Application of Proximal and Remote Sensing Techniques for the Rapid Assessment of Soil Properties

Rapid evaluation of soil properties is essential for managing soil health and supporting sustainable agriculture, particularly in semi-arid regions such as Saskatchewan’s Riverhurst Irrigation District. This thesis applies reflectance spectroscopy and remote sensing techniques to assess three key soil properties: soil organic carbon (SOC), total nitrogen (TN), and electrical conductivity (EC). First, a low-cost Nanoquest near-infrared spectrometer was evaluated for predicting SOC and TN using machine learning models. Despite its limited spectral range and resolution, the instrument showed strong potential for soil assessments where high precision is not required. To broaden the scope, a salinity model was developed to estimate electrical conductivity (EC) in landscapes with subtle salinity gradients, which normally are challenging for remote sensing approaches. By integrating spectral indices with terrain metrics, the model captured EC variation across multiple depths, though further refinement is needed for irrigation-induced salinity detection. Building on this, a Local Salinization Risk (LSR) assessment tool was developed to forecast areas vulnerable to future salinization. The LSR tool combined climate, terrain, surface cover, salinity indicators, and a novel drainage classification layer derived from remote sensing and machine learning. Results identified aridity and drainage as key salinization drivers, and the tool more precisely delineated high-risk zones compared to existing regional models. Collectively, the tools developed in this thesis offer a practical, scalable framework for soil monitoring at both field and regional levels, supporting data-informed decisions for sustainable soil management in semi-arid agricultural systems.