simulations of the seasonal variations of the thermosphere and ionosphere using a coupled,...

Journal ofdmo.~phsrzc and Terrestriul Physics, Vol. 50, No. IO/ I I, pp. 903-930, 1988. W21-9169/88 $3.00+ .oO

I’rmted in Great Britain. Pergamon Press plc

Simulations of the seasonal variations of the thermosphere and ionosphere using a coupled, three-dimensional, global model,

including variations of the interplanetary magnetic field

D. REES, T. J. FULLER-R• WELL

Department of Physics and Astronomy, University College London, Gower St., London WC1 E6BT. U.K.

S. QUEGAN, R. J. MOFFETT and G. J. BAILEY

Department of Applied and Computational Mathematics, University of Sheffield, Sheffield SlO 2TN, U.K.

(Received,for publimtion 20 Mqa 1988)

Abstract-The University College London Thermospheric Model and the Sheffield University Ionospheric Convection Model have been integrated and improved to produce a self-consistent coupled global thermospheric/high latitude ionospheric model. The neutral thertnospheric equations for wind velocity, composition, density and energy are solved, including their full interactions with the evolution of high latitude ion drift and plasma density, as these respond to convection, precipitation, solar photoionisation and changes of the thermosphere, particularly composition and wind velocity. Four 24 h Universal Time (UT) simulations have been performed. These correspond to positive and negative values of the IMF BY component at high solar activity, for a level of moderate geomagnetic activity, for each of the June and December solstices. In this paper we will describe the seasonal and IMF reponses of the coupled ionosphere/thermosphere system, as depicted by these simulations. In the winter polar region the diurnal migration of the polar convection pattern into and out of sunlight, together with ion transport, plays a major role in the plasma density structure at F-region altitudes. In the summer polar region an increase in the proportion of molecular to atomic species, created by the global seasonal therrnospheric circulation and augmented by the geomagnetic forcing, controls the plasma densities at all Universal Times. The increased destruction of F-region ions in the summer polar region reduces the mean level of ionization to similar mean levels seen in winter, despite the increased level of solar insolation. In the upper thermosphere in winter for BY negative, a tongue of plasma is transported anti-sunward over the dusk side of the polar cap. To effect this transport, co-rotation and plasma convection work in the same sense. For IMF BY positive, plasma convection and co-rotation tend to oppose so that, despite similar cross-polar cap electric fields, a smaller polar cap plasma tongue is produced, distributed more centrally across the polar cap. In the summer polar cap, the enhanced plasma destruction due to enhancement of neutral molecular species and thus a changed ionospheric composition, causes F-region plasma minima at the same locations where the polar cap plasma maxima are produced in winter.

1. INTRODUCTION

Numerical simulation of the thermosphere from first

principles requires that the majority of the most important physical processes, which occur within the thermosphere are treated correctly in a mathematical and numerical sense (FULLER-R• WELL and REES, 1980, 1983; DICKINSON et al., 1981, 1984; ROBLE et al., 1982). It is possible to assume that a wide variety

of the energy and momentum sources which drive the thermosphere are predetermined and invariant to the response of the thermosphere. In many senses, this pragmatic assumption is valid : the thermosphere does not determine the nature of the solar UV and EUV inputs which provide important heat and ionisation sources. However, there are many instances where a very simplistic calculation indicates that the ther-

mosphere must react strongly to forcing. The wind, temperature and composition response of the polar thermosphere to ion convection and heating, associated with the aurora1 oval and polar cap, are specific examples which are now well documented by ground- based and spaceborne observation (HAYS et al., 1984; REES et al., 1983, 1985, 1986). Some of these responses of the thermosphere may change the nature or magnitude of the forcing itself.

Specific examples of interactions and feedback processes are numerous, however, in this paper, we will concentrate on the effects of some relatively straight- forward interactions which occur mainly in the polar regions, resulting from the forcing of the thermosphere by the aurora1 ionosphere. Magnetospheric convection drives plasma within the polar cap and amoral oval into rapid motion (HEPPNER, 1977;

903

904 D. REES et al.

HEPPNER and MAYNARD, 1987). These rapid motions force acceleration of the aurora1 thermosphere via the process commonly known as ion drag.

The resulting ion-neutral frictional drag causes direct heating of both ions and neutrals, commonly known as Joule heating (COLE, 1962, 1971). Induced winds may increase or decrease (but generally decrease) the ion drag and the resulting frictional heating. The induced winds (or more correctly, changed winds, since there is always a complex wind system in existence prior to a given geomagnetic disturbance) may induce a ‘back-EMF’, opposing the initial magnetospheric convective electric field. This wind system will also induce ion drifts (or changes in ion drifts) parallel to the local magnetic field. Such ‘parallel’ ion drifts will also induce a field aligned electron flow, to maintain quasi-charge neutrality. Thus, the entire vertical plasma distribution will respond to wind changes. This change of ion density distribution will modify the consequent ion drag on the neutrals and thus the wind acceleration terms and finally, the winds themselves.

Large induced changes of upper thermospheric composition are caused by combined advection and convection, induced by strong thermospheric heating in the polar regions (MAYR and VOLLAND, 1972). These composition changes (enhanced concentrations of molecular nitrogen) are particularly pronounced at F-region heights during geomagnetic storm periods and within the summer geomagnetic polar region. Enhanced concentrations of molecular nitrogen cause significant depletions of F-region plasma densities by greatly increasing the effective recombination coefficient, while the ionisation rates, due to the combination of solar photoionisation and aurora1 precipitation, are only slightly changed.

At E-region altitudes induced wind changes, particularly those caused by ion drag, reduce the effective EMF thus decreasing the effective ion drag acceleration (limiting the maximum induced winds). The induced winds also tend to decrease the local electrojet current (at all heights), since the electromotive force e(E + V,, x B) is decreased. This is independent of any plasma density and conductivity modifications, however, both processes (changed electromotive force and conductivity) will be noted by the thermosphere.

