Monitoring upstream dam construction and its hydrological consequences
Optical, SAR and altimetry data can reveal dam construction and impoundment in transboundary basins before any official announcement, giving downstream governments an independent hydrological record.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m resolution in Interferometric Wide Swath mode, 6-day repeat at mid-latitudes (12-day per satellite). Penetrates cloud and operates at night, making it the primary tool for tracking earthwork extent, cofferdam placement and progressive inundation regardless of season or weather.
- Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands, 20 m shortwave infrared, 5-day revisit with both satellites combined. Distinguishes open water from bare soil and vegetation, maps reservoir pool extent, and detects the turbidity plume that typically appears downstream of active construction.
- Planet SuperDove: 3–4 m resolution, daily revisit over most land areas. Commercial tasking allows targeted collection over a specific dam site once a Sentinel alert flags activity. Resolves individual construction features such as spillway channels and access roads that are sub-pixel at Sentinel resolution.
- SWOT KaRIn (NASA/CNES): Ka-band radar interferometer launched December 2022. Measures water-surface elevation with decimetric precision and a swath of roughly 120 km, revisiting most river reaches every 21 days. Capable of detecting the step-change rise in reservoir surface elevation that marks initial impoundment, even on reaches with no in-situ gauge.
What earthworks look like from orbit before concrete is poured
Dam construction leaves a distinctive sequence of signatures that precede any official announcement. The earliest detectable phase is access-road construction and the clearing of riparian vegetation, visible in Sentinel-2 imagery as a bright bare-soil scar extending toward the river. Within weeks, cofferdam placement begins: a temporary embankment, often of compacted rock or sandbags, diverts flow so the main foundation can be excavated in dry conditions. In SAR backscatter, a cofferdam appears as a high-return linear feature bisecting the river channel, typically 20–80 m wide depending on river width and dam class.
As the main embankment rises, the disturbed area grows. Time-series change detection on Sentinel-1 coherence images is particularly sensitive here: construction activity repeatedly disrupts the surface, reducing interferometric coherence to near zero over the active zone while the surrounding terrain retains coherence. This decorrelation footprint can be tracked week by week. For large projects, the embankment itself eventually becomes resolvable in Planet imagery, and its crest elevation can be estimated by shadow-length photogrammetry if the acquisition geometry is favourable.
The inundation signal: reading a filling reservoir
Once the main gates close, the reservoir begins to fill. This is the most unambiguous signal in the data record. Open water has a very low SAR backscatter in calm conditions (specular reflection away from the sensor) and a distinctive spectral signature in Sentinel-2: high reflectance in the blue band, near-zero in the near-infrared. Mapping the advancing water edge on a weekly Sentinel-1 composite gives a direct record of impoundment rate and total inundated area.
The rate of filling is itself informative. A reservoir that fills in one wet season implies a relatively small storage volume or an unusually wet year. One that takes three or four years to reach operating level suggests a large storage, which in turn implies a larger potential impact on downstream flow timing. Published studies of the Mekong and Nile basins have used this filling-rate logic to estimate active storage capacity from optical time series alone, cross-checked against topographic data from SRTM or Copernicus DEM.
Cloud cover is the main practical constraint on optical methods. In tropical basins with long monsoon seasons, Sentinel-2 may return only a handful of usable scenes per quarter. SAR is the fallback, but SAR water-edge mapping becomes ambiguous in wind-roughened conditions or where floating vegetation covers the margins. Honest interpretation requires both sensors, and accepting that some filling events will be reconstructed retrospectively rather than observed in real time.
Altimetry catches what imagery misses
Surface-area mapping tells you the reservoir is filling. Altimetry tells you by how much in elevation terms, which is the variable that actually governs storage volume. Before SWOT, analysts relied on the legacy radar altimeters aboard Envisat, ICESat-2 and the Sentinel-6/Jason series. These instruments fly a fixed ground track with a footprint of roughly 300 m to several kilometres, so they only yield usable readings when a track crosses the reservoir at sufficient width. Coverage was opportunistic.
SWOT changes the geometry. Its 120 km swath means it observes most reservoirs larger than roughly one square kilometre on every pass, with 21-day repeat. The KaRIn instrument measures water-surface elevation with a precision of a few centimetres over calm water, sufficient to resolve a step-change impoundment signal of even a modest dam. Crucially, SWOT observes the downstream river reach on the same pass, so the elevation drop across the dam structure is directly measurable without any in-situ gauge. For transboundary monitoring, this is the key capability: an independent, non-cooperative measurement of what is happening on both sides of the dam.
Downstream consequences written in the flow record
A dam does not destroy water; it reschedules it. During filling, downstream discharge falls, sometimes sharply. During operation, the seasonal hydrograph is altered: flood peaks are clipped, low-flow periods may be augmented or depleted depending on the dam's purpose. Hydropower dams tend to produce rapid sub-daily fluctuations in discharge as turbines cycle on and off, a signal detectable in high-frequency gauge data but not yet directly from orbit.
