THE POTENTIAL OF NIRS AS A TOOL FOR SPATIAL MAPPING OF SOIL COMPOSITIONS

Authors: Reeves Iii, James; McCarty, Gregory

Submitted to: Near Infrared Reflectance International Conference Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 6/17/1999
Publication Date: N/A
Citation: N/A

Interpretive Summary: Near-infrared reflectance spectroscopy (NIRS) is a technique using light which can be used to determine the composition of a wide variety of materials. The objective of this work was to investigate the potential of NIRS under conditions where variations in the inorganic constituents of the soil were present. Three sets of samples were investigated: First, a set of 64 samples randomly collected from a 20 ha agricultural field which was part of a small watershed and contained a wide variety of soils. Second, a set of 547 samples (one sample lost) collected in a grid pattern from the soil surface (0 - 10 cm) and sub-surface (10 - 30 cm) at 274 locations from the same 20 ha field. Third, a set of 179 samples collected for a previous study from two different sites, but with less diverse soils, for which NIRS had been shown to work very well. Results demonstrated that accurate determinations of carbon and to a lesser extent nitrogen can be developed using near-infrared spectra for diverse sets of soil samples. However, these preliminary results indicate that calibrations developed using less diverse sets of soil samples may be more accurate. Results also showed that NIRS worked considerably better when applied to ground soil samples. In conclusion, while further work is needed, results demonstrated that NIRS can be extremely useful in the analysis of the large sample sets required for determining soil composition over large areas.

Technical Abstract: The objective of this work was to investigate the potential of near-infrared reflectance spectroscopy (NIRS) under conditions where variations in the inorganic constituents of the soil were present. Three sets of samples were investigated: First, a set of 64 samples were randomly collected from a 20 ha agricultural field which comprised a substantial part of a toposequence within a small watershed. Second, a set of 547 samples (one sample lost) collected in a grid pattern from the soil surface (0 - 10 cm) and sub-surface (10 - 30 cm) at 274 locations from the same 20 ha field. Third, a set of 179 samples collected for a previous study from two different sites for which NIRS had been shown to work very well. All samples were analyzed by combustion for C and N, and scanned from 400 to 2498 nm using a scanning monochromator and rotating sample cup. Results of PLS calibrations demonstrated that accurate calibrations for C and to a lesser extent N can be developed using near-infrared spectra for diverse sets of soil samples. However, these preliminary results indicate that calibrations developed using less diverse sets of soil samples may be more accurate. Results also showed that, at least when scanned using a rotating sample, near-infrared calibrations worked considerably better when applied to ground soil samples. In conclusion, while further work is needed, results demonstrated that NIRS can be extremely useful in the analysis of the large sample sets required for spatial mapping studies.