Estimation of electrical conductivity of a soil solution from the monitored TDR data and an extracted soil solution

In order to establish sustainable agricultural practices and to avoid excess fertiliser application, it appears important to understand the process of water and solute transport. With a view to analysing transport through the soil, based on the data obtained by means of time domain reflectometry, the relationship between the volumetric water content, the apparent electrical conductivity, and the soil solution electrical conductivity should be known. This paper proposes a new method for estimating the three parameters relationship by optimising the parameters obtained through Rhoades model with the Levenberg-Marquardt method. The proposed method systematically determines the initial parameter set required to conduct nonlinear optimisation. The method was used to estimate the continuous apparent electrical conductivity data based on the time domain reflectometry dataset, obtained from the field and occasional measurement of soil solution electrical conductivity data of the soil water, which was extracted by means of a suction sampler installed in the field. Compared with the conventional method where the parameters of Rhoades model are calibrated with a 2-step linear regression by means of laboratory experiment, the soil solution electrical conductivity estimated with the proposed method was closer to the field data, yielding smaller root mean square error values. Supplementary use of the dataset obtained through a laboratory experiment under dry and wet conditions improved the accuracy of parameter estimation.