Method for building a subsurface model of velocity or other elastic property from seismic reflection data using tomography. The method uses velocity scans to pick a focusing velocity model at each image point (40). The focusing velocities are used to pick depth errors from tables (60) generated using a tomographic inversion matrix (30) and a suite of different velocity models (10). The depth errors are then reconstructed at each image point from the velocity scans based on the difference between the base velocity model and the most coherent velocity from the scan (70). The reconstructed depth errors are used to compute the velocity model update (80).