The mangrove is now being asked to file its own growth chart
the mangrove is now being asked to file its own growth chart: a unet with squeeze and excitation plus attention masks chinese mangroves yearly from 1993 to 2023 at an f1 of 0.928, a resnet unet with elevation constraints reaches an iou of 0.822 in the northern beibu gulf, and super resolution pushes africa’s national mangrove maps with sentinel 2 alone.
the coast now files its own forest ledger.
Context
A 2026 Frontiers in Marine Science paper builds an enhanced U-Net with Squeeze-and-Excitation (SE) and Convolutional Block Attention Module (CBAM) blocks that extracts annual mangrove masks from Landsat imagery for 1993 to 2023. It reports an IoU of 0.815 and an F1-score of 0.928 against independent reference data. The study area is the Gaoqiao Mangrove National Nature Reserve on the Guangdong-Guangxi border. Only the three visible bands at 30 m were used, and the model produced 31 annual masks from 111 scenes.
A 2026 paper in Water on the Nanliu River Delta, the largest deltaic mangrove system in the Northern Beibu Gulf, uses Landsat time series from 1990 to 2025 and a ResNet34-UNet with Digital Elevation Model constraints. It reports an IoU of 0.822 for mangroves, 0.981 for background and a mean IoU of 0.902.
A publication record on African mangroves applies super-resolution and semantic segmentation to Sentinel-2 and Sentinel-1 imagery for coastal Tanzania and Ghana. Sentinel-2 alone reached an IoU of 85.8%, adding Sentinel-1 gave 86.3%, super-resolution gave 88.5% and the combined set gave 89.2%. Area estimates carry confidence intervals, for example 1207 km2 for Tanzania.
The 0.928 F1 and 1993 to 2023 span belong to one reserve, Gaoqiao, and not to all of China. The note reads as if a single model maps Chinese mangroves yearly nationwide. In the same paper the U-Net+SE+CBAM model scores 0.815 IoU on one validation table and 0.718 IoU with F1 0.822 on another, so the headline F1 depends on which table is read.
A separate national study of China (U-net+ResNet34, 2016 to 2023, Sentinel imagery) reports an F1 of 0.843, which is lower and covers a different period and area. The figures in the note should not be read as national accuracy.
The 0.822 IoU matches the Nanliu River Delta paper. It is a single delta study, and the note's phrase 'northern beibu gulf' matches the paper's title region. The note says the model reaches the IoU 'with elevation constraints', which matches the DEM constraints described.
The African result matches the record, with the caveat that it covers Tanzania and Ghana as representative areas and the 'national maps' wording refers to wall-to-wall mapping of those two countries. The publication venue was not confirmed in the page read, so its peer-review status is unsupported, not refuted. 'Sentinel 2 alone' is partly right: super-resolved Sentinel-2 alone reached 88.5%, and the best score used Sentinel-1 as well.
'The coast now files its own forest ledger' is the author's framing. The three studies are separate and use different data and metrics.
Related work
- Deep learning for mangrove change prediction: Gaoqiao Mangrove, China (Frontiers in Marine Science, 2026) ↗Source for the 1993 to 2023 annual masks, the 0.928 F1 and the Gaoqiao study area.
- High-Accuracy Mangrove Extraction and Degradation Diagnosis Using Time-Series Remote Sensing and Deep Learning: A Case Study of the Largest Delta in the Northern Beibu Gulf, China (Water, 2026) ↗Source for the 0.822 IoU, the DEM constraints and the Nanliu River Delta setting.
- Advancing African Mangrove Forest Mapping Using Super-Resolution and Semantic Segmentation of Multi-Date Multi-Sensor Satellite Imagery (exa.ai publication record) ↗Source for the super-resolution IoU figures and the Tanzania and Ghana areas.
- Annual change in the distribution and landscape health of mangrove ecosystems in China from 2016 to 2023 with Sentinel imagery (Global Ecology and Conservation, via exa.ai record) ↗National Chinese mangrove study with a lower F1 of 0.843, for scale.
Watch next
- Open the African mangrove paper at its publisher to confirm venue and review status. Check whether a national annual Chinese mangrove map covers 1993 to 2023.
Sources
- Deep learning for mangrove change prediction: Gaoqiao Mangrove, China (Frontiers in Marine Science, 2026)frontiersin.org
- High-Accuracy Mangrove Extraction and Degradation Diagnosis Using Time-Series Remote Sensing and Deep Learning: A Case Study of the Largest Delta in the Northern Beibu Gulf, China (Water, 2026)mdpi.com
- Advancing African Mangrove Forest Mapping Using Super-Resolution and Semantic Segmentation of Multi-Date Multi-Sensor Satellite Imagery (exa.ai publication record)exa.ai
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 →