Ergodic site amplification models for active tectonic regions are conditioned on the time-averaged shear wave velocity in the upper 30 m (V-S(30)) and the depth to a shear wave velocity isosurface (z(x)). The depth components of such models are derived using data from sites within many geomorphic domains. We provide a site amplification model utilizing V-S(30) and depth, with the depth component conditioned on type of geomorphic province: basins, valleys, and mountain/hills. As with current models, the depth component of our model is centered with respect to the V-S(30)-scaling model using differential depth delta z(x), taken as the difference between a site-specific depth and a V-S(30)-conditioned average depth. Using data from southern California, we find that long-period site response for all sites combined exhibits relative de-amplification and amplification for negative and positive differential depths, respectively. Individual provinces exhibit broadly similar trends with depth, but amplification levels are on average stronger in basins such that little relative de-amplification occurs at negative differential depths. Valley and mountain/hill sites have, on average, weaker amplification levels but stronger scaling with delta z(x). Site-to-site standard deviations vary appreciably across geomorphic provinces, with basins having lower dispersions than mountain/hill sites and the reference ergodic model.