Soil salinisation mapping from mine waste leachate in surrounding farmland
Leachate from tailings facilities and heap leach pads elevates soil salinity in adjacent farmland in patterns visible in Sentinel-2 SWIR ratios and PRISMA hyperspectral data. Mapping salt-crust extent, severity, and seasonal wetting cycles builds the evidentiary record for liability attribution and remediation planning.
Sensors
- Sentinel-2 MSI: 10 m visible bands, 20 m SWIR bands (B11 at 1610 nm, B12 at 2190 nm); 5-day revisit at mid-latitudes from twin-satellite constellation. SWIR band ratios (B11/B4, B12/B8A) are the primary workhorse for salt-crust brightness and clay-hydroxyl absorption. Free archive from 2015.
- ASI PRISMA: Hyperspectral imager covering 400–2500 nm in 239 contiguous bands at ~30 m spatial resolution and roughly 29-day revisit. Resolves diagnostic absorption features of halite (~2170 nm), gypsum (~1750 nm and ~2220 nm), and sulphate minerals produced by mine-derived leachate, enabling mineralogical discrimination that broadband sensors cannot achieve.
- Landsat 8/9 OLI-TIRS: 30 m SWIR bands (Band 6 at 1570 nm, Band 7 at 2110 nm) with 16-day revisit per satellite, 8-day combined. Archive to 1982 (Landsat 5 TM) enables multi-decade baseline for pre-mine soil condition. TIRS thermal band (Band 10, 100 m resampled to 30 m) can indicate evaporative cooling over moist leachate zones.
- AVIRIS-NG (airborne): Airborne hyperspectral sensor covering 380–2510 nm in ~5 nm steps at spatial resolutions typically 1–8 m depending on flight altitude. Used for high-confidence ground-truth and mineralogical validation of satellite-derived salinity maps. Not a repeat-pass system; campaign-based.
Why salt from a tailings pad looks different from natural salinity
Natural soil salinity tends to accumulate in topographic lows, along drainage lines, and in irrigated fields with poor drainage. Mine-derived leachate follows a different logic. It migrates downslope from a tailings storage facility or heap leach pad, tracking subsurface hydraulic gradients and preferential flow paths, and it carries a chemical signature shaped by the ore being processed. Sulphate-dominated leachate from sulphide tailings produces different secondary minerals than carbonate-buffered drainage. Gypsum (calcium sulphate) and jarosite are common precipitates; halite appears where evaporation concentrates sodium from process water. These minerals have distinct shortwave infrared absorption features that broadband SWIR ratios can detect at the landscape scale, and that hyperspectral sensors can resolve into specific mineral identities.
The spatial pattern is the first diagnostic. A salt plume radiating from a facility boundary in a direction inconsistent with regional topographic salinity is already an anomaly worth investigating. Seasonal behaviour adds a second diagnostic: leachate-affected zones typically show elevated SWIR brightness after wet-season recharge events and a characteristic drying crust in the dry season, whereas naturally saline soils in the same landscape may behave differently. Tracking that wetting-drying cycle across multiple years is where the archive depth of Sentinel-2 and Landsat becomes genuinely useful.
What the SWIR bands actually measure, and where they fall short
Soil salinity does not have a single, clean spectral signature. What SWIR bands detect is a combination of surface salt-crust brightness (high reflectance across SWIR when halite or gypsum dominates the surface), suppressed vegetation vigour (low NDVI, elevated SWIR/NIR ratios in stressed crops), and clay-hydroxyl absorption features around 2200 nm that shift with mineralogy. The Salinity Index variants most commonly applied to Sentinel-2, such as SI = sqrt(B4 × B11) or the combination of B11 and B12 with visible bands, are empirical proxies rather than direct salinity measurements. They correlate with electrical conductivity in field studies, but the relationship is site-specific and requires ground-truth calibration to assign meaningful conductivity values.
