A geostatistical extension of the sectioning procedure for disaggregating soil information to the scale of functional models of soil processes

Soil information related to a region may only be available in aggregated form (e.g., a physical average value from bulked sampling or a mean value for a land facet or similar map unit). When this information is to be used as input to a model that describes processes in smaller units (e.g., the pedon) then it must be disaggregated to an appropriate scale. It is shown how the sectioning method of Addiscott and Wagenet [Addiscott, T.M., Wagenet, R.J., 1985. A simple method for combining soil properties that show variability. Soil Sci. Soc. Am. J. 49, 1365–1369] may be used for this purpose. The sectioning method is extended by using a (univariate or multivariate) linear model of (co)regionalisation of the variables to compute the variances and covariances of disaggregated soil properties. It is shown how the eigenvectors of the disaggregated covariance matrix can be used to compute equiprobable sections for correlated soil properties. The methods are demonstrated using the SLIM model, a functional model of leaching. The results obtained are comparable to those from Monte Carlo analysis, and emphasise the importance of accounting for the spatial variability that underlies aggregated soil information. It is proposed that cost-effective spatial sampling to determine variograms of soil properties and/or the use of average variograms from comparable regions could provide the information to make this a practicable method for disaggregating soil information for practical problems.