Sea-ice thickness from radar and laser altimetry
Sea-ice thickness cannot be measured directly from orbit, so satellites measure freeboard and work backwards through isostatic balance. The method is powerful but depends critically on snow-load assumptions that vary across ice types and seasons.
Sensors
- CryoSat-2 SIRAL (ESA, launched 2010): Ku-band (13.575 GHz) radar altimeter operating in SAR Interferometric mode over sea ice. Along-track resolution approximately 300 m, across-track roughly 1.5 km in SARIn mode. Radar pulse penetrates into the snow layer, so the dominant scattering horizon sits somewhere between the snow surface and the snow-ice interface, introducing a systematic bias that must be corrected. Repeat cycle 369 days, but sub-cycle sampling gives monthly Arctic basin composites at useful density.
- ICESat-2 ATLAS (NASA, launched 2018): Photon-counting lidar at 532 nm. Along-track sampling at 0.7 m intervals, with effective surface-height precision of a few centimetres over flat ice. Measures the snow surface directly, complementing radar. Three pairs of beams give six ground tracks per pass. Repeat cycle 91 days; polar orbits provide dense coverage above 70° latitude. Cloud cover is the principal operational constraint.
- Sentinel-3 SRAL (ESA/EUMETSAT, launched 2016 and 2018): Ku- and C-band dual-frequency radar altimeter, primarily designed for ocean topography. Ku-band freeboard retrievals over sea ice are feasible but spatial resolution (roughly 300 m along-track in SAR mode) and the orbit's 27-day exact repeat limit thickness mapping to coarser spatial scales than CryoSat-2. Useful for cross-calibration and extending the time series.
- ENVISAT RA-2 (ESA, 2002–2012, archive): Ku- and S-band altimeter providing the pre-CryoSat archive. Spatial resolution coarser than SIRAL (footprint roughly 2–10 km depending on surface), but the decade-long record is valuable for trend analysis. S-band data allowed some snow-penetration correction. Now archived; no new acquisitions.
What freeboard actually tells you, and what it does not
Sea ice floats in isostatic balance. Roughly 90 per cent of an ice floe sits below the waterline; the remaining fraction, the freeboard, protrudes above it. Measure freeboard accurately and you can infer total thickness, provided you know the densities of ice, snow, and seawater. The governing equation is straightforward: thickness equals freeboard multiplied by seawater density, divided by the difference between seawater and ice densities, with a snow-load correction term added. The maths is clean. The inputs are not.
The density of sea ice varies with brine content, temperature, and age. First-year ice typically ranges from about 900 to 920 kg/m³; multi-year ice, which has lost much of its brine, can fall to 880 kg/m³ or lower. A 20 kg/m³ error in assumed ice density propagates to a thickness error of roughly 0.2–0.4 m for typical Arctic ice. That matters when basin-mean thickness might be 1.5 m in a thin-ice year. Most operational products use climatological density values with acknowledged uncertainty bands rather than spatially resolved density fields, because the latter simply do not exist at basin scale.
The snow problem: why two wavelengths are better than one
Snow sitting on top of a floe adds weight, pushing the ice deeper and reducing the freeboard a satellite observes. Ignore the snow load and you underestimate thickness. The Warren climatology, derived from Soviet drifting-station data collected between the 1950s and 1980s, remains widely used for Arctic snow depth and density despite its age and sparse sampling. It systematically overestimates snow depth on first-year ice, which now dominates much of the Arctic following the multi-year ice loss of recent decades. This is a known, documented bias in operational CryoSat-2 products.
The two sensor types also see snow differently. ICESat-2's 532 nm laser reflects from the top of the snow surface. CryoSat-2's Ku-band radar penetrates into the snowpack, with the dominant backscatter horizon typically somewhere near the snow-ice interface on cold, dry snow but shifting upward on wet or layered snow. The vertical offset between these two reference surfaces, sometimes called the radar-laser offset, can be 0.1–0.5 m depending on snow properties. Combining ICESat-2 and CryoSat-2 acquisitions over coincident or near-coincident areas allows researchers to estimate this offset and thereby constrain snow depth independently, a technique demonstrated in published work from the Alfred Wegener Institute and NASA teams. It is promising but not yet operationally routine.
First-year ice versus multi-year ice: different animals, different errors
Multi-year ice has survived at least one melt season. It is thicker, less saline, and structurally more complex, with hummocks and ridges that create freeboard variability at scales smaller than any altimeter footprint. Radar altimeters return a waveform that represents the aggregate of many surface facets within the footprint; retracking algorithms that work well on flat first-year ice can misplace the range on rough multi-year ice by tens of centimetres.
First-year ice is thinner and flatter, which makes waveform retracking more reliable. But it is also more likely to be flooded at the snow-ice interface in high-snow years, turning the freeboard reference ambiguous. The practical consequence is that thickness uncertainty is not uniform across the ice pack. Published estimates from ESA's CryoSat-2 sea-ice products quote basin-mean monthly uncertainties of roughly 0.3–0.5 m for individual grid cells, with larger errors over deformed ice. That is honest and should inform any application that needs local rather than regional accuracy.
