We present a new self-consistent model of the transport of hot electrons in the earth's magnetosphere and of their precipitation in the auroral ionosphere. In this model, the electron transport is described by the same fluid equations as in the work by Fontaine and Blanc (1983). But the electric field distribution which drives electron motions is computed from the equations of ionospheric currents, instead of being described by an empirical model as previously. The two basic equations to be solved, the hyperbolic equation governing electron transport and the elliptic equation governing the ionosphere current flow and the distribution of the electrostatic potential on the ionospheric conductor, are coupled via the ionospheric conductivities, which are functions of the electron precipitation flux at the top of the ionosphere. The numerical model is run to simulate the evolution of the system from an initial state in which there is no hot electron in the inner magnetosphere, to a steady state situation in which electrons penetrate into the inner magnetosphere from the tail and are transported onward until they precipitate into the ionosphere. To achieve this, the dawn-to-dusk electrostatic potential drop which is induced by the solar wind across the magnetosphere is turned on at t=0 and maintained constant for the rest of the simulation. Using different values of this potential drop for each individual run, we simulate various levels of magnetic activity and compare our results with observations. The formation of a belt of electron precipitation in the auroral zones is very well reproduced by the model. Both its distribution in local time and the ionospheric location of its equatorward edge agree fairly well with statistical models derived from satellite observations of the auroral electron precipitation zones.

On this latter point the agreement with observations is improved relative to our previous study. The formation of the electron precipitation belt modifies the convection pattern as it generates a localized enhancement of ionospheric conductivities in the auroral zone. This modification is essentially negligible on the sunlit side of the earth, where convection remains controlled by the distribution of conductivities due to the solar illumination. But it is very sensitive on the nightside, where the equipotentials are strongly distorted at the poleward and equatorward edges of the belt of enhanced conductivities, and where the evening vortex expands toward middle latitudes. In the equatorial plane of the magnetosphere, the shape of the plasmapause is also well reproduced; but its absolute size is underestimated for high levels of magnetic activity. A careful comparison of our computed electric fields with the fields measured at auroral and middle latitudes by several ionospheric incoherent scatter radars indicates that the shape of the local time variation of the fields is reproduced reasonably well, given the uncertainties inherent to the statistical averages derived from observations, and the intrinsic variability of the geophysical context. However, there is some indication that our model underestimates the magnitude of the meridional electric field at high latitudes, overestimates it at mid-latitudes, and predicts a position in local time of the convection vortices which is two or three hours westward of the observed ones.

These discrepancies can receive the same explanation the lack of a trapped population of energetic ions in our model, which other studies have shown to be responsible for increasing the high-latitude electric field and decreasing appreciably the mid-latitude electric field. Our next step in magnetospheric convection modeling will be to develop a two-fluid theory in which both electron and ion transport equations will be self-consistently coupled to the equation of the ionospheric electrical circuit.