SOIL MOISTURE RETRIEVAL MODEL BASED ON DIELECTRIC MEASUREMENTS AND ARTIFICIAL NEURAL NETWORK

The detection of water content in porous materials (soils, buildings, . . .) using dielectric measurements is one of the challenges that interest many researchers. Microwave remote sensing is an efficient nondestructive technique used to detect the moisture content in a porous medium. It is based on the resolution of an inverse problem to estimate the moisture content from the dielectric measurements. However, dielectric measurements are sensitive not only to the water content but also to several factors (components of porous media, temperature, etc.), which makes the modeling of the variation of the dielectric measurements with the water content very complex. To overcome the complexity of classical analytical models, we have chosen in this paper to use the artificial neural network (ANN) method. This method is used to extract the value of the volumetric water content in a soil sample from the sensitivity of the refractive index (RI) and the normalized attenuation coefficient (NAC) for three frequencies. We have developed three ANN models with different input parameters (the organic matter content and temperature as additional parameters in the input layer). We have found that the three developed models were able to detect soil moisture with a low average error (around 10−2), and the model with the least amount of information in the input layer (only RI and NAC measurements as inputs) was able to perform the detection with almost the same performance as the two other models. © 2022 by Begell House, Inc.