Peatland and wetland heritage site water-table monitoring via SAR coherence
Drainage, drought and agricultural abstraction that lower peatland water tables expose millennia-old organic material to irreversible aerobic decay. Sentinel-1 InSAR coherence and backscatter change provide landscape-scale early warning, validated against dipwell records from Irish and Scandinavian raised bogs.
Sensors
- Sentinel-1 SAR (C-band, 5.405 GHz): Interferometric Wide swath mode at 5 x 20 m ground range resolution, 6-day repeat at mid-latitudes with both satellites active. Coherence between repeat passes degrades when surface moisture changes; backscatter intensity responds to surface roughness and dielectric constant shifts driven by water-table position. Cloud-independent, day-and-night acquisition.
- Sentinel-2 MSI: 10 m visible and near-infrared bands, 20 m shortwave infrared (SWIR, bands 11 and 12). SWIR reflectance is sensitive to surface and near-surface moisture in bare peat; NDWI and NDMI indices track surface wetness at 5-day revisit (combined Sentinel-2A and 2B). Cloud cover over Atlantic bogs limits usable acquisitions to roughly 30–50 % of passes in maritime climates.
- MODIS (Terra/Aqua): 250–500 m resolution land surface reflectance and the MODIS land surface temperature product. Useful for basin-scale moisture anomaly context and multi-decade trend baselines back to 2000, though spatial resolution is too coarse to resolve individual drainage ditches or small bog complexes under about 1 km².
- Sentinel-1 InSAR coherence stacks: Multi-temporal coherence matrices derived from 6-day or 12-day pairs. Persistent low coherence indicates stable open water or saturated peat; coherence recovery after a dry period indicates surface desiccation. Phase-based subsidence signals from peat compaction under drainage can reach several centimetres per year and are detectable at millimetre-per-year precision using PS-InSAR or SBAS methods over stable reference points.
What the water table is actually protecting
Waterlogged peat is anoxic. Bacteria that decompose wood, leather, textile and bone cannot operate without oxygen, so objects buried in saturated bog can survive for thousands of years in near-perfect condition. The Tollund Man, the Clonycavan Man, Bronze Age trackways across the Irish Midlands, and the wooden structures of the Somerset Levels all owe their survival to an unbroken column of water above them.
The threat is not dramatic. A drainage ditch dug a kilometre away, a dry summer, a farmer deepening a field drain: any of these can lower the local water table by tens of centimetres. Once the peat surface dries and cracks, oxygen penetrates. Decay that took millennia to not happen can proceed in years. The loss is invisible from the road and, until recently, invisible from orbit.
Why C-band SAR coherence responds to water-table change
Synthetic aperture radar at C-band (roughly 5.6 cm wavelength) interacts primarily with the top few centimetres of a vegetated or bare peat surface. The dielectric constant of water is approximately 80, compared with roughly 3–5 for dry peat. As the water table drops and the capillary fringe retreats from the surface, the bulk dielectric constant of the peat layer falls sharply. This changes the backscatter intensity (sigma-nought) and, critically, alters the phase and amplitude of the return between repeat passes.
Coherence, the correlation coefficient between two SAR acquisitions over the same area, is high when the scattering geometry is stable between passes. A saturated, flat bog surface with low vegetation is geometrically stable and tends to maintain moderate coherence. When the surface dries, it shrinks, cracks and subsides unevenly; coherence drops. Conversely, a surface that was desiccated and then rewetted by rainfall or management intervention shows a coherence signature distinct from one that remains dry. Published studies from Finnish and Irish raised bogs, including work using ERS-1/2 and later Sentinel-1, have demonstrated statistically significant correlations between 6-day coherence values and dipwell water-table depth, particularly in the range of 0 to 40 cm below surface.
The honest limit: coherence is not a direct water-table measurement. It is a proxy that conflates surface roughness change, vegetation phenology, frost, and rain-induced surface scatter. Separating these requires multi-temporal stacks, seasonal decomposition, and ideally at least a sparse network of in-situ dipwells for calibration. A single coherence image is ambiguous; a time series of 50 or more 6-day pairs is not.
Backscatter intensity as a complementary signal
Coherence captures phase stability. Backscatter intensity captures the raw energy returned to the sensor. Over bare peat, VV-polarisation backscatter at C-band increases with surface moisture because the higher dielectric constant of wet peat returns more energy. Dual-polarisation Sentinel-1 data (VV and VH) allows a simple ratio that partially suppresses wind-roughness effects on open water and isolates the dielectric signal.
In practice, backscatter and coherence are used together. A site showing simultaneously falling backscatter and recovering coherence is a strong candidate for surface desiccation. Either signal alone is equivocal. Sentinel-2 SWIR indices add a third, optically independent line of evidence when cloud permits, and MODIS provides the multi-decade context needed to distinguish a management-induced trend from natural interannual variability.
What the published record from Irish and Scandinavian bogs shows
Research published in Remote Sensing (MDPI) and related journals has tested Sentinel-1 coherence against dipwell networks on Irish raised bogs managed by the National Parks and Wildlife Service and on Scandinavian mires. The broad finding is that 6-day coherence in the VV channel correlates with water-table depth at r-values typically in the range of 0.5 to 0.75 depending on bog type, vegetation cover and season. The relationship is strongest in spring and autumn when vegetation is not actively growing and surface conditions are dominated by moisture rather than phenology.
