Reconfiguration and closure of lobe flux by reconnection during northward IMF: possible evidence for signatures in cusp/cleft auroral emissions

. Observations are presented of the response of the dayside cusp/cleft aurora to changes in both the clock and elevation angles of the interplanetary magnetic ﬁeld (IMF) vector, as monitored by the WIND spacecraft. The auroral observations are made in 630 nm light at the winter solstice near magnetic noon, using an all-sky camera and a meridian-scanning pho-tometer on the island of Spitsbergen. The dominant change was the response to a northward turning of the IMF which caused a poleward retreat of the dayside aurora. A second, higher-latitude band of aurora was seen to form following the northward turning, which is interpreted as the e(cid:128)ect of lobe reconnection which reconﬁgures open ﬂux. We suggest that this was made possible in the winter hemisphere, despite the e(cid:128)ect of the Earth’s dipole tilt, by a relatively large negative X component of the IMF. A series of ﬁve events then formed in the poleward band and these propagated in a southwestward direction and faded at the equatorward edge of the lower-latitude band as it migrated poleward. It is shown that the auroral observations are consistent with overdraped lobe ﬂux being generated by lobe reconnection in the winter hemisphere and subsequently being re-closed by lobe reconnection in the summer hemisphere. We propose that the balance between the reconnection rates at these two sites is modulated by the IMF elevation angle, such that when the IMF points more directly northward, the summer lobe reconnection site dominates, re-closing all overdraped lobe ﬂux and eventually becoming disconnected from the Northern Hemisphere.


Introduction
The possibility that reconnection would take place between northward-pointing interplanetary magnetic ®eld (IMF) and the geomagnetic ®eld in the tail lobes, poleward of the magnetic cusps, was ®rst suggested by Dungey (1963).His concept was that a draped IMF ®eld line in the magnetosheath could reconnect simultaneously at the sunward edge of the tail lobes of both hemispheres, thereby producing a closed dayside ®eld line.This is now understood to be just one of several possibilities, but unlikely to take place in practice.Russell (1972) pointed out that a more likely situation was that the draped northward IMF would reconnect in one hemisphere only, thereby driving a circulation of open ¯ux within that polar cap.A catalogue of other topological possibilities was presented by Cowley (1981Cowley ( , 1983)).The northward-pointing sheath ®eld could, in principle, reconnect with either open or closed geomagnetic ¯ux: this reconnection could occur in both hemispheres simultaneously, but for dierent sheath ®eld lines, or at dierent times for any one sheath ®eld line.The ®rst observational evidence that this northward-IMF lobe reconnection does indeed take place was sunward ¯ow in the polar cap ionosphere, ®rst deduced from geomagnetic observations (Maezawa, 1976) and con®rmed by direct observations of the ¯ow (e.g.Burke et al., 1979;Cumnock, 1992) and of the associated ®eldaligned current pattern (Zanetti et al., 1984).The sunward ¯ow can give a net negative transpolar voltage (where positive voltage is associated with a dawn-todusk electric ®eld, i.e. antisunward ¯ow) (Freeman et al., 1993) and reverse latitudinal dispersion of cusp ions (Woch and Lundin, 1992) and can exist in both polar caps (Rei, 1982), implying that the lobe reconnection can take place simultaneously in both hemispheres.In addition, particle distribution functions at the lobe magnetopause (Gosling et al., 1991) and the direction of accelerated ¯ows on the dayside magnetopause (Paschmann et al., 1990;Kessel et al., 1996) con®rm that magnetopause reconnection at the lobe boundary does indeed take place.
The key dierence between this northward-IMF, lobe reconnection (at latitudes above the magnetic cusps) and southward-IMF, low-latitude reconnection (between the two magnetic cusps), is that the latter generates open ¯ux whereas the former recon®gures ¯ux that is already open.Lobe reconnection instantaneously changes the point where an ``old'' open ®eld line threads the magnetopause from somewhere down the tail to the reconnection site.Observations show that some lowlatitude open ¯ux generation can continue when the IMF is northward.Freeman et al. (1993) found that antisunward polar cap convection turned to sunward motion when the IMF clock angle was such that the magnitude of the clock angle jhj was less than about 70°: we here de®ne IMF clock angle h and elevation angle / by: h f y ajf y j tan À1 jf y jaf z 1 / f x ajf x j tan À1 jf x jaf z X 2 Scurry et al. (1994) and Phan et al. (1996) found the magnetopause ¯ow signatures of low-latitude reconnection persisted for northward IMF provided the local magnetic shear angle did not fall below about 45°.A similar conclusion was reached by Fuselier et al. (1995Fuselier et al. ( , 1997) ) from electron and ion distribution functions close to the magnetopause.The exact relationship of the local magnetic shear at the magnetopause with the IMF clock angle, h, is not known but Sandholt et al. (1998) have found evidence that the dayside aurora associated with low-latitude reconnection (and which they termed ``type 1'') was present for jhj b 45 whereas that associated with the lobe reconnection (which they termed ``type 2'') was present for jhj `90 and thus both were seen when |h| was between about 45°and 90°.Anderson et al. (1997) have proposed that the continuing low-latitude reconnection, generating open ¯ux, when jhj is in the range 45±90°is caused by the compression of the magnetosheath ®eld in the plasma depletion layer.However, even though low-latitude reconnection can continue during northward IMF, we know that open ¯ux is not generated to any great extent under these conditions.We know this because substorms are generally not observed during periods of prolonged northward IMF (jhj `90 ) (Farrugia et al., 1993).Wygant et al. (1983) showed that the range of residual transpolar voltages during northward IMF (previously thought to be due to some unspeci®ed ``viscous-like'' momentum transfer to closed ®eld lines) dropped progressively with time since the northward turning of the IMF: this strongly indicates that such voltage is associated with open ¯ux produced by the prior period(s) of southward IMF.Poleward contraction of the polar cap (due to closure of open ¯ux in the tail) can mimic the dayside ¯ow patterns expected for a viscous-like mechanism, as was postulated by Lockwood and Cowley (1992).This eect was con®rmed by Fox et al. (1994), who showed examples where the apparent viscous-like voltage in-creased during polar cap contractions caused by weak substorms.Thus much of the voltage associated with antisunward ¯ow during northward IMF is due to ®eld line closure in the tail and is not caused by either a viscous-like interaction or ®eld line opening at the dayside magnetopause.
