Multivariate Statistical Analysis of Reflection Coefficients from Free Space Measurements of Saline Solutions

Salinity levels have a crucial impact on ecosystems, including plants, marine life, and agriculture. Antenna-based sensors are preferred over satellite remote sensing for salinity detection, as they are cost-effective, easy to deploy, and better suited for local measurements. This study employs a broadband horn antenna to measure the scattering parameters (S-parameters) of bottled water with varying salt content. The S-parameter results were evaluated using Principal Component Analysis (PCA). In PCA, the real and imaginary parts of the S-parameters were analyzed to reveal distinct differences between the salt concentrations. PC1 accounted for 58.5% and 66.54% of the variance for the real and imaginary parts of the S-parameters, respectively. Furthermore, the dataset, including the first derivatives of the real and imaginary of the S-parameters, was examined, with PC1 explaining 30.41% and 29.93% of the variance for the real and imaginary components, respectively. The reduced features obtained from PCA analysis can further be applied in machine learning models, including Artificial Neural Networks (ANN) and Deep Neural Networks (DNN), for both regression and classification tasks. This process has the potential for accurately identifying salt concentrations in water through a microwave sensing system.