Refugee and IDP camp expansion monitoring within heritage site buffer zones
Rapid informal settlement growth inside UNESCO and national heritage buffer zones is measurable from orbit at weekly cadence. Very-high-resolution optical change detection can map tent fabric, prefab shelters and earthworks without criminalising displacement.
Sensors
- Pleiades Neo: 30 cm native resolution panchromatic, 50 cm multispectral (blue, green, red, red-edge, NIR). Revisit of 1 to 2 days at mid-latitudes with constellation of four satellites. Fine enough to distinguish individual tent footprints and prefab panel seams, and to detect new earthwork berms as narrow as 1 m.
- WorldView Legion: 30 cm panchromatic resolution, up to 15 revisits per day over conflict-affected regions by design. Stereo acquisition supports DSM differencing to detect ground-levelling and foundation excavation, relevant where earthworks threaten subsurface archaeology.
- PlanetScope SuperDove: 3 m resolution, eight spectral bands including red-edge and NIR, near-daily global coverage. Ideal for rapid change flags and area-wide monitoring across large buffer zones where VHR tasking budget is limited. Minimum detectable new structure cluster is roughly 50 to 100 sq m.
- Sentinel-2 MSI: 10 m resolution in visible and NIR bands, 5-day revisit at the equator with both satellites. Free and openly archived since 2015. Useful for perimeter-level change detection and for maintaining a long baseline archive; too coarse to classify individual shelter types but adequate for tracking gross footprint growth in hectares.
What the buffer zone boundary actually protects, and why it fails under crisis conditions
UNESCO World Heritage buffer zones are legal instruments, not physical barriers. They exist on paper, in national legislation and in site-management plans. During a humanitarian crisis, a family arriving at night with a tarpaulin and tent poles has no access to a cadastral map. The result is that informal settlement expansion inside buffer zones is rarely deliberate encroachment; it is the geometry of desperation meeting the geometry of available flat ground.
Jordan's Azraq camp sits within reach of the Azraq wetland Ramsar site and several registered archaeological zones. Lebanon's Bekaa Valley hosts both displaced Syrian populations and some of the densest concentrations of Bronze Age and Roman-period sites in the Levant. South Sudan's Bentiu and Malakal displacement sites are proximate to areas with documented Iron Age and pre-colonial earthwork landscapes. In each case, the monitoring problem is the same: how fast is the built-up footprint growing, in which direction, and is it crossing a legally defined line?
Spectral and textural signatures that separate shelters from archaeology
Tent fabric, polyethylene sheeting and corrugated metal roofing have distinct spectral profiles in the visible and near-infrared. UNHCR-standard white LDPE sheeting is highly reflective across all visible bands, producing a bright, spectrally flat signature that stands out sharply against bare soil or sparse vegetation. Corrugated iron reads differently: lower overall reflectance, a slight thermal signature detectable in shortwave infrared on sensors with that capability. Prefabricated shelter panels, often painted cream or grey, fall between the two.
Textural classifiers applied to VHR imagery exploit the regular spacing and repeated geometry of planned shelter rows, contrasting with the irregular masonry or earthen mound signatures of archaeological features. Grey-level co-occurrence matrix (GLCM) features, in particular contrast and homogeneity at pixel scales of 0.3 to 1 m, have been used in published UNOSAT workflows to separate humanitarian infrastructure from pre-existing built fabric. Object-based image analysis (OBIA) then groups spectrally and texturally similar pixels into shelter objects, road clearances and earthworks, each classified separately.
The honest limit here is shadow. At VHR resolution, shadows from shelter walls and roofing can obscure adjacent features, and at low sun angles (common in winter at Levantine latitudes) shadow fraction can exceed 20 percent of the scene. Analysts must account for this when computing net new area, or risk underestimating footprint by a material margin.
Change detection at weekly to monthly cadence: what is achievable and what is not
A practical monitoring cadence for active crisis situations is weekly VHR tasking over the highest-risk buffer zone segments, with PlanetScope providing daily gap-fill for perimeter-level alerts. Bitemporal differencing on orthorectified imagery, with radiometric normalisation to account for varying sun angle and atmospheric conditions between acquisitions, is the standard published approach. The AAAS Science and Human Rights Coalition has documented this methodology in conflict-site monitoring work, and UNOSAT applies it operationally across multiple crisis geographies.
At Pleiades Neo 50 cm multispectral resolution, a newly erected tent cluster of roughly 10 to 15 shelters (approximately 200 to 300 sq m) is detectable in a single acquisition. A single tent is not reliably detectable as a classified object; it may appear as an anomalous bright pixel cluster but cannot be confidently distinguished from vehicle parking or debris. That is the honest resolution floor for this application.
Cloud cover is the principal operational constraint at Levantine and East African sites. Winter cloud frequency over the Bekaa Valley can exceed 60 percent of days between November and March, making weekly clear-sky VHR acquisition unreliable. SAR-based change detection can fill some of this gap, but SAR backscatter from tent fabric is weak and inconsistent, so SAR is better used for detecting earthworks and ground-levelling than for shelter classification. Analysts should plan for a mixed-sensor workflow rather than relying on any single source.
