Coal ash pond expansion and structural change monitoring

Fly ash impoundments expand, shift and occasionally fail with little surface warning. Sentinel-1 SAR coherence and Sentinel-2 optical time series can track pond boundary changes, new dyke construction and surface moisture anomalies that warrant geotechnical follow-up.

Sensors

Sentinel-1 SAR (C-band, ESA): 10 m ground range resolution in Interferometric Wide Swath mode; 6-day revisit at mid-latitudes with both satellites active. Coherence change detection identifies dyke surface disturbance and new earthworks; backscatter intensity distinguishes wet slurry from dry consolidated ash.
Sentinel-2 MSI (ESA): 10 m resolution in visible and near-infrared bands; 20 m in shortwave infrared. 5-day revisit (combined Sentinel-2A and 2B). SWIR bands 11 and 12 are sensitive to surface moisture, making them useful for mapping saturated zones on dyke faces or pond margins. Cloud cover is the principal operational constraint.
Landsat 8/9 OLI (USGS/NASA): 30 m multispectral resolution; 16-day single-satellite revisit. Less spatially precise than Sentinel-2 but provides a continuous archive back to 1972 across the Landsat family, enabling decade-scale pond expansion histories and baseline establishment.
Planet PlanetScope: 3 m resolution, daily revisit in clear conditions. Useful for resolving fine dyke crest geometry and detecting small-scale earthwork activity that 10 m sensors may miss. Requires a commercial licence; archive depth is shallower than Sentinel or Landsat.

What ash ponds leave on the surface

Coal combustion produces fly ash at roughly 10–15 % of fuel mass. At a large power station burning several million tonnes of coal per year, the resulting ash volume is substantial. Wet sluicing, still common in older plant designs, deposits slurry into lined or unlined impoundments that grow outward and upward as operators raise dyke walls in successive lifts. The surface record of this process is legible from orbit.

Dry ash is highly reflective in visible wavelengths and appears pale grey to white in Sentinel-2 true-colour imagery. Wet slurry is darker and absorbs more near-infrared radiation. The boundary between active deposition zones and older, consolidated cells shifts over time as operators rotate discharge points. These transitions are detectable as changes in the Normalised Difference Water Index and in Sentinel-1 backscatter intensity, which drops sharply over saturated fine-grained material.

SAR coherence as a dyke construction detector

Synthetic aperture radar coherence measures how similar the radar phase is between two passes over the same ground. Stable, undisturbed surfaces such as compacted earth or concrete maintain high coherence across a 6- or 12-day interval. Active construction, earthmoving and fresh slurry deposition decorrelate the signal almost completely, producing coherence values near zero.

A time series of Sentinel-1 coherence images over an ash impoundment will show the active pond cells as persistent low-coherence zones. When a new dyke lift begins, the previously stable crest loses coherence for several weeks, then recovers as compaction proceeds. This pattern is distinguishable from vegetation growth or seasonal moisture change by its spatial geometry: dyke crests are linear, narrow and follow the pond perimeter. Automated change detection on coherence stacks can flag new construction activity within one to two repeat cycles, giving roughly 6–12 days of latency from event to alert.

One honest caveat: SAR coherence detects surface change. It cannot assess whether a dyke core is draining correctly, whether internal erosion is progressing, or whether pore-water pressure is approaching a critical threshold. Those questions require in-situ instrumentation or geotechnical investigation. Remote sensing here is a screening tool, not a substitute for engineering assessment.

Reading moisture anomalies on dyke faces

Seepage through an earthen dyke typically manifests at the surface as a zone of elevated moisture on the downstream face, often accompanied by soft ground, slumping or vegetation die-off as alkaline leachate reaches the toe. Sentinel-2 SWIR bands are sensitive to liquid water in the top few centimetres of soil. A persistent high-moisture signature on a dyke face, particularly one that appears or intensifies after a wet season or after pond water levels rise, is a surface indicator that warrants closer inspection.

The Modified Normalised Difference Water Index (MNDWI), computed from Sentinel-2 Green and SWIR Band 11, is more effective than NDWI for distinguishing moist soil from open water on these complex surfaces. Thresholds must be calibrated site-specifically because ash chemistry and surface albedo vary between facilities. A single anomalous image is rarely conclusive; the value lies in a persistent or growing signature across multiple cloud-free acquisitions over weeks to months.

PlanetScope's 3 m daily imagery can resolve individual seepage stains or slope failures on dyke faces that a 10 m sensor would average away. Where budget allows, combining Sentinel-2 for systematic coverage with targeted PlanetScope tasking during anomaly periods is a practical approach.

Pond boundary mapping and expansion rate estimation

Mapping the active water or slurry surface in each acquisition and differencing successive boundaries gives a direct measure of pond expansion. Over a multi-year Sentinel-2 archive, this produces an expansion rate in hectares per year that can be compared against permitted capacity limits or against declared generation output. A pond expanding faster than plant output would imply, or one that stops expanding abruptly, is worth investigating.

The Landsat archive extends the timeline back to the 1980s for most operating facilities. Pond boundaries are generally detectable at 30 m resolution because the spectral contrast between wet ash slurry and surrounding land cover is strong. For facilities where a historical baseline matters, such as those subject to regulatory review or litigation, Landsat provides evidence that no commercial archive can match in depth.

