A multipoint study of a substorm occurring on 7 December, 1992, and its theoretical implications

. On 7 December 1992, a moderate substorm was observed by a variety of satellites and ground-based instruments. Ionospheric ﬂows were monitored near dusk by the Goose Bay HF radar and near midnight by the EISCAT radar. The observed ﬂows are compared here with magnetometer observations by the IMAGE array in Scandinavia and the two Greenland chains, the auroral distribution observed by Freja and the substorm cycle observations by the SABRE radar, the SAMNET mag-netometer array and LANL geosynchronous satellites. Data from Galileo Earth-encounter II are used to estimate the IMF B z component. The data presented show that the substorm onset electrojet at midnight was conﬁned to closed ﬁeld lines equatorward of the pre-existing convection reversal boundaries observed in the dusk and midnight regions. No evidence of substantial closure of open ﬂux was detected following this substorm onset. Indeed the convection reversal boundary on the duskside continued to expand equatorward after onset due to the continued presence of strong southward IMF, such that growth and expansion phase features were simultaneously present. Clear indications of closure of open ﬂux were not observed until a subsequent substorm intensiﬁcation 25 min after the initial onset. After this time, the substorm auroral bulge in the nightside hours propagated well poleward of the pre-existing convection reversal boundary, and strong ﬂow perturbations were observed by the Goose Bay radar, indicative of ﬂows driven by reconnection in the tail.


Introduction
The onset of a magnetospheric substorm, marking the transition from the growth to the expansion phase, is de®ned as the beginning of the explosive brightening and expansion of the aurora.As the Akasofu (1964) model had only two substorm phases, expansion and recovery, expansion, onset was the ®rst event during a substorm.With the addition of the growth phase (McPherron, 1970) it was recognised that the substorm cycle began not at onset, but typically 30±60 min earlier when the rate of dayside ¯ux production increased (usually due to a southward turning of the IMF).The term onset was retained for the commencement of the auroral and current features described by Akasofu (1964), and subsequently became associated with the commencement of the reconnection of open ¯ux in the tail, which closes the open ¯ux accumulated in the growth phase.The main assumptions of the near-Earth neutral line (NENL) substorm model were that the onset of near-Earth tail reconnection caused the onset of the expansion phase, and that this reconnection rate is suciently rapid that the plasmoid is pinched o within a few minutes such that open lobe ¯ux then begins to be closed (Hones, 1979).More recent evidence, however, suggests that there may be a signi®cant delay between the expansion phase onset and the closure of open ¯ux.Although Nagai et al. (1998) have shown that reconnection generally commences between 20±30 R E downtail within a few minutes of substorm onset, it appears nevertheless that a substantial interval may elapse before the resulting plasmoid is pinched o and open lobe ¯ux starts to be reconnected.For example, Maynard et al. (1997) using an extensive dataset including Geotail data, have inferred that open ¯ux closure did not begin until 15 min after onset in one event, and 40 min after onset in another.In a study of ionospheric ¯ows using the AMIE technique, Taylor et al. (1996) also inferred that open ¯ux typically did not start to decrease until 20 min after substorm onset.Correspondingly, the expansion of the auroral bulge is observed to originate on the most equatorward discrete arc, which is usually located well equatorward of the open/closed ®eld line boundary, as implied by Akasofu's (1964) original picture (see also Galperin and Feldstein, 1991), and subsequently con®rmed by Viking imager data (Murphree and Cogger, 1992;Elphinstone and Hearn, 1992;Cogger and Elphinstone, 1992).The latter show that the auroral bulge often reaches the open/closed ®eld line boundary towards the end of the expansion phase (see also, for example, Lopez et al., 1992), at which time the poleward edge often brightens and becomes more active whilst the aurora in the middle of the bulge often fade, leading to a ``double oval'' con®guration.
In this work we will analyse a substorm which occurred during the second Galileo Earth-encounter and will show that in this case open ¯ux closure was delayed relative to substorm onset by more than 25 min.This will be inferred from a dataset combining ground magnetic records, radar observations of ionospheric ¯ow and satellite measurements.

