Subglacial lake detection and volume change from altimetry
Active subglacial lakes beneath Antarctica and Greenland inflate and drain over months to years, deforming the ice surface by centimetres to metres. Repeat-track laser and radar altimetry from ICESat-2 and CryoSat-2 isolates these anomalies from background firn and dynamic signals.
Sensors
- ICESat-2 ATLAS (NASA, 2018–present): Photon-counting lidar at 532 nm. Six-beam configuration with 91-day exact-repeat ground track. Surface elevation precision better than 3 cm over flat ice for a single 100 m along-track segment. The repeat geometry is tight enough to detect lake-fill or drain signals of a few tens of centimetres against the background trend.
- CryoSat-2 SIRAL (ESA, 2010–present): Ku-band radar altimeter operating in SARIn mode over complex ice terrain. Footprint roughly 400 m in the along-track direction in SARIn mode. Elevation accuracy over ice sheets typically 10–30 cm depending on surface slope. Provides dense cross-track coverage at high latitudes and a 369-day sub-cycle that allows 30-day repeat at the poles.
- ERS-2 RA (ESA, 1995–2011, historical archive): Ku-band pulse-limited radar altimeter. Coarser footprint than CryoSat-2 (roughly 20 km pulse-limited), but the 16-year archive extending back to 1995 is irreplaceable for establishing long-term baseline elevation trends and identifying lakes that cycled before the modern era.
- Sentinel-6 Michael Freilich (Copernicus, 2020–present): Primarily an ocean altimeter, but its high-resolution mode provides useful ice-sheet elevation data at the margins. Not the primary instrument for subglacial lake work, though its open data policy and continuity with the Topex/Poseidon/Jason series extend the multi-decadal elevation record toward the coasts.
What a draining lake does to three kilometres of ice above it
A subglacial lake sits at the bed of an ice sheet under enormous pressure. When basal meltwater accumulates faster than it drains along the hydraulic gradient, the lake inflates and the ice surface above it rises. When it drains, the surface drops. The geometry is not subtle: documented events beneath the Antarctic ice sheet have produced surface deflections of one to nine metres over areas of tens to hundreds of square kilometres, persisting over months to years. Smaller events, detectable only with ICESat-2's centimetre-level precision, may be far more numerous.
The complication is that the ice surface is moving anyway. It thickens or thins due to accumulation, ablation, firn compaction and ice dynamics. Isolating a lake signal means fitting and removing those background trends from repeat elevation profiles. The standard approach fits a polynomial to the along-track elevation time series at each crossing point, then looks for spatially coherent residuals that match the expected footprint of a lake. The spatial coherence test is important: random altimeter noise does not organise itself into smooth, kilometres-wide bowls.
ICESat-2 and CryoSat-2 do different things well
ICESat-2's ATLAS instrument fires 10,000 laser pulses per second across six beams, returning individual photon counts that reconstruct surface elevation with a formal precision below 3 cm for 100 m segments over flat ice. That precision is what makes it sensitive to the smaller, slower lake events that earlier instruments missed entirely. The 91-day repeat means a lake that drains over six months will be sampled at least twice during the event, though catching the peak requires some luck with the timing.
CryoSat-2 trades vertical precision for spatial density. Its 369-day sub-cycle produces a much denser cross-track grid at high latitudes than ICESat-2's fixed six beams, which means it can map the planimetric shape of a lake anomaly more completely. The two instruments are genuinely complementary: ICESat-2 finds the signal, CryoSat-2 constrains the spatial extent. Several published studies have combined both to produce volume-change estimates, though those estimates carry the caveat described below.
The density problem: why volume is always an inference
Altimetry measures surface elevation change, not water volume. Converting one to the other requires knowing what is moving. If ice is flowing into the subsidence zone from the sides, the volume deficit is partly filled by ice rather than by water draining away. If the surface change reflects pure vertical displacement of a rigid ice block, the geometry is simpler. In practice, the truth lies somewhere between these end members and depends on the ice rheology, the lake geometry and the timescale of the event.
The standard assumption is that the surface elevation change equals the water volume change divided by the lake area, with a correction for ice overburden. Published estimates of water volume exchanged in major Antarctic drainage events range from roughly 0.2 to over 6 cubic kilometres per event, but those figures carry uncertainties of 20 to 50 per cent depending on the assumptions made. Any analytical product derived from altimetry alone should state those assumptions explicitly. Gravitational data from GRACE-FO can provide an independent mass-change constraint for the largest events, but its spatial resolution of roughly 300 km is too coarse to isolate individual lakes in most cases.
Where the catalogue stands, and where it has gaps
The Smith et al. (2009) catalogue, derived from ICESat GLAS data, identified 124 active subglacial lakes in Antarctica. Subsequent ICESat-2 work has extended and revised that list considerably, finding activity in regions previously thought quiescent and resolving finer-scale events. Greenland's subglacial hydrological system is less well mapped; the ice sheet is smaller, the bed is more complex topographically, and fewer repeat-track studies have targeted it systematically.
