Sentinel-3 radar altimetry for sea level and inland water
Sentinel-3 SRAL measures sea surface height, wave height and inland water levels using SAR-mode radar altimetry. This page explains the physics, the 27-day repeat cycle, and where the method breaks down.
Sensors
- Sentinel-3 SRAL: Ku-band (13.575 GHz) and C-band (5.41 GHz) dual-frequency SAR altimeter. Along-track resolution in SAR mode roughly 300 m; across-track footprint several kilometres. 27-day exact repeat cycle. Primary instrument for this use case.
- Sentinel-6 Michael Freilich: High-resolution mode interleaved SAR altimeter on the reference orbit (66-day sub-cycle, 10-day repeat with Jason-3 interleaving). Provides the precision sea-level climate record; along-track resolution approximately 300 m in HR mode.
- Jason-3: Conventional pulse-limited Ku/C altimeter on the same 9.9156-day repeat reference orbit as Sentinel-6 MF. Along-track footprint several kilometres; the heritage benchmark against which Sentinel-3 retrievals are cross-calibrated.
- ICESat-2 ATLAS: Photon-counting green lidar (532 nm). Not a radar altimeter, but its 91-day repeat passes over lakes and reservoirs provide water-level cross-reference points accurate to roughly 3 cm, useful for validating SRAL retracking over small water bodies.
What a radar pulse actually measures
Sentinel-3 SRAL fires microwave pulses downward and times the two-way travel of the return echo. The elapsed time, corrected for the speed of light and a chain of atmospheric delays, gives the range from satellite to surface. Subtract that range from a precisely tracked orbital altitude and you have sea surface height relative to a reference ellipsoid. Simple in principle; demanding in practice.
The instrument operates in Synthetic Aperture Radar mode over most surfaces, processing consecutive pulses coherently to sharpen the along-track footprint to roughly 300 m. That is a significant improvement over the several-kilometre footprint of conventional pulse-limited altimeters like Jason-3. The dual-frequency design matters because Ku-band and C-band signals are refracted differently by free electrons in the ionosphere. Comparing the two ranges allows the ionospheric delay to be estimated and removed, which is a correction of order 1 to 5 cm depending on solar activity.
From open ocean to river gauge: the retracking problem
Over open ocean the return waveform has a predictable shape. A model called the Brown model fits it well, and sea surface height, significant wave height and near-surface wind speed can all be extracted from a single waveform. Significant wave height comes from the slope of the leading edge of the echo; wind speed from the backscatter amplitude. Accuracy for sea surface height over open ocean is typically in the 2 to 4 cm range for a single pass.
Over inland water the waveform looks nothing like the ocean case. Rivers, lakes and reservoirs produce bright, specular returns that saturate the tracker or arrive at unexpected times. Retracking algorithms, which refit the waveform offline rather than relying on the onboard tracker, recover usable water levels from these distorted echoes. Threshold retracking, Ice-1 and OCOG (Offset Centre Of Gravity) are the published methods most widely applied. The Global Hydrological Altimetry Service HYDROWEB and the Database for Hydrological Time Series of Inland Waters (DAHITI) both distribute Sentinel-3 derived river and lake levels using variants of these approaches.
The honest limit: the across-track footprint of SRAL remains several kilometres wide. A river narrower than roughly 100 to 200 m will have its waveform contaminated by returns from the surrounding floodplain or vegetation. Retracking can partially compensate, but level retrievals over channels narrower than that carry larger uncertainties, sometimes exceeding 50 cm. ICESat-2 lidar, with its 17 m beam diameter, handles narrow channels better, though its 91-day repeat and cloud sensitivity introduce different trade-offs.
The 27-day repeat and what it means for monitoring
Sentinel-3A and Sentinel-3B fly the same ground track with an 8-day offset between the two satellites, giving a combined revisit at any given virtual station of roughly 13 to 14 days. Each individual satellite repeats its exact ground track every 27 days. For sea level trend monitoring this is adequate: the signal of interest is slow, and multi-year stacking reduces noise. For flood early warning over rivers it is not. A river in spate can rise and fall in 48 hours; a 13-day revisit misses the peak entirely.
This is why operational flood monitoring typically combines altimetry with other sources. SAR imagery from Sentinel-1 maps flood extent at 10 to 20 m resolution regardless of cloud cover. Altimetry then provides the level at the few virtual stations it crosses. Neither alone tells the full story.
The coastal zone: where accuracy degrades
Within roughly 10 km of the coastline, altimetric retrievals become unreliable. The reason is straightforward: the footprint straddles land and water simultaneously. Land returns are brighter and arrive at different times than water returns, corrupting the waveform. Standard ocean retracking algorithms, tuned for the clean Brown model, fail here. Dedicated coastal retracking algorithms such as ALES (Adaptive Leading Edge Subwaveform) and SAMOSA+ partially recover usable data down to 2 to 3 km from shore in favourable geometry, but this is an active research area and not all coastlines are equally tractable.
