Peatland surface oscillation and carbon-stock change detection
Intact peatlands oscillate vertically by several centimetres each season; drained ones subside irreversibly. InSAR time-series turns millimetre-scale surface motion into a proxy for water-table depth, carbon storage and degradation risk.
Sensors
- Sentinel-1 A/B (C-band, 5.6 cm wavelength): Interferometric Wide Swath mode at 5 x 20 m resolution; 6-day repeat at mid-latitudes with both satellites. C-band loses coherence rapidly over dense Sphagnum and sedge canopy, limiting usable interferograms to short temporal baselines of 6 to 12 days, especially in summer. Archive from 2014 onwards.
- ALOS-2 PALSAR-2 (L-band, 23.6 cm wavelength): L-band penetrates low-stature peatland vegetation far more effectively than C-band, maintaining coherence over vegetated surfaces at 14-day repeat and 3 x 3 m to 10 x 10 m resolution depending on mode. Published studies on Indonesian peat domes use PALSAR-2 specifically because C-band coherence collapses over tropical swamp forest canopy.
- Envisat ASAR (C-band archive, 2002-2010): Provides a pre-Sentinel baseline for long-term subsidence trend analysis. 35-day repeat is a coherence liability on vegetated peat, but the archive depth allows decadal trend separation from seasonal oscillation when combined with PS-InSAR or SBAS processing.
- TanDEM-X bistatic DEM: Single-pass X-band DEM at roughly 12 m posting (global) and 6 m in high-resolution mode. Differencing two TanDEM-X acquisitions separated by years can detect bulk surface lowering from drainage-induced oxidation, bypassing coherence issues entirely. Vertical accuracy of the global product is approximately 1 m absolute, but relative change between acquisitions is better.
Why peat moves, and what that movement means for carbon
Peat is a poroelastic material. Its solid matrix of partially decomposed organic matter is saturated with water, and the water table effectively holds the surface up. When the water table rises after rainfall, the surface inflates; when it drops in summer or after artificial drainage, the surface deflates. In an intact raised bog this seasonal oscillation typically spans 2 to 8 cm vertically, directly tracking water-table fluctuation measured by dipwells. The relationship is not perfectly linear but is consistent enough that InSAR-derived displacement time-series have been used as a proxy for water-table depth across sites where dipwell density is low.
Drainage changes the story entirely. Once a peat surface is drained below a critical threshold, aerobic decomposition of the organic matrix begins. The carbon fixed over millennia oxidises to CO2. The surface does not spring back. Published subsidence rates on drained agricultural peat in Finland and the Netherlands range from roughly 1 to 3 cm per year, with tropical peatlands in Sumatra and Kalimantan showing rates of 3 to 5 cm per year or higher under oil-palm cultivation. At those rates the peat column is physically disappearing, and the carbon accounting implications are substantial. Detecting the transition from reversible oscillation to irreversible subsidence is the core monitoring problem.
The coherence problem: why C-band struggles and L-band helps
Interferometric coherence is the statistical correlation between two SAR acquisitions. Vegetation scatters radar in ways that change between passes, decorrelating the phase signal and making displacement measurement impossible. C-band (5.6 cm wavelength) is particularly vulnerable because the wavelength is comparable to leaf and stem dimensions in Sphagnum, sedge and cotton-grass canopies. Studies using Sentinel-1 over Finnish bogs report coherence values dropping below 0.3 in summer growing season, with usable interferograms largely confined to frozen or snow-covered winter conditions when the canopy is dormant.
L-band at 23.6 cm penetrates low-stature peatland vegetation and scatters preferentially from the wet peat surface and standing water beneath. ALOS-2 PALSAR-2 coherence over Finnish open bogs remains workable across seasons. Over tropical peat swamp forest in Indonesia the situation is more complex: the tall canopy still decorrelates L-band at longer temporal baselines, but 14-day PALSAR-2 interferograms have yielded coherent phase over open drainage canals and cleared areas, allowing subsidence mapping at canal margins where degradation is most acute. The honest summary is that no single sensor solves the problem everywhere; boreal open bogs favour Sentinel-1 in winter, tropical forested peat requires L-band, and bare or drained surfaces work at either frequency.
