We propose here a methodology to build near-surface velocity models by joint inversion of traveltime and highresolution airborne EM (AEM) data. The resulting velocity and resistivity models are steered to be structurally similar through the inclusion of a cross-gradient term in the objective function. The inversion is stable and results in better-fitting velocity and resistivity models. The resulting velocity model is then used to compute statics corrections on pre-stack seismic data. We tested the method on highquality coincident 3D seismic and AEM data from Canada by computing three different near-surface velocity models: Model 1 is a traveltime tomography using the first breaks of all the seismic shots and receivers, Models 2 and 3 are a traveltime tomography and a joint seismic-AEM inversion with a limited number of shots. The resulting stacks using statics corrections from Models 1 and 3 are very similar but the stack using Model 2 is not as sharp as the others. Our results suggest that adding AEM data to a seismic dataset with fewer shots produces seismic images as good as when a large number of shots are included.