Tsunami inundation extent and coastal infrastructure damage mapping
Comparing pre- and post-event optical and SAR imagery reveals how far a tsunami ran inland and which coastal structures survived. The method rests on sediment deposition, vegetation removal and coherence loss, but the usable window is narrow.
Sensors
- WorldView-3: 0.31 m panchromatic, 1.24 m multispectral. Resolves individual building footprints, collapsed roof sections and sediment sheets on roads. Single-pass revisit over a given coastal strip is typically 1 to 4.5 days depending on latitude and tasking priority.
- Pleiades Neo: 0.30 m panchromatic, 1.2 m multispectral, with a stereo and tri-stereo capability that can generate a post-event DSM for run-up elevation estimation. Constellation of four satellites gives sub-daily revisit at mid-latitudes under clear skies.
- Sentinel-1 SAR (C-band, 5.6 cm): 6 m range resolution in IW mode, 12-day repeat (6-day with both satellites). Penetrates cloud and works at night, making it the primary workhorse when optical acquisition is blocked. Backscatter change and coherence loss both serve as damage proxies, though C-band is partially attenuated by standing water and wet sediment.
- ALOS-2 PALSAR-2 (L-band, 24 cm): L-band penetrates vegetation canopy and wet debris more effectively than C-band. Spotlight mode reaches 1 to 3 m resolution. JAXA routinely activates ALOS-2 under the International Charter, as it did after the 2018 Sulawesi and 2022 Tonga events. Revisit is 14 days in standard mode, shorter under emergency tasking.
What a receding wave leaves behind that satellites can read
Tsunami inundation produces three spectral and structural signatures that distinguish it from ordinary coastal flooding. First, sediment deposition: sand and silt carried inland by the wave and deposited as the water recedes creates a bright, spectrally flat layer visible in multispectral imagery, particularly in the red and near-infrared bands where healthy vegetation has high reflectance and bare sand does not. Second, vegetation removal: saltwater and physical scour strip ground cover abruptly, producing a sharp NDVI discontinuity at the run-up limit that corresponds closely to the maximum inundation line documented in field surveys after the 2004 Indian Ocean and 2011 Tōhoku events. Third, structural collapse: reinforced concrete and timber-frame buildings respond differently to hydrodynamic loading, and the debris signature in very-high-resolution imagery, combined with SAR coherence loss between pre- and post-event acquisitions, allows damage to be graded from intact through partially damaged to destroyed.
The physical boundary between inundated and non-inundated land is often sharper in satellite imagery than in flood mapping, because tsunami run-up is fast and the sediment line is deposited at the hydraulic limit rather than seeping gradually. That sharpness is an advantage for delineating the inundation polygon, but it also means the evidence is fragile: rainfall, wind and human clearance operations begin erasing the sediment signature within days of the event.
The time window is the problem
Field teams and satellite analysts are in a race with the same adversary: time. The sediment sheet that marks the inundation boundary can be washed away by the next rain event or bulldozed by recovery machinery within 48 to 96 hours of the wave. The 2011 Tōhoku response demonstrated that WorldView and GeoEye acquisitions taken within 72 hours of the event produced inundation polygons that matched field-surveyed run-up lines to within acceptable margins for emergency planning. Acquisitions taken a week later were already compromised by debris removal.
Cloud cover compounds the pressure. Tsunamis generated by subduction-zone earthquakes frequently occur in regions with persistent convective cloud, and a single cloudy overpass wastes an irreplaceable day. This is where SAR earns its place: Sentinel-1 and ALOS-2 can acquire through cloud, and their backscatter change products, while coarser than optical damage maps, provide a usable inundation boundary within hours of data receipt. The practical workflow pairs an early SAR acquisition for extent with a subsequent optical acquisition for damage grading, accepting that the two products will have different spatial precision.
Separating tsunami damage from what was already broken
Coastal infrastructure in tsunami-prone regions is rarely pristine before the event. Pre-existing storm damage, subsidence, informal construction and previous flood scarring all create false positives in change-detection analysis. The only reliable mitigation is a high-quality pre-event archive: ideally a very-high-resolution optical image acquired within the previous six to eighteen months, taken in the same season to avoid phenological NDVI differences that mimic vegetation removal.
Even with a good archive, ambiguity persists in two specific situations. Structures that were already partially collapsed before the event will show coherence loss that predates the tsunami. And in regions where cyclone damage preceded the tsunami by weeks or months, the sediment signature may be partially masked by storm debris. Analysts should flag these zones explicitly in delivered products rather than forcing a binary damaged or undamaged classification. Honest uncertainty polygons are more useful to a humanitarian coordinator than a clean map that is wrong.
How the International Charter shapes what imagery is available
The International Charter on Space and Major Disasters is the primary mechanism by which very-high-resolution commercial imagery becomes available to national civil protection agencies in the immediate aftermath of a major tsunami. Activation authorises tasking of commercial satellites, including Pleiades and WorldView, at no direct cost to the requesting government, with products routed through designated Project Managers to operational users. The 2004 Indian Ocean activation was one of the Charter's earliest large-scale tests; the 2011 Tōhoku activation produced more than 200 value-added products within the first two weeks.