Cloud cover is the persistent operational constraint. In humid tropical mining regions, Sentinel-2 may deliver only a handful of usable cloud-free observations per quarter over a given field. Seasonal compositing mitigates this but blurs the temporal precision of event attribution. At 20 m, Sentinel-2 SWIR also cannot resolve small-scale heterogeneity within a single field; PRISMA at 30 m is similar. For sub-10 m spatial detail, an AVIRIS-NG campaign or a commercial hyperspectral satellite such as those from the DESIS or future CHIME mission would be needed. PRISMA's 29-day revisit is too coarse to track rapid leachate events after heavy rainfall without supplementary Sentinel-2 data.
Hyperspectral discrimination: naming the minerals, not just the brightness
PRISMA's 239-band coverage across 400–2500 nm allows spectral unmixing and matched-filter approaches that identify specific evaporite and secondary sulphate minerals by their diagnostic absorption features. Gypsum shows doublet absorptions near 1750 nm and 2220 nm. Jarosite, a potassium iron sulphate common in acid mine drainage environments, has absorptions near 900 nm and 2270 nm. Halite is spectrally bland in the visible but shows a broad feature near 2170 nm. Identifying which minerals are present in a leachate plume is not merely academic: it constrains the geochemical source, supports or undermines alternative explanations (such as agricultural fertiliser runoff), and provides the kind of mineralogical specificity that environmental lawyers and regulators can work with.
Published studies using AVIRIS and AVIRIS-NG over mine-affected sites have demonstrated detection of sulphate mineral assemblages at spatial scales of a few metres, and PRISMA has been validated against ground spectroscopy in several European and South American salinisation contexts. The honest caveat is that atmospheric correction over bright salt surfaces is difficult: water vapour absorption overlaps with some mineral features, and adjacency effects from nearby dark vegetation can introduce artefacts. Careful application of ATCOR or ACORN atmospheric correction, with field spectroradiometer validation, is not optional in high-stakes liability contexts.
Building the seasonal wetting-drying record for attribution
A single image showing salt-affected soil near a tailings facility is suggestive. A time series showing that the affected area expanded after each wet season, tracks the facility's operational history, and correlates spatially with the facility's drainage gradient is considerably more compelling. Landsat's archive to 1982 allows a pre-mine baseline to be established for sites with long operational histories. Sentinel-2's 5-day revisit from 2015 provides the temporal density needed to track individual recharge events.
The practical workflow involves constructing seasonal composites (dry-season peak crust expression, wet-season minimum) for each year in the archive, then computing change in salinity index extent and intensity against the pre-facility baseline. Pixels that transition from non-saline to saline classifications and remain elevated across multiple dry seasons are the strongest candidates for leachate attribution. Pixels that recover after a mine closure or liner repair are equally informative for demonstrating remediation efficacy. Satellize applies this time-series approach on open constellations for clients requiring defensible spatial evidence, drawing on the same analytical methods developed for its Tonga crop-estimation programme, where multi-date compositing underpins area and condition estimates.
One genuine ambiguity deserves acknowledgement: irrigation-induced salinity can produce similar spectral signatures and similar spatial expansion patterns. Disentangling the two requires combining the spectral record with irrigation history, hydrogeological modelling, and where possible, field water sampling. Remote sensing narrows the hypothesis space; it rarely closes it alone.
From map to evidence: what the output needs to say
Regulators and courts require more than a coloured map. The deliverable needs to state the spatial resolution and its implications for minimum detectable patch size (a 20 m Sentinel-2 pixel cannot confirm salinity in a 5 m strip between crop rows), the confidence intervals on any area estimate, the atmospheric correction method applied, and the field validation dataset used to calibrate the salinity index to electrical conductivity. Uncertainty quantification is not a weakness in the report; its absence is.
Practically useful outputs include: a GIS polygon layer of salt-crust extent by severity class for each season in the archive, a pixel-level time series of salinity index values exported as a raster stack, a change-detection summary showing area gained and lost per year relative to the pre-facility baseline, and, where PRISMA data is available, a mineral abundance map distinguishing gypsum, jarosite, and halite zones. These layers feed directly into remediation planning, because the mineralogy informs which amendment strategy (lime application, gypsum leaching, drainage intervention) is appropriate for a given zone.
