Ice-sheet surface elevation change from photon-counting lidar
ICESat-2's photon-counting lidar resolves ice-surface elevation change to a few centimetres per year, giving glaciologists and policymakers the most precise public record of ice-sheet geometry available from orbit.
Sensors
- ICESat-2 ATLAS (Advanced Topographic Laser Altimeter System): Single-photon counting lidar at 532 nm; six beams in three pairs; 91-day exact repeat orbit; along-track photon spacing approximately 0.7 m; cross-track beam-pair separation roughly 3.3 km; elevation precision on flat ice better than 3 cm for 20 m segments (ATL06 product)
- CryoSat-2 SIRAL (Synthetic Aperture Interferometric Radar Altimeter): Ku-band radar altimeter; SARIn mode over ice-sheet margins; 369-day repeat; footprint roughly 300 m in SARIn mode; penetrates cloud but also penetrates dry snow by up to a metre, introducing a systematic bias that must be corrected before cross-validation with lidar
- TanDEM-X global DEM: X-band SAR interferometry; 12 m posting; used as a reference surface to correct slope-induced artefacts when ICESat-2 repeat tracks do not overlap exactly; static reference only, not a change-detection sensor
- Sentinel-1 SAR (C-band): 6-day repeat at high latitudes in interferometric wide-swath mode; provides ice-flow velocity fields via offset tracking or InSAR; velocity context is essential for separating dynamic thinning from surface mass balance changes
What a photon counter sees that a radar altimeter cannot
Radar altimeters such as CryoSat-2 SIRAL have measured ice-sheet elevation for decades, but they carry a persistent ambiguity: Ku-band pulses penetrate dry firn by tens of centimetres to over a metre, depending on snow density and temperature. When the snowpack warms or densifies, the radar scattering horizon shifts vertically even if the physical surface has not moved. Separating real surface change from changing penetration depth requires ancillary firn models, and those models carry their own uncertainties.
ICESat-2 ATLAS sidesteps this entirely. At 532 nm, photons scatter from the very top of the snow surface. There is no penetration ambiguity. The instrument fires 10,000 pulses per second and counts individual returning photons through a timing detector sensitive to single-photon arrivals, achieving along-track photon spacing of roughly 0.7 m. The ATL06 land-ice product aggregates these into 20 m segments with reported elevation precision better than 3 cm on smooth, flat ice. That figure degrades on rough glacier margins and in the presence of blowing snow or thin cloud, but on the interior ice sheets it is genuinely centimetre-class.
The repeat-track problem: slopes are not the same as sinking
ICESat-2 returns to within roughly 100 m of a prior ground track after 91 days. On a perfectly flat surface that is fine. On a sloping glacier, a 100 m cross-track offset translates directly into a spurious elevation difference. A 1° surface slope and a 100 m cross-track miss combine to produce a roughly 1.7 m apparent elevation change that has nothing to do with ice loss.
The standard correction uses a reference DEM, typically TanDEM-X at 12 m posting, to compute the slope at each measurement point and project both passes onto a common reference plane. This repeat-track analysis is described in the ATL11 product documentation and is the method that makes multi-year elevation-change time series credible. It does not eliminate the problem entirely: where surface slopes change rapidly, as at crevasse fields or calving fronts, the DEM-derived slope may not represent local conditions, and residual errors of several centimetres remain possible.
Cloud cover at high latitudes: the honest limitation
Photons at 532 nm do not pass through cloud. In polar regions, cloud cover is frequent and seasonally variable. Over the Antarctic interior, clear-sky fractions can exceed 70 % in summer, making ATLAS highly productive. Over the Greenland margins and the Antarctic Peninsula, cloud cover is heavier and more persistent, and a significant fraction of 91-day repeat passes return no usable surface returns at all.
The practical consequence is that the temporal sampling of any given ground track is irregular. A location might have clear returns in three consecutive 91-day cycles and then nothing for two cycles. Trend estimation therefore requires multi-year stacking rather than cycle-to-cycle differencing, and the ATL11 product is specifically designed for this: it accumulates repeat-track pairs across the mission lifetime and fits elevation-change rates with associated uncertainties. For users who need sub-annual signals, cloud gaps are a genuine constraint that no processing choice can remove.
Cross-track gaps and what falls between the beams
ATLAS has six beams arranged in three pairs. Within a pair, beams are separated by about 90 m. Between pairs, the separation is roughly 3.3 km. Between adjacent ground tracks at 91-day repeat, the cross-track spacing at mid-latitudes is several kilometres, closing toward the poles as tracks converge. The result is that ICESat-2 samples the ice sheet along lines, not as a continuous surface.
For the interior of the Greenland and Antarctic ice sheets, where surface topography is smooth and elevation changes are spatially coherent over tens of kilometres, this sampling is adequate for basin-scale mass-balance inference. At glacier outlets, where thinning can be highly localised, a glacier tongue might fall entirely between beam pairs on every pass. Combining ICESat-2 profiles with CryoSat-2's broader swath coverage is the accepted approach for closing these gaps, despite the penetration-depth caveats that CryoSat-2 carries.
