One of the more remarkable developments in water-utility tech over the last decade: a billion-dollar industry built almost entirely on satellites looking for water leaks from orbit. The technology is called synthetic-aperture radar (SAR), and the company most associated with it is Utilis (now part of Asterra). The technique is real, the precedent is real, and the data pipeline is freely available via Sentinel-1.
HydroDSS uses the same data pipeline, with two angles on it: SAR backscatter for leak signatures, and interferometric SAR (InSAR) for ground deformation. Both feed the platform's risk and evidence layers as named, citable inputs.
What synthetic-aperture radar actually sees.
Optical satellites see what photographs see — visible light reflecting off surfaces. SAR satellites do something different. They emit a microwave signal toward the ground and measure what reflects back. Three things matter:
- Backscatter intensity — how strongly the signal reflects, which depends on roughness, moisture, and material.
- Phase — the precise distance the signal travelled, accurate to a fraction of the wavelength (3-30 cm depending on band).
- Polarisation — how the signal's orientation changed on reflection, which depends on surface structure.
Optical imagery cares about colour. SAR cares about texture, moisture, and millimetre-scale distance. For water utilities, the second and third matter.
SAR backscatter for leak detection.
Sub-surface leaks change the moisture content of the soil above them. The change is too small to see optically. But SAR backscatter responds to soil moisture — saturated soil reflects differently than dry soil. With repeated SAR observations and the right processing, the anomaly stands out.
Utilis pioneered this. Their pipeline ingests Sentinel-1 imagery (10 m resolution, 6-day revisit, free) and surfaces probable leak locations to utility operators. Validation studies on multiple utilities have shown the technique can detect leaks before they surface — sometimes weeks before traditional minimum-night-flow analysis catches them.
The signal is noisy. SAR backscatter changes for many reasons — recent rainfall, vegetation cycles, agricultural irrigation, snow cover. So the technique works in combination with traditional methods, not as a replacement. In the platform's risk layer, SAR backscatter anomalies are one of several inputs to the operational-state evidence source. They feed precision-weighted Bayesian fusion alongside SCADA, weather, and asset records — never standing alone as a single point of failure.
InSAR for ground deformation.
A related but distinct technique: interferometric SAR, or InSAR. Where SAR uses backscatter intensity, InSAR uses phase — the precise distance the signal travelled. By comparing two SAR images taken from nearly the same satellite position at different times, the phase difference can resolve ground displacement at millimetre scale.
Why does this matter for water utilities? Differential ground settlement is one of the largest drivers of pipe failure. When the ground subsides unevenly, buried mains experience stress concentration; over years, that's the precursor to a break. InSAR sees subsidence at the same scale at which it stresses pipes — and across entire urban areas with weekly cadence.
Subsidence is also often caused by groundwater extraction. That's a direct feedback loop with the water utility's own operations: if a city is pumping aquifers, InSAR sees the resulting settlement, and HydroDSS can score asset-failure risk accordingly. The whole picture is in one decomposable evidence chain.
Three satellites, three bands.
HydroDSS integrates radar pipelines from three publicly accessible sources:
- Sentinel-1 (ESA Copernicus, free) — C-band, 5×20 m resolution, 6-day revisit, observing globally since 2014. The backbone of operational SAR for utilities.
- ALOS-2 PALSAR-2 (JAXA) — L-band (longer wavelength, better soil penetration), ~14-day revisit. The right band for sub-surface leak signal.
- NISAR (NASA/ISRO, launched 2024) — dual-band (L + S), free, 12-day revisit. The major new dataset entering the stack on its 2026 quarterly cadence.
All three feed via the same Google Earth Engine pipeline that handles climate reanalysis and surface-water dynamics — authoritative agency data, retrieved on its publishing cadence, fused into the platform's evidence layer.
What we layer on top.
The integration is deliberately non-glamorous. We don't run our own ML models on raw SAR data — we use the agency-published analytical products (analysis-ready data, radiometric terrain correction, time-series displacement). The credibility of the inputs traces back to ESA, JAXA, and NASA — not to an in-house algorithm of our own.
The output: an asset-failure-risk evidence source that surfaces InSAR-derived subsidence as a feature; a leak-signature evidence source that surfaces SAR backscatter anomalies. Both fold into the broader Bayesian fusion. Both decompose cleanly in the reasoning chain. Every value traces back to a named, citable satellite mission and a named, citable processing chain.
Active radar for water utilities is no longer experimental. It's production. The data is free (mostly). The processing is solved. The remaining work is integration — into existing risk models, existing alerting workflows, existing capital-renewal processes. That's where HydroDSS plays.