Three of the important possibilities for reconnection are presented here in Fig. 1.The left hand schematics show magnetic ®eld lines relative to noon-midnight cross sections of the magnetosphere, as viewed from dusk such that the Sun is to the left.Open, closed and interplanetary magnetic ®eld lines are labelled o, c and i, respectively, and open ®eld lines recon®gured by lobe reconnection are called ``over-draped lobe'' (Crooker, 1992) and are labelled ol.The dashed line is the magnetopause (MP) and dots marked X are active reconnection sites.The right hand schematics are views of the Northern Hemisphere polar cap with the Sun to the top, dawn to the right and dusk to the left: convection ¯ow stream lines are marked with arrows, thin solid lines are non-reconnecting segments of the open-closed boundary and thick solid lines map to magnetopause reconnection X-lines.In all cases, we show steady-state situations in the ionosphere for simplicity, but the considerations can readily be generalised for non-steady cases by allowing for the ¯ows associated with moving boundaries (Cowley and Lockwood, 1992).The bulk of the ¯ows shown in Fig. 1 are reconnection-driven: as discussed, the ``viscously-driven'' ¯ow cells labelled v are most likely to be driven predominantly by continuing tail reconnection and polar cap contraction.
The numbered ®eld-line positions in the left hand plots show the evolution of each ®eld line under the combined action of the magnetic curvature (``tension'') force and the antisunward magnetosheath ¯ow.The corresponding motion of the ionospheric footprint of such ®eld lines is shown in the right hand plots.In Fig. 1a, the reconnection is between the draped interplanetary ®eld lines in the sheath (i) and closed magnetospheric ®eld lines (c).This reconnection gives poleward ¯ow and a ``standard'' dispersion of cusp ions (energy decreasing with increasing latitude) as the cusp precipitation evolves with time elapsed since reconnection (Lockwood, 1995).From this discussion, we infer this type of reconnection can continue (at some local times at least) during B z > 0 when jhj exceeds roughly 45°.
In Fig. 1b, lobe reconnection takes place between a draped interplanetary magnetic ®eld, (i) and the ``old'' open ¯ux of the tail lobe (o).Note that the old open ¯ux was produced by a prior period of southward IMF and threads the boundary further down the tail and has a dierent orientation in the sheath and in interplanetary space to i.In this case, often referred to as ``lobe stirring'' (Rei, 1982), the old open ®eld lines like 1 evolve toward the reconnection site (to 2) before being recon®gured at X so that they thread the dayside magnetopause (such as 3: these recon®gured open ®eld lines have been called ``over-draped lobe'' ¯ux see Crooker, 1992).These move sunward under magnetic tension, before sliding around either the dawn or dusk ¯ank and being returned to the tail lobe by the magnetosheath ¯ow.Evolution back into the tail lobe may be slow as the magnetic curvature force is initially acting against the sheath ¯ow.This case can, in principle, exist in steady state with lobe ¯ux circulating around a ¯ow cell (L) within the open ®eld line region, as shown in the right hand schematic.However, in reality, this is unlikely to be a steady-state phenomenon and ¯ux is very unlikely to circulate all the way around because the lifetime of the ¯ow cells (controlled by the persistence of the IMF orientation) is shorter than the time for ¯ux tubes to circulate around the cells.After lobe recon®guration, the ®eld line threads the dayside magnetopause, allowing higher density sheath plasma to stream into the magnetosphere and precipitate into the cusp ionosphere, sunward of the magnetic projection of X (the thick solid line).The sunward convection will give a reversed cusp ion dispersion (Woch and Lundin, 1993), as the ion precipitation evolves with elapsed time since reconnection.The precipitation seen on the old open ¯ux, poleward of the solid line (at the foot of ®eld lines 1 and 2) would be classed as either mantle or polar cap (as prior to the recon®guration, the sheath plasma will ¯ow mainly tailward on the old open ®eld lines).Because some of the overdraped lobe ®eld lines move  1a±c.Schematic illustrations of (left) the evolution of reconnected ®eld lines in the magnetosphere, as seen from the dusk ¯ank and (right) the corresponding steady-state ¯ow in the Northern Hemisphere ionosphere, with noon to the top.a is for southward IMF whereas b and c are two of the possibilities for northward IMF in all cases with B y » 0. Field lines and regions that are open, closed, interplanetary and overdraped lobe are lobe are labelled, respectively, o, c, i, and ol.See text for details (adapted from Lockwood, 1998) only very slowly, due to the competing eects of the sheath ¯ow and tension force, some of the cusp could be relatively stagnant and the precipitation relatively dispersionless.Figure 1 is drawn for small IMF jf y j and so the lobe convection cells are roughly equal in size whereas one of these would dominate for large jf y j (Russell, 1972;Heelis, 1984).Crooker and Rich (1993) found that lobe convection cells L are predominantly a summer phenomenon which they explained as the eect of the dipole tilt towards the sun which favours lobe reconnection in the summer hemisphere.In the opposite hemisphere to the lobe reconnection site X (the Southern Hemisphere in Fig. 1b), the lobe reconnection has no eect (Freeman et al., 1993;Crooker and Rich, 1993;Knipp et al., 1993).
An alternative situation is shown in Fig. 1c where the overdraped lobe ¯ux produced by X (like ®eld line 3) is itself reconnected at X1 in the other hemisphere to produce a closed ®eld line (like ®eld line 4).This is similar to the original suggestion for B z > 0 by Dungey (1963), recently invoked by Song and Russell (1992), Song et al. (1994) as a way of producing a closed ®eldline low-latitude boundary layer containing magnetosheath plasma.The example shown in Fig. 1c is the most realistic in that ®eld lines are not simultaneously reconnected at both lobe sites.In the case shown, reconnection occurs ®rst in the Northern Hemisphere and only subsequently does the overdraped lobe ®eld line reconnect in the Southern Hemisphere.Note that the ¯ow cells with sunward polar cap ¯ow now cross between the open and closed ®eld line regions and do not remain in the region of open ¯ux as in Fig. 1b.If the reconnection rate at X1 exceeds that at X, the Northern Hemisphere overdraped lobe ¯ux will decay, the latitudinal width of the northern cusp would shrink.After the Northern Hemisphere ol ®eld has disappeared at any one local time, X1 would act to expand an overdraped lobe connected to the Southern Hemisphere.
