Informal tenure boundary delineation for land formalisation programmes

Very-high-resolution optical imagery and drone-derived surface models can map de-facto plot boundaries in informal settlements, giving land-titling programmes a geometric starting point without a full cadastral survey.

Sensors

Pleiades Neo: 30 cm native panchromatic resolution, 4-band multispectral at 1.2 m. Stereo and tri-stereo tasking produces DSMs with 50 cm horizontal accuracy and roughly 30 cm vertical RMSE over flat terrain, sufficient to resolve wall-top profiles and height discontinuities between plots.
WorldView-3: 31 cm panchromatic, 1.24 m multispectral, plus 8 SWIR bands at 3.7 m. The SWIR stack helps distinguish roofing materials and bare-earth pathways from vegetation, which aids boundary tracing in settlements with significant tree cover. Stereo pairs support DSM generation comparable to Pleiades Neo.
Drone orthophotography (UAV): Ground-deployed UAV surveys with consumer or survey-grade cameras routinely deliver 2–5 cm ground sampling distance orthophotos and structure-from-motion DSMs with sub-10 cm vertical accuracy. This is the highest-resolution option and the only one that resolves shared-wall boundaries reliably, but coverage per flight is limited to a few square kilometres per day.
SkySat: 50 cm native resolution, up to 12 frames per day over a target area. Useful for rapid change detection between survey epochs, confirming whether a boundary feature has been altered since the primary mapping flight. Less suited to DSM generation than stereo-tasked Pleiades Neo or WorldView-3.

What a fence line looks like from 600 kilometres up

A cadastral boundary in a formal suburb is a legal abstraction. In an informal settlement it is a physical object: a row of breeze blocks, a line of corrugated sheeting, a worn footpath that everyone treats as a property edge. At 30 cm resolution, Pleiades Neo and WorldView-3 resolve these features directly. A single-course brick wall casts a shadow roughly equal to its height; at midday in low latitudes that shadow is narrow, but at a solar elevation of 40 degrees a 1.5 m wall casts a shadow about 1.8 m long, which is detectable at sub-metre resolution even without a DSM.

The structure-from-motion DSM adds the third dimension. By matching thousands of tie-points across overlapping stereo images, photogrammetric processing produces a dense point cloud from which a digital surface model is derived. Height discontinuities of 0.5 m or more between adjacent roof planes, or between a roof plane and a pathway, mark probable plot edges. The combination of spectral contrast at ground level and height discontinuity in the DSM is more reliable than either source alone.

Where the method works and where it does not

The approach performs well where boundaries are open features: pathways, low walls, fences, or gaps between structures. In settlements with single-digit-metre lane widths and two-storey construction, the lanes themselves are often in shadow for most of the day, and the DSM collapses adjacent roof planes into a single surface. Shared internal walls between two structures that share a party wall are essentially invisible from above. In those conditions, satellite imagery can identify the block boundary but not the internal plot subdivision.

Cloud cover is a persistent constraint in tropical cities, which is where the majority of rapid informal urbanisation is occurring. A single usable Pleiades Neo collect requires a cloud fraction below roughly 20 percent over the target area. In cities with persistent convective cloud, a tasking campaign may need to run across several weeks to achieve full coverage. Drone surveys sidestep the cloud problem but introduce their own: airspace permissions, battery endurance, and the logistical challenge of flying safely over densely populated areas.

Accuracy figures from published photogrammetric studies of informal settlements suggest horizontal boundary placement errors of 0.3–1.0 m for drone-derived orthophotos and 0.5–2.0 m for satellite stereo DSMs, depending on ground control point density and terrain relief. Those figures are adequate for participatory mapping but fall short of the sub-20 cm accuracy that formal cadastral standards in many jurisdictions require. This is an honest ceiling, not a defect of the method.

Turning image features into a boundary layer

The processing pipeline has four steps. First, the stereo imagery is processed photogrammetrically to produce an orthorectified image and a DSM. Second, edge-detection algorithms (Canny, or morphological operators on the DSM gradient) extract candidate linear features. Third, a human analyst or a trained classifier filters those candidates against the spectral signature of pathway surfaces, wall-top materials, and vegetation lines. Fourth, the resulting vector layer is exported as a draft boundary dataset and brought into a participatory verification session with residents.

That last step is not optional. Satellite-derived boundaries are a geometric hypothesis. Residents confirm, correct, and annotate them with tenure history that no image can supply. The most effective programmes treat the satellite layer as a time-saving first draft, not a finished cadastral product. Several published land-administration projects in sub-Saharan Africa and South and Southeast Asia have used this model, reducing field survey time by 40–70 percent compared with full ground measurement, though those figures vary considerably with settlement density and community engagement quality.

Pathway networks as a proxy for occupancy structure

Pathways deserve separate attention. In informal settlements, the footpath network is often the most legible boundary system visible from above. Paths are bare earth or compacted aggregate, spectrally distinct from roofing materials and from vegetation. At 30 cm resolution they are individually traceable even when only 0.8–1.0 m wide. Extracting the path skeleton gives a rough Voronoi partition of the settlement that approximates plot ownership zones, even before wall lines are mapped.

