Schistosomiasis transmission risk from freshwater snail habitat mapping
Bulinus and Biomphalaria snails colonise slow, warm, vegetated freshwater bodies detectable from orbit. Sentinel-2 water indices and MODIS land surface temperature can map suitable habitat at sub-kilometre scale to focus mass drug administration where transmission risk is highest.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. NDWI (Green/NIR ratio) and MNDWI (Green/SWIR ratio) separate open water from soil and vegetation. Cannot resolve irrigation channels narrower than roughly 20 m.
- Landsat 8/9 OLI + TIRS: 30 m multispectral for water indices; 100 m thermal band (TIRS) resampled to 30 m for surface water temperature. 16-day revisit per satellite; combined Landsat 8 and 9 reduces this to roughly 8 days. Archive extends to 1972 for long-term habitat trend analysis.
- MODIS Terra/Aqua LST: Land surface temperature at 1 km spatial resolution, twice-daily overpass per satellite. MOD11A1 and MYD11A1 products provide day and night LST. Coarse resolution means urban-rural temperature gradients are captured but individual canals are not resolved. Useful for filtering out thermally unsuitable habitat at landscape scale.
- GPM IMERG: Global Precipitation Measurement Integrated Multi-satellitE Retrievals provide half-hourly rainfall estimates at 0.1-degree resolution. Seasonal rainfall drives canal and pond filling; IMERG anomalies signal the onset of transmission-season conditions with roughly 3-hour latency.
Why snails are a remote-sensing problem
Schistosomiasis infects an estimated 250 million people, the vast majority in sub-Saharan Africa, through skin contact with freshwater containing cercariae shed by infected snails. The snails themselves, principally Bulinus species for urogenital schistosomiasis and Biomphalaria species for intestinal schistosomiasis, are not directly visible from orbit. What is visible is the habitat they require: slow or still freshwater bodies, surface temperatures between roughly 15 and 35 degrees Celsius, and aquatic or emergent vegetation providing attachment surfaces and shade.
That combination of conditions is mappable. The challenge is that schistosomiasis transmission concentrates in agricultural landscapes where the relevant water bodies are irrigation canals, seasonal ponds, rice paddies and reservoir margins, not large lakes. Many of these features are narrow, ephemeral or both. That is where sensor choice and honest resolution accounting matter.
What a water index actually measures, and where it fails
The Normalised Difference Water Index (NDWI), calculated from Sentinel-2 Band 3 (green, 560 nm) and Band 8 (near-infrared, 842 nm), exploits the fact that liquid water absorbs strongly in the NIR while vegetation reflects it. The Modified NDWI (MNDWI) substitutes Band 11 (SWIR, 1610 nm) for NIR, which suppresses built-up area signal and performs better in turbid agricultural water. Both indices produce continuous float values; a threshold around 0.0 to 0.2 (depending on scene turbidity) separates water from non-water pixels.
The 10 m pixel size of Sentinel-2 MSI is the finest freely available multispectral resolution for this application, but it carries a hard geometric limit. A pixel classified as water must be predominantly water across its full 10 m extent. In practice, mixed pixels at canal edges mean that channels narrower than roughly 20 m are unreliably detected. Secondary irrigation channels in many African smallholder systems are 1 to 5 m wide. Those channels are effectively invisible to Sentinel-2 water mapping and require field survey or very-high-resolution commercial imagery to characterise. This is not a failure of the method; it is a physical constraint that any honest programme design must account for.
Seasonal cloud cover compounds the problem in the humid tropics. In the Congo Basin and parts of West Africa, cloud-free Sentinel-2 acquisitions during the wet season, precisely when snail populations peak, can be scarce. Compositing over 30-day windows helps but introduces temporal blur; a pond that fills and drains in two weeks may appear as a low-confidence partial-water pixel rather than a confirmed habitat.
Temperature as a habitat filter, not a habitat detector
Snail biology constrains the thermal envelope of transmission risk. Bulinus truncatus, the principal intermediate host of Schistosoma haematobium, has a lower developmental threshold of roughly 15 degrees Celsius and shows peak cercarial shedding between 25 and 30 degrees. Above 35 degrees, snail survival declines sharply. MODIS LST products, available twice daily at 1 km, allow a landscape-scale thermal mask to be applied before water mapping. Pixels where monthly mean LST consistently falls outside the 15 to 35 degree range can be excluded from habitat probability scoring without finer-resolution analysis.
This is a coarse filter, not a snail census. A 1 km MODIS pixel over a mosaic of irrigated fields, dry soil and canal water will report an average temperature that may not reflect the microclimate of a shaded canal margin where snails actually live. Landsat TIRS at 100 m (resampled to 30 m) provides a better spatial match to field-scale features, though its 16-day revisit limits temporal resolution during rapidly changing seasons.
Seasonal dynamics and the transmission window
Snail populations are not static. They expand rapidly after rainfall events that fill seasonal water bodies, peak several weeks into the wet season, and crash during dry-season desiccation. The lag between rainfall onset and peak cercarial shedding is epidemiologically important: mass drug administration timed just before peak transmission, rather than after, reduces infection incidence more efficiently.
GPM IMERG provides near-real-time rainfall accumulation at 0.1-degree resolution, sufficient to detect the onset of the wet season at district scale. Combining IMERG anomalies (current season versus a climatological baseline) with Sentinel-2 water extent change produces a seasonal habitat emergence signal. When new water bodies appear in thermally suitable areas, a transmission-risk alert can be generated within days of the triggering rainfall event. The 0.1-degree IMERG footprint is roughly 11 km at the equator, so the rainfall trigger is coarse; it flags districts rather than individual villages.
