V_p/V_s models provide important complementary information to V_p and V_s models, relevant to lithology, rock damage, partial melting, water saturation, etc. However, seismic tomography using body wave traveltime data from local or regional earthquakes does not constrain V_p/V_s well due to the different resolution of V_p and V_s models, with the V_p models usually better constrained than V_s. Since surface wave data are most sensitive to V_s, which leads to complementary strengths with respect to body wave data, we jointly invert body and surface wave data to better resolve the V_p/V_s models. In order to show the robustness of our joint inversion method, we compare the results to other approaches, including dividing V_p by V_s models and V_p/V_s parametrization with body wave or both body and surface wave data, using synthetic data and real data from the southern California plate boundary region. We confirm that V_p/V_s models from body wave inversion obtained by division tend to include artefacts, even when both V_p and V_s models seem reasonable. The joint inversion with V_p/V_s parametrization is found to improve the V_p/V_s model significantly and the resultant V_p/V_s model shows more geologically consistent features, such as high V_p/V_s along fault traces at shallow depths likely indicating fault-related damage. The V_p/V_s model also exhibits contrasts at intermediate depths along the Peninsular Range compositional boundary, and high V_p/V_s in the lower crust near the Salton Trough correlated with high heat flow and may indicate partial melting. The improved V_p/V_s as well as individual V_p and V_s models are useful for earthquake relocation, high-resolution Moho depth imaging and interpretation of other data and tectonic evolution in the region.