Storm surge and coastal inundation extent from SAR
When a tropical cyclone makes landfall, optical sensors go blind and the surge boundary moves by the hour. X-band SAR from ICEYE and Capella Space cuts through cloud to deliver sub-metre flood extent within hours, though mangrove and salt-marsh returns demand careful interpretation.
Sensors
- ICEYE X-band SAR constellation: Spotlight mode delivers approximately 0.5 m ground resolution; strip-map around 3 m. On-demand tasking with revisit to any point achievable within 1–3 hours under the full constellation, making it the primary choice for tracking a surge boundary that moves on tidal timescales. Operates at 9.65 GHz, penetrating cloud and rain with negligible attenuation at X-band rain rates below roughly 20 mm/hr.
- Capella Space X-band SAR: Spotlight mode to approximately 0.5 m, sliding-spotlight to 1 m, strip-map to 5 m. Similar on-demand tasking model to ICEYE. Useful for high-resolution verification of the inundation edge in built environments and for detecting individual structures above the waterline.
- Sentinel-1 IW (C-band, ESA): 10 m resolution in Interferometric Wide Swath mode, 250 km swath, six-day repeat at the equator (three days with both satellites). Free and open archive back to 2014. The six-day cadence is too slow to track a fast-moving surge front, but the wide swath and archive depth make it the reference layer for pre-event baseline and post-event extent once conditions stabilise.
- RADARSAT Constellation Mission (C-band, CSA): Three-satellite constellation with a four-day exact repeat and sub-daily revisit at high latitudes. Medium-resolution mode at 16 m, compact polarimetry available. Provides a middle ground between Sentinel-1 archive depth and commercial on-demand responsiveness, and is routinely activated under the International Charter on Space and Major Disasters.
Why the surge boundary is the hardest flood edge to map
River floods expand over hours to days. A storm surge arrives and retreats within a single tidal cycle, sometimes in under twelve hours. The inundation edge at peak surge can be kilometres inland of the edge recorded six hours later. Any sensor with a revisit measured in days is not mapping the event; it is mapping the aftermath, which is a different and less operationally useful thing.
Optical sensors compound the problem. Tropical cyclones carry dense convective cloud that persists for 24 to 72 hours after landfall. MODIS, Sentinel-2 and commercial very-high-resolution optical systems are effectively unavailable during the period of maximum interest. SAR is the only orbital sensor class that works through this cloud reliably at the frequencies used by current civilian constellations.
Open water reads cleanly. Mangroves do not.
Over bare agricultural land or urban streets, SAR flood mapping is straightforward: open water returns very low backscatter because the specular surface deflects the radar pulse away from the sensor, producing a dark signature easily distinguished from surrounding land. Detection of inundation under open canopy is also well-established, where the double-bounce between the water surface and vertical tree trunks actually increases backscatter relative to dry conditions.
Mangrove and salt-marsh environments break both of those rules simultaneously. The pneumatophores, prop roots and dense low canopy of a mangrove forest scatter the radar signal in multiple directions at once. Flooded mangrove may appear brighter than dry mangrove in some geometries and darker in others, depending on incidence angle, polarisation and canopy density. Salt marsh introduces similar ambiguity: the standing water is partially obscured by grass stems, producing intermediate backscatter values that sit between the clear-water and dry-land signatures. Analysts working these environments must treat the inundation boundary as a probability zone rather than a sharp line, and cross-validate against tide-gauge records and any available pre-event vegetation maps.
X-band is more sensitive to fine surface structure than C-band, which helps resolve individual features but also means the backscatter is more sensitive to wind-roughened water surfaces. A surge arriving under high winds can produce backscatter from open water that mimics dry land. Acquisition timing relative to the peak wind speed matters.
On-demand versus systematic: the revisit argument in numbers
Sentinel-1's six-day repeat is not a design flaw; it is the cost of a free, global, 250 km swath archive. For mapping the stable post-event extent of a slow-moving river flood, it is entirely adequate. For a surge that peaks and retreats in under twelve hours, six days is irrelevant.
ICEYE's published constellation capability allows tasking of any point on Earth within a window that, under current fleet size, can be as short as one to three hours. Capella Space operates on a comparable model. The practical consequence is that an emergency operations centre can request an acquisition timed to coincide with predicted surge peak, receive processed imagery within roughly an hour of acquisition, and have a georeferenced flood-extent layer before the water has fully receded. That is a qualitatively different operational capability, not merely a faster version of the same thing.
The trade-off is cost and swath width. ICEYE strip-map covers roughly 30 km by 50 km per scene in standard mode; spotlight is smaller still. A major cyclone affecting 200 km of coastline requires multiple tasked acquisitions, which multiplies cost and coordination overhead. Sentinel-1 covers the same coastline in a single pass. The practical answer for most activations is a hybrid: Sentinel-1 or RCM for the wide-area baseline, commercial X-band for the critical sub-regions where the surge boundary is most consequential.
