SAR coherence mapping of seasonal waterlogging and anaerobic preservation zones
Repeat-pass SAR coherence drops sharply when soil moisture rises between acquisitions. Mapping that seasonal loss identifies ground where anaerobic conditions preserve organic archaeological material, giving heritage managers a spatially explicit preservation envelope.
Sensors
- Sentinel-1 C-band SAR (ESA): 5.6 cm wavelength; Interferometric Wide Swath mode at 5 x 20 m ground range resolution, 250 km swath. Six-day exact repeat at mid-latitudes when both satellites are operational, giving coherence pairs at 6, 12 and 24-day intervals. C-band is sensitive to surface and shallow subsurface moisture but attenuates in dense canopy.
- ALOS-2 PALSAR-2 L-band SAR (JAXA): 23.6 cm wavelength penetrates vegetation canopy and dry topsoil more effectively than C-band. Fine-beam single polarisation mode reaches 3 m resolution; standard mode 10 m. Repeat cycle 14 days. L-band coherence is more stable over vegetated ground, making moisture-driven decorrelation easier to isolate from canopy motion noise.
- RADARSAT-2 C-band SAR (MDA): Flexible multi-polarisation modes from 3 m (Ultra-Fine) to 100 m (ScanSAR). Exact repeat 24 days, though non-exact repeats can be tasked at shorter intervals at the cost of larger perpendicular baseline, which complicates coherence interpretation. Useful for targeted acquisition over specific sites where Sentinel-1 geometry is suboptimal.
- Sentinel-1 archive (ESA Copernicus): Continuous acquisitions since April 2014 for Sentinel-1A, October 2016 for Sentinel-1B (now inactive). The multi-year archive allows seasonal coherence stacks spanning many hydrological cycles, which is necessary to distinguish persistent waterlogging from single-event flooding.
Why coherence falls when the ground gets wet
Repeat-pass SAR coherence measures how similar the phase of a radar signal is between two acquisitions taken from nearly the same orbital position. When the ground surface or shallow subsurface changes between passes, phase relationships break down and coherence falls toward zero. Soil moisture is one of the most reliable drivers of this decorrelation because water has a dielectric constant of roughly 80, compared with around 3 to 5 for dry mineral soil. A modest increase in volumetric moisture content, from 10 to 30 percent, can raise the bulk dielectric constant enough to shift the effective scattering depth by several centimetres, randomising the phase return.
The practical consequence for archaeology is that ground which floods seasonally, or which maintains a perched water table through winter, produces a distinctive low-coherence signature during the wet season and recovers higher coherence when it dries. The spatial pattern of that seasonal swing, computed from coherence pairs spanning wet and dry acquisitions, approximates the boundary of the anaerobic preservation envelope. Organic materials, timber piles, leather, basketry and pollen assemblages, survive in the archaeological record precisely where oxygen is excluded by sustained waterlogging. The coherence map does not detect the artefacts; it identifies the ground conditions that allow them to persist.
Sentinel-1 versus L-band: choosing the right frequency for the site
Sentinel-1 C-band is the default starting point. Free data, six-day repeat and a decade-long archive make seasonal coherence stacking straightforward. The limitation is penetration depth. At 5.6 cm wavelength, C-band interacts primarily with the top few centimetres of bare soil and with the canopy of any vegetation present. Over grassland or arable fields, canopy motion between acquisitions adds temporal decorrelation that is unrelated to moisture, compressing the dynamic range of the moisture signal. Short coherence pairs, six days, minimise this noise but also reduce the moisture contrast if the wet-to-dry transition is slow.
ALOS-2 PALSAR-2 L-band addresses the canopy problem. At 23.6 cm, the signal passes through grass swards and low crops with far less attenuation, reaching the soil surface more reliably. The 14-day repeat is longer than Sentinel-1's, which means each coherence pair spans more of the drying or wetting cycle, but it also means more temporal decorrelation from vegetation growth. The practical approach is to use Sentinel-1 for initial mapping across large areas and bring in PALSAR-2 to resolve ambiguous boundaries, particularly on sites with permanent pasture or scrub cover. RADARSAT-2 is worth considering where a specific acquisition geometry is needed to reduce layover or shadow over earthwork topography.
Building a seasonal coherence stack: method and honest limits
The standard workflow registers a time series of single-look complex SAR images to a common reference geometry, then computes interferometric coherence for pairs spanning the wet season (typically November to March in temperate northwest Europe) and the dry season (June to August). The difference between mean wet-season and mean dry-season coherence, sometimes called the coherence change index, produces a raster in which high positive values indicate ground that decorrelates strongly when wet. Thresholding that raster, usually guided by field soil-moisture observations or ancillary drainage maps, gives a candidate preservation zone boundary.
