Hotspots are commonly assumed to be the surface expressions of mantle plumes that originate deep in the mantle. The relative stability of hotspots has led to the interpretation that plumes are rooted in a relatively stationary part of the deep mantle that serves as a reference frame for plate motions. Nevertheless, convection in the deep mantle will move plume sources, and plumes must ascend through a convecting mantle, which may divert them horizontally as they rise.
There are some indications of the apparent motions of hotspots from plate motion reconstructions, and relative to a hotspot reference frame there is a net rotation of the lithosphere. However, if the source region of plumes is moving substantially, or if the plumes are moved horizontally by mantle flow, then the apparent motions of hotspots may be biased.
We have calculated the trajectories of plumes during their ascent in the mantle to address several questions: 1. What are the conditions for the relative stability of hotspots with respect to plate motions? 2. What are the biases that mantle flow may introduce in inferences of plate motion and net rotation of the lithosphere? 3. What are the implications of the models for flow and properties of the deep mantle? 4. Does the distribution of hotspots reflect lateral variations in the source region of plumes? 5. Can we identify the locations of plumes at depth in the mantle?
We calculate the viscous flow in the mantle using recent seismic tomographic models with assumed thermodynamic relations among seismic properties and density as well as viscosity structure; these models also include the observed plate motions over the last ~ 100 My. The trajectories of plumes that rise from the lower mantle to the surface through the ambient flow field in the mantle are calculated for a range of ascent velocities. For hotspots to be relatively stationary with respect to one another requires that lower mantle flow velocities must be small compared to plate motions; this requires an increased viscosity in the lower mantle as well as relatively modest temperature variations driving the flow. The viscosity of the upper mantle must be substantially lower in order to account for the sharp bend in the Hawaii-Emperor hotspot trace.
The return flow in the lower mantle beneath the Pacific plate biases the apparent motion of the Pacific plate relative to the hotspots to a high value. For our preferred model, we obtain coherent motion of Pacific hotspots in the mean lithospheric reference frame, as well as relative to African hotspots, as a result of return flow in the lower mantle antiparallel to plate motion. Coherent motion of groups of hotspots can largely explain the relative motion of Pacific and African hotspots during the last 43 million years. Before that, it is necessary to invoke additional Pacific-Antarctic motion at an unknown plate boundary. Mean lithospheric rotation is substantially reduced from that of previous models, but it is not eliminated.
Plumes that are severly tilted by flow in the mantle may not survive to cause hotspots; model calculations show that regions where plumes would be tilted by more than 60° from vertical correspond to regions where few hotspots are observed. The source regions of plumes deep in the mantle are usually displaced from the positions of hotspots on the surface by up to ~ 20°; they also tend to be closer to regions of slow seismic velocities in the lower mantle than are the surface hotspots.
The resulting model of mantle viscosity and flow may bear on the extent of flow and mass transfer between the upper and lower mantle. The rates of flow and mixing in the upper mantle can be estimated from plate motions, where toroidal components of flow promote mixing. The flow in a higher viscosity lower mantle would be substantially slower, and primarily poloidal. At present rates, the time scale for the slab flux to fill the mantle is several billion years. (The time scale for passing the mass of the mantle beneath mid-ocean ridges is roughly the same.) However, ridges sweep over the surface and sample the upper mantle on a time scale of 500 My. Thus the upper mantle should be sampled repeatedly by ridge processes while parts of the lower mantle may never have been sampled. Recent flow models (by M. Manga, U. Oregon) show that heterogeneous regions (blobs) in the lower mantle that have higher viscosity (due to composition or mineralogy) may persist relatively undeformed for several billion years. Any high viscosity heterogeneities in the early mantle thus could have survived to provide the source regions of noble gases in plumes. Plumes and other convective processes would then only slowly remove the gases from the blobs, thus accounting for the slow transfer of such material from the "lower" to "upper" mantle geochemical reservoir. Such a model may help reconcile geochemical observations with seismic and geodynamical models.