Peatland drainage and subsidence monitoring using InSAR

Drained peatlands subside at rates of centimetres per year, releasing stored carbon invisibly. InSAR time series turns that subsidence into a spatially continuous, millimetre-scale record without a single ground benchmark.

Sensors

Sentinel-1 (C-band SAR, ESA): 5.6 cm wavelength; 5 x 20 m ground range resolution in Interferometric Wide Swath mode; 6-day repeat at mid-latitudes when both satellites are active. Useful over bare or sparsely vegetated peat surfaces but suffers rapid coherence loss over dense tropical vegetation, limiting interferogram quality to days rather than weeks.
ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength penetrates vegetation canopy and scatters from the soil-root zone rather than the leaf layer, preserving interferometric coherence over vegetated peatlands for temporal baselines of 14 to 42 days. Stripmap mode delivers 3 x 3 m resolution; ScanSAR wide mode gives 60 m at 350 km swath. Revisit is 14 days for a single satellite.
NISAR (L/S-band SAR, NASA/ISRO, forthcoming): Dual-frequency L-band (24 cm) and S-band (9 cm) system scheduled for launch in 2024-2025. Global 12-day repeat in L-band; designed explicitly for deformation time series. Will provide the first systematic global L-band InSAR dataset at consistent cadence, removing the ad-hoc tasking constraint that limits ALOS-2 coverage of remote peatlands.
Copernicus DEM (TanDEM-X derived): Global 30 m and 90 m digital elevation model derived from TanDEM-X bistatic X-band acquisitions. Used as the external DEM to remove topographic phase contribution from peatland interferograms. Vertical accuracy is approximately 4 m absolute and better than 1 m relative over flat terrain, which is adequate for peatland InSAR processing where slopes are gentle.

Why subsidence is the carbon signal you actually want

Peatlands store roughly 600 gigatonnes of carbon globally, accumulated over millennia in waterlogged, oxygen-depleted conditions. Drain them and the water table drops. Aerobic oxidation begins. The peat matrix compacts and oxidises, and the surface descends. That descent is not incidental: it is a direct physical proxy for carbon loss. Studies of drained tropical peatlands in Southeast Asia have recorded subsidence rates of 2 to 5 cm per year under oil palm and acacia plantations, implying carbon emissions of tens of tonnes per hectare annually.

Ground-based precise levelling can measure this with sub-millimetre accuracy, but it requires benchmarks, access roads and field teams. In the peat domes of Kalimantan or the boreal mires of western Siberia, none of those exist at the spatial density needed to map drainage impacts across millions of hectares. InSAR fills that gap by treating the entire illuminated surface as a continuous network of measurement points.

The coherence problem: why C-band struggles in the tropics

Interferometric SAR works by comparing the phase of radar returns from the same patch of ground across two acquisition dates. If the scatterers in that patch move or change between passes, the phase relationship degrades. This loss of coherence is the central practical constraint in peatland monitoring.

At C-band (Sentinel-1, 5.6 cm wavelength), the radar interacts primarily with vegetation canopy elements: leaves, small branches, the top of the herb layer. Tropical vegetation grows, sways and changes moisture content rapidly. Coherence over dense peat-swamp forest typically collapses to unusable levels within 6 to 12 days. That is precisely Sentinel-1's repeat interval, so many tropical peatland pixels yield no usable interferometric signal at all. Over bare, recently burned or drained peat with little standing vegetation, C-band performs better and can detect subsidence at rates above roughly 1 cm per year.

L-band radar at 23 cm wavelength penetrates the canopy and interacts with larger woody stems and the soil surface. The scatterers are physically more stable between passes. ALOS-2 PALSAR-2 studies over Sumatran and Bornean peatlands have demonstrated coherence retention sufficient for time-series analysis at 42-day baselines, enabling displacement maps across forested peat that C-band cannot touch. The trade-off is coarser coverage cadence and, until NISAR launches, limited systematic acquisition outside JAXA research campaigns.

From interferograms to subsidence time series: the processing chain

A single interferogram gives a snapshot of relative displacement between two dates, expressed as phase cycles. Atmospheric water vapour introduces phase delays that mimic ground deformation at magnitudes of 5 to 15 mm, which is comparable to or larger than the peat signal over short periods. Time-series methods, principally Small Baseline Subset (SBAS) and Persistent Scatterer InSAR (PS-InSAR), stack many interferograms and separate the temporally correlated deformation signal from the spatially correlated atmospheric noise.

Over vegetated peatlands, PS-InSAR is largely ineffective because there are few stable point scatterers. SBAS, which works with distributed scatterers and spatially filtered coherence, is the standard approach. Typical processing uses the Copernicus DEM or TanDEM-X for topographic phase removal, followed by interferogram network construction, phase unwrapping and inversion to produce a displacement time series at each coherent pixel. Vertical displacement rates in the line-of-sight direction are then decomposed assuming predominantly vertical motion, which is physically reasonable for compaction-driven peat subsidence.

Detection limits depend on coherence quality and atmospheric correction. Over bare or lightly vegetated peat with good Sentinel-1 coherence, SBAS time series can resolve annual subsidence rates of around 5 mm per year. Over vegetated peat with L-band data and adequate atmospheric correction using ERA5 reanalysis or GACOS delay maps, detection thresholds of 10 to 20 mm per year are realistic. Slower signals are recoverable only with long archives and dense temporal sampling.

Boreal peat versus tropical peat: different problems, same physics

Boreal peatlands in Scandinavia, Canada and Russia present a different coherence environment. Vegetation is lower and sparser than tropical peat-swamp forest, so C-band Sentinel-1 retains useful coherence over longer baselines, particularly in winter when frozen ground stabilises the surface. Seasonal freeze-thaw cycles introduce their own deformation signals that must be separated from drainage-driven subsidence, typically by analysing only summer acquisitions or by modelling the thermoelastic component explicitly.

