Testing the geomagnetic dipole and reversing dynamo models over Earth’s cooling history

•We use numerical dynamo models to test the geocentric axial dipole (GAD) hypothesis.•We compare dynamo models with IGRF and CALS10k geomagnetic field models.•Inclination anomalies are proportional to the relative buoyancy flux at the CMB versus ICB.•Latitudinal buoyancy flux variation likely leads to only small inclination anomalies.•Moderately shallow inclinations are predicted prior to inner core nucleation. We use the inclination test of the geocentric axial dipole hypothesis to interpret observational magnetic field models and Earthlike polarity-reversing numerical dynamo models with uniform and latitudinally variable buoyancy flux boundary conditions. Dynamo models with uniform buoyancy flux represent three convective states of the mantle and core: (1) present era Earth, driven thermochemically at the inner core boundary; (2) mantle overturn, with elevated heat flux at the core–mantle boundary, and (3) ancient Earth prior to inner core nucleation, with buoyancy production solely at the core-mantle boundary. By verifying the effect of latitudinal heat flux variation on reversal frequency, we show that polar cooling at the core-mantle boundary is likely to yield only small inclination anomalies. Instead, we find that radial heat flow structure can explain magnetic field morphologies over broad eras of Earth’s cooling history. Consistent with Earth’s present magnetic field, dynamos driven by buoyancy due to inner core growth are nearly dipolar. In contrast, elevated core–mantle boundary heat flow yields small to moderate inclination flattening due to a persistent octupole that reverses synchronously with the dipole. We find that reversal frequency and inclination anomalies can indicate the convective state of the mantle–core system. In particular our results, along with evidence of a young inner core, are consistent with a previous suggestion that an entirely liquid core contributed to shallow inclinations in Precambrian time.