The large-scale electric potential patterns, describing ionospheric convection, are estimated for northern high latitudes during January 18--19, 1984, from combined incoherent-scatter radar and ground magnetometer observations, using the technique of Richmond and Kamide (this issue). The patterns usually have a dominant two-cell characteristic, although the intensities, orientations and shapes of the cells undergo considerable changes with time. Often evident during substorm expansive phases is a ''tongue'' of low electric potential extending toward the east along the low-latitude edge of the high potential cell at night. Time-series plots of the maximum and minimum electric potentials show that they can respond rapidly to changes in the interplanetary magnetic field Bz component.

Total estimated potential drops for this 2-day period range from about 15 kV up to 108 kV. The influence of the different types of data on the resultant estimated electric potential patterns is analyzed. Where available, the direct electric field observations by the radars primarily control the characteristics of the estimated potential patterns, while the magnetometer data have their greatest influence in regions where direct electric field measurements are unavailable. We also employ the statistical electric potential model of Foster et al. (1986) to help fill in the patterns in data-sparse regions. For the present study, data coverage is often good enough that the statistical model plays only a secondary role in determining the estimated convection patterns. The ionospheric electrical conductance observations from the Sondrestrom and EISCAT radars are very important in helping modify the statistical conductance model of Fuller-Rowell and Evans (1987) to yield modified conductance distributions suitable for interrelating the electric fields, currents, and magnetic perturbations. Analysis of the statistical uncertainty in the estimated large-scale electric field patterns shows the uncertainty to exceed 50% in the polar cap and subauroral regions and to be less than 20% only in the vicinity of the radar electric field observations. ¿ American Geophysical Union 1988