Salar lithium brine surface extent and seasonal change mapping

SAR backscatter and optical indices distinguish brine pools, halite crust and clay margins on lithium salars through cloud and seasonal flood cycles, giving operators and regulators a consistent surface-extent record.

Sensors

Sentinel-1 C-band SAR (ESA): 5.405 GHz C-band, 10 m ground range resolution in Interferometric Wide Swath mode, 6-day repeat at the equator (12-day single-satellite). Specular reflection from smooth brine surfaces drives backscatter down to roughly -20 dB or below, giving a reliable open-brine proxy regardless of cloud cover. Dry halite crust returns 6-10 dB higher.
Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands, 20 m shortwave-infrared (SWIR) bands, 5-day revisit with two satellites. SWIR bands (1610 nm and 2190 nm) respond to brine salinity and surface moisture; NDWI and modified NDWI variants separate open water from wet salt. Cloud cover over Andean salars during the wet season (December-March) can exceed 60% of acquisitions, making optical-only monitoring unreliable for that period.
ALOS-2 PALSAR-2 (JAXA): L-band (1.27 GHz) SAR at 3-10 m resolution depending on mode. Longer wavelength penetrates thin dry-salt crusts and interacts with subsurface moisture, complementing C-band surface returns. Useful for distinguishing saturated subsurface brine from surface-dry areas that C-band reads as dry. Revisit is 14 days; tasking is commercial.
Landsat-9 OLI (USGS/NASA): 30 m multispectral, 15 m panchromatic, 16-day revisit. Free archive back to 1972 (Landsat 1-9 combined) provides the longest continuous surface-extent record for any salar, essential for baseline and decadal trend analysis. SWIR band 6 (1570 nm) and band 7 (2110 nm) are particularly useful for salt and brine discrimination.

Why a smooth brine pool is a radar dark spot

C-band radar illuminates the salar surface at an incidence angle of roughly 20-46 degrees in Sentinel-1 IW mode. When the surface is flat and liquid, the outgoing pulse reflects specularly away from the sensor, much like a mirror angled away from your eye. The instrument records near-zero backscatter. Dry halite crust, by contrast, is rough at centimetre scales and scatters energy back strongly. The contrast between the two states is typically 8-12 dB, large enough to classify reliably even with a simple threshold.

This physics-based contrast is the reason SAR outperforms optical sensors during the Andean wet season. Cloud cover, which is optically opaque, is effectively transparent at C-band. An operator monitoring a salar in the Atacama or Puna regions between December and March can expect usable Sentinel-1 imagery every 6-12 days regardless of weather. Optical sensors may go weeks without a cloud-free view of the same area.

What a floating roof gives away: reading the halite polygon structure

Salars are not uniform. A mature lithium salar typically presents several distinct surface units: open brine pools (natural or managed), a polygonal halite crust of varying thickness, wet clay and silty margins fed by alluvial fans, and evaporite terraces at higher elevations. Each unit has a distinct SAR backscatter signature and a distinct optical index profile.

Halite polygons form because brine evaporates and salt precipitates at the polygon edges, raising rims above the interior. In Sentinel-2 false colour, the polygon interiors appear slightly darker in SWIR because residual moisture persists there longer. In SAR, the raised rims scatter slightly more than the interiors. Mapping these sub-units matters operationally: clay margins are geotechnically distinct from halite crust, and confusing them has consequences for infrastructure siting and brine extraction well placement.

The combination of Sentinel-1 backscatter intensity and Sentinel-2 SWIR-based indices, fused at 10 m, resolves polygon structure down to features of roughly 20-30 m across. Smaller features, such as individual brine seeps or narrow drainage channels, require commercial tasking at 1-3 m resolution and are outside the scope of open-constellation mapping.

Seasonal change: the wet-season flood cycle and what it means for operators

Andean salars experience pronounced seasonal flooding driven by the South American Monsoon. Rainfall in the catchment, primarily between December and March, raises the brine table and inundates low-lying halite surfaces. The flooded area can expand by tens of square kilometres on a large salar within a few weeks. By September the surface is typically at its driest annual extent.

For a lithium operation, this cycle matters in several ways. Evaporation pond management, brine extraction scheduling and road access across the salar surface all depend on knowing where the wet-dry boundary sits week by week. Sentinel-1's 6-day repeat provides a time series dense enough to track the flood front advance and retreat. A multi-year SAR archive, now extending back to Sentinel-1A's launch in 2014, allows operators to characterise inter-annual variability and identify anomalous flood years driven by El Niño or La Niña conditions.

One honest limit: SAR backscatter over very thin brine films on halite (a few millimetres deep) can be ambiguous. The surface may be wet enough to reduce backscatter moderately without reaching the specular minimum associated with open pools. Optical NDWI from Sentinel-2, when cloud-free acquisitions are available, resolves this ambiguity by confirming surface reflectance in the green and near-infrared bands.

Archive depth and the long baseline problem

Lithium project timelines span decades. Environmental impact assessments, water-use permits and closure planning all require understanding of natural variability before operations began. Landsat's archive is the only freely available source with coverage back to the early 1970s at consistent spatial resolution. Landsat 5 TM, Landsat 7 ETM+ and Landsat 8-9 OLI together provide a 50-year surface-extent record for most South American salars.

