Soil salinity mapping and salt-affected cropland degradation

Spectral signatures of halite, gypsum and carbonate crusts in the shortwave-infrared expose salt-affected soils that look unremarkable to the naked eye. Combining hyperspectral PRISMA data, Sentinel-2 SWIR bands and ALOS-2 L-band SAR dielectric returns produces salinity maps that field surveys alone cannot match at scale.

Sensors

ASI PRISMA (PRecursore IperSpettrale della Missione Applicativa): Hyperspectral imager covering 400–2500 nm in 239 contiguous bands at roughly 30 m spatial resolution. The SWIR range resolves diagnostic absorption features of halite near 2170 nm, gypsum near 1750 nm and 2220 nm, and carbonate near 2340 nm, enabling mineral-specific discrimination unavailable from broadband sensors. Revisit is approximately 29 days at the equator for a given point, improving with off-nadir tasking.
ESA Sentinel-2 MSI: Multispectral imager at 10–20 m resolution with SWIR bands at 1610 nm (B11) and 2190 nm (B12). These support salinity indices such as SI, NDSI and combinations with visible bands. Free 5-day revisit (two-satellite constellation) makes it the workhorse for bare-soil composite construction over cloud-free windows. Cannot resolve individual mineral species but performs well for broad salinity class mapping.
JAXA ALOS-2 PALSAR-2: L-band SAR at 1.27 GHz, with full-polarimetry modes at 6 m range resolution (stripmap) or 3 m (spotlight). L-band penetrates dry surface layers and responds to the dielectric constant of the soil, which rises with soluble salt and moisture content. Provides a salinity-related signal independent of surface mineralogy, useful where crusts are thin or partially obscured. All-weather, day-night acquisition removes the cloud constraint that limits optical sensors.
ESA Sentinel-1 C-band SAR: C-band at 5.405 GHz, 10 m IW mode, 6-day revisit. Shallower penetration than L-band limits its sensitivity to deep salt accumulation, but backscatter from rough salt crusts and surface dielectric contrasts still contributes useful contextual information. Freely available archive back to 2014.
USGS Landsat 8/9 OLI-TIRS: 30 m multispectral with SWIR1 (1565–1651 nm) and SWIR2 (2107–2294 nm) bands, plus thermal infrared. An 8-day combined revisit and a 40-year archive allow long-term degradation trend analysis. Useful for historical baselines where PRISMA or Sentinel-2 data do not yet exist.

What the shortwave-infrared gives away about salt

Salt-affected soils leave fingerprints in reflected light that are invisible to the human eye but obvious in the shortwave-infrared. Halite (sodium chloride) has a characteristic absorption near 2170 nm. Gypsum shows doublet features near 1750 nm and 2220 nm. Carbonate minerals absorb near 2340 nm. These features are well documented in laboratory spectroscopy libraries, and PRISMA's 239 contiguous bands from 400 to 2500 nm can resolve them directly at 30 m on the ground. Sentinel-2 cannot match that spectral resolution, but its B11 and B12 SWIR bands are broad enough to capture the integrated reflectance elevation that salt crusts produce, making index-based mapping practical at 10–20 m.

The catch is bare soil. Vegetation absorbs and scatters light in ways that swamp the subtle salt signal, so the method requires either naturally sparse cover or a bare-soil composite: a pixel-by-pixel selection of the lowest-NDVI observations from a seasonal time series, retaining only the moments when the ground is genuinely exposed. In densely cropped landscapes, those windows may be short or absent, and salinity estimates under a full canopy should be treated as unreliable regardless of which sensor is used.

L-band SAR adds a second, independent line of evidence

Radar backscatter at L-band responds to the dielectric constant of the soil rather than its mineralogy. Soluble salts in the soil solution raise the dielectric constant, which increases backscatter return in ways that can be distinguished from moisture effects when multi-temporal data are available. ALOS-2 PALSAR-2 full-polarimetry modes add decomposition parameters (particularly the volume scattering component) that help separate salt-induced dielectric changes from structural roughness caused by cracking and crust formation.

This matters operationally because cloud cover routinely breaks the optical time series in irrigated deltas during monsoon seasons. ALOS-2 acquires regardless of weather. A combined workflow that fuses PRISMA or Sentinel-2 mineralogical indices with ALOS-2 dielectric estimates produces a more defensible salinity classification than either source alone, and it allows the mapping season to extend across periods when optical data are unavailable. The honest caveat: disentangling salinity from soil moisture in SAR data is genuinely difficult, and validation against field electrical conductivity measurements remains essential before management decisions are made.

Salinity indices: what the published literature actually supports

Several spectral salinity indices have been proposed and tested in peer-reviewed literature. The Salinity Index (SI = sqrt(Green × Red)) and its variants exploit the high reflectance of salt crusts in visible bands. The Normalised Difference Salinity Index (NDSI = (Red − NIR) / (Red + NIR)) captures the contrast between salt-bright and vegetation-dark surfaces. SWIR-based ratios using Sentinel-2 B11 and B12 improve discrimination of specific crust types. Published studies in arid irrigated systems, including work in the Aral Sea basin and the Indus plains, report overall classification accuracies in the 75–90% range for broad salinity classes (non-saline, slightly saline, moderately saline, strongly saline) when validated against field measurements of electrical conductivity.

Those figures come with important context. Accuracy degrades in mixed pixels where salt patches are smaller than the sensor's ground resolution. Atmospheric correction quality is critical in SWIR bands, where small errors in water vapour correction shift reflectance values enough to misclassify soil types. And no index yet published performs well across all soil textures and parent materials without local recalibration. A map produced for one irrigated basin should not be applied uncritically to another without ground-truth adjustment.

