River delta sediment compaction and relative sea-level rise

Major river deltas subside at rates that dwarf global mean sea-level rise, driven by sediment compaction, fluid withdrawal and starved fluvial supply. InSAR from Sentinel-1 and ALOS-2 quantifies these rates at millimetre-per-year precision, and combined with tide-gauge records, yields the relative sea-level rise figure that coastal planners actually need.

Sensors

Sentinel-1 A/B (C-band SAR, ESA): 5.6 cm wavelength, 20 m ground range resolution in Interferometric Wide Swath mode, 250 km swath, 6-day repeat at the equator with both satellites. The workhorse for delta-wide deformation time-series using SBAS or PS-InSAR. Archive from 2014 onwards.
ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength penetrates vegetation canopy better than C-band, reducing decorrelation over rice paddies and mangroves. Stripmap mode at 3 m resolution; ScanSAR at 60 m over 350 km swath. 14-day repeat. Particularly valuable for the Mekong and Ganges-Brahmaputra deltas where dense vegetation degrades C-band coherence.
Envisat ASAR archive (C-band SAR, ESA): Operational 2002 to 2012. Provides the pre-Sentinel baseline for long-term trend separation. Image Mode at 30 m resolution. Coherence is lower than Sentinel-1 due to longer revisit intervals, but the archive extends the deformation record by a decade, critical for distinguishing secular compaction from episodic events.
GNSS continuous stations co-located with tide gauges: Absolute vertical land motion from GNSS, typically precise to 0.3 to 1 mm/year over multi-year records, is subtracted from tide-gauge records to isolate oceanographic sea-level change from local subsidence. The Permanent Service for Mean Sea Level (PSMSL) maintains the global tide-gauge archive; SONEL coordinates GNSS-at-tide-gauge data.
Sentinel-1 coherence matrix (temporal decorrelation analysis): Not a separate sensor but an analytical product derived from Sentinel-1 pairs. Short temporal baselines (6 or 12 days) preserve coherence over bare sediment and urban areas; longer baselines reveal vegetation seasonality that must be masked before subsidence retrieval.

Why deltas sink: three mechanisms, one compounding problem

River deltas are built from sediment deposited faster than it consolidates. The weight of overlying material expels pore water from deeper layers, a process called primary consolidation, which continues for decades to centuries after deposition. Secondary compaction, driven by organic matter decomposition and grain rearrangement, adds further settlement at shallower depths. These natural processes produce background subsidence rates of roughly 1 to 5 mm per year across the delta plain, varying with sediment grain size, organic content and depositional age.

Anthropogenic drivers amplify this considerably. Groundwater extraction reduces pore pressure, accelerating consolidation and causing the surface to drop at rates that can exceed 20 to 30 mm per year in heavily pumped areas, as documented in parts of the Mekong delta. Hydrocarbon withdrawal has similar mechanics. Upstream dams intercept the fluvial sediment supply that would otherwise replenish the delta surface, so the natural offset against compaction disappears. The Mississippi delta loses roughly 25 to 35 km² of land per year, a figure tied directly to the sediment deficit created by levee construction and river management since the mid-20th century.

What InSAR actually measures, and what it does not

InSAR measures the change in slant-range distance between a SAR satellite and the ground surface across two or more acquisition dates. A single interferometric pair gives a snapshot of displacement projected along the radar line of sight, typically within about 20 to 40 degrees of vertical for the satellite geometries used over deltas. Multi-temporal methods, principally Persistent Scatterer InSAR (PS-InSAR) and Small Baseline Subset (SBAS), stack dozens to hundreds of interferograms to extract a time-series of surface motion at each coherent pixel, suppressing atmospheric noise and improving velocity precision to roughly 1 to 2 mm per year over multi-year stacks.

