Post-flood crop and infrastructure damage assessment
SAR coherence loss and optical NDVI change detection together identify damaged cropland, collapsed structures and debris-covered roads after a flood. Pre-event baseline archives are the critical dependency that determines whether the analysis is possible at all.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range detected resolution in Interferometric Wide Swath mode; 6-day repeat at the equator with both satellites active, 12-day with one. Provides backscatter change and interferometric coherence, the latter being particularly sensitive to structural damage and surface disruption.
- Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands; 20 m red-edge and shortwave infrared. Five-day revisit with both satellites. Delivers pre- and post-event NDVI, NDWI and false-colour composites for crop condition and inundation extent, but is blocked by cloud cover, which is common in post-flood conditions.
- Planet SuperDove: 3 m resolution, 8-band PlanetScope constellation of roughly 200 satellites. Near-daily revisit globally. Useful for field-level crop damage classification where cloud windows exist, and for detecting debris patterns on roads and in built areas at sub-parcel scale.
- COSMO-SkyMed (ASI): X-band SAR with selectable modes from 3 m (Stripmap) to 1 m (Spotlight). Shorter wavelength than Sentinel-1 makes it more sensitive to small structural features and surface roughness changes on roads and rooftops. Revisit is constellation-dependent and typically requires tasking; not freely available.
What the physics gives you after a flood
Two independent physical signals become readable once floodwaters drop. The first is SAR backscatter change: a standing crop or intact building returns radar energy in a predictable pattern; flattened crop, collapsed masonry and debris scatter it differently, producing a measurable shift in backscatter intensity between a pre-event image and a post-event one. The second signal is interferometric coherence. Sentinel-1 coherence between two acquisitions six days apart is high over stable surfaces and collapses toward zero wherever the surface has been disturbed. Structural damage, soil displacement and debris deposition all destroy coherence. Studies using Sentinel-1 coherence over earthquake and flood events have consistently shown that coherence loss below roughly 0.3 in built-up areas correlates with significant structural damage, though the threshold is scene-dependent.
Optical sensors add what SAR cannot easily distinguish: crop physiology. NDVI computed from Sentinel-2 red and near-infrared bands separates healthy, stressed and dead vegetation at 10 m. A field that appears geometrically intact in SAR but shows NDVI collapse two weeks after flooding has experienced root-zone damage or sediment burial rather than physical destruction. The two data streams answer different questions and are most useful together.
The baseline archive is not optional
Post-event analysis depends entirely on what was collected before the event. Without a pre-flood SAR acquisition from the same orbital geometry, coherence cannot be computed. Without a cloud-free optical image from the same crop growth stage in a prior season, NDVI change is ambiguous: a field at harvest looks spectrally similar to a flood-damaged one. This is the central operational constraint that most buyers underestimate.
The Copernicus programme archives Sentinel-1 and Sentinel-2 data continuously and globally. For most locations in Europe, South and Southeast Asia, and parts of Africa, there are multi-year SAR stacks available through the Copernicus Data Space Ecosystem. For areas outside routine acquisition zones, the archive may have gaps of weeks or months. Knowing your coverage status before an event is the difference between a 48-hour damage assessment and a two-week scramble for substitute data.
Commercial X-band tasking from COSMO-SkyMed or similar systems can fill resolution gaps, but only if tasked promptly after the event. The window before agricultural fields are replanted or roads cleared is often shorter than the procurement cycle for unplanned commercial imagery.
How the Copernicus Emergency Management Service activates
The Copernicus Emergency Management Service (CEMS) can be activated by eligible national authorities, international organisations and EU institutions. Once activated, CEMS produces grading maps, delineation maps and damage assessment products, typically within hours to a few days depending on data availability and event complexity. The products are delivered as vector and raster GIS files and are publicly released after a short embargo period.
CEMS uses a defined workflow: a reference product is created from pre-event data, then a monitoring or grading product is created from post-event acquisitions. Damage grades follow a standardised classification from no damage through to destroyed. The service primarily uses Sentinel-1 and Sentinel-2, supplemented by very-high-resolution commercial imagery where available. Response times are constrained by satellite overpass scheduling and, for optical data, cloud cover. A CEMS activation is not a substitute for a client-specific analysis: the service covers a defined area of interest and uses standard product specifications that may not match an insurer's or government ministry's exact requirements.
Crop damage: what the signal can and cannot resolve
Field-level crop damage mapping is feasible at 10 m with Sentinel-2 NDVI differencing and at 3 m with Planet SuperDove, provided cloud-free acquisitions exist within roughly two to three weeks of the event. Beyond that window, crop recovery or replanting begins to obscure the damage signal. Sediment-smothered fields are detectable by their characteristic low NDVI combined with elevated SWIR reflectance. Waterlogged but physically intact crops show a more ambiguous signature: NDVI may recover partially within days, making the damage appear less severe than it is.