In this study, we present optical observations of the dayside cusp/cleft aurora made from Spitsbergen, one of the Svalbard islands, near the December solstice and at 9 UT.Observations of the cusp aurora are possible in the dark magnetic-noon, winter conditions at Svalbard (sun more than 10°below the horizon) for, roughly speaking, a two-month observing season about the December solstice and around 9 UT, when the viewing area is close to 12 MLT.Selecting these dark conditions at noon means that we study an extreme magnetic topology in the cusp region, with the Earth's rotational axis tipped away from the Sun at an angle of 23.5°at winter solstice, and with a further tip of the northern magnetic axis.At this UT, the angle of the magnetic axis (projected onto the GSE XZ plane) with respect to the Z axis is d = 28°.The dipole tilt would be greatest at 5 UT (d » 35°), when a location roughly 60°to the east of Svalbard is at magnetic noon.
The dipole tilt eect on the dayside cusp precipitation (Newell and Meng, 1989) and magnetopause ®eld topology was invoked by Crooker and Rich (1993) as the reason that the lobe reconnection cells are seen predominantly in the summer hemisphere.In the winter hemisphere, a large B x component of the IMF would be required to drape a northward-pointing IMF such that it produced an anti-parallel ®eld con®guration on the lobe boundary, whereas such a con®guration is much more readily achieved in the summer hemisphere.Thus the IMF B x component is likely to be of relevance as it should favour lobe reconnection in one hemisphere (B x < 0, B z > 0 should favour the Northern Hemisphere, whereas B x > 0, B z > 0 should favour the Southern Hemisphere).
Figure 2 illustrates how the IMF B x eect on lobe reconnection can be characterised by the elevation angle of the IMF / (as de®ned by Eq. 2). Figure 2 shows selected geomagnetic ®eld lines predicted by the Tsyganenko T96-01 model at 11 UT at equinox (d = 0) and 9 UT at December solstice (d = 28°).At the equinoxes, 0</ <90°would be an IMF orientation that favoured Southern Hemisphere lobe reconnection (at the site LX in Fig. 2a) whereas the range (A90°)</< 0 would favour northern lobe reconnection (Fig. 2b).However, for the observations discussed here at 9 UT and near December solstice, the dipole axis is tipped at d = 28°and Fig. 2c shows how we would expect the southern lobe to be favoured for most values of / (roughly speaking for the range Ad < / < 90°).For Northern Hemisphere lobe reconnection to occur would require A90°< / < Ad and so calls for / to be close to (A 90°) (Fig. 2d).
In this work, we ®rst present observations of changes in the interplanetary medium, made by the WIND satellite, and contrast with observations of the response of the dayside cusp/cleft aurora, as detected by meridian-scanning photometers at Ny A Ê lesund, Svalbard and an all-sky camera at Longyearbyen, Svalbard.In the subsequent section, we interpret the optical observations in terms of the appearance and disappearance of dierent reconnection sites and relate these to the changes in the IMF orientation.These observations were made during northward IMF near winter solstice when the dipole axis is tipped at d = 28°.Hence lobe reconnection should be favoured in the Southern Hemisphere for most values of the IMF elevation angle.However, we argue that for sucient magnitude of negative IMF B x values, lobe reconnection should occur in the Northern Hemisphere (Fig. 2).

IMF and solar wind observations
Figure 3 shows observations of the IMF components, in GSM coordinates, made by the WIND satellite on 12 January 1997, between 07:30 and 09:00 UT.At this time the satellite was located near (X = 103 R E , Y = A55 R E , and Z = A6R E ) in GSM co-ordinates.The solar wind speed in this interval was approximately constant at 530 km s A1 .From this we estimate that the propagation delay from the satellite to the magnetopause is 23 3 min.This estimate includes a propagation time of 17 min from WIND to a bow shock at X = 15 R E , plus a time of 6 min to cross the magnetosheath.The uncertainty is derived from the fact that the orientation of the changes in the interplanetary medium is not known and we use the observed ¯uctuation level of the IMF longitudinal angle of 15°to estimate the lag uncertainty.To this we add a further 2 min to allow for the minimum electron ¯ight times from the magnetopause to the ionosphere and the average radiative lifetime of the 630 nm atomic oxygen emission line.Thus we deduce an average delay of about 25 3 min between the observation of an IMF or solar wind change by WIND and the associated change seen in the 630 nm cusp/cleft aurora.Notice, however, that high ¯ux precipitation can persist for several tens of minutes down open ®eld lines following reconnection: hence the delay may be as much as 35 min between an IMF change that causes a cessation of reconnection and the conclusion of the consequent fading of an auroral structure.
The solar wind density, like the solar wind speed, was relatively steady.It ¯uctuated in the range 7±8.5 cm A3 in the interval 07:30±09:00.Important changes, however, did take place in the IMF.The B z component increases from southward to northward in a relatively smooth manner, giving a gradual fall in the IMF clock angle, but with a discontinuous change through B z = 0 at 08:23.There is a brief return to a southward orientation lasting 2 min around 08:44 UT.At 08:00 the B y component is negative (A1.5 nT), but this turns positive shortly thereafter and increases to about +5 nT before declining again.The Bx component is negative after 08:00 but ¯uctuates between A4 nT and A2 nT.The elevation angle of the IMF / (bottom panel of Fig. 3) is initially close to A180°, but increases through A90°with the northward turning at 08:23 UT.Subsequently / remains between A70°and A90°until the brief excursion to B z < 0 at 08:44.After 08:45 / is between A40°and A60°.For the time of the observations, d = 28°t herefore prior to 08:44 UT, / is signi®cantly smaller than (Ad ) (by over 40°) whereas after 08:45 / is within 30°of (Ad).During this time, the clock angle h decreases gradually from near 180°at 08:00, falling below 45°only brie¯y near the very end of the interval (08:58).Figure 4 shows the behaviour of the IMF in this interval in the form of two hodograms: the three dots are for 08:00, 08:30 and 09:00.In both the XZ and YZ GSM planes, the ®eld shows the systematic changes described already.
