Regionalization of soil water retention curves in a highly variable soilscape, I. Developing a new pedotransfer function

Geostatistically interpolated soil properties were combined with a pedotransfer function (PTF) to predict the three-dimensional variability of water retention curves (WRCs) in a highly variable soilscape. A new PTF had to be developed to account for the extreme variation in soil parameters: texture varying between gravel and clay, organic C content up to 81 g kg^-1, and bulk density from 0.80 to 1.85 Mg m^-3. A common procedure to generate such a PTF is first to parameterize the WRCs with a function, and then to calculate regression equations, linking the function s parameters with soil properties. This procedure could not be used, however, because of the overparametrization of possible functions with respect to the eight measured data points of the WRCs. Therefore, the parameters of a Van Genuchten-type function, θ(s), θ(r), α, and n, were substituted by linear equations relating these parameters with soil properties in a physically meaningful way: θ(s) = f (porosity, clay), θ(r) = f (clay, organic C), α = f (d(g)), and n = f (1/σ(g)). That is, the particle-size distribution parameters d(g) and σ(g), were assumed to be related to the pore-size distribution parameters α and n. The substituted Van Genuchten function was then fitted to all WRC, data to estimate the slopes and intercepts of these relations. More than 99% of the WRCs' variation could be explained by this model. The suitability of the model as a PTF was tested with two additional data sets. It produced reliable predictions within the study area as well as when transferred to other soils. Compared with another PTF, the new PTF improved the prediction of WRCs by 60% within the study area. This improvement was mainly caused by accounting for skeletal soils and soils with low density and high organic matter content. Due to its wide range of validity and its inclusion of physically meaningful relations, this new PTF may be reliably applied to other soilscapes. Future efforts to improve the prediction of WRCs should concentrate on developing simple methods to measure the pore-size distribution.