Coastal cholera risk from sea surface temperature and phytoplankton bloom mapping
Vibrio cholerae persists in marine reservoirs tied to phytoplankton and zooplankton blooms. SST anomalies and chlorophyll-a signals from MODIS, VIIRS and Sentinel-3 can precede clinical case surges by one to three weeks, giving health authorities a narrow but real warning window.
Sensors
- MODIS Aqua (ocean colour): 250 m to 1 km resolution depending on band; daily global coverage. Chlorophyll-a retrieved via OC3M algorithm. Archive from 2002 provides two decades of bloom climatology for anomaly detection.
- VIIRS (SNPP / NOAA-20): 750 m imagery bands, 375 m day-night band. Ocean colour products include chlorophyll-a and inherent optical properties. Near-daily revisit at tropical latitudes; data continuity with MODIS record.
- GHRSST (multi-sensor SST composite): Group for High Resolution Sea Surface Temperature products blend infrared and microwave radiometers to produce daily 1 km and 0.25° composites. Microwave retrievals penetrate thin cloud; infrared retrievals are masked by thick cloud cover.
- Sentinel-3 OLCI: Ocean and Land Colour Instrument; 300 m resolution, 21 spectral bands from 400 to 1020 nm. Revisit roughly 1.4 days at the equator with both Sentinel-3A and 3B combined. Better spectral discrimination of phytoplankton functional types than MODIS in clear water.
Why the ocean is a cholera reservoir, not just a transport route
Vibrio cholerae is autochthonous to estuarine and coastal marine environments. It does not simply wash in from human sewage and disappear. The bacterium attaches to the chitinous exoskeletons of copepods and to phytoplankton cell surfaces, where it finds both physical substrate and dissolved nutrients. When phytoplankton bloom, copepod populations follow, and Vibrio concentrations in the water column rise with them. This is not a hypothesis: studies in the Bay of Bengal, published in peer-reviewed literature and underpinning operational warning systems in Bangladesh, have documented the association repeatedly since the early 2000s.
The practical consequence is that bloom onset is a leading indicator of elevated human exposure risk, particularly in communities that drink or cook with surface water, fish in coastal shallows, or handle raw catch without refrigeration. The lag between detectable bloom conditions and the first clinical case presentations runs roughly one to three weeks, depending on water temperature, population density near the shore, and the pre-existing immunity profile of the community. One to three weeks is enough time to pre-position oral rehydration salts, alert district health officers, and intensify water treatment.
What SST thresholds and chlorophyll signals actually tell you
The temperature threshold most commonly cited in published cholera-environment models sits around 20°C as a lower bound for Vibrio proliferation, with growth rates accelerating substantially above 25°C. GHRSST composites can resolve SST anomalies of roughly 0.2 to 0.5°C relative to a seasonal climatology, which is sufficient to flag warming events that push marginal coastal waters above the growth threshold. The anomaly matters as much as the absolute value: a bay that is already warm but stable is less informative than one showing a rapid positive departure from its 30-year mean.
Chlorophyll-a concentration is the standard proxy for phytoplankton biomass. MODIS and VIIRS retrieve it optically, using the ratio of water-leaving radiance in blue and green bands. In clear open water, the OC3M algorithm performs well down to concentrations of roughly 0.05 mg m⁻³, and the upper end of a significant bloom can reach 10 to 50 mg m⁻³. Sentinel-3 OLCI adds value through its finer spectral resolution, which helps distinguish phytoplankton from coloured dissolved organic matter, a persistent source of confusion in retrieval algorithms.
Where the retrieval breaks down: turbidity, cloud and the river-plume problem
River-influenced coastal waters present a serious retrieval problem that any honest analysis must state plainly. The Ganges-Brahmaputra delta, the Rufiji delta in Tanzania, and similar high-discharge systems carry enormous sediment loads that optically dominate the water-leaving signal. Standard blue-green chlorophyll algorithms fail in these conditions, producing artefactual high-chlorophyll retrievals where the actual signal is suspended sediment. This is precisely the geography where cholera risk is highest.
Corrections exist. Regional bio-optical algorithms tuned to Case 2 waters, and inherent-optical-property inversion methods, can partially recover chlorophyll-a in turbid conditions, but they introduce additional uncertainty and require local validation data to be trusted. Cloud cover compounds the problem: the Bay of Bengal and East African coast experience monsoon seasons during which optical sensors may be blocked for days to weeks at a stretch. GHRSST microwave SST retrievals are less affected by cloud, but microwave radiometers cannot retrieve chlorophyll-a. The honest position is that satellite-derived cholera risk indices are most reliable in the pre-monsoon and post-monsoon shoulder seasons, and should be treated with greater caution during peak cloud cover.