Minimum detectable damage area is approximately one to two Sentinel-2 pixels, so roughly 100 to 400 square metres. Below that threshold, field-level losses are invisible without commercial tasking. For smallholder agriculture, where individual plots may be under 0.1 hectares, this is a real limitation. Planet SuperDove at 3 m reduces that floor substantially, but the constellation's radiometric calibration for NDVI is less mature than Sentinel-2's. Honest damage statistics should carry an uncertainty band that reflects the resolution floor and the cloud-cover fraction of post-event acquisitions.
Infrastructure damage: roads, bridges and structures
Debris on roads shows up in two ways: as a backscatter anomaly in SAR, where the rough, heterogeneous debris surface returns more energy than a smooth road surface, and as a spectral anomaly in optical imagery, where pale silt or dark debris contrasts with the road's pre-event colour. The combination is more reliable than either alone. Sentinel-1 at 10 m can detect road blockages on major routes; detecting a single landslide deposit on a minor road requires 1 to 3 m resolution from COSMO-SkyMed or a commercial optical constellation.
Bridge damage is harder. A bridge that has partially collapsed but retains its deck may show no coherence loss if the surface geometry is largely unchanged. Only a very-high-resolution optical or SAR acquisition, or a repeat-pass InSAR deformation product, will reveal structural compromise that does not manifest as surface change. Buyers assessing critical infrastructure should treat SAR change detection as a triage tool, not a structural survey.
Turning the analysis into a usable product
A damage assessment that arrives three weeks after a flood is largely useless for emergency response and only marginally useful for insurance adjustment. The operational target for a first-pass product is 24 to 72 hours after the first post-event SAR acquisition. That requires pre-staged processing pipelines, pre-loaded baseline archives and automated change detection that flags anomalies for human review rather than waiting for manual interpretation.
Satellize runs change-detection workflows on open Sentinel data and adds commercial tasking on client licence. The approach is similar to what underpins the Tonga crop-estimation programme: automated index computation against a maintained baseline, with outputs delivered as GIS layers and structured reports rather than raw imagery. For flood damage, the critical pre-work is agreeing the area of interest and confirming archive coverage before the flood season begins, not after the levee breaks.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m ground range detected; 20 m azimuth (multi-looked) |
| SAR spatial resolution (COSMO-SkyMed Spotlight) | 1 m (single-look); tasked acquisition required |
| Optical spatial resolution | 10 m (Sentinel-2 visible/NIR); 3 m (Planet SuperDove) |
| Sentinel-1 revisit (both satellites) | 6 days at equator; shorter at higher latitudes |
| Sentinel-2 revisit (both satellites) | 5 days globally; cloud cover may extend effective revisit to weeks |
| Coherence sensitivity threshold (indicative) | Coherence below ~0.3 in built-up areas associated with significant damage; scene-dependent |
| Minimum detectable crop damage area | ~100–400 m² (Sentinel-2); ~9–25 m² (Planet SuperDove) |
| Sentinel archive depth | Sentinel-1 from 2014; Sentinel-2 from 2015; coverage varies by acquisition zone |
| Typical first-pass product latency | 24–72 hours after first post-event SAR pass, given pre-staged baseline |
| Delivery formats | GeoTIFF damage rasters, GeoJSON/Shapefile vector grading layers, PDF summary report |
Analytics Satellize can run
| SAR backscatter change map | Log-ratio of pre- and post-event Sentinel-1 sigma-nought in VV and VH polarisations; thresholded by Otsu or scene-specific method | GeoTIFF raster with change magnitude; vector polygons of significant change areas |
| Interferometric coherence loss map | Complex coherence computed between pre- and post-event Sentinel-1 SLC pairs; coherence drop below threshold flagged as damage indicator | GeoTIFF coherence layer; built-up area damage grade vector file |
| Crop damage classification | NDVI differencing between post-event and seasonal baseline (same phenological stage, prior year); classes: healthy, stressed, severely damaged, destroyed | Field-level GIS layer with damage class and area statistics per administrative unit |
| Road accessibility assessment | SAR backscatter anomaly detection on road network buffer; optical false-colour review of debris signature; cross-referenced against OpenStreetMap road network | Road segment status shapefile (passable/blocked/unknown) for logistics planning |
| Multi-sensor damage grade composite | Fusion of SAR coherence loss, backscatter change and optical NDVI change using decision-tree or weighted overlay; grading follows CEMS-compatible classification schema | Standardised damage grade GIS layer; tabular loss estimates by crop type and administrative boundary |
| Baseline archive audit | Automated query of Copernicus Data Space Ecosystem and commercial archive holdings for area of interest; gap identification by season and orbital track | Coverage report identifying archive gaps and recommended pre-season tasking schedule |
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.