This path-skeleton approach has been used in academic mapping exercises in Nairobi, Dar es Salaam, and Mumbai, among others, and is well documented in the remote sensing literature. It is particularly useful as a rapid reconnaissance product: a path network can be extracted in a few hours of processing and gives programme managers a first sense of block structure before detailed boundary work begins.

Data vintage, change, and the titling timeline

Land-titling programmes take years. A boundary survey conducted in year one may be partially obsolete by the time titles are issued in year three, because residents subdivide plots, extend structures, and negotiate boundary adjustments continuously. Multi-epoch satellite coverage, using SkySat for quarterly or monthly revisits between the primary mapping flights, allows analysts to flag areas where significant structural change has occurred and prioritise re-survey. This is cheaper than resurveying the entire settlement and more defensible than issuing titles against stale geometry.

Archive depth matters here. Pleiades (the predecessor to Pleiades Neo) has been collecting since 2011, and WorldView-3 since 2014. In many cities, usable archive imagery predates the current settlement extent, which lets analysts reconstruct how the settlement grew, identify which plots were established earliest, and provide a temporal record that can inform adverse-possession or first-occupancy claims.

What a programme manager should specify before commissioning imagery

Four parameters determine whether a satellite-based boundary mapping exercise will be fit for purpose. The first is target resolution: 30 cm or better for boundary tracing; 50 cm is marginal; anything coarser than 70 cm will miss most boundary features in dense settlements. The second is sun angle at collect time: a solar elevation between 35 and 55 degrees gives useful shadow length without excessive shadow overlap between adjacent structures. The third is stereo geometry: a base-to-height ratio of 0.4–0.6 is standard for Pleiades Neo and WorldView-3 stereo pairs and produces a DSM adequate for wall detection. The fourth is ground control: at least four well-distributed GCPs per square kilometre, collected by GNSS, are needed to bring horizontal accuracy below 1 m.

If drone surveys are part of the plan, the UAV flight should be designed for 80 percent forward overlap and 60 percent side overlap at minimum, with GCPs at the same density. Processing in Agisoft Metashape, Pix4D, or OpenDroneMap, all of which have published accuracy benchmarks, will produce results that are directly comparable with satellite-derived products and can be merged into a single boundary dataset. Commissioning a pilot block of 10–20 hectares before scaling is strongly advisable: it surfaces airspace, community access, and processing issues before they affect programme timelines.

Typical figures

Best available spatial resolution (satellite)	30 cm GSD (Pleiades Neo panchromatic, WorldView-3 panchromatic)
Best available spatial resolution (drone)	2–5 cm GSD, survey-grade UAV with nadir camera
DSM vertical accuracy (satellite stereo)	0.3–0.5 m RMSE over flat terrain with adequate GCPs (Pleiades Neo / WorldView-3 published figures)
DSM vertical accuracy (drone SfM)	5–10 cm RMSE with GCPs; degrades to 15–30 cm without
Minimum detectable boundary feature	Wall or fence of approximately 0.5 m height and 0.3 m width at 30 cm resolution with favourable sun angle
Satellite tasking revisit	Pleiades Neo: 1–2 day revisit at mid-latitudes; SkySat: up to 12 collects per day
Cloud constraint	Usable collect requires <20% cloud fraction over target; tropical cities may require multi-week tasking campaign
Archive depth	Pleiades (predecessor): from 2011; WorldView-3: from 2014; Pleiades Neo: from 2021
Horizontal boundary placement accuracy	0.3–1.0 m (drone); 0.5–2.0 m (satellite stereo), depending on GCP density
Delivery formats	GeoTIFF orthophoto, GeoTIFF DSM, Shapefile / GeoPackage boundary vectors, PDF field-verification sheets

Analytics Satellize can run

Draft plot boundary vector layer	Edge detection and morphological filtering on pan-sharpened VHR imagery combined with DSM gradient analysis; candidate lines classified by spectral and height signatures	GeoPackage polygon layer with per-boundary confidence score, ready for community verification session
Pathway network skeleton	Spectral segmentation isolating bare-earth and compacted-surface pixels; medial-axis thinning to produce centreline vectors; manual QA against imagery	GeoPackage line layer of path centrelines with estimated width attribute, usable as block-structure reference
Settlement block structure map	Voronoi partitioning of path skeleton combined with building footprint outlines; blocks labelled by estimated plot count	GeoPackage polygon layer of blocks with plot-count estimates, delivered as GIS layer and A3 PDF map sheets
Multi-epoch change flags	Pixel-level and object-level change detection between primary mapping epoch and subsequent SkySat collects; changed objects flagged for re-survey	Quarterly GeoPackage update layer highlighting areas of structural change since baseline, with change-type attribute
Historical occupancy timeline	Visual and automated interpretation of commercial archive imagery (Pleiades, WorldView-3) to date first appearance of structures; results linked to boundary polygons	Attribute table appended to boundary layer with estimated first-occupancy year and image-date provenance
Boundary ambiguity report	Spatial analysis identifying polygons where DSM and spectral evidence are contradictory or where shared-wall conditions prevent automated delineation	PDF report and flagged GIS layer identifying areas requiring priority ground-truth or community arbitration

Sources

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.