Mapping Soil Properties from Satellite Imagery: What's Possible and What's Not

A soil scientist colleague once joked that remote sensing of soils should be called "remote sensing of soil when there's no vegetation in the way." She wasn't wrong. The fundamental limitation of optical soil mapping from satellites is that you can only see the soil when nothing is covering it.

But when soil is exposed — after harvest, before planting, in fallow periods, in arid regions — satellite data reveals more about soil composition than most people expect.

What Soil Properties Are Visible from Space

Soil Organic Carbon (SOC)

Organic matter darkens soil. High-SOC soils absorb more light across visible and near-infrared wavelengths, appearing darker in satellite imagery. The relationship is strongest in the visible range (400-700 nm).

The approach: On bare soil images, darker pixels generally indicate higher organic carbon. More quantitatively, regression models using visible and NIR reflectance can estimate SOC with R² values of 0.5-0.7 at regional scales.

Limitations: The relationship breaks down when other factors also darken the soil — moisture (wet soil is darker regardless of SOC), iron oxides (dark minerals), and surface roughness all confound the SOC signal.

Iron Oxides

Iron minerals (hematite, goethite, ferrihydrite) produce diagnostic spectral features:

• Hematite: Strong absorption in blue (450 nm), giving soil a red color
• Goethite: Absorption in blue-green (480 nm), producing yellow-brown color
• General iron: Broad absorption feature around 900 nm (NIR)

Sentinel-2's blue and red bands, combined with NIR, enable basic iron oxide mapping. The ratio B4/B2 (Red/Blue) is a simple iron oxide index — high values indicate iron-rich soils.

Clay Minerals

Clay minerals (kaolinite, illite, montmorillonite) have absorption features in the SWIR region, particularly around 2200 nm. Sentinel-2's B12 (2190 nm) captures the edge of this absorption.

The clay index B11/B12 provides a coarse indicator of clay content, though the spectral resolution of Sentinel-2 is too broad for reliable clay mineral identification. Hyperspectral sensors (like EnMAP or PRISMA) with numerous narrow SWIR bands perform much better for clay mineralogy.

Soil Moisture

As discussed in the SAR soil moisture article, water dramatically changes soil's spectral and dielectric properties. In optical imagery, wet soil appears darker than dry soil across all bands. In SAR, wet soil returns stronger backscatter.

Calcium Carbonate

Carbonate-rich soils (calcareous soils) are typically light-colored with high reflectance across the visible spectrum. They show subtle absorption features around 2340 nm, just beyond Sentinel-2's spectral range but detectable by hyperspectral sensors.

Surface Texture (Roughness)

While not a compositional property, soil surface roughness affects how light is scattered and can be inferred from SAR backscatter. Freshly plowed soil is rough; crusted soil is smooth. This matters for erosion risk assessment and seedbed quality evaluation.

Bare Soil Composites

The key to soil mapping from satellites is obtaining cloud-free, vegetation-free imagery. The Bare Soil Composite approach:

1. Collect all available Sentinel-2 images over several years (3-5 years)
2. Apply cloud masking (SCL-based)
3. Apply vegetation masking: Exclude pixels with NDVI > 0.25 (vegetation present)
4. Composite the bare soil observations: For each pixel, select the observation(s) with the lowest NDVI (most bare soil)

The result is a synthetic "bare soil image" where every pixel shows the soil surface, even if it's only bare for a few weeks per year. This composite maximizes spatial coverage and provides the cleanest soil spectral signal.

This technique works particularly well in agricultural landscapes where fields are periodically bare between crops. In permanently vegetated areas (forests, permanent grassland), bare soil composites can't help.

Digital Soil Mapping with Satellite Data

Modern digital soil mapping integrates satellite-derived soil indices with:

• Terrain variables: Elevation, slope, curvature, wetness index (from DEMs)
• Climate data: Temperature, precipitation, evapotranspiration
• Geological maps: Parent material information
• Existing soil samples: Sparse field measurements as calibration/validation

Machine learning models (Random Forest, gradient boosting) trained on these combined datasets produce continuous soil property maps at 10-30 m resolution. The satellite data provides the spatial detail that sparse soil samples alone cannot.

Accuracy Expectations

Be realistic about what satellite-based soil mapping can achieve:

Property	Satellite Approach	Typical R²	Notes
Organic carbon	Visible/NIR darkness	0.5-0.7	Best in homogeneous landscapes
Iron oxide content	Red/Blue ratio	0.6-0.8	Strong spectral signal
Clay content	SWIR ratios	0.4-0.6	Sentinel-2 too broad; hyperspectral better
Soil moisture	SAR backscatter	0.5-0.7	Bare soil only at C-band
Texture class	Combined indices	0.3-0.5	Very approximate
pH	Indirect (via SOC, carbonate)	0.3-0.5	Weak satellite signal

These accuracies are sufficient for:

• Identifying spatial patterns and variability zones
• Guiding targeted soil sampling (sample where the satellite indicates variability)
• Monitoring temporal changes (SOC trends, erosion)

They're insufficient for:

• Replacing soil laboratory analysis
• Meeting regulatory requirements for soil classification
• Making precise fertilizer recommendations based on nutrient content

Practical Applications

Erosion Risk Mapping

Combining bare soil composite brightness (indicating SOC/texture) with terrain slope identifies erosion-prone areas. Light-colored soil on steep slopes is at highest risk — likely low organic matter, poor structure, and gravitational stress.

Carbon Stock Estimation

Climate change mitigation requires large-scale SOC monitoring. Satellite-derived SOC maps, calibrated with field measurements, provide estimates of soil carbon stocks across agricultural regions. Temporal monitoring (comparing bare soil composites from different years) can track SOC changes, though detecting small changes requires many years of data.

Precision Soil Sampling

Instead of sampling on a regular grid, use satellite soil variability maps to place samples strategically — more samples where the satellite shows high variability, fewer where conditions appear uniform.

Geological/Mineral Exploration

In arid regions where soil is permanently exposed, satellite spectral analysis aids mineral exploration by mapping surface mineralogy across large areas. Iron oxide and clay alteration patterns visible in Sentinel-2 SWIR data can indicate subsurface mineralization.

The bottom line: satellite soil mapping is a powerful reconnaissance tool that reveals spatial patterns invisible from field-level observation. It doesn't replace soil sampling — it makes sampling smarter, more targeted, and more cost-effective. And in an era where we need to monitor soil health across entire continents, satellite data is the only practical way to achieve that spatial coverage.