Transboundary river flow monitoring without in-country gauge access
When data-sharing agreements fail, satellite observations of water-surface slope, inundation width and total water storage can reconstruct cross-border river discharge independently of any upstream gauge network.
Sensors
- SWOT KaRIn: Ka-band radar interferometer measuring water-surface elevation to roughly 10 cm (rms) over 1 km² reaches and river width for channels wider than about 100 m; 21-day exact repeat with near-global coverage to ±78° latitude. The primary source of slope-based discharge inference without in-country gauges.
- Sentinel-1 SAR (C-band): 6- to 12-day revisit (constellation of two satellites when both operational); 10 m resolution in IW mode. Detects open water by backscatter contrast, mapping inundation extent and active channel width even under cloud cover or at night. Width-discharge rating curves can be calibrated from historical gauged periods.
- GRACE-FO: Monthly total water storage anomalies at roughly 300–400 km spatial resolution (the fundamental limit of the gravity-gradient measurement). Detects basin-scale storage change in equivalent water height to approximately 1–2 cm, integrating groundwater, soil moisture, snow and surface water. Useful for detecting sustained upstream abstraction over multi-month windows, not for event-scale discharge.
- Sentinel-2 MSI: 10 m visible and near-infrared bands; 5-day revisit at mid-latitudes with both satellites. Provides cloud-free water-surface mapping for width retrieval and channel planform, and detects turbidity shifts that can indicate upstream impoundment or large releases. Blocked by cloud, unlike SAR.
Why the gauge network fails at borders
Of the world's 310-odd international river basins, fewer than half have any active bilateral data-sharing arrangement for real-time hydrological data, according to assessments by the UN Economic Commission for Europe. In practice, the gap is worse: agreements exist on paper but transmission lapses during political disputes, precisely when the data matters most.
A downstream state facing reduced flows, unexpected floods or suspected upstream diversion has historically had no independent means of verification short of diplomatic protest or litigation. Satellite observations change that arithmetic, not completely, but enough to support treaty monitoring, early warning and negotiating evidence.
What a water-surface slope gives away
The SWOT satellite, launched in December 2022 as a joint NASA/CNES mission, carries the Ka-band Radar Interferometer (KaRIn), which measures water-surface elevation across a 120 km swath with a 20 km nadir gap. For rivers wider than roughly 100 m, SWOT can resolve reach-averaged slopes over segments of 10 km or more. Slope and width together feed Manning-based or hydraulic geometry discharge estimates without any in-country gauge.
The method is not frictionless. Slope errors grow on low-gradient rivers, where a 10 cm elevation uncertainty across a 10 km reach translates to a slope error of 10 µm/m, which is comparable to the signal itself on very flat systems such as the lower Amazon or Mekong delta. SWOT's 21-day repeat also means it captures a snapshot, not a continuous record. Interpolation between passes requires ancillary modelling or the width time-series from Sentinel-1 to fill gaps.
Building discharge from width alone: the Sentinel-1 workhorse
Before SWOT, width-discharge relationships were the primary satellite-based discharge proxy. Active-microwave backscatter from Sentinel-1 distinguishes open water from vegetated or dry floodplain at 10 m resolution regardless of cloud, making it the operational backbone for continuous monitoring. Hydraulic geometry theory, established in the published literature since Leopold and Maddock (1953), predicts that at-a-station width scales as a power function of discharge. Where a gauged period exists anywhere on the reach, even decades-old gauge records, that relationship can be calibrated and then applied to the satellite width time-series.
The honest limit: width saturates at bankfull discharge. Once the river overtops its banks, width expands onto the floodplain and the simple power-law breaks down. Sentinel-1's 6-day revisit (when both satellites are operational; single-satellite revisit is 12 days) is adequate for capturing seasonal cycles but will miss fast-rising flood pulses between passes. For channels narrower than about 30 m, the 10 m pixel size introduces significant width uncertainty.
GRACE-FO as a basin-scale audit
GRACE-FO measures month-to-month changes in the Earth's gravity field caused by mass redistribution, of which water is the dominant signal over most continental basins. The coarse spatial footprint (effective resolution around 300–400 km) means it cannot localise where in an upstream basin water is being stored or abstracted. What it can do is close the water balance: if precipitation input from satellite products (IMERG, ERA5) minus modelled evapotranspiration does not account for the storage change GRACE-FO observes, the residual implicates either measurement error or unaccounted human abstraction.
This residual approach has been used in published studies of the Indus, Tigris-Euphrates and Amu Darya basins to quantify groundwater depletion attributable to irrigation. Applied to a transboundary dispute, it provides a basin-scale consistency check rather than reach-level evidence. Courts and arbitration panels require precision; GRACE-FO provides direction and order of magnitude.
