How well do geomagnetic indices predict storm evolution in observations and simulations?

The geomagnetic indices are widely used to monitor and predict the magnetospheric state and storm evolution. Observationally, it is difficult to assess how well they perform and when they are at their best, due to lack of global observations. On the other hand, simulations provide us complete coverage of the dynamics, albeit with limited physics description. We compare the statistical properties of geomagnetic storms using solar wind data, geomagnetic indices, and over 100 storms simulated with the the Space Weather Modeling Framework (SWMF) Geospace code. Starting from the magnetospheric boundary, the subsolar magnetopause location shows good agreement with the Shue (1998) model, but the magnetotail flaring is consistently much smaller in the simulation than that predicted by the Shue model. In the magnetotail, the ring current (defined as flowing inside of 8 Re) is closely correlated with the Dst index during the recovery phase, but there are substantial deviations between the two during the main phase, caused by the tail current contribution to Dst. In the auroral oval region, the simulation produces consistently lower AL values than those observed. Further, we examine the solar wind coupling functions and their correlation to the geomagnetic indices as observed and evaluated from the simulation, and find that the scatter of the distributions is almost the same. This may indicate a possibility to improve the coupling functions by examining the time series of the global evolution in the simulation.