Reservoir and lake surface-level monitoring by radar altimetry

Satellite radar altimetry and photon-counting lidar measure water-surface elevation in reservoirs and lakes to centimetre accuracy, with no in-country infrastructure required. Virtual-station time series now extend to bodies as small as 250 m across.

Sensors

Sentinel-6 Michael Freilich (Poseidon-4 altimeter): Primary operational altimeter for ocean and large inland water bodies. Ku-band and C-band dual-frequency radar; along-track footprint roughly 300 m in high-resolution closed-loop mode. 10-day exact repeat orbit. Absolute range accuracy over calm open water approximately 2–3 cm. Coverage limited to lakes large enough to fill the footprint cleanly; performance degrades near complex shorelines.
SWOT KaRIn (Ka-band Radar Interferometer): Wide-swath interferometric altimeter launched December 2022. Dual 50 km swaths separated by a 20 km nadir gap; spatial resolution approximately 50–100 m in pixel terms, enabling water-surface elevation retrieval over water bodies as small as 250 m across. 21-day exact repeat. Height accuracy target 10 cm (rms) over 1 km² averaging areas. A step-change in small-lake and reservoir coverage globally.
ICESat-2 ATLAS (photon-counting lidar): 532 nm green lidar with six beams in three pairs. Along-track spacing ~70 cm; ground footprint ~17 m per photon cluster. Inland water-surface elevation retrievable to a few centimetres where surface is calm and cloud-free. 91-day exact repeat limits temporal density. Provides independent cross-validation of radar altimetry and is unaffected by radar signal contamination from vegetation fringing the shoreline.
Envisat RA-2 (archive): Ku- and S-band altimeter operational 2002–2012. 35-day repeat. Now a historical archive only, but its decade-long record underpins virtual-station time series that pre-date Sentinel-3 and Sentinel-6. Essential for trend analysis on reservoirs where no in-situ gauge record exists.
Sentinel-3A/3B (SRAL altimeter): Ku- and C-band SAR-mode altimeter on a 27-day repeat orbit. Two satellites in the same orbital plane reduce effective revisit to roughly 14 days over fixed ground tracks. Accuracy over inland water 3–5 cm on large targets, degrading on smaller bodies. Provides continuity between Envisat and Sentinel-6 and is freely available via Copernicus.

What the radar is actually measuring

A nadir-pointing altimeter emits a short microwave pulse and times its return. The two-way travel time, corrected for atmospheric delays (ionospheric, wet and dry tropospheric), tidal loading and instrument biases, gives the distance from the satellite to the water surface. Subtract that from the precisely known satellite orbit and you have water-surface elevation relative to a reference ellipsoid, typically WGS84 or EGM2008 geoid. The figure that emerges is not a depth; it is an absolute height. To convert it into something operationally useful, you either compare successive passes at the same location (a virtual gauge) or difference it against bathymetric surveys to estimate stored volume.

Ku-band (roughly 13.6 GHz) is the standard frequency for altimetry over water. It penetrates light rain and reflects well from calm surfaces. Ka-band, used by SWOT's KaRIn, reflects even more strongly from water but is more sensitive to rain. The dual-frequency design on Sentinel-6 allows the ionospheric delay to be estimated and removed directly, which matters most at low latitudes where the ionosphere is most active.

Virtual stations: how a ground track becomes a time series

A virtual station is simply a geographic window, typically a polygon drawn over a water body, intersected with a satellite's repeat ground track. Every time the satellite passes over that window, it contributes one elevation estimate. Repeat intervals range from 10 days (Sentinel-6) to 27 days (Sentinel-3) to 91 days (ICESat-2). For a reservoir managed on weekly or daily cycles, a 10-day revisit is often too coarse to catch rapid drawdown events; that is an honest constraint that no amount of processing can fix.

The archive depth matters enormously for trend detection. Combining Envisat RA-2 (2002–2012), Sentinel-3 (2016–present) and Sentinel-6 (2020–present) produces a 20-year virtual-station record over thousands of lakes globally, without a single gauge ever having been installed. Databases such as USDA's G-REALM and the Hydroweb service maintained by CNES/LEGOS have been doing exactly this for major reservoirs since the early 2000s. The science is not new; the improvement is in spatial coverage and accuracy as new missions arrive.

SWOT changes the size threshold

Before SWOT, satellite altimetry was largely limited to lakes and reservoirs several kilometres across, because the footprint of a conventional nadir altimeter is too large to avoid contamination from surrounding land. SWOT's KaRIn instrument uses interferometry across two antennas separated by a 10 m boom to produce a two-dimensional map of water-surface elevation across a 120 km swath (minus the 20 km nadir gap). The published minimum detectable water body is approximately 250 m in at least one dimension, with height accuracy targets of 10 cm rms over 1 km² averaging areas.

That threshold matters for reservoir monitoring in mountainous or heavily fragmented terrain, where many operationally critical reservoirs are small. The 21-day repeat is a limitation; SWOT is not a rapid-response tool. But for monthly or seasonal storage accounting, it covers water bodies that were previously invisible to altimetry entirely. Early validation results published by the SWOT science team show good agreement with in-situ gauges on lakes above roughly 0.5 km², with performance degrading on smaller and more irregularly shaped bodies.

