From the review above, it can be summarised that NIRS techniques can be used to quantify soil constituents that are “near-infrared active” constituents (covalent bonds of small atoms, particularly, O, H, C and N), and other soil constituents which are featureless in the near-infrared but well correlated to one or more of the “near-infrared active” soil components. Field root density measurements using NIRS have not been reported, although roots are “near-infrared active”. The only reported laboratory root density measurements using NIRS were carried out on peat samples to quantify
Ericales roots for palaeoecological studies (McTiernan et al. 1998). Field measurements have not been made of root density in pastoral and arable soils using NIRS. Root density measurement using conventional techniques (e.g. wet sieving prior to counting or weighing)is tedious, and requires separating the roots from the soil.
Related to the identified gaps in the use of NIRS, the first study in this thesis aims to develop a method for the non-destructive measurement of root density in cores of
intact soil using Vis-NIRS. Root density measurements of ryegrass grown in a pot trial are reported, assessing the ability of Vis-NIRS in predicting root density. For this study, development of a purpose-built soil probe suitable for close distance spectral acquisition was needed.
Assessment of root density prediction using Vis-NIRS from a pot trial is followed by root density prediction in the field, in permanent pasture. Because many soil chromophores e.g. water, iron oxides, decomposed organic matter and soil aggregates may influence soil reflectance of field soils, the ability of Vis-NIRS to predict root content from spectral reflectance needs to be examined. The utility of calibration models developed from glasshouse data to predict field root density is also a focus of this study. In addition, as soil organic C is partly formed from decomposed plant root, whether Vis-NIRS predicts roots independently from soil organic matter needs to be evaluated, in order to identify potential C sequestration from root production. This study required a purpose-built soil probe that allowed reflectance spectra to be acquired from a soil corer.
Deep rooting plants are useful in e.g. extracting subsoil water and nitrate leached past shallower roots, and in sequestrating carbon deeper down in the soil which will prolong its residence time. As part of studies of such deep rooting plants, field assessment of root density using Vis-NIRS was extended to arable land and maize root density assessment down to 600 mm depth. This also involved the use of a soil probe attached to a soil corer.
There is no published literature on the use of NIRS for in situ field measurements of C and N in New Zealand soils. Hill Laboratories in New Zealand has used NIRS techniques, but only using dried ground soil samples in the laboratory, although the results have not yet been published. Moreover, there are limited numbers of publications assessing the ability of NIRS techniques in real-field soil conditions where there are variations in other soil chromophores e.g. water, clay and non-clay mineral types, iron oxides and parent materials. To address these concerns, the next objective of this thesis was to assess in situ “Vis-NIRS-soil core” technique in quantifying soil C and N from various New Zealand soils e.g. Allophanic, Pumice and Tephric and Recent soils.
Another area that needs further attention is the basis for selection of the spectral data samples (out of a larger population of spectra data) in which the reference property will be analysed for use in the chemometric calibration procedure. There are limited numbers of publications using the spectral data as a basis for this selection of the calibration and validation sets. This is an important piece of work because if accurate predictions of soil properties are to be achieved from spectral reflectance, the chemometric calibration may have to be restricted to unknown sample with similar spectral attributes. So the influence of sample selection techniques for the calibration and validation sets on the NIRS prediction accuracy is another focus of this study.
Even though excellent prediction of soil C and N using NIRS can be achieved (Chang and Laird 2002; Moron and Cozzolino 2004), this was carried out using dried ground sieved soil samples. Excellent accuracy for soil C and N measurement in the field has not been reported, but may be attempted by developing appropriate field techniques. This study is focused on developing and testing such spectral acquisition techniques. Spectral reflectance acquired from flat horizontal cross-sections of soil cores are compared with spectra collected from the outer surface of the curved, vertical wall of cylindrical soil cores. Being able to predict soil C concentrations to depth will be important because the IPCC (Intergovernmental Panel on Climate Change) protocol for soil C accounting recommends C measurement to 300 mm depth. So, accuracy of the two techniques in predicting soil C and N will be assessed from soil samples collected from 15 mm to 315 mm depth.
It has been explained in the Sections 2.2.3.1 and 2.2.3.6 that water and iron oxides considerably affect soil spectral reflectance (Baumgardner et al. 1985; Ben-Dor 2002). Seasonal fluctuation of soil moisture deficit together with fluctuation of water table, reduction/oxidation reactions and the iron oxides formed also influence the spectral reflectance which may in turn influence the accuracy of field Vis-NIRS measurement. The objective of the last study reported in this thesis was to assess the temporal robustness of calibration models used for predicting soil C and N. The robustness of the model was evaluated using data collected under different field moisture conditions (in dry and slightly wet months). The best spectral acquisition technique from the earlier studies was used to collect spectral reflectance from 15 to 315 mm soil depth.