Any modifications of the horizontal current (usu- ally decreases) due to the induced winds will change the capacity of the ionosphere to carry field aligned currents (FAC) which connect to the magnetosphere. Intuitively, the feedback effects of the induced winds on the total electromotive force and the modified capacity of the aurora1 ionosphere to transmit or connect the FAC might be expected to cause some sig-

nificant effects on the magnetosphere at times of large disturbances (when the E-region winds are known to reach 50% of the E x B ion drift velocity, driven by magnetospheric electric fields, i.e. RJXS, 1971, 1973 ; PEREIRA et al., 1980). Although some numerical experiments in these areas are in progress (i.e. HAREL et al., 1981), it is reasonable to say that the theoretical and experimental exploration of these problems is at a very preliminary phase : we may not even be asking the correct questions at the present time. We do not understand whether limits of the availability of charge carriers (ionospheric or magnetospheric) are important, whether slight changes of the FAC are matched by changes in field aligned potentials, or whether slight shifts occur in the patterns of overall magnetospheric convection, or its mapping to the ionosphere, to compensate for thermospheric or ionospheric feedback processes (FULLER-R• WELL et al., 1987a). In principle, feedback processes could affect a number of magnetospheric processes which have often been thought of as purely magnetospheric/plasma physics phenomena.

The processes by which interactions between the polar ionosphere, thermosphere and magnetosphere occur and the major routes by which feedback can occur, are easy to describe intuitively. However, the thermosphere and ionosphere are so closely inter- woven by chemistry, dynamics and energetics, that quantitative modelling cannot rely on an intuitive approach.

The earliest numerical global 3-dimensional time- dependent models of the thermosphere (FULLER- Rown~~and REES, 1980,1983; DICKINK~N~~~Z., 1981, 1984; ROBLE et al., 1982) used theoretical or simple empirical models of the ionosphere, based on the Cmu (1975) global model. The lack of any response of the Chiu ionospheric model at high latitudes to geomagnetic processes (precipitation, convection) caused a gross underestimation of the magnitude of ion drag and of Joule/frictional heating at E-region altitudes. The E-region plasma densities were not so seriously underestimated and the Chiu model, used in con- junction with the 3-D T-D (or GCM) models simulated more realistic E-region winds and temperatures of the thermosphere (REES et al., 1985, 1986).

However, the underestimate of E-region forcing caused very serious problems in simulations of the geomagnetic response in neutral chemistry, as well as the E-region winds, etc. Thus a series of numerical experiments were started to find methods of coping more realistically with the underlying difficulty : how to simulate the mutual interactions of the polar ionosphere and thermosphere (REES et al., 1985, 1986; Rnns and FULLER-R• ~ELL, 1987a).

Simulations of the seasonal variations of the thermosphere and ionosphere 905

The first serious efforts to create an interactive model for the polar ionosphere and thermosphere were published by QUEGAN et al. (1982) and FULLER- ROWELL et al. (1984). In this model, data sets from a global thermosphere (UCL) and the ‘Sheffield’ polar ionosphere model (UT-independent) were iteratively exchanged until stability was achieved. The effects of the model iterations showed that significant changes in plasma density were caused by the effects of induced winds. The aurora1 oval plasma densities were greatly enhanced compared with those of the Chiu model. As a result, induced winds and heating were generally greatly increased compared with previous simulations using the global CHIU (1975) model.

However, this simple model could not be univ- ersally applied to study UT variations, let alone the effects of variable solar, geomagnetic and seasonal conditions. The next stage, developing a fully interactive thermosphere and polar ionosphere model, was complex and is described in some detail in a series of papers by FULLER-R• WELL et al. (1987b, 1988). This fully coupled model exchanges ionospheric and thermospheric parameters throughout the region poleward of 55” geomagnetic latitude. At lower latitudes, the numerical model is presently still dependent on empirical ionospheric descriptions. Thus, the major purpose of this paper will be to discuss thermosphere/ionosphere interactions at high geomagnetic latitudes, where the physics of the major ionospheric-thermospheric interactions is now con- tained within the coupled model. It will be seen that many of the ad hoc requirements for additional ‘geomagnetic’ energy sources, to explain observations, are unnecessary when the ‘coupled’ model is used : the additional energy and momentum occurs naturally through the medium of the coupled and interactive physics and the new model takes care of the feedback processes in a credible and correct way.

2. THE COUPLED GLOBAL THERMOSPHERE/POLAR IONOSPHERE MODEL

The UCL Three-Dimensional Thermospheric Model (or GCM) simulates the time dependent structure of the vector wind, temperature, density and composition of the neutral atmosphere by numerically solving the non-linear equations of momentum, energy and continuity (FULLER-R• WELL and REEs, 1980) and a time-dependent mean mass equation (FULLER-R• WELL and REES, 1983). The global atmosphere is divided into a series of elements in geographic latitude, longitude and pressure. Each grid point rotates with the Earth to define a non-inertial frame of reference in a spherical polar coordinate

system. The latitude resolution is 2”, the longitude resolution is 18” and each longitude slice sweeps through all local times, with a 1 min time step. In the vertical direction the atmosphere is divided into 15 levels in log (pressure), each layer is equivalent to one scale height thickness, from a lower boundary of 1 Pa at 80 km height.

The time dependent variables of southward and eastward neutral wind, total energy density and mean molecular mass are evaluated at each grid point by an explicit time stepping numerical technique. After each iteration the vertical wind is derived, together with temperature, heights of pressure surfaces, density and atomic oxygen and molecular nitrogen concentrations. The data can be interpolated to fixed heights for comparison with experimental data or with empirical models. The momentum equation is non-linear and the solutions fully describe the horizontal and vertical advection, i.e. the transport of momentum.

The neutral atmosphere numerical model uses an Eulerian approach. However, the ionospheric code (QUEGAN et al., 1982, FULLER-R• WELL et al., 1987b, 1988) has to be evaluated in a Lagrangian system. The complex convection patterns imposed by a magnetospheric electric field on plasma movements within the polar regions are referenced to a fixed Sun-Earth frame, assuming pure E x B drifts. The electric field is derived by merging a model of magnetospheric convection (for example one of the models presented by HEPPNER and MAYNARD, 1987) with the co-rotation potential (induced by the Earth’s rotation). Parcels of plasma are traced along their convections paths, which are often complex.

In the ionospheric code, atomic (Hi and Of) and molecular ion concentrations are evaluated over the height range from 100 to 1500 km and used in the thermospheric code within 35” magnetic latitude of the north and south magnetic poles. The use of the self-consistent ionosphere at high latitudes and an empirical description at low and mid-latitudes can result in a discontinuity in the vicinity of the plasma- pause. The ionospheric code is presently being extended to include the self-consistent calculation at mid- and low latitudes, including computation of the equatorial anomaly, but these new results will not be discussed here.

To avoid further repetition of the details of the numerical codes, the reader is referred to the series of papers previously cited for details of the UCL thermospheric model (FULLER-R• WELL and REES, 1980, 1982), for information on the Sheffield ionospheric code (QUEGAN et al., 1982; QUEGAN et al., 1986; WATKINS, 1978 and ALLEN et al., 1986), and for information on the procedures used to couple the

906 D. Ram et al.

UCL thermospheric and Sheffield polar ionospheric models (FULLER-R• WELL et cd., 1987b, 1988).