Permafrost thaw in boreal and subarctic peatlands adds a further signal: thermokarst subsidence can exceed 10 cm per year in areas of rapid permafrost degradation, well above InSAR detection thresholds. Here Sentinel-1 SBAS time series have proven effective in published studies over Siberian and Alaskan sites, with the frozen-season acquisitions providing the most coherent interferograms. The physics is the same as tropical drainage subsidence, but the driver is thermal rather than hydrological.

What the maps cannot tell you, and what fills the gap

InSAR measures surface displacement in the radar line-of-sight direction. It does not directly measure water table depth, carbon flux or peat thickness. Converting a subsidence rate to a carbon emission estimate requires assumptions about peat bulk density, oxidation fraction and the relationship between water table drawdown and surface lowering. Published empirical relationships exist, notably from long-term monitoring sites in Southeast Asia, but they carry substantial uncertainty when applied to sites without ground calibration.

Layover and shadow in steep terrain are not a concern over flat peatlands, but dense tropical forest can still suppress coherence to the point where large areas produce no valid pixels. Those gaps are not random: they tend to coincide with the most intact, least-drained peat, which is precisely the area of highest conservation value. Interpreting a map that shows subsidence only in degraded areas and nothing in intact forest requires care; absence of signal is not the same as absence of change.

Satellize integrates InSAR displacement time series with optical indices and, where clients hold ground-truth water table records, builds site-specific calibration models. The approach mirrors the analytics framework applied in the Tonga crop-estimation programme: open-constellation data processed systematically, with honest uncertainty bounds attached to every output layer.

Practical acquisition planning for a peatland monitoring programme

For tropical peatlands, the first decision is whether the target area retains enough C-band coherence to use Sentinel-1 at all. A quick coherence test over a representative 12-day pair costs nothing using the Copernicus Data Space archive. If mean coherence over forested peat falls below 0.3, ALOS-2 tasking or NISAR planning is the correct path. For boreal sites, Sentinel-1 winter stacks are generally sufficient and the archive now extends back to 2014, providing a decade of displacement history.

Spatial resolution matters less than temporal sampling for subsidence time series. A 20 m pixel that moves 3 cm per year is detectable; a 5 m pixel with incoherent phase is not. Prioritise acquisition frequency and archive depth over resolution when specifying a programme. For sites where subsidence rates are expected to be slow, below 1 cm per year, plan for at least two years of systematic acquisition before drawing conclusions about trend.

Typical figures

Spatial resolution (Sentinel-1 IW mode)	5 x 20 m ground range; typically multi-looked to 40-80 m for InSAR time series
Spatial resolution (ALOS-2 Stripmap)	3 x 3 m single-look; typically processed at 10-30 m for peatland SBAS
Revisit period	6 days (Sentinel-1 dual satellite, mid-latitudes); 14 days (ALOS-2 single satellite); 12 days (NISAR, planned)
Radar frequency / wavelength	C-band 5.405 GHz / 5.6 cm (Sentinel-1); L-band 1.2 GHz / 23.6 cm (ALOS-2, NISAR)
Minimum detectable annual subsidence rate	~5 mm/yr over bare/low-vegetation peat (C-band SBAS); ~10-20 mm/yr over vegetated peat (L-band SBAS with atmospheric correction)
Atmospheric phase correction	ERA5 reanalysis or GACOS tropospheric delay maps; residual error typically 5-15 mm per interferogram
Reference DEM	Copernicus DEM GLO-30 (30 m, ~4 m absolute vertical accuracy, <1 m relative over flat terrain)
Archive depth	Sentinel-1: October 2014 to present; ALOS-2: 2014 to present (campaign-based); ALOS-1 PALSAR: 2006-2011
Swath width	250 km (Sentinel-1 IW); 50 km Stripmap / 350 km ScanSAR (ALOS-2)
Deliverable formats	GeoTIFF displacement rate maps, NetCDF time-series stacks, GIS-ready shapefiles with uncertainty polygons

Analytics Satellize can run

Annual subsidence rate map	SBAS InSAR time-series inversion over multi-year Sentinel-1 or ALOS-2 stack	GeoTIFF raster of line-of-sight displacement rate (mm/yr) with per-pixel standard deviation layer
Drainage-event detection	Change-point analysis on pixel-level displacement time series to identify onset of accelerated subsidence coinciding with canal construction	Dated polygon layer marking areas where subsidence rate increased beyond a client-defined threshold, with confidence score
Carbon-loss proxy index	Subsidence rate converted to indicative CO2-equivalent emission using published bulk-density and oxidation-fraction relationships from tropical peat literature	Tabular summary by land-cover zone with explicit uncertainty range; not a certified carbon account
Coherence quality assessment	Mean coherence mapping over candidate monitoring area using representative 12-day and 42-day interferogram pairs	Advisory report on C-band versus L-band suitability, with recommended acquisition strategy and expected detection limits for the specific site
Multi-epoch deformation time series	SBAS network inversion with atmospheric correction using ERA5 or GACOS, referenced to stable non-peat benchmark pixels	NetCDF stack of cumulative displacement per acquisition date, suitable for import into client GIS or time-series visualisation tools
Intact-versus-degraded peat classification	Fusion of InSAR subsidence rate, Sentinel-1 backscatter amplitude and optical land-cover indices to stratify peat condition	Annual GIS layer distinguishing stable, slowly subsiding and rapidly subsiding peat zones, updated on each new SAR acquisition cycle

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.