River blackfly breeding habitat mapping for onchocerciasis transmission risk

Simulium blackflies breed only in fast, well-oxygenated river reaches. Combining Sentinel-1 SAR, Sentinel-2 vegetation indices and DEM-derived stream-power models lets public-health teams map those reaches and focus larviciding where it counts.

Sensors

Sentinel-1 SAR (C-band, ESA): 10 m ground resolution in Interferometric Wide Swath mode, 6-day repeat at the equator (12-day per satellite). Backscatter intensity and coherence change detect surface roughness differences between turbulent and still water, providing a flow-character proxy where optical sensors are cloud-blocked.
Sentinel-2 MSI (ESA): 10 m (visible/NIR) and 20 m (red-edge/SWIR) resolution, 5-day revisit at the equator. NDVI maps riparian canopy density; NDWI (Green-NIR) delineates open water extent. Cloud cover over tropical river corridors is the principal operational constraint, often exceeding 70 % of acquisitions in wet season.
SRTM / Copernicus DEM (30 m): Global 30 m digital elevation model used to derive slope, upstream contributing area and stream-power index (SPI = discharge proxy × slope). SPI identifies river segments with the gradient and flow energy that favour Simulium attachment substrates. Vertical accuracy is ±16 m RMSE globally, which limits reach-scale discrimination on very low-gradient rivers.
Landsat 8/9 OLI (USGS/NASA): 30 m resolution, 16-day revisit per satellite (8-day combined). Band 3 (green) and Band 4 (red) ratios provide turbidity estimates; clearer, faster water tends toward lower turbidity signatures, complementing the SAR roughness signal. Archive extends to 1972 for multi-decadal reach-change analysis.

Why the river itself is the diagnostic unit

Onchocerciasis control has always been a hydrological problem dressed as an entomological one. Simulium damnosum and its sibling species do not breed in ponds, slow backwaters or irrigation channels. They require fast-flowing, well-oxygenated water, typically attaching egg masses and larvae to submerged rocks, trailing vegetation and woody debris in reaches where surface velocity exceeds roughly 0.3 to 0.5 m/s. That physical specificity is what makes satellite mapping tractable: you are not looking for an insect, you are looking for a hydraulic signature.

The World Health Organisation's African Programme for Onchocerciasis Control (APOC) and its predecessor, the Onchocerciasis Control Programme (OCP), demonstrated over decades that larviciding fast-water breeding sites with Bacillus thuringiensis israelensis (Bti) or temephos can suppress transmission. The constraint was always ground survey coverage across river networks spanning millions of square kilometres in West, Central and East Africa, and in Yemen's Tihama plain. Satellite-derived habitat maps do not replace entomological ground surveys, but they reduce the search space dramatically before a single field team is deployed.

What a floating roof gives away: reading turbulence from orbit

C-band SAR backscatter responds to centimetre-scale surface roughness. Calm water returns very little energy toward the sensor (specular reflection away from the antenna), producing a dark signature. Turbulent, fast-flowing water scatters energy back more diffusely, appearing brighter. The contrast is detectable at Sentinel-1's 10 m pixel size on river reaches wider than roughly 20 to 30 m, which covers the majority of Simulium-productive rivers in the Volta, Niger and Nile basins.

Coherence between repeat-pass SAR acquisitions adds a second dimension. Stable water surfaces maintain moderate coherence; highly turbulent or rapidly changing surfaces decorrelate strongly. Combining amplitude and coherence in a two-feature space helps distinguish fast rapids from wind-roughened lakes, which can otherwise produce similar backscatter values. The honest caveat is that very narrow rivers, below about 20 m width, fall below the effective resolution threshold and require either commercial very-high-resolution SAR (such as ICEYE or Capella) or field verification.

Stream-power index: letting the terrain do the filtering

Before any image analysis, a DEM-derived stream-power index can pre-screen the entire river network. SPI is calculated as the product of the upstream contributing area (a proxy for discharge) and the local slope angle. High SPI values identify reaches where water is both voluminous and steep, the hydraulic conditions that generate the surface velocities Simulium requires. Applied to the Copernicus 30 m DEM, this computation is near-instantaneous across a country-scale river network and produces a ranked list of candidate breeding reaches before a single satellite image is examined.

The limitation is real and worth stating plainly. DEM-derived slope at 30 m resolution smooths over short rapids and waterfalls. A reach that is hydraulically fast over 50 m may appear as a modest slope in the averaged DEM. Ground-truthed calibration, matching known breeding sites against SPI thresholds, is necessary in each river basin because the relationship between SPI and actual velocity varies with channel morphology and substrate. Published studies in the Volta basin suggest SPI thresholds in the range of 10³ to 10⁵ (dimensionless, depending on area units) capture the majority of confirmed breeding sites, but false-negative rates at low-gradient reaches remain a known weakness.

Riparian vegetation as a habitat modifier

Dense riparian canopy does more than shade the water. Overhanging vegetation provides trailing substrates for larval attachment and moderates water temperature. Sentinel-2 NDVI at 10 m resolution maps riparian greenness with sufficient detail to classify corridor width and density along river banks. Reaches with narrow or degraded riparian strips, common near agricultural encroachment, tend to support lower blackfly populations even when hydraulic conditions are otherwise suitable.

