Glacier mass balance and volume change monitoring
The geodetic method compares digital elevation models from different epochs to quantify glacier surface-elevation change, then converts volume to mass. Satellite stereo imagery, SAR interferometry and laser altimetry now make this possible at near-global scale.
Sensors
- TanDEM-X InSAR (TerraSAR-X / TanDEM-X pair): Bistatic X-band SAR interferometry; global DEM at 12 m posting, absolute height accuracy approximately 2 m, relative accuracy better than 1 m over flat terrain. The 2010–2015 global acquisition provides the standard geodetic reference surface for mass-balance studies. Repeat acquisitions enable differencing at roughly annual cadence where tasked.
- ICESat-2 ATLAS: Photon-counting lidar at 532 nm; along-track ground spacing approximately 0.7 m, 91-day exact repeat, 1387-day sub-cycle. Surface-elevation precision better than 0.03 m over flat ice on a single pass. Provides independent validation of DEM-differencing results and captures elevation change on narrow outlet glaciers too small for SAR coherence.
- ASTER VNIR stereo: 15 m resolution nadir and aft-looking bands 3N/3B at 27.6-degree convergence angle. Stereo-derived DEMs typically carry 10–30 m horizontal accuracy and 5–15 m vertical uncertainty depending on terrain relief and image quality. Free archive back to 2000 makes it the workhorse for multi-decadal geodetic mass-balance studies in regions without SAR coverage.
- Copernicus DEM (GLO-30 / GLO-90): Derived from TanDEM-X acquisitions, released at 30 m and 90 m global posting. Freely available as a consistent reference epoch (nominally 2011–2015 acquisitions). Differencing a newer stereo or InSAR DEM against Copernicus DEM is now a standard approach for decadal change detection.
- Sentinel-2 MSI: 10 m visible and near-infrared bands, 5-day revisit at mid-latitudes. Not a DEM source, but used to delineate glacier outlines (via NDSI and manual editing) that define the spatial mask for volume-change calculations. Accurate outlines matter: a 5% error in glacier area propagates directly into the mass-balance estimate.
What a differenced DEM actually measures
The geodetic method is straightforward in principle. Acquire a DEM of a glacier at time T1, acquire another at time T2, subtract one from the other pixel by pixel, and you have a map of surface-elevation change. Multiply by area and a density assumption, and you have mass change in gigatonnes or millimetres of sea-level equivalent.
In practice, the method carries several compounding uncertainties. Co-registration error between the two DEMs is the dominant noise source; even a one-pixel shift over sloped terrain introduces spurious elevation differences. The standard correction, developed by Nuth and Kääb (2011) and now widely implemented, regresses elevation difference against terrain aspect over stable off-glacier ground to solve for the horizontal offset before any glaciological interpretation begins. Residual biases from different radar penetration depths between epochs, or from different atmospheric conditions during stereo acquisition, can add systematic offsets of 1–3 m that look indistinguishable from real ice loss at the individual-glacier scale.
The density problem: firn is not ice
Converting a volume change to a mass change requires knowing what material was lost or gained. Ice has a density of approximately 917 kg/m³. Fresh snow is closer to 100–400 kg/m³. The firn layer, the intermediate compacted snow that covers accumulation zones, sits somewhere between 550 and 830 kg/m³ and is itself compacting under its own weight even when no net mass is being lost.
The standard practice, following Huss (2013), is to apply a bulk density of 850 ± 60 kg/m³ when a glacier is losing mass over a multi-year period and the change integrates across both accumulation and ablation zones. This assumption introduces an uncertainty of roughly 7% in the mass estimate, which is acceptable for regional aggregates but becomes significant for individual small glaciers where the firn fraction of total volume is large. When a glacier is in near-balance or gaining mass, the appropriate density is far less certain, and honest reporting requires widening the error bars accordingly.