From elevation change to mass balance: what the numbers mean
Elevation change is not mass change. Converting dh/dt to dm/dt requires knowing whether the volume change is ice (density roughly 917 kg/m³) or firn compaction (density 400 to 600 kg/m³). A metre of surface lowering driven by firn compaction after a warm summer represents far less mass loss than a metre of dynamic thinning at a calving glacier. This density ambiguity is the largest source of uncertainty in lidar-derived mass-balance estimates and is typically addressed by coupling elevation-change maps with firn compaction models such as IMAU-FDM or GSFC-FDM.
Sentinel-1 velocity fields enter here as a diagnostic. Rapid acceleration of an outlet glacier, visible in Sentinel-1 offset-tracking products at 6-day repeat, signals dynamic thinning rather than surface mass balance change. That distinction matters for projecting future sea-level contribution, because dynamic instabilities can persist and accelerate whereas surface mass balance anomalies tend to be reversible on decadal timescales.
Satellize runs repeat-track elevation-change analysis on ATL11 data as a standard analytics layer, applying the same slope-correction and firn-density workflow used in published cryosphere studies. The methodology is the same one underpinning our Tonga crop-estimation programme's philosophy: open-data rigour, no proprietary black boxes.
What a buyer actually receives, and what they should not expect
A well-configured ICESat-2 analytics pipeline delivers gridded elevation-change-rate maps at roughly 500 m posting for smooth interior ice, with per-cell uncertainty estimates, updated on each 91-day cycle as new ATL11 data are released by NASA. Outlet glacier profiles can be extracted at 20 m along-track resolution. Anomaly alerts are straightforward: flag any grid cell where the current-cycle elevation change exceeds the multi-year trend by more than two standard deviations.
What the data cannot deliver: wall-to-wall coverage at high temporal frequency, sub-centimetre precision on rough terrain, or any signal through cloud. For users whose primary interest is a single fast-changing outlet glacier that happens to sit between beam pairs, the honest answer is that ICESat-2 alone may not be sufficient and a CryoSat-2 or airborne lidar campaign should be considered alongside it. Requests for a specific glacier assessment are the right starting point.
Typical figures
| Along-track photon spacing | ~0.7 m (ATLAS, 10 kHz pulse rate) |
| ATL06 segment length | 20 m (land-ice elevation product) |
| Elevation precision (flat ice, ATL06) | <3 cm on smooth interior ice; degrades to several decimetres on rough margins |
| Repeat cycle | 91 days (exact repeat); 91-day sub-cycle data also available |
| Cross-track beam-pair separation | ~3.3 km between pairs; ~90 m within a pair |
| Wavelength | 532 nm (green, photon-counting; no firn penetration) |
| Cloud limitation | No surface returns through cloud; clear-sky fraction varies 30–70 % depending on region and season |
| Archive depth | ICESat-2 operational from October 2018; ATL11 multi-cycle products from 2019 onward |
| Primary products | ATL06 (land-ice elevation, 20 m segments); ATL11 (repeat-track time series) |
| Delivery formats | HDF5 (NASA standard); derivative GeoTIFF and GeoPackage on request |
Analytics Satellize can run
| Basin-scale elevation-change-rate map | Repeat-track analysis on ATL11, slope correction via TanDEM-X reference DEM, multi-cycle least-squares trend fitting | GeoTIFF grid at ~500 m posting, updated per 91-day cycle, with per-cell dh/dt and 1-sigma uncertainty |
| Outlet-glacier thinning profile | ATL06 along-track extraction, cross-track offset correction, firn-compaction model overlay (GSFC-FDM or IMAU-FDM) | CSV and GIS polyline with elevation, elevation change, and density-corrected mass-change estimate per 20 m segment |
| Elevation-change anomaly alert | Cycle-to-cycle deviation from multi-year ATL11 trend; threshold set at ±2 standard deviations | Automated alert report (PDF + GeoJSON) flagging anomalous cells within 72 hours of NASA data release |
| Dynamic vs. surface-mass-balance thinning attribution | Sentinel-1 offset-tracking velocity fields co-registered with ATL06 elevation-change profiles; flux-divergence analysis | Annotated map distinguishing dynamic thinning zones from firn-compaction-driven lowering, with supporting velocity vectors |
| CryoSat-2 cross-validation layer | Co-located CryoSat-2 SIRAL SARIn elevations differenced against ATL06; penetration-depth bias estimated from residuals | Bias-correction table and spatial map of radar–lidar elevation offsets, useful for extending the record pre-2018 |
| Multi-decadal elevation-change context report | ICESat-2 ATL11 trends spliced with ICESat (2003–2009) and ERS/Envisat radar altimetry literature values for historical framing | Narrative report with time-series figures situating current change rates within the longer instrumental record |
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.