Irrigation canal network mapping as rural fibre routing corridors
Irrigation canal banks are graded, maintained access routes that already connect rural population centres across South Asia, the Middle East and sub-Saharan Africa. Satellite imagery and SAR extract their geometry, bank width and seasonal flood risk to support least-cost fibre route planning.
Sensors
- Maxar WorldView-3: Panchromatic resolution of 0.31 m and multispectral at 1.24 m enables extraction of individual canal bank edges, measurement of bank-top width to roughly 1 m accuracy, and detection of access tracks running along the berm. Revisit at mid-latitudes is typically 1 to 4.5 days depending on tasking priority.
- TanDEM-X: The global DEM product at 12 m posting (High Resolution DEM, commercially licensed) captures the cross-sectional relief of raised canal banks above surrounding field level, typically 0.5 to 3 m of relative height, supporting duct-burial depth planning and drainage gradient assessment. Absolute vertical accuracy is around 2 m LE90 over flat terrain.
- Sentinel-1 SAR (C-band, 5.405 GHz): Interferometric Wide swath mode delivers 10 m ground range resolution with 6-day repeat at the equator (3-day with both satellites). Open water returns near-zero backscatter in C-band, making Sentinel-1 the standard tool for mapping seasonal inundation extent along canal margins. Archive runs from 2014.
- Sentinel-2 MSI: Ten-metre visible and near-infrared bands support water-index mapping (NDWI, MNDWI) for canal delineation during cloud-free periods and vegetation mapping along banks. Five-day revisit with both Sentinel-2A and 2B. Cloud cover is the primary constraint in monsoon-affected regions, where SAR must substitute.
Why canal banks are already half a fibre route
Irrigation infrastructure is built to move water efficiently across flat agricultural land. That means straight alignments, minimal gradient change, and regular maintenance access, often a graded track running the full length of the bank. In the Indus basin, the Nile Delta, the Tigris-Euphrates plain and across large parts of sub-Saharan Africa, these networks extend for tens of thousands of kilometres, passing through precisely the rural areas where fibre penetration is lowest and where conventional road-based route surveys are expensive.
The planning problem is not whether canal banks are physically usable. It is knowing their geometry in detail before committing to a route survey. Bank-top width determines whether a duct can be buried safely clear of the canal structure. The gradient of the bank face determines drainage risk to a trench. And the seasonal behaviour of the canal itself determines whether a buried duct will sit in saturated ground for three months of the year. Satellite data answers all three questions from a desk, before anyone travels to the site.
Extracting canal geometry from sub-metre imagery
WorldView-3 panchromatic imagery at 0.31 m is sufficient to resolve the water surface, the inner bank face, the bank crest and the outer berm of most irrigation canals wider than roughly 2 m. Automated extraction uses a combination of spectral water indices to locate the water surface and edge-detection or active contour methods to delineate the bank edges. Bank-top width can then be measured directly from the orthorectified image. For canals carrying less than a few cubic metres per second, water surface width is typically 3 to 15 m; the maintained berm alongside may add another 3 to 6 m of usable surface.
TanDEM-X at 12 m posting adds the third dimension. Raised canal banks in flat terrain produce a clear ridge signature in the DEM. Differencing the bank crest elevation from the adjacent field level gives the embankment height, which matters for duct burial because a shallow embankment over clay-rich soils may have limited depth before hitting the puddled clay lining of the canal itself. The combination of optical edge detection and DEM cross-section analysis produces a canal centreline vector with associated bank-width and embankment-height attributes, deliverable as a GIS polyline layer.
What a floating roof gives away about seasonal inundation
The title is borrowed from oil-tank monitoring, but the principle applies here. Sentinel-1 SAR detects standing water by its near-zero backscatter return: smooth open water acts as a specular reflector and returns almost no signal to the sensor. In practice, thresholds for water detection in IW mode imagery are well established in the literature, and the ESA-supported HASARD and JRC Global Surface Water products both use Sentinel-1 time series for this purpose.
For canal routing, the relevant question is not whether the canal itself floods, but whether the surrounding field parcels adjacent to the bank become inundated during peak irrigation season or monsoon events. A buried duct on a bank that is surrounded by flooded fields for 90 days per year faces sustained hydrostatic pressure on any imperfect joint. Sentinel-1 time series from 2014 onwards allows characterisation of the worst-case inundation extent at each point along a proposed route, expressed as a percentage of years in which that bank segment was flanked by standing water. That is a directly usable input to duct-specification and jointing-interval decisions.
Cloud cover is not a constraint for SAR. This matters enormously in Bangladesh, the Punjab and the Nile Delta during the kharif or summer irrigation season, when optical imagery may be cloud-obscured for weeks at a time.
From geometry to least-cost path
Canal network extraction produces a graph: nodes at junctions and offtakes, edges representing bank segments with attributes for width, embankment height, inundation frequency and proximity to population centres. Least-cost path analysis across this graph, weighted by those attributes alongside conventional terrain cost factors, identifies the route that minimises civil works while maximising population coverage.