Charter activations do not guarantee cloud-free imagery, and they do not override the physics of orbital revisit. A government that has pre-negotiated tasking agreements and archive access through its own programme is better positioned to receive imagery in the first 24 hours than one that depends entirely on Charter allocation. Sovereign space programme design increasingly accounts for this: building national archive depth and emergency tasking priority into standing contracts rather than treating them as post-disaster procurement problems.
From pixels to products: what the analysis actually delivers
The primary deliverable is a georeferenced inundation polygon with a confidence classification tied to the data source. SAR-derived polygons carry wider uncertainty bands than optical-derived ones; both should be delivered with metadata that states the acquisition time relative to the event, the pre-event reference date and the method used. A secondary deliverable is a per-structure damage grade for the built environment within the inundation zone, typically expressed as a four-class scheme (no visible damage, minor, major, destroyed) applied to building footprints from a pre-event cadastral layer or from the satellite imagery itself.
Run-up elevation estimation is possible where Pleiades Neo stereo or an existing LiDAR DEM is available. The inundation boundary polygon is intersected with the DEM to extract a maximum water-surface elevation contour, which serves as input to hydrodynamic model calibration. This is a post-processing step, not a direct satellite measurement, and its accuracy depends entirely on the quality of the underlying elevation data. In low-lying Pacific island contexts, where the DEM may be derived from SRTM at 30 m resolution, the elevation uncertainty can exceed the run-up height itself.
Satellize structures this analysis as a staged delivery: a SAR-based extent layer within six hours of data receipt, followed by an optical damage-grade layer as soon as cloud-free imagery is confirmed, and a final consolidated report integrating both. The Tonga crop-estimation programme gave us direct experience with the island-chain data environment that shapes Pacific tsunami response, including the archive gaps and connectivity constraints that affect product delivery.
Honest limits of the method
Resolution sets a hard floor on what can be detected. At Sentinel-1's 6 m IW resolution, individual small structures are below the detection threshold; damage mapping at that scale is statistical rather than per-building. WorldView-3 at 0.31 m resolves individual roof panels, but its swath width is 13.1 km, meaning that a 50 km stretch of coastline requires multiple passes or a mosaic from different acquisition times, introducing temporal inconsistency.
SAR coherence loss is a sensitive but non-specific damage indicator. Wet sediment, standing water and even dense debris fields all reduce coherence without necessarily indicating structural collapse. C-band coherence loss in particular should be interpreted alongside optical imagery rather than as a standalone damage metric. Finally, the method says nothing about casualties, displaced population or the condition of infrastructure that is not visible from above: underground utilities, bridge foundations and subsurface drainage are outside the scope of any current satellite-based analysis.
Typical figures
| Best optical resolution (WorldView-3 / Pleiades Neo) | 0.31 m panchromatic; 1.2–1.24 m multispectral |
| SAR resolution (Sentinel-1 IW / ALOS-2 Spotlight) | ~6 m / 1–3 m |
| Revisit under emergency tasking (optical) | Sub-daily to 4.5 days depending on constellation and latitude |
| Sentinel-1 standard revisit | 6 days (both satellites); 12 days (single satellite) |
| Latency from acquisition to delivered extent layer | 2–6 hours for SAR; 4–12 hours for optical, depending on processing pipeline |
| Key spectral bands for inundation mapping | Red, NIR (NDVI change); SWIR (moisture/sediment); C-band and L-band SAR backscatter |
| Minimum detectable inundation feature (optical) | ~1–2 m width at 0.31 m resolution; ~10–15 m at 6 m SAR resolution |
| Pre-event archive depth (Sentinel-1) | From 2014; freely accessible via Copernicus Data Space |
| Damage classification scheme | Four-class: no visible damage / minor / major / destroyed (per building footprint) |
| Primary delivery formats | GeoTIFF, GeoPackage, Shapefile, KMZ; optional WMS/WMTS feed |
Analytics Satellize can run
| SAR-derived inundation extent polygon | Backscatter change detection (pre/post amplitude ratio) and coherence differencing on Sentinel-1 IW or ALOS-2 ScanSAR pairs | GeoPackage polygon layer with confidence class; delivered within 6 hours of SAR data receipt |
| Optical inundation boundary with sediment and vegetation indicators | NDVI change map and spectral angle mapping on pre/post multispectral pairs (WorldView-3 or Pleiades Neo) | GeoTIFF change layer and vector inundation polygon; confidence rated by cloud-free coverage percentage |
| Per-building damage grade | Object-based image analysis on very-high-resolution optical imagery, with SAR coherence loss as secondary indicator; graded to four-class scheme | Shapefile with damage attribute per building footprint; tabular summary by administrative unit |
| Run-up elevation profile | Inundation polygon intersected with available DEM (SRTM 30 m, Copernicus DEM 10 m or client-supplied LiDAR); uncertainty bounds stated explicitly | Elevation contour GeoJSON with DEM source and uncertainty range noted in metadata |
| Coastal infrastructure damage inventory | Manual and semi-automated feature extraction from 0.30–0.31 m optical imagery; comparison against pre-event cadastral or OpenStreetMap baseline | Structured damage report with geolocated records for roads, jetties, sea walls and critical buildings; PDF and GIS layer |
| Time-series evidence package for Charter or insurance use | Curated pre-event archive stack (Sentinel-1, Sentinel-2, commercial optical) with post-event acquisitions; provenance metadata preserved | Zipped archive of georeferenced imagery and derived layers with acquisition timestamps; suitable for legal or insurance submission |
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.