This study explores the full 3D earthquake location for the Australian continent, exploiting the recent 3D Australian Seismological Reference Model (AuSREM). Seismic velocities from AuSREM were used as input to precompute finely spaced P‐ and S‐travel‐time grids for each station in the Australian National Seismograph Network using the multistage fast marching method. Travel times from anywhere in the grid to the corresponding station can then be computed by interpolation. The location search using these travel times is based on matching observed and computed arrival times using the neighborhood algorithm. All computations involved can be performed in practical time frames on a single processor computer.
The performance of the 3D approach relative to location using the 1D global ak135 velocity model was assessed by locating a set of recent earthquakes. The arrival‐time residuals for P and S arrivals are significantly smaller when using the 3D AuSREM model. The improvements over ak135 are especially large in the 10°–18° distance range, in which a distance bias is strongly reduced and for those paths where the ak135 residuals are large. A small set of six ground‐truth events was used to assess to what extent the reduction in travel‐time residuals leads to better absolute location accuracy. The 3D location offset from the ground‐truth position is typically half that of the ak135 offset. The patterns of offsets suggest that the already fast mantle wavespeeds in western Australia need to be even faster than in AuSREM.