B.: Ionospheric flow during extended intervals of northward but By-dominated IMF, Ann

Abstract. We present SuperDARN radar observations of the nightside high-latitude ionospheric flow during two 6-hour intervals of quasi-steady northward interplanetary magnetic field (IMF). During both intervals (01:30–07:30 UT on 2 December and 21:00–03:00 UT on 14/15 December 1999), the solar wind and IMF remained relatively steady with Bz positive and By negative, such that the IMF clock angle was ~ - 50° to - 60°. Throughout both intervals the radar data clearly indicate the presence of a highly distorted By-dominated twin cell flow pattern, indicative of an open magnetosphere, which is confirmed by DMSP and auroral data. Estimates of the changes in open flux present during each interval indicate approximately balanced dayside and nightside reconnection at rates of ~ 30–35 kV over the full 6 h. However, strong bursts of flow with speeds of over ~ 1000 ms-1 are observed near magnetic midnight on time scales of ~ 1 h, which are associated with increases in the transpolar voltage. These are indicative of the net closure of open flux by recon-nection in the tail. During one large flow burst, the night-side reconnection rate is estimated to have been ~ 1.5 times the dayside rate, i.e. ~ 45–60 kV compared with ~ 30–40 kV. Magnetic bays, which would indicate the formation of a sub-storm current wedge, are not observed in association with these bursts. In addition, no low-latitude Pi2s or geostationary particle injections were observed, although some local, small amplitude Pi2-band (5–50 mHz) activity does accompany the bursts. Coincident measurements of the flow and of the low amplitude magnetic perturbations reveal nightside ionospheric conductances of no more than a few mho, indicative of little associated precipitation. Therefore, we suggest that the flow bursts are the ionospheric manifestation of bursty reconnection events occurring in the more distant geomagnetic tail. The main implication of these findings is that, under the circumstances examined here, the convection cycle is not equivalent to the usual substorm cycle that occurs for southward IMF. Key words. Ionosphere (plasma convection; ionosphere-magnetosphere interactions) – Magnetospheric Physics (magnetotail)


Introduction
The dynamics of the Earth's coupled magnetosphereionosphere system is largely influenced by the orientation of the interplanetary magnetic field (IMF).More precisely, it depends primarily on the strength and orientation of the transverse GSM y and z components of the field, whose direction is given by the "clock angle", θ , with respect to north.Thus, a 0 • clock angle corresponds to a transverse field purely in the +z direction (northward), ±90 • to ±y (respectively), and ±180 • to −z (southward).When the interplanetary field points south, i.e. when |θ |>90 • , the dynamics of the system is dominated by Dungey-cycle flow and the substorm cycle (Dungey, 1961;Hones, 1979;Farrugia et al., 1993;Baker et al., 1996).Reconnection at the lowlatitude magnetopause between the geomagnetic field and the IMF results in the formation of open flux tubes, which are transported anti-sunward by the solar wind and stored in the lobes of the geomagnetic tail.This process is dominant during the growth phase of substorms, and is associated with twin-vortex flow in the polar ionosphere, which is strongest on the dayside and is related to a concurrent expansion in the area of open flux in the polar cap (Siscoe and Huang, 1985;Cowley and Lockwood, 1992).Rapid excitation of polar convection during intervals of southward IMF is now a well-established phenomenon (e.g.Etemadi et al., 1988;Ruohoniemi et al., 1993;Ridley et al., 1998;Khan and Cowley, 1999), and even the sub-structure of individual magnetopause reconnection events ("flux transfer events") is wellresolved in dayside flow and auroral data (e.g.Lockwood et al., 1989;Elphic et al., 1990;Pinnock et al., 1993;Provan et al., 1998;Neudegg et al., 1999;Milan et al., 2000).Subsequently, reconnection of lobe flux tubes in the tail current sheet results in open flux closure and field line contraction back towards the Earth, thus completing the Dungey cycle.
It has long been supposed that the tail process dominates during the expansion and recovery phases of substorms.Recent results from the Geotail spacecraft indeed show that tail reconnection typically begins in the dusk sector plasma sheet at down-tail distances of ∼20-30 R E within a few minutes of expansion phase onset signatures appearing on the ground, and subsequently expands into the dusk and midnight sectors (Nagai et al., 1998;Petrukovich et al., 1998;Machida et al., 1999).This process should also excite twin-vortex flow in the polar ionosphere (additional to that driven simultaneously by magnetopause processes), but with the flow now being strongest on the nightside and related to a concurrent contraction in the area of the open flux in the polar cap.Evidence for flow excitation during substorm expansions has recently been presented by Opgenoorth and Pellinen (1998) and Grocott et al. (2002), though matters are generally complicated by additional flow suppression and polarisation effects associated with the strong conductivity gradients in the substorm-disturbed nightside ionosphere (e.g.Kirkwood et al., 1988;Morelli et al., 1995;Weimer, 1999;Gjerloev and Hoffman, 2001;Khan et al., 2001).
