Hydropower reservoir water level and storage estimation from satellite altimetry
Satellite altimetry from ICESat-2, Sentinel-6, SWOT and Sentinel-3 measures reservoir surface elevation to within centimetres, converting to storage volume via area-elevation curves. Coverage is geometry-dependent: not every reservoir is tractable.
Sensors
- ICESat-2 ATLAS (NASA): Photon-counting lidar at 532 nm. Ground track footprint approximately 17 m diameter, shot spacing 0.7 m along track. Elevation accuracy better than 3 cm over flat water. Repeat cycle 91 days; exact-repeat tracks pass over the same ground reference track, so a given reservoir is measured only if a track crosses it. Six beam pairs give some cross-track sampling but beam separation is roughly 3.3 km.
- Sentinel-6 Michael Freilich (ESA/EUMETSAT/Copernicus): Ku-band and C-band radar altimeter in a 10-day exact-repeat orbit at 1336 km altitude. Nadir-only measurement; footprint roughly 2 km diameter over inland water. Absolute sea-surface height accuracy better than 2 cm; over reservoirs, range noise is higher due to off-nadir land returns, but level retrievals of 5–10 cm accuracy are documented for reservoirs wider than roughly 500 m at the track crossing.
- SWOT KaRIn (NASA/CNES): Ka-band radar interferometer with a 120 km wide swath (two 50 km swaths separated by a 20 km nadir gap). Nominal water surface height accuracy 10 cm for water bodies larger than 1 km². Repeat cycle 21 days. Launched December 2022; SWOT is the first mission that can measure reservoir level without requiring a track to pass directly overhead, making it tractable for a far wider range of reservoir sizes and shapes.
- Sentinel-3 SRAL (ESA/Copernicus): Ku/C dual-frequency radar altimeter, 27-day repeat, nadir-only. Water level accuracy 10–30 cm over inland targets depending on reservoir geometry. Best suited to large reservoirs (surface area above roughly 50 km²) where the return waveform is not dominated by surrounding land. The Copernicus Global Land Service Hydrology portal publishes near-real-time Sentinel-3 water level time series for several hundred monitored lakes and reservoirs.
Why elevation is the number that matters
Hydropower generation capacity is a direct function of hydraulic head: the vertical distance between the water surface and the turbine intake. A reservoir dropping two metres does not lose two metres of storage uniformly; because reservoir basins are rarely cylindrical, the storage loss depends on the area-elevation curve, which is specific to each dam's bathymetry. Getting the water level right is therefore the prerequisite for every downstream calculation, from available energy to flood-release scheduling.
Gauge stations measure level precisely, but they are expensive to maintain, politically sensitive to share, and frequently absent on reservoirs in lower-income river basins. Satellite altimetry sidesteps the data-sharing problem entirely. The measurement is made from orbit, requires no in-country infrastructure, and the physics are the same whether the reservoir is in Norway or on the Mekong.
Which reservoirs are actually measurable
Track spacing is the central constraint. ICESat-2 and Sentinel-3 SRAL are nadir-pointing instruments: they only measure where a ground track crosses the water surface. ICESat-2's 91-day repeat gives excellent elevation accuracy but sparse spatial coverage. A reservoir must be wide enough at the track crossing to return a clean water surface signal; in practice this means at least 200–300 m of open water perpendicular to the track for ICESat-2, and 500 m or more for Sentinel-3. Many smaller or narrow reservoirs simply do not have a usable track crossing.
SWOT changes this calculus. Its 120 km swath covers most of the globe every 21 days, and its interferometric approach measures water surface height across the entire visible water body rather than along a single nadir line. For reservoirs larger than roughly 1 km² of surface area, SWOT can deliver level estimates without the track-crossing lottery. For reservoirs between 0.1 and 1 km², accuracy degrades and results should be treated as indicative. Reservoirs smaller than 0.1 km² are not reliably tractable with any current spaceborne altimeter.
Geometry matters beyond size. Long, narrow reservoirs in steep gorges (common in Himalayan and Andean hydropower) can be difficult even when they are large, because surrounding terrain contaminates the radar return. ICESat-2's lidar is less susceptible to this problem; its 532 nm photon-counting approach distinguishes water returns from land returns more cleanly, making it particularly valuable for canyon reservoirs where radar altimetry struggles.
Converting a level reading to a storage volume
A water level measurement becomes operationally useful only when paired with an area-elevation curve, sometimes called a hypsometric curve. This relates water surface elevation to the planimetric area of the reservoir at that elevation, and by integration, to total storage volume. These curves are derived from bathymetric surveys, pre-impoundment topographic maps, or, increasingly, from optical satellite imagery combined with digital elevation models. Landsat and Sentinel-2 time series can reconstruct area-elevation relationships by tracking the waterline at different observed levels across years of imagery.
The conversion introduces its own uncertainty. Bathymetric surveys age: sedimentation progressively reduces active storage, and a curve derived from a 1980s survey may overestimate current storage by 10–30% in heavily silted reservoirs. Where updated bathymetry is unavailable, the storage estimate carries that uncertainty explicitly. Honest reporting requires stating it.
Latency, archive depth and operational use
ICESat-2 data are publicly available through NASA Earthdata, typically within a few days of acquisition. Sentinel-3 near-real-time water level products for monitored sites are published by the Copernicus Global Land Service with latency of roughly 3 days. SWOT science data are distributed through NASA's Physical Oceanography DAAC (PO.DAAC) and CNES; operational latency for hydrological products is currently in the range of days to weeks as the mission matures.
