Ice-covered polar ocean bathymetry from satellite gravity inversion
Satellite radar altimetry over open-water leads measures marine gravity anomalies that, when inverted, reveal seafloor topography beneath Arctic and Southern Ocean ice at roughly 5 km resolution, filling gaps that echo-sounding ships may never close.
Sensors
- CryoSat-2 SIRAL: Synthetic-aperture radar altimeter operating at Ku-band (13.575 GHz). Over open leads it achieves sea-surface height precision of roughly 2–3 cm per pass. Its 369-day exact-repeat cycle and drifting sub-cycles give dense ground-track coverage poleward of 88°N, which no other operational altimeter reaches. SAR and SARIn modes resolve along-track surface features to approximately 300 m, though the gravity-inversion product is limited to ~5 km by the inversion physics, not the altimeter footprint.
- ICESat-2 ATL12 (ocean-surface height product): Photon-counting lidar at 532 nm. ATL12 retrieves sea-surface height in open leads with centimetre-level single-pass precision where returns are clean. Provides an independent, cross-calibration check on radar-altimeter gravity fields, particularly valuable in the marginal ice zone where Ku-band waveform retracking is ambiguous. Coverage extends to 88° latitude; cloud cover is the principal data-loss mechanism.
- Geosat geodetic mission archive (1985–1986): U.S. Navy altimeter mission whose declassified geodetic-mission data, released in 1995, provided the first dense global gravity grid. Ground-track spacing of roughly 4 km at mid-latitudes gave gravity anomaly resolution adequate to resolve bathymetric features above ~10 km scale. Polar coverage is limited to approximately 72°N/S by orbital inclination, so Arctic and Antarctic shelves require supplementation from higher-inclination missions.
- ERS-1 geodetic mission archive (1994–1995): ESA's 168-day geodetic phase produced a complementary gravity grid with ~8 km track spacing at the equator and coverage to 81.5° latitude, extending useful gravity data well into the Arctic and toward the Antarctic coast. Combined with Geosat, the two archives underpin the SIO V32 and DTU global gravity models that most published bathymetric inversions use as their starting point.
- Sentinel-3A/B SRAL: ESA Ku/C dual-frequency radar altimeter continuing the long-term sea-surface height record in SAR mode. Revisit of roughly 27 days per satellite (combined ~14 days). Adds contemporary passes to the gravity stack and helps detect interannual sea-surface variability that must be removed before gravity inversion. Coverage to 81.35° latitude.
What a gravity anomaly tells you about the seafloor
The principle is old and the physics is unambiguous. Seafloor topography displaces mantle and crustal rock relative to water. That density contrast produces a measurable perturbation in the local gravitational field, which in turn deforms the ocean's equipotential surface, the geoid, by a few centimetres per kilometre of relief. A radar altimeter measuring sea-surface height over a calm open-water lead is, in effect, reading that geoid undulation. Differentiate twice and you have the gravity anomaly; invert the anomaly with an assumed density contrast and you recover an estimate of the depth to the seafloor.
The density contrast between seawater (roughly 1,025 kg/m³) and oceanic crust (roughly 2,700–2,900 kg/m³) is well constrained, which is why the method works at all. The limiting factor is not the physics but the spatial resolution of the gravity field itself. Gravity anomalies attenuate with depth: a feature 4 km below the surface produces a gravity signal spread over a horizontal scale several times its own width. In practice, bathymetric features smaller than about 5–6 km horizontal extent are unresolvable from satellite gravity at abyssal depths, regardless of how precisely the altimeter measures sea-surface height. Shallower shelves allow slightly finer resolution; deeper basins do not.
Why leads matter more than open ocean for polar gravity
Over sea ice, a radar altimeter cannot measure the sea surface. It measures the ice surface, which carries no geoid signal useful for gravity inversion. The method therefore depends entirely on open-water leads: the fractures, polynyas and seasonal melt zones where the ocean is briefly exposed. In the central Arctic, leads occupy perhaps 1–5% of the surface in winter. That is enough, given a sufficiently dense archive of altimeter passes accumulated over years, to reconstruct a continuous gravity field by interpolation and along-track filtering.