The honest caveat is that satellite data cannot detect subsurface conditions. Canal banks in older irrigation systems often contain informal drainage channels, animal burrows or root systems that create voids. A satellite-derived route plan reduces the survey area from hundreds of kilometres to the specific segments that score well on all measurable criteria. Ground truthing of those segments remains necessary before civil works begin. The satellite product narrows the problem; it does not eliminate fieldwork.
Coverage, limits and what the archive cannot tell you
Sentinel-1 and Sentinel-2 cover the relevant geographies globally, with Sentinel-1 archive depth back to 2014 providing a decade of seasonal behaviour. WorldView-3 tasking is available over most of South Asia, the Middle East and Africa, though cloud-free acquisition in tropical regions requires patience and sometimes multiple passes. TanDEM-X global coverage at 12 m is commercially available through Airbus Defence and Space.
Resolution limits matter for small canals. Tertiary and quaternary irrigation channels below roughly 2 m water-surface width are not reliably detectable in 10 m Sentinel-2 imagery and may be marginal even in WorldView-3 panchromatic. In densely cultivated areas with tree cover along banks, canopy obscures the bank edge in optical imagery and SAR backscatter becomes ambiguous between flooded soil and wet vegetation. In those situations, lidar would be the preferred sensor, but no open-access spaceborne lidar at the required density currently covers agricultural canal networks. GEDI (Global Ecosystem Dynamics Investigation) provides canopy height globally but at 25 m footprint spacing, which is not suited to narrow linear feature mapping.
Satellize runs this analysis on Sentinel open data combined with commercial WorldView-3 and TanDEM-X tasking on client licence. The workflow is similar in structure to the crop-area estimation approach used in the Kingdom of Tonga programme, adapted for linear infrastructure rather than field-parcel classification.
What the output actually looks like
The primary deliverable is a GIS polyline dataset of the canal network within the study area, attributed with bank-top width (in metres, with a stated uncertainty of roughly plus or minus 1 m for WorldView-3 derived measurements), embankment height above adjacent ground, inundation frequency (percentage of Sentinel-1 observation epochs showing flanking water), and a composite civil-works cost score. A secondary raster layer shows seasonal maximum inundation extent at 10 m resolution, derived from the Sentinel-1 time series.
These layers feed directly into route-planning tools such as QGIS cost-distance analysis or proprietary network planning platforms. The format is standard: GeoPackage or shapefile for vectors, GeoTIFF for rasters, with metadata conforming to ISO 19115. Delivery timescale from tasking confirmation to first-pass output is typically four to eight weeks for a study area of 50,000 square kilometres, depending on cloud-free acquisition availability.
Typical figures
| Canal bank edge resolution (optical) | 0.31 m panchromatic (WorldView-3); bank-top width measurable to approx. ±1 m |
| Elevation model posting (TanDEM-X) | 12 m; vertical accuracy ~2 m LE90 over flat terrain |
| SAR inundation mapping resolution | 10 m ground range (Sentinel-1 IW mode) |
| Sentinel-1 revisit | 6 days at equator (3 days with both satellites); archive from 2014 |
| Sentinel-2 revisit | 5 days (combined A+B); 10 m visible/NIR bands |
| Minimum detectable canal width (optical) | ~2 m water surface in WorldView-3 panchromatic; ~10 m in Sentinel-2 |
| Inundation frequency archive depth | Up to 10 years (Sentinel-1 from 2014) |
| Study area coverage | No hard limit; typical engagement 10,000–100,000 km² |
| Delivery formats | GeoPackage / shapefile (vectors), GeoTIFF (rasters), ISO 19115 metadata |
| Indicative delivery timescale | 4–8 weeks from tasking confirmation (cloud-free acquisition dependent) |
Analytics Satellize can run
| Canal network centreline and junction graph | Spectral water-index edge detection on WorldView-3 and Sentinel-2; active contour / morphological thinning | GIS polyline layer with canal ID, length, order and junction nodes |
| Bank-top width and embankment height profile | WorldView-3 orthorectified edge measurement combined with TanDEM-X cross-section analysis | Attributed polyline layer; per-segment width and height statistics in tabular form |
| Seasonal inundation frequency map | Sentinel-1 C-band backscatter thresholding over multi-year time series; water body classification per epoch | 10 m GeoTIFF raster of inundation frequency (% of epochs) and worst-case extent shapefile |
| Duct-burial risk score per route segment | Composite scoring of bank width, embankment height and inundation frequency; configurable weighting by client specification | Attributed polyline layer with risk score and component sub-scores; PDF summary report |
| Least-cost fibre path recommendation | Graph-based cost-distance analysis on canal network with civil-works cost weights; population node demand from ancillary datasets | Recommended route GIS layer with cost breakdown; alternative routes ranked by score |
| Cloud-gap-filled inundation composite | SAR-optical fusion: Sentinel-1 fills Sentinel-2 cloud gaps using decision-tree or random-forest classifier trained on clear-sky overlap periods | Monthly inundation composites as GeoTIFF stack for monsoon-affected study areas |
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.