Flood-damaged cropland extent mapping for multi-peril crop insurance settlement
Fusing Sentinel-1 SAR backscatter with Sentinel-2 vegetation indices lets loss adjusters delineate flood-affected parcels within days of an event, even when monsoon cloud cover makes optical-only workflows impossible for weeks.
Sensors
- Sentinel-1 C-band SAR (ESA): 10 m ground resolution in Interferometric Wide Swath mode, 250 km swath, 6-day revisit at the equator with both satellites combined (12-day per satellite). Provides inundation mapping regardless of cloud or darkness. Open water returns very low backscatter (typically below -15 dB in VV polarisation); flooded vegetation shows a characteristic double-bounce enhancement that distinguishes it from bare-soil inundation.
- Sentinel-2 MSI (ESA): 10 m (visible/NIR) and 20 m (red-edge, SWIR) resolution, 5-day revisit with both satellites. NDVI time series detects canopy collapse or chlorophyll loss in the days after flood recession. NDWI (using Green and NIR bands) and MNDWI (Green and SWIR) map residual surface water and waterlogged soil. Cloud cover during monsoon seasons can render individual acquisitions unusable for two to four weeks; multi-date compositing partially compensates.
- RADARSAT Constellation Mission (CSA): Three-satellite C-band SAR constellation, 1 to 100 m resolution depending on mode, daily revisit of any point in Canada and near-daily revisit globally in compact polarimetry modes. Compact polarimetry (RCM's distinguishing feature) improves discrimination between flooded vegetation and open water relative to single-polarisation systems, relevant for rice paddies and other structured crops.
- Landsat 8/9 OLI/TIRS (USGS/NASA): 30 m multispectral resolution, 16-day revisit per satellite (8-day combined). SWIR bands (Band 6 at 1.57 µm, Band 7 at 2.11 µm) penetrate thin cloud and haze better than visible bands, aiding post-flood soil-moisture mapping. The long Landsat archive (1972 onwards) provides pre-event seasonal baselines essential for change-detection approaches.
What flooded soil and flattened crops look like to a radar
C-band SAR at 5.4 GHz interacts with agricultural surfaces in ways that are physically predictable and therefore exploitable. Open standing water acts as a near-perfect specular reflector: microwave energy bounces away from the sensor and backscatter drops sharply, typically to below -15 dB in VV polarisation. That signal is distinct from dry bare soil, which scatters diffusely and returns values in the -8 to -12 dB range depending on roughness and moisture.
Flooded vegetation complicates this picture usefully. When a crop canopy is still upright but the field is inundated beneath it, a double-bounce mechanism operates: the signal travels down through the canopy, reflects off the water surface, bounces off the vertical plant stems, and returns strongly to the sensor. This effect is well documented in rice paddies and makes flooded standing crops brighter in SAR imagery than either open water or healthy dry crops. Once the canopy collapses under the weight of water or lodges in current, the double-bounce disappears and backscatter drops. Tracking that temporal sequence across multiple Sentinel-1 passes gives a timeline of inundation onset and canopy damage severity that no optical sensor can replicate through cloud.
Why optical indices are still necessary, and when they are not enough
SAR maps water presence with confidence but does not directly measure crop stress or yield loss. NDVI, computed from Sentinel-2 red and NIR bands, tracks photosynthetic activity. A field that was inundated for four days and then drained may show no SAR anomaly by the time the next overpass arrives, but NDVI will register chlorophyll degradation for one to three weeks afterwards as root damage and anaerobic stress manifest. That temporal offset is the optical layer's main contribution to indemnity-based settlement: it extends the evidentiary window beyond the inundation event itself.
The honest constraint is cloud cover. In the Indo-Gangetic Plain during the kharif season, or across the Mekong Delta during the southwest monsoon, Sentinel-2 cloud-free acquisitions can be absent for three to six consecutive weeks. In those conditions, SAR must carry the loss-adjustment case alone. A fusion workflow that treats optical data as primary and SAR as supplementary will fail exactly when losses are largest. The correct architecture inverts that priority: SAR defines the inundation polygon; optical indices, when available, quantify stress depth and support differentiation from pre-existing drought or pest damage in adjacent parcels.
Separating flood damage from drought stress and pest damage in neighbouring parcels
Multi-peril crop insurance policies cover several distinct loss causes that can produce superficially similar spectral signatures. A field stressed by drought shows gradual NDVI decline over weeks; a flood-damaged field shows abrupt NDVI collapse following a discrete SAR-detected inundation event. Pest or disease damage typically produces spatially irregular patterns within a parcel rather than the field-boundary-conforming or topographically-driven extent that flooding generates.
The combination of event timing (SAR inundation date), spatial pattern (low-lying parcels, proximity to drainage channels, conformance with digital elevation model contours), and spectral trajectory (sudden NDVI drop post-event versus gradual pre-event decline) is usually sufficient to attribute cause. Where ambiguity remains, Sentinel-1 coherence change detection can help: flood-induced crop lodging destroys interferometric coherence between pre- and post-event SAR pairs, whereas standing drought-stressed crops maintain higher coherence. This is not a universal discriminator, but it adds a physically grounded dimension to the attribution argument.
From satellite polygons to insured-parcel loss ratios
The satellite output is a raster or vector layer of affected area. Converting that to a loss ratio for a specific policy requires intersection with the insurer's cadastral parcel database and crop-type map. Crop type matters because inundation tolerance varies substantially: rice can withstand several days of standing water at certain growth stages, while maize suffers yield loss from as little as 48 hours of root-zone saturation. Without knowing the crop and its phenological stage at flood onset, affected area alone cannot support an indemnity calculation.