Typical figures
| Primary spatial resolution (SWIR) | 20 m (Sentinel-2 B11/B12); 30 m (Landsat OLI Band 6/7, PRISMA); 1–8 m (AVIRIS-NG, altitude-dependent) |
| Revisit frequency | 5 days (Sentinel-2 twin constellation, mid-latitudes); 8 days combined (Landsat 8+9); ~29 days (PRISMA); campaign-only (AVIRIS-NG) |
| Spectral bands relevant to salinity | SWIR 1 (~1570–1610 nm), SWIR 2 (~2110–2190 nm), NIR, Red; full 400–2500 nm hyperspectral for PRISMA and AVIRIS-NG |
| Minimum detectable salt-crust patch | ~1–2 ha reliably at 20 m Sentinel-2 SWIR; sub-hectare with AVIRIS-NG at 4 m resolution |
| Archive depth | Sentinel-2: 2015 to present; Landsat: 1982 to present (TM/ETM+/OLI); PRISMA: 2019 to present |
| Latency (open data) | Sentinel-2 L2A: typically 1–3 days after acquisition via Copernicus Data Space; Landsat Collection 2: 1–2 days |
| Salinity index type | Empirical SWIR-based indices (SI, NDSI variants); requires site-specific calibration against field ECe measurements for quantitative output |
| Atmospheric correction requirement | Essential for SWIR and hyperspectral; ATCOR, Sen2Cor (Sentinel-2), or ACORN for PRISMA/AVIRIS; field spectroradiometer validation recommended for liability use |
| Delivery formats | GeoTIFF raster stacks, polygon shapefiles or GeoJSON by severity class, CSV time-series tables, PDF evidence report with uncertainty quantification |
Analytics Satellize can run
| Seasonal salt-crust extent map | SWIR band ratio compositing (Sentinel-2 B11/B4, B12/B8A) with threshold classification calibrated to field ECe; cloud-masked seasonal median composite | Annual dry-season and wet-season GeoTIFF and polygon layer, severity-classified (low/moderate/high salinity) |
| Multi-decade change detection against pre-facility baseline | Landsat TM/ETM+/OLI time-series analysis; pixel-level salinity index trend using Mann-Kendall or linear regression on annual dry-season composites | Change raster and summary table showing area (ha) gained per severity class per year; PDF trend report with confidence intervals |
| Mineral species map (gypsum, jarosite, halite) | PRISMA hyperspectral spectral unmixing or matched-filter against laboratory reference spectra (USGS spectral library); atmospheric correction via ATCOR | Mineral abundance GeoTIFF per species; interpreted polygon overlay distinguishing mine-derived sulphate minerals from background evaporites |
| Leachate migration pathway delineation | Multi-date SWIR salinity index combined with digital elevation model flow-accumulation analysis; spatial correlation of salinity expansion direction with facility drainage gradient | GIS vector layer of inferred migration pathways; statistical test of directional association with tailings facility |
| Vegetation stress co-mapping | NDVI and red-edge chlorophyll index (Sentinel-2 B5/B4, B7/B5) time series; identification of pixels showing persistent NDVI suppression coincident with elevated salinity index | Crop stress overlay aligned with salinity extent map; area estimate of agricultural land affected above defined ECe proxy threshold |
| Wetting-drying cycle event log | High-frequency Sentinel-2 SWIR time series (5-day revisit); change-point detection on pixel-level salinity index to identify recharge events following rainfall | Per-pixel event log CSV; alert layer flagging pixels with anomalous SWIR increase within 10 days of a rainfall event above defined threshold |
| Remediation progress verification | Before/after salinity index comparison at intervention polygons; PRISMA mineral map update to confirm reduction in sulphate mineral abundance post-treatment | Biannual verification GeoTIFF with change statistics; compliance summary table for regulator submission |
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.