Resolution and Error Analysis of Lower Mantle Tomographic Models


C. Reif, G. Masters, G. Laske

Institute of Geophysics and Planetary Physics, University of California San Diego, La Jolla, CA

creif@ucsd.edu


Most regions of anomalously fast or slow seismic velocity in the mantle can be explained purely by thermal effects. However, there are regions of anomalously slow shear velocity in the lower mantle, such as the African and western Pacific superplumes, that can not be explained by thermal effects alone. While this result seems to be robust, we need to be mindful that seismic tomographic models are subject to inadequacies in data coverage and modelling parameterization. Performing error analysis on the very large matrices inherent in seismic tomography is too computationally expensive to use conventional methods. We have employed the method of Nolet et. al. (GJI 1999) where an approximation to the inverse matrix is calculated using a one-step backprojection. This approximate inverse is then used to compute the resolution and a posteriori covariance of our tomographic models. Comparison with conventional resolution/covariance estimates using an SVD analysis and Monte Carlo techniques indicate that the approximate covariance matrix is of the correct magnitude though some details of the error maps are inaccurate. Our results show that at the base of the mantle the errors are larger in the southern hemisphere (where the superplumes are located), but the model errors are still much smaller than the inferred seismic anomalies. The resolution and error maps do allow the robustness of individual anomalies to be assessed and provide a useful tool to the tomographer.


Back to the list of abstracts.
Back to main page.