A testbed for geomagnetic data assimilation

Geomagnetic data assimilation (DA) merges observations of Earth's magnetic field with numerical geodynamo models to estimate the dynamic state of the core and initialize forecasts. While geomagnetic DA has become a very active field over the last decade, many questions remain concerning the appropriate DA techniques for this setting. The complexity and computational expense of numerical geodynamos limits the extent of numerical experiments that can be performed with a geomagnetic DA system. For this reason, we present a new, computationally inexpensive "proxy model'' of the geodynamo that can be used to extensively test techniques and explore the capabilities and limitations of geomagnetic DA. The approach of using proxy models to develop DA systems has been widely employed in, e.g., atmospheric and oceanographic DA. We present a large set of systematic numerical experiments with the proxy. The results illustrate the importance of developing techniques in geomagnetic DA known as "localization" and "inflation", which are used to tailor a DA system to a particular problem and have been critical to the success of atmospheric and oceanographic DA. We evaluate candidate localization and inflation methods with the proxy and propose techniques for use in operational geomagnetic DA systems. Additionally, the results highlight challenges in determining uncertainties for estimates of the core state and forecasts of the geomagnetic field.