Urban utilities corridor planning with satellite data
Satellite-derived land cover, elevation models and nighttime-light data can screen feasible utility corridors before a survey crew sets foot on the ground, cutting wasted mobilisation in rapidly urbanising areas where base maps are unreliable or absent.
Sensors
- Sentinel-2 MSI: 13 spectral bands at 10–60 m resolution, 5-day revisit at the equator with both satellites. The red-edge and SWIR bands separate built surfaces, bare soil, vegetation and standing water reliably enough for corridor-scale land-cover classification. Free and open.
- TanDEM-X DSM: Global digital surface model at 12 m posting (90 m public release; 12 m commercial). Vertical accuracy is typically better than 2 m absolute in open terrain but degrades under dense tree canopy and in radar shadow on steep slopes. The primary elevation source for slope and drainage analysis in corridor screening.
- VIIRS Day/Night Band (DNB): Panchromatic low-light sensor on Suomi-NPP and NOAA-20, roughly 750 m pixel size. Detects artificial light at night down to about 2 × 10⁻⁹ W cm⁻² sr⁻¹. Used to map electrification extent and identify unserved dark zones that justify new distribution-line corridors.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, daily revisit capacity. Used to sharpen corridor candidates identified by coarser sensors: resolves individual structures, compound walls and road widths before ground survey is commissioned. Commercial tasking, not free.
Why satellite data enters utilities planning at all
Utilities engineers designing a water main or a medium-voltage distribution line need corridor widths, slope profiles, crossing points and land-use constraints. Traditionally that information comes from cadastral surveys and topographic maps. In cities growing at 3–5% per year across sub-Saharan Africa, South Asia and parts of Latin America, those maps either do not exist or were last updated before the current built footprint appeared. Sending survey teams into unmapped peri-urban areas without any prior screening is expensive and occasionally dangerous.
Satellite data does not replace the survey. It narrows the problem. A corridor screening exercise using Sentinel-2 land cover and a TanDEM-X slope model can eliminate obviously infeasible routes (steep gullies, dense built fabric, seasonal wetlands) and rank remaining options before anyone leaves the office. That is the honest scope of what remote sensing contributes at this stage of a project.
What a floating roof gives away: reading land cover for corridor constraints
Sentinel-2's 10 m bands (blue, green, red, near-infrared) and its 20 m red-edge and SWIR bands support supervised classification of built surfaces, bare soil, low vegetation, tree canopy and water. In practice, classification accuracy in dense informal settlements sits at roughly 80–88% overall in published studies, which is sufficient for screening but not for detailed design. The practical output is a constraint map: pixels classified as built fabric or water are obstacles; pixels classified as bare soil or low scrub are candidate corridor zones.
SWIR reflectance is particularly useful for distinguishing impervious surfaces from dry bare soil, a distinction that matters when routing a trench through a peri-urban fringe. The red-edge bands help separate sparse tree cover from open grassland, which affects wayleave negotiation. Neither distinction is possible with a standard RGB aerial photograph unless it was taken recently and at high resolution.
Slope, drainage and the TanDEM-X surface model
Gravity governs sewer design. A gravity sewer needs a minimum gradient, typically 1:150 to 1:300 depending on diameter, and cannot cross a ridge without a pumping station. The TanDEM-X 12 m DSM provides enough vertical resolution to compute slope profiles and catchment boundaries at corridor scale. Least-cost path algorithms applied to the slope-derived cost surface produce candidate alignments that respect gradient constraints before any surveyor visits the site.
The honest caveat is that the DSM captures the surface, not the ground. In areas with dense tree canopy, the model surface may sit 10–20 m above actual ground level. In those zones, TanDEM-X slope estimates are unreliable and the corridor analysis must flag them as requiring LiDAR or ground-truthed profiles. Similarly, the 12 m posting means the model cannot resolve a 3 m retaining wall or a culvert. Those features are discovered by survey, not by satellite.
Nighttime light as a proxy for electrification gaps
VIIRS DNB monthly composites, produced by the Earth Observation Group at the Colorado School of Mines and distributed via eogdata.mines.edu, show radiance values that correlate with grid electricity access at neighbourhood scale. A settlement that appears dark in VIIRS composites across multiple months is almost certainly unconnected to the grid. That is a straightforward input to a utility's capital planning exercise: where are the dark zones, how large are they, and which existing grid infrastructure is closest?
The 750 m pixel size means VIIRS cannot distinguish a single unlit street from an unlit district. It is a planning-scale instrument. A cluster of dark pixels covering several square kilometres justifies commissioning a Pléiades Neo tasking to count structures and estimate connection density before designing a distribution line. The two sensors work in sequence, not in competition.