3. SIMULATIONS FOR JUNE AND DECEMBER

SOLSTICES, HIGH SOLAR ACTIVITY

Results from four thermospheric simulations will be presented here. These simulations have been generated for each of the June and December solstices, for IMF BY positive and negative conditions, for a geomagnetic activity level corresponding to approxi- mately K, = 3+ and for a moderately high level of

solar activity (P,,., cm = 185). The simulations are time dependent, that is they are UT dependent, and the results are diurnally reproducible. However, the external solar or geomagnetic inputs are not time dependent in these particular simulations.

The characteristic UT variations of the summer and winter polar regions are dependent on the offset of

the geomagnetic poles from the geographic poles. During the UT day, at all seasons, the geomagnetic polar caps are carried into and out of sunlight. There

is, therefore, a large modulation of the solar photoionisation and UV/EUV heating of the geomagnetic polar regions, which causes quite large UT variations

in plasma density, conductivity, ion drag and Joule and solar heating of the thermosphere. There are consequent large modulations of the thermospheric and ionospheric response. To avoid repetition of results presented in previous papers, these characteristic UT variations of thermospheric and ionospheric structures, and the associated thermospheric- ionospheric interactions, will not be discussed further here and the reader is referred to FULLER-R• ~ELL et rd. (1988) for a full description of the appropriate

modelling and results. Within the scope of this paper, it is not feasible to

catalogue the simulated variations of all parameters at all pressure levels. A sample of the seasonal and IMF BY variations at two pressure levels will be given, one at pressure level 12 characteristic of the F-region and the other characteristic of the E-region (pressure level 7, approx 125 km height). Data presented at these levels illustrate some of the strongest coupling processes. In particular, the height dependent response to convection models appropriate to situ- ations when the IMF BY component is either strongly positive or strongly negative, can be observed.

4. HIGH LATITUDE BEHAVIOUR OF THE UPPER AND

LOWER THERMOSPHERE AT SOLSTICE

4.1. IMF BY positive, Northern Hemisphere winter

Figure 1 (a, b) shows the plasma density and wind velocity (a) and temperature, mean molecular mass,

atomic oxygen density and molecular nitrogen density distributions (b), respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation

for the December solstice for IMF BY positive. The

data are presented at pressure level 12, for the northern polar region (poleward of 50” geographic latitude) and at 1800 UT.

Figure 2 (a, b) shows the plasma density and wind

velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density distributions (b), respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation

for the December solstice for IMF BY positive. The data are presented at pressure level 7, for the northern polar region (poleward of 50” geographic latitude) and at 1800 UT.

4.2. IMF BY positive, Northern HenzisphercJ summer


velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density

distributions (b) respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation for the June solstice for IMF BY positive. The data are presented at pressure level 12, for the northern polar region (poleward of 50” geographic latitude) and at 18 UT.


velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density

distributions (b), respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation for the June solstice for IMF BY positive. The data are presented at pressure level 7, for the northern

polar region (poleward of 50” geographic latitude) and at 1800 UT.

4.3. IMF BY negative, Northern Hemisphere winter

Figure 5 (a, b) shows the plasma density and wind velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density distributions (b), respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation for the December solstice for IMF BY negative. The data are presented at pressure level 12, for the north-

ern polar region (poleward of 50” geographic latitude) and at 1800 UT.


velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density distributions (b), respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation for the December solstice for MIF BY negative.

Fig.

1.

Sho

ws

the

plas

ma

dens

ity

and

win

d ve

loci

ty

(a),

and

te

mpe

ratu

re,

mea

n m

olec

ular

m

ass,

at

omic

ox

ygen

de

nsity

an

d m

olec

ular

ni

trog

en

dens

ity

dist

ribu

tions

(b

), r

espe

ctiv

ely,

ta

ken

from

the

UC

L

coup

led

iono

sphe

re/th

erm

osph

ere

GC

M

sim

ulat

ion

for

the

Dec

embe

r so

lstic

e fo

r IM

F B

Y p

ositi

ve.

The

da

ta a

re p

rese

nted

at

pre

ssur

e le

vel

12. f

or

the

no

rth

ern

p

ola

r re

gin

(pol

cv.a

rd

of 5

0 ge

ogra

phic

la

titud

e)

and

at 1

800

UT

.

D. REES et al.

*_ : __-- “2

u

.-,I

909

D. REES et al.

__i .-. k_

D. bES et al.

;.-: . . fi

t-- i.---~i _- .-_-_

:! : E

ck _i

-. ! --! .’ :_c

-s ; / ._-

: i -., -.-7: I-. !

,

:, .,i -- .: ../ : j

-_.: ;;

---

:-,.: ._ E !

: z i-L

D. REES et al.

<.:.“:2

D. F&ES et ai.

.T

j? ____ :.. .*T%:;::::- __.- _,.. l:_. 1 --..__ _*=

__” ---__

-G__+___--= _..--

,y- : .Z”

,?p; .-._,_

:y? y

t: t G. : -i_ .,_

: -3

I----__--__-__ ._.

-_-__i

.<I

._, -..

--

917

D. REES et al.

1- u -.

.’ ,. : -. : : .__:

D. REES et al.

--- -1‘. ---- _S+._“---

< --. :5&X :-=.--. :,i_:

r-- _ -- . ._:-i:

D. REES et al.


The data are presented at pressure level 7, for the

northern polar region (poleward of 50” geographic latitude) and at 1800 UT.

4.4. IMF BY negative, Northern Hemisphere summer

Figure 7 (a, b) shows the plasma density and wind velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density

distributions (b), respectively, taken from the UCL coupled ionosphere/thermosphere GCM simulation for the June solstice for IMF BY negative. The data are presented at pressure level 12, for the northern polar region (poleward of 50” geographic latitude)

and at 1800 UT. Figure 8 (a, b) shows the plasma density and wind

velocity (a) and temperature, mean molecular mass, atomic oxygen density and molecular nitrogen density distributions (b), respectively, taken from the UCL

coupled ionosphere/thermosphere GCM simulation for the June solstice for IMF BY negative. The data

are presented at pressure level 7, for the northern

polar region (poleward of 50” geographic latitude) and at 1800 UT.