Building a risk index: from pixels to a health-authority alert
Operational cholera early-warning systems have combined SST anomaly and chlorophyll-a anomaly into composite risk indices, sometimes adding sea surface height anomalies from altimetry as a proxy for coastal upwelling or downwelling that concentrates nutrients. The canonical approach, documented in work by Colwell, Lobitz and colleagues and subsequently extended by groups at the International Centre for Diarrhoeal Disease Research in Bangladesh, flags coastal grid cells where both SST and chlorophyll-a exceed defined anomaly thresholds simultaneously. The resulting binary or probabilistic risk map is then intersected with population exposure data to produce district-level alerts.
The one-to-three-week lag is not a fixed number. It varies with local hydrology, the route from coastal water to drinking water, and the infectious dose required for symptomatic disease. Calibrating the lag for a specific coastline requires historical case data from district health systems. Without that calibration, the satellite signal is suggestive but not actionable at the precision a health ministry needs. This is why the satellite component is necessary but not sufficient: it must be paired with ground-truth case reporting to be operationally useful.
Satellize has run satellite analytics for food-security applications, including the Kingdom of Tonga crop-estimation programme, using a similar workflow of open-constellation data combined with anomaly detection. The same analytical infrastructure applies directly to ocean colour and SST monitoring.
Archive depth and what it is good for
MODIS Aqua has operated since 2002, giving more than two decades of ocean colour and SST data at consistent quality. VIIRS on SNPP extends the record from 2012. That archive depth supports two things that a single-season deployment cannot: climatological baseline construction and interannual trend analysis. A ministry of health can ask not just whether this month's bloom is anomalous relative to last year, but whether the frequency of high-risk bloom events has increased over a decade, which is a different and more politically significant question.
Sentinel-3 OLCI data is freely available through the Copernicus Data Space Ecosystem. MODIS and VIIRS ocean colour products are distributed by NASA Earthdata. GHRSST composites are accessible through PODAAC. None of these require a satellite licence to access, which matters for health agencies operating on constrained budgets. The analytical cost is in processing, calibration and interpretation, not in data acquisition.
Typical figures
| Chlorophyll-a spatial resolution | MODIS: 1 km (OC bands); VIIRS: 750 m; Sentinel-3 OLCI: 300 m |
| SST spatial resolution | GHRSST L4 composites: 1 km (infrared blend) to 0.25° (~25 km, microwave) |
| Revisit frequency | MODIS Aqua: ~1 day; VIIRS SNPP + NOAA-20: ~1 day; Sentinel-3A+B combined: ~1.4 days at equator |
| Chlorophyll-a detection range | Approximately 0.05 to >50 mg m⁻³ in clear Case 1 water; accuracy degrades in turbid Case 2 water |
| SST anomaly sensitivity | ~0.2 to 0.5°C relative to climatological baseline (GHRSST multi-sensor composite) |
| Key spectral bands | Blue (443 nm), green (547–555 nm) for OC retrieval; thermal infrared 10–12 µm for SST; microwave 6–37 GHz for all-weather SST |
| Cloud limitation | Optical chlorophyll-a retrieval blocked by cloud; microwave SST unaffected by thin-to-moderate cloud |
| Archive depth | MODIS Aqua: 2002 to present; VIIRS SNPP: 2012 to present; Sentinel-3A: 2016 to present |
| Epidemiological lead time | Published models: 1 to 3 weeks between bloom onset and clinical case surge; site-specific calibration required |
| Data access | Open: NASA Earthdata (MODIS, VIIRS), Copernicus Data Space (Sentinel-3), PODAAC (GHRSST) |
Analytics Satellize can run
| Weekly SST anomaly map | Pixel-wise departure from 20-year GHRSST climatological baseline; threshold flagging above defined anomaly magnitude | GeoTIFF layer and PDF bulletin, delivered weekly per defined coastal region of interest |
| Chlorophyll-a anomaly map | OC3M or regional Case 2 algorithm applied to MODIS/VIIRS/Sentinel-3 OLCI; anomaly relative to monthly climatology | GeoTIFF layer with per-pixel confidence flag indicating turbidity interference |
| Composite bloom risk index | Simultaneous SST and chlorophyll-a anomaly exceedance scoring, following published Bay of Bengal early-warning model structure | District-level risk score table and map, updated on each cloud-free overpass |
| Turbidity contamination flag layer | Remote-sensing reflectance ratio screening to identify river-plume-influenced pixels where chlorophyll-a retrieval is unreliable | Mask layer delivered alongside every chlorophyll-a product; documented in data dictionary |
| Interannual bloom frequency trend report | Time-series analysis over MODIS archive (2002 to present); Mann-Kendall trend test on annual bloom event count per coastal cell | Annual PDF report with trend maps and statistical summary for ministry or donor reporting |
| Alert feed for health authority integration | Threshold-triggered notification when composite risk index exceeds calibrated cutoff; lag-adjusted to local epidemiological parameters where case data provided | JSON alert feed or email bulletin, configurable per district boundary shapefile supplied by client |
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.