Boundary detection accuracy depends on the clarity of the water-land transition. Facilities with gradual drawdown zones or partially dried peripheral cells present ambiguous edges. Reported accuracy in published water-body delineation studies using Sentinel-2 is typically within one to two pixels of manually digitised boundaries under good conditions, which equates to roughly 10–20 m positional uncertainty on the pond perimeter.

Limits the method cannot cross

Subsurface dyke integrity is invisible to optical and SAR sensors at the resolutions available from the systems discussed here. Piping failures, internal erosion and foundation instability produce no reliable surface signature until they are already advanced. The 2008 Kingston Fossil Plant failure in Tennessee and the 2014 Dan River spill in North Carolina both involved structural failures that were not preceded by surface anomalies detectable from orbit.

Cloud cover is the dominant operational constraint for optical sensors. Sentinel-2 and PlanetScope imagery is unavailable during overcast periods, which can be prolonged in tropical or monsoon climates where many large coal-fired stations operate. SAR is cloud-transparent, making Sentinel-1 the more reliable monitoring backbone in those environments, with optical imagery used opportunistically.

Spatial resolution also sets a floor on what is detectable. Small cracks in dyke crests, minor slope failures covering a few square metres, or early-stage seepage stains narrower than roughly 10 m will not be resolved by Sentinel sensors. PlanetScope narrows but does not eliminate this gap.

Turning a monitoring programme into a decision tool

The practical output of this kind of monitoring is not a single map but a change signal delivered on a schedule that matches the pace of the hazard. For a facility with active deposition, monthly boundary updates and quarterly coherence-change summaries are a reasonable baseline. For a facility under regulatory scrutiny or showing prior anomalies, weekly SAR coherence alerts are achievable given Sentinel-1's repeat cycle.

Satellize structures analytics of this type as time-series feeds delivered as GIS layers or tabular reports, with threshold-based alerts when a defined change metric exceeds a site-specific baseline. The Tonga crop-estimation programme uses a similar architecture: systematic open-data ingestion, site-specific calibration and periodic structured output rather than ad-hoc image delivery. The same approach applies here, with the additional option of commercial PlanetScope tasking when a Sentinel anomaly warrants higher-resolution confirmation.

Regulators, insurers and plant operators all have legitimate reasons to want an independent, continuous record of surface change at ash impoundments. A well-documented satellite time series is also a defensible evidentiary record if a facility's compliance history is later disputed. That is a different kind of value from operational monitoring, but it is real and increasingly requested.

Typical figures

Spatial resolution (SAR)	10 m (Sentinel-1 IW mode); 3–5 m with commercial SAR on request
Spatial resolution (optical)	10 m visible/NIR, 20 m SWIR (Sentinel-2); 30 m (Landsat 8/9); 3 m (PlanetScope)
Revisit (SAR)	6 days at mid-latitudes with Sentinel-1A and 1B; single satellite gives 12 days
Revisit (optical)	5 days (Sentinel-2A+B combined); 16 days (Landsat); daily (PlanetScope, cloud-permitting)
Spectral bands used	C-band 5.4 GHz (SAR); Sentinel-2 Bands 3, 8, 11, 12 (Green, NIR, SWIR); Landsat OLI Bands 3, 5, 6, 7
Minimum detectable boundary shift	Approximately 10–20 m positional uncertainty on pond perimeter (Sentinel-2); ~30 m (Landsat)
Archive depth	Sentinel-1 from 2014; Sentinel-2 from 2015; Landsat from 1972; PlanetScope from approximately 2016
Alert latency	6–12 days from surface change event to SAR coherence anomaly detection
Coverage	Global; Sentinel-1 coverage varies by acquisition mode and region; Sentinel-2 covers all land surfaces
Delivery formats	GeoTIFF change layers, GeoJSON pond boundaries, tabular expansion metrics, PDF summary reports

Analytics Satellize can run

Pond boundary time series	MNDWI and NDWI thresholding on Sentinel-2 and Landsat multispectral stacks; manual QA on ambiguous edges	Annual or quarterly GeoJSON boundary layers with expansion area and rate statistics
Dyke construction activity alerts	Sentinel-1 coherence change detection; decorrelation on linear dyke-crest geometry flagged against stable-surface baseline	Automated alert with flagged image chip and coherence difference map; 6–12 day latency
Surface moisture anomaly map	MNDWI and SWIR Band 11 anomaly detection on Sentinel-2 time series; persistent-anomaly filtering across minimum three cloud-free acquisitions	GeoTIFF moisture anomaly layer with site-specific threshold annotation; monthly or event-triggered
Multi-decadal expansion history	Landsat archive water-body delineation from 1970s to present; change-point analysis on annual boundary centroids	PDF report with mapped expansion stages and tabular area-by-year data; suitable for regulatory or insurance review
High-resolution anomaly confirmation	Targeted PlanetScope tasking following Sentinel-flagged events; visual and spectral comparison against baseline	3 m orthorectified imagery with annotated change polygons; delivered within 48 hours of tasking window
Backscatter intensity trend analysis	Sentinel-1 sigma-nought time series over defined pond zones; wet-to-dry transition detection using published backscatter thresholds for fine-grained saturated soils	Time-series chart and GIS layer showing backscatter regime shifts; quarterly summary report

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.