Where gauge data are unavailable or withheld, satellite proxies fill part of the gap. River width measured from Sentinel-2 or Planet imagery correlates with discharge through hydraulic geometry relationships, with uncertainties of roughly 20–40 percent depending on channel morphology and calibration. SWOT adds water-surface slope, which combined with width allows a Manning-equation discharge estimate. Neither method matches a well-maintained gauge, but both provide an independent record that a downstream state can present as evidence of changed flow conditions.
Sediment is a related consequence often overlooked. Reservoirs trap sediment, so downstream reaches typically show reduced turbidity after impoundment. This is detectable in Sentinel-2 band-ratio indices sensitive to suspended sediment. A persistent reduction in downstream turbidity, coinciding with the inundation signal upstream, is corroborating evidence of active storage.
Transboundary use: building an independent evidence record
The political context for this use case is specific. Downstream states on transboundary rivers often lack access to upstream gauge data, environmental-impact assessments or construction schedules. The 1997 UN Watercourses Convention establishes notification obligations, but compliance is uneven. Satellite observation fills the information gap without requiring the upstream state's cooperation.
An effective monitoring programme for this purpose combines three elements. First, a baseline archive review: Sentinel-1 and Sentinel-2 data are available from 2014 and 2015 respectively, and Landsat extends the record to 1972. Establishing what the river and its tributaries looked like before any construction is essential for attribution. Second, a near-real-time alert layer keyed to SAR coherence change and optical water-extent anomalies. Third, a periodic altimetry summary from SWOT passes, tied to the reservoir and the downstream reach.
Satellize structures this kind of programme as a standing analytics service, drawing on open Sentinel and SWOT data and adding commercial Planet tasking when a site warrants closer inspection. The methodology is the same class used in the Tonga crop-estimation programme: open-constellation baselines, change-detection alerts, and periodic analytical reports that a non-specialist decision-maker can act on. The output is not a legal instrument, but it is a documented, reproducible, independently verifiable record.
Honest limits of the method
Several constraints are worth stating plainly. Small dams, particularly run-of-river weirs with minimal storage, may produce no detectable inundation footprint and only a modest downstream flow change. The detection threshold for reservoir area is roughly 0.1 square kilometres for Sentinel-2 under good conditions; smaller impoundments are likely missed. Underground diversions or inter-basin transfers leave no surface water signature at all.
Attributing a downstream flow reduction to a specific upstream dam, rather than to drought or land-use change, requires careful baseline analysis and ideally a hydrological model. Satellite data alone cannot close that attribution loop with certainty. And SWOT, despite its capabilities, has a 21-day repeat: a rapid filling event or a sudden gate operation can occur between passes and be reconstructed only approximately. This is a monitoring system, not a continuous sensor.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m ground range × 10 m azimuth |
| Optical spatial resolution (Sentinel-2) | 10 m (VIS/NIR), 20 m (SWIR) |
| Very-high-resolution optical (Planet SuperDove) | 3–4 m |
| Altimetry water-surface precision (SWOT KaRIn) | Decimetric over calm water; ~21-day repeat |
| Sentinel-1 revisit at mid-latitudes | 6 days (two-satellite constellation) |
| Minimum detectable reservoir area (Sentinel-2) | ~0.1 km² under clear-sky conditions |
| Archive depth (Sentinel-1/2) | From 2014/2015; Landsat extends to 1972 |
| Cloud-penetration capability | SAR (Sentinel-1) all-weather; optical sensors cloud-limited |
| Downstream discharge proxy uncertainty | ~20–40% via satellite hydraulic geometry |
| Typical delivery format | GeoTIFF change layers, GIS-ready vector polygons, PDF analytical report |
Analytics Satellize can run
| Construction-phase alert | SAR coherence change detection on Sentinel-1 time series; decorrelation over active earthwork zones | Automated alert with scene date, affected area polygon and thumbnail, delivered within 48 hours of Sentinel-1 acquisition |
| Reservoir inundation time series | Sentinel-1 backscatter thresholding and Sentinel-2 modified normalised difference water index (MNDWI) mapping, weekly composites | GIS layer stack showing weekly water-extent polygons and area-time chart from first detection to current date |
| Impoundment elevation signal | SWOT KaRIn water-surface elevation extraction over reservoir and downstream reach, differenced against pre-construction baseline | Elevation time series chart with step-change annotation and estimated storage-volume range based on Copernicus DEM hypsometry |
| Downstream turbidity trend | Sentinel-2 band-ratio suspended sediment index on downstream reach, pre- and post-impoundment comparison | Tabular summary of mean turbidity index by month, with before/after boxplots and annotated map of sampling transects |
| Satellite-proxy discharge anomaly | River-width extraction from Sentinel-2 and Planet imagery using hydraulic geometry regression; SWOT slope-width Manning estimate where track coverage allows | Monthly discharge-proxy time series with uncertainty bounds, flagged against climatological baseline |
| Baseline archive review | Landsat 5/7/8/9 and Sentinel-1/2 retrospective analysis of river corridor from 1990 to present; land-cover and water-extent change classification | PDF report with decadal snapshots, annotated change map and narrative summary suitable for policy briefing |
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.