It is the aim of this work to study the eects of these ®eld rotations on the dayside cusp auroral, as seen in 630 nm light at magnetic noon.

Photometer observations
Figure 5 shows the intensity of 630 nm (left) and 557.7 nm (right) auroral emissions seen by the meridi-an-scanning photometer at Ny A Ê lesund, Svalbard.This instrument is situated at 78.9°N and 11.9°E, and is at a magnetic latitude of 75.9°.In both panels, time runs from 08:30 UT at the top to 09:32 UT at the bottom and the intensity is shown as a function of zenith angle (with 80°to the north to the left and 75°south to the right) along the magnetic meridian for each UT.The latitudes corresponding to these zenith angles depend on the emission altitude, as given by Lockwood et al. (1993).The MLT of the observing site increases from about 11:15 to 12:15 in this interval, so the observations are made close to magnetic noon.At the start of the interval, the aurora was relatively weak and unstructured and peaked at about 63°south of zenith.This corresponds to a magnetic latitude near 74°, consistent with the lagged IMF having been predominantly southward (average B z » A5 nT) since about 06:15 UT.A weak intensi®cation, followed by a poleward-moving event, was seen commencing about 08:32.At 08:40 there was an intensi®cation (labelled C), after which a second band of aurora broke away from the poleward edge and moved poleward.We here adopt the terminology  wood et al., 1993).Auroral forms seen by the 630 nm all-sky camera, are marked for the on left panel if and when they appear in the merdian scanned by the photometer, using the same lettering code as Fig. 6 introduced by Sandholt et al. (1996) and call the new more poleward band ``type 2'' and the more equatorward band ``type 1''.The type 1 persisted and intensi®ed around 08:55 (E).Shortly after this time, a strong, new auroral form G was seen in the higher-latitude, type 2 band in both 630 nm and 557.7 nm light (this is con®rmed to have been a new event by the all-sky camera data, see next section) and moved equatorward.
As this faded, the type 1 moved poleward and brightened intermittently in both the red and the green light (J).The type 2 aurora brightened twice more (H and I) and then persisted at a lower intensity at a constant latitude.By the end of the period the two emission bands had almost merged into one band of 630-nm dominant emission, well to the north of the observing site (type 1 only).This major poleward motion of the 630-nm noon aurora appears to have been the response to the northward rotation of the IMF shown in Figs. 3  and 4. In addition to this evolution over about 45 min, there are changes which may be associated with shorterlived IMF variations.For example, the IMF returned brie¯y southward near 08:44 (Fig. 3) and this may have caused of the decay of the type 2 band and the intensi®cation of the type 1 at around 09:08 UT.Auroral forms that were also seen in the all-sky camera data (see next section) are labelled in Fig. 5: C, E, and J are part of the type 1 aurora, whereas H and I are intensi®cations of the type 2 aurora.The camera data reveal that these are two out of 5 events that form at the poleward edge of the type 2 aurora and migrate equatorward to join the type 1 aurora before fading.This motion takes place to the west of the scanned meridian and is not seen by the photometer, except for event G that just extends suciently far eastward to touch the meridian.This event is brightest shortly after it forms (at 08:58) in the type 1 band and just before it fades (at 09:12) near the type 2 band: between these two times the event is much less intense at the meridian, but its equatorward motion can still be discerned in Fig. 5. Equatorward-moving type 2 transient auroral events during northward IMF were ®rst reported by Sandholt (1991).
The right-hand panel of Fig. 5 shows that many of the features seen in the 630-nm emissions are seen in the 557.7-nm emissions also.In general, the 630 nm emissions are more intense, consistent with soft cusp/cleft electron precipitation.However, some events show enhanced 557.7 emission, particularly the events G, H and I when they ®rst form at the poleward boundary of the type 2 aurora.

All-sky camera observations
Figure 6 shows selected 1-s all-sky images recorded during this auroral sequence, as seen by the 630 nm allsky camera at Longyearbyen, 120 km (1°of magnetic latitude) to the south of Ny A Ê lesund along the magnetic meridian.The line-of-sight intensities have been mapped onto a geographic grid by assuming a constant and single emission altitude of 250 km.Note that the intensities seen in the bottom right of each plot are the eect of scattered sunlight within the ®eld-of-view (f.o.v.).In the ®rst image, frame1 (for 08:33:14 UT), the aurora form a weak and thin band at almost constant L values, well to the south of Svalbard.The most intense aurora in this band is to the west of the magnetic meridian scanned by the Ny A Ê lesund.This auroral form is labelled A in Fig. 6.In image 2, for 08:49:38 (~16 min later), four major changes have taken place: (a) the bright patch A to the west has faded, but the low-latitude band can still be seen, (b) a second band B has appeared to the west at higher latitudes, having emerged from the poleward edge of the pre-existing band; (c) a major intensi®cation C is seen to the east (inspection of the images between images 1 and 2 shows that this has spread along the pre-existing band from the east); and (d) there has been a poleward-moving transient event D which also expanded into the f.o.v.from the east and is now fading at higher latitudes to the east.
In image 3 (08:58:40 UT, 522 s later), the high luminosity to the east, C, has retreated eastward somewhat and the poleward, type 2, aurora B has moved slightly equatorward.There is a brief equatorward expansion of the aurora with (rayed) activity on the lower band to the south of Spitsbergen (E) and a pair of intensi®cations have recently commenced on the higher latitude band very close to Ny A Ê lesund (forms F and G which appeared at 08:56 and 08:58).The second of these is the ®rst of the events to be seen at the longitude of the photometer scan (giving the event labelled G in Fig. 5).
In image 4 (09:03:21 UT, 281 s later), all activity on the equatorward band has ceased to the west of Spitsbergen but the feature B has faded and moved a considerable distance equatorward to just poleward of where the low-latitude band used to be.The features F and G have expanded by spreading to the north and to the west.Just to the east of G, a new feature H is seen forming.
