Antarctic ice-shelf calving-front monitoring and basal melt
SAR imagery and laser altimetry reveal calving-front retreat, rift propagation and basal melt on Antarctic ice shelves year-round, through cloud and polar night. Together they quantify the processes that govern future sea-level rise.
Sensors
- Sentinel-1 A/B C-band SAR: 5 x 20 m resolution in Interferometric Wide Swath mode, 250 km swath, 6-day exact repeat at the equator and typically 1-3 days over Antarctica due to orbit convergence. Operates through cloud and polar night; the primary workhorse for calving-front delineation and InSAR-derived ice velocity.
- RADARSAT-2 C-band SAR: Selectable modes from 3 m (Spotlight) to 100 m (ScanSAR); the Canadian Ice Service and scientific campaigns use it for high-resolution rift mapping. Tasked commercially; not free-to-access like Sentinel-1.
- ICESat-2 ATLAS photon-counting lidar: 532 nm green laser, six beams in three pairs, 91-day exact repeat, 0.7 m along-track shot spacing. Measures surface-height change to better than 3 cm per 1 km segment under clear sky; the primary source for ice-shelf thinning rates and grounding-line migration detection.
- Landsat 8/9 OLI: 15 m panchromatic, 30 m multispectral, 16-day repeat. Useful for calving-front mapping and rift tracking during polar summer when cloud permits, and for validating SAR-derived front positions. Free archive to 1972 (Landsat 1 MSS) provides the longest optical baseline for retreat measurement.
Why ice shelves matter more than the ice above them
Antarctic ice shelves are the floating extensions of the grounded ice sheet. They do not themselves raise sea level when they melt or calve, but they act as buttresses, slowing the flow of grounded glaciers behind them. Remove the buttress and the grounded ice accelerates seaward. The collapse of the Larsen B ice shelf in 2002 was followed within months by a two-to-eight-fold speed increase in the glaciers it had restrained. That is the mechanism that connects ice-shelf health to sea-level projections.
Basal melt, driven by relatively warm Circumpolar Deep Water intruding beneath the shelf, is now understood to be the dominant mass-loss process for many shelves. Calving is episodic and dramatic; basal melt is chronic and largely invisible from above. Satellite observation must address both.
SAR sees what optical cannot: calving fronts in the dark
Antarctica spends roughly half the year in polar night and is cloud-covered for much of the rest. Optical sensors, including Landsat OLI and Sentinel-2 MSS, are simply unavailable for months at a time over the most dynamic shelves. C-band SAR, by contrast, is indifferent to darkness and cloud. Sentinel-1's Interferometric Wide Swath mode covers the entire Antarctic coastline on a 6-day cycle, and orbital convergence at high latitudes means many shelf fronts are imaged every one to three days.
The calving front appears in SAR imagery as a sharp backscatter boundary between the coherent ice surface and the rougher or open ocean beyond. Automated edge-detection algorithms, trained on the contrast between shelf and sea, can delineate front positions to within a few pixels, roughly 20-40 m in IW mode. Rift interiors show as low-backscatter linear features cutting across the shelf; their propagation rate can be tracked by differencing sequential images. The publicly documented CALFIN dataset, which uses deep-learning segmentation on Landsat and SAR imagery, demonstrates front-position uncertainty of order 100-300 m depending on image quality and front type.
InSAR velocity fields and the grounding line you cannot see directly
The grounding line, where floating ice detaches from the bed, is not visible at the surface. But it has a measurable signature. Floating ice flexes tidally; grounded ice does not. Repeat-pass InSAR, using the phase difference between two SAR acquisitions separated by one tidal cycle, reveals a zone of differential vertical motion that brackets the grounding line to within a few hundred metres. ESA's RETREAT and MEaSUREs programmes have published grounding-line positions for most of Antarctica derived from this method using ERS, Envisat and Sentinel-1 data.
Velocity fields derived from offset tracking or InSAR coherence also reveal the dynamics of the shelf interior. Acceleration upstream of a rift, or deceleration where a pinning point grounds the shelf, are diagnostic signals. A shelf that is thinning from below will typically show surface drawdown in ICESat-2 repeat tracks before any visible calving event occurs.
Measuring what you cannot see: basal melt from altimetry and mass budget
ICESat-2 ATLAS measures surface-height change with centimetre-level precision over 1 km segments. For a floating shelf in hydrostatic equilibrium, a surface lowering of 1 cm corresponds to roughly 9 cm of basal thinning, because ice is about 10% less dense than seawater. Repeat ICESat-2 tracks, combined with firn-compaction models to separate dynamic thinning from densification, can resolve basal melt rates of a few metres per year over areas of tens of square kilometres. The Pine Island and Thwaites shelves, the two most closely watched, have published melt rates of 25-100 m per year in their fastest-melting zones, derived from this method.
An independent cross-check comes from the mass-budget approach: measure ice flux across a gate near the grounding line using InSAR velocities and ice-thickness estimates, then subtract accumulation. The residual is basal melt plus calving. Neither method is clean in isolation. ICESat-2 requires a firn model; the mass budget requires bed-topography data, which carries its own uncertainty. Honest analysis presents both and quantifies the disagreement.