5. SEASONAL VARIATIONS-F-REGION

5.1. Composition

The seasonal variation of neutral thermospheric composition at a fixed height level is also well estab- lished by experimental observations (i.e. HEDIN, 1983,

1987). At middle latitudes, enhanced concentrations of [N2] relative to [0] in the summer hemisphere

between 250 and 350 km feeds back into the ionospheric chemistry by increasing the recombination of the dominant ion Of, decreasing the NmF, (peak electron density at the peak of the F,-layer). Conversely, in the winter hemisphere, the enhanced

[O/N21 ratio decreases the recombination rate of the F-region 0 + ions and causes a general increase in NmF2, despite the overall decrease of solar insolation.

These general features are well illustrated by comparisons between Figs. lb and 3b, for the F-region (IMF BY positive) or Figs. Sb and 7b (IMF BY negative). The highest values of mean molecular mass

which occur poleward of 50” latitude in the winter

polar F-region equal the lowest value in the equivalent summer polar region. At high winter mid-latitudes, the mean molecular mass is close to 16, indicating a composition which is nearly pure atomic oxygen. There is then a geomagnetic polar plateau, due to heating, upwelling and outflow, where the mean molecular mass reaches 20.

This winter plateau value of mean molecular mass

is very dependent on geomagnetic activity. At very

low activity levels the mean molecular mass may be as low as 1617, while during extended geomagnetic

storm conditions the mean molecular mass may reach

values as high as 22. In the summer polar cap, the lowest values of mean molecular mass above 50” are 20, and the highest values within the geomagnetic

polar cap reach 2425. At this constant pressure level (12), this implies a 4 fold reduction in the amount of atomic oxygen going from high summer mid-latitudes

to the pole, and atomic oxygen concentrations a factor of 10 lower than those found in the high winter mid- latitude region. The variation of molecular nitrogen

density is in direct anti-phase and compensates for the atomic oxygen changes.

There are significant changes in the compositional variations caused by changes in the IMF BY sense.

These detailed changes in mean molecular mass are modest (1-2 units) compared with the larger changes

(34 units) induced by the combination of seasonal solar insolation changes and the heating and conse-

quent overturning of the thermosphere resulting from

geomagnetic forcing. However, the F-region plasma density is highly responsive to the BY-induced compositional modulation.

5.2. Winds

Geomagnetic control of the polar regions in the ionospheric and neutral thermospheric constituents has been well demonstrated by observations reported from the Dynamics Explorer spacecraft (DE-2) (HAYS

et al., 1984; KILLEEN et al., 1983; REES et a/., 1983, 1985, 1986).

At F-region altitudes, only relatively small differ-

ences in the wind patterns can be ascribed to seasonal effects. In the winter, the winds in the dawn aurora1 oval show only a slight tendency to follow the rela-

tively weak sunward ion convection (BY positive). In the summer, there is no significant indication of sunward wind acceleration at all in the dawn cell for

BY positive. Only minor seasonal differences occur in the strong clockwise circulation wind cell which follows the strong sunward ion convection in the dusk aurora1 oval and anti-sunward ion flow over the dawn side of the polar cap. Equatorward of the nightside aurora1 oval the wind flow is distinctly equatorward

in the June simulation and rather weaker in the December simulation (100 rather than 200 m s- ‘).

5.3. Plasma density

Comparing Figs. la and 3a and recalling that it is only in the region higher than 55” magnetic latitude (offset 10” toward the 12 LST meridian at 1800 UT)

924 D. kES et al.

the high latitude seasonal F-region ionospheric

anomaly is apparent : the winter high latitude plasma densities exceed those of the equivalent summer, sunlit region. Peak densities in the winter sunlit cusp region

(which is just in sunlight at 1800 UT) are about a factor of 2-3 higher than the equivalent summer region. The winter polar cap, at 1800 UT, is filled with high density plasma which is rapidly convected from the dayside through the cusp region. There is a well developed sub-aurora1 trough, from 1600 to 0800 LST at this UT (but which encircles the entire aurora1

oval at the winter solstice for much of the rest of the UT day). At UT times when the cusp is not in sunlight, the plasma density within the polar cap tends to fall to significantly lower values (FULLER-R• WELL et al., 1988).

In contrast, the summer sunlit polar cap is a region where the electron recombination rate is high, so that plasma densities are low despite the combination of

photoionisation and electron precipitation. Regions of lowest plasma density have a one-to-one relation-

ship with regions of the highest mean molecular mass. The regions of lowest plasma density within the summer polar cap are responsive to thermospheric compositional changes, induced by convection and Joule heating. The details of these responses are highly sen- sitive to the convection response to IMF BY changes and these will be discussed in detail later.

There is no feature which corresponds to the winter sub-aurora1 trough in the summer hemisphere. There is sufficient solar photoionisation in the sub-aurora1 regions (at least at 1800 UT) so that sub-aurora1 troughs do not develop in regions of plasma stagnation (which are otherwise similar to those of the equivalent winter polar regions). There is enhanced plasma destruction compared with the equivalent winter regions, but the values of mean molecular mass (2&21) are too low to cause the extreme plasma destruction rates which occur within the polar cap, where the mean molecular mass reaches 24-25. The only caveat concerning this overall conclusion is related to the southern summer polar region. At Uni- versal Times when the geomagnetic polar cap is rotated furthest from the dayside, and due to the greater offset of the southern geomagnetic pole from the geographic pole, there is no photoionisation on the nightside in the sub-aurora1 oval. A partial summer sub-aurora1 trough may occur under such conditions. This feature will, however, never develop to the extent seen in the winter hemisphere. For most of the UT day, in winter, the sub-aurora1 trough completely or nearly completely encircles the aurora1 oval. The situation depicted at 1800 UT in all these simulations, where the winter northern cusp is sunlit only

occurs for about 6 h in the UT day. Sub-aurora1 troughs which develop in the summer

hemisphere are always likely to be the fossils of an

earlier, expanded, aurora1 oval. Intense heating,

forcing enrichment of molecular species by convective overturning and advection will lead to the rapid destruction of plasma once the aurora1 source dimin- ishes at the end of the disturbance. This is likely to be the mechanism creating such troughs, which is therefore quite distinct from the plasma stagnation mechanism which is mainly responsible for the winter

sub-aurora1 trough. There are no previously published ionospheric

modelling results which are directly comparable with the present studies. The work of the Utah State Uni- versity group (SOJKA et al., 1979, 1981, 1985; SOJKA and SCHLJNK, 1983) provides a range of simulations under generally low solar activity conditions. Their

simulations use simplified polar plasma convection patterns, compared with the patterns used in these simulations. In their ionospheric computations they

also use very simplified anti-sunward, cross polar cap thermospheric wind patterns. In their computations, the MSIS series of empirical thermospheric models are used. The MSIS models certainly show the broad- scale seasonal, UT and geomagnetic responses of thermospheric composition and temperature etc., but do not resolve the detailed compositional response to geomagnetic activity. In the Utah model there is no feedback between vertical and horizontal winds and neutral composition, a feedback process which we

show has a very significant control of ionospheric behaviour. The wind patterns used in the Utah computations are not modified in response to conductivity changes, with consequent changes of ion drag and

Joule heating. Nevertheless, the Utah studies correctly simulate and identify broad scale features such as the

sub-aurora1 trough and the summer polar trough, features which we are able to examine in considerably greater detail with the feedback effects included in these present simulations.

