Constraints on the Structure of Mantle Convection Using Seismic Observations, Flow Models, and the Geoid

Abstract

The establishment of the theory of plate tectonics in the late 1960s has left little doubt that the mantle is convecting. The plates themselves form the cold upper thermal boundary layer of the mantle convection system; the cooling of oceanic plates as they move away from midoceanic ridges provides the mechanism by which the Earth loses most of its heat (e.g., Sclater et al., 1980; O'Connell and Hager, 1980). The mantle is in turn cooled by the cold slabs that plunge into Earth's interior at subduction zones. Although plate tectonics implies that convective motions in the mantle are the dominant mechanism for heat transport, and we can measure the surface motions associated with it, we are remarkably ignorant of even the gross features of the interior flow field associated with this mantle circulation. Only at subduction zones, where seismicity presumably marks the particle trajectories of the cold descending boundary layer, do we have direct evidence for the interior flow pattern and state of stress. Most of what is understood, or thought to be understood, about convection in the Earth's interior is based on comparison of simplified models to observations taken at the surface. Examples of these models of mantle convection are given in the other chapters of this book, as well as in the general geophysical literature. These include studies of convection in media with uniform rheology (Busse, this volume; Jarvis and Peltier, this volume), interpretation of travel time anomalies from deep earthquakes in terms of simple thermal models of subducted slabs (Jordan eta!., this volume), interpretation of geochemical anomalies in terms of models of the distribution of mantle heterogeneities (Hart and Zindler, this volume), and interpretation of changes in the Earth's shape and rotational parameters in terms of models of mantle rheology (Peltier, this volume). In order to be useful, models must be simple enough to understand, and yet contain enough of the essential physics to be applicable. The line between oversimplification and overwhelming complexity is a fine one, and its positioning is a matter of subjective judgement, particularly when some observations have a fairly small signal to noise ratio. The ultimate test of a particular model is whether it can satisfy, within their uncertainties, the observations. If it cannot it must be rejected, although unfortunately, the converse is not true. The more types of observations a model can satisfy, however, the more likely it is to be correct.