Accessible, explicit, high resolution soil information is crucial for land management, resource allocation, and agriculture. This thesis investigates how to build a comprehensive methodological framework to improve Canada’s soil maps by updating existing soil property maps for Middlesex County, Ontario. First, the most accurate soil depth functions to standardized depths per soil property are determined. Next, the highest accuracy covariates per soil property are defined. Finally, interpolation and machine learning algorithms are explored, to find the highest accuracy soil property maps per soil property. Geostatistical and deterministic algorithms can work well to interpolate soil organic matter data; the equal area quadratic spline function accurately standardizes soil profile depths - all horizons being present; and different covariates and soil property prediction methods are necessary for accurate 3D soil property maps for different soil properties at different depths. This methodological framework can be used to refine soil maps for Ontario, Canada.