Building a degradation trajectory, not just a snapshot

A single salinity map is useful. A degradation trajectory is what irrigation managers actually need. Landsat's archive from 1984 onwards allows decadal trend analysis using consistent SWIR bands, identifying which fields have crossed from slight to moderate salinity over a 20-year irrigation history. Sentinel-2's 5-day revisit, available since 2015, then provides the annual and seasonal resolution to track whether reclamation measures, drainage improvements or changed irrigation scheduling are having any measurable effect.

The practical workflow is a change-detection stack: annual bare-soil composites derived from the lowest-NDVI Sentinel-2 observations in each growing season, with salinity indices computed per pixel and compared year on year. Pixels that cross a defined index threshold in consecutive years are flagged as progressively degrading. This is not an alarm system with sub-week latency; the signal accumulates over seasons. But it is far cheaper and more spatially complete than the field surveys it replaces, and it produces evidence that holds up in irrigation authority reporting.

Honest limits and what they mean for buyers

Several constraints are worth stating plainly before any procurement decision. First, PRISMA's 29-day revisit and limited archive depth (the satellite launched in 2019) restrict its use for long-term trend analysis; it contributes spectral precision, not temporal density. Second, ALOS-2 commercial tasking is not free, and full-polarimetry acquisitions over large areas carry data costs that should be scoped carefully. Third, cloud cover in humid irrigated regions can reduce usable Sentinel-2 observations to a handful per year, compressing the bare-soil composite window dangerously. Fourth, the method says nothing about the root cause of salinity accumulation: whether it is rising water tables, poor drainage, or saline irrigation water. That diagnosis requires agronomic ground data alongside the spectral output.

Satellize builds salinity mapping workflows on open Sentinel-2 and Landsat archives, with PRISMA and ALOS-2 tasking added where spectral precision or all-weather acquisition justify the cost. The approach mirrors the bare-soil composite and spectral index methodology used in the Tonga crop-estimation programme, adapted for the very different problem of salt-crust mineralogy. Buyers who want a pilot over a specific irrigation scheme can request a scoped assessment against a defined area of interest and a target classification accuracy.

From map to management decision

A salinity map that sits in a GIS folder changes nothing. The output needs to connect to irrigation scheduling, drainage investment prioritisation, or crop selection decisions at the field level. The most useful deliverable is a ranked field list: fields ordered by salinity severity index, with year-on-year change rate attached, so an irrigation authority can direct leaching and drainage resources to the plots deteriorating fastest rather than spreading effort uniformly.

At national scale, the same data layer feeds food security early warning, identifying regions where irrigated productivity is structurally declining rather than suffering a single bad season. That distinction matters enormously for policy: a drought year recovers; a salinised aquifer does not.

Typical figures

Spatial resolution (optical)	10–20 m (Sentinel-2 MSI), 30 m (PRISMA hyperspectral, Landsat OLI)
Spatial resolution (SAR)	3–6 m (ALOS-2 PALSAR-2 stripmap/spotlight), 10 m (Sentinel-1 IW)
Revisit frequency	5 days (Sentinel-2 two-satellite), ~29 days (PRISMA, tasked), 14 days (ALOS-2 standard), 8 days (Landsat 8+9 combined)
Key spectral bands for salinity	SWIR 1610 nm and 2190 nm (Sentinel-2 B11/B12); 400–2500 nm continuous (PRISMA); L-band 1.27 GHz (ALOS-2)
Minimum detectable salinity class	Broad classes (non-saline to strongly saline) reliably separable; sub-class precision requires field EC validation. Patch sizes below ~0.5 ha are likely to be mixed-pixel artefacts at 30 m resolution.
Archive depth	Landsat: 1984–present. Sentinel-2: 2015–present. PRISMA: 2019–present. ALOS-2: 2014–present.
Cloud sensitivity	Optical sensors fully blocked by cloud; SAR unaffected. Bare-soil composite construction requires at least 4–6 cloud-free scenes per season.
Vegetation constraint	Method loses sensitivity at NDVI above approximately 0.3; dense canopy renders spectral salinity indices unreliable.
Typical classification accuracy (published studies)	75–90% overall accuracy for 4-class salinity scheme, validated against field electrical conductivity measurements in arid irrigated systems.
Delivery formats	GeoTIFF salinity class rasters, field-level ranked CSV, annual change-detection stack, PDF degradation report

Analytics Satellize can run

Bare-soil salinity index composite	Pixel-wise minimum-NDVI temporal compositing from Sentinel-2 time series, followed by NDSI and SWIR-ratio index computation on bare-soil observations	Annual GeoTIFF salinity index raster per season, clipped to irrigated area mask
Mineral-specific crust classification	Spectral angle mapper or partial least-squares regression applied to PRISMA hyperspectral imagery against published spectral libraries for halite, gypsum and carbonate	GeoTIFF classified crust-type map with per-class confidence layer
SAR dielectric salinity indicator	Multi-temporal ALOS-2 PALSAR-2 backscatter and polarimetric decomposition to isolate dielectric anomalies associated with elevated soil salt content	GeoTIFF dielectric anomaly layer, fused with optical salinity index for combined confidence score
Multi-year degradation trajectory	Annual salinity index stack from Landsat and Sentinel-2 archives; per-pixel linear trend and threshold-crossing detection	Field-level ranked CSV showing salinity severity score and year-on-year change rate; PDF trend report with mapped hotspots
Reclamation effectiveness monitoring	Before/after salinity index comparison over fields where drainage or leaching interventions have been recorded, using paired seasonal composites	GIS polygon layer with intervention-response attribution and statistical significance flags
Irrigated-area salinity risk screening	Salinity class map intersected with crop-type and irrigation-scheme boundaries to compute area-weighted risk scores per management unit	Tabular risk summary per irrigation block, suitable for authority reporting and investment prioritisation

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.