The bloom is now being asked to file its own sighting report
the bloom is now being asked to file its own sighting report: a vision transformer maps coastal algal blooms from 30 meter landsat sentinel 2 images for the first time, a benchmark of commercial vision language models found they cry wolf on 73 to 93 percent of bloom free satellite images while a multispectral svm stayed reliable, and a transformer plus bilstm forecasts lake champlain cyanobacteria intensity up to 14 days out.
the algae now files its own bloom log.
Context
The vision transformer work is an arXiv paper (2606.17242, June 2026), Landsat-Sentinel-2 Algal Bloom Mapping Using Vision Transformers. The abstract calls it the first successful implementation of vision transformer-based coastal algal bloom mapping using 30 m Landsat-Sentinel-2 images, built on a globally distributed bloom patch dataset.
The benchmark is Bloom or Bluff? Benchmarking Vision-Language Models Against Classical Machine Learning (2026). Its key findings say commercial VLMs (GPT-4o, GPT-5.5, Claude Sonnet 4.6) flagged harmful algal blooms with 73 to 93% false positive rates on bloom-absent satellite images, and that a multispectral SVM on 10 Sentinel-2 bands performed best overall (F1 0.833, 79.5% accuracy, 27% false positive rate). The forecast is From Remote Sensing to Multiple Time Horizons Forecasts (arXiv 2512.06598), a Transformer plus BiLSTM that predicts CyanoHAB intensity in Lake Champlain up to 14 days ahead from satellite cyanobacterial index and temperature data.
The numbers match their sources. The SVM 'stayed reliable' in the note is relative: the benchmark's own figure for it is a 27% false positive rate and 79.5% accuracy, better than the language models but not a clean result. 'First' in the vision transformer claim is the authors' own statement. The 14 days is the maximum horizon, and the excerpt read does not give skill at 14 days, so skill by horizon is not claimed. The three items are separate studies.
Watch next
- Forecast skill by lead day for the Lake Champlain model. Whether the VLM benchmark was repeated on other regions.
Sources
- Landsat-Sentinel-2 Algal Bloom Mapping Using Vision Transformers (arXiv 2606.17242)arxiv.org
- Bloom or Bluff? Benchmarking Vision-Language Models Against Classical Machine Learningexa.ai
- From Remote Sensing to Multiple Time Horizons Forecasts: Transformers Model for CyanoHAB Intensity in Lake Champlain (arXiv 2512.06598)arxiv.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 10:03 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 →