The ethical architecture of this kind of monitoring
Satellite monitoring of displaced populations is not ethically neutral. The same change-detection product that tells a heritage authority which parcels are newly occupied can, in the wrong hands, tell a security force exactly where people are sleeping. This is not a hypothetical concern. The AAAS Human Rights Program has published explicit guidance on the dual-use risk of high-resolution settlement mapping in conflict contexts, and UNOSAT operates under UN data-governance frameworks that restrict who receives site-specific coordinate data.
The appropriate safeguard architecture has several components. Spatial aggregation: delivering area statistics and perimeter measurements rather than geocoded shelter centroids. Access tiering: heritage protection authorities receive full spatial products; public-facing outputs show only buffer-zone-level summaries. Temporal lag: in active conflict settings, a deliberate delay between acquisition and release reduces operational utility for hostile actors. And explicit purpose limitation in data-sharing agreements, specifying that the product is for heritage impact assessment, not population tracking.
Satellize operates on the same principle applied to its Tonga crop-estimation programme: the client defines the governance framework, and the analytics are scoped accordingly. For heritage buffer monitoring, that means working with the site management authority and the relevant humanitarian coordination body jointly, not supplying one without the other.
What the output actually looks like for a site manager
A practical deliverable for a heritage authority is a monthly change report covering the buffer zone perimeter, expressed as: net new built-up area in hectares since the previous acquisition, broken down by shelter type (fabric, rigid, earthwork); a vector layer showing newly detected footprints clipped to the buffer zone boundary; and a simple traffic-light status for each defined sub-zone (no change, marginal encroachment, significant encroachment requiring field verification).
Field verification remains essential. Satellite classification at this scale carries a false-positive rate that varies with scene complexity, but published OBIA workflows on VHR imagery typically achieve overall accuracies of 85 to 92 percent for built-up versus non-built-up in informal settlement contexts, based on studies published in Remote Sensing (MDPI) and similar peer-reviewed venues. That means roughly one in ten flagged objects warrants ground-truthing before any formal heritage-protection action is taken. The satellite product narrows the search area; it does not replace the site inspector.
Typical figures
| Finest spatial resolution (VHR) | 30 cm panchromatic (Pleiades Neo, WorldView Legion) |
| Multispectral resolution (VHR) | 50 cm (Pleiades Neo); 1.2 m (WorldView Legion) |
| Medium-resolution optical | 3 m (PlanetScope SuperDove); 10 m (Sentinel-2 MSI) |
| Revisit cadence | Daily to near-daily (PlanetScope); 1 to 2 days (Pleiades Neo); 5 days (Sentinel-2 at equator) |
| Minimum detectable new structure cluster | ~200 to 300 sq m at VHR (10 to 15 shelters); ~50 to 100 sq m area change at 3 m resolution |
| Spectral bands used | Blue, green, red, red-edge, NIR (primary); SWIR for roofing material discrimination where available |
| Archive depth | Sentinel-2 from 2015; PlanetScope from ~2016; Pleiades from 2012; WorldView from 2009 |
| Cloud-cover constraint | Significant November to March at Levantine sites; SAR gap-fill recommended for earthwork detection |
| Change detection latency | 24 to 72 hours from clear acquisition to classified change layer, depending on processing pipeline |
| Delivery formats | GeoTIFF classified rasters, GeoPackage or Shapefile vector layers, PDF summary report, optional GIS web service |
Analytics Satellize can run
| Buffer zone encroachment map | Bitemporal OBIA change detection on orthorectified VHR multispectral imagery with GLCM textural features; radiometric normalisation between dates | Monthly GeoPackage vector layer showing new built-up footprints clipped to buffer zone boundary, with shelter-type classification attributes |
| Built-up area growth time series | Pixel-based change detection on PlanetScope SuperDove 3 m imagery; NDBI and spectral mixture analysis for impervious surface fraction | Monthly hectare-count table per buffer sub-zone, delivered as CSV and chart in PDF report |
| Shelter-type classification | Supervised random-forest classifier trained on spectral and GLCM features at VHR resolution, distinguishing fabric tents, rigid prefab panels and earthworks from pre-existing masonry | Classified raster (GeoTIFF) with per-class area statistics; overall accuracy reported with confusion matrix |
| Earthwork and ground-disturbance detection | DSM differencing from stereo VHR acquisitions (WorldView Legion or Pleiades Neo stereo pairs); elevation change thresholded at 0.5 m vertical | Elevation-change raster and vector polygons flagging ground-levelling or excavation events within buffer zone |
| Perimeter alert feed | Automated daily PlanetScope NDVI and brightness-change threshold trigger on buffer zone perimeter band (configurable width, typically 200 m inside boundary) | Email or API alert within 24 hours of a flagged change event, with thumbnail image and coordinates of the alert polygon |
| Aggregated status report for heritage authority | Spatial aggregation of classified outputs to sub-zone level; traffic-light status assignment based on thresholds agreed with site manager | Monthly PDF report with no geocoded individual shelter centroids in the public-facing version; full spatial data in access-controlled GIS portal |
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.