Objective and quantitative soil information is crucial for pedological investigations and to inform diverse decision making processes. New techniques are required so that soil information can be ascertained in a timely manner to support sampling at finer spatial and temporal resolutions. Currently, no single technique can provide information on all of the properties of interest. This research investigated the conjoint use of visible near-infrared diffuse reflectance spectroscopy (VisNIR) and portable X-ray fluorescence spectroscopy (pXRF) for the in situ investigation of soil properties, profile variability and description. Fifteen soil pits across New South Wales, Australia, were selected for their diverse representation of soil properties. Sampling at these sites involved scanning three vertical with sensor readings taken at 2.5 cm intervals to a depth of 1 m within each transect. Soils were described by traditional pit description techniques and horizon based sampling was conducted to characterise the soil in terms of mineral composition, OC, TC, TN, CEC, EC, pH and PSA. A data fusion approach involving model averaging, and a mass balance was implemented to characterise the mineral composition of soils, including phyllosilicates sesquioxides, carbonate, gypsum, quartz and feldspars. Results were validated against X-ray diffraction analysis. To explore the predictive capability of scans taken in situ, existing spectral libraries were used to calibrate VisNIR and pXRF models and identify the best use of proximal sensor data to maximise soil information gain. As not all properties of interest have detectable spectral activity by either VisNIR or pXRF, a spectral soil inference system (SPEC-SINFERS) to augment the number of predicted properties. This system involved the propagation of sensor and model uncertainties through one hundred independent simulations for each calculation, and allowed the integration of both regression models and machine learning techniques.