NDWI (the Green-NIR normalised difference) delineates the active water surface and its seasonal fluctuation. In the dry season, many West African rivers contract to isolated pools; breeding is concentrated in the remaining fast-water sections. Tracking NDWI through the Sentinel-2 archive identifies which reaches retain flow year-round, a critical input for planning perennial larviciding versus seasonal interventions. Cloud contamination in the wet season, when transmission risk peaks, is the persistent operational problem. SAR-optical fusion, using Sentinel-1 to fill Sentinel-2 cloud gaps, is the standard mitigation, though it introduces its own registration and calibration steps.

Assembling a transmission-risk layer a field team can actually use

The analytic pipeline combines four inputs: SPI-ranked candidate reaches from the DEM, SAR roughness and coherence scores, Sentinel-2 NDVI and NDWI for riparian context, and Landsat-derived turbidity to flag sediment-loaded reaches where blackfly density is typically suppressed. Each reach segment, typically 100 to 500 m in length, receives a composite suitability score. The output is a GIS vector layer of river reaches ranked by breeding suitability, with confidence flags where cloud cover or DEM artefacts reduce certainty.

This is not a blackfly-density forecast. It is a hydraulic and vegetation habitat suitability map. Actual fly populations depend on temperature, seasonal flow regime, historical larviciding pressure and local species composition, none of which satellite data resolves directly. The map is most valuable as a prioritisation tool: directing entomological sampling teams to the highest-ranked reaches first, and providing a spatial framework for planning larviciding flight paths or ground-application routes. Satellize's analytics team can structure this pipeline for a specific river basin, drawing on the same open-constellation approach used in the Tonga crop-estimation programme, adapted here to hydrological rather than agricultural feature extraction.

For Yemen's Tihama coastal plain, where onchocerciasis persists in the Wadi Harad and adjacent drainage systems, the same methodology applies with one additional complication: political access constraints make ground survey sporadic, which raises the value of satellite-derived prioritisation and simultaneously makes field validation harder to obtain. Honest programme design must account for that gap.

What this approach cannot do, and what comes next

Satellite habitat mapping does not replace vector surveillance. It cannot detect adult fly populations, measure biting rates or confirm transmission intensity. Cytotaxonomic identification of Simulium sibling species, which differ in vectorial capacity, requires microscopy. The maps inform where to look, not what will be found.

Revisit frequency is a genuine constraint for operational larviciding support. Bti treatments must be reapplied every seven to ten days during the breeding season. A 5-day Sentinel-2 revisit, reduced to perhaps 2 to 3 usable acquisitions per month by cloud cover, is insufficient for treatment-cycle monitoring. SAR can partially compensate, but neither sensor provides the near-daily cadence that larviciding logistics ideally require. Commercial SAR constellations with daily revisit are an option for high-priority reaches, at additional cost. Programmes should plan for this gap in their monitoring design rather than assume satellite data will track every treatment window.

Typical figures

Primary SAR resolution	10 m (Sentinel-1 IW mode)
Primary optical resolution	10 m visible/NIR, 20 m red-edge/SWIR (Sentinel-2)
DEM resolution	30 m (Copernicus GLO-30 / SRTM)
Sentinel-1 revisit	6 days (two-satellite constellation at equator)
Sentinel-2 revisit	5 days (two-satellite constellation at equator); effective usable acquisitions reduced by cloud cover
Landsat 8/9 combined revisit	8 days at equator; 30 m resolution
Minimum detectable river width (SAR)	~20-30 m; narrower reaches require commercial VHR SAR
Stream-power index vertical accuracy constraint	SRTM ±16 m RMSE globally; limits slope precision on low-gradient reaches
Sentinel archive depth	Sentinel-1 from 2014; Sentinel-2 from 2015; Landsat from 1972
Deliverable format	GIS vector (GeoPackage / Shapefile / GeoJSON), reach-ranked CSV, cloud-optimised GeoTIFF rasters

Analytics Satellize can run

Stream-power index reach ranking	DEM hydrological analysis (flow accumulation × slope); Copernicus GLO-30 or SRTM input	Vector polyline layer of river network with SPI score per 100-500 m segment; GIS layer
SAR turbulence probability score	Sentinel-1 C-band backscatter amplitude and repeat-pass coherence classification over water pixels	Raster probability layer (0-1) of fast/turbulent surface water per acquisition; cloud-optimised GeoTIFF
Riparian vegetation density profile	Sentinel-2 NDVI time-series along buffered river centrelines; seasonal compositing to reduce cloud gaps	Per-reach riparian NDVI statistics (mean, seasonality amplitude); tabular report and GIS layer
Active water surface seasonal extent	Sentinel-2 NDWI thresholding with SAR gap-fill; dry-season vs. wet-season composite comparison	Binary water-presence raster per season; perennial-flow reach classification layer
Composite breeding habitat suitability map	Weighted multi-criteria overlay of SPI rank, SAR roughness score, NDVI corridor width and turbidity flag; weights calibrated against published Simulium habitat studies	Ranked reach suitability map with confidence flags; GIS vector layer and PDF summary report
Larviciding priority zone delineation	Suitability map thresholded to high-confidence breeding reaches; buffered to ground-accessible treatment zones	Operational priority polygons for larviciding flight or ground-application planning; GeoJSON and printable field map
Multi-year habitat change detection	Landsat archive time-series for turbidity and NDWI; Sentinel-2 archive from 2015 for NDVI trend; change-point detection on reach-level statistics	Annual reach-condition time series; report identifying reaches with significant habitat change

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.