Flood-driven waterborne disease risk from SAR inundation mapping
SAR satellites map flood extents through monsoon cloud cover within hours of acquisition. Overlaid with sanitation infrastructure and population data, those extents become actionable cholera, typhoid and hepatitis A risk surfaces for public health responders.
Sensors
- Sentinel-1 C-band SAR (ESA / Copernicus): 10 m ground range detected resolution in Interferometric Wide Swath mode, 250 km swath, 6-day repeat at the equator and 1 to 3 days at mid-latitudes when both satellites are active. Penetrates cloud and rain with no optical degradation, making it the primary operational sensor for monsoon and cyclone flood mapping. C-band (5.4 GHz) backscatter drops sharply over open water, enabling reliable inundation detection on flat agricultural and peri-urban terrain. Dense forest canopy and multi-storey urban fabric attenuate the signal and can mask standing water beneath them.
- ALOS-2 PALSAR-2 L-band SAR (JAXA): L-band (1.2 GHz) penetrates vegetation canopy significantly better than C-band, improving detection of inundation under rice paddies and light forest cover. 3 to 10 m resolution in spotlight and stripmap modes; 14-day repeat. Tasked by JAXA under its disaster response protocol, which means access is not routine but is available for declared emergencies. Particularly valuable for riverine floodplains with dense riparian vegetation where Sentinel-1 underestimates extent.
- Copernicus Emergency Management Service (CEMS) Rapid Mapping: CEMS activates within hours of a disaster declaration and delivers delineation products, grading maps and change maps derived primarily from Sentinel-1 and contributing commercial SAR. Products include reference, monitoring and post-event delineation layers in GeoPackage and GeoTIFF. Activation latency is typically 3 to 12 hours from request to first product delivery, depending on acquisition scheduling. Products are publicly released and carry documented accuracy assessments.
- MODIS Terra/Aqua near-infrared flood proxy (NASA): 250 m resolution near-infrared and shortwave-infrared bands allow rapid water-body delineation in cloud-free conditions using the Modified Normalised Difference Water Index. Daily global coverage. Useful for large-scale extent estimation and for cross-checking SAR products, but cloud cover during active monsoon events renders it unreliable precisely when it is most needed. Best treated as a corroborating layer rather than the primary detection tool.
What SAR actually sees when a city floods
Synthetic aperture radar works by measuring the microwave energy that bounces back from the surface. Open water is specular: it reflects energy away from the antenna and returns almost nothing. That dark signature, set against the brighter return from roads, buildings and vegetation, is how inundation is detected. The physics is reliable. The geometry is not always cooperative.
Sentinel-1's C-band signal struggles in two common urban flood scenarios. First, dense multi-storey construction creates radar shadow and double-bounce effects that can mimic or obscure flooded ground. Second, water beneath a closed forest canopy is largely invisible to C-band because the canopy scatters the signal before it reaches the surface. ALOS-2 PALSAR-2's L-band partially addresses the second problem, penetrating light to moderate canopy, but urban shadow remains a fundamental limitation of any side-looking radar. Analysts should always report the proportion of the area of interest that falls within these uncertainty zones.
From inundation polygon to disease risk surface
A flood extent polygon is not a disease risk map. The conversion requires at least three additional data layers: the location and type of sanitation infrastructure (pit latrines, septic tanks, sewage outfalls), the proximity of drinking-water sources (hand pumps, surface intakes, open wells), and population density at sub-kilometre resolution. When floodwater overtops or submerges a latrine, faecal contamination enters the water column. The distance that contamination travels before it reaches a drinking-water source, and the number of people using that source, determines exposure probability.
The standard published approach, used in humanitarian response by organisations including WHO and UNICEF, assigns risk tiers by combining inundation depth or duration estimates with sanitation infrastructure density and population data. Depth is not directly observable from SAR amplitude alone; it requires either a digital elevation model differenced against the inundation boundary, or water-level gauge data. The Global Surface Water dataset from the Joint Research Centre provides a useful baseline of normal water extent, against which anomalous inundation can be measured. None of this tells you how much Vibrio cholerae, Salmonella typhi or hepatitis A virus is actually in the water. Pathogen load requires field sampling. Satellite analysis can identify where to send the sampling teams.
The 12-to-24-hour acquisition window that determines operational value
During a cyclone or monsoon event, optical satellites are blind. Cloud cover over a major flood event can persist for days to weeks. Sentinel-1's 6-day revisit at the equator sounds adequate until you consider that the operationally critical window for early public health intervention is the first 48 to 72 hours after peak inundation, when contaminated water is still in contact with sanitation infrastructure and drinking sources.
In practice, Sentinel-1's dual-satellite configuration (Sentinel-1A and 1B, though 1B suffered a power anomaly in 2021 and has been offline since) reduced to a single satellite means the 6-day repeat is the realistic baseline for most tropical latitudes. CEMS can sometimes coordinate ascending and descending pass acquisitions to shorten the effective revisit to 3 days. For events in regions with high-incidence disease histories, pre-positioning analysis pipelines so that processing begins automatically on the next available acquisition is the difference between a useful product and a retrospective report.
