Spectroscopy-Based Models Outperform Pedotransfer Functions for Estimating Soil Hydraulic Properties

ORCID

Gamagedara: https://orcid.org/0000-0002-6826-8420; Feng: https://orcid.org/0000-0001-7783-1644; Tagert: https://orcid.org/0000-0002-0745-2099; Martins: https://orcid.org/0000-0003-3802-0368; Wijewardane: https://orcid.org/0000-0001-8962-9451

MSU Affiliation

College of Agriculture and Life Sciences; Department of Agricultural and Biological Engineering; James Worth Bagley College of Engineering

Creation Date

2026-06-30

Abstract

Soil-water relationships are critical for water retention and movement in soil, influencing agriculture, environmental sustainability, and irrigation planning. Direct measurement of soil hydraulic properties (SHPs) is labor-intensive, costly, and time-consuming. Pedotransfer functions (PTFs) like Rosetta 3 predict SHPs using readily available soil properties, while spectroscopy offers enhanced sensitivity to soil chemical and physical characteristics, although its application for SHP estimation remains underexplored. This study developed robust spectroscopy-based models using visible & near-infrared (vis-NIR) and mid-infrared (MIR) spectral regions from 335 soil samples (0–5 and 5–10 cm depths) collected in Texas and Mississippi to estimate Mualem-van Genuchten soil hydraulic parameters and derived properties. Both PTFs from Rosetta 3 and spectroscopic-based models were evaluated against the centrifuge laboratory reference method. Overall, spectroscopy-based models outperformed Rosetta 3, particularly for predicting field capacity (FC) and permanent wilting point (PWP). The MIR region showed superior performance, with R2 values of 0.80 for both FC and PWP, compared with the vis-NIR region (R2 = 0.65 for FC and 0.70 for PWP). Predictions of PWP were more consistent across soil scanning conditions than those of FC, with MIR spectra exhibiting greater stability. Among modeling approaches, partial least squares regression and ridge regression generally outperformed tree-based models, including random forest and categorical boosting regression. Fine-ground and non-fine-ground soil samples yielded better predictions than fresh soils, emphasizing the importance of sample preparation. These findings highlighted that MIR spectroscopy, combined with appropriate modeling and sample preparation, provides an accurate, scalable, rapid, and cost-effective approach for large-scale estimation of SHPs.

Publication Date

4-18-2026

Publication Title

CATENA

Publisher

Elsevier

Rights

© 2026 The Authors

Link to Full Text

Share

COinS
 

Digital Object Identifier (DOI)

https://doi.org/10.1016/j.catena.2026.110099