What satellite data can and cannot prove in a treaty dispute
This is the question buyers rarely ask explicitly but always need answered. Satellite discharge estimates carry uncertainty ranges that are honest and documentable: SWOT-based discharge estimates in published validation studies show root-mean-square errors of roughly 20–30% on gauged reaches, improving on rivers with strong slope signals. Width-based estimates from Sentinel-1 perform similarly where calibration data exist. These figures are defensible in technical annexes to diplomatic submissions, provided the uncertainty is stated.
What satellite data cannot prove, at least not directly, is intent. A dam closing its gates produces a detectable flow reduction downstream, visible in both width and SWOT elevation. Whether that closure violates a treaty depends on the treaty's specific flow-guarantee language, which is a legal question. The satellite record establishes the physical fact; the legal interpretation belongs to counsel. Satellize structures its analytical outputs explicitly to separate the physical observation from any causal inference, which matters when the output may be submitted as evidence.
A further limit: reservoir operations upstream may affect flow timing without changing total annual volume, and SWOT's 21-day repeat may not resolve sub-monthly manipulation. Combining SWOT passes with daily Sentinel-1 width retrievals is the best currently available approach to detect operational rather than structural changes.
Putting the sensors together operationally
A practical monitoring system for a downstream state combines three layers. The first is a near-real-time Sentinel-1 width time-series, updated every 6 to 12 days, with anomaly flags when width falls or rises beyond seasonal norms. The second is a SWOT pass archive, ingested at each 21-day repeat, converting slope-and-width pairs into discharge estimates with explicit uncertainty bounds. The third is a quarterly GRACE-FO water-balance audit that checks whether basin-scale storage trends are consistent with the reach-level observations.
Sentinel-2 optical imagery adds a fourth, opportunistic layer: cloud-permitting, it can detect turbidity plumes from dam releases, sediment pulses from new construction and channel planform shifts that precede detectable flow changes. None of these sensors was designed for treaty monitoring. That they can serve the purpose reflects the physical specificity of what they measure, not any diplomatic mandate.
Satellize applies this multi-sensor framework on open Copernicus and NASA archives, with commercial tasking added where higher-cadence coverage is needed. The analytical approach mirrors the water-balance methods developed for the Tonga crop-estimation programme, adapted here to discharge rather than soil moisture. Outputs are structured for ingestion by legal and technical teams, not just hydrologists.
Typical figures
| SWOT KaRIn spatial resolution | Reach-averaged: ~1 km² minimum water body; river width retrievable for channels >100 m |
| SWOT repeat cycle | 21-day exact repeat; near-global coverage to ±78° latitude |
| SWOT elevation accuracy | ~10 cm rms over 1 km² water surface (published mission specification) |
| Sentinel-1 spatial resolution | 10 m (IW mode); minimum detectable channel width ~30 m with acceptable uncertainty |
| Sentinel-1 revisit | 6 days (two-satellite constellation); 12 days single satellite |
| GRACE-FO spatial resolution | Effective ~300–400 km; total water storage anomaly to ~1–2 cm equivalent water height |
| GRACE-FO temporal resolution | Monthly gravity solutions; ~30-day latency for standard products |
| Sentinel-2 spatial resolution | 10 m (VIS/NIR); 5-day revisit at mid-latitudes; cloud-limited |
| Discharge estimate uncertainty | Typically 20–30% rms on validated reaches (SWOT-based); higher on low-gradient systems |
| Archive depth | Sentinel-1: from 2014; Sentinel-2: from 2015; GRACE/GRACE-FO: from 2002; SWOT: from early 2023 |
Analytics Satellize can run
| Reach-scale discharge time-series | Hydraulic geometry and Manning-based inversion using SWOT slope and Sentinel-1 width; uncertainty propagation from sensor error budgets | GeoJSON reach layer with discharge estimates and confidence intervals at each SWOT pass; CSV export for legal annexes |
| Sentinel-1 width anomaly alerts | Automated backscatter thresholding and active-channel delineation; z-score anomaly detection against rolling seasonal baseline | Near-real-time alert feed (6–12 day cadence) flagging statistically significant width departures; PDF summary for non-specialist briefing |
| Quarterly water-balance audit | GRACE-FO storage anomaly differenced against IMERG precipitation and ERA5-derived evapotranspiration; residual attributed to unaccounted storage change | Quarterly report with basin-scale water-balance closure table and residual trend chart |
| Dam-operation event detection | Change-point analysis on Sentinel-1 width and SWOT elevation time-series; corroborated by Sentinel-2 turbidity mapping where cloud-free | Event log with date, estimated flow reduction and supporting imagery; formatted for technical annex submission |
| Historical baseline reconstruction | Sentinel-1 width-discharge rating curve calibrated against any available historical gauge records; applied retrospectively to full Sentinel-1 archive from 2014 | 10-year discharge proxy record per reach; uncertainty-bounded trend analysis |
| Uncertainty-qualified evidence package | Structured separation of physical observation, retrieval uncertainty and causal inference; formatted to published hydrological reporting standards | Standalone technical report with methods appendix, suitable for submission to arbitration panels or inter-governmental technical committees |
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.