Environmental health risk mapping in refugee and displacement settlements
Satellite imagery from Sentinel-1, Sentinel-2 and commercial VHR sensors can map sanitation proximity, vegetation loss and flood-driven waterborne risk across displacement settlements, giving WASH planners spatial evidence they rarely have on the ground.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. NDVI and NDWI indices derived from bands 4, 8 and 11 characterise vegetation loss and surface-water extent. Cloud cover over tropical zones can interrupt usable acquisitions for weeks at a time.
- Sentinel-1 SAR (C-band, 5.405 GHz): 6-day repeat in Interferometric Wide Swath mode at 10 m ground range resolution. Penetrates cloud and rain, making it the primary tool for mapping post-rainfall surface-water accumulation when optical data are unavailable. Backscatter change between dry and wet acquisitions reliably delineates inundated ground.
- Maxar WorldView series (VHR optical): Panchromatic resolution of 0.31 m (WorldView-3), multispectral at 1.24 m. Sufficient to distinguish individual latrine structures, open defecation areas and water-point infrastructure. Tasked on demand; archive depth varies by site but many humanitarian zones have coverage from 2011 onwards.
- Planet SuperDove: 3 m multispectral, near-daily revisit globally. Eight spectral bands including red-edge and near-infrared. Useful for tracking rapid vegetation clearance and bare-soil expansion week by week as a settlement grows, though cloud persistence remains a constraint in the Congo Basin and South Asian monsoon zones.
Why a settlement's layout is a health document
When a displacement settlement is established quickly, the spatial relationship between latrines, water points and drainage channels is rarely planned. It is determined by slope, available ground and the order in which people arrive. That geometry encodes risk. A latrine pit upslope of a hand-pump within 30 metres is a contamination pathway; the Sphere Handbook sets a minimum separation of 30 metres precisely because that distance matters in sandy or laterite soils. Satellite imagery at VHR resolution can measure that distance across thousands of structures in a single pass.
The same imagery reveals secondary signals: bare soil expanding outward from a settlement boundary as wood is collected for fuel; drainage channels blocked by shelters erected after the original layout; low-lying areas that collect standing water after rain. None of these require field survey to detect. They do require the right sensor at the right moment, which is where the sensor mix matters.
What a floating roof gives away, and what it hides
VHR optical imagery from WorldView-3 can resolve a latrine superstructure at 0.31 m panchromatic resolution, but it cannot see inside or confirm whether the pit is lined. It can confirm the structure exists, estimate its footprint and measure its distance to the nearest mapped water point. That is already more spatial data than most camp managers hold in a GIS.
The honest limit: optical imagery sees rooftops, not what is beneath them. A tarpaulin shelter and a latrine slab look different spectrally and in shape, but distinguishing a functional latrine from an abandoned one requires either field validation or a time series long enough to observe usage patterns. Analysts should treat VHR-derived latrine counts as upper-bound estimates and flag structures for ground-truthing rather than reporting them as confirmed facilities.
Vegetation indices from Sentinel-2 add a different layer. NDVI time series over a settlement and its surrounding buffer quantify fuel-wood pressure objectively. A sustained NDVI decline of 0.1 to 0.2 units in a 2 km radius, measured against a pre-settlement baseline, is consistent with progressive tree cover loss documented in published studies of settlements in Uganda and Ethiopia, though the precise threshold varies with baseline vegetation type.
SAR fills the gap that cloud creates
The settlements most in need of monitoring are disproportionately located in tropical zones: the Sahel's wet season, the East African highlands, Bangladesh's flood plains. These are also the zones where optical sensors are most frequently blocked. In Kakuma, Kenya, and Cox's Bazar, Bangladesh, usable Sentinel-2 acquisitions during the rainy season can drop to fewer than one per month.
Sentinel-1 C-band SAR changes that calculus. Rain does not attenuate C-band radar in any operationally meaningful way at the intensities typical of tropical storms. The sensor acquires regardless. Post-rainfall surface-water mapping using change detection between a dry-season reference and a post-event acquisition can delineate inundated ground at 10 m resolution within 24 to 48 hours of an overpass, depending on processing pipeline latency. That is fast enough to inform decisions about shelter relocation before waterborne disease risk compounds.
SAR does introduce its own ambiguities. Dense shelter arrangements create layover and shadow artefacts in high-incidence-angle acquisitions. Smooth wet soil and standing water both produce low backscatter, which can be confused in single-date analysis. A two-date change approach, comparing against a known dry baseline, substantially reduces false positives, though it requires a usable archive acquisition from a dry period, which is not always available for newly established sites.
Building a WASH risk index from spatial layers
A spatial WASH risk index aggregates multiple satellite-derived inputs into a per-grid-cell score that field teams can act on. Typical inputs include: latrine-to-water-point proximity derived from VHR classification; surface-water accumulation frequency from SAR time series; bare-soil fraction from Sentinel-2 NDVI; and drainage network connectivity estimated from a digital elevation model. Each layer is normalised and weighted according to the epidemiological literature on faecal-oral transmission pathways.