Instrumentation
In this study we will combine together data from a network of ground-based instruments with satellite observations to study the evolution of a substorm on 7 December, 1992.Data from the IMAGE, Greenland and SAMNET magnetometer chains and Los Alamos National Laboratory (LANL) spacecraft in geosynchronous orbit are used to detect the substorm onset and subsequent intensi®cation.The location and motion of the convection reversal boundary were monitored by the EISCAT incoherent scatter radar, Goose Bay HF radar, SABRE VHF radar and the Defense Meteorological Satellite Program (DMSP) satellites.Unfortunately, there are no interplanetary magnetic ®eld (IMF) data for the interval studied owing to a gap in the IMP-8 data record.However, Galileo was located in the magnetosheath during its second Earth-encounter and the data obtained can be used to deduce the orientation of the IMF. Figure 1 shows the positions of the ground stations used in this study together with the footprints of the LANL satellites.This ®gure illustrates the broad coverage of the observations, extending over 6 h of magnetic local time (MLT) and from sub-auroral latitudes to within the polar cap.A brief description of the data sources is given.
1. Data from the EISCAT radar were obtained using the SP-UK-CONV experiment, in which the VHF radar was used in a split-beam mode.One pair of panels of the antenna were phased such that the beam pointed along an azimuth 344.8°to the east of geographic north (west beam) and the beam of the remaining pair to 359.8°( east beam), both at an elevation of 30°.In this mode, therefore, the VHF system provides two independent radars operating with two dierent ®xed beam directions.The data are taken simultaneously at each azimuth with 10 s resolution.Echoes are received from range gates 65.3 km in length along the beam passing through the F-region ionosphere, covering invariant latitudes 71°±78°.The data are analysed using an ``Lshell ®tting'' technique to estimate the ionospheric ¯ow magnitude and direction from line-of-sight velocities, assuming that the ¯ow is the same along the L-shells between the two azimuths and that the magnetic ®eld parallel ¯ow is negligible.The data have been analysed for the 10 s preintegration periods with no post integration.The MLT of the CONV ®eld-of-view is approximately UT + 2.25 h for azimuth 344.8°andUT + 2.75 h for 359.8°.
2. The Goose Bay HF coherent backscatter radar has been described by Greenwald et al. (1985) and now forms part of the SuperDARN network (Greenwald et al., 1996).The radar measures the power and line-ofsight Doppler spectral characteristics of signals backscattered from irregularities in the polar F-region.The Goose Bay radar ®eld-of-view covers the range of invariant latitudes 65°±85°and spans 4 h of local time.The centre of the ®eld-of-view is at a MLT given approximately by UT A 3 h.Data from the 16 beam directions are recorded over successive 6 s integration periods, such that the ®eld-of-view is scanned every 96 s.Flow vectors are derived where sucient backscatter intensity is received by the L-shell ®tting technique (Ruohoniemi et al., 1989;Freeman et al., 1991).
3. At the time of this study the SABRE radar comprised a monostatic VHF radar located at Wick, with a ®eld-of-view centred on an MLT of $ UT + 1 h and covering an invariant latitude range of 61°±65°.The temporal resolution was typically 20 s.The signal received by a VHF coherent radar has been scattered by E-region irregularities which have a drift speed related to the convection ¯ow speed.A more complete description of the SABRE auroral backscatter system is given by Nielsen et al. (1983).
4. SAMNET is an array comprising seven triaxial ¯uxgate magnetometers which cover midlatitudes in western Europe (Yeoman et al., 1990).These data are used to identify the occurrence of Pi2 pulsations and thus the onset of substorms.
5. The Greenland magnetometer network consists of two latitudinal chains of magnetometers located along the east and west coasts of Greenland (Wilhjelm and Friis-Christensen, 1976).The east coast chain stretches from 70°to 81°in geomagnetic latitude and 56°to 111°i n geomagnetic longitude.The corresponding ranges for the west coast chain are from 68°to 86°in latitude and 36°to 47°in longitude.Each station has a three-component ¯uxgate magnetometer which operates at 20 s resolution.The perturbations from the quiet-day diurnal variations are given here in terms of the three component directions, H, E and Z, where H is positive towards the north along the local magnetic meridian, E is positive towards the east perpendicular to the local meridian, and Z is positive in the vertically downwards direction.
6.The IMAGE magnetometer network is located in Scandinavia with stations from mid-latitudes up to Svalbard (LuÈ hr et al., 1998).The entire network extends latitudinally from 56°to 75°and longitudinally from 106°to 119°(geomagnetic coordinates).Each of the instruments in the network is a triaxial ¯uxgate magnetometer similar to those used in the Greenland chain, but the ®eld perturbation components given here are the three orthogonal components, X, horizontal geographic northward, Y, horizontal geographic eastward, and Z, vertically downwards.
7. The Los Alamos particle analyser instruments CPA and SOPA obtained the geosynchronous orbit energetic particle data presented here.The spacecraft used in this study are 1984-129 at a local time of UT + 06:54 h and 1990-095 at UT + 00:31 h.The CPA instruments on 1984-129 measure electrons in the energy range 30±300 keV and ions with energies 72± 573 keV.The SOPA instruments aboard 1990-095 measure electrons in the energy range 50±315 keV and ions with energies 113±670 keV.A more detailed description of the Los Alamos particle instruments is given by Higbie et al. (1978) and Belian et al. (1992).In this study we use the data to observe particle injections which identify onset (Baker et al., 1978).
8. During the interval reviewed here, magnetic ®eld data from the Galileo spacecraft, at the time of its second Earth encounter, were used to monitor the magnetic ®eld in the Earth's magnetosheath.Galileo was located at GSM coordinates (X, Y, Z) $ (A70, 40, 26 R E ), suciently far from the centre of the tail that the magnetic ®eld data can be used as a reasonable proxy for the IMF.The Galileo magnetometers are described in Kivelson et al. (1992).9.The DMSP spacecraft are a series of polar orbiting satellites designed to observe the weather on Earth and to monitor the near-Earth space environment at 840 km.We have used ionospheric velocity measurements obtained by the F10 and F11 spacecraft to derive the potential along the spacecraft track and hence the convection reversal boundaries.The orbital period of the satellites is $103 min and each of the polar passes (magnetic latitudes exceeding 50°) takes 18±20 min.
10. FREJA, a Swedish and German scienti®c satellite, was launched on 6 October, 1992.Freja orbits at an altitude ranging between about 600 and 1750 km with an orbital inclination of 63°.The payload includes an auroral imager comprising two cameras.The one presented here monitors the molecular nitrogen Lyman-Birge-Hop®eld (LBH) emissions and has a passband from 134 to 180 nm (Murphree et al., 1994).The image contained here is a composite of ®ve separate images taken at two minute intervals, where each of the separate images has an overall ®eld-of-view of 29.4°by 109.2°.