CryoSat-2's SAR mode is particularly effective here because its Doppler processing sharpens the along-track footprint to roughly 300 m, allowing it to lock onto narrow leads that a conventional pulse-limited altimeter would miss or contaminate with off-nadir ice returns. The waveform shape over a lead is also distinctive, a sharp specular return compared with the diffuse return over rough ice, so automated lead detection and retracking algorithms can select valid sea-surface measurements with reasonable reliability. ICESat-2 adds an independent check: its 532 nm photon-counting beam is insensitive to the dielectric ambiguity that complicates Ku-band retracking over thin new ice.
From gravity anomaly to depth grid: the inversion pipeline
The standard published workflow, described in detail by Sandwell, Müller, Smith and colleagues in their SIO global gravity series, proceeds in several steps. First, along-track sea-surface heights from multiple missions and multiple years are cross-calibrated, de-trended for ocean variability and differenced along-track to produce along-track gravity gradients. These are then gridded, typically at 1 arc-minute (roughly 1.8 km), using minimum-curvature or spline methods. The resulting free-air gravity anomaly grid is the core input.
Inversion to depth uses a Parker-Oldenburg iterative scheme or its spectral equivalent. A reference depth is assumed (often taken from GEBCO ship-track data where it exists), and the inversion iterates until the predicted gravity field matches the observed anomaly within a noise threshold. The output is a bathymetric grid at effectively 5–6 km resolution in data-sparse polar regions, degrading gracefully where gravity track density is lower. Published models such as GEBCO_2023 incorporate these satellite-derived estimates explicitly for the roughly 20% of the global ocean floor that has never been directly sounded.
What the method cannot see, and why that matters operationally
Honest accounting of the limits is necessary if the product is to be used correctly. The 5 km resolution floor means that seamounts narrower than that are invisible, or appear as smeared, attenuated bumps. Sediment-filled basins present a different problem: thick soft sediment has a density close to water, so a sediment-draped canyon may produce almost no gravity signal despite being hundreds of metres deep. The inversion assumes a single density contrast; where geology departs from that assumption, depths can be wrong by hundreds of metres.
Polar-specific complications include dynamic ocean topography from under-ice currents, which can reach 20–30 cm and masquerade as geoid signal if not corrected using ocean circulation models. Lead availability is seasonal and spatially uneven: the Beaufort Sea and parts of the East Siberian shelf have fewer leads than the Transpolar Drift corridor, so gravity track density is lower there and inversion errors are larger. For submarine-ridge identification, tidal-channel routing and deep-water formation modelling, a 5 km grid is often sufficient. For mine-countermeasure surveys or submarine navigation at fine scale, it is not.
Applications that depend on getting the depth right
Arctic bathymetry is not an academic curiosity. The Lomonosov and Gakkel ridges constrain the exchange of Atlantic Water between sub-basins and control where cold, dense bottom water forms and flows. Errors in ridge height of even 200–300 m propagate into ocean circulation models as significant biases in heat transport, which in turn affect sea-ice forecasts on decadal timescales. Tidal dissipation models for the Arctic are similarly sensitive: tidal energy is concentrated at shelf breaks and ridge crests, and a misplaced ridge shifts the dissipation pattern in ways that affect both tidal prediction and internal-wave mixing estimates.
For the Southern Ocean, accurate bathymetry beneath the Filchner-Ronne and Ross ice shelves is directly relevant to basal melt modelling. Warm Circumpolar Deep Water intrudes along specific bathymetric channels; if those channels are mislocated or their depths are wrong, melt-rate projections for the West Antarctic Ice Sheet carry corresponding uncertainty. Governments investing in long-range climate projections or Arctic sovereignty claims need the best available bathymetric grid, and satellite gravity inversion is, for most of the polar ocean, the only method that can deliver one at continental scale.
Satellize incorporates published gravity-derived bathymetric grids, including GEBCO and SIO products, as baseline layers in sovereign-programme analytics packages, combining them with current CryoSat-2 and ICESat-2 lead-height time series to flag regions where new data density is sufficient to warrant a local inversion update.
