Aeolian sand dune migration rate and sediment flux mapping
Active sand dunes migrate at metres to tens of metres per year, threatening roads, pipelines and farmland. SAR amplitude offset tracking and sub-metre optical feature tracking now make systematic, basin-wide measurement routine.
Sensors
- Sentinel-1 A/B (C-band SAR): 5 m x 20 m ground range detected resolution in IW mode; 6-day repeat at mid-latitudes with both satellites. SAR amplitude offset tracking cross-correlates image patches across acquisition pairs to measure surface displacement of 0.1 pixel or better, resolving dune crest movement of roughly 1 m or more per tracking interval. Unaffected by cloud or darkness.
- Pléiades 1A/1B (optical, sub-metre): 0.5 m panchromatic, 2 m multispectral. On-demand tasking with same-day stereo capability for DEM generation. Optical normalised cross-correlation between repeat passes resolves dune crest displacement to sub-pixel precision, typically 0.3–0.5 m at 0.5 m resolution. Best results when acquisition geometry is consistent and wind has not altered dune shape between passes.
- COSMO-SkyMed (X-band SAR): Spotlight mode delivers 1 m resolution; stripmap at 3–15 m. X-band backscatter is sensitive to surface roughness at millimetre scale, improving discrimination between active slip-face and armoured lag surface. Revisit of 1–4 days for tasked acquisitions. Useful for high-cadence tracking of fast-moving barchans.
- Landsat 8/9 (OLI): 30 m multispectral, 15 m panchromatic. 16-day repeat; archive extends to 1972 for earlier missions. Too coarse for individual dune crest tracking except on very large draa or mega-dunes, but indispensable for decadal sand-sea boundary change and for calibrating migration trends against long climate records.
Why dunes move at the speeds they do
Dune migration rate is controlled by two competing factors: the sand flux arriving from upwind, and the dune's own volume. A barchan dune on the Skeleton Coast or in the Qaidam Basin migrates faster when it is small, because the same wind-driven flux moves a smaller mass. The relationship, first formalised by Bagnold in 1941 and confirmed repeatedly by field and remote-sensing studies, predicts that migration velocity scales inversely with dune height. Published rates for barchans in the Sahara range from roughly 10 to 100 m per year; transverse dunes in the Taklamakan average 5–30 m per year depending on fetch and wind regime.
Sand flux itself depends on the cube of wind speed above a threshold, which means that a relatively modest increase in storm frequency can disproportionately accelerate a dune field. This non-linearity is why infrastructure engineers cannot simply extrapolate from a single measured rate: a road that was safe for thirty years can be buried within a decade if regional wind patterns shift.
What offset tracking actually measures, and where it fails
SAR amplitude offset tracking divides two co-registered images into small chips, typically 64 x 64 or 128 x 128 pixels, and finds the sub-pixel shift that maximises cross-correlation between each chip pair. The result is a dense displacement field covering the entire scene. For Sentinel-1 IW mode, the practical detection floor is roughly 1–2 m of horizontal displacement over a 12-day pair at mid-latitudes, improving to about 0.5 m with 6-day pairs when both satellites are operational. Published studies on the Sahara using Sentinel-1 have resolved crest migration rates of 8–60 m per year with this method.
The method has genuine limits. Dunes that change shape between acquisitions, rather than simply translating, degrade correlation and produce noisy or missing vectors. Slip-face avalanching, which is episodic, can shift the crest position discontinuously. Very flat sand sheets with no coherent texture return poor correlation. Optical feature tracking with Pléiades avoids the shape-change ambiguity somewhat, because visible texture is richer, but optical data is blocked by cloud and by the haze that accompanies dust storms, which are precisely the events that drive the largest dune displacements.
From displacement vectors to sediment flux
Measuring where a dune crest has moved is only the first step. Estimating volumetric sediment flux requires a cross-sectional area. That area comes from a digital elevation model, either photogrammetric (from Pléiades stereo pairs, achieving vertical accuracy of 0.5–1 m in low-relief desert terrain) or from TanDEM-X, which provides a global 12 m DEM with roughly 2 m vertical accuracy over sand surfaces. Multiplying the mean cross-sectional area of a dune by its measured migration velocity gives a bulk volumetric flux in cubic metres per metre width per year, a figure directly useful for planning sand-exclusion barriers or estimating burial timelines for buried infrastructure.
The conversion carries uncertainty. Dune porosity (typically 0.35–0.40 for aeolian sand) must be assumed to convert bulk volume to mass flux. Dune height varies along the crest, so a single profile underestimates total volume. Studies on Arabian Peninsula dunes using this approach report flux values of 10–80 m³ per metre width per year, with the highest values on exposed coastal corridors such as the Wahiba Sands.