Oceanographic proxies, specifically ocean temperature and salinity at depth from Argo floats and ship-based CTD casts, inform the forcing side of the equation. Warmer Circumpolar Deep Water shoaling onto the continental shelf correlates with elevated melt rates, though the coupling is not instantaneous and the sub-shelf cavity geometry modulates the response in ways that remain poorly constrained.
What the data cannot tell you, and why that matters
SAR front-delineation fails when sea ice fills the calving bay and the backscatter contrast between shelf and frozen ocean collapses. This is not a rare edge case; it happens every austral winter on several shelves. Analysts must flag ambiguous front positions rather than interpolate silently through them.
ICESat-2's 91-day repeat means it captures a snapshot of surface elevation, not a continuous record. Between repeat cycles, a rapid thinning event or a calving that removes a large area can go undetected until the next pass. Cloud is not the constraint for lidar that it is for optical sensors, but the 532 nm beam is attenuated by thick cloud, and the instrument flags degraded returns. Spatial coverage is also not wall-to-wall: the six-beam pattern leaves gaps between tracks that widen toward the equator, though at Antarctic latitudes the track spacing is typically 2-3 km at 88°S.
Basal melt rates derived from altimetry carry uncertainties of 20-40% in regions where firn-compaction models are poorly constrained. For policy or engineering decisions that depend on precise melt estimates, those error bars must be stated explicitly.
Turning observations into operational intelligence
For most government clients, the practical output is not a raw SAR scene but a time-series of front-position polygons, rift-length measurements and thinning-rate maps delivered as GIS layers against a defined baseline. Change alerts, triggered when a rift exceeds a threshold length or a front retreats beyond a defined isoline, convert continuous monitoring into actionable notifications without requiring a remote-sensing specialist to watch every image.
Satellize runs exactly this kind of analytics pipeline on open constellations, combining Sentinel-1 SAR, ICESat-2 altimetry and Landsat OLI into structured outputs. The same architecture that underpins the Kingdom of Tonga crop-estimation programme, ingesting open data and delivering structured analytics rather than raw imagery, applies directly to polar ice monitoring for national Antarctic programmes or infrastructure operators with exposure to sea-level risk. The Overhead column has covered several Antarctic calving events as they happened, using publicly available Sentinel-1 acquisitions.
If your organisation needs a defined calving-front baseline, a rift-propagation alert service, or a basal-melt time-series for a specific shelf, the right starting point is a scoping call to define the shelf extent, the required temporal resolution and the acceptable uncertainty budget.
Typical figures
| Calving-front spatial resolution (SAR) | 5 x 20 m (Sentinel-1 IW mode); front-position uncertainty typically 100-300 m depending on contrast |
| SAR revisit over Antarctica | 1-3 days (Sentinel-1, due to orbital convergence at high latitudes); 6-day exact repeat |
| ICESat-2 surface-height precision | Better than 3 cm per 1 km segment under clear sky; 91-day exact repeat |
| ICESat-2 along-track shot spacing | 0.7 m; six beams in three pairs, beam separation ~3.3 km across-track |
| Landsat OLI resolution | 15 m panchromatic, 30 m multispectral; 16-day repeat; usable only in polar summer under clear sky |
| Minimum detectable basal thinning (ICESat-2) | Approximately 0.3 m/yr surface lowering over 1 km segments; equates to ~2.7 m/yr basal melt |
| SAR archive depth (Sentinel-1) | From 2014; Landsat optical archive from 1972 (Landsat 1 MSS) |
| Grounding-line position uncertainty (InSAR) | Typically a few hundred metres; dependent on tidal amplitude and image-pair temporal baseline |
| Basal melt rate uncertainty (altimetry method) | 20-40% in poorly constrained firn-compaction regions; lower where firn models are validated |
| Delivery formats | GeoTIFF, GeoPackage/Shapefile (front polygons, rift polylines), NetCDF (thinning grids), change-alert JSON feed |
Analytics Satellize can run
| Calving-front position time-series | Automated edge detection and deep-learning segmentation on Sentinel-1 IW SAR and Landsat OLI imagery, cross-validated against each other where polar summer permits | Quarterly GIS polygon layer with front-position uncertainty estimate; annual retreat-rate summary report |
| Rift propagation tracking | Sequential SAR image differencing and manual verification; rift tip position extracted per acquisition | Rift-length time-series CSV and alert notification when propagation rate exceeds a client-defined threshold |
| Ice-shelf surface velocity field | SAR offset tracking and InSAR coherence on Sentinel-1 repeat pairs; mosaicked to full-shelf coverage | Seasonal velocity GeoTIFF mosaic; anomaly flag where acceleration exceeds baseline by defined margin |
| Grounding-line position mapping | Tidal differential InSAR on Sentinel-1 pairs spanning one tidal cycle; flexure zone delineated as grounding-line proxy | Grounding-line polyline shapefile with uncertainty buffer; comparison against published MEaSUREs baseline |
| Ice-shelf thinning rate map | ICESat-2 repeat-track altimetry differencing corrected for firn compaction using published firn-compaction model outputs; hydrostatic conversion to basal melt rate | Annual thinning-rate NetCDF grid with per-pixel uncertainty; shelf-integrated volume-change summary |
| Calving-event alert | Near-real-time SAR scene ingestion; front-position comparison against previous baseline polygon; area-loss calculation | Alert notification within 48 hours of Sentinel-1 acquisition confirming calving; estimated iceberg area and front-retreat distance |
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.