Irrigation canal and pipeline leakage detection by InSAR soil-moisture anomaly
Persistent-scatterer and small-baseline InSAR time-series detect millimetre-scale ground deformation caused by soil saturation or void formation next to leaking buried irrigation infrastructure, confirmed by L-band backscatter and passive microwave soil-moisture products.
Sensors
- Sentinel-1 IW SLC (C-band, 5.6 cm): Interferometric Wide swath in Single Look Complex mode; ground range resolution roughly 5 × 20 m, 12-day repeat per satellite (6-day with both A and B), free global archive from 2014. C-band phase is sensitive to surface and near-surface moisture changes but attenuates in dense vegetation and is susceptible to atmospheric delay artefacts requiring correction.
- ALOS-2 PALSAR-2 (L-band, 23.6 cm): L-band penetrates dry soil 0.5–2 m depending on texture and moisture content, making it more sensitive to subsurface void formation and deeper saturation fronts than C-band. Stripmap mode delivers 3 × 3 m resolution; ScanSAR mode 60–100 m at wider swath. Repeat is 14 days; data are not free but archived scenes are licensable.
- NISAR L-band SAR (launch 2024, ISRO/NASA): Dual-frequency (L and S band) with 12-day global repeat and free open data policy. L-band channel provides subsurface sensitivity comparable to PALSAR-2; the planned systematic global acquisition strategy will build dense time-series suitable for PS-InSAR within a year of operations. Specifications are pre-launch; performance figures are mission-design values, not yet validated at scale.
- SMAP L3 passive microwave soil moisture: Provides daily global surface soil-moisture estimates (top ~5 cm) at 36 km passive resolution and 9 km active-passive blend. Too coarse to pinpoint a canal leak, but useful for distinguishing a localised InSAR anomaly from a regional wetting event that would affect the whole scene. Freely available via NASA Earthdata.
What a leaking canal does to the ground beneath it
When water escapes from a buried irrigation pipe or seeps through a cracked canal lining, it saturates the surrounding soil. Saturated soil is denser than air-filled soil and slightly less dense than the mineral skeleton it has displaced. The net mechanical effect depends on soil type: in loose alluvial material, saturation causes consolidation and millimetre-scale subsidence; in expansive clays, it causes uplift of a similar magnitude; in poorly compacted backfill around a pipe, it can cause void migration and more abrupt settlement. All three signatures are, in principle, detectable by InSAR phase time-series before any surface expression is visible to a camera.
The physical link between moisture and radar phase works through two mechanisms. First, the dielectric constant of wet soil is much higher than dry soil, which shifts the effective scattering surface and alters the round-trip path length measured by the radar. Second, and more useful for buried infrastructure, the mechanical deformation of the ground surface caused by subsurface saturation or void growth produces a genuine displacement signal in the interferometric phase. Separating these two contributions is non-trivial and is one reason corroborating datasets matter.
PS-InSAR and SBAS: the two main processing routes
Persistent-scatterer InSAR (PS-InSAR) identifies individual pixels that maintain phase coherence across a long stack of interferograms, typically man-made structures, rock outcrops or coarse gravel. Along an irrigation canal corridor, concrete culverts, valve chambers and road crossings often serve as natural persistent scatterers. A time-series of displacement at these points, resolved to roughly 1 mm per year in ideal conditions, can reveal localised subsidence bowls centred on leak points. The method requires at least 20 to 30 acquisitions to be statistically reliable; with Sentinel-1 at 6-day repeat, that is achievable in three to six months.
Small-baseline subset (SBAS) processing uses distributed scatterers rather than point targets, averaging phase over small spatial windows to improve signal-to-noise. It performs better in agricultural or semi-vegetated terrain where point scatterers are sparse, but at the cost of spatial resolution: effective resolution in SBAS outputs is typically 20 to 100 m depending on window size and coherence. Neither method is immune to atmospheric delay artefacts. Tropospheric water vapour can introduce apparent displacement signals of several centimetres, easily masking or mimicking a real leak signal. Correction using ERA5 reanalysis or GACOS atmospheric delay maps is standard practice and should be applied before drawing conclusions.