Observations
Data from the Galileo Earth-encounter II for 21 UT on 7 December to 01 UT on 8 December, 1992, are plotted in Fig. 2. When these data were taken, the satellite was located in the magnetosheath $72 R E downtail.By taking a value of 450 km s A1 for the solar wind speed (from the Geotail record), a delay of $20 min can be estimated as the time taken for eects at the nose of the magnetosphere to be detected by Galileo.For the ®rst hour of this interval, the magnetic ®eld is mainly northward pointing with B y negative and B x positive.Shortly after 2200 UT, a southward turning occurred in the B z component ($2140 UT at the subsolar magne-topause).Subsequently B z decreased slowly until near 2235 UT ($2215 UT at the subsolar magnetopause) when it suddenly dropped to A20 nT.Prior to this time the B y component was strongly negative (around A25 nT) while B x was near zero.After 2235 UT B y declined sharply towards zero such that the ®eld was directed nearly due south in the GSM frame for about 20 min.At 2255 UT ($2235 UT at the subsolar magnetopause), B z suddenly turned northward and remained so for the following 25 min, except when it brie¯y returned to zero at $2257 UT.From 2315 to 0015 UT the B z ®eld was variable but typically weakly negative.During the whole of the period after $2255 UT, B x was positive and B y strongly negative.
For the remainder of this study these times will be quoted in terms of the eect at the subsolar magnetopause and not detection by Galileo.
Turning now to the corresponding ground-based data, we shall now examine the magnetometer and radar During the time period reviewed, Galileo was located in the magnetosheath at GSM coordinates (X, Y, Z $ A70, 40, 26 R E ), suciently far from the centre of the tail that the magnetic ®eld data can be used to indicate the strength and orientation of the IMF.The bottom panel indicates the merging index.A value of 1 denotes that the conditions are favourable for low-latitude dayside reconnection, i.e. the IMF B z is southward or the magnetic shear angle between the subsolar magnetospheric ®eld and the IMF was only $70°w hen the IMF B z was northward observations.The backscatter intensity detected by the SABRE radar between 21 and 23 UT is plotted in Fig. 3 as a function of range and UT.It shows an intensi®cation of the backscatter at 2130 UT, which rapidly expanded equatorward starting at 2200 UT.We infer that this expansion was a response to the southward turning of the IMF with a delay of 20 min, and interpret the equatorward motion as a growth phase signature.A similar phenomenon was reported by Lester et al. (1993) during the SUNDIAL campaign when the onset of backscatter was detected before any substorm onset signatures and was thus assumed to be associated with the southward turning of the IMF.The apparent bifurcation of the backscatter band which appears at around 2210 UT is probably due to an instrumental eect known as Lloyd's mirror (Mattin and Jones, 1985).The remainder of this ®gure will be described later, after we have introduced the adjacent magnetic observations at the IMAGE chain.
The X-and Z-components of the magnetic ®eld detected by the IMAGE magnetometers are plotted in Fig. 4.These data show that there was very little magnetic activity during the initial period of modest southward IMF B z , 2140±2215 UT (Fig. 2), including the nightside response to the growth phase (starting at 2200 UT) described already.The onset of the ®rst bay in the X-component can be seen at 2215 UT, indicated by the ®rst dash-dot line in Fig. 4a.This occurred around the time of the large increase in southward IMF B z to A20 nT at the subsolar magnetopause (Fig. 2) and the onset of the Pi2 s detected by SAMNET, shown in Fig. 5.Both the X-and Z-components (Fig. 4a, b) of the magnetic de¯ections seen by the IMAGE array provide evidence that the electrojet was approximately centred at station PEL (62.4°latitude) at this time; as well as the peak X component being detected at PEL, there was a positive Z de¯ection in the stations north of this station and a negative de¯ection to the south.Shortly after the expansion phase onset, the poleward edge of the backscatter band at SABRE (Fig. 3) contracted equatorward, starting at about 2220 UT.This may have been an eect of the substorm electrojet entering the radar ®eld of view to the west of IMAGE (Fig. 1) where a reduction in ¯ow speed inhibits the generation of E-region irregularities and hence radar backscatter (Inhester et al., 1981).Since the SABRE ®eld of view was located just prior to midnight at this time, and the reduction of the backscatter encroached from the east, the implication is that the onset began in the post-midnight sector in the vicinity of the IMAGE chain and propagated westward to SABRE.
Returning now to the IMAGE data, it can be seen that the initial disturbance intensi®ed over an interval of $15 min before starting to decay.This was followed by an intensi®cation at $2240 UT (second dash-dot line in Fig. 4a) centred around MAS (64.9°), which coincided approximately with the northward turning of the IMF at the nose, as detected by Galileo (Fig. 2).An associated Pi2 enhancement was also seen in the D component of the SAMNET data (Fig. 5).This intensi®cation resulted in a poleward expansion of the substorm disturbed region, which reached the Svalbard magnetometers (BJO (70.1°),HOP (71.4°),HOR (72.8°) and NAL (74.9°)) after $2250 UT.At about this time, the magnetic disturbance at the latitude of the original intensi®cation, centred at MAS, started to recover, while those poleward and equatorward were intensifying so that a ``double oval'' con®guration in the magnetic perturbations had formed by $2300 UT.Recovery of the poleward part of the magnetic disturbance was in progress by 2340 UT and of the equatorward currents by 2350 UT.
