Estimation Accuracies of near Infrared Spectroscopy for General Soil Properties and Enzyme Activities for Two Forest Sites along Three Transects

Abstract

Visible and near infrared spectroscopy (vis-NIRS) is an established method for estimating the contents of soil organic carbon (SOC) and soil nitrogen (N). Recent studies have suggested that it may also be useful for estimating enzyme activities, however this potential has not been explored in detail. Objectives were to determine estimation accuracies of vis-NIRS for general soil properties (SOC, N, pH, texture) and nine enzyme activities for two forest sites, one on a sandy soil (Grinderwald) and the other on a loess soil (Rildershausen). For each site, a calibration sample consisting of two transects sampled down to 185 cm (Grinderwald: 128 units, Rildershausen: 64 units) and an independent validation sample consisting of one transect (Grinderwald: 64 units, Rildershausen: 32 units) was obtained and their absorbance spectra recorded. Chemometric approaches included the standard partial least squares (PLS) regression and PLS regression with a genetic algorithm (GA-PLS) for variable selection, which may improve estimation accuracies. This study confirmed the marked usefulness of vis-NIRS for an estimation of SOC and N contents in independent transects of a field scale. Estimation accuracy of soil pH and texture in independent transects was variable and mainly dependent on the range of measured data. GA-PLS markedly, improved estimation accuracies compared to PLS in the cross-validation, but generally not in the validation transects. Few enzyme activities could be estimated in independent validation, but there was almost no additional benefit of vis-NIRS for their estimations compared to estimations using measured contents of main properties; however, these contents of main properties may also be estimated with vis-NIRS. Overall, for the studied sites, we do not see a benefit of vis-NIRS for a direct estimation of enzyme activities compared to laboratory methods, most likely due to a lack of sufficiently strong specific impacts on the measured spectral signals. (C) 2016 Elsevier B.V. All rights reserved.

Publication
GEODERMA