Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range resolution in Interferometric Wide Swath mode, 6-day repeat at mid-latitudes (12-day per orbit). Cloud-independent. Calm water appears as a very low backscatter region (typically below -15 dB in VV polarisation), giving reliable water/land discrimination year-round. Wind roughening on large surfaces can reduce contrast, so calm-condition acquisitions are preferred.
- Sentinel-2 MSI (ESA): 10 m visible and NIR bands, 20 m SWIR bands, 5-day revisit with both satellites. NDWI (Green minus NIR, divided by Green plus NIR) or MNDWI (Green minus SWIR) delineates shorelines to sub-pixel accuracy in cloud-free conditions. Optical only; cloud cover is a hard limit and can be persistent over months in tropical or montane corridors.
- Landsat 8/9 OLI (USGS/NASA): 30 m multispectral, 16-day repeat per satellite (8-day combined). MNDWI performs well for permanent and seasonal water body mapping. The Landsat archive extends to 1972 (Landsat 1 MSS), making it the deepest public record for historical seasonal extent, useful for characterising worst-case inundation years.
- Planet SuperDove: 3 m resolution, near-daily revisit over most land areas. Useful for refining shoreline geometry around reservoir edges and narrow river reaches where 10 m pixels are ambiguous. Commercial tasking; archive and revisit terms depend on client licence. No SAR capability, so cloud cover remains a constraint.
Why a lake is not just terrain
Standard microwave propagation tools, including ITU-R P.526 and most commercial path-profile packages, treat the Earth's surface as a terrain-plus-clutter problem. They account for diffraction over ridges, vegetation attenuation and building obstructions. What they do not model by default is the specular reflection geometry that arises when a calm water surface sits within the Fresnel zone of a point-to-point link.
A calm lake or reservoir behaves as a near-perfect reflector at microwave frequencies. The reflected ray travels a slightly longer path than the direct ray. When the path-length difference equals an odd multiple of half a wavelength, the two rays arrive at the receive antenna in destructive interference. The resulting fade can exceed 20 dB on an otherwise clear line-of-sight link, and it recurs as atmospheric conditions shift the effective ray angles. Engineers who discover this during commissioning, rather than during planning, face expensive antenna height changes or passive repeater installations.
Locating the specular point: the geometry planners need
The specular reflection point is not simply the midpoint of the link. Its position along the path depends on the antenna heights at each end and the link distance. For a flat reflecting surface, the specular point lies at the position where the angle of incidence equals the angle of reflection. If the two towers are at different heights, the specular point shifts toward the lower antenna. For a 20 km link with a 30 m tower at one end and a 60 m tower at the other, the specular point falls roughly one-third of the path from the taller tower, not at the midpoint.
The first Fresnel zone radius at the specular point determines how large a water surface must be to contribute meaningfully to the reflected ray. At 7 GHz over a 20 km link, the first Fresnel zone radius at mid-path is approximately 90 m. A reservoir 200 m wide at the specular point is therefore a significant reflector; a narrow irrigation canal of 10 m width is not. This geometry must be computed for each candidate link before deciding whether water mapping is warranted, and it changes with frequency: at 23 GHz the same geometry produces a Fresnel zone radius of roughly 50 m, tightening the target but also meaning smaller water bodies become relevant at higher bands.
What a floating roof gives away: reading SAR for water
Sentinel-1 C-band SAR detects open water through the absence of backscatter. Land surfaces, vegetation and built structures return energy to the sensor; calm water reflects the radar pulse away from the sensor, appearing dark. The contrast is typically large enough to apply a simple threshold in VV polarisation, though automated methods such as the Otsu threshold or change-detection against a land-cover prior are more reliable at ambiguous shorelines.
The practical limitation is wind. A surface wind of around 3 m/s is sufficient to roughen water enough to raise backscatter by several decibels, reducing contrast with surrounding land. For propagation planning this is actually informative: a wind-roughened surface also produces a less coherent specular reflection, so the worst-case multipath condition corresponds to the calm-water SAR signature. Acquiring multiple Sentinel-1 passes and selecting the lowest-backscatter epoch gives the engineer the worst-case reflector geometry, not just the average.
Seasonal variation matters. Many reservoirs in sub-Saharan Africa, South Asia and inland Australia vary by tens of metres in shoreline position between wet and dry seasons. A link planned against dry-season imagery may cross a water surface that did not exist when the terrain model was built. The Landsat archive, combined with the Global Surface Water dataset published by the European Commission's Joint Research Centre (derived from Landsat), provides monthly water occurrence probabilities back to 1984, giving a statistically grounded worst-case extent.