5.4. Ionosphere-thermosphere coupling

Enhanced anti-sunward winds over the geomagnetic polar cap, strong sunward winds in the dusk aurora1 oval, enhancement of N2 : 0 ratios in the sum-

mer polar cap (due to enhanced polar upwelling) all require modifications (enhancements) of the polar ionosphere plasma density associated with aurora1 precipitation, to produce observed effects.

The most important conclusion which comes from this study, related to F-region behaviour and ionosphere-thermosphere coupling, is the distinction


between the seasonal behaviour. In the winter polar

region it is primarily the plasma convection which

determines the distribution of plasma density and so

modulates the local frictional heating and the acceleration of winds by ion drag. The transport of F- region plasma by convection and the combination of convection and co-rotation control the existence and location of tongues of high plasma density within the

polar cap and troughs of low plasma density within the polar cap and in the sub-aurora1 trough. Neutral

wind effects are significant in the vertical transport of plasma. The convective overturning of the neutral atmosphere and associated advection causes significant compositional changes which cause notice-

able, but not dominant effects on the F-region plasma distribution.

In the summer hemisphere, the effects of combined solar insolation and geomagnetic heating force a much greater convective/advective overturning of the

thermosphere. In the summer polar cap, at levels of moderate geomagnetic activity, the greatly increased

proportion of molecular species causes a fundamental change in the F-region ionospheric chemistry. As a result, compositional changes, driven by the detailed

response of the thermosphere to geomagnetic forcing, dominate the detailed ionospheric response. Regions where in the winter pole the plasma density is enhanced by rapid convection from the dayside, tend to become regions of low plasma density in summer. The combination of plasma convection and co- rotation are still important. However, in the summer polar cap, the plasma density distribution is controlled by the composition distribution. The composition is

modified by two processes : changes in the location of

Joule heating and consequent upwelling, and horizontal transport by the neutral wind. Plasma

depletions are signatures of an increase in mean molecular mass.

6. SEASONAL VARIATION%.!?-REGION

6.1. Composition

Comparisons of the seasonal variations in composition of the E-region can be seen in Figs. 2b and 4b (IMF BY positive) and 6b and 8b (for IMF BY negative). The mean molecular mass increases from winter high latitude values of around 22-26, to sum-

mer polar values of 2627 at times of moderate geomagnetic activity. The gradient of composition (mean molecular mass) over the winter pole is much greater than that over the summer pole, although the latter values are consistently higher than the highest values simulated for the winter polar region.

It is interesting that in complete contrast to the

behaviour of the polar F-region, the temperature

structures of the E-region in both summer and winter

polar regions are actually very similar. The ranges in both summer and winter polar regions, almost irres- pective of IMF BY sense, are from about 450-650 K. The highest temperatures are closely associated with regions of fast ion convection and strong Joule heating: ion temperatures follow rather similar dis-

tributions. In the F-region, peak values of mean molecular

mass are normally associated with the highest neutral temperatures. In the E-region, although there is a general polar plateau of high mean molecular mass, the fine details of the thermal and compositional dis-

tributions are not so closely associated. For example, there are consistent peaks of mean molecular mass on the dawn side of the polar cap, in a region where the E-region neutral temperature tends to be rather low.

This feature, both in temperature and in composition, is due to the anti-clockwise (cyclonic) ion flow vor-

tex. The curvature and Coriolis acceleration terms (FULLER-R• WELL and REES, 1984) have the same direction in this region and the neutral gas is spun outward, in addition to the tendency for induction of

anti-clockwise wind flow by ion drag. The induced outflow causes enhanced upwelling, yet another example of the complexity of the variety of interactions between the polar ionosphere and thermosphere and the respective roles of plasma convection, precipitation, and the neutral gas response.

6.2. Winds

The major changes of winds of the E-region result

from the distinctly different pole to pole thermal struc-

ture at the June and December solstices. There is a pronounced cyclonic vortex around the winter polar region, which is observable at high mid-latitudes, superimposed on a generally weak flow away from the sub-solar point. This circulation is replaced by a pronounced anti-cyclonic vortex around the summer polar region (again best seen at high mid- latitudes).

The wind patterns within the aurora1 oval and polar

cap show little change with season for the same con-

vection and precipitation patterns. The major wind features result from ion drag and are thus responsive to convection changes and plasma density enhancements (resulting from combined particle precipitation and solar photoionisation). The basic aurora1 oval wind pattern follows that of ion convection, as in the F-region. The wind velocities are about a factor of 2- 3 lower than at F-region altitudes.

926 D. REES et al.

A highly satisfactory feature of the present self- consistent simulations is that they show E-region winds, at times of moderate geomagnetic disturh- antes, which are of similar magnitudes and structures to those reported in experimental studies of E-region winds, such as REES (1971, 1972) and PEREIRA et al. (1980). Until we developed this self-consistent code, such correspondence of simulated E-region winds with the real world could only be obtained by exten- sive and artificial tuning of E-region plasma density distributions (even in the earlier code of QUEGAN et al., 1982 and FULLER-R• WFZLL et al., 1984).

6.3. Plasma density Under all the conditions of these four simulations,

the E-region ionospheric chemistry is not funda- mentally changed, despite the seasonal modulation of mean molecular mass described earlier. The plasma density distribution in regions outside the aurora1 oval shows an increase (as expected) in the summer hemisphere, as the result of increased photoionisation. Values on the nightside in the winter hemisphere are predicted to be low (2-S x 10’) in places. Within the geomagnetic polar cap there is a greater decrease of plasma density in the winter hemisphere than in the summer hemisphere. This is the direct result of additional solar photoionisation within the summer polar cap. On average, the E-region values within the winter polar cap are about a factor of 4 lower than in the summer polar cap.

