Bartsch et al. (2025) Soil laboratory and satellite spectral data filtering: A Spectral Quality Protocol (SQuaP)

Identification

• Journal: Remote Sensing of Environment
• Year: 2025
• Date: 2025-11-20
• Authors: Bruno dos Anjos Bartsch, José A.M. Demattê, Nikolaos Tziolas, Jorge Tadeu Fim Rosas, Gabriel Pimenta Barbosa de Sousa, Nicolas Augusto Rosin, Giannis Gallios, Borges Marfrann Dias Melo
• DOI: 10.1016/j.rse.2025.115144

Research Groups

• Department of Soil Science, Luiz de Queiroz College of Agriculture, University of S˜ao Paulo, Piracicaba, S˜ao Paulo, Brazil
• Southwest Florida Research and Education Center, Department of Soil, Water and Ecosystem Sciences, Institute of Food and Agricultural Sciences, University of Florida, Immokalee, FL 34142, USA
• Faculty of Agronomy and Veterinary Medicine, University of Brasília, 70910-970, Brasília, DF, Brazil

Short Summary

This study introduces the Spectral Quality Protocol (SQuaP) to filter laboratory and satellite soil spectral data, significantly improving the reliability and predictive performance of soil property models for clay and soil organic carbon.

Objective

• To develop and validate the Spectral Quality Protocol (SQuaP) for filtering soil spectral datasets from laboratory and satellite sources to enhance data quality and improve the predictive performance of soil property models.

Study Configuration

• Spatial Scale: Regional (Brazil, with 9261 soil samples) to Global (tested on worldwide open datasets and bare soil satellite observations).
• Temporal Scale: Not explicitly defined for data collection, but the protocol is designed for broad applicability across various datasets.

Methodology and Data

• Models used: Rule-based checks (crop residues, reflectance trend), PCA-Mahalanobis, soil line regression, Isolation Forest, Random Forest, Kernel Density Estimation (KDE).
• Data sources:
  • Laboratory hyperspectral data (350–2500 nm) from 9261 Brazilian soil samples.
  • Sentinel-2 multispectral data paired with the Brazilian samples.
  • Worldwide open datasets.
  • Bare soil satellite observations identified by the Geospatial Soil Sensing System (GEOS3).

Main Results

• SQuaP significantly enhanced spectral data reliability and modeling performance for clay and soil organic carbon (SOC).
• For hyperspectral data, SQuaP increased R² by 17.55 % for clay and 1.58 % for SOC, and reduced RMSE by 10.41 g kg⁻¹ for clay and 2.06 g kg⁻¹ for SOC.
• For satellite data, improvements were more pronounced, with an increase of 15.76 % for clay and 13.38 % for SOC on R², alongside a reduction in RMSE.
• At the worldwide scale, predictive performance improved after SQuaP implementation in nearly all continents, with R² gains ranging from -2.15 % to 13.40 % across evaluated scenarios.
• The protocol successfully removed outliers while preserving data variability, as confirmed by Kernel Density Estimation (KDE) analysis.

Contributions

• Introduction of SQuaP, a novel and comprehensive filtering protocol for enhancing the quality of both laboratory and satellite soil spectral datasets.
• Demonstrated significant and quantifiable improvements in the reliability and predictive accuracy of soil property models (clay, SOC) across different spectral data types and spatial scales.
• Provides a flexible and adaptable protocol that can be tailored to specific datasets and applications, thereby advancing digital soil mapping and precision agriculture.
• Ensures the statistical integrity of datasets by effectively removing outliers without compromising data variability.

Funding

• Not explicitly mentioned in the provided text.

Citation

@article{Bartsch2025Soil,
  author = {Bartsch, Bruno dos Anjos and Demattê, José A.M. and Tziolas, Nikolaos and Rosas, Jorge Tadeu Fim and Sousa, Gabriel Pimenta Barbosa de and Rosin, Nicolas Augusto and Gallios, Giannis and Melo, Borges Marfrann Dias},
  title = {Soil laboratory and satellite spectral data filtering: A Spectral Quality Protocol (SQuaP)},
  journal = {Remote Sensing of Environment},
  year = {2025},
  doi = {10.1016/j.rse.2025.115144},
  url = {https://doi.org/10.1016/j.rse.2025.115144}
}