Continuing advancements in subsurface electrical resistivity tomography (ERT) are increasing its capabilities for understanding shallow subsurface properties and processes. The inability of ERT imaging data to resolve unique subsurface structures and the corresponding need to include constraining information remains one of the greatest limitations, yet provides one of the greatest opportunities for further advancing the utility of the method. We propose a new method of incorporating constraining information into an ERT imaging algorithm in the form of discontinuous boundaries, known values, and spatial covariance information. We demonstrated the approach by imaging a uranium-contaminated wellfield at the Hanford Site in southeastern Washington State, USA. We incorporate into the algorithm known boundary information and spatial covariance structures derived from the highly resolved near-borehole regions of a regularized ERT inversion. The resulting inversion provides a solution which fits the ERT data (given the estimated noise level), honors the spatial covariance structure throughout the model, and is consistent with known bulk-conductivity discontinuities. The results are validated with core-scale measurements, indicating a significant improvement in accuracy over the standard regularized inversion and revealing important subsurface structure known to influence flow and transport at the site.