H-component data from the Greenland magnetometer chain are plotted in Fig. 6.No perturbations were observed at any of the Greenland chain stations in response to the substorm onset detected by IMAGE at 2215 UT.This shows that the initial onset electrojet did not extend westward to the 20 MLT sector, monitored by the lower latitude Greenland stations (NAQ and FHB).The east coast stations are located at latitudes too high to observe the onset (since the perturbations did not reach BJO at 70°at IMAGE), but examination of data from the Leirvogur station on Iceland (not shown) at 66°invariant latitude and MLT of UT-0:30, however, also did not observe any clear signatures associated with the initial onset, showing that the electrojet was con®ned eastward of about 21:45 MLT.The onset of negative bays was seen after $2245 UT in the lowest latitude stations of the east coast Greenland chain (AMK $70°, and SCO $72°) in the pre-midnight sector, about 5 minutes after the intensi®cation detected by IMAGE as discussed (Fig. 4).The disturbance subsequently propagated poleward similarly to the IMAGE chain, reaching DNB ($75°) at 2305 UT and DMH ($77°) shortly thereafter.The westward electrojet did not reach station NRD (81°), however.The magnetometer stations on the west coast ($21±22 MLT) observed no signi®cant perturbations before $2300 UT when weak negative bays appeared at latitudes between 70 and 75°.This shows that the substorm westward electrojet did not expand into this magnetic local time sector until late in the expansion phase development.
The relationship between the magnetic observations described already and the substorm disturbed region is  7, where the location of the magnetometer stations are superposed, together with the ®eldsof-view of the various radars used in this study, on an auroral image from the Freja spacecraft.The image is a composite of consecutive UV images obtained in the period 2329±2339 UT.This period corresponds to the interval just prior to the onset of recovery of the disturbance at the IMAGE meridian when the double current structure was still well developed but starting to decline (Fig. 4).The images cover a region from the post-noon oval on the left, through dusk, and towards midnight on the right, though not quite reaching the meridian of the IMAGE chain in the early postmidnight hours ($2 MLT).It can be seen that the nightside auroral distribution was very broad, extending from an equatorward boundary at a magnetic latitude of $60°poleward to $73°, this spanning essentially the same latitude range as the two-component current system observed on the IMAGE meridian just to its east.It can also be seen that the auroral bulge does not quite reach the west coast stations of the Greenland magnetometer chain, thus explaining the lack of major magnetic perturbations on this meridian.It does, however, encompass the equatorward part of the eastern chain, although not extending to the poleward stations.
The LANL geosynchronous satellites detected particle injections around the time of the onset signatures seen by IMAGE and SAMNET at 2215 UT (Figs. 4 and  5).At $2218 UT, spacecraft 1990-095 (22:51 MLT) detected a dispersionless proton injection (Fig. 8b), followed by a weak dispersed electron injection (Fig. 8a) which implies that the satellite was located just to the west of the injection region.The positioning of this spacecraft with respect to the injection region is in good agreement with the fact that the Iceland and Greenland west magnetometer chains were too far west to observe the perturbations detected by IMAGE associated with the substorm onset at 2215 UT.We saw that at $2240 UT there was an intensi®cation of magnetic perturbations at IMAGE, and a Pi2 at SAMNET.The geosynchronous satellites also detected particle injections around this time.Spacecraft 1990-095 detected a rise in the proton ¯ux at $2242 UT and electron ¯ux at $2248 UT (23:19 MLT).Both the electron and proton injections appeared to be dispersionless (Fig. 8a,  b).This implies that the satellite was located within the MLT of the injection region at this time.Spacecraft 1984-129 located at 05:44 MLT (data not shown) also detected dispersed electron and proton injections two minutes later at 2250 UT.The dispersion curves for these injections appeared to be associated with the injections detected by 1990-095 at $2248 UT.
Summarising the data discussed in Figs.2±8, we can deduce that a cycle of substorm activity occurred between $2140 UT and $0030 UT.The start of the growth phase was determined to be at $2140 UT when we infer that the B z component of the IMF turned south at the subsolar magnetopause (Galileo data in Fig. 2), such that subsequently the SABRE radar (Fig. 3) detected a band of increased backscatter which expanded equatorward, starting at 2200 UT.Expansion phase onset signatures were seen by the IMAGE and SAM-NET magnetometer chains at $2215 UT with the LANL satellite 1990-095 detecting a dispersionless ion injection slightly later at 2218 UT.This onset coincided with the IMF B z component dropping to A20 nT.The IMF then remained negative until $2235 UT when it became strongly positive.Shortly after this at 2240 UT, there was intensi®cation in the magnetic perturbations and Pi2 s observed by IMAGE and SAMNET in the post-midnight sector, and a little later by Greenland east and Iceland in the pre-midnight sector.Particle injection signatures were also detected by the LANL satellites 1990-095 and 1984-124 at this time.The recovery phase commenced around 2340 UT.
We now turn to examine the ¯ow data from the Goose Bay HF radar, located in the dusk sector to the west of the substorm disturbed region (see Fig. 7).A summary of the ¯ow data deduced from this radar is presented in Fig. 9, where we show the vector ¯ow We have thus examined the vector velocities over the entire ®eld-of-view of the radar (15 beams) and used the method described to ®nd the invariant latitude of the CRB for the interval 2130±00 UT (1830±2100 MLT) as shown in Fig. 10.The lack of scatter, or of clear ¯ow often makes identi®cation of the CRB latitude dicult and the vertical error bars indicate these uncertainties.Following an initial equatorward motion between 2130 and 2135 UT, the CRB remained at $70°until $2200 UT (Fig. 10).An equatorward motion was then evident continuing across the onset of the substorm at 2215 UT and accelerating prior to the intensi®cation at 2240 UT.This behaviour closely mirrors the behaviour of the backscatter band seen by the SABRE radar (Fig. 3) which showed a similar equatorward motion between 2200 and at least 2235 UT.From this we infer that the region of open magnetic ¯ux continued to expand for $25 min after substorm onset, presumably due to the fact that the IMF B z was $A20 nT during this period, such that the rate of reconnection on the dayside considerably exceeded that on the nightside.During this interval, therefore, typical signatures of both growth and expansion phases were simultaneously present.Following the substorm intensi®cation at $2240 UT (and northward turning of the IMF) the CRB within the Goose Bay ®eld-of-view remained at a roughly constant latitude of $68°until about 2305 UT when it appeared to jump poleward to 72°, where it remained located for the remainder of this interval.Given that there was a substorm intensi®cation at 2240 UT together with the presence of northward IMF since 2235 UT, such that dayside reconnection will have been substantially reduced, it is perhaps surprising that the poleward contraction of the CRB at Goose Bay did not occur until 2305 UT.This point will be discussed later.
