Mangrove extent and blue-carbon stock mapping

Mangroves sequester carbon at rates far exceeding terrestrial forests, yet their extent is routinely overstated in project baselines. Combining L-band SAR with multispectral time series gives auditable, cloud-independent maps of canopy extent, zonation and above-ground biomass for Verra VM0033 and IUCN blue-carbon accounting.

Sensors

ALOS-2 PALSAR-2: L-band SAR at 1.27 GHz. The long wavelength penetrates the canopy and produces a strong double-bounce return from vertical stems over standing water or saturated soil, the physical signature that separates mangrove from saltmarsh and mudflat. Single-polarisation ScanSAR mode delivers 10 m resolution; fine-beam dual-polarisation reaches 6 m. Revisit is 14 days.
Sentinel-1 C-SAR: C-band SAR at 5.405 GHz, 10 m ground range resolution in IW mode, 6-day revisit at mid-latitudes with both satellites operational. The shorter wavelength interacts more with upper-canopy structure than with stems, so double-bounce is weaker than L-band but still detectable in mature stands. Useful for change detection and inundation timing; less reliable for sparse or young regenerating fringe.
Sentinel-2 MSI: 10 m visible and near-infrared bands, 20 m red-edge and SWIR, 5-day revisit. NDVI, EVI and red-edge chlorophyll indices separate mangrove from surrounding vegetation, but tidal stage at acquisition time shifts the apparent forest boundary by tens of metres in low-gradient settings. Cloud cover over tropical coasts routinely exceeds 70 percent, so time-series compositing across multiple acquisitions is essential before any extent measurement is taken.
Planet SuperDove: 3 m resolution, 8 spectral bands including red-edge, near-daily revisit. Resolves individual canopy gaps and small clearing events that are sub-pixel in Sentinel-2. Useful for validating SAR-derived extent boundaries and for detecting early disturbance in project areas, though the commercial archive and tasking costs require client licence.

What a floating roof gives away

Mangrove canopies sit on a lattice of prop roots and pneumatophores above tidal water or saturated sediment. That geometry is the key to radar mapping. When a microwave pulse at L-band strikes a vertical stem and reflects from the water surface beneath, it returns to the sensor with roughly twice the path delay of a direct surface return. This double-bounce mechanism produces backscatter values that are distinctly elevated relative to open water, mudflat or even saltmarsh, which lacks the stem-water geometry. ALOS-2 PALSAR-2 exploits this most cleanly: published studies using JERS-1 and ALOS PALSAR data have demonstrated mangrove classification accuracies above 90 percent in HH polarisation, a figure that holds across cloud-free and heavily overcast conditions alike because SAR is indifferent to cloud.

C-band Sentinel-1 produces a weaker but still usable double-bounce signal in mature stands where stems are thick relative to the 5.6 cm wavelength. In young regenerating fringe, where stems are thin and sparse, the signal collapses toward the noise floor. This is an honest limitation: Sentinel-1 alone will undercount recovering mangrove in the first three to five years after replanting, which matters considerably for restoration project baselines.

The tidal-stage problem in optical imagery

Sentinel-2 and Planet imagery of mangrove edges are not snapshots of a fixed boundary. They are snapshots of a boundary at a particular tidal stage. At low tide, mudflat and pneumatophore zones are exposed; the apparent forest edge retreats landward. At high tide, water floods the understorey and the same forest edge appears to advance seaward. In a low-gradient intertidal setting, this tidal excursion can shift the mapped boundary by 30 to 80 metres, an error that translates directly into project area and therefore carbon credit volume.

The practical remedy is to filter optical acquisitions by tidal height before compositing. Tide gauge records or global tidal models such as TPXO can be used to tag each Sentinel-2 granule with an estimated local water level at acquisition time. Composites built from acquisitions within a narrow tidal window (typically mean sea level ± 0.2 m) are far more consistent. Where tidal records are absent, SAR-derived inundation frequency maps provide an independent check on the optical boundary.

Zonation, species and the biomass gradient

Mangroves are not a uniform block. They typically show a landward-to-seaward zonation: tall closed-canopy Rhizophora or Avicennia at the forest interior, grading to shorter fringe species and eventually to sparse pneumatophore fields at the tidal flat margin. Above-ground biomass density tracks this gradient closely, with interior stands sometimes exceeding 200 Mg ha⁻¹ and fringe stands falling below 50 Mg ha⁻¹. A project that maps only the dense interior will overstate average stock density; one that includes the pneumatophore zone will overstate area.

Red-edge bands on Sentinel-2 and SuperDove are sensitive to canopy chlorophyll content and can partially separate species assemblages, though species-level classification from space remains uncertain in mixed or degraded stands. The more reliable approach for VM0033 purposes is to use SAR-derived canopy height proxies (backscatter intensity correlates loosely with stand height in L-band) combined with field-calibrated allometric equations to produce a continuous biomass map, then stratify it by zone for accounting.

One quantity satellite imagery cannot supply is below-ground biomass, which in mangroves includes extensive root systems and can equal or exceed above-ground stocks. VM0033 and the IUCN framework both require below-ground estimation, but that figure must come from published root-to-shoot ratios applied to satellite-derived above-ground estimates, not from direct remote sensing. Auditors should treat the below-ground component as a modelled quantity with explicit uncertainty bounds, not a measured one.

