Journal article
Usefulness of near-infrared spectroscopy to determine biological and chemical soil properties: Importance of sample pre-treatment



Publication Details
Authors:
Terhoeven-Urselmans, T.; Schmidt, H.; Jörgensen, R.; Ludwig, B.
Publication year:
2008
Journal:
Soil Biology and Biochemistry
Pages range:
1178-1188
Volume number:
40
Issue number:
5
Start page:
1178
End page:
1188
ISSN:
0038-0717

Abstract
Near-infrared reflectance spectroscopy (NIRS) is known for its inexpensiveness, rapidity and accuracy and may become a useful tool for the assessment of soil quality. Objectives were (i) to evaluate the ability of NIRS to predict several chemical and biological properties of organically managed arable soils as well as the properties of grain yield from winter cereals for a closed population and (ii) to test whether the use of field-moist and pre-treated (quick-freezing followed by freeze-drying and grinding) samples will generate similar results. One hundred and sixteen soil samples from nine organically managed farms from Germany sampled in 2005 and 2006 were used for this investigation. Spectra of the near-infrared region (including the visible range, 400-2500 nm) from field-moist (<2 mm) or pretreated soil samples were recorded. A modified partial least-square regression method and cross-validation were used to develop an equation over the whole spectrum (first-third derivation). For the pre-treated soils, good predictions were obtained for pH, contents of organic C, total N, plant-available P (Olsen) and exchangeable K (calcium-acetate-lactate (CAL)), contents of microbial biomass C and N (C-mic and N-mic) and ergosterol, basal respiration, metabolic quotient, the ratio of organic C/total N, the grain yield of winter cereals and grain nitrogen uptake. The RSC (the ratio of standard deviation of laboratory results to standard error of cross-validation) was greater than 2.0, the correlation coefficients (r) of a linear regression (measured against predicted values) were greater than or equal to 0.9 and the regression coefficients (a) ranged from 0.9 to 1.1. Similar good predictions were obtained if field-moist samples were used, with the exception of P (Olsen), K (CAL), metabolic quotient, grain yield of winter cereals and grain nitrogen uptake (satisfactory predictions) and ergosterol content (unsatisfactory prediction). Good predictions of the contents of Mg (CaCl2) and microbial biomass P (P-mic) were achieved for field-moist but not for pre-treated samples. Despite sample preparation, only satisfactory predictions were obtained for the ratios of C-mic/N-mic and ergosterol/C-mic and grain nitrogen content (1.4 <= RSC <= 2.0, r >= 0.8 and 0.8 <= a <= 1.2). However, unsatisfactory predictions for field-moist and pre-treated samples were achieved for the content of P (CAL), the nitrogen mineralisation rate and the ratios of C-mic/P-mic and basal respiration/nitrogen mineralisation rate. Our results demonstrate that biological soil properties can be predicted with NIRS for closed populations in both sample states. The pre-treatment should be used if samples have to be stored prior to infrared measurements for periods longer than a month. (C) 2007 Elsevier Ltd. All rights reserved.

Last updated on 2018-22-11 at 14:25

Share link