Coverage has genuine blind spots. ICESat-2's six beams sample only a narrow swath on each pass; a lake that inflates and drains between repeat cycles will be missed entirely unless it leaves a lasting surface expression. Cloud cover does not affect radar altimetry but does block the laser, so ATLAS data gaps cluster in persistently cloudy coastal regions. Interior Antarctica, with its stable, clear atmosphere, is the best-observed region. The deep interior of Greenland and the margins of both ice sheets remain less certain.
Connecting lakes to ice dynamics and sea-level risk
Subglacial lakes matter beyond glaciology for two reasons. First, they lubricate the ice bed: a sudden drainage event can accelerate ice flow downstream, which in turn affects outlet glacier discharge and, ultimately, sea-level contribution. Several documented drainage cascades in Antarctica have been linked to transient speed-up of ice streams. Second, the water itself eventually reaches the ocean, contributing a small but measurable freshwater flux that affects local oceanography.
For governments and infrastructure operators with interests in Antarctic logistics, ice-sheet stability assessments, or sea-level projections, a systematic lake-monitoring programme provides early indicators of changing basal hydrology. Satellize runs analytics on open altimetry archives to produce elevation-anomaly time series and lake-activity reports; the same analytical framework that underpins the Tonga crop-estimation programme's change-detection logic applies here, adapted to elevation rather than spectral signals. The deliverable is a structured dataset of lake locations, activity periods, estimated volume changes and associated uncertainties, updated on each new ICESat-2 and CryoSat-2 data release.
What this method cannot do
Altimetry cannot image the lake directly. It infers the lake's existence from the surface expression of water movement beneath three kilometres of ice. A lake that is full and stable produces no detectable signal. A lake that drains slowly enough to be masked by firn-compaction variability may also escape detection. The minimum detectable volume change depends on lake area and the noise floor of the instrument: for ICESat-2, events producing surface anomalies below roughly 20 cm over areas smaller than a few square kilometres are at the edge of reliable detection.
Distinguishing a subglacial lake signal from an ice-dynamic anomaly (a surge, a sticky-spot transition, a change in basal friction) requires additional evidence. Satellite radar interferometry can show whether the surface deformation pattern is consistent with a buoyant block being lifted, which is the expected signature of lake inflation. Without that corroboration, an elevation anomaly is a hypothesis, not a confirmed lake. Honest reporting of subglacial lake detections should always separate confirmed, probable and possible events.
Typical figures
| Along-track elevation precision (ICESat-2 ATLAS, flat ice) | Better than 3 cm per 100 m segment |
| Radar altimeter elevation accuracy (CryoSat-2 SARIn mode) | Typically 10–30 cm depending on surface slope |
| ICESat-2 repeat cycle | 91 days (exact repeat ground track) |
| CryoSat-2 sub-cycle at polar latitudes | 30-day approximate repeat; 369-day exact sub-cycle |
| Minimum detectable surface anomaly (ICESat-2) | ~20 cm over areas of several km², larger for noisier surfaces |
| Archive depth | ERS-2 from 1995; ICESat GLAS 2003–2009; CryoSat-2 from 2010; ICESat-2 from 2018 |
| Polar coverage | ICESat-2 to 88°N/S; CryoSat-2 to 88°N/S; ERS-2 to 82°N/S |
| Volume-change uncertainty | 20–50% depending on density and rheology assumptions |
| Delivery formats | GeoTIFF elevation anomaly grids, CSV time series, GeoJSON lake polygons, PDF activity reports |
Analytics Satellize can run
| Elevation anomaly time series per lake | Repeat-track differencing with polynomial trend removal (standard published method for ICESat/ICESat-2 subglacial lake studies) | CSV time series of surface height residuals at each lake centroid, updated per new data release |
| Active lake catalogue with confidence classification | Spatial coherence filtering of elevation residuals; confirmed/probable/possible tiering following published classification frameworks | GeoJSON polygon layer with activity status, dates of detected inflation/drainage, and confidence tier |
| Volume-change estimate with uncertainty bounds | Geometric integration of surface anomaly over lake area with stated density assumptions; uncertainty propagated from altimeter noise and rheology range | Tabular report per event with central estimate, 1-sigma bounds, and assumption log |
| Drainage event alert | Threshold exceedance on rolling elevation residual; triggered when anomaly exceeds 3-sigma of background variability over the lake footprint | Email or API alert with lake ID, estimated onset date, and magnitude |
| Multi-instrument fusion elevation grid | Cross-calibrated combination of ICESat-2 and CryoSat-2 tracks to improve spatial coverage of lake planimetric shape | GeoTIFF elevation-change grid at 500 m posting for each detected event period |
| Long-term basal hydrology baseline | Trend analysis across ERS-2, ICESat GLAS, CryoSat-2 and ICESat-2 archives to characterise multi-decadal lake activity cycles | PDF summary report with annotated elevation time series and inter-mission cross-calibration notes |
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.