Dam and reservoir integrity monitoring after seismic or flood events

SAR coherence loss, InSAR displacement, and optical change detection give emergency managers the first quantitative read on dam integrity within hours of an earthquake or extreme flood, before ground teams can safely approach.

Sensors

Sentinel-1 SAR (C-band, ESA): Interferometric Wide Swath mode at 5x20 m ground range resolution; 6-day repeat at mid-latitudes, 12-day globally. Coherence change detection and InSAR displacement at millimetre-to-centimetre scale. Free archive from 2014. Cloud-independent.
ICEYE SAR (X-band, commercial): Spotlight mode delivers approximately 0.5 m resolution single-look imagery; strip mode revisit can be tasked within hours of an event trigger. X-band penetrates light vegetation better than optical but is sensitive to surface moisture changes that can mask displacement signals.
Capella Space SAR (X-band, commercial): Spotlight products at approximately 0.5 m resolution with flexible tasking geometry. Useful for high-resolution breach characterisation and crest-line change detection where ICEYE coverage is unavailable or redundancy is required.
Sentinel-2 MSI (multispectral, ESA): 10 m visible and near-infrared bands, 20 m shortwave infrared. 5-day revisit at equator with both satellites. Detects turbid water plumes, spillway erosion scars, and downstream inundation extent in cloud-free conditions. Free and open archive from 2015.

What a dam failure looks like from orbit before anyone sees it from the ground

Earthfill and rockfill embankments fail in recognisable physical sequences: internal erosion (piping), slope instability, overtopping, or foundation liquefaction under seismic loading. Each stage leaves a different signal in satellite data. Piping and early slope movement produce millimetre-scale surface displacement detectable by InSAR days before visible cracking. Overtopping and breach produce coherence loss across the downstream face and a turbid sediment plume in optical bands. Foundation liquefaction after an earthquake can produce subsidence of the dam crest measurable as a centimetre-scale fringe in a single interferogram.

The critical window is the first 24 to 72 hours after the triggering event. Ground inspection is often impossible: roads are cut, the dam is in a restricted zone, or aftershock risk keeps teams away. Satellite data fills that gap not as a substitute for engineering assessment but as a triage tool that tells responders which of twenty dams in an affected region needs a helicopter first.

SAR coherence loss: the slope disruption signal

Synthetic aperture radar coherence measures how similar the radar backscatter from a patch of ground is between two passes. A stable surface, compacted embankment fill, concrete apron, or dry riprap, maintains high coherence. Surface disruption, whether from slumping, cracking, saturation, or debris deposition, decorrelates the signal. The coherence difference between a pre-event and post-event Sentinel-1 pair, computed over the embankment slope, is one of the fastest indicators of surface disturbance available.

The method has real limits. Vegetation on embankment slopes decorrelates naturally between passes regardless of any structural event, producing false positives in forested or heavily vegetated catchments. Temporal decorrelation also increases with the interval between acquisitions: a 12-day baseline in a tropical environment may show coherence loss that has nothing to do with the dam. Analysts must use pre-event coherence time series to establish a baseline decorrelation rate before interpreting post-event drops as structurally significant. This is not optional; it is the difference between a useful alert and a false alarm that wastes emergency resources.

InSAR displacement: measuring what the embankment body is doing

Differential InSAR compares the phase of two SAR acquisitions to detect line-of-sight displacement at sub-centimetre precision over stable, coherent surfaces. Applied to a concrete-faced or compacted earthfill dam, it can detect crest subsidence, upstream face heave from pore-pressure increase, or differential settlement across the abutments. The Mosul Dam in Iraq has been monitored by InSAR time series in published studies using Sentinel-1 data, showing measurable ongoing deformation that informed maintenance decisions. That is a documented, peer-reviewed application of the technique.

Post-earthquake InSAR is more demanding. Seismic shaking itself decorrelates the SAR signal over disturbed terrain, so the interferogram may be uninterpretable immediately after a large event. The practical approach is to use the first coherent post-event acquisition against a pre-event image, accepting that the displacement measurement integrates over the entire period including the event itself. For embankments that survive the initial shaking, subsequent passes can track ongoing creep or consolidation settlement. Sentinel-1's six-day revisit at mid-latitudes makes this feasible within a single week.

Optical change detection: breach, spillway erosion, and the downstream plume

When a dam overtops or breaches, the optical signature is unambiguous in cloud-free conditions: a gap in the crest line, erosion scarring on the downstream face, and a turbid plume of suspended sediment extending kilometres into the receiving river or valley. Sentinel-2's 10 m bands resolve crest-line changes of tens of metres. The shortwave infrared band (Band 11, 1610 nm) is particularly useful for distinguishing sediment-laden water from clear water and for mapping wet versus dry embankment surfaces.

Cloud is the hard constraint. Extreme rainfall events, which cause the majority of dam overtopping incidents globally, are almost always accompanied by persistent cloud cover that blocks optical sensors for days. This is precisely why SAR is the primary sensor for emergency response: it sees through cloud at C-band and X-band. Optical imagery is most useful in the days following an event when cloud breaks, for damage characterisation, downstream inundation mapping, and communicating the situation to decision-makers who are less comfortable interpreting SAR products.