For operators interested in sea level at specific ports or lagoons, tide-gauge networks remain the reference. Altimetry is most powerful for the open ocean and for large inland water bodies where no gauge exists.
Combining altimetry with other observations for operational products
Sea surface height from Sentinel-3 SRAL feeds into ocean circulation models alongside data from Sentinel-6 MF and Jason-3. The merged, gridded sea level anomaly products distributed by the Copernicus Marine Service (CMEMS) typically combine three or more altimeters to improve spatial coverage. A single altimeter samples the ocean sparsely; the merged product fills in the mesoscale eddy field at roughly 0.25-degree resolution.
For inland water, the practical workflow runs from raw Level-1b waveforms through retracking to Level-2 water levels at virtual stations, then into hydrological models or simple time-series dashboards. Satellize applies this workflow to client-specific river basins, integrating altimetric levels with optical flood-extent layers where the monitoring question demands it. The Tonga crop-estimation programme is a separate analytics engagement, but the underlying data-fusion logic is similar: open constellation data processed against a specific operational question rather than a generic product.
One underappreciated application is reservoir storage estimation. Water level from SRAL, combined with a bathymetric curve derived from historical imagery or survey data, gives volume change without any in-country infrastructure. For governments managing water security in data-sparse regions, that is a meaningful capability.
What the archive enables and what it cannot settle
Sentinel-3A has been operational since 2016, giving roughly eight years of continuous altimetric record over the same ground tracks. That is long enough to detect interannual variability in lake levels driven by rainfall anomalies, or to track the slow acceleration of sea level rise at regional scale. It is not long enough on its own to separate the decadal signal from natural variability; for that, the Sentinel-3 record must be spliced onto the Jason series going back to TOPEX/Poseidon in 1992.
The physics of radar altimetry also sets a hard floor on what can be detected. Ice sheets, wetlands with emergent vegetation and urban water bodies all produce waveforms that current retracking cannot fully interpret. Progress is being made, but any vendor claiming centimetre-level accuracy over a reed-fringed lake is overstating the state of the art.
Typical figures
| Instrument | Sentinel-3 SRAL, dual-frequency Ku-band (13.575 GHz) and C-band (5.41 GHz) |
| Along-track resolution (SAR mode) | ~300 m |
| Across-track footprint | Several kilometres (pulse-limited in cross-track direction) |
| Repeat cycle (single satellite) | 27 days exact repeat |
| Combined revisit (Sentinel-3A + 3B) | ~13 to 14 days at any ground track crossing |
| Sea surface height accuracy (open ocean, single pass) | Approximately 2 to 4 cm RMS |
| Coastal degradation zone | Retrievals unreliable within ~10 km of shore; partial recovery to ~2–3 km with ALES/SAMOSA+ retracking |
| Minimum river width for reliable retracking | Approximately 100 to 200 m; narrower channels carry uncertainties that may exceed 50 cm |
| Archive depth | Sentinel-3A from 2016; Jason-3 from 2016; TOPEX/Poseidon heritage back to 1992 |
| Data access | Level-1b and Level-2 products via Copernicus Data Space Ecosystem; near-real-time and non-time-critical streams available |
Analytics Satellize can run
| River virtual station time series | Waveform retracking (OCOG or Ice-1) applied to Level-1b SRAL data at user-defined crossing points | CSV or GeoJSON time series of water level anomaly at named gauging locations, updated each satellite pass |
| Reservoir storage change estimate | SRAL-derived level combined with client-supplied or publicly available hypsometric curve to compute volume change | Monthly storage anomaly report with uncertainty bounds, delivered as PDF and structured data feed |
| Regional sea level anomaly map | Along-track Level-2 sea surface height anomaly gridded and merged with CMEMS multi-mission product | NetCDF or GeoTIFF layer at 0.25-degree resolution, suitable for ingestion into ocean model boundary conditions |
| Significant wave height climatology | Statistical aggregation of SRAL Level-2 SWH retrievals over user-defined region and time window | Seasonal and annual percentile maps for offshore infrastructure planning, delivered as GIS layers |
| Flood-peak level reconstruction | SRAL virtual station levels cross-referenced with Sentinel-1 SAR flood extent to reconstruct inundation depth at crossing points | Event report combining level time series and flood extent polygons, with caveats on revisit gaps |
| Long-term lake level trend | Linear trend fitting over multi-year SRAL virtual station archive, with seasonal decomposition | Annual trend summary (mm/year with 95% confidence interval) and interactive time-series chart |
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.