Another important feature of the Goose Bay ¯ow data can also be seen in Fig. 9, in that there was a major change in the ¯ow pattern at 2245 UT, from an eastward ¯ow in the polar cap to strong northeast ¯ows.These were observed for three consecutive 96 s scans until the backscatter from the poleward part of the ®eld-of-view disappeared.An example of the velocities deduced over the whole radar ®eld-of-view during this interval is shown in Fig. 11.Comparing Fig. 11 with the Freja image in Fig. 7, shows that the ¯ows in the poleward region of the Goose Bay radar are in fact directed towards the poleward expanding substorm bulge near midnight.It seems reasonable to infer that these ¯ows Fig. 7.A composite of ®ve consecutive UV images taken from the Freja satellite (orbit 828) between 2329±2339 UT on 7 December, 1992.The image shows the auroral oval in the late expansion phase over Canada, Greenland and the north Atlantic.The ®elds-of-view of the Goose Bay, EISCAT and SABRE radars and the position of the SAMNET, IMAGE and Greenland magnetometer chains are overlaid.The 55°line is geomagnetic latitude, and 19 is the magnetic local time, both calculated using the eccentric dipole model (IGRF 1992.937).The northern end of the 19 MLT line is the magnetic pole were associated with the substorm intensi®cation at 2240 UT, and were generated by the dipolarisation in the tail and subsequent closure of open ¯ux (Cowley et al., 1998).
Additional information about the ionospheric ¯ow and the motion of the CRB was provided by data from three passes across the northern polar cap by the DMSP satellites F10 and F11.These are shown in Fig. 12.The UT quoted on the plots corresponds to the time the satellites were closest to the pole during each pass.These were 2133 UT (Fig. 12a), corresponding to the interval just prior to the start of the growth phase, 2249 UT (Fig. 12b), shortly after the substorm intensi®cation, and 2314 UT (Fig. 12c), corresponding to the late expansion phase, prior to the commencement of recovery.The position of the CRBs are listed in Table 1 as determined from the maximum and minimum values of the potential, together with the local time at which the boundary was detected and the UT when the observation was made.There are two passes of satellite F10 in the Northern Hemisphere (Fig. 12a, c) which ¯y through the 21 MLT meridian close to the ®eld-of-view of the Goose Bay radar.The latitudes of the 21 MLT CRB for these orbits are plotted in Fig. 10.Early in the sequence in Fig. 10, at 2130 UT ($18:30 MLT), the CRB seen by Goose Bay is at a similar latitude to that seen at 21:17 MLT by DMSP F10.At 2309 UT F10 observed the CRB at 21:16 MLT at 70.38°in the Northern Hemisphere.The centre of the Goose Bay ®eld-of-view is at 20:09 MLT at this time, i.e. the satellite was at a meridian one hour later in MLT.This therefore, is consistent with the minimum latitude estimate from the Goose Bay radar.
Turning now to the ¯ow data shown in Fig. 12, it can be seen that the ¯ow prior to the start of the growth phase contained surprisingly large ¯ows on the nightside, possibly driven by reconnection in the distant tail.There is also evidence of an anti-clockwise vortex in the polar cap possibly driven by lobe reconnection in the presence of IMF B y negative and B z positive (Fig. 2).At 2249 UT, as shown in Fig. 12b, the IMF had returned to almost similar values but now the dayside ¯ows show a well-developed twin-vortex pattern.The presence of these strong dayside ¯ows implies that they were being driven by nightside reconnection and consequent poleward contraction of the open/closed ®eld line boundary associated with the substorm intensi®cation at 2240 UT.Again strong dayside ¯ows are shown in Fig. 12c, with anti-Sunward ¯ow in the polar cap.Given that the IMF B z is inferred from Fig. 2 to have been weakly negative but with a very strong negative B y at this time, these ¯ows were likely to be driven by dayside reconnection as indicated by the strong eastward ¯ows near noon, together with some possible continuing contribution from the nightside.
Flow vectors near midnight corresponding to the period just prior to the growth phase (2125± 2155 UT = MLT $23:55±00:25) derived from the EI-SCAT radar observations are shown in Fig. 13.These vectors were derived using the ``beam-splitting'' technique described in Sect. 2. Note that to avoid conges- tion, the vectors have been rotated clockwise through 90°so northward ¯ows appear as vectors pointing to the right of the ®gure.At the start of the interval, the dusk cell CRB is apparent as a rotation of the ¯ow from eastward and equatorward in the poleward part of the ®eld-view, to westward and equatorward in the lower latitude region.This reversal took place at about 72.7°i nvariant latitude which is in good agreement with the location of the CRB observed in the dusk sector at this time by both the Goose Bay radar (Fig. 10) and the DMSP satellite F10 (Fig. 12a).After 2140 UT, however, the ¯ows become variable, but predominantly eastward until 2327 UT over the entire ®eld-of-view.The relationship of these ¯ows to the position of the open/closed ®eld line boundary is unclear during the ®rst part of this interval, but at 2250 UT the eects of the substorm bulge were observed expanding poleward across the radar beams, clearly establishing the radar ®eld-of-view as being on closed ®eld lines after that time.Data for this interval are not shown as they will be the subject of a further paper together with the detailed observations of this poleward expansion.Subsequently the dawn CRB propagated equatorward across the ®eldof-view starting at the furthest range gate at 2327 UT and reaching the nearest range gate just before 2339 UT.This established westward ¯ow across the ®eld-of-view.We interpret this motion as the equatorward relaxation of the substorm bulge in the immediate post-midnight sector, corresponding to the onset of substorm recovery.

