Reservoir sedimentation and storage capacity loss from orbit

Repeat satellite altimetry, optical surface-area mapping and ICESat-2 photon-counting lidar combine to track how sediment infill shrinks live storage, without a single diver or echo sounder.

Sensors

ICESat-2 ATL03 / ATL13: Photon-counting lidar at 532 nm; ATL13 provides inland water-surface heights to roughly 5–10 cm vertical precision. In optically clear shallows, photons penetrate to around 1–2 m depth, giving bathymetric returns near inflow deltas where sediment accumulation is fastest. Repeat ground-track spacing is 91 days for the exact repeat cycle, with denser coverage at higher latitudes.
Sentinel-3 SRAL: Ku/C-band radar altimeter; over inland water bodies larger than roughly 0.5–1 km across, it tracks water-surface elevation to 5–10 cm RMS. A 27-day exact repeat cycle provides a consistent time series for constructing elevation-area-volume (hypsometric) curves. Free and open via EUMETSAT.
Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at mid-latitudes with both satellites. Used to delineate the water-surface boundary at each altimetry epoch, map turbidity plumes via the red and red-edge bands, and detect delta-fan growth at the pixel scale. Cloud cover is the primary operational constraint.
Landsat 8 / 9 OLI: 30 m multispectral; 8-day combined revisit. The archive extends to 1984 (Landsat 5 TM), making it the longest consistent optical record for historical storage-loss reconstruction. Band 3 (green) and band 4 (red) support empirical turbidity retrieval; the panchromatic band at 15 m sharpens delta-fan delineation.
Sentinel-1 SAR C-band: 6–12 day revisit, 10 m IW-mode resolution; cloud-independent water-body mapping. SAR backscatter contrast between open water and exposed sediment bars allows surface-area estimates on days when optical sensors are obscured. Useful for monsoon or consistently cloudy catchments.

What a shrinking bathtub ring actually measures

Every reservoir has a hypsometric curve: the relationship between water-surface elevation and stored volume. As sediment accumulates on the bed, the curve shifts. A given elevation now corresponds to less water. That shift is the signal satellites are reading, indirectly but systematically.

The method pairs two independently observable quantities. Radar or lidar altimetry gives the water-surface elevation at a known date. Optical or SAR imagery, processed to extract the precise shoreline at that same date, gives the surface area. Plot enough elevation-area pairs over time and you can fit a hypsometric curve, then compare curves from different years to infer volume change. The arithmetic is straightforward; the difficulty is in the measurement precision and the consistency of the shoreline extraction across sensors and seasons.

ICESat-2 and the shallow-water problem

Conventional altimeters measure the water surface. They say nothing about what is happening below it. ICESat-2's 532 nm photon-counting lidar changes that in one specific zone: the shallow inflow delta, where sedimentation rates are highest and where the water column is often thin enough for green light to reach the bed and return a detectable signal.

ATL13 inland water-surface heights carry roughly 5–10 cm vertical precision under good conditions. In clear or moderately turbid shallows, ATL03 raw photon returns can be processed to recover bathymetric profiles along the ground track, typically to 1–2 m depth before turbidity attenuates the signal completely. That is shallow, but it is precisely where newly deposited delta fans grow. The 91-day repeat cycle means seasonal coverage is sparse; analysts typically combine multiple years of passes to build a composite bathymetric picture rather than a true time series.

One honest caveat: ICESat-2 ground tracks are fixed and narrow. They sample the delta only where the track crosses it. Spatial interpolation between tracks introduces uncertainty that grows with distance from the nearest pass.

Turbidity plumes as a proxy for sediment supply

Optical imagery cannot measure what has already settled, but it can observe what is arriving. Suspended sediment increases water-leaving radiance in the red and near-infrared bands in a physically predictable way. Sentinel-2 band 4 (665 nm) and band 5 (705 nm) are particularly useful for empirical turbidity retrieval in inland waters, with published studies reporting retrieval errors of roughly 20–40 percent against in-situ measurements depending on sediment type and atmospheric correction quality.

Tracking plume extent, frequency and intensity across a multi-year image archive gives a qualitative indicator of sediment loading: large, persistent plumes after flood events suggest high trap efficiency and accelerating infill. This does not replace a bathymetric survey, but it adds physical context that pure elevation-area analysis lacks. A reservoir that shows declining turbidity plumes over time, despite similar upstream rainfall, may be trapping less sediment because the delta fan has prograded far enough to reduce the settling distance.

Delta-fan growth: what high-resolution imagery resolves

At low reservoir levels, exposed sediment bars and advancing delta fans become directly visible in optical imagery. Sentinel-2 at 10 m and Landsat at 30 m can map the planimetric extent of these features. Repeated mapping at comparable water levels, using altimetry to condition the comparison, allows analysts to quantify how far the fan has prograded and to estimate the volume of material deposited above the water line.

The Landsat archive is particularly valuable here. A reservoir impounded in the 1970s or 1980s will have dozens of low-water images across four decades. Digitising the exposed delta in each gives a long-run record of fan growth that no in-situ programme could match retrospectively. The spatial resolution limit is real: features smaller than roughly 15–30 m are not reliably resolved, and shallow subaqueous bars remain invisible until exposed.

