Numerical Dynamo Simulations Reproduce Paleomagnetic Field Behavior

Numerical geodynamo simulations capture several features of the spatial and temporal geomagnetic field variability on historical and Holocene timescales. However, a recent analysis questioned the ability of these numerical models to comply with long‐term paleomagnetic field behavior. Analyzing a suite of 50 geodynamo models, we present here the first numerical simulations known to reproduce the salient aspects of the paleosecular variation and time‐averaged field behavior since 10 Ma. We find that the simulated field characteristics covary with the relative dipole field strength at the core‐mantle boundary (dipolarity). Only models dominantly driven by compositional convection, with an Ekman number (ratio of viscous to Coriolis forces) lower than 10−3 and a dipolarity in the range 0.34–0.56 can capture the observed paleomagnetic field behavior. This dipolarity range agrees well with state‐of‐the‐art statistical field models and represents a testable prediction for next generation global paleomagnetic field model reconstructions. Plain Language Summary Earth's magnetic field varies on a wide range of timescales, from less than a year to hundreds of millions of years and longer. Such variations are produced by the complex fluid dynamic processes in the liquid iron core, which are generally studied using three‐dimensional computer simulations. While these simulations reproduce several features of the geomagnetic field on relatively short timescales (less than 10 kyr), their compliance with the field characteristics on longer timescales has been recently questioned. Here, we present the first simulations known to reproduce the salient features of the geomagnetic field behavior over the last 10 Myr. Analyzing a large suite of simulations, we demonstrate that the most Earth‐like ones employ buoyancy sources modeling the release of light elements from the inner core, have a low enough viscosity and a magnetic field morphology which is sufficiently, but not too strongly, dipolar. Our estimates of the degree of dipole dominance agree well with those obtained from observational field models. Our findings can be employed by future studies to reliably simulate long‐term geomagnetic field behavior, hence improving our understanding of the Earth's core and its evolution. Key Points We present the first numerical geodynamo simulations known to reproduce the main features of paleomagnetic field variability since 10 Ma All simulated characteristics of paleomagnetic behavio