its dataset table, so exact totals are not repeated here. The Bohemian article mixes planned-method wording with completed results and contains malformed text; its findings warrant verification from code, data and a clearer methods record before strong reliance.
Watch next
- Reproducible code and field-confirmed green-attack labels. Clarification of annotation totals in the distillation preprint. Independent geographic tests and errors per tree, not only per pixel.
Sources
- deadtrees.earth-aerial: A Multi-Resolution Aerial Image Dataset for Tree Cover and Mortality Detectionarxiv.org
- Cross-Domain Dead Tree Detection via Knowledge Distillationarxiv.org
- Accuracy of Deep Learning-Based Satellite Image Analysis in Early Detection of Insect Infestation-Induced Tree Mortality: A Comparative Analysis with Conventional Remote Sensing Methodsopastpublishers.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 06:03 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 →