What published results from Finland and Indonesia actually show
Finnish studies using Sentinel-1 SBAS time-series over Lakkasuo and similar raised bogs have demonstrated seasonal displacement amplitudes of 3 to 6 cm correlating with dipwell records at r-squared values above 0.7. The signal is clearest in winter interferograms. Subsidence on drained agricultural margins adjacent to the same bogs is detectable as a secular trend separable from the seasonal oscillation once a time-series of two or more years is assembled.
Indonesian work using ALOS-2 PALSAR-2 over Sumatra and Kalimantan has mapped subsidence bowls around drainage canals, with rates consistent with published ground-truth measurements from levelling surveys. A key finding across multiple published studies is that the subsidence gradient away from canals follows a predictable decay curve: highest within 200 to 500 m of the canal, diminishing toward the undrained dome centre. This spatial pattern is itself diagnostic of drainage-induced degradation rather than natural compaction or loading.
Processing choices: SBAS, PS-InSAR and the seasonal-trend decomposition
Two InSAR time-series approaches dominate peatland work. Small Baseline Subset (SBAS) processing selects interferogram pairs with short temporal and spatial baselines to maximise coherence, then inverts the network to recover a displacement time-series at each coherent pixel. Persistent Scatterer InSAR (PS-InSAR) identifies individual stable reflectors, useful on drained peat where bare soil, infrastructure or dead vegetation provides point-like scattering, but sparse over intact vegetated bog.
The analytical challenge is decomposing the time-series into a seasonal oscillation component and a long-term trend. A simple sinusoidal model fitted to the seasonal signal leaves residuals that, if they accumulate monotonically, indicate irreversible subsidence. The separation requires at least two full annual cycles to be reliable; shorter time-series conflate a dry year's water-table drawdown with permanent loss. Atmospheric phase delay is a persistent nuisance: peat bogs are often in regions of variable humidity, and tropospheric artefacts at the centimetre level overlap with the signal of interest. ERA5 reanalysis data or GACOS correction is standard practice for mitigation, though residual errors of a few millimetres remain after correction.
Connecting surface motion to carbon: what the proxy can and cannot say
The poroelastic relationship between water-table depth and surface elevation has been parameterised in published studies well enough to estimate water-table depth from InSAR displacement, with uncertainties of roughly plus or minus 5 to 10 cm in water-table depth depending on site-specific peat properties. Water-table depth is itself a strong predictor of CO2 and methane flux: published flux-tower studies show that keeping the water table within 10 cm of the surface suppresses aerobic decomposition and maintains net carbon sequestration.
The honest limit is that InSAR gives surface elevation change, not a direct measurement of carbon flux. Converting displacement to carbon-stock change requires assumptions about peat bulk density, decomposition depth and the fraction of subsidence attributable to oxidation versus compaction. Those assumptions carry substantial uncertainty. What the satellite record can do reliably is flag sites where irreversible subsidence is occurring, rank them by rate, and track whether rewetting interventions are arresting the trend. Satellize's analytics pipeline applies this logic operationally; the Tonga crop-estimation programme is a different domain, but the underlying time-series decomposition methodology is shared infrastructure.
TanDEM-X DEM differencing adds a complementary check. Where repeat acquisitions exist, bulk surface lowering at the metre scale over a decade is unambiguous even without interferometric coherence. It is a coarser tool but one immune to the coherence constraints that limit InSAR on vegetated surfaces.
