Despite obvious limitations as a proxy for site amplification, the use of time-averaged shear-wave velocity over the top 30 m (VS30) remains widely practiced, most notably through its use as an explanatory variable in ground motion prediction equations (and thus hazard maps and ShakeMaps, among other applications). As such, we are developing an improved strategy for producing VS30 maps given the common observational constraints. Using the abundant VS30 measurements in Taiwan, we compare alternative mapping methods that combine topographic slope, surface geology, and spatial correlation structure. The different VS30 mapping algorithms are distinguished by the way that slope and geology are combined to define a spatial model of VS30. We consider the globally applicable slope-only model as a baseline to which we compare two methods of combining both slope and geology. For both hybrid approaches, we model spatial correlation structure of the residuals using the kriging-with-a-trend technique, which brings the map into closer agreement with the observations. Cross validation indicates that we can reduce the uncertainty of the VS30 map by up to 16% relative to the slope-only approach.