Several limits deserve stating plainly. First, coherence is sensitive to any change between acquisitions, not just moisture. Cultivation, grazing, construction and even heavy rainfall on dry soil all decorrelate the signal. Multi-year stacking reduces these transient effects but does not eliminate them. Second, the method maps surface and near-surface moisture; it cannot confirm that waterlogging extends to the depth of archaeological deposits, which may lie one to three metres below ground. Third, C-band penetration in clay-rich soils is shallow enough that the signal may not reach the water table at all. The coherence envelope is a probabilistic indicator, not a confirmed preservation boundary, and it must be ground-truthed against borehole or auger data before management decisions are made. Fourth, dense woodland produces near-zero coherence regardless of soil moisture, making the method blind beneath closed canopy.
What the archive already holds
Sentinel-1A has been acquiring data since April 2014. That is more than ten complete hydrological years over most of Europe, available at no cost through the Copernicus Data Space Ecosystem. For a heritage manager assessing a site, this means a seasonal coherence climatology can be computed without a single new acquisition. The archive also captures anomalous years: the exceptionally dry summer of 2018 across northern Europe, for instance, produced coherence values over normally wet ground that would be invisible in a single-year analysis. Identifying sites where coherence remained low even in drought years is a strong indicator of genuinely deep, sustained anaerobic conditions.
PALSAR-2 archive depth is shorter and access is through JAXA's AUIG2 portal, which requires registration. Coverage is not global on a fixed schedule in the way Sentinel-1 is, so L-band archive availability over a specific site should be checked before committing to an L-band workflow. RADARSAT-2 archive access is commercial.
From coherence map to heritage management output
The immediate deliverable is a seasonally resolved coherence-change raster, typically at 10 to 20 m pixel spacing for Sentinel-1 or 10 m for PALSAR-2, clipped to the study area and classified into preservation-probability zones. That raster integrates directly with GIS layers of known site locations, scheduled monument boundaries and developer red-line boundaries. The combination allows a heritage authority to flag which proposed development footprints intersect candidate anaerobic preservation zones, and to specify targeted evaluation methods, principally coring and palaeoenvironmental sampling, rather than blanket area excavation.
Satellize can run this workflow on the Sentinel-1 archive for any site where a client provides a boundary polygon, producing a coherence climatology report with GIS deliverables. The analytical approach is the same class of repeat-pass coherence processing used in the Tonga crop-estimation programme, adapted from agricultural to heritage contexts. A useful next step for any heritage authority is to define the study polygon and specify the hydrological seasons relevant to their climate, since the wet-dry contrast that drives the method varies considerably between Atlantic, continental and Mediterranean regimes.
Typical figures
| Spatial resolution (Sentinel-1 IW mode) | 5 x 20 m ground range; coherence typically computed at 10 x 10 m after multi-looking |
| Spatial resolution (PALSAR-2 Fine Beam) | 3 m single-look; 10 m standard mode |
| Revisit period | 6 days (Sentinel-1, mid-latitudes, single satellite); 14 days (PALSAR-2) |
| Radar frequency | C-band 5.405 GHz (Sentinel-1); L-band 1.2 GHz (PALSAR-2); C-band 5.405 GHz (RADARSAT-2) |
| Coherence pair baseline | 6, 12 or 24 days (Sentinel-1); 14 days (PALSAR-2); shorter pairs reduce temporal decorrelation from vegetation |
| Archive depth | Sentinel-1: April 2014 to present (10+ years); PALSAR-2: 2014 to present (site-dependent coverage) |
| Effective soil penetration depth (C-band) | A few centimetres in moist mineral soil; increases in dry sandy soil |
| Minimum detectable moisture contrast | Volumetric moisture change of roughly 10 percentage points produces measurable coherence change; finer contrasts may be within noise |
| Cloud sensitivity | None; SAR is cloud-independent |
| Delivery format | GeoTIFF coherence rasters, classified shapefiles or GeoPackage, PDF report with methodology notes |
Analytics Satellize can run
| Seasonal coherence-change index raster | Repeat-pass interferometric coherence estimation from registered SLC stacks; wet-season minus dry-season mean coherence | GeoTIFF at 10 m pixel spacing, classified into low/medium/high preservation-probability zones |
| Multi-year coherence climatology | Stacking 10+ years of Sentinel-1 seasonal coherence pairs to derive per-pixel mean, standard deviation and anomaly years | GeoTIFF stack with summary statistics table; PDF report identifying persistently low-coherence zones |
| Preservation zone boundary polygon | Thresholding of coherence-change index, optionally constrained by ancillary drainage or soil-type layers | GeoPackage polygon layer compatible with ArcGIS and QGIS, with confidence classification attribute |
| Developer impact screening layer | Spatial intersection of preservation zone polygons with submitted development footprints | Intersection report in PDF and GIS format, flagging footprints that overlap candidate anaerobic zones |
| C-band versus L-band coherence comparison | Parallel processing of Sentinel-1 and PALSAR-2 coherence stacks over the same site; difference analysis to identify canopy-masked zones where C-band is unreliable | Side-by-side raster comparison with narrative interpretation of discrepancies |
| Annual water-table proxy time series | Per-pixel coherence time series extracted from Sentinel-1 six-day pairs, smoothed and plotted against calendar date to approximate seasonal inundation onset and recession | CSV time series per sample polygon; line-chart PDF for reporting |
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.