Discussion
We have discussed an interval of substorm activity using multipoint observations encompassing a broad range of latitude and local time in the nightside sector.The growth phase appears to have been initiated by a southward turning of the IMF which reached the nose of the magnetosphere at 2140 UT as inferred from magnetosheath data from the Galileo spacecraft.The production of open ¯ux lead to an expansion of the polar cap which was observed as an equatorward motion of the CRB in the Goose Bay radar ®eld-ofview and an equatorward expansion of the SABRE backscatter band starting at about 2200 UT.A schematic of the ¯ow con®guration expected at this time is shown in Fig. 14, based on the theoretical models of Cowley and Lockwood (1992)    Although this disturbance may have involved an interval of reconnection in the near-Earth tail, we infer that it did not involve the closure of open ¯ux.Consistent with this, the CRB at Goose Bay and the backscatter band at SABRE continued to expand equatorward after onset; indeed this motion was intensi®ed after onset, presumably associated with the strongly enhanced negative IMF B z that was present during this time.The ¯ow pattern is, therefore, likely to be similar to that shown in Fig. 14a, except that the ¯ow speeds may be enhanced, as seen by the Goose Bay radar shown in Fig. 9.There may also be nightside ¯ow perturbations mainly on closed ®eld lines associated with the dipolarisation of the tail ®eld (Cowley et al., 1998), although we have no direct data to support this.Despite the absence of open ¯ux closure during the initial expansion phase, the later substorm intensi®cation at 2240 UT clearly did involve substantial closure of open ¯ux.We infer that open ¯ux was closed because ®rstly the magnetic disturbance reached high magnetic latitudes up to 75±77°on the IMAGE and Greenland east chain, latitudes that considerably exceeded that of the CRB at both Goose Bay and EISCAT seen prior to substorm expansion.Second, we observed a region of strong poleward ¯ow directed towards the midnight sector in the polar cap in the Goose Bay radar ®eld-ofview.A composite of the radar ¯ow observations is shown in Fig. 15, comprising vectors from both the EISCAT and Goose Bay radars at $2245 UT.A schematic of the ¯ows occurring at this time consistent with these observations is shown in Fig. 14b.These show the presence of enhanced nightside ¯ows driven by tail reconnection consistent with the poleward ¯ows seen in the higher latitude region of the Goose Bay radar in the dusk sector.These strong ¯ows continue  into the dayside sector driven by the contraction of the open/closed ®eld line boundary as observed by the DMSP F11 spacecraft.The schematic also shows a possible anti-clockwise lobe cell vortex which may be driven by the IMF B y negative and B z positive which was present at this time.This pattern provides a possible explanation of the sudden poleward motion of the CRB observed at Goose Bay at 2305 UT.Prior to this time, the observed CRB may have been located in the viscous cell on closed ®eld lines.At 2305 UT the radar could then have moved into the ¯ow cell region driven by nightside reconnection such that the observed CRB would be located much closer to the open/closed ®eld line boundary, as shown in Fig. 14b.A schematic of the recovery phase is shown in Fig. 14c which is related to equatorward relaxation motion of the boundary in the midnight sector as observed by EISCAT and weaker poleward motions elsewhere.In addition, asymmetric ¯ow is also shown as being driven on the dayside during this interval by open ¯ux production associated with the weaker southward IMF with B y negative which was again present during this interval, as indicated by the DMSP data in Fig. 12c.

