UPPER MANTLE STRUCTURE FROM LONG-PERIOD DIFFERENTIAL TRAVEL-TIMES AND FREE OSCILLATION DATA

We have performed an inversion to model upper mantle structure using SS-S differential traveltime residuals in conjunction with previously reported structure coefficient estimates obtained from free oscillation frequency and attenuation measurements. The model we obtain is consistent with both the differential traveltime and structure coefficient data sets and provides nearly the same level of fit as independent inversions of these data sets. We find that a relatively simple model parametrization appears to be adequate for the data used. Upper mantle velocity perturbations and lateral variations in the radii of boundaries at 400 and 670 km depth are parametrized with spherical harmonic expansions up to degree and order 8. The expansions are fixed at specific depth knots (separated by roughly 200 km) spanning the range 20-670 km. Radial variations in velocity are obtained by linear interpolation between these depth knots. Smoothness constraints were applied as part of the inversion procedure, as well as corrections for the thickness and velocity variations of the crust. The smoothness constraints were adjusted to require a smoother structure in the bottom portion (400 to 670 km depth) of the model, where the structure coefficients have the greatest sensitivity. The upper part of the model is primarily constrained by the differential traveltime data. The resulting misfit of the final model shows that, on average, the individual structure coefficients and SS-S measurements are fit to within approximately 1.5 standard deviations. We suggest that some of the misfit is due to lower mantle structure which is not included in the modelling. The power as a function of harmonic degree for the top portion of the model (20 to 220 km depth) does not fall off significantly at the higher harmonic degrees, suggesting that the differential traveltime data may require models which include structure for degrees greater than 8. Boundary perturbations at depths of 400 and 670 km trade off strongly with the velocity variations. Allowing variations of the order of +/- 10 km in the radii of the discontinuities at 400 and 670 km provides a slight improvement to the data fit. Comparisons of several depths of our upper mantle velocity model to the geoid show the strongest correlation at degree 2 for structure near 670 km depth. The distribution of slab structures is highly correlated to the degree 2 components of the model for depths of 200 to 670 km. Examination of the model in the vicinity of hotspots shows no consistent hotspot related signal.