What we see from this study is that E-region plasma densities around the aurora1 oval are primarily enhanced as the result of precipitation. The seasonal modulation and the effects of plasma wind transport are generally of little importance. Vertical plasma transport due to the horizontal winds is likely to be the most important single effect, and this may be particularly important in connection with the for- mation and subsequent behaviour of sporadic-E layers, resulting from metallic ions. Such long-lived ions are not subject to the rapid destruction of the molecular ions considered in this study and thus horizontal as well as vertical transport effects will be much more significant.

7.IMF YCOMPONENTVARIATIONS--F-REGION

7.1 Winds

There are two major responses of high latitude F- region winds to changes of IMF BY sense, firstly, within the polar cap and secondly, in the dawn aurora1 oval. A region of high-velocity anti-sunward flow

occurs on the dusk side of the polar cap for BY negative. The high velocity anti-sunward winds shift to the dawn side of the polar cap for BY positive. This has been previously reported (MCCORMACK and SMRH, 1985 ; REJLS et nl., 1986) in observations from the DE- 2 spacecraft and from ground based interferometers. The winds of the dawn aurora1 oval also show a characteristic change. For IMF BY positive, there is very little indication of sunward wind flow in response to the weak dawn aurora1 oval/polar cap convection cell. For IMF BY negative, there are significant sunward winds, however, these are always weaker than the strong sunward winds of the dusk aurora1 oval. The primary reasons for this dusk/dawn asymmetry have been discussed previously (FULLER-R• WELL and

REES, ~~~~;FULLER-ROWELL~~ al., 1984).Theasym- metric response is caused by the ‘resonance’ of the clockwise wind vortex of the dusk aurora1 oval (driven by sunward convection) and anti-sunward ion convection over the polar cap.

Coriolis and curvature accelerations balance within the clockwise vortex, of which the dusk aurora1 oval is part. In the dawn aurora1 oval the conditions for this resonance do not exist. As a result, for the same plasma densities and convection velocities there is a much smaller effective sunward acceleration. This effect has been well observed by DE-2 (&ES et al., 1983, 1985 ; HAYS et al., 1984) and was predicted by much simpler simulations than those we are describing here. However, it is perhaps reassuring to see that the asym- metric dusk/dawn wind response is still observed in the present rather more sophisticated simulations, where the possibility of a complex feedback process between the thermosphere and ionosphere, which might change the nature or magnitude of the wind acceleration and response can now be discounted, since the coupled model accounts for all the major feedback mechanisms with the possible exception of feedback processes involving the magnetosphere/ ionosphere coupling.

7.2. Plasma density

In a comparison between Figs. la (BY positive) and 5a (BY negative) the major response of the winter F- region plasma density in response to IMF BY changes can be seen. For BY negative (5a), a long extended tongue or plume of plasma is rapidly transported anti- sunward from the dayside and cusp. This rapid transport is produced by the situation that the strong anti- sunward plasma convection on the dusk side of the polar cap and the effect of co-rotation in that location are in the same direction. On the dawn side of the polar cap, the weak anti-sunward plasma convection


is opposed by co-rotation so that a deep plasma hole develops (no sunlight, little precipitation).

When IMF BY is positive, the rapid anti-sunward plasma transport on the dawn side of the polar cap is

opposed by co-rotation. The corresponding tongue or plume is thus weaker and shorter and tends to lie in the central polar cap, rather than the dawn side. The co-rotation velocity is not opposed to convection in the central polar cap region. Co-rotation also aids the

otherwise weak anti-sunward transport on the dusk side of the polar cap, another effect which broadens the width of the anti-sunward plasma tongue or

plume. The effect of combined convection and co-rotation

for IMF BY negative is sufficiently strong that for the same precipitation, the dusk aurora1 oval plasma densities are roughly double those for IMF BY posi-

tive. This is entirely due to the shorter time required to transport plasma from the dayside cusp through the polar cap and into the dusk aurora1 oval for IMF BY negative.

There are relatively few changes of the sub-aurora1 trough which are produced by the response to IMF

BY. For IMF BY positive, the dusk trough is somewhat broader in latitudinal extent than for BY nega-

tive, due to the stronger sunward convection in the

dusk oval, but the effect is relatively minor and must be dependent on the coincidence (or non-coincidence) of convection and precipitation boundaries.

The plasma density structures and values within the winter aurora1 oval do not appear to be strongly dependent on IMF BY. In the general sense, there is rather little enhancement of F-region plasma density associated with the aurora1 oval, although a range of characteristic signatures in ion and electron tem-

perature, as well as changes of composition, are produced.

In the summer F-region polar cap, the differences

are quite dramatic. The deep troughs forming in regions where there is strong photoionisation and some particle precipitation are caused by the com-

positional response to strong heating: the stronger the ion convection, the stronger the heating of ions and, generally, the stronger the heating of the neutral atmosphere. The rapid convective upwelling, combined with horizontal advection causes the very large increase in F-region mean molecular mass, which then causes the deep plasma troughs.

8. IMF Y COMPONENT VARIATIONS--E-REGION

8.1. Winds

Seasonal changes of E-region winds are relatively small, however the response to IMF/plasma con-

vection changes is very marked. Obviously this

response is also modulated by the precipitation, which

causes the large increases of ion drag coupling via enhanced E-region plasma density enhancements.

For IMF BY positive, there is a large clockwise swirl, sunward around the dusk aurora1 oval and anti- sunward over the dawn side of the polar cap. The peak wind speeds in the dusk aurora1 oval reach 20& 250 m s- ‘, which is quite consistent with observations. There is a relatively weak sunward wind flow in the dawn aurora1 oval (less than 50 m s- ‘), combined with a significant equatorward component of how

(So-100 m s-‘). For IMF BY negative, the sunward flow in the dusk

aurora1 oval is still strong, except that it does not extend to earlier than about 14 LST. The anti-sunward flow over the polar cap is more generally spread over

the entire width of the polar cap, rather than being confined to the dusk side, as tends to happen in the F-region (see earlier discussion). The sunward winds of the dawn aurora1 oval, for IMF BY negative are significantly increased to nearly 100 m sP ‘, and there

is a generally more coherent flow, following the ion

convection of the dawn aurora1 oval. The asymmetry of the wind response in the dusk

and dawn parts of the aurora1 oval is again related

to the conditions required for resonance, discussed earlier and by FULLER-R• WELL and REES (1984).

