A method is described that enables the boundary conditions of numerical models to be calibrated to individual or groups of stress measurements. The method was developed to interpret stress measurements made in mines where it is not possible to locate the measurement points far enough away from excavations to obtain a direct measurement of the pre-mining stress field. It can also be used to analyze measurements that are influenced by surface topography. The stress field at any point is assumed to be comprised of gravitational and tectonic components. The tectonic component is assumed to act entirely in the horizontal plane in the far-field and at the model boundary. Unit normal and shear tractions are applied to the model boundaries and the response is computed at the location of the measurement points in the model. An optimization procedure is used to compute the proportions of each unit response tensor that is required, in addition to the gravitational stress, to reproduce the measured stress at the measurement point in the model. Scaling of the measured stress tensor can be included in the optimization to account for incorrect rock modulus scaling of the measured strains. The method is demonstrated using a synthetic set of stresses from a numerical model in which the measurements are influenced by a nearby excavation and topography.