Phenological stage is recoverable from the pre-event Sentinel-2 NDVI time series. A field that had reached its NDVI peak (grain fill in cereals, pod fill in legumes) before inundation faces near-total loss; a field still in vegetative growth may partially recover if drainage is rapid. Pairing the satellite-derived inundation polygon with the pre-event NDVI trajectory and a crop-specific damage function produces a per-parcel loss estimate that is defensible, auditable, and reproducible. Satellize applies this fusion approach in its crop-estimation work, including the Kingdom of Tonga programme, and the same parcel-level pipeline transfers directly to insurance loss adjustment at scale.
One practical caution: satellite-derived parcel boundaries and ground-truth cadastral boundaries rarely align perfectly. A 10 m SAR pixel straddles field bunds. Losses near parcel edges carry geometric uncertainty that should be disclosed in any settlement report, particularly for small-holder agriculture where individual parcels may be only one to three hectares.
Latency, archive depth and what adjusters should realistically expect
Sentinel-1 data are typically available in the Copernicus Data Space within 24 hours of acquisition. Processing to a calibrated, terrain-corrected backscatter product adds a few hours with automated pipelines. A first inundation map covering a flood event can therefore reach an adjuster's desk within 48 hours of the relevant overpass, which may itself arrive within six days of the event depending on orbital geometry and latitude.
Archive depth is a significant asset. Sentinel-1 data run continuously from April 2014 (Sentinel-1A) and Landsat from 1972. Pre-event baselines spanning multiple years allow statistical characterisation of normal backscatter ranges for each parcel under different seasonal conditions, which tightens the anomaly-detection threshold and reduces false positives. For a new insurance programme covering a region with no prior ground survey, the satellite archive effectively substitutes for years of field data collection.
What adjusters should not expect: SAR cannot resolve individual plant rows at 10 m resolution, so within-field variability is averaged. Cloud-free optical confirmation during peak monsoon may genuinely be unavailable for the critical window. And no satellite sensor can determine whether a crop was already compromised before the flood, unless a clean pre-event time series exists. Transparency about these limits is not a weakness in the evidence; it is what makes the evidence credible.
Typical figures
| Primary SAR spatial resolution | 10 m (Sentinel-1 IW mode); 3–100 m (RADARSAT Constellation Mission, mode-dependent) |
| Primary optical spatial resolution | 10 m visible/NIR, 20 m red-edge/SWIR (Sentinel-2); 30 m (Landsat 8/9) |
| SAR revisit (Sentinel-1 A+B combined) | 6 days at equator; shorter at higher latitudes due to orbit overlap |
| Optical revisit (Sentinel-2 A+B combined) | 5 days at equator; cloud-free acquisition frequency much lower in monsoon regions |
| Typical data latency from acquisition | 12–36 hours to Copernicus Data Space; processed inundation map deliverable within 24–48 hours of acquisition |
| Minimum detectable inundated parcel | Approximately 0.1 ha at 10 m resolution; sub-hectare parcels carry higher edge-pixel uncertainty |
| SAR frequency and polarisation | C-band 5.4 GHz; VV, VH dual-polarisation (Sentinel-1); compact polarimetry available on RADARSAT Constellation Mission |
| Archive depth | Sentinel-1 from April 2014; Landsat from 1972; enables multi-year pre-event baseline construction |
| Optical spectral indices used | NDVI (Red, NIR), NDWI (Green, NIR), MNDWI (Green, SWIR1), SWIR-based soil-moisture proxies |
| Deliverable formats | GeoTIFF raster, GeoPackage or Shapefile vector parcels, CSV loss-ratio table, PDF adjuster report |
Analytics Satellize can run
| SAR inundation extent map | Thresholded VV/VH backscatter change detection against multi-date pre-event baseline; Otsu or histogram-based threshold selection | GeoTIFF and polygon layer of inundated area with date-stamped acquisition metadata |
| Flooded-vegetation discrimination layer | Dual-polarisation ratio (VV/VH) and temporal backscatter trajectory analysis to separate open water from double-bounce flooded canopy | Classified raster distinguishing open inundation, flooded standing crop, and post-flood lodged crop |
| NDVI anomaly time series per insured parcel | Sentinel-2 NDVI computed at each cloud-free acquisition; z-score anomaly against same-season multi-year baseline; parcel-level aggregation | Per-parcel NDVI trajectory chart and tabular anomaly scores for adjuster review |
| Flood-versus-drought attribution score | Event-timing cross-reference (SAR inundation date versus NDVI decline onset), spatial pattern analysis, DEM-based topographic conformance check, optional SAR coherence change layer | Per-parcel cause-attribution confidence rating included in settlement report |
| Phenological stage at flood onset | Pre-event NDVI trajectory fitted to crop-specific phenological calendar; growth stage estimated from NDVI value and rate-of-change at inundation date | Tabular output of estimated growth stage per parcel, used as input to crop-specific damage functions |
| Per-parcel loss ratio estimate | Inundation duration (from multi-pass SAR) combined with phenological stage and crop-specific flood-tolerance damage function; intersected with insurer cadastral database | CSV and PDF loss-ratio table per insured parcel, with uncertainty bounds flagged for edge-pixel and cloud-gap cases |
| Rapid first-alert inundation bulletin | Automated SAR backscatter threshold applied within 24 hours of Sentinel-1 acquisition; no optical dependency | Email alert with preliminary affected-area estimate and map, issued before optical confirmation is available |
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.