Paul Roberts:

Why is the Earth magnetic? Why is its field predominantly dipolar? Why do the geomagnetic poles tend to be near the geographic poles? Why nevertheless are they persistently offset from them? What determines the strength of the field? Why has the field been at much the same strength, for most of Earth's history? What is the reason for the slow "secular variation" of the field? Why do flux patches on the core surface was and wane? Why do recognizable features of the field tend to drift westward? What determines the speed of the drift? Why does the field change its polarity at irregular intervals separated by intervals of order a million years. Why statistically does the field show no preference for one polarity over the other? Why does the reversal process itself take only a few thousand years? What causes polarity excursions?

Some of these questions have been asked for decades and some for more than a millenium. It has long been realized that the field must be due to electric currents flowing in the electrically conducting core. Becasue of the statistical indifference of the field to the sense of polarity it seems certain that these currents are driven by self-excited dynamo action. The fluid motions are most probably driven by thermal and chemical buoyancy as the core cools and the inner core grows. In 1995 and 1996, Gary Glatzmaier, a visitor to IGPP, and Paul Roberts produced simulations of the convective geodynamo that behaved much like the real thing. Their model self-reversed irregularly like the Earth's field; excursions occurred between reversals. Flux patches waxed and waned on the core surface as they drifted westward at much the right speed. They also confirmed an old idea of Hide's that the westward drift is a wave phenomenon, and that the actual motion deep in the core is eastward. For this reason they PREDICTED that the inner core would be found to drift westward; predictions are rare things in geophysics and physics. And so far seimic studies have not ruled this prediction out. The Glatzmaier-Roberts simulation is still the only one that incorporates Earth cooling in a self consistent way.