Long-term Landsat archives, continuous from 1984 in practical terms, allow habitat persistence classification. A water body that appears in Landsat imagery every wet season for 30 years is a different transmission risk than one that appeared for the first time after a dam was constructed. Persistence maps inform where permanent snail control infrastructure (molluscicide application points, canal lining) is worth the investment.
From habitat map to mass drug administration targeting
The operational output of this analysis is a habitat suitability layer, typically a probability surface at 10 to 30 m resolution, combined with population exposure data to produce a transmission-risk score by administrative unit. National neglected tropical disease programmes use such layers to prioritise which districts receive praziquantel under preventive chemotherapy campaigns. The World Health Organisation's 2030 NTD road map explicitly calls for geospatial data to support this targeting.
The honest limit of satellite-derived habitat mapping is that it identifies where snails could live, not where they do live. A high suitability score requires field validation: snail sampling by trained teams using standard quadrat methods remains the ground truth. Satellite analysis is most valuable in reducing the area requiring field survey, not in replacing it. In a country the size of Ethiopia or Nigeria, satellite pre-screening can reduce the survey burden by an order of magnitude.
Satellize has built seasonal water-body monitoring pipelines for agricultural applications, including the Kingdom of Tonga crop-estimation programme, and the same compositing and change-detection architecture applies directly to snail habitat tracking. Ministries of health or NTD programme managers wanting a habitat suitability baseline for a specific river basin or irrigation scheme can commission a defined analytical deliverable rather than a broad research engagement.
What the satellite cannot see
Several factors that strongly influence snail abundance are not detectable from orbit. Molluscicide application history, predator communities, water chemistry (pH, conductivity, dissolved oxygen) and human water-contact behaviour all modulate actual transmission risk independently of habitat suitability. A satellite-derived map showing high habitat suitability in a district where annual mollusciciding has been sustained for a decade will overestimate current risk.
Submerged aquatic vegetation, which Biomphalaria snails preferentially colonise, is detectable in shallow clear water using Sentinel-2 red-edge bands (Band 5 at 705 nm, Band 6 at 740 nm), but only where water depth is less than roughly 1 to 2 metres and turbidity is low. In the turbid irrigation water typical of smallholder agriculture, the spectral signal from submerged vegetation is lost. Floating and emergent vegetation remains detectable. These are not reasons to abandon the approach; they are reasons to be precise about what the deliverable claims.
Typical figures
| Primary water mapping resolution | 10 m (Sentinel-2 MSI); 30 m (Landsat 8/9 OLI) |
| Minimum detectable water body | Approximately 400 m² (single pixel at 20 m effective detection limit); channels narrower than ~20 m unreliably detected |
| Revisit frequency | 5 days (Sentinel-2 A+B combined); ~8 days (Landsat 8+9 combined); daily (MODIS Terra+Aqua) |
| Thermal resolution (LST) | 1 km (MODIS MOD11A1/MYD11A1); 100 m native / 30 m resampled (Landsat TIRS) |
| Rainfall input resolution | 0.1 degree (~11 km at equator), 30-minute accumulation (GPM IMERG) |
| Spectral bands used | Green (560 nm), NIR (842 nm), Red-edge (705, 740 nm), SWIR (1610 nm) for water/vegetation; TIR (10.6–12.5 µm) for LST |
| Archive depth | Sentinel-2: from 2015; Landsat: from 1984 (OLI from 2013); MODIS LST: from 2000 |
| Latency (operational product) | Sentinel-2 Level-2A typically available within 3–5 hours of acquisition; MODIS LST daily product within ~24 hours; IMERG Early Run ~4 hours |
| Thermal suitability window | 15–35 °C surface water temperature; MODIS LST used as landscape-scale proxy |
| Delivery formats | GeoTIFF habitat probability raster, GeoPackage vector polygons by administrative unit, CSV risk scores, seasonal change report (PDF) |
Analytics Satellize can run
| Seasonal water-body extent map | MNDWI thresholding on Sentinel-2 monthly cloud-free composites; Otsu adaptive thresholding for scene-specific water/non-water separation | GeoTIFF raster and vector polygon layer per season, updated monthly |
| Thermal suitability mask | MODIS MOD11A1 monthly mean LST filtered to 15–35 °C snail developmental range; Landsat TIRS used for sub-kilometre refinement where cloud permits | Binary suitability raster at 1 km and 30 m, delivered as GeoTIFF overlay |
| Habitat persistence classification | Landsat annual water-body composites (1984–present) classified into permanent, seasonal and ephemeral categories using frequency-of-inundation thresholds | 30 m persistence map with 3-class legend, GeoTIFF and GeoPackage |
| Transmission-season onset alert | GPM IMERG 10-day accumulation anomaly (versus 20-year climatological baseline) combined with Sentinel-2 new-water-body detection trigger | District-level alert table (CSV) and map layer, issued within 5 days of triggering rainfall event |
| Snail habitat suitability probability surface | Logistic or MaxEnt-class habitat model combining MNDWI water extent, NDVI vegetation density, LST, elevation and seasonal flood frequency as predictor variables | Continuous probability raster (0–1) at 30 m, with uncertainty band layer; GeoTIFF |
| Population-weighted transmission risk score by administrative unit | Habitat suitability surface overlaid with WorldPop or national census population grid; area-weighted mean suitability computed per district | Ranked district table (CSV) and choropleth GIS layer for MDA prioritisation |
| Aquatic vegetation extent (shallow clear water only) | Sentinel-2 red-edge band ratio (Band 5/Band 4) to detect submerged and emergent macrophytes in water bodies where depth < ~1.5 m and turbidity is low | Vegetation presence layer flagged with turbidity confidence score; GeoTIFF |
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.