Practical limits a buyer should know before commissioning a survey
Cloud cover does not affect SAR. That is one genuine advantage over optical monitoring in the persistently cloudy tropics. But temporal decorrelation, atmospheric noise and the seasonal coherence window are real constraints that determine what is deliverable. Over intact tropical peat swamp forest, expect coherent pixels to be sparse and concentrated on open water, canals and cleared patches. Over boreal open bog, expect good results in winter and degraded results in summer. Revisit gaps in the ALOS-2 archive over specific sites can be substantial; PALSAR-2 does not offer the same systematic global coverage that Sentinel-1 does.
Minimum detectable displacement rates depend on the number of interferograms and atmospheric correction quality. Under favourable conditions, SBAS time-series can resolve secular trends of 3 to 5 mm per year over multi-year stacks. Faster subsidence rates of 1 cm per year or more are detectable within a single year. Spatial resolution of the displacement map is typically 20 to 100 m after multi-looking, which means individual drainage features narrower than roughly 20 m may not be resolved in their full spatial extent.
Typical figures
| Primary SAR frequency (C-band) | 5.405 GHz, 5.6 cm wavelength (Sentinel-1) |
| Primary SAR frequency (L-band) | 1.2565 GHz, 23.6 cm wavelength (ALOS-2 PALSAR-2) |
| Sentinel-1 ground resolution (IW mode) | 5 x 20 m (range x azimuth); 250 km swath |
| ALOS-2 PALSAR-2 resolution | 3 x 3 m (Spotlight) to 100 x 100 m (ScanSAR); standard stripmap 10 x 10 m |
| Revisit period | Sentinel-1: 6 days (dual satellite, mid-latitudes); ALOS-2: 14 days |
| Minimum detectable displacement rate (SBAS stack) | 3 to 5 mm/year under favourable atmospheric conditions; ~1 cm/year within a single year |
| Seasonal oscillation signal range (intact bog) | Typically 2 to 8 cm peak-to-trough; site-dependent on peat type and climate |
| Archive depth | Sentinel-1: from April 2014; ALOS-2: from 2014; Envisat ASAR: 2002 to 2010 |
| Coherence constraint (C-band, vegetated) | Usable interferograms largely confined to 6 to 12-day baselines; winter preferred in boreal sites |
| Delivery formats | GeoTIFF displacement maps, NetCDF time-series, shapefile subsidence-rate polygons, PDF site report |
Analytics Satellize can run
| Seasonal oscillation amplitude map | SBAS InSAR time-series with sinusoidal seasonal decomposition; Sentinel-1 winter interferogram stack | GeoTIFF raster of peak-to-trough displacement amplitude per pixel, updated annually |
| Irreversible subsidence rate map | Linear trend extraction from SBAS residuals after seasonal component removal; minimum two-year stack | Shapefile of subsidence-rate zones (mm/year) with confidence intervals; PDF interpretation report |
| Water-table depth proxy | Poroelastic inversion of InSAR displacement using published site-specific or literature-derived compressibility parameters | Gridded water-table depth estimate (cm below surface) with stated uncertainty bounds; GeoTIFF |
| Drainage-impact spatial extent | Canal-distance decay modelling of subsidence gradient from PALSAR-2 SBAS; L-band preferred for tropical sites | Polygon layer delineating degraded peat zones by distance from drainage infrastructure; GIS layer |
| Rewetting intervention effectiveness assessment | Before/after SBAS time-series comparison across canal-blocking or rewetting dates; trend inflection detection | Time-series chart per intervention site with displacement trend before and after; PDF report |
| Bulk surface-lowering estimate from DEM differencing | TanDEM-X or SRTM-to-TanDEM-X differencing; co-registration and bias correction using stable off-peat reference areas | GeoTIFF of net elevation change (m) over the differencing period; summary statistics table |
| Multi-decadal degradation trend (archive fusion) | Envisat ASAR SBAS chain extended through Sentinel-1; inter-sensor offset calibration using overlapping acquisition periods | Long-term displacement time-series from 2002 to present; NetCDF and GeoTIFF |
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.