The kelp is now being asked to keep its own canopy map
the kelp is now being asked to keep its own canopy map: a unet and vgg16 hybrid reads planet dove satellites to draw california’s kelp forests in a detail old monitoring missed, an open source python tool called skema kelp classifies sentinel 2 imagery, and ucla put the statewide maps on kelpwatch.
the forest of seaweed now files its own survival record.
Context
A Communications Earth & Environment paper (published January 2026) trains a VGG16-U-Net, a U-Net that uses pre-trained VGG16 weights, on Planet Dove imagery and creates annual 3 m maps of floating kelp canopy across California from 2017 to 2024. The model classifies pixels as kelp, seawater, waves or land and reached an overall mean test accuracy of 89.59%, with 86.4% for kelp. The paper contrasts this with the public Landsat product at 30 m resolution.
UCLA's newsroom (March 18, 2026) says the team applied deep learning to Planet's Dove constellation to build a map 10 times more detailed than earlier standard satellite records, that Sonoma and Mendocino counties lost more than 90% of kelp after the 2014 to 2016 marine heatwave, and that the maps are public through KelpWatch.org with The Nature Conservancy.
SKeMa is an open source Python tool on GitHub that classifies kelp in Sentinel-2 images with a PyTorch segmentation model, through a command line interface, installable with pip as skema-kelp. The model is on Hugging Face.
The note's U-Net and VGG16 hybrid matches the paper's VGG16-U-Net. 'Planet Dove satellites' and 'statewide maps on KelpWatch' match the paper and UCLA. The 'detail old monitoring missed' fits the 3 m versus 30 m comparison.
The note describes SKeMa as a tool that 'classifies sentinel 2 imagery', which matches the README. SKeMa is from a different group, so it is a separate tool and not the method behind the California maps. Which kelp species or regions SKeMa was trained on was not read here, and is unsupported, not refuted.
The 89.59% accuracy is an overall mean across four classes. Waves scored lowest at 79.2%, so the kelp-specific number is the better one to cite.
'The forest of seaweed now files its own survival record' is the author's opinion.
Related work
- High-resolution Planet Dove data identify local drivers of kelp canopy persistence (Communications Earth & Environment, 2026) ↗Primary source for the model, the 3 m maps and the accuracy figures.
- How AI deep learning is helping UCLA scientists protect California's coastal ecosystems (UCLA Newsroom, March 18, 2026) ↗Source for the 10 times claim, the marine heatwave losses and the KelpWatch link.
- m5ghanba/skema: Satellite-based Kelp Mapping (GitHub) ↗Source for the SKeMa tool, its command line interface and pip install.
- Kelpwatch.org ↗Public map platform where the statewide maps are shown; the landing page was not read in full here.
Watch next
- Read the SKeMa Hugging Face model card for its training regions and accuracy. Compare the 3 m California maps with SKeMa's 10 m Sentinel-2 output on the same coast.
Sources
- High-resolution Planet Dove data identify local drivers of kelp canopy persistence (Communications Earth & Environment, 2026)nature.com
- How AI deep learning is helping UCLA scientists protect California's coastal ecosystems (UCLA Newsroom, March 18, 2026)newsroom.ucla.edu
- m5ghanba/skema: Satellite-based Kelp Mapping (GitHub)github.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 3 October 2026 at 22:35 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 →