Conclusions
We have examined observations over a substorm cycle from a variety of data sources located over 6 h of local time in the nightside sector.The observations from the Goose Bay and SABRE radars have enabled us to investigate the behaviour of the dusk convection cell while the EISCAT radar and IMAGE magnetometer chain simultaneously detected activity in the immediate post-midnight sector.The onset and development of the substorm were monitored by the IMAGE and Greenland magnetometer chains, Pi2 data at midlatitudes, and energetic particle data from two geosynchronous satellites.The IMF conditions for this interval were determined by the data from the Galileo satellite in the nightside magnetosheath.There was an initial growth phase of about 30 min during which the IMF B z component was $A5 to 10 nT.The B z component subsequently dropped to A20 nT in concert with the substorm expansion phase onset signatures.IMF B z then remained at this large negative value for another 20 min until it suddenly turned to a positive value of $20 nT.This northward turning occurred shortly before a substorm intensi®cation, and commencement of the CRB contraction.The recovery phase commenced at around 2340 UT.
The most important results that we have found from our study are as follows.Firstly, the substorm onset electrojet was con®ned to closed ®eld lines equatorward of the pre-existing CRBs observed by the Goose Bay and EISCAT radars.No evidence of substantial closure of open ¯ux was observed following this substorm onset.Indeed the CRB in the duskside continued to expand equatorward following onset due to the continued presence of strong southward IMF, such that growth and expansion phase features were simultaneously present.Clear indications of closure of open ¯ux were not observed until after the subsequent intensi®cation at 2240 UT, i.e. more than 25 min after the initial onset.After this time, the substorm auroral bulge in the nightside hours propagated well poleward of the preexisting CRB, and strong ¯ow perturbations were observed by the Goose Bay radar, indicative of ¯ows driven by reconnection in the tail.These data thus provide strong evidence that the closure of open ¯ux in the tail can be delayed signi®cantly relative to the onset of the substorm expansion phase.