What a client actually receives
The deliverable is not a raw altimeter archive. It is a depth grid in a specified projection and resolution, accompanied by an uncertainty layer that maps estimated inversion error as a function of gravity track density and assumed geological variance. Where ship-track soundings exist in the GEBCO or IBCAO databases, those are used to anchor and validate the satellite-derived surface; the uncertainty layer reflects the distance from the nearest sounding.
Clients with specific corridor interests, a proposed Arctic shipping lane, a submarine cable route, a resource-licence boundary, can receive a targeted inversion run over their area of interest using the latest available altimeter data, with the uncertainty quantified honestly rather than papered over. The update cycle is constrained by CryoSat-2's 369-day exact-repeat and the accumulation of drifting sub-cycle passes: meaningful improvement in a specific region typically requires one to three years of additional data, a timeline that should be stated clearly at the outset.
Typical figures
| Effective bathymetric resolution | ~5–6 km horizontal at abyssal depths; ~3–4 km on shallow shelves where gravity signal is stronger |
| Depth uncertainty (open ocean, sparse sounding) | ±100–300 m typical; ±500 m or more in sediment-filled basins or low-track-density regions |
| Gravity anomaly grid resolution (input) | 1 arc-minute (~1.8 km) for SIO/DTU global products; polar regions may be coarser where lead coverage is thin |
| Altimeter sea-surface height precision (per pass, over leads) | ~2–3 cm for CryoSat-2 SIRAL SAR mode; ~1–2 cm for ICESat-2 ATL12 in clear conditions |
| Polar coverage ceiling | 88°N/S for CryoSat-2; 88° for ICESat-2; ~81.5° for ERS-1 archive; ~72° for Geosat archive |
| CryoSat-2 exact-repeat cycle | 369 days; drifting sub-cycles provide denser but irregular passes between exact repeats |
| Archive depth (geodetic missions) | Geosat geodetic phase 1985–1986; ERS-1 geodetic phase 1994–1995; both declassified/released and publicly archived |
| Minimum detectable feature (relief) | ~200–300 m vertical relief for features wider than 10 km; smaller features attenuate below noise floor |
| Delivery format | GeoTIFF depth grid + NetCDF uncertainty layer; ESRI-compatible shapefile for ridge/trough centrelines on request |
| Baseline global product | GEBCO_2023 (incorporates SIO V32 satellite gravity inversion for unsounded areas) |
Analytics Satellize can run
| Polar bathymetric depth grid with uncertainty layer | Parker-Oldenburg spectral inversion of free-air gravity anomalies derived from multi-mission altimeter sea-surface heights over leads | GeoTIFF depth grid at 1 arc-minute, paired NetCDF uncertainty raster, delivered per area-of-interest polygon |
| Gravity track density assessment | Spatial binning of CryoSat-2 and Sentinel-3 lead-detection passes over a defined epoch; comparison against minimum density threshold for inversion convergence | PDF report with maps showing track coverage, lead-hit rate by season and estimated wait time for density improvement |
| Ridge and seamount identification layer | Automated positive-anomaly extraction from gravity grid followed by spectral depth-conversion; cross-referenced against IBCAO and GEBCO ship-track records | GIS polygon layer of candidate ridge crests and seamount summits with confidence scores and estimated height above surrounding seafloor |
| Bathymetric channel routing for deep-water intrusion modelling | Least-cost path analysis on depth grid to identify lowest-sill pathways connecting shelf break to sub-ice-shelf cavities or inter-basin passages | Shapefile of candidate intrusion corridors with sill depths and uncertainty bounds, formatted for import into MOM6 or NEMO ocean model grids |
| Inversion update triggered by new lead-altimetry accumulation | Monitoring of CryoSat-2 drifting sub-cycle pass count within client-defined bounding box; re-inversion when track density crosses a pre-agreed threshold | Revised depth grid with change layer showing cells where depth estimate shifted by more than one sigma, delivered as versioned GeoTIFF |
| Tidal dissipation sensitivity map | Barotropic tidal model (TPXO or FES) run over updated bathymetric grid; comparison of energy dissipation fields against run using GEBCO baseline to isolate bathymetry-driven differences | NetCDF dissipation-difference grid and summary table of affected tidal constituents, suitable for input to ice-ocean coupled models |
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.