Nutrient accumulation in soils beneath confined livestock areas is a potential source of groundwater contamination. Electromagnetic induction (EMI) has become a practical method to assess nutrient content, with multiple linear regression (MLR) as the statistical method often employed to translate EMI readings into nutrient content. The purpose of this research is to compare and contrast the performance of spatially referenced MLR models that include secondary, 'easy-to-acquire' predictor variables such as spatial coordinate locations, soil water content and elevation information with MLR models based solely on EMI readings. Six feedlot areas were surveyed with an EM38 conductivity meter and between 6 and 12 sites at each feedlot were sampled at five different depths. The electrical conductivity (EC(e)), nitrate (NO3-) and phosphate (PO4(3-)) concentrations were measured and used as response variables. Analyses were performed using two different approaches: the response variables in individual layers and response variables by combining the layers within the soil profile. The results of both MLR methods were comparable in most instances because the models preferentially incorporated predictors derived from EM38 readings. Differences between the models were more evident when predicting NO3- and PO4(3-), even though prediction of these two analytes by either method was generally poor. Combined profile analysis was more effective for defining nutrient build-up because by-layer analysis gave non-significant or poor models in many instances.