Isostatic and dynamic models of Earth's surface topography can provide important insights into the driving processes of tectonic deformation. We analyze these two estimates for the tectonically-active western United States using refined structural models derived from EarthScope USArray. For the crust, use of recent Moho depth measurements and crustal density anomalies inferred from passive source seismology improve isostatic models. However, seismically determined lithospheric thickness variations from “lithosphere–asthenosphere boundary” (LAB) maps, and lithospheric and mantle density anomalies derived from heat flow or uppermost mantle tomography, do not improve isostatic models substantially. Perhaps this is a consequence of compositional heterogeneity, a mismatch between thermal and seismological LAB, and structural complexity caused by smaller-scale dynamics. The remaining, non-isostatic (“dynamic”) component of topography is large. Topography anomalies include negative residuals likely due to active subduction of the Juan de Fuca plate, and perhaps remnants of formerly active convergence further south along the margin. Our finding of broad-scale, positive residual topography in the Basin and Range substantiates previous results, implying the presence of anomalous buoyancy there which we cannot fully explain. The Colorado Plateau does not appear dynamically anomalous at present, except at its edges. Many of the residual topography features are consistent with predictions from mantle flow computations. This suggests a convective origin, and important interactions between vigorous upper mantle convection and intraplate deformation.