Recent studies have represented time variations in the Earth's axial magnetic dipole field as a stochastic process, which comprises both deterministic and random elements. To explore how these elements are affected by the style and vigor of convection in the core, as well as the core-mantle boundary conditions, we construct stochastic models from a set of numerical dynamo simulations at low Ekman numbers. The deterministic part of the stochastic model, the drift term, characterises the slow relaxation of the dipole back to its time average. We find that these variations are predominantly accommodated by the slowest decay mode, enhanced by turbulent diffusion to enable a faster relaxation. The random part—the noise term—is set by the amplitude and timescale of variations in dipole field generation, including contributions from both velocity and internal magnetic field variations. Applying these interpretations to the paleomagnetic field suggests that reversal rates are very sensitive to rms variations in the source. Less than a 50 percent reduction in rms source variations is sufficient to prevent reversals for the recent magnetic field.

This dataset was generated using the numerical geodynamo simulation software CALYPSO (https://geodynamics.org/cig/software/calypso/).