Soil Reflectance Composites—Improved Thresholding and Performance Evaluation

Summary

This paper presents HISET, a novel data-driven methodology for automatically deriving spectral index thresholds from multispectral satellite imagery to identify bare soil pixels in temporal composites without manual intervention or site-specific calibration. The authors validated the approach using Sentinel-2 data and three different spectral indices, demonstrating that NBR2-based composites achieved the highest spectral fidelity to reference soil spectra and broadest spatial coverage. The work addresses a key operational constraint in large-scale soil monitoring, where existing threshold-derivation approaches rely on expert knowledge or labour-intensive manual procedures unsuitable for continental or global processing workflows.

UK applicability

The HISET methodology is directly applicable to UK soil monitoring programmes and could support improved mapping of soil organic carbon across arable and grassland areas using freely available Sentinel-2 data. UK agricultural and environmental agencies could integrate this approach into national soil health monitoring systems, though validation against UK-specific land cover types and climatic conditions would be advisable.

Key measures

Spectral similarity to LUCAS reference spectra; spatial coverage of bare soil pixels; number of valid observations per pixel; accuracy metrics for bare soil classification; performance comparison across three spectral index approaches

Outcomes reported

The study developed and validated the HISET (HIstogram SEparation Threshold) methodology for automatically deriving spectral index thresholds to identify bare soil pixels in Sentinel-2 satellite imagery. The methodology was tested on six soil reflectance composites using different spectral indices (NDVI, NBR2, and PV+IR2) to assess spatial coverage, spectral accuracy and performance for large-scale soil constituent modelling.

Topic tags