UNHCR's own technical guidance and published academic work on Cox's Bazar (one of the most extensively studied displacement settings from a remote-sensing perspective) have used variants of this approach to prioritise sanitation interventions spatially. The output is not a disease incidence map. It is a relative risk surface: areas where the physical conditions are most conducive to transmission. That distinction matters when communicating to public health partners who may otherwise over-interpret the product.
Satellize has applied analogous spatial index methods in agricultural contexts, including the Kingdom of Tonga crop-estimation programme. The underlying approach of combining multi-sensor inputs into a single decision-relevant layer transfers directly to WASH risk scoring, with the domain knowledge adjusted accordingly.
Temporal change is the real product
A single-date analysis of a settlement is a snapshot. A time series is a story. The most operationally useful satellite product for a settlement manager is not a map of current conditions but a change layer showing where conditions have deteriorated since the last assessment. That might be a new cluster of structures encroaching on a drainage buffer, a water point that has become surrounded by bare soil and foot traffic, or a vegetation buffer that has been cleared.
Planet SuperDove's near-daily revisit at 3 m makes it the most capable open-commercial option for tracking this kind of incremental change. The practical constraint is data volume and processing time: a large settlement such as Bidi Bidi in Uganda covers roughly 250 square kilometres, generating substantial imagery volume per day. Automated change detection pipelines, rather than manual interpretation, are necessary at that scale.
Archive depth also matters. Sentinel-2 data are available from 2015, Sentinel-1 from 2014. For settlements established after those dates, a pre-settlement baseline is available for both sensors, which is the prerequisite for any meaningful change analysis. For settlements predating the archive, analysts must work with what exists, which may mean using Landsat 8 imagery (30 m, available from 2013) as a coarser baseline.
Typical figures
| Optical spatial resolution (Sentinel-2) | 10 m (visible and NIR), 20 m (SWIR and red-edge) |
| Optical spatial resolution (VHR, WorldView-3) | 0.31 m panchromatic, 1.24 m multispectral |
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m ground range (after multi-looking) |
| Revisit cadence | Sentinel-2: 5 days (equator, both satellites); Sentinel-1: 6 days; Planet SuperDove: near-daily; WorldView-3: tasked on demand, typically 1-4 days |
| Minimum detectable structure (VHR) | Individual latrine superstructures detectable at WorldView-3 panchromatic resolution; functional status requires field validation |
| Surface-water detection floor (SAR) | Patches of approximately 0.01 ha and above in open terrain; dense shelter shadow reduces sensitivity in built areas |
| Archive depth | Sentinel-2 from 2015; Sentinel-1 from 2014; Landsat 8 from 2013 (30 m) for earlier baselines |
| Cloud penetration | Sentinel-1 SAR: unaffected by cloud and rain. Optical sensors: unusable under cloud; tropical sites may yield fewer than one clear acquisition per month during wet season |
| Typical processing latency | SAR flood maps: 24-48 hours post-overpass in operational pipelines; Sentinel-2 indices: 3-5 days for cloud-screened composites |
| Delivery formats | GeoTIFF raster layers, GeoJSON vector overlays, GIS-ready risk index grids, PDF summary reports |
Analytics Satellize can run
| Latrine-to-water-point proximity map | VHR object-based image analysis (OBIA) to classify latrine and water-point structures; Euclidean distance calculation against Sphere Handbook 30 m threshold | GeoJSON layer with per-structure proximity scores and threshold-breach flags for field prioritisation |
| Post-rainfall surface-water accumulation layer | Sentinel-1 SAR backscatter change detection (wet acquisition versus dry-season reference); water/non-water binary classification using thresholded sigma-naught difference | GeoTIFF inundation extent map delivered within 48 hours of overpass, with shelter overlap count |
| Vegetation loss and fuel-wood pressure index | Sentinel-2 NDVI time series with pre-settlement baseline subtraction; per-pixel trend analysis over 2 km settlement buffer | Monthly NDVI anomaly raster and summary chart showing cumulative canopy loss in hectares |
| Composite WASH risk index | Weighted spatial overlay of latrine proximity, surface-water frequency, bare-soil fraction and drainage-network proximity; normalised to 0-100 relative risk scale | Grid-cell risk surface (GeoTIFF) and ranked priority-zone report for field team deployment |
| Settlement growth and encroachment monitoring | Planet SuperDove or Sentinel-2 bi-weekly change detection using normalised difference built-up index (NDBI) and bare-soil spectral signatures | Change polygon layer showing newly occupied ground, with area statistics and drainage-buffer encroachment alerts |
| Seasonal flood-risk forecast layer | DEM-derived flow accumulation combined with historical SAR inundation frequency to produce per-cell flood-probability surface for wet-season planning | Annual pre-season risk map with shelter-count exposure estimates, formatted for UNHCR camp planning tools |
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.