The honest limits matter. InSAR measures relative displacement between coherent points; it does not give absolute elevation. Over agricultural deltas, coherence drops sharply during flooding or dense crop growth, leaving gaps precisely where subsidence is often fastest. L-band (ALOS-2) mitigates this but does not eliminate it. Atmospheric water vapour introduces artefacts at the 5 to 20 mm level per acquisition, partially corrected using ERA5 reanalysis or GACOS tropospheric delay products but never fully removed. Subsidence faster than roughly half the radar wavelength per revisit interval (about 28 mm per 6-day cycle for Sentinel-1) causes phase unwrapping failure, which is a real problem in areas of rapid groundwater-driven collapse.

Published results: Mekong, Ganges-Brahmaputra and Mississippi

Studies using Sentinel-1 over the Mekong delta have reported spatially variable subsidence averaging 10 to 20 mm per year across the Vietnamese delta plain, with localised peaks exceeding 30 mm per year near Can Tho and other urban centres where groundwater extraction is intensive. The signal correlates spatially with well density and pumping depth, providing a mechanistic link rather than a statistical coincidence. ALOS-2 observations confirm these rates and extend coherent coverage into the mangrove fringes where Sentinel-1 loses phase.

The Ganges-Brahmaputra delta, shared between Bangladesh and India, presents a different pattern. Natural compaction of the thick Holocene sediment package drives background rates of 2 to 5 mm per year, but anthropogenic extraction in Dhaka and surrounding areas produces local anomalies of 20 to 50 mm per year, documented in studies using both Envisat ASAR and Sentinel-1 time-series. The relative sea-level rise experienced by coastal Bangladesh is therefore the sum of eustatic rise (currently around 3 to 4 mm per year from satellite altimetry) and local subsidence, giving effective rates of 5 to 10 mm per year or more across much of the delta.

The Mississippi delta is the longest-studied case. Envisat and Sentinel-1 time-series, cross-validated against levelling surveys and GNSS, show subsidence of 5 to 10 mm per year across the natural delta plain and up to 20 mm per year in areas of historical hydrocarbon extraction. The combination of sediment starvation, fluid withdrawal and eustatic rise produces relative sea-level rise rates that are among the highest on any inhabited coastline, with some tide-gauge records showing effective rise exceeding 9 mm per year.

Separating natural from anthropogenic signal: the analytical challenge

A subsidence map alone does not tell a planner which lever to pull. Separating primary consolidation from groundwater extraction requires spatial correlation with hydrogeological data, pumping records and stratigraphic models. Where the InSAR velocity field shows sharp spatial gradients coinciding with well-field boundaries, the anthropogenic interpretation is defensible. Where subsidence is diffuse and correlates with Holocene sediment thickness, natural compaction is the more parsimonious explanation. In practice, both signals overlap, and decomposition requires iterative modelling rather than a single-pass analysis.

GNSS stations co-located with tide gauges provide the ground truth for absolute vertical land motion, but their spatial density on most deltas is sparse. A typical approach uses GNSS to anchor the InSAR velocity field to an absolute reference frame, then interpolates using the dense InSAR coverage to produce a spatially continuous relative sea-level rise map. The uncertainty on this product is typically 1 to 3 mm per year, dominated by atmospheric noise in the InSAR stack and by the short GNSS records at many delta stations.

From deformation map to coastal planning number

A relative sea-level rise rate for coastal infrastructure design is not simply the InSAR velocity. It requires adding the local eustatic trend from the nearest reliable tide gauge (corrected for glacial isostatic adjustment using published GIA models), accounting for interannual variability in ocean dynamic sea level, and propagating all measurement uncertainties honestly into a planning range rather than a point estimate. The output is a probabilistic envelope: for example, 8 to 14 mm per year of effective sea-level rise at a given delta location over the next 30 years, under current extraction rates.

