The soil water retention curve plays a fundamental role in the development of studies on the dynamics of soil water, modeling of physical soil processes and plant growth. The retention curve is usually obtained by simultaneously measuring water content (θ) and soil water potential (ψ) in a single sample. An alternative procedure is to use several samples per ψ to describe the retention curve. The use of this procedure requires that the variation factors which exist among the samples are incorporated into the parameters of the mathematical functions used to describe the retention curve. The objective of this study was to obtain the retention curve using the latter procedure and fitting the θ(ψ) data using two different nonlinear functions. Undisturbed soil samples (0.05 m in diameter and 0.05 m in height) were collected from a Rhodic Ferralsol (Typic Hapludox) cropped with corn by no-tillage and conventional tillage. Ninety-six samples per soil tillage were taken at a depth of 0-0.10 m, from two positions: along the crop row and between crop rows. The retention curve was obtained using 12 matric potentials, with 16 samples per ψ: eight per tillage system and four per sampled position. Data were adjusted using the Genuchten (1980) model, VG, and the function proposed by Hutson & Cass (1987), HC. Mathematical functions relating the model parameters with the independent variables (soil tillage, sampling position and soil bulk density - Bd) substituted the model parameters in the fitting of the data. The tillage systems and sampling position exerted no statistically significant influence (p > 0.05) on the fitting of the data. With the VG function, Bd produced significant effects on the n parameter, which was described by a quadratic function of Bd. A similar result was obtained with the "a" parameter of the HC function. The retention curve proved to be sensitive to variations of Bd. The adopted procedure offered methodological advantages, involving substantially less time and lower costs to obtain the retention curve. The precision of the models was practically the same, but the model of HC possesses smaller number of empiric parameters than the model of VG.
soil water; physical soil properties; pedotransfer functions; non-linear regression; soil porosity