Temporal coverage and the seasonal blind spot
Radar altimetry over sea ice works best in winter, when the snowpack is cold and dry and the dominant backscatter horizon is stable near the snow-ice interface. From roughly May to September, surface melt creates wet snow and melt ponds that scatter radar energy unpredictably, making freeboard retrievals unreliable or impossible. The operational CryoSat-2 Arctic thickness product therefore covers October to April. ICESat-2 lidar is less affected by surface wetness per se, but melt ponds and specular reflections from open water can saturate the detector or introduce noise.
This seasonal gap is not a minor inconvenience. Ice is thinnest at the end of summer, precisely when altimetry is least reliable. Volume estimates at the September minimum rely on extrapolation from spring observations combined with thermodynamic models, adding another layer of uncertainty to what is already an indirect measurement.
What the data can and cannot support in practice
Basin-scale monthly thickness maps are the core operational product, and they are genuinely useful for climate monitoring, shipping-season planning, and ice-mass-balance tracking. The ESA CryoSat-2 sea-ice thickness dataset extends back to 2010; combined with the ENVISAT RA-2 archive, a radar-altimetry record reaching back to 2002 exists, though inter-mission calibration adds uncertainty to long-term trends.
Local or sub-weekly thickness estimates are a different matter. A single CryoSat-2 pass samples a swath; filling in a regional grid to useful density takes weeks of accumulation. For route-specific ice-condition assessment at the scale of a ship corridor, altimetry thickness provides context rather than operational precision. Thickness products are best used alongside ice-concentration and ice-type maps derived from passive microwave or SAR imagery, not as a standalone source. Satellize integrates CryoSat-2 and ICESat-2 derived thickness grids with Sentinel-1 SAR ice-type classification in its polar analytics workflows, applying the same open-method rigour it uses in programmes such as the Tonga crop-estimation work.
Anyone commissioning thickness analytics should ask specifically which snow climatology is used, how multi-year ice fraction is handled, and what the stated per-cell uncertainty is. These are not pedantic questions. They determine whether the product is fit for a given decision.
Typical figures
| CryoSat-2 SIRAL along-track resolution (SARIn mode) | ~300 m |
| ICESat-2 ATLAS along-track sampling interval | 0.7 m; effective height precision ~2–3 cm over flat ice |
| CryoSat-2 repeat cycle | 369 days exact; monthly basin composites achievable via sub-cycle sampling |
| ICESat-2 repeat cycle | 91 days |
| Operational thickness retrieval season (radar) | October to April (winter only; melt season excluded) |
| Typical basin-mean monthly thickness uncertainty (CryoSat-2) | 0.3–0.5 m per grid cell; larger over deformed multi-year ice |
| Radar frequency (CryoSat-2 / Sentinel-3) | Ku-band 13.575 GHz; Sentinel-3 adds C-band 5.41 GHz |
| Archive depth | ENVISAT RA-2 from 2002; CryoSat-2 from 2010; ICESat-2 from 2018 |
| Polar coverage limit | CryoSat-2 to 88°N/S; ICESat-2 to 88°N/S |
| Primary operational data formats | NetCDF (ESA, NASA standard products); GeoTIFF grids available via processing |
Analytics Satellize can run
| Monthly Arctic or Antarctic thickness grid | CryoSat-2 freeboard retracking with TFMRA50 or similar threshold retracker; isostatic balance with Warren or modified snow climatology | GeoTIFF or NetCDF thickness grid at 25 km EASE-Grid resolution, with per-cell uncertainty layer |
| Snow-depth estimate from radar-laser offset | Differencing of ICESat-2 surface elevation and CryoSat-2 retracked range over coincident or temporally close acquisitions | Gridded snow-depth field with confidence flags, delivered as GIS layer |
| Multi-year versus first-year ice thickness separation | Ice-type classification from Sentinel-1 SAR or passive microwave applied as mask; separate density assumptions per class in isostatic model | Stratified thickness report with separate uncertainty estimates per ice type |
| Interannual volume anomaly trend | Time-series integration of monthly thickness grids over ice-extent mask; linear trend fitting with breakpoint analysis across ENVISAT/CryoSat-2 record | Annual trend report with basin-mean volume time series chart and statistical confidence intervals |
| Corridor-specific thickness profile for route planning | Along-track extraction of CryoSat-2 passes within defined shipping corridor; aggregated to weekly mean with ice-type context from SAR | Weekly PDF or GeoJSON corridor report showing thickness distribution and proportion of ice exceeding defined thickness thresholds |
| Seasonal forecast input layer | End-of-winter (April) thickness map ingested as initialisation field for thermodynamic sea-ice model; uncertainty propagation documented | Gridded spring-thickness initialisation file with metadata for model ingestion |
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.