Tidal correction: why the gauge matters as much as the satellite
A SAR image records the water surface at a single moment. Without knowing the tidal state at that moment, the inundation extent cannot be separated into the meteorological surge component and the normal tidal inundation. A coastline that floods at mean high water every day is not a disaster; the same coastline flooded two metres above mean high water is.
The correction is conceptually simple: subtract the predicted astronomical tide at acquisition time from the observed water level at the nearest co-located tide gauge, and the residual is the surge. In practice, tide-gauge networks are sparse along many of the coastlines most vulnerable to cyclone surge, particularly in the Pacific and parts of the Bay of Bengal. Where gauges exist, their records are often transmitted with latency that exceeds the operational window. Satellite altimetry from missions such as Sentinel-6 Michael Freilich can supplement gauge data along open coasts, but the along-track sampling is coarse relative to the spatial scale of surge variability in embayments and estuaries.
The honest position is that tidal correction improves the scientific quality of the inundation map but introduces its own uncertainty when gauge coverage is poor. Analysts should report the tidal state at acquisition time alongside the flood-extent product, and flag areas where the correction is based on interpolated rather than measured water levels.
What the data pipeline looks like under time pressure
Emergency SAR activations under the International Charter on Space and Major Disasters can deliver processed imagery within hours of a tasking request, depending on the authorised user and the Charter project officer's response time. Outside Charter activations, direct commercial tasking with ICEYE or Capella Space follows a similar timeline if contracts are pre-positioned before the event season.
Processing to a flood-extent layer involves orthorectification against a digital elevation model, speckle filtering (Lee or refined Lee filters are standard), thresholding or change-detection against a pre-event reference scene, and vectorisation of the inundation boundary. The DEM choice matters: SRTM at 30 m is freely available but was acquired at a single epoch and does not capture post-2000 coastal change. TanDEM-X at 12 m is more current and more accurate in low-relief coastal terrain, though it carries a licensing cost.
Satellize integrates commercial SAR tasking into its analytics workflows on client licence, alongside open-constellation baselines. The Overhead column has covered surge-mapping methodology in the context of recent Pacific cyclone seasons.
Typical figures
| Spatial resolution (ICEYE/Capella spotlight) | Approximately 0.5 m |
| Spatial resolution (Sentinel-1 IW) | 10 m (range) × 10 m (azimuth) after multi-looking |
| Revisit (ICEYE/Capella on-demand) | 1–3 hours to any point under current fleet size |
| Revisit (Sentinel-1 IW, equatorial) | 6 days single satellite; approximately 3 days with both Sentinel-1A and 1C |
| Radar frequency | X-band: 9.65 GHz (ICEYE, Capella); C-band: 5.405 GHz (Sentinel-1, RCM) |
| Swath width | ICEYE strip-map ~30 km; Sentinel-1 IW 250 km; RCM medium-resolution ~125 km |
| Minimum detectable inundation patch | Approximately 0.1 ha in open terrain at 3 m resolution; larger in vegetated environments |
| Latency from acquisition to processed layer | Typically 1–4 hours for commercial X-band with pre-positioned processing pipeline |
| Archive depth (Sentinel-1) | 2014 to present, globally accessible via Copernicus Data Space |
| Delivery formats | GeoTIFF (backscatter), GeoJSON/Shapefile (flood extent polygon), COG for web streaming |
Analytics Satellize can run
| Peak surge inundation extent polygon | Change-detection thresholding against pre-event Sentinel-1 or commercial SAR baseline; morphological filtering to remove speckle artefacts | GeoJSON polygon layer with acquisition timestamp and tidal-state metadata, delivered within 2–4 hours of SAR scene receipt |
| Surge depth estimate over open terrain | Intersection of inundation boundary with bare-earth DEM (TanDEM-X or SRTM); uncertainty band reported where DEM vertical accuracy exceeds 1 m | Raster depth grid (GeoTIFF) with per-pixel confidence class; summary statistics by administrative unit |
| Inundation time series across the event window | Multi-temporal stack of all available SAR acquisitions; per-pixel water-probability score using empirical backscatter distributions | Animated GIF and GeoTIFF stack showing flood extent at each acquisition epoch; suitable for situation reports |
| Mangrove and salt-marsh ambiguity flag | Overlay of inundation boundary against Global Mangrove Watch or equivalent vegetation map; flagged pixels reported separately with reduced confidence class | Annotated polygon layer distinguishing high-confidence open-water inundation from ambiguous vegetated-coast pixels |
| Tidal-corrected surge residual map | Subtraction of predicted astronomical tide (from co-located gauge or FES2014 tidal model) at acquisition epoch from inundation boundary elevation | Surge-residual raster with gauge-coverage quality flag; tabular summary of peak surge height by coastal segment |
| Population and infrastructure exposure count | Spatial intersection of inundation extent with WorldPop gridded population and OpenStreetMap road/building layers | Tabular exposure report by district: estimated affected population, road-km inundated, building count within flood boundary |
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.