Salt flat and dry lake bed reflectivity mapping for specular microwave interference

Salt pans and dry playas create specular microwave reflection paths that cause destructive interference on terrestrial links. Sentinel-2 SWIR, Landsat surface-reflectance archives, and TanDEM-X roughness data let engineers map these surfaces, track their seasonal extent, and quantify specular risk before a link goes live.

Sensors

Sentinel-2 MSI: 10 m resolution in visible bands, 20 m in SWIR bands (1610 nm and 2190 nm). Five-day revisit at the equator with two satellites. SWIR reflectance distinguishes evaporite crusts from dry soil and sand with high reliability; seasonal mosaics reveal wet-season flooding that temporarily suppresses specular risk.
Landsat 8/9 OLI: 30 m multispectral resolution including SWIR1 (1570 nm) and SWIR2 (2110 nm). Sixteen-day revisit per satellite, eight days combined. Surface-reflectance products (Collection 2 Level-2) provide a consistent archive back to 1984, enabling multi-decade extent and albedo trend analysis for planning-grade assessments.
TanDEM-X: Global DEM at 12 m posting (0.4 arcsecond), with relative vertical accuracy around 2 m in flat terrain. Interferometric coherence and backscatter data also yield surface roughness estimates at C/X-band wavelengths, which correlate directly with the Rayleigh roughness criterion used to classify surfaces as specular or diffuse at microwave frequencies.
SRTM: 30 m global DEM, freely available, sufficient for link geometry modelling and Fresnel zone clearance calculations where TanDEM-X is not licensed. Vertical accuracy degrades in very flat basins due to phase noise, so SRTM is best used as a fallback or sanity check rather than a primary roughness source.

Why a flat white surface is a microwave engineer's quiet adversary

Specular reflection from a flat, high-albedo surface differs fundamentally from the diffuse scattering that dominates urban multipath. When a salt pan or dry playa lies within the first Fresnel zone of a terrestrial microwave link, it can redirect a coherent copy of the transmitted signal toward the receive antenna with a phase shift determined by geometry. If that phase shift is close to 180 degrees, the reflected ray partially cancels the direct ray. The effect is not random fading: it is a stable, geometry-driven notch that can persist for hours or days and does not average away with antenna diversity.

The Rayleigh roughness criterion defines the boundary between specular and diffuse reflection. A surface is effectively specular when its RMS height variation is less than the wavelength divided by 8 times the sine of the grazing angle. At 7 GHz with a grazing angle of two degrees, that threshold is roughly 15 cm. Salt crusts on dry playas routinely meet this condition. The problem is therefore not exotic: it affects any link whose geometry places the Fresnel zone over a salt flat, and it is entirely predictable if the surface is mapped.

What a floating roof gives away: reading salt flats in SWIR

Evaporite minerals, primarily halite and gypsum, have distinctive shortwave-infrared reflectance signatures. In Sentinel-2 Band 11 (1610 nm) and Band 12 (2190 nm), dry salt crusts maintain reflectance values above 0.5 and often above 0.7, well above dry sand (typically 0.3 to 0.5) and far above vegetation or wet soil. A simple normalised difference salinity index, computed from red and SWIR bands, separates evaporite surfaces from surrounding terrain with reasonable reliability in arid environments, though carbonate-rich playas and gypsiferous soils can produce false positives that require visual confirmation.

Seasonal change matters enormously. Many salt flats flood partially during wet seasons or after episodic rainfall, transforming from specular reflectors into shallow brine lakes. A flooded surface has a very different microwave reflection coefficient and a roughness dominated by wind-driven capillary waves, which typically suppresses specular coherence. Landsat's archive depth, extending to 1984 for OLI predecessors and to 1972 for earlier missions, allows analysts to characterise the full seasonal and interannual envelope of a surface rather than relying on a single cloud-free image that may represent an atypical state.

From roughness map to reflection probability

TanDEM-X provides the geometric input the Rayleigh criterion needs. From the global 12 m DEM, analysts derive RMS height variation over the Fresnel zone footprint for a given link geometry. That figure, combined with the link frequency and grazing angle, yields a binary or probabilistic specular classification. The coherence layer from TanDEM-X interferometric processing adds a second dimension: low coherence over a nominally flat surface often indicates temporal surface change between the two SAR acquisitions, which is itself a proxy for seasonal instability.

The Fresnel zone footprint is elliptical and depends on link distance, frequency, and antenna heights. At 7 GHz over a 40 km link with antennas at 30 m, the first Fresnel zone radius at the midpoint is roughly 160 m. A salt flat need not be enormous to be dangerous; a few hundred metres of smooth crust centred on the specular point is sufficient. Mapping at 10 to 30 m resolution is therefore adequate for most practical assessments, provided the specular point location is calculated accurately from the link geometry.

