Statistical assessment of soil-water characteristic curve models for geotechnical engineering

Abstract

A number of empirical equations have been proposed for the soil-water characteristic curve. A nonlinear, least squares method was used to determine best-fit parameters for several empirical equations that were best-fit to 230 water content versus soil suction data sets. In addition, two proposed correction methods to accommodate high soil suctions up to 1 000 000 kPa were applied to the various soil-water characteristic curve equations. The data sets of water content versus soil suction were arranged into one of the USDA soil classifications based on their relative amounts of sand, silt, and clay (only eight soil classifications had sufficient data for later analysis). The quality of fit for each model was compared using the Akaike Information Criterion. A series of conclusions were arrived at regarding (i) the relationship between two- and three-parameter equations, (ii) the relationship between exponential and sigmoidal type equations, and (iii) the value of correction factors for the high soil suction range.Key words: soil-water characteristic curve, unsaturated soil, soil suction, regression analysis, SWCC models, Akaike Information Criterion.