A realistic detection threshold for a single leak event is deformation of 3 to 5 mm accumulated over several acquisition cycles, provided atmospheric correction is applied and the area contains coherent scatterers. Smaller or slower leaks accumulate detectable signal over months rather than weeks. This is a monitoring tool, not an alarm system.
L-band adds depth; passive microwave adds context
C-band radar (Sentinel-1, 5.6 cm wavelength) interacts primarily with the top 5 to 10 cm of dry soil, shallower still when the soil is wet. L-band (PALSAR-2, NISAR, 23.6 cm wavelength) penetrates to 0.5–2 m in typical agricultural soils, which is the depth at which many buried irrigation mains and secondary distribution pipes sit. For subsurface void detection, L-band backscatter change over time is a more direct indicator than C-band phase alone. A pipe failure that creates a growing void will produce a progressive decrease in backscatter as the void scatters energy away from the sensor.
SMAP passive microwave data, despite its coarse 36 km resolution, provides a scene-wide moisture baseline. If an InSAR anomaly appears at the same time as a region-wide SMAP wetting signal, the most parsimonious explanation is rainfall or flood irrigation, not a point leak. If SMAP shows dry regional conditions while a localised InSAR anomaly persists or grows, the case for a structural leak strengthens considerably. This two-dataset logic is simple but effective at reducing false-positive investigations.
Resolution floors and the corroboration requirement
The spatial resolution limit is the most important honest caveat for this technique. Sentinel-1 IW SLC has a ground range resolution of roughly 5 m in azimuth but 20 m in range; after multi-looking for phase quality, effective resolution in most InSAR products is 20 to 40 m. A leak from a 200 mm diameter buried pipe affects a soil volume that may be only 2 to 5 m across at the surface in its early stages. That footprint is sub-pixel. The signal only becomes detectable once the saturated zone has grown to occupy a significant fraction of a resolution cell, which may take weeks to months depending on soil permeability and leak rate.
Features narrower than roughly 20 m are ambiguous without corroborating data. The practical workflow for a confirmed leak location therefore combines InSAR anomaly mapping to identify candidate zones, followed by high-resolution optical imagery (to look for surface ponding, vegetation stress or disturbed soil) and, where resources allow, ground-penetrating radar or electromagnetic induction surveys to confirm subsurface saturation. Satellite InSAR narrows the search area; it rarely eliminates the need for field verification entirely.
Canal lining materials also affect detectability. Concrete-lined canals with intact linings leak less and produce smaller signals. Unlined earthen canals leak continuously, which means the surrounding soil may already be near saturation, reducing the differential signal from an incremental increase. The most detectable scenario is a sudden failure in a previously dry corridor, such as a buried main that ruptures under a road or field.
Practical workflow for an irrigation authority
A credible monitoring programme starts with a baseline InSAR velocity map derived from one to two years of archived Sentinel-1 data. This establishes which parts of the canal network are already showing slow deformation and which are stable. New acquisitions are then stacked against this baseline; anomalies that exceed the local noise floor by more than two standard deviations over three or more consecutive epochs are flagged for review.
Priority ranking of flagged zones can incorporate ancillary data: pipe age and material from asset registers, soil type from national soil surveys, and historical maintenance records. A 30-year-old asbestos cement main in expansive clay, flagged by InSAR, warrants faster physical inspection than a recently relaid HDPE pipe in sandy loam showing a marginal signal.
Satellize has run similar time-series analytics workflows in agricultural contexts, including the Kingdom of Tonga crop-estimation programme, and can configure InSAR processing chains on open Sentinel-1 and ALOS-2 archives for irrigation authorities seeking to prioritise inspection budgets without deploying ground sensors across hundreds of kilometres of buried infrastructure. The starting point is a scene-selection and coherence assessment over the client's network, which takes a few days of processing rather than months of fieldwork.