In image 5 (09:05:41 UT, 140 s later) the events F and G have continued to move west, but have also moved increasingly equatorward.The faint remnants of the original type 2 feature, B, can still just be seen as a very weak patch at the equatorward edge of the western end of the visible auroral band.The feature to the north of Svalbard, H has intensi®ed and a new event I has formed to the east of it.The south-west motion of events F, G, H and I continues in images 6 (09:07:02, 81 s later) and 7 (09:08:22, a further 80 s later).
In image 6, form F brightens considerably and the form G is approaching it at its eastern end.This merging is almost complete in image 7, and thus the equatorward edge of the aurora to the west of Svalbard is made up of forms F and G, which initially formed at the poleward edge of the type 2 aurora.In addition, a small form J has suddenly brightened just equatorward of G.This feature crosses the photometer meridian and so can also be seen in Fig. 5.It drifts poleward over the next 15 min.
Image 8 (09:14:23 UT, 6 min later) shows that the patch C to the east has ®nally faded away.By this time, the 630-nm aurora forms a weak single band at latitudes to the north of Ny A Ê lesund, with slightly stronger intensities seen to the far west of the ®eld of view.This enhanced emission is the remnants of the last two of the southwestward moving events, H and I.
Thus the images show that there was a great deal of structure and equatorward motion, which was not generally seen by the photometer.In addition, they reveal the westward motion of the auroral forms.A key point to note that is the type 2 structure which forms to the north of the pre-existing (southward IMF, type 1) auroral band develops into a series of events.Their elongated nature almost certainly re¯ects a rayed-arc structure.In considering their true motion we must remember the viewing geometry and that the emission comes from a range of altitudes.They appear to rotate around the observation site only because they are moving south-westward and to the north west of the camera.The sequence of images 3, 4, 5 and 6 are chosen to be close together in time to show that this poleward activity develops into a series of events that drift west and to the south and always end by moving into the equatorward edge of the lower latitude auroral band to the west of Svalbard and fading.The ®rst event to do this, B, was ®rst seen at around 08:49 UT (when it formed the new poleward auroral band to the west of Svalbard) and it disappeared (after fading and brightening again during its south westward motion) into the equatorward auroral edge at 09:06, giving it a total lifetime of 17 min.The second event to follow this sequence (F) formed at about 08:56 and faded at 09:11 so it had a shorter lifetime of 15 min.Event G formed at 08:58 and faded at 09:12, giving it a lifetime of 14 min.Event H formed at about 09:03 and faded at 09:15, making its lifetime 12 min and event I formed at 09:05 and also faded at 09:15, so its lifetime was 10 min.
Thus there is a sequence of ®ve of these events, which form at the poleward edge of the type 2 aurora and migrate south and west before fading at the equatorward edge of the type 1 aurora.Their formation times are separated by 7, 2, 5 and 2 min and their lifetime gets progressively shorter from 16 to 10 min.They form only in the interval 08:49±09:05 UT.

Interpretation of the images
The auroral intensities are, at most times, dominated by the 630 nm emission, the relatively low emission of 557.7 nm light showing that most of the emission is caused by soft magnetosheath-like particles.The exceptions to this are all short-lived and associated with the onset of the equatorward-moving events in the type 2 aurora.Lockwood et al. (1993) found that dominant 557.7 nm emissions in transient events were in the regions of upward ®eld-aligned current associated with the ¯ows.
The orientation of the IMF is known to vary during the interval and so the location of magnetopause reconnection sites is expected to move, whether one envisages anti-parallel merging (Crooker, 1979) or tilted subsolar reconnection during southward IMF, giving way to lobe reconnection during northward IMF (e.g.Gosling et al., 1990Gosling et al., , 1991)).Furthermore, the variation of the soft-electron precipitation seen subsequent to each reconnection, will depend on how the opened ®eld lines evolve over the magnetopause (Onsager et al., 1993;Lockwood, 1995) and this too will depend critically on the IMF orientation.In addition, the magnetic mapping of the reconnection site to the ionosphere is altered by the penetration of IMF components (particularly B y ) into the magnetosphere (Cowley et al., 1991).
One clear eect observed in Figs. 5 and 6 is the poleward motion of the cusp/cleft aurora in response to the northward turning of the IMF.This is expected from statistical studies of the latitude of cusp precipitation (e.g.Burch, 1973;Candidi et al., 1983;Newell et al., 1989) and from studies of the dayside cusp/cleft aurora (Horwitz and Akasofu, 1977;Sandholt et al., 1998).These ®gures also show, however, that the cusp/cleft aurora is narrow in latitudinal width when the IMF was both strongly southward and strongly northward.This is somewhat surprising as statistical studies have suggested the cusp and cusp/cleft regions are much wider in latitude under northward IMF conditions (Carbary and Meng, 1988;Newell and Meng, 1987).The precipitation seen here does initially spread in latitude following the northward turning, but then decreases again in latitudinal width.This is particularly apparent in the photometer data shown in Fig. 5.We therefore conclude that cusp latitudinal width is not a simple function of IMF B z but also depends on other factors.
The formation of type 2 events and structures poleward of the original type 1 aurora indicates that lobe reconnection has set in the Northern Hemisphere lobe (Weiss et al., 1995;Sandholt et al., 1996Sandholt et al., , 1998a, b;, b;éieroset et al., 1997).These features initially form at the poleward edge of the pre-existing type 1 aurora, as noted by éieroset et al. (1997).It is somewhat surprising that such features can be seen, considering that this is not favoured at this time near winter solstice because of the Earth's dipole tilt away from the sun in the Northern Hemisphere.It has been suggested that the dipole tilt causes lobe reconnection (and associated sunward ¯ow) to be seen only in the summer hemisphere (Crooker and Rich, 1993).However in this case, the strongly negative B x component of the IMF could have overcome this dipole tilt eect when B z is only weakly positive, giving / close to A90°and draping magnetosheath ®eld lines over the Northern Hemisphere lobe (as in Fig. 2d).The IMF seen by WIND had such an orientation for an interval of roughly 08:26±08:42, which could therefore have been responsible for the formation of the type 2 auroral events at 08:51±09:07 if the propagation delay from the satellite to the ionosphere was the nominal 25 min.This was the period in which type 2 events were seen to form.It is also the period in which the poleward part of the aurora migrates poleward in Fig. 5.