Archive depth is meaningful for trend analysis. ICESat-2 data run from October 2018. Sentinel-3A has operated since 2016, giving nearly a decade of level records for monitored large reservoirs. For sites where older radar altimeters (Envisat, Jason-2) passed over, virtual station records extending back to the early 2000s exist in databases such as the USDA G-REALM and the Copernicus Global Land Service hydrology portal. This multi-decadal context is directly relevant to energy traders and project financiers assessing drought risk.
Honest limits: what altimetry cannot do
Cloud cover does not affect radar altimetry or ICESat-2 in the same way it affects optical sensors, but it is not irrelevant. ICESat-2 is a lidar; dense cloud can attenuate the signal enough to prevent a clean water return, particularly in tropical regions during monsoon season. Sentinel-3 and SWOT radar altimetry are essentially all-weather, but heavy rain can introduce path-delay errors.
Altimetry measures surface elevation, not total storage. If a reservoir has significant dead storage below the lowest outlet, that volume is invisible to the satellite. Similarly, if the area-elevation curve is poorly constrained in the upper range (because the reservoir has rarely been full), storage estimates near capacity carry larger uncertainty. Sub-centimetre gauge precision is not replicated by any current spaceborne altimeter over inland water; for operational dispatch decisions at a single plant, in-situ gauges remain superior. The satellite's advantage is breadth: monitoring dozens of reservoirs across a basin simultaneously, without negotiating data-sharing agreements with each riparian operator.
From raw elevation to an energy intelligence product
The analytic chain from altimeter return to generation-capacity estimate involves several steps: waveform retracking to extract surface elevation, cross-referencing against a geoid model to convert ellipsoidal height to orthometric height, matching to the area-elevation curve, integrating to storage volume, and finally applying a plant-specific head-to-power relationship. Each step has documented methods in the peer-reviewed literature; none is proprietary.
What varies between providers is the degree to which this chain is automated, validated against available gauge data, and delivered at the cadence and format an energy trader or grid operator can actually use. Satellize runs this pipeline on open-constellation data, including Sentinel-3 and ICESat-2, with commercial SWOT tasking added where clients require higher-frequency coverage. The Tonga crop-estimation programme demonstrated the same underlying approach of converting physical remote-sensing signals into decision-relevant commodity metrics. For hydropower clients, the deliverable is a reservoir-level and storage time series, updated at each satellite overpass, flagged when level crosses operator-defined thresholds, and exported in formats compatible with energy management systems.
If you manage or trade generation capacity across a multi-reservoir basin, the first practical step is a tractability assessment: which of your reservoirs have usable altimeter tracks, what is the achievable accuracy at each site, and where does SWOT coverage fill the gaps. That assessment takes days, not months, and it determines whether the investment in an ongoing monitoring feed is justified before any commitment is made.
Typical figures
| ICESat-2 elevation accuracy (flat water) | Better than 3 cm (published mission specification; ATLAS instrument) |
| Sentinel-6 / Sentinel-3 level accuracy (large reservoirs) | 5–30 cm depending on reservoir width at track crossing and surrounding terrain |
| SWOT KaRIn level accuracy | ~10 cm for water bodies ≥1 km²; degrades for smaller targets |
| Repeat cycle | ICESat-2: 91 days; Sentinel-3: 27 days; Sentinel-6: 10 days; SWOT: 21 days |
| Minimum tractable reservoir size | ~1 km² for SWOT; ~0.05 km² for ICESat-2 where a track crosses; ~50 km² for reliable Sentinel-3 retrieval |
| Swath / coverage mode | ICESat-2, Sentinel-3, Sentinel-6: nadir-only. SWOT: 120 km swath (two 50 km swaths, 20 km nadir gap) |
| Data latency (public archives) | ICESat-2: ~3 days; Sentinel-3 near-real-time hydrology products: ~3 days; SWOT science products: days to weeks (mission maturing) |
| Archive depth | ICESat-2 from October 2018; Sentinel-3A from 2016; legacy radar altimeter virtual stations from early 2000s (Envisat, Jason series) |
| Storage volume uncertainty (conversion step) | Dependent on area-elevation curve quality; sedimentation can introduce 10–30% bias in older surveys |
| Cloud / weather sensitivity | Radar altimetry (Sentinel-3, Sentinel-6, SWOT): all-weather. ICESat-2 lidar: signal loss possible under dense cloud |
Analytics Satellize can run
| Reservoir water level time series | Waveform retracking (radar) or photon-filtering (ICESat-2 lidar) to extract surface elevation; geoid-referenced to orthometric height | CSV or GeoJSON time series updated at each satellite overpass, with per-observation uncertainty flag |
| Storage volume estimate | Integration of area-elevation (hypsometric) curve against measured level; curve sourced from bathymetric data or optically derived waterline regression | Storage volume in million cubic metres per observation epoch, delivered as time series with uncertainty range |
| Tractability assessment for a reservoir portfolio | Overlay of ICESat-2 reference ground tracks, Sentinel-3 virtual station database, and SWOT swath coverage against client reservoir geometries | Per-reservoir feasibility report: achievable sensor, expected accuracy, revisit frequency, and recommended monitoring configuration |
| Threshold alert: level below critical head | Comparison of current level estimate against operator-defined minimum head for generation; alert triggered when level crosses threshold on two consecutive overpasses | Automated alert (email or API webhook) with current level, distance to threshold, and trend over prior 30 days |
| Area-elevation curve derivation from optical imagery | Waterline extraction from Sentinel-2 or Landsat multispectral imagery at multiple historical levels; regression against coincident altimeter readings to build hypsometric relationship | Area-elevation curve table and fitted polynomial, with residual statistics, delivered as a one-time GIS dataset |
| Multi-reservoir basin storage dashboard | Aggregation of per-reservoir level and storage estimates across a defined river basin; normalised against historical percentile distribution | Weekly basin-level storage report in PDF and interactive web format, showing each reservoir as percentage of historical median |
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.