Relative arrival-time upper-mantle tomography and the elusive background mean

The interpretation of seismic tomographic images of upper-mantle seismic wave speed structure is often a matter of considerable debate because the observations can usually be explained by a range of hypotheses, including variable temperature, composition, anisotropy, and the presence of partial melt. An additional problem, often overlooked in tomographic studies using relative as opposed to absolute arrival-times, is the issue of the resulting velocity models zero mean. In shield areas, for example, relative arrival-time analysis strips off a background mean velocity structure that is markedly fast compared to the global average. Conversely, in active areas, the background mean is often markedly slow compared to the global average. Appreciation of this issue is vital when interpreting seismic tomographic images: high and low velocity anomalies should not necessarily be interpreted, respectively, as fast and slow compared to normal mantle. This issue has been discussed in the seismological literature in detail over the years, yet subsequent tomography studies have still fallen into the trap of mis-interpreting their velocity models. I highlight here some recent examples of this and provide a simple strategy to address the problem using constraints from a recent global tomographic model, and insights from catalogues of absolute traveltime anomalies. Consultation of such absolute measures of seismic wave speed should be routine during regional tomographic studies, if only for the benefit of the broader Earth Science community, who readily follow the red = hot and slow, blue = cold and fast rule of thumb when interpreting the images for themselves.