Fig. 1 .
Fig. 1.Geomagnetic latitude and longitude of the ground-based stations used.The MLT is indicated in brackets corresponding to 2245 UT, which is one of the principal periods of study.The ®eld-of-view of the Goose Bay radar is labelled GB, the two beams for the EISCAT CP-4-B experiment E and the SABRE beam at

Fig. 2 .
Fig.2.Magnetic data from Galileo Earth-encounter II in GSM co-ordinates.The data shown are from 21UT, 7 December, 1992, to 01 UT, 8  December, 1992  and are averaged over 20 s.During the time period reviewed, Galileo was located in the magnetosheath at GSM coordinates (X, Y, Z $ A70, 40, 26 R E ), suciently far from the centre of the tail that the magnetic ®eld data can be used to indicate the strength and orientation of the IMF.The bottom panel indicates the merging index.A value of 1 denotes that the conditions are favourable for low-latitude dayside reconnection, i.e. the IMF B z is southward or the magnetic shear angle between the subsolar magnetospheric ®eld and the IMF was only $70°w hen the IMF B z was northward

Fig. 3 .
Fig. 3. Range time intensity plot produced using data from the SABRE radar for 21 UT, 7 December 1992±01 UT, 8 December 1992.Only one of the central beams, beam 5, of the Wick radar is

Fig
Fig. 4a, b.IMAGE magnetometer chain data for 7 December±8 December 1992.Four hours of data are shown as indicated by the UT at the bottom of each panel.a Shows the X-component of the magnetic ®eld and b the Z-component.The station names and their magnetic latitudes are shown between the plots.The vertical dash-dot lines

Fig. 5 .
Fig. 5. Data from the SAM-NET magnetometers for 21 UT, 7 December 1992±01 UT, 8 December 1992.The data are the ®ltered H (upper panel) and D (lower panel) components of the magnetic ®eld, using a band-pass ®lter between 40 and 200 s

Fig. 6 .
Fig. 6.The two panels show magnetic ®eld data from the Greenland magnetometer chain for 21 UT, 7 December, 1992±01 UT, 8 December, 1992.The top panel is H-component data from the west coast chain (invariant latitudes 66.3±85.4°)and the bottom panel from the east coast chain (latitudes 52.7±81.0°)

Fig
Fig. 8a, b.Geosynchronous energetic particle data for 21 UT, 7 December, 1992±01 UT, 8 December, 1992, from spacecraft 1990-095 (local time = UT + 0031).a Is energetic electron data and b is energetic proton data.The ¯ux ranges are indicated to the left of the two panels . The lines with arrows show the plasma stream lines.Those without arrows show the open-closed ®eld line boundary, where the solid portion indicates the adiaroic section and the dashed portion the merging gap.The large arrows indicate the motion of the open-closed ®eld line boundary.The ¯ows are shown as being driven by steady unbalanced dayside reconnection with the open/ closed ®eld line boundary expanding equatorward at all local times.We also show the possible co-existence of viscously-driven ¯ow cells at latitudes lower than that of the open/closed ®eld line boundary.Substorm expansion phase onset occurred at 2215 UT, at which time the IMF became even more strongly

Fig. 10 .
Fig. 10.Graph showing the latitudinal position of the CRB deduced from consecutive ``fan'' plots of Goose Bay data plotted against time.The latitude of the poleward UV auroral boundary is marked for Freja orbit 828 as F1 and the latitude of the CRB as determined from the DMSP-F10 data are denoted by solid triangles.The times of substorm onset and the start of enhanced poleward ¯ows detected by the Goose Bay radar are also indicated

Fig. 13a±c .
Fig. 13a±c.The three panels show the vectors deduced from the EISCAT CP-4-B experiment, which represent the velocity and direction of ¯ow in the ionosphere.The vectors are rotated clockwise through 90°in order to reduce the congestion of the predominantly east/west ¯ows.The panels are for time intervals a 2125±2155 UT, b 2240±2310, and c 2310±2340 UT, 7 December, 1992

Fig. 15 .
Fig. 15.EISCAT and SuperDARN ¯ow vectors at 22:44:51 UT ± the time of onset of the poleward ¯ows seen at Goose Bay

Fig. 14a±c .
Fig.14a±c.Schematics of convection patterns and boundary motions during the substorm cycle described based on theoretical models ofLockwood (1990) andCowley and Lockwood (1992).In each of the three plots, the merging gaps are indicated by dashed lines and the large arrows indicate the boundary movement.a Shows the scene after 2215 UT, when substorm onset signatures were detected and the IMF B z ®eld has decreased to A20 nT.Although this disturbance may have involved an interval of reconnection, we infer that no open ¯ux was closed, and hence the polar cap area continues to expand.b Represents the ¯ows after the IMF B z has turned northward.A lobe cell is present ± the direction of circulation determined by the negative B y ®eld.Enhanced nightside ¯ows are shown driven by tail reconnection, consistent with the poleward ¯ows detected by the Goose Bay radar at high latitudes.The strong ¯ows are seen in the dayside sector driven by the contraction of the open/closed ®eld line boundary.c Shows the scenario during the substorm recovery phase with the equatorward relaxation of the boundary in the midnight sector, together with the weaker poleward motions elsewhere.Asymmetric ¯ow on the dayside is present as detected by the DMSP F10 satellite (Fig. 12c)

Table 1 .
The latitudinal position of the CRB, the local time location and the UT at which it was detected are indicated in this table, for consecutive polar passes of satellites F10 and F11 in the Northern Hemisphere ig. 12a±c.Ionospheric ¯ow data from polar passes of the DMSP satellites F10 and F11 for the period 2133±2314 UT, 7 December, 1992.The plots are for Northern Hemisphere passes and are arranged in chronological order regardless of which satellite they represent