Constraining a numerical geodynamo model with 100 years of surface observations

Abstract

SUMMARY To constrain numerical geodynamo models with surface geomagnetic observations, we develop an assimilation scheme for mapping surface geomagnetic data into numerical dynamo solutions. This approach is similar to the optimal interpolation that has been used in atmospheric data assimilation for many years. The scheme is numerically stable and computationally efficient, and has been tested in a series of observing system simulation experiments with either simplified MHD systems or synthetic data from numerical dynamo simulations. In this study, the algorithm is integrated with our MoSST core dynamics model and is tested with 100-yr surface geomagnetic observations, starting in 1900. These experiments use 5 and 20-yr analysis intervals, and forecasts are made every 5 yr starting from 1905. We show that forecast errors gradually reduce over the course of the assimilation. Shorter assimilation runs (starting in 1960 and 1980) result in much larger forecast errors. We investigate the causes of the error reduction by examining how the assimilation impacts the unobserved variables. We discuss the implications for the ability of much longer assimilation runs to constrain the physical fields within the Earth's outer core.