Performance of TDR calibration models as affected by soil texture

Time domain reflectometry (TDR) is widely used to measure the soil water contents. The effect of soil composition on TDR calibration has to be quantified to diminish the need in re-calibration of TDR. Our objectives were: (a) to evaluate the ability of published models to fit TDR calibration data for the soils of different texture, and (b) to explore using a piecewise-constant distribution to approximate the continuous dependence of the dielectric constant on the position of water within soil pore space. We calibrated TDR in sand-kaolin mixtures, dernovo-podzolic soil, gray forest soil, and leached chernozem. Distinct changes in slopes of the calibration curves were observed for the fine-textured samples in the range of volumetric water contents from 0.13 to 0.27. The ratio of the lack-of-fit mean square error and the mean square pure error was used as a statistic for the error comparison to conclude whether a model provided a satisfactory fit of data. Published models that did not have parameters to adjust fitted well data for coarse-textured mixtures and soils but failed to fit data for fine textured mixtures and soils. The piecewise-constant distribution model with three adjustable parameters fitted well all data. Parameters of this model were related to clay content in the samples. (C) 1999 Elsevier Science B.V. All rights reserved.