Extracting a consistent time series from this archive requires careful handling of sensor differences, atmospheric correction and the shift from 60 m (early Landsat) to 30 m (Landsat 4 onwards) pixel sizes. Pre-1984 data is sparser and noisier. Still, even a coarse pre-operational baseline is far more defensible in a regulatory proceeding than no baseline at all. Sentinel-1 and Sentinel-2 data from 2014 and 2015 onwards respectively provide the dense, high-quality time series that overlaps with most current project development phases.

Honest limits of the method

SAR-based brine mapping has three well-documented failure modes. First, wind roughens brine pool surfaces and raises backscatter, potentially causing the classifier to read open brine as wet crust. Acquisitions during high-wind events should be flagged and, where possible, cross-checked against wind records from nearby stations. Second, C-band cannot distinguish lithium-rich brine from other saline water bodies on backscatter alone; the method maps surface extent, not brine grade. Grade estimation from spectral data is a separate and more uncertain problem, covered on the adjacent evaporation pond brine chemistry page in this library. Third, at 10 m resolution, narrow access causeways and small test wells are below the reliable detection threshold.

L-band PALSAR-2 helps with the wind ambiguity because its longer wavelength is less sensitive to small-scale surface roughness, but it carries a commercial tasking cost and a 14-day revisit. For most operational monitoring programmes, the practical approach is to use Sentinel-1 as the primary cloud-independent layer, Sentinel-2 as the cloud-permitting optical check, and PALSAR-2 for targeted acquisitions during ambiguous periods or for subsurface moisture questions.

Satellize runs this multi-sensor fusion workflow on open constellations and can add commercial tasking under client licence. The Tonga crop-estimation programme demonstrated the same time-series architecture, adapted here to a very different physical problem.

From pixels to a usable surface-extent product

The standard processing chain starts with Sentinel-1 Ground Range Detected (GRD) scenes in VV and VH polarisation, terrain-corrected using a SRTM or Copernicus DEM, and filtered with a refined Lee or Gamma-MAP speckle filter. A backscatter threshold, calibrated against known open-brine and dry-crust training areas within the salar, produces a binary flood mask. That mask is then refined using Sentinel-2 NDWI and SWIR band ratios on cloud-free dates, and the two layers are combined into a surface-unit classification: open brine, wet crust, dry halite, clay margin.

Change products are generated by differencing classified maps across dates, producing area statistics for each surface unit per epoch. Delivered as GeoTIFF rasters and vector polygon layers, these feed directly into GIS environments used by environmental and operations teams. Alert thresholds, for example when open brine extent exceeds a permitted boundary, can be set against the time series and triggered automatically on new acquisitions.

Typical figures

Primary SAR spatial resolution	10 m (Sentinel-1 IW mode, ground range); 3-10 m (ALOS-2 PALSAR-2, mode-dependent)
Primary optical spatial resolution	10 m visible/NIR, 20 m SWIR (Sentinel-2); 30 m multispectral (Landsat-9)
SAR revisit (Sentinel-1)	6 days at the equator with two satellites; 12 days single-satellite
Optical revisit (Sentinel-2)	5 days with two satellites; cloud-free acquisition rate over Andean salars drops to roughly 40% during wet season (December-March)
SAR frequency and polarisation	C-band 5.405 GHz, VV and VH (Sentinel-1); L-band 1.27 GHz, HH/HV (ALOS-2)
Minimum detectable open-brine feature	Reliably ~0.5 ha at 10 m SAR resolution; sub-hectare features require commercial tasking
Archive depth	SAR: Sentinel-1 from 2014; optical: Landsat from 1972 (Landsat 1), continuous 30 m from 1984 (Landsat 5)
Typical processing latency	24-72 hours after scene acquisition for standard GRD products; near-real-time pipelines achievable with pre-configured workflows
Delivery formats	GeoTIFF (classified raster), GeoPackage or Shapefile (vector polygons), time-series CSV (area statistics per class per date)
Coverage	Global; all major South American salars within Sentinel-1 and Sentinel-2 systematic acquisition zones

Analytics Satellize can run

Seasonal surface-extent map	Sentinel-1 backscatter threshold classification (VV polarisation, terrain-corrected GRD), refined with Sentinel-2 NDWI	GeoTIFF and vector polygon layer per acquisition date, with area statistics per surface class
Wet-season flood front tracking	Multi-temporal SAR change detection; binary flood mask differenced across 6-day Sentinel-1 epochs	Time-series animation and CSV of flooded area (ha) per date, with anomaly flags for exceedance of defined thresholds
Surface-unit classification (brine / halite / clay margin)	SAR backscatter intensity combined with Sentinel-2 SWIR band ratios and NDWI; supervised classification trained on salar-specific ground-truth or spectral library	Multi-class GeoTIFF and polygon layer; per-class area table
Inter-annual variability report	Annual dry-season and wet-season composites from Landsat and Sentinel archives; linear trend and anomaly detection over available record	PDF report with time-series charts and maps; GeoTIFF stack of annual composites
Permitted-boundary exceedance alert	Automated comparison of classified brine extent against operator-defined permit polygon; triggered on each new Sentinel-1 acquisition	Email or API alert with exceedance area (ha) and difference map attached
L-band subsurface moisture supplement	ALOS-2 PALSAR-2 HH backscatter analysis for shallow subsurface brine saturation, cross-referenced against C-band surface classification	Supplementary GeoTIFF layer flagging areas where L-band and C-band classifications diverge, indicating possible subsurface saturation under dry surface crust

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.