Use of pedotransfer functions and DEM data to upscale soil hydraulic properties from the pedon to the catchment scale

The recent trend toward large-scale distributed-process modeling has generated a pressing need for detailed information on spatial distribution of input data, especially the parameters that describe the hydrologic behavior of soils. Such parameters, the so-called soil hydraulic properties, are the water retention θ( h ) and the hydraulic conductivity K (θ) functions that relate the volumetric water content θ, pressure head h, and unsaturated hydraulic conductivity K. It is often the case that soil hydraulic properties are not available from standard sources or are not in a suitable format for computer processing. Moreover, a good assessment of soil hydraulic properties and their inherent spatial variability plays a crucial role as this information may influence the simulated water fluxes even significantly. Practical applications of large-scale hydrological models also require an assessment of equivalent soil hydraulic parameters, which can be defined as areally averaged values over an appropriate spatial domain. Soil hydraulic properties can be determined using direct methods under either laboratory or field experimental conditions. Such methods are time-consuming, require complex measuring devices and skilled operators, and hence are very costly. To carry out a soil hydraulic characterization in a cost-effective way, pedotransfer functions (PTFs) are being used increasingly to estimate soil hydraulic parameters from easily measurable or already available soil data. Because of economic and logistic constraints, in most cases only limited point data are available. Several studies have shown that the distribution of soil properties can be explained, to a certain extent, by landscape variables. Digital terrain analysis can thus provide a systematic basis for deriving topographic attributes and relating them to soil variables in order to improve the prediction of the key soil hydraulic parameters. Up to now PTFs usually consider as input information only soil physical and chemical variables measured at the same spatial scale (pedon scale) of the dependent soil hydraulic variables. A natural step forward is to take, in addition to variables, such as, bulk density, texture, organic matter content, etc., also landscape variables into account. Therefore, a major objective of this study is to examine the possibility to improve the performances of published PTFs by including information coming from terrain attributes, such as elevation, slope, aspect, and wetness index. This additional information can be easily extracted from a digital elevation model (DEM) of the study area. Two PTFs were tested in this study: the PTF developed by Rawls and Brakensiek (1989) and the PTF of Vereecken et al. (1989). Evaluations were carried using soil data sampled in the ‘Fiumarella di Corleto’ catchment (Basilicata, southern Italy). Equally spaced sampling locations were established along transects in different soil-landscape units. At each location, soil cores were collected from the uppermost soil horizon and subject to laboratory measurements. Soil water retention data for the sampled area manifest evident variations with distance along the transects. Even if the tested PTFs seem to generate a reasonable picture of the average water retention behavior of the area, we improve the PTF-predictions by including landscape information in the published regression equations that define the tested PTFs. Multivariate analyses was used for this purpose, whereas statistical indicators of goodness-of-fit helped in evaluating the effectiveness of the proposed procedure.