Upper limits on geomagnetic field model errors using geomagnetic data assimilation

Recent advances in geomagnetic data assimilation have led to improvements in forecasting of the geomagnetic field. Validation of the data assimilation system (DAS) has been done for historical data from the past 400 years by comparing differences between observations and forecasts (O-F) at the CMB. However, this approach is limited by the lack of knowledge about the accuracy of the observations, which come from geomagnetic field models (GFM). But we can also use geomagnetic data assimilation to estimate upper limits on GFM uncertainty by making some simplifying assumptions. We show that large forecast biases, which degrade the optimality of the assimilation, can be easily removed. This leaves just the much smaller random component of the forecast uncertainty and the remaining (O-F) from which it is much easier to the extract a maximum value for the observation uncertainty.