Interactive map
3-D LiDAR point cloud
Point size
1 · Study area & data
The AOI is the convex hull of the 13 mangrove polygons from the prior study (~4.4 ha). AlphaEarth embeddings (64-band, 10 m, annual 2017–2025) and Sentinel-2 (15-band, 10 m) provide the satellite record; a DJI Zenmuse L1 drone (Dec 2025) provides a centimetre orthophoto and 61 M-point LiDAR cloud.
2 · Coverage & change (AlphaEarth)
One k-means is fit on all hull pixels × years and applied per year, so cluster identities stay consistent; the cluster most overlapping the prior polygons is labelled mangrove. Coverage . Growth occurs on the periphery of a stable core, with effectively no decay.
3 · Sentinel-2 cross-check
An independent clustering on a 15-band Sentinel-2 stack (10 m + 20 m bands plus NDVI/NDRE/NDWI/MNDWI/NDMI) reproduces the same expansion (), confirming the result is not an artefact of one feature set.
4 · Biomass & carbon
ETH 10 m canopy height (2020) → AGB via the Simard 2019 mangrove allometry gives a reference biomass; a Sentinel-2 spectral-index model extends it to all years. Carbon stock and sequestration rate follow.
5 · Drone orthophoto & clustering
An OpenDroneMap orthophoto from 114 DJI Zenmuse L1 images is clustered with density-based HDBSCAN on RGB + colour indices (ExG, VARI) at half-metre scale — a fully independent, very-high-resolution mangrove delineation.
6 · LiDAR validation
The L1 LiDAR (Dec 2025) is the structural ground truth: a 10 m CHM built on the satellite grid. Extent agreement is strong (); per-pixel structure correlates only weakly — satellite maps where mangroves are, LiDAR resolves their structure.
7 · Final integration — LiDAR biomass (two masks)
The most accurate estimate fuses every source, side by side over two mangrove delineations. LiDAR points in the cluster-defined ground cells build the ground surface (interpolated under the closed canopy); LiDAR points give canopy height (Z − ground) and tree density; height → biomass via the Simard allometry. Only the mangrove mask differs: the cm drone orthophoto (HDBSCAN) versus the 10 m AlphaEarth satellite mask.
| Mangrove mask | Area (ha) | Mean H (m) | Trees/ha | AGB (Mg) | Carbon (Mg C) |
|---|
The LiDAR reveals a predominantly young, short stand (mean canopy ~2.6–3.2 m) — consistent with the rapid 2017→2025 expansion. Both masks converge on ~33–35 Mg C, far below the satellite/ETH height-allometry estimates (≈112 Mg C), which overestimate for short canopy.
8 · Synthesis
Four independent sensors — AlphaEarth, Sentinel-2, a drone orthophoto, and LiDAR — agree that the Hiyagon mangrove stand expanded markedly over 2017–2025, sequestering carbon as it grew. Satellite methods reliably map extent and change at 10 m; the drone and LiDAR add centimetre structure for the current state, and the LiDAR×clustering fusion gives the most accurate, measured biomass — revealing a young, still-establishing stand. Caveats: clusters are unsupervised (anchored to prior polygons); the satellite biomass series is a spectral extrapolation of a single-epoch canopy-height product; 2017 Sentinel-2 is unavailable over Japan.