TanDEM-X as a reference surface, and its limits
The TanDEM-X global DEM, acquired primarily between 2010 and 2015, has become the de facto geodetic baseline for glacier studies. Its 12 m posting and sub-metre relative vertical accuracy over ice are substantially better than the SRTM (2000, 30 m, C-band) it largely replaced for cryospheric work. The Copernicus DEM GLO-30 product, freely distributed via the Copernicus Data Space, is derived from TanDEM-X and is the practical entry point for most users.
X-band radar penetrates dry snow and firn by up to several metres, depending on snow grain size and liquid water content. This means the TanDEM-X surface over an accumulation zone sits below the true snow surface by an unknown amount, typically 1–5 m on temperate glaciers and potentially more on cold polar firn. When differencing a newer optical DEM against TanDEM-X, the penetration bias shifts the apparent elevation change. Studies address this by restricting comparisons to ablation zones where bare ice is exposed and penetration is negligible, or by applying a penetration correction derived from ancillary data. Neither approach eliminates the uncertainty entirely.
ICESat-2 as a validator, not just a dataset
ICESat-2's ATLAS instrument fires 10,000 laser pulses per second across six beams arranged in three pairs, producing along-track elevation profiles with centimetre-level precision over flat ice. Its 91-day repeat allows direct comparison of the same ground track across years. For large ice sheets this is the primary measurement. For mountain glaciers, the geometry is more awkward: a single pass may cross a narrow glacier at an oblique angle, sampling only a thin strip of the surface.
The value of ICESat-2 for the geodetic method is validation rather than primary mapping. Where an ICESat-2 repeat track crosses a glacier that has also been covered by DEM differencing, the two estimates can be compared directly. Agreement within 0.2–0.5 m/year is generally considered satisfactory. Systematic disagreement usually points to a co-registration error or a penetration bias in the DEM pair, and the ICESat-2 number, being a direct physical measurement, is the one to trust. The NASA NSIDC distributes the ATL06 land-ice elevation product, which is the standard input for this validation workflow.
Cloud cover over maritime glaciers: the practical ceiling
Optical stereo methods, whether ASTER or commercial high-resolution imagery, require cloud-free acquisitions over the glacier at both epochs. Maritime glaciers, the ones in Patagonia, coastal Alaska, Iceland, Norway and New Zealand's Southern Alps, sit in some of the cloudiest airspace on Earth. Annual cloud-free acquisition rates over parts of Patagonia can fall below 10%, meaning years can pass without a usable stereo pair.
SAR interferometry sidesteps cloud cover entirely, but introduces its own constraints. InSAR-derived DEMs require coherence between the two acquisitions; wet snow and rapid surface change can decorrelate the signal within days. TanDEM-X bistatic acquisitions, taken simultaneously from two spacecraft, avoid temporal decorrelation by definition, which is why the global TanDEM-X DEM is coherent even over dynamic ice. For change detection using repeat-pass InSAR, coherence over wet maritime glaciers is often poor, and the method defaults back to optical stereo when conditions allow. The honest position is that no single sensor solves the maritime glacier problem completely; combining SAR reference surfaces with opportunistic optical acquisitions and ICESat-2 profiles is the current best practice.
Satellize's analytics pipeline applies this multi-sensor approach to open constellations, with commercial tasking added on client licence where archive gaps exist. The workflow is the same one we use for the Tonga crop-estimation programme: ingest, validate against independent elevation data, and deliver a quantified uncertainty alongside the headline number.
What the output looks like, and what to do with it
A completed geodetic mass-balance analysis delivers a gridded elevation-change map (typically at 30–100 m posting depending on input DEM resolution), a per-glacier or per-basin mass-change time series in gigatonnes per year with explicit uncertainty bounds, and a hypsometric breakdown showing where elevation change is concentrated by altitude band. The hypsometric profile distinguishes a glacier losing mass primarily at its terminus from one thinning across the entire accumulation zone, which has different implications for future trajectory.
