In 2004, Raffaella Montelli and co-workers published the first global tomographic model for P velocity that used finite-frequency theory (Montelli et al., 2004). At the time, we had only two frequency bands available for P waves: the low-frequency cross-correlation observations from Scripps (courtesy of Guy Masters), and the published ISC onset times for P, which we assumed to be sufficiently `high frequency' to be modeled with ray theory (the idea was to extract the extra information available from the different resolution).

The models showed, for the first time, the existence of a dozen lower mantle plumes. At the time, Morgan's `plume hypothesis' was under attack. Moreover, many seismologists were understandably not eager to throw their (ray-theoretical) software in the garbage bin. Consequently, the new global model was not everywhere favourably received... In answer to Adam Dziewonski who challenged us to show that S velocity models give the same result, we also analysed S data - though this data set is only reliable for the low frequency band, and therefore not as well constrained. We also redid the analysis for P waves and corrected a (minor) error in the crustal corrections. The resulting models were published by (Montelli et al., 2006). These are the models you can download from this page.

To download the models, click on this figure:- The models were a result of the first attempts to see how much
improvement can be obtained in the
*lower mantle*using finite-frequency theory. Ray theory is less questionable in the upper mantle, and we made no effort to improve upon existing ray-theoretical models for the upper mantle. - In particular, no surface wave or normal mode data have been used to construct the S-velocity models and low-frequency body wave data are all for teleseismic distances,
- The model parameterization is for an unordered grid, using tetrahedral interpolation. Software to extract cross-sections from the grid for plotting with GMT is provided in the tar file,
- The GLOBALSEIS project aims to construct global models from more complete data sets. Stay tuned...