Spaceborne lidar ice sheet and terrain elevation change detection
Repeat-track spaceborne lidar measures surface elevation change over ice sheets, glaciers and bare terrain by differencing profiles separated in time, revealing mass gain or loss without SAR phase ambiguity. Honest limits apply: cross-track separation, slope geometry and waveform saturation over fresh snow all require careful correction.
Sensors
- ICESat-2 ATLAS (NASA, 2018–present): Photon-counting lidar at 532 nm, six beams in three pairs, 91-day repeat cycle, along-track shot spacing ~0.7 m, footprint ~17 m diameter. Produces ATL06 land-ice surface heights with reported single-pass precision of ~3 cm over flat ice. Cross-track beam separation of ~3.3 km between pairs means exact repeat is rarely achieved over sloped terrain.
- GEDI (NASA/ISS, 2019–present): Full-waveform lidar at 1064 nm, eight beams, ~25 m footprint, non-repeat ISS orbit covering 51.6° latitude. Gridded L3 elevation products at 1 km. Coverage excludes polar ice sheets above ~51.6° N/S; most useful for mid-latitude glaciers and high-altitude terrain. Waveform saturation occurs over highly reflective fresh snow.
- CryoSat-2 SIRAL (ESA, 2010–present): Ku-band radar altimeter with SARIn mode over ice-sheet margins, ~300 m across-track resolution in SARIn, continuous coverage to 88° latitude, 369-day exact repeat. Complements lidar where cloud is persistent and photon-counting is interrupted; penetrates dry snow by a few centimetres, introducing a bias that lidar avoids.
- ArcticDEM / REMA (stereo optical, PGC): High-resolution digital elevation models derived from commercial stereo optical imagery (WorldView, GeoEye) by the Polar Geospatial Center. ArcticDEM covers the Arctic at 2 m posting; REMA covers Antarctica at 2–8 m. Temporal coverage is opportunistic rather than systematic, but differencing multi-epoch strips reveals elevation change at metre-scale vertical precision over stable terrain.
Why lidar, not radar, for elevation differencing
Synthetic aperture radar interferometry measures surface displacement along the line of sight with millimetre-scale precision, but over ice sheets it accumulates phase ambiguity wherever the surface decorrelates between passes, and the radar signal penetrates dry firn by a variable depth that shifts with temperature and density. That penetration bias can exceed half a metre at Ku-band, making it difficult to separate true surface lowering from a change in the dielectric properties of the snowpack.
Lidar carries no such ambiguity. A photon reflected from the ice surface returns at a time that encodes distance directly. ICESat-2's ATLAS instrument sends 10,000 pulses per second at 532 nm and records individual photon arrival times, giving along-track elevation profiles with a single-pass precision of roughly 3 cm over flat, bright ice. Differencing two profiles from different epochs over the same ground location gives a clean elevation-change signal, provided the two tracks actually cross the same point. That proviso is the central practical problem.
The cross-track separation problem and how slope correction handles it
ICESat-2 has a 91-day exact-repeat cycle, but 'exact' is nominal. The actual ground track repeats within a corridor roughly 100 m wide, so two passes over a glacier separated by one year may be laterally offset by tens of metres. On a flat ice sheet this barely matters. On an outlet glacier sloping at 1–5°, a 50 m cross-track offset translates to an apparent elevation difference of 0.9–4.4 m that has nothing to do with mass change.
The standard correction uses a reference DEM (often ArcticDEM or REMA) to compute the local surface slope and aspect, then adjusts each elevation observation to a common reference track before differencing. The correction is well-established in the literature and implemented in the NASA ATL06 processing chain, but it propagates DEM error into the result. Where the reference DEM is coarse or outdated, the residual slope-correction error can exceed the signal being measured on small, steep glaciers. GEDI's non-repeat ISS orbit makes the problem worse: there is no repeat track to difference at all, so GEDI elevation is used for gridded climatological products rather than point-to-point change detection.
CryoSat-2 sidesteps the problem differently. Its SARIn mode uses two antenna phase centres to locate the point of closest approach across-track, giving a position fix that allows elevation retrieval even off the nadir point. The trade-off is coarser spatial resolution and the radar penetration bias noted above.
Waveform saturation over fresh snow: a limit worth naming
Fresh dry snow reflects at 532 nm with an albedo that can exceed 0.98. ICESat-2's photon-counting detector is designed for the low return rates expected over dark ocean or vegetated land; over a bright, flat snowfield it can receive so many photons per pulse that the detector enters a dead-time state and misses subsequent photons in the same pulse. This is detector saturation, and it biases the apparent surface elevation upward by a few centimetres because the first detected photon arrives slightly before the centroid of the return.
Full-waveform systems like GEDI saturate differently. The analogue waveform digitiser clips at high return power, truncating the peak and making the surface appear lower than it is. Both effects are correctable using instrument housekeeping data, but the corrections add uncertainty. Over fresh Antarctic interior snow, where returns are brightest, the saturation correction can approach the magnitude of the annual elevation-change signal on the high plateau, where accumulation rates are only a few centimetres of water equivalent per year. Analysts should flag saturated returns and treat those elevation estimates with wider error bars.