Two factors then become of particular interest.Firstly, the type 2 auroral events were seen to drift equatorward into the equatorward boundary of the persistent type 1 aurora, where they faded.This suggests that the open ®eld lines that were recon®gured by lobe reconnection may have subsequently been re-closed by reconnection in the other lobe.This has been proposed by Song and Russell (1992) and Song et al. (1994) as a way of populating a closed low-latitude boundary layer with magnetosheath plasma during northward IMF.Secondly, the nature of the poleward retreat of the type 1 (low-latitude) aurora after 09:07 is signi®cant because it was not accompanied by a poleward motion of the more northerly type-2 aurora and so represents a narrowing of the latitudinal width of the total band of 630 nm aurora (as seen in Fig. 5).This narrowing also implies that re-con®gured lobe ¯ux in the Northern Hemisphere was being closed by lobe reconnection in the Southern Hemisphere.Lastly, one should consider how the poleward motion of a dayside aurora can occur.In general it represents the closure of open ¯ux.That closure could be taking place in the tail giving a polar cap contraction as discussed by Lockwood et al. (1990) and Cowley and Lockwood (1992).However, were this to have been occurring here, we would expect all of the dayside aurora (both type 1 and type 2) to have both migrated poleward, and Fig. 5 shows that this was clearly not the case in this example.Thus this strongly implies that in this case we were observing the re-closure of open ¯ux on the dayside by lobe reconnection, such that the equatorward type 1 aurora migrates poleward, but the poleward type 2 aurora does not.
This interpretation is consistent with the observed IMF orientation changes seen by WIND.After 08:35, The B x component remained roughly constant but the average B z component increased from about +2 nT at 08:40 to about +4 nT at 09:00, corresponding to a change in the IMF orientation in the GSM X-Z plane from about 80°to 40°with respect to the Z direction (last panel, Fig. 3).Given that the tilt of the magnetic dipole at this time, d was close to 28°in this frame.This rotation may have been sucient to cause the reconnection in the northern (winter) lobe to slow and possibly cease and to have caused that in the southern (summer) lobe to commence and become dominant.The rotation of the IMF will not directly change the orientation of the interplanetary part of overdraped lobe (ol) ¯ux; however, the accumulation of draped IMF over the overdraped lobe ¯ux in the magnetosheath will compress it onto the Southern Hemisphere lobe and so promote its reconnection.This argument is illustrated graphically in the next section.Using the satellite-tomagnetopause delay of 25 min.discussed above, these changes would be seen over the interval 09:05±09:25 which, as shown by the photometer data (Fig. 5), is indeed the interval in which the activity on the poleward part of the oval decays and in which the equatorward type 1 aurora migrates poleward.Note that the migration proceeds most rapidly after each of a series of intensi®cations of this auroral band suggesting that the Southern Hemisphere lobe reconnection closing the overdraped lobe also occurs in bursts.
We note also that the appearance of the small feature J and the fading of the type 2 aurora follow the short excursion to southward IMF around 08:44 by the predicted lag.We suggest that event J is similar to event E and that this brief swing was sucient to cause a brief burst of low-latitude reconnection whilst marking the end of the lobe reconnection in the Northern Hemisphere (which never returned because of the drift in the IMF elevation angle / to lower values).

Schematic illustration of the proposed interpretation
Figure 7 is a series of schematic illustrations to indicate the likely changes in the reconnection topology and how they are re¯ected in the dayside aurora seen at Longyearbyen.In this ®gure, the left-hand plots are views of the magnetosphere from the dusk side of the Earth, the middle plots are views of the magnetosphere from the sun and the right-hand plots are views of the noon polar ionosphere from above.In the left-hand plots, the magnetopause is shown as a dotted line and a segment of an active reconnection X-line as a dot.In the middle plots, lobe ®eld lines have been omitted to avoid complicating the diagram and active reconnection X-lines are shown as heavy lines.In the right hand plots, the circle is the ®eld of view of the all-sky camera (the centre of which moves from 11:30 MLT to 12:30 MLT during the period of interest), ¯ow stream lines are arrows, the thin solid segments are non-reconnecting segments of the open-closed boundary and the heavy solid lines are merging gaps which map to active reconnection sites on the magnetopause.The IMF orientations are taken from the WIND data for the nominal lag of 25 min.The motion of boundaries and the positions of the reconnection sites are inferred from locations of the 630-nm aurora.The directions of convective ¯ow are inferred from the orientation of the precipitation plumes in the case of quasi-steady reconnection, and from the direction of event motion in the case of transient events.In addition, dominant 557.7-nm emission is taken to show regions of upward ®eld-aligned current.
Figure 7a corresponds to the start of the interval of auroral data (08:25 UT).For the inferred WIND to ionosphere lag of about 25 min, this IMF impinging on the magnetosphere at this time would be that seen by WIND near 08:00 and have B z < 0, B y < 0 and B x < 0 and the orientation shown in the left and centre plots of Fig. 7a.This allows ``low-latitude'' reconnection (by which we here mean reconnection of closed ®eld lines anywhere between the magnetic cusps which generates open ¯ux) and because of the IMF orientation in the XY plane is at 90°to the garden hose direction we suggest reconnection is favoured on the dawn ¯ank of the magnetosphere.This is re¯ected in the location of the merging gap (labelled a) to the west of Svalbard.The weakly negative B y component at this time moves the newly-opened ®eld lines to the east, but we suggest that this is not sucient to bring high precipitation ¯uxes into most of the camera f.o.v., and the brightest 630 nm aurora (A) would have been to the west of the f.o.v.This thus can explain the situation seen in image 1.The low intensity band would then be the low-latitude boundary layer, LLBL (on either closed or open ®eld lines) to the east of the merging gap. Figure 7b is for 08:40 (corresponding to WIND observations at about 08:15), for which the B y component has changed polarity giving B z < 0, B y > 0 and B x < 0 and the IMF orientation shown.For this, and all subsequent times, the IMF makes a typical garden hose angle near 45°degrees in the XY plane and we suggest this favours reconnection more to the dawnside that in Fig. 7a.The aurora A does not migrate eastward across the noon meridian, but a new form C expands from the east and so we suggest that the X-line a has ceased reconnecting and a new one b has formed and spread into the f.o.v.from the east.The polarity change in IMF B y was seen by WIND at 08:05, for which the onset of the easterly aurora C would be about 08:30 (it is ®rst seen in the camera f-o-v at 08:38) and the westerly aurora A would fade over the subsequent 10±20 min, as observed.
