Representative elementary area issue in soil spectral measurements

Abstract

Soil spectral characteristics from laboratory and in situ measurements are an effective source of information about various soil properties. However, features of scanned surfaces are unified by the drying and grounding of soil samples solely for laboratory measurements. Spectra scanned in situ are perturbed to varying degrees due to spatial variation of the soil’s moisture and roughness that results for the most part from cultivation practices and intrinsic soil aggregation. Even intensive data pre-processing and advanced modeling techniques cannot fully eliminate these signal modifications even though field spectra can be interpreted if and when the representativeness criterion is reached. This work is aimed at examining the effects of soil surface roughness on the size of viewing areas that fulfill the representative elementary area (REA) criterion for a stable spectra acquisition. Surface horizon’s samples with natural soil aggregates and undisturbed clods from five different soils were prepared on trays into three different soil surface roughness levels. Spectral measurements were carried out under laboratory conditions while incrementally increasing sensor height above the reference level at nadir, orthogonal to the target surface, and illuminated along the one plane at three zenithal angles (θs). Roughness of surfaces was measured using a laser scanner and described in terms of height standard deviation (HSD) of the soil surface area within its delineated basic digital elevation model (DEM) unit. To express equally the size and roughness of the viewed area, a single qualifier DRR (diameter to roughness ratio) was proposed. This is the ratio of the diameter of field of view (DFOV) to the surface roughness index HSD, and can be interpreted as numbers of structural elements placed along DFOV, where higher values indicate higher representativeness of the viewed area. In general, a stable spectrum is scanned at DRR starting from 60, and this criterion should be complied with at a minimum during proximal spectral measurements of soils. Furthermore, DRR above 60 is the only minimum criterion. Results of the work suggest the stabilization of spectral measurement at DRR close to 60, as well as the elimination of noise error at DRR close to 200.