Honest limits of the method

Cloud cover is the primary operational constraint. Sentinel-2 and Landsat are passive optical sensors, and persistent cloud over wet-season playas is precisely when surface state is most uncertain. Synthetic aperture radar, including Sentinel-1 C-band, can penetrate cloud but does not directly measure optical reflectance; it measures backscatter, which is a function of both roughness and dielectric constant. A flooded salt flat looks very different in SAR than a dry one, which is useful, but the relationship between C-band backscatter and microwave-link reflection coefficient at 7 or 11 GHz is not straightforward and requires careful physical reasoning rather than direct substitution.

Landsat Collection 2 surface-reflectance products apply atmospheric correction, but residual aerosol errors in dusty arid environments can shift SWIR reflectance by several percent, which matters when the classification threshold is tight. TanDEM-X roughness estimates integrate height variation over the DEM posting scale, not at the millimetre scale relevant to the Rayleigh criterion at high microwave frequencies. For links above 18 GHz, where wavelengths are short enough that even modest surface texture becomes diffuse, the roughness classification becomes conservative and may overstate specular risk.

Finally, this analysis identifies surfaces that can create specular interference. Whether a given link actually suffers depends on antenna height, tilt, and the precise geometry of the specular point. The satellite-derived surface map is an input to propagation modelling, not a substitute for it.

Putting the analysis into a planning workflow

The practical output is a seasonal surface classification raster, typically three states: dry specular-risk, seasonally variable, and non-specular. This is overlaid with the Fresnel zone footprint for each candidate link route, computed from the link geometry and frequency plan. Where the specular point falls within a dry-specular-risk zone, the analyst flags the link for further propagation modelling and potentially recommends antenna height adjustment, space diversity, or route realignment.

Satellize runs this type of surface characterisation on open-constellation data, combining Sentinel-2 seasonal mosaics with Landsat archive depth and TanDEM-X roughness layers. The workflow draws on the same spectral processing infrastructure used in the Tonga crop-estimation programme, adapted for arid-surface classification rather than agricultural phenology. Deliverables are GIS-ready rasters and link-specific risk summaries, not raw imagery dumps.

For network operators planning backhaul across the Sahel, the Arabian Peninsula, the Atacama, the Altiplano, or Central Asian basins, the cost of a specular fade event on a high-capacity link is orders of magnitude larger than the cost of a pre-deployment surface assessment. The physics is deterministic. The surface data is largely free. The gap is the analysis.

Typical figures

Optical mapping resolution	10 m (Sentinel-2 visible), 20 m (Sentinel-2 SWIR), 30 m (Landsat OLI SWIR)
Elevation / roughness resolution	12 m posting (TanDEM-X global DEM); 30 m (SRTM fallback)
Sentinel-2 revisit	5 days at equator (two-satellite constellation)
Landsat 8/9 combined revisit	8 days at equator
Key spectral bands	SWIR1 ~1610 nm, SWIR2 ~2110 nm (salt/evaporite discrimination); Red ~665 nm (normalised indices)
TanDEM-X relative vertical accuracy (flat terrain)	~2 m (1-sigma); sufficient for Fresnel-zone clearance and RMS roughness estimation
Landsat archive depth	Surface-reflectance Collection 2 Level-2 from 1984 (OLI predecessors TM/ETM+ extend to 1972)
Minimum mappable salt flat extent	~1 ha reliably at 20 m SWIR resolution; sub-hectare features may be missed
Deliverable formats	GeoTIFF seasonal classification rasters, GeoPackage / Shapefile extent polygons, link-specific PDF risk summaries

Analytics Satellize can run

Seasonal salt-flat extent classification	Normalised difference salinity / SWIR-ratio thresholding on Sentinel-2 and Landsat surface-reflectance time series; change detection across wet and dry seasons	Three-class GeoTIFF raster (dry specular, seasonally variable, non-specular) per season, covering the link corridor buffer
Surface roughness classification per Rayleigh criterion	RMS height variation derived from TanDEM-X 12 m DEM over Fresnel-zone footprint; evaluated against frequency- and grazing-angle-specific roughness threshold	Per-link specular probability score and roughness map layer in GeoPackage format
Fresnel zone specular point overlay	Geometric computation of first Fresnel zone ellipse from link endpoints, antenna heights, and frequency; intersection with surface classification raster	Vector layer of specular point locations and Fresnel ellipses, colour-coded by surface risk class, in GIS-ready format
Multi-year albedo trend and interannual variability report	Landsat Collection 2 SWIR surface-reflectance time series analysis; percentile-based characterisation of dry-season albedo stability over the archive period	PDF report with time-series plots and summary statistics per identified salt flat within the study area
Wet-season inundation probability map	Landsat and Sentinel-2 water index (MNDWI) applied across multi-year archive; frequency-of-inundation raster derived from annual wet-season composites	Inundation frequency GeoTIFF and tabular summary of months per year each salt flat is likely flooded
Link-specific specular interference risk summary	Integration of surface classification, roughness score, Fresnel geometry, and seasonal variability into a structured risk matrix per candidate link route	Structured PDF or spreadsheet report flagging high-risk links, recommended mitigation options, and confidence levels

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.