Pneumatophore zones: the mapping gap no one advertises

Pneumatophores are the pencil-thin aerial roots that project upward from the sediment surface in Avicennia and Sonneratia stands. They are ecologically part of the mangrove system and contribute to sediment carbon accumulation, but they are essentially invisible to optical sensors at any commercially available resolution, and their SAR return is dominated by the mudflat beneath rather than by the root structure above.

This is not a solvable problem with current satellite constellations. The pneumatophore zone boundary must be established by field survey and then held fixed between monitoring epochs unless a structural change is clearly visible in high-resolution imagery. Project developers using satellite MRV should document this gap explicitly in their methodology and apply a conservative buffer that excludes ambiguous fringe pixels from the accounting perimeter.

Building an audit-grade time series

Carbon registries require that baseline extent and subsequent monitoring epochs be derived from consistent methods. That means the same sensor, the same tidal-stage filter, the same classification algorithm and the same accuracy-assessment protocol applied at each monitoring date. Archive depth matters: ALOS PALSAR data from the JAXA Global Mangrove Watch programme extends back to 1996, providing a 25-plus year baseline against which project-era change can be measured. Sentinel-1 provides consistent global coverage from 2014 onward.

Change detection between epochs should be reported as net area change with a confidence interval derived from classification accuracy matrices, not as a single point estimate. Where disturbance is detected, the date of disturbance can often be bracketed to within one revisit cycle (6 days for Sentinel-1, 14 days for PALSAR-2), which supports permanence monitoring and reversal buffer calculations.

Satellize structures its mangrove analytics around exactly this audit trail: each delivered map layer carries provenance metadata linking it to the source granules, tidal-stage filter parameters and classifier version, so that a third-party verifier can reproduce the result from the same inputs. The methodology is analogous to the approach used in our Tonga crop-estimation programme, where reproducibility of the analytic chain was a contractual requirement.

Honest limits of the satellite record

Three limitations deserve plain statement. First, above-ground biomass estimates from SAR backscatter saturate at roughly 100 to 150 Mg ha⁻¹ in C-band and at higher values in L-band, but the saturation threshold in dense mangrove is not well constrained and varies by stand structure. High-biomass interior stands will be underestimated. Spaceborne lidar (GEDI, ICESat-2) can help calibrate the upper end of the biomass distribution, though GEDI's 25 m footprint and sparse sampling are poorly suited to narrow fringe mangroves. That is a separate page in this library.

Second, spatial resolution limits the minimum detectable clearing. At 10 m, Sentinel-1 and Sentinel-2 will miss clearings smaller than roughly 0.01 ha, which is relevant for small-scale artisanal harvesting. Planet SuperDove at 3 m lowers this threshold substantially but requires commercial tasking.

Third, sediment carbon, the largest component of the blue-carbon stock in many mangrove systems, is entirely beyond what any current satellite can measure directly. The satellite contribution to blue-carbon MRV is real and growing, but it covers above-ground biomass and canopy extent. Everything below the sediment surface remains a field-measurement and modelling problem.

Typical figures

Spatial resolution (SAR, L-band)	6–10 m (ALOS-2 PALSAR-2 fine-beam to ScanSAR)
Spatial resolution (SAR, C-band)	10 m (Sentinel-1 IW mode)
Spatial resolution (optical)	3 m (Planet SuperDove); 10–20 m (Sentinel-2 MSI)
Revisit cadence	6 days (Sentinel-1, dual-satellite); 14 days (ALOS-2); near-daily (Planet)
SAR frequency	1.27 GHz L-band (ALOS-2); 5.405 GHz C-band (Sentinel-1)
Minimum detectable clearing	~0.01 ha at 10 m SAR; ~0.001 ha at 3 m optical
Above-ground biomass saturation (L-band)	Approximately 150–200 Mg ha⁻¹; site-dependent
Archive depth	JAXA Global Mangrove Watch: 1996–present; Sentinel-1: 2014–present
Cloud sensitivity	SAR: none. Optical: requires multi-date compositing with tidal-stage filter
Delivery formats	GeoTIFF biomass raster, GeoPackage extent polygons, CSV change-detection summary, PDF verification report

Analytics Satellize can run

Mangrove canopy extent map	L-band double-bounce threshold classification combined with optical NDVI time-series compositing filtered by tidal stage	GeoPackage polygon layer with per-patch area, confidence class and acquisition metadata
Above-ground biomass raster	Empirical backscatter-to-biomass regression calibrated against published allometric equations; stratified by SAR-derived canopy height proxy	GeoTIFF at 10 m resolution with uncertainty band raster; summary statistics by project stratum
Zonation classification	Unsupervised clustering of L-band HH/HV ratio and Sentinel-2 red-edge index to separate interior, fringe and pneumatophore-margin zones	GeoTIFF zone map with accuracy matrix; zone-level biomass summary table
Multi-epoch change detection report	Bi-temporal SAR coherence and backscatter differencing; optical NDVI trajectory analysis; change area reported with 95% confidence interval	PDF verification report with change polygons, epoch dates and tidal-stage provenance metadata
Tidal-stage corrected optical composite	Sentinel-2 granule filtering against TPXO tidal model; median compositing within ± 0.2 m water-level window	GeoTIFF composite with per-pixel acquisition count and tidal-height range metadata
VM0033-aligned baseline area and stock estimate	Stratified area estimation following Cochran (1977) design-based inference; biomass stocks derived from above-ground raster with published root-to-shoot ratios applied for below-ground modelling	Structured data package formatted for Verra VM0033 monitoring report submission, including uncertainty tables

Sources

Who does the work

We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.