Mantle plumes: Dynamic models and seismic images

Different theories on the origin of hot spots have been debated for a long time by many authors from different fields, and global-scale seismic tomography is probably the most effective tool at our disposal to substantiate, modify, or abandon the mantle-plume hypothesis. We attempt to identify coherent, approximately vertical slow/hot anomalies in recently published maps of P and S velocity heterogeneity throughout the mantle, combining the following independent quantitative approaches: (1) development and application of a "plume-detection" algorithm, which allows us to identify a variety of vertically coherent features, with similar properties, in all considered tomographic models, and (2) quantification of the similarity between patterns of various tomographic versus dynamic plume-conduit models. Experiment 2 is complicated by the inherent dependence of plume conduit tilt on mantle flow and by the dependence of the latter on the lateral structure of the Earth's mantle, which can only be extrapolated from seismic tomography itself: it is inherently difficult to disentangle the role of upwellings in "attracting" plumes versus plumes being defined as relatively slow, and thus located in regions of upwellings. Our results favor the idea that only a small subset of known hot spots have a lower-mantle origin. Most of those that do can be associated geographically with a few well-defined slow/hot regions of very large scale in the lowermost mantle. We find evidence for both secondary plumes originating from the mentioned slow/hot regions and deep plumes whose conduits remain narrow all the way to the lowermost mantle. To best agree with tomographic results, modeled plume conduits must take into account the effects of advection and the associated displacement of plume sources at the base of the mantle.