Honest limits: what the method cannot resolve

Capacity loss below roughly 2–5 percent per decade sits at or below the combined uncertainty of the elevation-area-volume method as currently practised from open constellations. Altimeter precision, shoreline extraction errors and seasonal volume fluctuations all contribute noise that can mask slow, steady infill. Reservoirs smaller than about 0.5–1 km across are too small for Sentinel-3 SRAL to track reliably; ICESat-2 may still provide useful surface heights if a ground track crosses the water body, but coverage is not guaranteed.

Cloud cover is a persistent problem in tropical catchments, precisely where many heavily sedimented reservoirs sit. SAR fills some gaps, but SAR-derived shorelines carry their own uncertainties, particularly in wind-roughened conditions or where emergent vegetation blurs the water edge. The method works best as a screening tool: it identifies reservoirs where capacity loss is large enough to warrant a targeted echo-sounder survey, rather than replacing that survey entirely.

Satellize applies this workflow on open Sentinel and Landsat data, with ICESat-2 passes incorporated where ground-track geometry is favourable. The approach is similar in structure to the crop-estimation analytics we run for the Kingdom of Tonga: open-constellation data, physically grounded retrieval, and explicit uncertainty bounds delivered alongside the estimate.

Turning the curve into an operational number

The end product an operator needs is not a hypsometric curve. It is a number: live storage remaining, expressed in million cubic metres, with a confidence interval and a trend rate. Deriving that from the satellite inputs requires fitting a curve to the elevation-area pairs, integrating under the curve between the dead-storage sill and the spillway crest, and propagating measurement uncertainties through the integration.

Published studies on large reservoirs, including work on the Indus system and African rift-valley lakes, suggest that where Sentinel-3 altimetry and Landsat area estimates are combined over multi-year windows, volume change can be detected to within roughly 5–15 percent for reservoirs above a few hundred million cubic metres. Smaller reservoirs require either higher-resolution commercial altimetry or in-situ augmentation. Reporting the uncertainty honestly is not a weakness in the product; it is the information a dam operator needs to decide whether a bathymetric survey is worth commissioning.

Typical figures

Water-surface elevation precision	Sentinel-3 SRAL: 5–10 cm RMS on reservoirs wider than ~0.5–1 km; ICESat-2 ATL13: ~5–10 cm under favourable conditions
Optical spatial resolution	Sentinel-2 MSI: 10 m (visible/NIR); Landsat 8/9 OLI: 30 m multispectral, 15 m panchromatic
SAR spatial resolution	Sentinel-1 IW mode: 10 m (range × azimuth ~5 × 20 m native, multi-looked to 10 m)
Revisit cadence	Sentinel-2: 5 days (both satellites); Landsat 8+9 combined: 8 days; Sentinel-3 SRAL: 27-day exact repeat; ICESat-2: 91-day exact repeat
Minimum reservoir size for altimetry	~0.5–1 km width for Sentinel-3 SRAL; ICESat-2 track-dependent, no hard minimum but coverage not guaranteed
ICESat-2 bathymetric depth limit	~1–2 m in clear to moderately turbid water at 532 nm; shallower in high-turbidity inflow zones
Volume-change detection threshold	Approximately 5–15% for reservoirs above a few hundred Mm³; below ~2–5% per decade, in-situ augmentation is required
Archive depth	Landsat: 1984 to present (Landsat 5 TM onward); Sentinel-2: 2015 to present; Sentinel-3: 2016 to present; ICESat-2: 2018 to present
Turbidity retrieval uncertainty	Approximately 20–40% against in-situ reference, depending on sediment type and atmospheric correction
Delivery formats	GeoTIFF hypsometric rasters, CSV elevation-area-volume tables, GeoJSON shoreline polygons, PDF capacity-loss reports

Analytics Satellize can run

Multi-epoch hypsometric curve	Paired Sentinel-3 or ICESat-2 elevation with Sentinel-2/Landsat optical shoreline extraction; polynomial or power-law curve fitting to elevation-area pairs	CSV and chart showing elevation vs. area vs. volume at each epoch, with uncertainty bounds; updated annually or on request
Live-storage volume trend	Integration of hypsometric curves between dead-storage sill and spillway crest; linear or piecewise trend fitting across available epochs	Annual capacity-loss rate in Mm³/yr and percent of original design capacity, with 90% confidence interval; PDF summary report
Delta-fan planimetric growth map	Multi-date Landsat/Sentinel-2 classification of exposed sediment at comparable low-water elevations; change detection between earliest and most recent epochs	GeoTIFF and GeoJSON showing fan extent at each epoch; area growth table; suitable for import into dam-management GIS
Turbidity plume frequency composite	Empirical suspended-sediment retrieval from Sentinel-2 red and red-edge bands across all cloud-free scenes; frequency-of-exceedance mapping	Annual raster showing percentage of observations above a defined turbidity threshold; identifies persistent high-load inflow corridors
ICESat-2 shallow bathymetry profile	ATL03 photon-cloud processing along available ground tracks crossing the inflow delta; refraction-corrected depth retrieval in water shallower than ~2 m	Point-cloud GeoJSON with depth estimates and quality flags along each track; advisory note on track geometry and coverage gaps
Historical storage reconstruction	Landsat archive processing from 1984 onward; low-water image selection conditioned on gauge or altimetry records to ensure comparable pool levels	Decadal capacity estimates back to first available imagery; PDF narrative with methodology and caveats; flags reservoirs warranting in-situ survey

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.