Time varying effects are often not well modeled. Many of the research models are run under steady boundary conditions to produce diurnally reproducible results. Yet the solar wind, the magnetosphere and the ionosphere vary considerably on time scales of hours and even minutes. The MSFM provides a good time dependent calculation for some purposes, but it is a regional model which is not internally self-consistent and does not properly account for inter regional coupling. Two research models have recently been run with time varying inputs to attempt to provide more realistic simulations. On a global scale, Fedder et al. [1994] have used an extended series of IMP--8 solar wind data to drive their magnetohydrodynamic (MHD) model of the magnetosphere in an attempt to simulate the dynamics of a substorm. Results show promise; however, conditions under which this type of run can be made are limited. At mesoscales (hundreds to thousands of km), the Utah State Time Dependent Ionospheric Model (TDIM) has been run with time varying electric field drivers to produce patches of enhanced ionization in the polar cap [ Sojka et al., 1994], similar to those which have been observed to cause increased scintillation in radio signals and communication disruptions during negative IMF BZ conditions. TDIM has also been used to model the transient ionospheric conditions of traveling convection twin vortices which have been observed in the vicinity of the cusp and boundary layers [ Schunk et al., 1994]. Accurate simulation of time varying phenomena at scales of a hundred km to global is an essential part of space weather prediction.
A second related problem is the triggering mechanism of substorms. The current space weather model of the inner magnetosphere (MSFM) is time dependent but is unable to trigger a substorm. The event must be empirically determined and then the resulting action followed. There is a lack of agreement in the community on the physical processes that cause substorm initiation although focus has moved toward the inner magnetosphere in the last few years (the reader is referred to the proceedings of the 1992 Kiruna conference Substorms 1 for a discussion of various models). MHD models have added ad-hoc resistance to trigger substorm-like effects [see Hesse and Birn, 1992].
Before their untimely death, Goertz and associates [ Goertz et al., 1993] were able to use solar wind data to predict the AE, AL and AU indices over a specific limited data set suggesting direct driving by the solar wind. These auroral activity indices are derived from ground magnetometer measurements of the magnetic field due to ionospheric currents from 11 to 14 stations spaced around the auroral region, and represent the maximum perturbation from the eastward electrojet (AU), the maximum from the westward electrojet (AL) and the sum of the magnitudes (AE). The simple analytical model of Goertz et al., derived on the basis of frontside reconnection, force balance in the tail and Alfvén wave coupling to the ionosphere (magnetohydrodynamic waves associated with transverse motion of lines of magnetic force), optimized 6 free parameters for the data set studied. McPherron and Rostoker [1993] were unsuccessful in reproducing their results on other data sets and suggested that the free parameters were unique to that data set and that there are also storage release contributions to the currents. The release of the stored energy in an unloading process may or may not be directly tied to the driver. Klimas et al. [1992] have proposed a low-dimensional analog model for the substorm unloading process. Baker and colleagues [see Vassiliadis et al., 1990, 1993; Vassiliadis, 1994] have used locally linear prediction filtering (of non-linear phenomena) and deterministic chaos theory to characterize the magnetospheric response. A low number of filter coefficients (4 to 6) describe the magnetospheric behavior, and a few dominant degrees of freedom characterize the influence of the solar wind. In the relevant state space there are a limited number of dynamic trajectories, and this recent work shows that the magnetosphere tends to remain on one trajectory for hours. The substorm ``unloading'' or energy release is built into these trajectories. This explains why the Goertz et al. method (which did not contain any loading--unloading or other feedback features) worked once but was not reproducible for other data sets. Thus, by finding the proper state condition, AE, AL, and AU can be predicted for as long as the magnetosphere remains on that trajectory in input state space. One to two hour predictions with 80 to 90% confidence factor are possible, and limited accuracy extensions may also be possible. This technique could provide an alternative means for following substorm activity until the physics of substorm onset are sorted out.