EUVIMEDScopri EUVIMED
Crossref

Predicting Soil Health Indicators in Florida Sandy Soils Using Diffuse Reflectance Mid‐Infrared Fourier Transform Spectroscopy

Franky Celestin, Leonardo Deiss, Ryan E. Champiny, Gabriel Maltais‐Landry, Jose C. B. Dubeux, Rao Mylavarapu, Yang Lin

Soil Use and Management · 2026-01 · 42 (1)

DOI: 10.1111/sum.70184 ↗

At a glance

Journal article in Soil Use and Management, published online on 2026-02-26. Written by Franky Celestin and 6 co-authors. Cited 1 times so far (Crossref); 71 references.

AbstractDati bibliograficiCiteRelated studies

Abstract (original language)

Measuring soil health indicators at scale remains a major challenge to implementing soil health management, given the high cost and labour intensity of traditional laboratory analysis. Mid‐infrared diffuse reflectance spectroscopy (mid‐DRIFTS) offers a promising alternative to conventional wet chemistry by enabling rapid, high‐throughput prediction of key soil indicators, while capturing mechanistic links between soil biochemical and mineral composition with soil health properties. This study evaluated the predictive and interpretive potential of mid‐DRIFTS for soil health assessment. We compared the predictive performance of three modelling approaches: partial least squares (PLS), support vector machines (SVM) and neural networks (NN) and tested the added benefit of spectral partitioning, a novel approach in soil Fourier transform infrared (FTIR) spectroscopy. Results demonstrated that spectral partitioning consistently improved model performance across indicators, with root mean squared error (RMSE) reduction of up to 37%, and it was especially beneficial for non‐linear machine learning models (SVM and NN). Indicators including soil organic matter (SOM), permanganate oxidizable carbon (POX‐C) and autoclaved citrate‐extractable (ACE) protein were predicted with high accuracy, while indicators such as mineralizable carbon (Cmin), water‐extractable nitrogen (WEN) and extractable nutrients showed moderate to poor predictive performance. These findings suggest that integrating spectral partitioning into modelling routines is a viable approach for accurately predicting biochemical indicators with strong spectral signals. Furthermore, we examined correlations between mineral and organic functional groups estimated as peak areas in mid‐DRIFTS and soil health indicators. POX‐C and ACE protein were positively associated with aliphatic C–H groups and negatively related to aromatic and clay‐derived mineral functional groups, highlighting the role of biochemical composition and mineral–organic interactions in regulating dynamic soil health properties. Overall, mid‐DRIFTS approaches enable estimation of key soil health indicators while revealing mechanistic links between soil composition and soil health.

Cite

APA

Celestin, F., Deiss, L., Champiny, R. E., Maltais‐Landry, G., Dubeux, J. C. B., Mylavarapu, R., & Lin, Y. (2026). Predicting Soil Health Indicators in Florida Sandy Soils Using Diffuse Reflectance Mid‐Infrared Fourier Transform Spectroscopy. Soil Use and Management, 42(1). https://doi.org/10.1111/sum.70184

Vancouver

Celestin F, Deiss L, Champiny RE, Maltais‐Landry G, Dubeux JCB, Mylavarapu R, et al. Predicting Soil Health Indicators in Florida Sandy Soils Using Diffuse Reflectance Mid‐Infrared Fourier Transform Spectroscopy. Soil Use and Management. 2026;42(1). doi:10.1111/sum.70184

BibTeX
@article{Celestin2026,
  author = {Celestin, Franky and Deiss, Leonardo and Champiny, Ryan E. and Maltais‐Landry, Gabriel and Dubeux, Jose C. B. and Mylavarapu, Rao and Lin, Yang},
  title = {Predicting Soil Health Indicators in Florida Sandy Soils Using Diffuse Reflectance Mid‐Infrared Fourier Transform Spectroscopy},
  journal = {Soil Use and Management},
  year = {2026},
  volume = {42},
  number = {1},
  doi = {10.1111/sum.70184}
}
Start for free →