What the method cannot do
InSAR does not measure soil moisture directly. It measures phase, which is a function of geometry, dielectric properties and atmospheric delay all mixed together. Attributing a phase anomaly to a leak rather than to a passing weather system, a passing heavy vehicle, or seasonal thermal expansion of a concrete structure requires careful multi-temporal analysis and, in ambiguous cases, independent corroboration.
Dense vegetation breaks coherence. A canal corridor lined with mature trees or passing through rice paddies during the growing season will lose coherence in C-band and degrade significantly even in L-band. Monitoring in these environments requires either longer baselines between acquisitions (accepting lower temporal resolution) or a shift to SBAS with aggressive spatial averaging. Revisit rates also matter: a fast-developing void that causes ground collapse within days may not be captured between 6-day or 12-day acquisitions at a useful stage. InSAR is best suited to slow, chronic leakage rather than catastrophic pipe burst.
Typical figures
| Spatial resolution (Sentinel-1 IW SLC, after multi-look) | 20–40 m effective for InSAR products; 5 × 20 m native pixel |
| Spatial resolution (ALOS-2 PALSAR-2 Stripmap) | 3 × 3 m native; typically 10–20 m after InSAR multi-look |
| Revisit period | 6 days (Sentinel-1 A+B combined); 14 days (ALOS-2); 12 days (NISAR, planned) |
| Radar frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1); L-band 1.27 GHz / 23.6 cm (ALOS-2, NISAR) |
| Minimum detectable displacement (PS-InSAR, ideal conditions) | ~1 mm yr⁻¹ velocity; ~3–5 mm cumulative per event after atmospheric correction |
| Minimum detectable spatial footprint | ~20 m; features smaller than one resolution cell require corroborating optical or field data |
| Soil-moisture context product resolution (SMAP) | 36 km (passive); 9 km (active-passive blend) |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A); continuous to present |
| Latency from acquisition to processed interferogram | 1–5 days for near-real-time processing pipelines; hours for pre-configured automated chains |
| Delivery formats | GeoTIFF displacement maps, shapefiles of anomaly zones, time-series CSV per point, PDF inspection-priority report |
Analytics Satellize can run
| Baseline InSAR velocity map | PS-InSAR or SBAS time-series over 1–2 year archive stack (StaMPS, MintPy or equivalent open-source chain) | GeoTIFF mean line-of-sight velocity map with uncertainty layer; stable-zone mask identifying coherent scatterers along canal corridor |
| Anomaly detection and change alerts | Sequential interferogram stacking with ERA5 or GACOS atmospheric correction; two-sigma threshold flagging against rolling baseline | Shapefile of flagged anomaly zones updated per acquisition cycle; optional email or API alert feed |
| L-band backscatter change map | Multi-temporal SAR backscatter differencing (ALOS-2 or NISAR) to detect dielectric change consistent with subsurface saturation or void growth | GeoTIFF change-intensity raster overlaid on canal network; ranked list of candidate leak zones by magnitude of change |
| SMAP moisture context layer | NASA SMAP L3 passive daily composites co-registered to scene; regional moisture index compared against InSAR anomaly timestamps | Time-series chart per anomaly zone showing SMAP regional moisture vs. local InSAR displacement; flag for anomalies inconsistent with regional wetting |
| Inspection-priority ranking | Weighted scoring combining InSAR displacement magnitude, L-band backscatter change, SMAP context, pipe age and soil-type data from client asset register | Tabular priority list with map; each candidate zone scored and annotated with recommended inspection method (optical follow-up, GPR survey, or field visit) |
| Multi-year deformation trend report | Long-arc SBAS time-series over full Sentinel-1 archive (2014 to present) to identify chronic slow-leak zones predating current monitoring period | PDF report with annotated displacement time-series plots for up to 20 priority network segments; suitable for asset management planning |
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.