EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Interpretable Additive Modeling of Soil Organic Carbon and Its Fractions From Mid‐Infrared Spectra Using MCR ‐ ALS

Mikhail Borisover, David Yalin, William Mlelwa, Nadezhda Bukhanovsky, Dan Yakir, José M. Grünzweig

European Journal of Soil Science · 2026-09 · 77 (5)

DOI: 10.1111/ejss.70435 ↗

Kurz gefasst

Fachartikel in der Zeitschrift European Journal of Soil Science, online erschienen am 12.10.2026. Verfasst von Mikhail Borisover und 5 weiteren Autor:innen. Die Arbeit nennt 79 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Jewish National Fund.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Modeling soil organic carbon (OC) contents from mid‐infrared (MIR) spectra remains challenging, and conventional multivariate spectroscopic models often lack chemical interpretability. The objective of this study was to examine whether an additive model based on non‐negative multivariate curve resolution—alternating least squares (MCR‐ALS) could reproduce variations in OC contents of whole soils and size‐separated fractions while providing a chemically interpretable description of that variation. The approach was implemented on MIR spectra of forest and woodland soils in Israel, including fine (< 50 μm) and coarse (> 50 μm) fractions representing mineral‐associated organic matter (MAOM) and particulate organic matter (POM), respectively. The OC content was represented as a sum of contributions from nonnegative MCR‐ALS spectral components, and the effects of soil fractionation and spectral range selection (4000–400 cm −1 versus optimized subranges) were evaluated. Using the full MIR spectral range, the additive model reproduced OC contents more effectively for the POM‐associated fraction than for the whole soils, explaining 71.7% and 38.2% of the observed variance, respectively, whereas no improvement was obtained for the MAOM‐associated fraction (16.1%). Optimization of the spectral range increased the explained variance to 69.1% for the whole soils and 56.1% for the MAOM‐associated fraction. Leave‐one‐out cross‐validation indicated the predictive capability of the optimized models ( Q 2 = 0.625, 0.526 and 0.512 for the POM‐associated fraction, whole soils and the MAOM‐associated fraction, respectively), while bootstrap analysis supported the stability of several optimized parameter estimates. These results demonstrate the potential of spectral range optimization for effective OC modeling using nonnegative MCR‐ALS. Analysis of model performance versus the number of MCR components revealed transition‐like behavior, supporting the identification of chemically meaningful components. Overall, this study establishes a physically grounded and interpretable framework for MCR‐ALS‐based modeling of soil OC from MIR spectra, highlighting the roles of soil fractionation, optimization of continuous spectral windows, and interpretability of spectral components.

Förderung

  • Jewish National Fund

Zitieren

APA

Borisover, M., Yalin, D., Mlelwa, W., Bukhanovsky, N., Yakir, D., & Grünzweig, J. M. (2026). Interpretable Additive Modeling of Soil Organic Carbon and Its Fractions From Mid‐Infrared Spectra Using MCR ‐ ALS. European Journal of Soil Science, 77(5). https://doi.org/10.1111/ejss.70435

Vancouver

Borisover M, Yalin D, Mlelwa W, Bukhanovsky N, Yakir D, Grünzweig JM. Interpretable Additive Modeling of Soil Organic Carbon and Its Fractions From Mid‐Infrared Spectra Using MCR ‐ ALS. European Journal of Soil Science. 2026;77(5). doi:10.1111/ejss.70435

BibTeX
@article{Borisover2026,
  author = {Borisover, Mikhail and Yalin, David and Mlelwa, William and Bukhanovsky, Nadezhda and Yakir, Dan and Grünzweig, José M.},
  title = {Interpretable Additive Modeling of Soil Organic Carbon and Its Fractions From Mid‐Infrared Spectra Using MCR ‐ ALS},
  journal = {European Journal of Soil Science},
  year = {2026},
  volume = {77},
  number = {5},
  doi = {10.1111/ejss.70435}
}
Kostenlos starten →