8.2. Plasma density

Related to the rapid recombination rates typical of

molecular ions at E-region altitudes there are few significant features of the E-region plasma distribution which can be attributed to convection and thus to IMF BY changes. This would not be the situation for longer lived ions, such as the metallic species found in sporadic-E layers. However, with the excep-

tion of such localised and probably ephemeral

features, the present simulations indicate that primarily the E-region plasma distribution is controlled by the combination of solar photoionisation and ener- getic electron precipitation. The plasma density enhancements are a critical factor in producing observed winds and conductivity enhancements.

The actual values of ion drag wind acceleration terms, ionospheric conductivities, electric currents and the implicit distribution of field aligned electric currents, which power the entire high latitude system, are therefore determined primarily by precipitation : typical peak E-region ion density values within the moderately disturbed aurora1 oval are a factor of 3 greater than those of the surrounding sunlit, but non- auroral, ionosphere in summer time. In winter time

92x D. REES et nl.

the corresponding enhancement factor due to electron

precipitation is between a factor of 10 and 100 times

above the non-sunlit E-region ionosphere.

8.3. Zonosphere-thermosphere coupling

As has been just demonstrated, the entire nature of ion-neutral coupling within the polar regions, particularly in the winter hemisphere, is dominated by the plasma density distribution. Increased or decreased precipitation will strongly modulate all of the major features of ionosphere/thermosphere coupling : ion drag wind acceleration, horizontal and field aligned ionospheric currents.

The aurora1 oval wind values which are reported here are consistent in magnitude and direction with observational data. This is not the case when simu-

lations using non self-consistent codes are used. We have presented steady state results and we

should note that the necessity for self-consistency is considerably greater for strongly time dependent simulations. For similar reasons, the importance of self-consistency is paramount for simulations of geomagnetic activity levels which are higher than those reported here.

The interactions between the polar ionosphere and the magnetosphere have not been considered in these simulations. We can note that there is considerable

modulation of conductivity and winds and therefore of electric currents and induced ‘back-EMF’. These modulations are significant, with features corresponding to both seasonal and IMF responses.

SUMMARY

Data is presented from a series of four numerical simulations of the coupled thermosphere and ionosphere. The simulations are time dependent, but are diurnally reproducible for stable solar and geomagnetic conditions, with Universal Time being the only variable. The data from high latitudes is shown, at high (300 km) and low (125 km) altitudes, to illustrate the response of the coupled system to changes in the orientation of the IMF BY component at both the summer and winter solstices. Each simulation is for conditions of high solar activity (F,,,, cm = 185) and moderate geomagnetic activity and the data are presented for 1800 UT.

In the winter F-region the plasma distribution is strongly controlled by transport. As the polar cap (including its convection electric field) rotates into sunlight for a period of about 6 h centred on 1800 UT, solar produced ionization is ‘picked-up’ and carried over the polar cap through the cusp. For BY

negative a tongue of plasma extends over the dusk

side of the polar cap, the convection electric field being

assisted by the natural tendency of co-rotation. The

lack of transport to the dawn side of the polar cap,

where there is no sunlight and negligible particle ion-

ization, creates a significant polar plasma hole. Plasma carried over the dusk sector polar cap con-

tinues into the dusk aurora1 oval, where number densities exceed those in the dawn sector aurora1 oval by at least a factor of two. For BY positive, the tendency of the convection electric field is to transport the plasma to the dawn side of the polar cap. The opposing influence of co-rotation in this case restricts the flow, resulting in a much smaller tongue over the central polar cap. Aurora1 oval ion densities in this case remain symmetric. There is therefore a significant enhancement (factor of 2) of dusk amoral oval F- region plasma density when BY is negative, compared

with when IMF BY is positive. For both winter cases (BY positive and BY nega-

tive), a sub-aurora1 trough encircles the nightside for the entire UT day. For some 18 h of the UT day, the trough extends onto the dayside as well, isolating the entire polar and aurora1 regions.

There is also a characteristic neutral wind response in the upper thermosphere to the changing direction of the IMF BY. For BY negative, an enhanced anti- sunward wind jet (600 m s- ‘) occurs on the dusk side of the polar cap, and modest (200 m s- ‘) sunward winds are generated in the dawn aurora1 oval. For BY positive, the region of peak anti-sunward winds moves to the dawn side of the polar cap and the region of sunward winds in the dawn aurora1 oval is virtually eliminated. In both cases, however, strong (40& 600 m s- ‘) sunward winds occur in the dusk aurora1 oval. The peak dusk aurora1 oval sunward winds are somewhat stronger when BY is negative, due to the higher ion densities resulting from enhanced cross polar cap plasma convection from the dayside, coun- teracting the slightly weaker sunward plasma convection.

In summer the vertical plasma distributions are flatter, are generally lower in concentration and the sub-aurora1 trough is lacking, at least in the 1800 UT period presented here. It is probably only in the southern summer that, for a short part of the UT day, a limited nightside sub-aurora1 trough can appear due to plasma stagnation. Transport is much less important in defining the plasma distribution in the summer polar region. At the level of geomagnetic activity used for these simulations, the summer time plasma distribution is controlled by the peaks and troughs in neutral mean molecular mass. Plasma minima cor- relate strongly with peaks in the mean molecular mass,


a strong contrast with the situation that peaks in ion density occur at almost identical geomagnetic locations in winter. The generally higher mean molecular mass (i.e. enhanced molecular species) in the summer F-region is driven by the global seasonal neutral circulation pattern. This is mainly caused by hem- ispherically assymetric solar insolation, which is then enhanced by rather stronger Joule heating in the summer polar region. At high latitudes, the distribution of mean molecular mass is modulated by maxima and minima of upwelling due to Joule heating. The characteristic BY dependent composition distribution is finally imposed by the transport effect of the horizontal neutral wind. The polar cap and aurora1 oval neutral wind patterns maintain their BY dependence in summer and these change very little with season.

In the E-region, at this level of geomagnetic activity, the ion density in the aurora1 oval is dominated by the distribution of aurora1 (electron) precipitation, which for these simulations has no BY or seasonal dependence. Small changes do occur due to the seasonal change in solar ionization. Outside the aurora1 oval increased solar illumination in summer generates ion densities exceeding those in winter by about a factor of three. Within the polar cap, the summertime E-region plasma densities exceed those of the equivalent winter region by a factor of about 34, again due to solar photoionisation.