Sanitation data is the weakest link, not the satellite
The accuracy of a disease risk surface is bounded by the quality of the sanitation infrastructure map it is overlaid on. In many high-risk countries, latrine and well locations are recorded in household surveys conducted years apart, with spatial accuracy measured in hundreds of metres. Open-defecation prevalence data from WHO/UNICEF Joint Monitoring Programme reports gives national and subnational statistics but not georeferenced point locations.
Urban growth compounds the problem. Informal settlements expand rapidly, and their sanitation infrastructure, such as it is, is rarely mapped at all. Satellite-derived building footprints from high-resolution commercial imagery can partially substitute: settlement extent and density are a reasonable proxy for latrine density in the absence of direct data. The honest position is that the satellite component of this analysis is the most reliable part. The ground-truth data it is combined with is often the primary source of uncertainty in the final risk product.
What a risk tier map can and cannot justify
A well-constructed flood-driven disease risk surface can support several concrete decisions: prioritising oral rehydration salt pre-positioning, directing water-quality testing teams to highest-exposure communities, identifying which health facilities are themselves inundated and therefore offline, and communicating evacuation or boiling-water advisories to specific administrative units. These are decisions with measurable humanitarian value.
What the map cannot do is predict case counts. Cholera outbreaks depend on pre-existing immunity, vaccination coverage, the speed of water-system restoration and the initial pathogen load in the floodwater. All of these are outside the satellite's observational scope. Presenting risk tiers as incidence forecasts overstates the method and erodes trust with epidemiologists who know the difference. The correct framing is prioritised surveillance zones, not predicted outbreak locations.
Satellize runs this combined SAR-plus-infrastructure analysis for government public health agencies and disaster management authorities, drawing on the same open-constellation pipeline used in its Tonga crop-estimation programme. The analytic output is a tiered risk GeoTIFF and a structured briefing note, delivered within hours of a CEMS or manual SAR acquisition.
Archive depth and its use in preparedness planning
Sentinel-1 data extends back to April 2014 for Sentinel-1A. That archive supports retrospective flood frequency mapping: identifying which communities have been inundated in two or more events in a decade, and therefore carry compounded infrastructure damage and elevated baseline disease risk. Combined with historical cholera case data where it exists, frequency maps can inform pre-event vaccine stockpiling and oral cholera vaccine deployment decisions.
The MODIS record stretches back to 2000, providing a 25-year surface water anomaly history at 250 m resolution for basins where cloud-free acquisitions exist. For preparedness rather than response, this longer archive is often more useful than the higher-resolution but shorter SAR record. The two are complementary, not competing.
Typical figures
| Primary sensor spatial resolution | Sentinel-1 IW mode: 10 m ground range detected; ALOS-2 PALSAR-2 stripmap: 3 to 10 m |
| Swath width | Sentinel-1 IW: 250 km; ALOS-2 stripmap: 50 to 70 km |
| Revisit interval (Sentinel-1A single satellite) | 12 days at equator; 6 days at mid-latitudes; CEMS coordination can achieve ~3-day effective revisit |
| Acquisition-to-product latency (CEMS Rapid Mapping) | 3 to 12 hours from activation to first delineation layer |
| Radar frequency | Sentinel-1: C-band, 5.405 GHz; ALOS-2 PALSAR-2: L-band, 1.2 GHz |
| Minimum detectable open-water patch | Approximately 0.1 ha on flat terrain at 10 m resolution; larger under canopy or in urban shadow |
| Cloud penetration | Full: SAR is unaffected by cloud or rain at C- and L-band |
| Sentinel-1 archive depth | April 2014 to present (Sentinel-1A); MODIS flood proxy archive from 2000 |
| Known detection gaps | Dense urban fabric (radar shadow/double-bounce), closed forest canopy (C-band), water depth not directly observable from amplitude alone |
| Output formats | GeoTIFF, GeoPackage, Shapefile, COG; CEMS products also available as WMS/WFS |
Analytics Satellize can run
| SAR inundation extent layer | Thresholded backscatter change detection against pre-event reference image; Otsu or histogram-based threshold selection on sigma-naught difference image | GeoTIFF polygon of flooded area with uncertainty mask for canopy and urban shadow zones |
| Flood duration surface | Multi-temporal SAR stack differenced against Global Surface Water permanent water baseline (JRC); duration estimated from number of acquisitions showing inundation | Raster layer of inundation duration in days per pixel, delivered as COG |
| Sanitation infrastructure exposure index | Spatial join of inundation polygon with georeferenced latrine, well and sewage outfall points; buffer analysis at 50 m, 100 m and 500 m from inundated area boundary | Point feature layer with exposure classification and counts by administrative unit, in GeoPackage |
| Population-weighted disease risk tier map | Overlay of inundation extent, sanitation exposure index and WorldPop or LandScan population density; risk scored by inundation-to-source proximity and population count per tier | Tiered risk raster (four classes) and tabular summary by district, delivered as GeoTIFF plus PDF briefing note |
| Offline health facility alert | Intersection of inundation polygon with WHO Health Facility Registry or national facility point dataset | Alert table listing facility name, type, inundation status and nearest non-inundated alternative, delivered as CSV and structured alert feed |
| Flood frequency and repeat-exposure map | Retrospective SAR archive analysis (2014 to present) counting inundation events per pixel; overlay with current event to flag communities with two or more prior events | Frequency raster and ranked community list for pre-event vaccine or supply pre-positioning decisions, as GeoTIFF and report |
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.