Honest limits and the forested catchment problem

Dense forest over an embankment or its abutments is the single biggest obstacle to SAR-based dam monitoring. Forest canopy scatters C-band radar and decorrelates rapidly between passes, masking the embankment signal entirely. X-band from ICEYE or Capella is worse in this respect, not better: shorter wavelength means even less canopy penetration. L-band SAR (ALOS-2 PALSAR-2, or the forthcoming NISAR mission) penetrates forest more effectively, but L-band commercial tasking for emergency response is not as readily available as C- or X-band.

Resolution also matters for small structures. A tailings dam or a small irrigation bund with a 10 m crest width is at the limit of Sentinel-1 resolution and essentially invisible to InSAR coherence analysis. High-resolution commercial SAR is required, and even then the coherence analysis needs careful handling because the embankment may occupy only a handful of pixels. Satellite monitoring is most reliable for large embankment dams with crest lengths of hundreds of metres or more, clear zones around the dam body, and a history of SAR acquisitions that allows a pre-event baseline to be established.

Satellize structures rapid-response analytics for dam safety using Sentinel-1 open data as the baseline layer, with commercial SAR tasking added where resolution or revisit requirements exceed what the open constellation provides. The approach draws on the same InSAR and coherence-change methods used in Satellize's ongoing work with Pacific island governments, including the Tonga crop-estimation programme, where understanding how land surface change signals behave in data-sparse environments is a practical daily concern.

From acquisition to engineering decision: what the workflow actually produces

The output of a post-event dam monitoring analysis is not a map. It is a structured assessment: which structures in the affected region show anomalous coherence loss, which show measurable displacement, and what is the confidence level given the pre-event baseline. That assessment feeds directly into the decision about which sites to inspect physically and in what order.

Latency is the governing variable for emergency use. Sentinel-1 data is typically available on the Copernicus Data Space within a few hours of acquisition. Processing an interferogram and coherence map adds one to three hours of analyst time. A first-pass assessment of a region with ten to twenty dams can realistically be delivered within six to twelve hours of a Sentinel-1 overpass, assuming pre-event imagery and a prepared processing baseline exist. If they do not, add several hours for baseline construction. This is why pre-event preparation, establishing the archive, the processing pipeline, and the communication chain before any disaster occurs, is not a procedural nicety. It is what determines whether the analysis arrives in time to be useful.

Typical figures

SAR spatial resolution (Sentinel-1 IW mode)	5 x 20 m (ground range x azimuth); 10 m after multi-looking
SAR spatial resolution (ICEYE / Capella Spotlight)	Approximately 0.5 m single-look; 1 m typical product
Optical resolution (Sentinel-2 MSI)	10 m (VIS/NIR), 20 m (SWIR)
Revisit interval	Sentinel-1: 6 days at mid-latitudes, 12 days globally. ICEYE/Capella: taskable within hours, subject to constellation load
InSAR displacement sensitivity	Millimetre-scale line-of-sight displacement over coherent surfaces; centimetre-scale practical threshold after atmospheric correction
Minimum detectable crest-line change (optical)	Approximately 10-20 m with Sentinel-2; approximately 1-2 m with 0.5 m SAR spotlight
SAR frequency	C-band (5.4 GHz, Sentinel-1); X-band (9.65 GHz, ICEYE and Capella)
Archive depth	Sentinel-1: from April 2014. Sentinel-2: from June 2015. Commercial SAR: varies by operator
Cloud penetration	SAR: full cloud penetration at C- and X-band. Optical: blocked by cloud; SWIR partially penetrates thin haze
Typical analysis latency (emergency response)	6-12 hours from SAR acquisition to first-pass assessment, assuming pre-prepared baseline

Analytics Satellize can run

Coherence-change map of embankment slopes	Sentinel-1 interferometric coherence differencing between pre-event stack and first post-event acquisition	GeoTIFF and PDF report flagging structures with statistically anomalous coherence loss, ranked by severity
InSAR displacement map of dam body	Differential InSAR (two-pass or multi-temporal) using Sentinel-1 or commercial SAR; atmospheric phase screen correction applied	Line-of-sight displacement GeoTIFF with uncertainty bounds; annotated cross-section through dam crest
Optical breach and erosion assessment	Sentinel-2 multispectral change detection using pre/post NDWI and SWIR band differencing	Classified change layer (breach, erosion, inundation, no change) with area statistics; suitable for situation reports
Downstream inundation extent	SAR backscatter thresholding (Sentinel-1 or ICEYE) combined with Sentinel-2 NDWI where cloud-free	Inundation polygon layer in GeoPackage format, timestamped, with estimated area in hectares
Multi-structure triage ranking	Composite scoring of coherence anomaly, displacement magnitude, and optical change across all monitored structures in an affected region	Prioritised inspection list with per-structure confidence ratings; delivered as structured PDF and CSV within 12 hours of first post-event SAR pass
Time-series deformation monitoring (post-event)	Persistent Scatterer or Small Baseline Subset InSAR time series over subsequent Sentinel-1 acquisitions	Weekly displacement time-series chart per structure; GIS layer updated each new acquisition cycle for up to 90 days post-event

Sources

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.