Combining ICESat-2 ATL06 with GEDI L3 and airborne campaigns
No single spaceborne lidar system covers all glaciated terrain at adequate density. ICESat-2 provides dense along-track profiles but sparse cross-track sampling; its six beams separated by 3.3 km leave large swaths of a glacier unmeasured between passes. GEDI's 25 m footprints sample mid-latitude mountain glaciers but cannot reach the Greenland or Antarctic ice sheets. The standard practice is to combine both datasets on a common grid, fill spatial gaps with kriging or plane-fitting, and then validate the gridded elevation-change field against airborne surveys.
NASA's Operation IceBridge flew airborne lidar (ATM instrument) and radar over Greenland and Antarctica from 2009 to 2019, producing a validation dataset that bridges the gap between ICESat (2003–2009) and ICESat-2 (2018–present). Airborne campaigns typically achieve vertical accuracy of 10–20 cm over rough terrain, coarser than ICESat-2 over flat ice but more spatially complete across a single glacier basin. Mass-balance estimates derived from elevation change require a density assumption to convert volume change to mass: 917 kg/m³ for solid ice loss, but something between 300 and 600 kg/m³ for firn compaction changes, and the choice of density model is often the largest source of uncertainty in published mass-balance figures.
Satellize can ingest ATL06 granules, apply slope correction against a client-specified reference DEM, and deliver gridded elevation-change maps at user-defined epochs.
What the data can and cannot tell a government buyer
A ministry managing water resources from glaciated catchments wants to know how much ice is left and how fast it is going. Spaceborne lidar gives a credible answer for large glaciers (area above roughly 1 km²) where the along-track sampling is dense enough to characterise the surface. For small cirque glaciers below a few hundred metres across, a single ICESat-2 beam may cross the glacier only once per 91-day cycle, and a single bad pass due to cloud or saturation can corrupt the seasonal signal.
Cloud is the other honest limit. Lidar at 532 nm cannot penetrate cloud. In persistently overcast regions such as the southern Andes or coastal Alaska, the fraction of cloud-free ICESat-2 passes may be low enough that multi-year averaging is needed before a trend emerges with statistical confidence. CryoSat-2 radar altimetry fills that gap at the cost of the penetration bias. A well-designed monitoring programme uses both, treating the lidar as the ground truth and the radar as the gap-filler, with explicit uncertainty budgets for each.
Typical figures
| Along-track shot spacing (ICESat-2 ATLAS) | ~0.7 m |
| Footprint diameter (ICESat-2 ATLAS) | ~17 m |
| Single-pass elevation precision over flat ice (ICESat-2) | ~3 cm (reported; degrades over rough or sloped terrain) |
| Repeat cycle (ICESat-2) | 91 days nominal; cross-track offset within ~100 m corridor |
| GEDI footprint / latitude coverage | ~25 m / 51.6° N to 51.6° S only |
| CryoSat-2 SARIn across-track resolution | ~300 m; repeat 369 days; coverage to 88° latitude |
| ArcticDEM / REMA posting | 2 m (ArcticDEM); 2–8 m (REMA); vertical accuracy ~0.2–1 m over stable terrain |
| Minimum detectable elevation change (ICESat-2, annual repeat) | ~5–10 cm over flat ice after slope correction; larger on steep glaciers |
| Archive depth | ICESat-2 from September 2018; CryoSat-2 from April 2010; ICESat (legacy) 2003–2009 |
| Delivery formats | HDF5 (ATL06 native), GeoTIFF gridded change maps, CSV elevation profiles, NetCDF time series |
Analytics Satellize can run
| Glacier-basin elevation-change map | ICESat-2 ATL06 repeat-track differencing with slope correction against ArcticDEM or REMA reference DEM | GeoTIFF raster of dh/dt (m/year) per epoch pair, with per-pixel uncertainty layer |
| Volume and mass-balance estimate | Spatial integration of gridded dh/dt, density-assumption model (ice vs. firn column), error propagation | Tabular report of Gt/year or km³/year with confidence intervals, updated per 91-day ICESat-2 cycle |
| Cross-track offset and slope-correction diagnostic | Beam-pair geometry analysis, local slope extraction from reference DEM, residual error quantification | QA report flagging passes where slope-correction uncertainty exceeds a client-defined threshold |
| Saturation-flagged elevation time series | ATL06 quality flag parsing, dead-time correction application, outlier removal | Cleaned CSV time series of surface elevation per beam per epoch, with saturation-event log |
| Multi-sensor blended elevation-change grid | Co-registration of ICESat-2 profiles and CryoSat-2 SARIn swaths, kriging to common 500 m grid, radar penetration bias correction | NetCDF gridded product covering cloud-gap periods, with separate uncertainty fields for lidar and radar contributions |
| Airborne validation comparison report | Co-location of ATL06 points with client-supplied or NASA IceBridge ATM profiles, bias and RMSE statistics | PDF validation report with scatter plots, spatial residual maps and recommended bias corrections |
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.