Satellize can assemble this pipeline for delta authorities or infrastructure investors, drawing on open Sentinel-1 and ALOS-2 archives, PSMSL tide-gauge records and ERA5 atmospheric correction. The approach is the same family of methods used in our Tonga crop-estimation programme: open-constellation data processed through published analytical frameworks, delivered as decision-ready products rather than raw imagery. The concrete output is a georeferenced subsidence velocity grid, a relative sea-level rise map with uncertainty bounds, and a written interpretation of the anthropogenic versus natural signal split, formatted for submission to national coastal adaptation planning processes.

Typical figures

Spatial resolution (Sentinel-1 IW mode)	20 m range × 22 m azimuth ground resolution; PS-InSAR point density typically 50 to 500 scatterers per km² over urban areas, lower over agricultural land
Spatial resolution (ALOS-2 PALSAR-2 Stripmap)	3 m × 3 m; ScanSAR mode 60 m × 60 m over 350 km swath
Revisit interval	6 days (Sentinel-1 A+B combined at equator); 14 days (ALOS-2); 35 days (Envisat ASAR archive)
Minimum detectable velocity (multi-year stack)	Approximately 1 to 2 mm per year for PS-InSAR over coherent surfaces with 3+ years of data; 3 to 5 mm per year for SBAS over agricultural areas
Radar frequency and wavelength	C-band 5.405 GHz / 5.6 cm (Sentinel-1, Envisat); L-band 1.2575 GHz / 23.6 cm (ALOS-2)
Archive depth	Sentinel-1: 2014 to present; ALOS-2: 2014 to present; Envisat ASAR: 2002 to 2012; ALOS-1 PALSAR: 2006 to 2011
Atmospheric correction uncertainty	5 to 20 mm per acquisition epoch; reducible to 2 to 5 mm using ERA5 or GACOS tropospheric delay models
Tide-gauge vertical land motion precision (GNSS)	0.3 to 1 mm per year for continuous GNSS records exceeding 5 years; degrades significantly for records under 3 years
Delivery formats	GeoTIFF velocity grids, shapefiles of subsidence anomaly zones, PDF technical report, CSV time-series per point-of-interest

Analytics Satellize can run

Delta-wide subsidence velocity map	SBAS-InSAR time-series analysis on Sentinel-1 IW SLC stack, with ERA5 atmospheric correction and phase unwrapping using minimum cost flow	GeoTIFF raster of mean line-of-sight velocity (mm/year) with uncertainty layer, covering the full delta extent
Urban and peri-urban PS-InSAR deformation time-series	Persistent Scatterer InSAR on Sentinel-1 or ALOS-2, anchored to co-located GNSS reference station for absolute vertical conversion	Point shapefile of PS velocities and displacement time-series, formatted for import into coastal infrastructure asset registers
Relative sea-level rise map with uncertainty bounds	InSAR vertical velocity combined with PSMSL tide-gauge eustatic trend and published GIA correction (e.g. ICE-6G model), propagated through Monte Carlo uncertainty analysis	Georeferenced map of effective relative sea-level rise (mm/year) with 5th to 95th percentile bounds, suitable for submission to national adaptation planning processes
Anthropogenic versus natural subsidence decomposition	Spatial regression of InSAR velocity field against groundwater well density, Holocene sediment thickness (from published stratigraphic surveys) and hydrocarbon field extents	Written technical interpretation report with spatial attribution maps, identifying priority zones for extraction management
Coherence loss and data-gap assessment	Temporal coherence matrix analysis across the full Sentinel-1 and ALOS-2 archive, identifying seasons and areas where phase retrieval is unreliable	Coherence mask layer and recommended acquisition strategy for supplementary ALOS-2 tasking to fill C-band gaps
Long-term trend baseline from Envisat ASAR archive	SBAS processing of Envisat ASAR Image Mode scenes (2002 to 2012) cross-calibrated to Sentinel-1 velocity field to extend the deformation record	Decadal velocity comparison table (Envisat epoch vs Sentinel-1 epoch) highlighting acceleration or deceleration of subsidence trends

Sources

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.