Water-resource planners care about the downstream freshwater flux, which requires converting mass change to runoff volume and distributing it across a melt season. Hydropower operators in glacier-fed basins use multi-year mass-balance trends to anticipate the 'peak water' inflection point, the decade when accelerating melt temporarily increases runoff before the glacier shrinks too far to sustain it. Infrastructure developers and reinsurers use the same data to assess proglacial lake growth and the associated outburst flood hazard. The satellite record, now stretching back to ASTER's 2000 launch and SRTM's February 2000 acquisition, is long enough to detect statistically significant trends for most glaciers larger than a few square kilometres.
Typical figures
| Primary DEM spatial resolution | TanDEM-X: 12 m posting; Copernicus GLO-30: 30 m; ASTER stereo: 15–30 m native, resampled to 30 m for analysis |
| Vertical accuracy (reference DEM) | TanDEM-X relative: better than 1 m (flat terrain); absolute: approximately 2 m; ASTER stereo: 5–15 m depending on terrain |
| ICESat-2 elevation precision | Better than 0.03 m per pass over flat ice; 91-day exact repeat; 1387-day sub-cycle for dense coverage |
| Revisit for change detection | ASTER: up to 16 days (cloud-limited in practice); TanDEM-X repeat tasking: approximately annual; Sentinel-2 outline mapping: 5 days |
| Minimum detectable elevation change | Approximately 0.3–0.5 m/year over multi-year periods after co-registration correction; limited by DEM noise floor and density uncertainty |
| Density assumption uncertainty | 850 ± 60 kg/m³ (Huss 2013 bulk assumption for net-loss periods); introduces approximately 7% uncertainty in mass estimate |
| Archive depth | ASTER stereo: 2000 to present; SRTM reference: February 2000; TanDEM-X reference: 2010–2015; ICESat-2: 2018 to present |
| Radar penetration bias (X-band over firn) | 1–5 m on temperate glaciers; potentially greater on cold polar firn; negligible over bare ablation-zone ice |
| Glacier outline mapping accuracy | Sentinel-2 NDSI at 10 m; manual editing required for debris-covered ice; outline error propagates linearly into volume estimate |
| Output formats | GeoTIFF elevation-change grids; CSV mass-balance time series; GeoPackage glacier outlines; PDF uncertainty report |
Analytics Satellize can run
| Geodetic mass-balance map | DEM differencing (Copernicus GLO-30 or TanDEM-X vs. newer ASTER or commercial stereo DEM), Nuth-Kääb co-registration correction, 850 kg/m³ bulk density conversion | GeoTIFF elevation-change grid at 30 m posting, per-glacier mass-change table with uncertainty bounds, PDF summary report |
| Multi-epoch mass-balance time series | Sequential DEM differencing across ASTER archive epochs from 2000 to present, hypsometric integration by 50 m altitude band | CSV time series of annual or biennial mass change in Gt/year and mm sea-level equivalent, with propagated uncertainty per epoch |
| ICESat-2 validation cross-check | ATL06 land-ice elevation product extracted along repeat ground tracks, differenced against DEM-derived elevation-change field at coincident locations | Validation report quantifying agreement or systematic offset between geodetic and altimetric estimates, flagging penetration-bias candidates |
| Glacier outline change layer | Sentinel-2 NDSI thresholding plus manual editing for debris-covered margins, compared across available cloud-free epochs | GeoPackage polygon layer of glacier extents per epoch, area-change statistics, and terminus retreat distance for selected outlet glaciers |
| Proglacial lake growth monitoring | Sentinel-2 water-body detection (NDWI) combined with elevation-change maps to identify new or expanding ice-contact lakes | Annual lake-area and estimated volume time series, flagged alerts when area growth exceeds a client-defined threshold |
| Peak-water trajectory assessment | Multi-decadal mass-balance trend extrapolation using published degree-day or temperature-index melt scaling, combined with glacier hypsometry | Basin-level report estimating the approximate decade of peak meltwater runoff and the post-peak runoff trajectory, with explicit scenario uncertainty |
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.