Where accuracy breaks down

Over large, open, calm water the numbers are impressive: Sentinel-6 achieves 2–3 cm range accuracy under good conditions. Reality is messier. Wind-roughened surfaces shift the apparent reflection point. Vegetation fringing a shoreline contaminates the radar return, biasing the elevation estimate high. Narrow reservoirs cause the altimeter footprint to straddle land and water simultaneously, producing a mixed return that waveform retracking algorithms handle imperfectly. Mountain reservoirs with steep valley walls create additional multipath problems.

Cloud has no effect on radar altimetry, which is one of its genuine advantages over optical methods. But rain does attenuate the signal, particularly at Ka-band. ICESat-2's lidar, by contrast, is blocked by any significant cloud cover, limiting its utility in persistently overcast regions. Neither system provides the daily revisit that a dam operator managing flood risk actually wants. That gap is typically filled by combining satellite altimetry for absolute elevation with gauge telemetry for high-frequency variation, where gauges exist at all.

From elevation to stored volume

Water-surface elevation is not the same as stored volume. The conversion requires a hypsometric curve, the relationship between surface elevation and reservoir volume, which comes from bathymetric surveys or, increasingly, from pre-impoundment topographic data combined with optical imagery to map the shoreline at multiple known elevations over time. Once the hypsometric curve is established, each altimetric elevation estimate can be converted to a volume estimate. Uncertainty compounds: a 10 cm elevation error translates to a volume error that depends entirely on the shape of the reservoir at that elevation. In a steep-sided canyon reservoir it may be trivial; in a shallow, wide impoundment it can be substantial.

Satellize builds virtual-station time series and hypsometric volume products over client-specified reservoirs using open-constellation data from Sentinel-3, Sentinel-6 and SWOT, cross-validated against ICESat-2 passes where available. The Tonga crop-estimation programme is a different analytics domain, but the underlying pipeline for multi-mission data fusion is the same. Clients receive a structured dataset and a monthly briefing rather than raw altimetry files.

Practical questions before commissioning a monitoring programme

Three questions determine whether satellite altimetry is the right tool for a given reservoir. First, how large is the water body? Below 250 m in its narrowest dimension, even SWOT struggles. Second, what temporal resolution does the decision-maker actually need? If weekly drawdown tracking is required, altimetry alone cannot deliver it. Third, is there an existing gauge record to anchor the virtual-station series? Absolute elevation from a satellite is useful; relative change is more reliable, and both are most valuable when tied to at least some ground truth.

If the answer to all three questions is favourable, satellite altimetry offers something a gauge network cannot: simultaneous coverage of every reservoir in a basin, with no maintenance burden, no risk of vandalism or flood damage to instruments, and a historical record that predates any decision to start monitoring. For water-stressed governments managing dozens of reservoirs across remote terrain, that combination is genuinely difficult to replicate by other means.

Typical figures

Along-track footprint (Sentinel-6 high-resolution mode)	~300 m
Swath width (SWOT KaRIn)	2 × 50 km (20 km nadir gap)
Minimum water body (SWOT)	~250 m in at least one dimension
Height accuracy, large open water (Sentinel-6)	2–3 cm rms under good conditions
Height accuracy target (SWOT, 1 km² average)	10 cm rms
Repeat interval	Sentinel-6: 10 days; Sentinel-3: 27 days (14 days with A+B); SWOT: 21 days; ICESat-2: 91 days
Radar frequency	Ku-band (~13.6 GHz) for Sentinel-3/6; Ka-band (~35 GHz) for SWOT KaRIn; 532 nm lidar for ICESat-2
Archive depth (combined missions)	Envisat RA-2 from 2002; continuous Sentinel-3 from 2016; Sentinel-6 from 2020
Cloud sensitivity	None for radar altimetry; ICESat-2 lidar blocked by significant cloud
Latency (Sentinel-6 near-real-time product)	~3 hours from acquisition to NRT product; science-grade product within ~60 days

Analytics Satellize can run

Virtual-station elevation time series	Multi-mission altimetry waveform retracking and orbit/atmospheric correction; stations defined by ground-track intersection with user-specified reservoir polygons	Monthly CSV or GeoJSON feed of water-surface elevation per reservoir, with per-pass uncertainty flags
Reservoir storage volume estimate	Hypsometric curve construction from multi-date optical shoreline mapping combined with altimetric elevation series; volume computed by integration	Monthly volume estimate (in km³ or Mm³) with confidence interval; updated hypsometric curve as new elevation-shoreline pairs accumulate
Anomaly and drought-stress alert	Z-score comparison of current elevation against the climatological distribution for the same calendar period across the archive record	Automated alert (email or API webhook) when reservoir falls below a user-defined percentile threshold
Basin-wide reservoir inventory and ranking	SWOT and Sentinel-3 virtual-station extraction across all detectable water bodies within a defined basin polygon; elevation change ranked by magnitude	Quarterly GIS layer and ranked table of reservoirs by storage-change severity, suitable for water-ministry briefing
Long-term trend report	Linear and seasonal decomposition of 20-year virtual-station series (Envisat through Sentinel-6); Mann-Kendall trend test applied per station	PDF trend report per reservoir or basin, with annotated time-series charts and statistical significance statements
ICESat-2 cross-validation layer	Co-location of ICESat-2 ATLAS surface-elevation photon returns over reservoir surfaces with concurrent radar altimetry estimates; bias and scatter statistics computed	Validation summary table confirming or flagging systematic offsets in the radar virtual-station series; recommended corrections where applicable

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.