Source depth estimation is a key process in earthquake location, hazard assessment, and nuclear monitoring. For instance, hypocentral depth is an important parameter affecting the spatial attenuation of ground motions resulting from an earthquake and an important indicator of natural versus man‐made events. Resolving transition zone phase complexity is important in sparsely monitored areas, for improved location and magnitude estimation of a seismic event at regional‐to‐teleseismic distances. In response to this need, we demonstrate a new approach to first‐order depth estimation. Depth phases are notoriously difficult to identify, specifically for shallower events. The usual way of determining the depth of focus of an earthquake is to measure the difference Δ t P − pP between the arrival time of the direct P wave and the arrival time of the surface reflection pP . Once Δ t P − pP is calculated, the depth of the focus can then be estimated closely, assuming that the velocity structure between the surface and the event focus is satisfactorily known. Although this method sounds simple, one major challenge remains: depth‐phase identification. For this reason, many monitoring agencies pick few depth phases from earthquakes shallower than 70 km. For example, depth‐phase observations were reported for only about 5% of the analyst‐reviewed events listed in the revised event bulletins of the Prototype International Data Centre during the Group of Scientific Experts Technical Test No 3 (GSETT‐3) experiment (Murphy and Barker, 2006). Routine depth‐phase identification is complicated by issues including (1) focal mechanism (affecting the relative P and pP or sP amplitude), (2) differential attenuation effects, (3) variations of surface topography and of subsurface structure at the depth‐phase bounce points, (4) earthquake epicentral distance (e.g., complex coda of P phases bottoming at the transition zone of the upper and lower mantle makes pP identification difficult), (5) rupture duration for large earthquakes ( M …