The assimilative mapping of ionospheric electrodynamics technique has been used to derive the large-scale high-latitude ionospheric convection patterns simultaneously in both northern and southern hemispheres during the period of January 27-29, 1992. When the interplanetary magnetic field (IMF) B• component is negative, the convection patterns in the southern hemisphere are basically the mirror images of those in the northern hemisphere. The total cross-polar-cap potential drops in the two hemispheres are similar. When B• is positive and IB•I > B•, the convection configurations are mainly determined by B• and they may appear as normal "two-cell" patterns in both hemispheres much as one would expect under southward IMF conditions. However, there is a significant difference in the cross-polar-cap potential drop between the two hemispheres, with the potential drop in the southern (summer) hemisphere over 50% larger than that in the northern (winter) hemisphere. As the ratio of decreases (less thn one), the convection configuration in the two hemispheres may be significantly different, with reverse convection in the southern hemisphere and weak but disturbed convection in the northern hemisphere. By comparing the convection patterns with the corresponding spectrograms of precipitating particles, we interpret the convection patterns in terms of the concept of merging cells, lobe cells, and viscous cells. Estimates of the "merging cell" potential drops, that is, the potential ascribed to the opening of the dayside field lines, are usually comparable between the two hemispheres, as they should be. The "lobe cell" provides a potential between 8.5 and 26 kV and can differ greatly between hemispheres, as predicted. Lobe cells can be significant even for southward IMF, if IBl > IBI. To estimate the potential drop of the "viscous cells," we assume that the low-latitude boundary layer is on closed field lines. We find that this potential drop varies from case to case, with a typical value of 10 kV. If the source of these cells is truly a viscous interaction at the flank of the magnetopause, the process is likely spatially and temporally varying rather than steady state. New Zealand. 6491 6492 LU ET AL.: HIGH-LATITUDE IONOSPHERIC CONVECTION PATTERN Pedersen and Hall conductance models are obtained by combining the auroral conductance model of Fuller-Rowell and Evans [1987] with an empirical model of conductance produced by solar extreme ultraviolet radiation based on Chatanika radar observations. The statistical electric potential model is based on Millstone Hill radar observations [Foster et al., 1986]. Both conductance and potential models are parameterized by the hemispheric power index (HPI) [Foster e! al., 1986]. A very important feature of AMIE is its ability to give quantitative information about the uncertainty in the resultant patterns, so that features mapped reliably can LU ET AL.' HIGH-LATITUDE