The resolution gap and what it means for project workflow
The gap between what satellites observe and what utilities engineers need for detailed design is real and should not be papered over. A 10 m Sentinel-2 pixel cannot tell you whether a compound wall is 1.5 m or 3 m high, whether a road reserve is 6 m or 10 m wide, or whether an apparent open space is public land or a private yard. Pléiades Neo at 30 cm narrows that gap considerably but still cannot replace a measured survey for design purposes.
The correct workflow positions satellite analysis as a pre-survey screening stage. Satellite data identifies three or four candidate corridors from a dozen possible routes. Ground survey then characterises those three or four in detail. That sequence typically reduces survey mobilisation costs by eliminating routes that would have been rejected anyway. Satellize structures its corridor-screening analytics around this workflow, delivering ranked corridor options with associated constraint layers rather than a single recommended route.
For clients with no existing base map at all, the Pléiades Neo tasking adds a current-condition photographic record that can anchor subsequent ground observations. It is not a cadastral survey, but it is a defensible starting point for a project that would otherwise begin with nothing.
Archive depth and change detection as a planning asset
Sentinel-2 archive runs from June 2015 for Sentinel-2A and from March 2017 for Sentinel-2B. That eight-plus-year record allows a planner to see how fast the built footprint has expanded along a proposed corridor, which matters for way-leave negotiation and for forecasting future demand. A corridor that was open scrubland in 2017 and is now partially built tells a different story than one that has been stable agricultural land for a decade.
VIIRS DNB archive extends to 2012 on the Suomi-NPP platform. Comparing 2012 and current nighttime radiance along a proposed distribution corridor shows whether electrification demand is growing or static, and at roughly what rate. Neither data set replaces a demand survey, but both reduce the uncertainty that planners carry into that survey.
Typical figures
| Sentinel-2 spatial resolution | 10 m (visible/NIR), 20 m (red-edge/SWIR), 60 m (atmospheric bands) |
| Sentinel-2 revisit | 5 days at equator (both satellites combined); 2–3 days at mid-latitudes |
| TanDEM-X DSM posting | 12 m (commercial); 90 m (public release). Vertical accuracy typically <2 m absolute in open terrain |
| VIIRS DNB pixel size | ~750 m. Minimum detectable radiance ~2 × 10⁻⁹ W cm⁻² sr⁻¹ |
| Pléiades Neo resolution | 30 cm panchromatic, 1.2 m multispectral; daily revisit capacity |
| Sentinel-2 archive depth | From June 2015 (2A) and March 2017 (2B) |
| VIIRS DNB archive depth | From 2012 (Suomi-NPP); NOAA-20 from 2018 |
| Land-cover classification accuracy (Sentinel-2, peri-urban) | Typically 80–88% overall accuracy in published studies; lower in dense informal settlements |
| Delivery formats | GeoTIFF, GeoPackage, Shapefile, COG; optionally via WMS/WFS for GIS integration |
| Cloud limitation | Sentinel-2 optical; cloud cover renders individual acquisitions unusable. Multi-date compositing mitigates but does not eliminate the problem in persistently cloudy climates |
Analytics Satellize can run
| Land-cover constraint map | Supervised classification of Sentinel-2 multispectral imagery (random forest or support vector machine on spectral bands including red-edge and SWIR) | GeoTIFF raster and polygon layer classifying built fabric, bare soil, vegetation, water and mixed/uncertain zones across the study area |
| Slope and drainage cost surface | Slope, aspect and flow-accumulation analysis derived from TanDEM-X 12 m DSM using standard DEM processing (ArcGIS or GDAL/GRASS workflows) | Raster cost surface and vector drainage network; flagged zones where canopy cover makes DSM-derived slopes unreliable |
| Least-cost corridor candidates | Least-cost path analysis combining land-cover constraint map and slope cost surface, with configurable weighting for different utility types (gravity sewer vs. overhead line vs. buried fibre) | Ranked vector corridor alignments (typically three to five options) with per-segment constraint summaries in PDF and GIS format |
| Electrification-gap map | VIIRS DNB monthly composite analysis; radiance thresholding to delineate persistently dark zones correlated with settlement footprint from Sentinel-2 classification | Polygon layer of unelectrified settlement clusters with estimated structure counts from Pléiades Neo where tasked |
| Urban growth trajectory along corridor | Multi-date Sentinel-2 land-cover change detection (2015/17 to present) using post-classification comparison or spectral change vector analysis | Time-series chart and change map showing built-footprint expansion rate within a user-defined corridor buffer |
| Pléiades Neo corridor strip | Commercial tasking of Airbus Pléiades Neo along selected candidate corridor; orthorectification and pansharpening to 30 cm | Orthorectified image strip in GeoTIFF with associated RPC metadata; suitable as base layer for ground-survey 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.