Sentinel-2 NDWI: sharper edges, honest limits
Once SAR has confirmed that a water body intersects the specular zone, Sentinel-2 NDWI refines the shoreline geometry to 10 m. This matters because the width of the water surface at the specular point determines the effective reflection aperture and therefore the severity of the fade. A reservoir that is 500 m wide at the specular point presents a very different risk from one that is 80 m wide.
The honest limit: NDWI is useless under cloud. In equatorial regions, months can pass without a cloud-free Sentinel-2 acquisition over a given 100 km corridor. The practical workflow is to use Sentinel-1 SAR for all-weather detection and extent estimation, then apply Sentinel-2 NDWI when a cloud-free acquisition is available to sharpen the shoreline for the specular-point calculation. Planet SuperDove at 3 m can resolve narrow river channels and reservoir inlets that 10 m pixels smear, but it adds cost and is not always necessary once the Fresnel-zone geometry has been computed.
From water map to fade margin: the planning output
The analytic output for a network planner is not simply a map of water. It is a structured assessment of each candidate link segment: does a water surface of sufficient width intersect the first Fresnel zone at the specular point, and if so, what is the worst-case seasonal extent? That output feeds directly into ITU-R P.530 two-ray model calculations, where the engineer can assess whether the existing antenna heights provide adequate clearance or whether the link geometry needs adjustment.
Satellize runs this workflow on Sentinel-1 and Sentinel-2 open-constellation data, with optional Planet tasking for high-resolution shoreline confirmation. The same pipeline underpins the seasonal-extent work done for the Kingdom of Tonga crop-estimation programme, where distinguishing inundated paddies from dry land under variable cloud cover required the same SAR-optical fusion approach.
A word on what satellite data cannot do: it cannot measure the dielectric constant of the water surface, which affects reflection coefficient, nor can it resolve sub-metre ripple structure that determines whether a surface is truly specular at a given frequency. Those parameters require local meteorological data and engineering judgement. Satellite mapping defines where the water is and how large it is. The propagation physics still require the engineer.
Typical figures
| Sentinel-1 spatial resolution | 10 m (IW mode, ground range); 20 m (EW mode) |
| Sentinel-1 revisit | 6 days at mid-latitudes (both satellites); 12 days per satellite |
| Sentinel-2 water index resolution | 10 m (NDWI using Band 3 and Band 8); 20 m (MNDWI using Band 11) |
| Sentinel-2 revisit | 5 days (both satellites combined); cloud cover limits usable acquisitions |
| Landsat 8/9 resolution and revisit | 30 m multispectral; 8-day combined revisit; archive to 1972 |
| Planet SuperDove resolution | 3 m; near-daily revisit; commercial licence required |
| Minimum detectable water body (SAR) | Approximately 0.5 ha in calm conditions at 10 m resolution; smaller features unreliable |
| Seasonal archive depth | Sentinel-1 from 2014; Landsat from 1972; Global Surface Water monthly occurrence from 1984 |
| Delivery format | GeoTIFF water-extent rasters, GeoJSON shoreline polygons, per-link specular-point assessment report (PDF) |
| Frequency applicability | Specular geometry analysis applicable from 2 GHz to 80 GHz; Fresnel zone radius scales with inverse square root of frequency |
Analytics Satellize can run
| Seasonal water extent map along link corridor | Sentinel-1 SAR thresholding (Otsu or histogram-based) across multiple epochs; worst-case and median extent derived | GeoTIFF raster stack with per-pixel water occurrence probability; corridor-clipped GeoJSON polygons |
| Specular-point location and Fresnel zone intersection assessment | Two-ray geometry calculation (ITU-R P.526 formulation) using antenna heights, link distance and frequency; intersection tested against water-extent polygons | Per-link tabular report: specular point coordinates, first Fresnel zone radius at that point, water surface width at intersection, worst-case and median season |
| Refined shoreline geometry at specular zone | Sentinel-2 MNDWI (Green minus SWIR2) computed from cloud-free acquisitions; sub-pixel shoreline vectorisation | GeoJSON shoreline polygon at 10 m accuracy; flagged as cloud-limited where no clear acquisition available within the analysis window |
| Historical worst-case inundation extent | Landsat-derived Global Surface Water monthly occurrence product (JRC); maximum observed extent extracted for each water body intersecting the corridor | Maximum-extent GeoJSON polygon with year of occurrence; comparison against current Sentinel-1 extent |
| High-resolution shoreline confirmation (narrow channels and inlets) | Planet SuperDove 3 m NDWI; applied selectively where 10 m pixels are ambiguous at the specular point | 3 m GeoTIFF water mask and vectorised shoreline; flagged sections where width uncertainty affects fade-margin calculation |
| Change alert for reservoir level shifts post-commissioning | Sentinel-1 monthly water-extent monitoring against baseline; area-change threshold triggers alert | Monthly GIS layer update with change-flagged email alert when water extent at specular zone exceeds defined threshold |
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.