Peat subsidence detected by PS-InSAR over drained bogs in the Netherlands and East Anglia has been measured at rates of 1 to 3 cm per year, consistent with oxidation and compaction losses documented in the ground record. These rates are well within Sentinel-1's detection capability. For heritage managers, a subsidence signal of this magnitude sustained over two or three years is a credible indicator that the anaerobic preservation environment is being compromised.
The approach does not work equally well everywhere. Forested bogs scatter radar energy from the canopy rather than the surface, masking the peat moisture signal almost entirely. Dense Sphagnum hummock microrelief at centimetre scale introduces spatial heterogeneity that 5 x 20 m pixels average over. Sites smaller than roughly two to three Sentinel-1 pixels in extent are effectively unmonitorable by this method alone and require in-situ instrumentation.
Building an operational monitoring workflow
An operational system ingests every 6-day Sentinel-1 IW acquisition over the site, computes coherence and backscatter change maps relative to a wet-season baseline, and flags grid cells where coherence has recovered above a threshold or backscatter has fallen below one. Thresholds are site-specific and require an initial calibration season against dipwell data. Where no dipwells exist, the first year of satellite data establishes a relative baseline rather than an absolute one.
Alert latency is principally a function of SAR processing time and the 6-day repeat. A heritage manager can realistically expect to receive a flag within 10 to 14 days of a water-table change that is large enough to alter the surface dielectric signature. That is not fast enough to prevent a single drought event but is fast enough to trigger emergency rewetting interventions, such as ditch blocking, before a dry season compounds into a second or third year of exposure.
Satellize runs coherence change analytics on open Sentinel-1 stacks and can integrate dipwell telemetry feeds where clients have installed them. The Tonga crop-estimation programme demonstrated that time-series calibration against sparse ground truth is operationally viable at national scale; the same statistical framework applies here. A sensible first step for a heritage body is a one-season pilot covering a single priority site with at least four dipwells, generating a calibration curve before committing to landscape-scale deployment.
Honest limits and what they mean for heritage managers
Satellite monitoring cannot replace dipwells. It can prioritise where dipwells are needed most urgently across a large landscape. A national heritage agency managing dozens of bog sites cannot instrument every one; coherence mapping identifies the three or four sites where the signal has changed most sharply and concentrates ground resources there.
Cloud cover is not a problem for SAR but is a problem for the Sentinel-2 optical corroboration layer. In persistently cloudy maritime climates, optical validation may be available only a handful of times per year. The SAR signal stands alone most of the time, which is acceptable given its physical basis but means anomalies should be treated as alerts requiring ground-truth rather than confirmed diagnoses.
Finally, the method monitors the current state of the water table. It cannot retroactively assess how long a site has been desiccated before monitoring began. Establishing an archive baseline using the Sentinel-1 record back to 2014 is therefore the first analytical act, not an optional enhancement.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 5 m range × 20 m azimuth (ground range); commonly processed to 10 × 10 m or 20 × 20 m pixels |
| Sentinel-1 revisit interval | 6 days at mid-latitudes with both Sentinel-1A and 1B active; 12 days with one satellite |
| Sentinel-2 optical resolution | 10 m (VIS/NIR), 20 m (SWIR bands 11 and 12 used for moisture indices) |
| Sentinel-2 revisit interval | 5 days combined (2A + 2B); usable cloud-free passes 30–50 % in Atlantic climates |
| SAR frequency and polarisation | C-band 5.405 GHz; dual-pol VV + VH in IW mode |
| Coherence sensitivity depth | Surface to approximately 5 cm in vegetated peat; deeper penetration in dry bare peat |
| InSAR subsidence detection limit | Approximately 1–3 mm per year using PS-InSAR or SBAS over stable reference points |
| Sentinel-1 archive depth | From April 2014 (Sentinel-1A launch); continuous near-global coverage from 2016 |
| MODIS contextual archive depth | From 2000 (Terra) and 2002 (Aqua); 250–500 m resolution |
| Alert latency (operational) | 10–14 days from acquisition to flagged coherence anomaly, subject to processing pipeline |
Analytics Satellize can run
| 6-day coherence change map | Sentinel-1 IW interferometric coherence estimation from SLC pairs; change relative to wet-season baseline | GeoTIFF raster layer per acquisition cycle, clipped to site boundary, with anomaly pixels flagged |
| Backscatter trend index | Multi-temporal VV and VH sigma-nought time series; linear trend and seasonal decomposition per pixel | Annual trend raster and site-level summary report (PDF + CSV) |
| Peat surface subsidence map | PS-InSAR or SBAS processing of Sentinel-1 SLC stack; line-of-sight displacement converted to vertical using incidence angle | Displacement velocity raster (mm/year) and time-series plots for user-selected points; GIS layer |
| Optical moisture index time series | NDMI and NDWI derived from Sentinel-2 SWIR and NIR bands; cloud-masked composites | Monthly cloud-free composite rasters and pixel-level time-series chart export |
| Water-table proxy alert | Threshold exceedance on coherence recovery or backscatter drop, calibrated against dipwell records where available | Email or API alert with site ID, pixel count affected, and magnitude of anomaly; updated each 6-day cycle |
| Multi-decade desiccation baseline | Retrospective coherence and backscatter analysis of full Sentinel-1 archive (2014–present) plus MODIS moisture anomaly context (2000–present) | Historical desiccation chronology report with annotated timeline of significant anomaly episodes |
| Priority site ranking | Composite score from coherence trend, subsidence rate and optical moisture anomaly across a portfolio of bog heritage sites | Ranked site table with supporting evidence layers; suitable for capital allocation decisions |
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.