Leveraging soil spectral libraries for property prediction: global versus local mid-infrared models and the determinants of predictability

Malithi Weerasekara

doi:10.1016/j.geoderma.2026.117950

Leveraging soil spectral libraries for property prediction: global versus local mid-infrared models and the determinants of predictability

2026-07-22

0

OA

AI

M

Malithi Weerasekara

A

Alfred E. Hartemink

J

Jonathan J. Maynard

E

Elliott Demain

Y

Yakun Zhang *

DOI：10.1016/j.geoderma.2026.117950

Original

Original request for help

Save

Abstract

En 中文

• Global and local models were compared for predicting soil property from MIR spectra. • Global nonlinear models achieved the highest accuracy and transferability. • Stratification mainly benefited linear models like PLSR. • Model choice reflects interactions of algorithm, sample size, and data distribution. • A practical framework for selecting calibration scope and model class was proposed.

Keywords:

Soil spectral library

Soil property prediction

Calibration strategies

Linear and nonlinear models

Model transferability

Mid-infrared spectroscopy

Leveraging soil spectral libraries for property prediction: global versus local mid-infrared models and the determinants of predictability

Abstract

Journal

Organization

Cited Papers

Citing Papers