The neutral wind in the lower thermosphere main- tains the signature of the IMF BY that was the hall- mark of the F-region winds. The E-region values are lower than those of the F-region by a factor of about 2-3 (pressure level 7, 125 km, rather than 12, 320 km). This is partly due to reduced ion drift velocities as a result of increased ion-neutral collisions and partly due to the increased inertia of the denser neutral gas compared with higher levels. There is a noticeable, but not dominant, seasonal variation of E-region winds due to the increased summertime ion density by the background solar illumination. This is particularly important in polar regions where there are significant ion convection velocities, but where the local electron precipitation is low.

AcknowledgementsWe are indebted to Dr FRED RICH for provision of the Heppner and Maynard polar electric fields in the form of harmonic coefficients. We would also like to express our particular thanks to JOHN HARMER, HILARY HUGHFS and ISRAEL PERLA for their assistance in preparing, running and processing the computer simulations using the UCL/Sheffield coupled ionospheric/thermospheric model. Computer time was made available by the University of London Computer Center (CRAY I-S), and the CRAY- XMP-48 at the Rutherford Appleton Laboratory. The research was supported by grants from the U.K. Science and Engineering Research Council, and from the European Office of Aerospace Research and Development (AFOSR- 86-341).

ALLEN B. T., BAILEY G. J. and MOFFETT R. J. COLE K. D. COLE K. D. CHN Y. T. DICKINSON R. E., RIDLEY E. C. and ROBLE R. G. DICKINSON R. E., RIDLEY E. C. and ROBLE R. G. FULLER-R• WELL T. J. and REEs D. FULLER-R• WELL T. J. and REES D. FULLER-R• WELL T. J. and REES D. FULLER-R• WELL T. J., REES D., QUEGAN S., BAILEY

G. J. and MOFFETT R. J. FULLER-R• ~ELL T. J., REES D., QUEGAN S.,

MOFFETT R. J. and BAILEY G. J.

FULLER-R• WELL T. J., QUEGAN S., REES D., MOFFETT R. J. and BAILEY G. J.

FULLER-R• ~ELL T. J., REES D., QUEGAN S., MOFFE~T R. J. and BAILEY G. J.

HAREL M., WOLF R. A., RElFF P. H., SPIRO R. W., BURKE W. J., RICH F. J. and SMIDDY M.

HAYS P. B., KILLEEN T. L., SPENCER N. W., WHARTON L. E., ROBLE R. G., EMERY B. A., FULLER-R• WELL T. J., REE!S D., FRANK L. A. and CRAVEN J. D.

HEDIN A. E. HEDIN A. E.

REFERENCES

1986 Annls Geophys 4A, 97. 1962 Aust. J. Phys. 15,223. 1971 Planet. Space Sci. 19, 59. 1975 J. atmos. terr. Phys. 31, 1563. 1981 J. geophys. Res. 86, 1499. 1984 J. atmos. terr. Phys. 41,205. 1980 J. atmos. Sci. 37,2545. 1983 Planet. Space Sci. 31, 1209. 1984 Planet. Space Sci. 32, 69. 1984 Planet. Space Sci. 32,469.

1987a Symposium on Quantitative Modeling of Magneto- sphere-ionosphere Coupling Processes, p. 20. Kyoto Sangyo University.

1987b J. geophys. Res. 92,7744.

1988 Pure Appl. Geophys. 121, 189.

1981 J. geophys. Res. 86,221l.

1984 J. geophys. Rex 89,5547.

1983 J. geophys Res. 88, 10170. 1987 J. geophys. Res. 92, 4649.

930 D. RED et al.

HEPPNER J.P. HEPPNER J. P. and MAYNARD N. C. KILLEEN T. L., HAYS P. B., SPENCER N. W. and

WHARTON L. E. MA~R H. G. and VOLLAND H. MCC~RMAC F. G. and SMITH R. W. PEREIRA E., KELLEY M. C., REES D., MIKKEL~ON

I. S., JORGENSEN T. S. and FULLER-R• WELL T. J. QUEGAN S., BAILEY G. J., MOF~ R. J., HEELIS

R. A., FULLER-R• WELL T. J., REES D. and SPIRO R. W.

QUEGAN S., BAILEY G. J., MOFFETT R. J. and WILKINSON L. C.

REES D. REES D.

REEs D., FULLER-R• WELL T. J., GORDON R., KILLEEN T. L., HAYS P. B., WHARTON L. E. and SPENCER N. W.

1983

REES D., GORDON R., FULLER-R• WELL T. J., SMITH M. F., CARIGNAN G. R., KILLEEN T. L., HAYS P. B. and SPENCER N. W.

1985 Planet. Space. Sci. 33, 617.

REES D., FULLER-R• WELL T. J., GORDON R., SMITH M. F., HEPPNER J. P., MAYNARD N. C., SPENCER N. W., WHARTON L. E., HAYS P. B. and KILLEEN T. L.

1986 Planet. Space Sci. 34, 1.

REES D. and FULLER-R• WELL T. J. 1987a REEs D. and FULLER-R• WELL T. J. 1987b ROBLE R. G., DICKINSON R. E. and RIDLEY E. C. 1982 SOJKA J. J., RAINS W. J. and SCHLJNK R. W. 1979 SOJKA J. J., RAITT W. J. and SCHLJNK R. W. 1981 SOJKA J. J., RAINS W. J. and SC~NK R. W. 1981 SOJKA J. J. and SCHUNK R. W. 1983 SOJKA J. J., RAITT W. J., SCHUNK R. W., PARISH 1985

J. L. and RICH F. J. WATKINS B. J. 1978

1971 J. geophys. Rex 82, 1115. 1987 J. geophys. Res. 92,4467. 1983 Adv. Space Res. 2, 133.

1972 Planef. Space Sci. 20, 379. 1984 Geophys. Rex L&t. 11, 935. 1980 J. geophys. Res. 85,2935.

1982 J. atmos. terr. Phys. 44,619.

1986 J. atmos. terr. Phys. 48, 25.

1971 1973

J. Br. interplanet. Sot. 24, 233. Physics and Chemistry of Upper Atmospheres

(MCCORMAC B. M., Ed.), p 22. Reidel, Dordrecht. Planet. Space Sci. 31, 1299.

Adt?. Space Res. 7, 259. CIRA 87, chapter 2. Pergamon Press, Oxford. J. geophys. Res. 87, 1599. J. geojhys. Res. 84,5943. J. aeovhvs. Res. 86.609. J. geo>h$s. Res. 86; 609. J. geophys. Res. 88,2112. Planet. Space. Sci. 33, 1375.

Planet. Space Sci. 26, 559.

simulations of the seasonal variations of the thermosphere and ionosphere using a coupled,...

Documents