Development of new dielectric models for soil moisture content using mixture theory, empirical methods, and artificial neural network

Environmental, geotechnical, agriculture, and water resources engineers all rely on accurate measurements of soil moisture content. The most widely used technique for determining soil moisture content is the electromagnetic method, which employs dielectric models to relate soil dielectric properties to its moisture levels. This paper introduces an innovative electromagnetic sensor designed to measure the dielectric properties of moist soil. The dielectric properties of seventeen coarse-grain soil samples and seventy-five samples with both coarse and fine grains at varying moisture contents, textures, and densities were measured. The findings were used to evaluate the effectiveness of the existing most common theoretical and empirical models for soil moisture measurement. The results show that all existing models have difficulties with accurately quantifying the soil moisture content. In response, this study developed three new types of dielectric models: a theoretical volumetric model, a general empirical model that addresses the shortcomings of existing models, and an artificial neural network (ANN) model, which demonstrated a higher potential for accurately predicting soil moisture content. The best new theoretical volumetric model was the power model, with a power of 0.9 for the dielectric constant and 1.4 for the loss factor. The best new general empirical model developed in this study considered soil density, texture, and moisture, achieving correlation coefficients of 97.6% for the dielectric constant and 97.2% for the loss factor. The developed ANN models to predict the dielectric properties of moist soil provided high correlation coefficients of more than 98.5%.