Published results from the Sahara, Arabia and Taklamakan
The public literature on satellite-derived dune kinematics is now substantial. Vermeesch and Drake (2008) used ASTER and Landsat to track barchans in the Western Sahara, finding rates of 20–50 m per year consistent with field surveys. Subsequent Sentinel-1 studies, published in Remote Sensing (MDPI) and similar journals, have confirmed and refined those figures at finer spatial resolution. In the Taklamakan, where China's Southern Xinjiang Highway is a chronic engineering challenge, published tracking results show transverse dunes advancing at 5–25 m per year, with the fastest movement concentrated in topographic corridors that funnel northeasterly winds.
On the Arabian Peninsula, barchan fields near the Rub' al Khali margins have been tracked using COSMO-SkyMed spotlight imagery at 1 m resolution, resolving individual dune horn advance at rates of 15–40 m per year. These figures feed directly into maintenance scheduling for oil and gas infrastructure, where sand encroachment on access roads is a recurring operational cost. None of this is exotic research: the methods are published, the sensors are available, and the gap is usually in systematic operational deployment rather than technical feasibility.
What a programme actually needs to deliver useful numbers
A credible dune-monitoring programme requires three things that are often treated as afterthoughts. First, a baseline DEM of sufficient vertical accuracy to support flux estimation. A 30 m SRTM DEM is not adequate for dunes below about 10 m height; Pléiades stereo or TanDEM-X is the practical minimum. Second, a consistent acquisition geometry across the time series. SAR offset tracking degrades sharply when incidence angles differ between passes, and optical tracking is sensitive to sun-angle differences that alter shadow length and apparent crest position. Third, wind data co-registered in time. Without knowing the wind forcing between acquisitions, a measured displacement cannot be extrapolated to an annual rate with any confidence.
Satellize structures dune-migration analytics around these three requirements, combining open Sentinel-1 time series with commercial Pléiades tasking for baseline DEM generation and periodic high-resolution validation. The approach is the same one applied to agricultural land-cover monitoring in the Kingdom of Tonga crop-estimation programme: open data for systematic coverage, commercial data for precision where it matters. The analytic outputs are GIS layers and structured reports, not raw imagery, because the client's engineer needs a migration rate and a confidence interval, not a stack of SAR scenes.
Typical figures
| SAR tracking spatial resolution | 5 x 20 m (Sentinel-1 IW); 1–3 m (COSMO-SkyMed Spotlight/Stripmap) |
| Optical tracking spatial resolution | 0.5 m panchromatic (Pléiades); 15 m pan-sharpened (Landsat 8/9) |
| Minimum detectable displacement (SAR offset tracking) | ~0.5–2 m per 6–12 day pair (Sentinel-1); ~0.1–0.3 m per pair (COSMO-SkyMed Spotlight) |
| Revisit cadence | 6 days (Sentinel-1 A+B combined); 1–4 days tasked (COSMO-SkyMed); 16 days (Landsat 8/9) |
| DEM vertical accuracy over sand | 0.5–1 m (Pléiades stereo photogrammetry); ~2 m (TanDEM-X 12 m product) |
| Archive depth | Sentinel-1: 2014–present; Landsat: 1972–present; Pléiades: 2012–present (tasked archive) |
| Cloud/dust sensitivity | SAR: unaffected. Optical: blocked by cloud and dense dust haze |
| Typical migration rate detection range | ~1–200 m per year (SAR); ~0.5–200 m per year (sub-metre optical) |
| Sediment flux estimation uncertainty | ±15–30% depending on DEM accuracy and assumed porosity (0.35–0.40) |
| Delivery formats | GeoTIFF displacement rasters, vector crest-position shapefiles, CSV flux tables, PDF reports |
Analytics Satellize can run
| Annual dune crest migration map | SAR amplitude offset tracking (normalised cross-correlation on Sentinel-1 IW pairs); displacement vectors averaged over 12-month stack | GeoTIFF raster of migration velocity (m/yr) with vector confidence layer; updated annually or on request |
| High-resolution crest displacement (sub-metre) | Optical normalised cross-correlation on Pléiades 0.5 m panchromatic pairs; co-registration to <0.1 pixel | Vector displacement field shapefile; per-dune crest advance table in CSV |
| Volumetric sediment flux estimate | Migration velocity multiplied by dune cross-sectional area from Pléiades stereo DEM; porosity correction applied | Structured report with flux values (m³/m/yr) per dune or dune-field zone, with uncertainty bounds |
| Decadal sand-sea boundary change | Landsat 8/9 OLI time series classification using band ratio indices (e.g. sand index from SWIR/NIR); change detection against 1984 baseline | Multi-epoch GeoTIFF stack and area-change summary table; suitable for land-use planning input |
| Infrastructure encroachment risk layer | Extrapolation of measured migration vectors to 5- and 10-year horizons; intersection with infrastructure GIS data supplied by client | Risk-ranked shapefile of infrastructure assets with estimated years to sand contact; PDF briefing for engineering teams |
| Wind-forcing correlation report | Co-registration of ERA5 reanalysis wind vectors with measured displacement time series; regression of sand flux against wind power proxy | PDF report with scatter plots and R² statistics; identifies dominant transport seasons for maintenance scheduling |
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.