A measure of soil structure derived from water retention properties: A Kullback-Leibler Distance approach

Water retention curves of structured soils reflect the combined effects of pore systems associated to a given distribution of particle sizes (texture) and those that are the result of rearrangement of soil particles into soil structural units. The main hypothesis of this work was that the entropic distance between structural and textural soil pore systems can be a measure of soil structure. It was also hypothesized that such distance can be derived from water retention curves by assuming that both pore systems follow lognormal distributions and that textural pore systems are the result of random arrangements of particles sizes. The entropic difference between the distributions of the two pore systems considered was derived as a Kullback-Leibler Distance or KLD using an explicit equation that uses the geometric means and standard deviations of pores and particle size distributions derived from water retention data and from information on clay, silt and sand content. Data on water retention and texture obtained for this study was supplemented with data from the literature and with data on 1468 soils from the US National Pedon Characterization database. The KLD concept was tested by comparing physically disturbed samples with undisturbed samples over a range of soil structure conditions. Values of KLD from younger soils along two chronosequences, and from compacted and/or degraded soils were smaller than those from older soils and non compacted or degraded soils. Also, KLD values of disturbed samples were lower than those of undisturbed samples. The use of KLD in a large dataset showed that KLD was linearly related to saturated hydraulic conductivity, which is an important hydraulic property. Also, KLD was an important grouping factor in a regression tree analysis for estimation of water content at -33 kPa and in clusters defined from a combination of field and morphological variables. The KLD measure is a promising tool to characterize soil structure. Future studies should consider incorporating KLD into pedotransfer functions or other predictive schemes aimed at improving the estimation of hydraulic properties, which are sensitive to soil structure.