Quantifying short-range variation of soil texture and total carbon of a 330-ha farm

Abstract

We quantified the short-range variation (<5 m) of soil texture and total carbon at a 330-ha arable farm in Wisconsin, USA. The farm is located in a young landscape (<12 ky), and geomorphic units include unglaciated parts, outwash plain, end moraine, and a former lacustrine bed. At 100 locations, three soil cores, in a 90-cm triangle configuration, were collected to 90-cm depth. The soil cores were subsampled at 10-cm depth interval (total 2,241 samples). Clay, sand, silt, and total carbon were predicted for all the 2,241 samples using cubist spectral models developed on 25 locations (197 samples) with the vis-NIR and PXRF spectra. The coefficient of variation (CV) of each soil property was calculated for the three cores at each location and at each depth increment. The soil short-range variation was quantified for soils of the four geomorphic units. In addition, within-horizon variation (CVs) of soil properties was calculated. It was found that parent materials affected the short-range variation more than land use. Soils developed from lacustrine deposits had higher sand content and a slightly larger short-range variation for clay and silt content. Soils developed in the outwash plain had large range of CVs in the deep soil. The short-range variation and within-horizon variation were smaller in the A horizons than in the B and C horizons. The short-range variation affected the spatial structure (variograms) of the soil properties. In these young landscapes, short-range variation of soils is substantial and appropriate sampling schemes should be developed. We summarize three strategies for calculating or reducing soil short-range variation: (1) composite sampling, (2) increasing sampling density using proximal sensing instruments, (3) utilize prior information of the study area.