On seismological reference models and the perceived nature of heterogeneity

A key ingredient in understanding Earth structure is the integration of seismological information with results from mineral physics for both the dominant radial dependence of material properties and the 3D structure revealed by seismic tomography. In each case an important role is played by reference models for the internal structure of the Earth originally constructed to meet seismological needs. Representations of the travel times of seismic phases are required for the location of seismic events and these have been supplemented by models that match the frequencies of the free oscillation of the Earth, which have played a major role in characterising seismic sources through the centroid moment tensor derived from long-period records. Parameterised models were introduced to combat the computational demands of inversion and have been extensively used; however, in such models (e.g. PREM) the form of the seismic parameters is prescribed to have a functional behaviour, such as a cubic, based on mathematical convenience rather than physical requirements. Any 1D models face the challenge of the significant lateral heterogeneity at the top and bottom of the mantle, as well as the more subtle variations in between. It may well be possible to capture the long wavelength components of 3D heterogeneity, but the nature of available data sources means that it will be difficult to achieve comparable P and S definition for the whole mantle.In many circumstances reference models are used for comparison with models constructed for different data sets. It is then best that, e.g., a cratonic reference model is used to judge results from cratons. Regionalisation of travel-time observations provides a means of constructing such reference models for a broad range of tectonic environments. When coupled with suitable path corrections it is possible to make a good account of regional 2D variations.It is desirable that reference models used for interpretation of seismic tomography are tied as clearly as possible to mineral physics results, with a representation based on the bulk modulus K, shear modulus G and density rather than just the P and S wavespeeds. Comparator models for different mineral compositions and temperature regimes have the potential to aid the interpretation of tomographic images, particularly in the separation of the influences of temperature and composition.