Predicting soil properties over a region using sample information from a mapped reference area

Publisher Website
Google Scholar
Abstract
Summary: A method for mapping soil properties at a regional scale with acceptable precision and cost was investigated. It combines soil classification and interpolation, and uses sample information from a reference area and simple soil observations over the region. The method consists of two stages. First is the prediction of soil properties at a set of sites covering the region by classifying each site according to the soil classification of the reference area, and then by assigning to each site the values of soil properties measured at the representative profile of its soil class. The second is by interpolating the predictions of the properties from the classes at the observation sites. Three methods were considered: kriging, inverse squared distance and nearest neighbour. The performance of classification combined with interpolation was evaluated for mapping water content at wilting point in an area of 1736 ha in a physiographic region of Southern France. The method was compared with conventional methods, namely kriging with actual data and prediction from a soil map at a scale of 1:100000. Estimates from classification combined with kriging were more precise than those from the 1:1OOOOO soil map in all instances, and close to those of ordinary kriging with actual data when prediction points were near the observation points. Classification combined with inverse squared distance or with nearest neighbour interpolation was always less precise than classification with kriging. However, it was more precise than classification at the 1:100 000 scale except when prediction points were far from the observation points.