Figure 7c is for 09:00 (corresponding to 08:35 at WIND) for which the B z component has changed polarity giving B z > 0, B y > 0 and B x < 0. The elevation angle / of the IMF in the ZX plane (between A70°and A90°in GSM) allows it to be draped over the northern winter lobe such that, despite the dipole tilt, a lobe reconnection site c is possible there.This produces ®eld lines that initially convect sunward and have one of the topologies suggested by Cowley (1981Cowley ( , 1983) ) and which were called ``overdraped lobe'' by Crooker (1992).The large B y component at this time means that the ionospheric footprint of ®eld lines reconnected by c would move strongly westward as well as slightly sunward.The rotation of the IMF has caused the lowlatitude reconnection X-line b to split into two small parts b¢ and b¢¢ which have rotated to higher magnetic latitudes near the cusp, as suggested for the anti-parallel reconnection hypothesis by Crooker (1979).The presence of b¢ is indicated by the maintenance of the aurora C to the east of Svalbard, although intensities may be lower as the open ®eld lines produced by b¢ thread the boundary further away from the high density sheath at the nose of the magnetosphere than did those produced by b in Fig. 7b (see Lockwood, 1997).However the curvature (``tension'') force will tend to bring these back towards the dayside slightly, allowing some moderate precipitation ¯uxes to be maintained.The ®eld lines connected to the Northern Hemisphere in the out¯ow region from b¢¢ will be moved away from the nose in the antisunward direction and so ¯uxes will be lower and auroral intensities lower.The existence of b¢¢ is, however, indicated by the brief ¯aring of aurora E in the small erosion event to the west of Svalbard in image 3 of Fig. 6.In addition, we could associate the formation of event J in image 7 of Fig. 6 with a brief re-activation of b¢¢ during the brief swing to southward IMF seen by WIND around 08:44.Although the production of new open ®eld lines by b¢¢ would only last 1 or 2 min, the precipitation (and this event J) would remain until the newly opened ®eld lines were swept into the tail.The lobe reconnection explains the formation of the ®ve southwestward moving events discussed in the previous section (B, F, G, H and I).The formation of the X-line c is ®rst indicated by the formation of auroral form B and of the type 2 aurora that breaks away from the poleward edge of the type 1.These occur at 08:49, corresponding to 08:24 in the WIND observations when the IMF ®rst turned northward.Lobe reconnection recon®gures old open ¯ux (which thread the magnetopause far down tail and in which no precipitation is seen at low altitudes) to make overdraped lobe ¯ux which threads the magnetopause sunward of the lobe reconnection site and down which high magnetosheath ¯uxes will precipitate.
In Figs.7d, e and f, the reconnection X-lines b¢ and b¢¢ have disappeared and so no more open ¯ux is being generated by low-latitude reconnection.The auroral form to the east of Svalbard, C, has almost disappeared in image 8, its ®nal fade beginning at about 09:18.Given that the last ®eld lines to be opened by this reconnection site will continue to experience high ¯uxes of magnetosheath ions and electrons for at least 10 min after reconnection, this suggests that the X-line b¢ ceased reconnecting some time before 09:10, which corresponds to WIND observations at about 08:45, when the IMF clock angle has decreased to near 55°.
It is not unambiguously clear from the data if lobe reconnection in the Southern Hemisphere (at d) commences before or after that in the Northern Hemisphere (at c) ceases.Figure 7d assumes that it is before, but this stage may never be present if this is not the case.It is, however, included here for completeness.The ®rst evidence for the re-closure of overdraped lobe ¯ux is the arrival of the ®rst of the south-west drifting features at the equatorward edge of the auroral band at about 09:10.The last of the southwest drifting events, I, formed before this at 09:05.This implies that, infact, c may well have ceased reconnecting before d commenced and the phase shown in Fig. 7d was never actually present.Were the reconnection at the lobe sites c and d to take place at the same rate, the amount of overdraped lobe ¯ux would remain constant, however, the rotation of the IMF towards northward (increasing /) will increasingly favour the summer hemisphere and so d will start to dominate over c and the amount of overdraped lobe ¯ux will decrease.Figure 5 shows that peak latitudinal width of the 630 nm aurora is reached at about 09:08, corresponding to WIND observations at about 08:43, close to the time that the IMF elevation / increases over a suggested threshold of about A60°for the ®rst time.The right-hand panel in Fig. 7d shows the situation which would exist shortly after the formation of reconnection line d, if c were still dominant.The reconnection at c would drive a weak dusk lobe cell as well as the larger dawn cell because B y , although still positive, is decreasing in magnitude.There is also a ¯ow stream line which re¯ects the re-closure of overdraped ¯ux at d.This streamline crosses a non-reconnecting (``adiaroic'', see Siscoe and Huang, 1985) segment of the boundary where ¯ux, plasma and the boundary all move together poleward, as shown by the large arrow.
Figure 7e is for 09:10 (corresponding to roughly 08:45 at WIND) for which both the clock and the elevation angles have changed as the IMF turns more northward to the orientation shown.The reconnection at d is now closing overdraped ®eld lines and causing the observed poleward motion of the equatorward band of the aurora, while events formed by c (before it ceased reconnection) are still migrating to the south and west (forms F, G, H and I). Figure 7f marks the end of this process when the last of the overdraped ¯ux is re-closed.The X-line d is now purely reconnecting the Southern Hemisphere lobe ¯ux and is not connected at all to the Northern Hemisphere.Thus all directly reconnectiondriven auroral luminosity ceases on a highly contracted oval.This is close to happening in image 8 of Fig 6 .The loss of overdraped ¯ux is also consistent with the increasingly shorter lifetimes of the equatorward-moving patches as they have a shorter transit time between the merging gaps c and d.

Discussion and conclusions
We have presented some observations of the interplanetary magnetic ®eld and of the associated dayside cusp/ cleft aurora.The rotation of the IMF to northward caused the poleward contraction of the aurora and the appearance of type 2 aurora at the poleward edge of the type 1 (southward-IMF) aurora.This is consistent with previous observations reported by Sandholt et al. (1996Sandholt et al. ( , 1998a,b) ,b) (1997).The type 2 aurora subsequently developed into a series of ®ve events which moved west and equatorward.
The latitudinal width of the cusp/cleft aurora would increase as the lobe reconnection generates overdraped lobe ¯ux.The rate at which this ¯ux is moved away from near noon (largely towards dawn because of the dominant positive B y component of the IMF) by the tension force and the sheath ¯ow would be relatively slow, in which case the poleward motion of the poleward edge of the aurora re¯ects the lobe reconnection rate.The poleward edge moves from a zenith angle of 50°to the south at 08:45 to 60°to the north at 09:06, for an assumed emission height of 250 km this corresponds to a motion over 408 km in 21 min, i.e. an average speed of v = 325 m s A1 , corresponding to a reconnection rate, mapped into the ionosphere, of E = vB i = 16 mV m A1 .
There is evidence in this case that overdraped lobe ¯ux was lost at the equatorward edge of the cusp/cleft aurora.This means that the overdraped lobe ¯ux was reconnected in the Southern Hemisphere to become closed ¯ux, as suggested by Crooker (1992), Song and Russell (1992) and Song et al. (1994) as a source of magnetosheath plasma in a northward-IMF, closed LLBL.There are two pieces of evidence for this.Firstly, the type 2 aurora events propagate south and west (consistent with the observed positive IMF B y ) and were seen, in the all-sky camera data, to disappear at the equatorward edge of the pre-existing type 1 aurora to the west of Svalbard.Secondly, as the IMF elevation angle / increased, the width of the cusp/cleft aurora decreased, implying the loss of overdraped lobe ¯ux.The maximum width of the cusp/cleft aurora was achieved at about 09:08, corresponding to WIND observations at about 08:43, very close to the time that the IMF elevation / increased to near (Ad) for the ®rst time.We interpret this as the threshold at which we would expect the Southern Hemisphere lobe reconnection (which destroys Northern Hemisphere overdraped lobe ¯ux) to have started to dominate over Northern Hemisphere lobe reconnection (which generates the Northern Hemisphere overdraped lobe ¯ux).
The evolution of the cusp/cleft aurora was therefore consistent with lobe reconnection in both hemispheres.Furthermore, the ratio of the reconnection rates at the two lobes appears to have been controlled by the IMF elevation angle / for the prevailing dipole tilt of the Earth's magnetic ®eld.
Fig. 1a±c.Schematic illustrations of (left) the evolution of reconnected ®eld lines in the magnetosphere, as seen from the dusk ¯ank and (right) the corresponding steady-state ¯ow in the Northern Hemisphere ionosphere, with noon to the top.a is for southward IMF whereas b and c are two of the possibilities for northward IMF in all cases with B y » 0. Field lines and regions that are open, closed, interplanetary and overdraped lobe are lobe are labelled, respectively, o, c, i, and ol.See text for details (adapted fromLockwood,  1998) Fig. 2a±d.Schematic illustration of lobe reconnection for equinox (a and b) and December solstice (c and d) conditions with various IMF elevation angles /.In a B x > 0, B z > 0 (0 < / < 90°) and this should favour a lobe reconnection site (LX) in the Southern Hemisphere.In the other 3 cases, B x < 0, B z > 0 (A90°< / < 0).At equinox, this favours the Northern Hemisphere if (as in b), but at December solstice would favour the Southern Hemisphere if / > Ad (as in c), where d is the tilt of the Earth's dipole axis with respect to the Z axis in the XZ plane: northern lobe reconnection would require / to be close to A90°(as in d)

Fig. 3 .
Fig. 3. IMF observations by the WIND satellite on 12 January 1997, between 07:30 and 09:00 UT.The estimated satellite-to-ionospheric cusp propagation delay is about 25 min.The panels show (from top to bottom) the B x , B y and B z IMF components (in the GSM frame and in nT), the clock angle h, de®ned by Eq. (1), and the elevation angle /, de®ned by Eq. (2) (both in degrees)

Fig. 5 .
Fig. 5. Scans by the meridianscanning photometers at Ny A Ê lesund, Svalbard on 12 January 1997.630 nm (left) and 557.7 nm (right) intensity is shown as a function of time (increasing down the page) and zenith angle according to the scale bars given at the top of each column.The zenith angles vary between 80°to the north of Ny A Ê lesund and 70°to the south.The latitude of the emissions depends on the emission pro®le with altitude (seeLockwood et al., 1993).Auroral forms seen by the 630 nm all-sky camera, are marked for the on left panel if and when they appear in the merdian scanned by the photometer, using the same lettering code as Fig.6

Fig. 6 .
Fig. 6.A series of all-sky images by the 630 nm camera at Longyearbyen, Svalbard on 17 January 1997: (1) 08:33:14 UT; (2) 08:49:38; (3) 08:58:40; (4) 09:03:21; (5) 09:05:41; (6) 09:07:02; (7) 09:08:22; and (8) 09:14:23.Various auroral forms (including both long-lived structures and transient events) are labelled A-I b Fig. 7a±f.A series of schematics for a 08:25; b 08:40; c 09:00; d 09:07 (although, in fact this situation is unlikely to be present at all); e 09:10; f 09:14.The left hand plots are cross sections of the magnetosphere in a plane parallel to the XZ plane, viewed form dusk with the Sun to the left.The centre plots are views from the Sun and the right hand plots are views of the Northern Hemisphere ionosphere from above with noon to the top, dawn to the right and dusk to the left.Note that in the middle plots the lobe reconnection sites (c and d) are shown, but lobe and overdraped lobe ®eld lines are omitted for clarity.Reconnection sites are labelled a, b, c and d and are dots in the lefthand ®gures and solid lines in the centre ®gures: their projections into the Northern Hemisphere ionosphere are solid lines in the right-hand ®gures.Lines with arrows are magnetic ®eld lines in the left and centre ®gures, and ¯ow stream lines in the right-hand ®gures.The circle shows the ®eld-ofview of the 630 nm all-sky camera at Longyearbyen and éieroset et al. (1997) and with a lobe reconnection site which initially mapped to the openclosed ®eld line boundary, as predicted by éieroset et al.