When the interplanetary field points north, it is well established that Dungey-cycle flow and substorm activity are reduced (e.g.Fairfield and Cahill, 1966;Reiff et al., 1981), though observations in the dayside ionosphere suggest that open flux tube production does not switch off entirely until the clock angle falls below ∼30 • -40 • (e.g.Sandholt et al., 1998a, b).At the same time, high-latitude reconnection between lobe field lines and the IMF begins, exciting additional flow cells, particularly on the dayside (e.g.Dungey, 1963;Russell, 1972;Reiff and Burch, 1985;Bristow et al., 1998;Sandholt et al., 1999Sandholt et al., , 2000Sandholt et al., , 2001)).On the nightside under these conditions a variety of phenomena have recently been reported, whose physical connections have yet to be established.Huang et al. (2001) reported observations of quasisinusoidal nightside flow oscillations in SuperDARN radar data during an extended interval of modest northward IMF and small (typically <45 • ) clock angle.The flows had amplitudes ∼500 m s −1 and a period of ∼1 h, and were associated with the formation of large-scale clockwise flow vortices in the post-midnight sector which subsequently propagated toward dawn.The authors associate these flows with global oscillations of an essentially closed tail.Walker et al. (1998Walker et al. ( , 2002)), on the other hand, have reported the occurrence of latitudinally-restricted (∼1 • north-south) bursts of highspeed (∼2 km s −1 ) westward flow in the pre-midnight sector under similar interplanetary conditions (weak IMF with small clock angle).These flow bursts had a ∼12 min recurrence time (four main cycles were observed), and appear to have occurred at the equatorward edge of the dusk convection cell.Walker et al. interpret the bursts as being due to sporadic energy release and field dipolarisation in the geomagnetic tail associated with a viscously-driven twin-vortex flow system.It was noted by Sutcliffe (1998) that Pi2 and weak (few tens of nT) midnight-sector bay activity occurred during these events.
In yet another apparent class of nightside phenomena, bursty flow events having larger spatial scales and longer time scales have been reported.De la Beaujardière et al. (1994) report the occurrence of bursts of equatorwarddirected flow in nightside Sondrestrom radar data, which took place during a "quiet-time" interval in which the transverse components of the IMF were small and directed mainly northward.These flow bursts were initiated near the nightside open-closed field line boundary, had amplitudes up to several 100 m s −1 , and recur at ∼1 h intervals, lasting on each occasion for a few tens of minutes.Simultaneous auroral data show that each flow pulse was accompanied by an activation of the poleward-most arc system, in which a new east-west aligned arc formed poleward of the existing boundary, and subsequently moved equatorward, together with existing precipitation structures.This behaviour strongly suggests an origin in bursts of reconnection in the "quiet-time" tail.Poleward boundary auroral intensifications (PBIs) and arc "bifurcations" have subsequently been shown to occur commonly under a wide variety of conditions, both during "quiet" times, and in the growth, expansion and recovery phases of substorms (Lyons et al., 1999), often leading to the ejection of one or more north-south aligned auroral forms (called "auroral streamers") towards lower latitudes (Henderson et al., 1998).These longitudinally-restricted "streamers" have themselves been associated with the occurrence of narrow "bursty-bulk flow" (BBF) channels of earthwarddirected flow in the plasma sheet (Baumjohann, 1990;Angelopoulos et al., 1992), Pi2-band (5-50 mHz) activity, and longitudinally-localised injections of energetic particles at geostationary orbit (Sergeev et al., 1999).The relationship between the "streamer" structures and the large-scale ionospheric flow has yet to be investigated in detail.However, Senior et al. (2002) have recently reported the occurrence of large-scale bursty flows in the "quiet time" nightside ionosphere during an interval of "intermediate" clock angles >45 • .These flows have a recurrence time of ∼1 h, with substructure on tens of minutes time scales, similar to the bursts reported by de la Beaujardière et al. (1994), and take the form of surges of westward "return" flow in the dusk convection cell, several degrees wide in latitude, and extending in local time from post-midnight towards dusk.
From the above discussion it will be clear that the nightside ionosphere and geomagnetic tail for northward IMF is often anything but "quiet", and is instead characterized by a wealth of flow and precipitation phenomena having a variety of characteristics.However, the relationships between many of these phenomena have not yet been clarified, and no overall picture has yet emerged of the behaviour of the tail and conjugate ionosphere under these conditions.In this paper we aim to contribute to this issue through the study of Su-perDARN radar observations of nightside flows and related parameters during two extended intervals in which the IMF clock angle remained relatively steady at intermediate values θ ∼−50 • to −60 • .Thus, IMF B z was positive, but the largest transverse field component was IMF B y , which was consistently negative.These intervals (on 2 and 14/15 December 1999) were sufficiently extended (∼6 h each) that we may reasonably expect the magnetosphere to have reached a state entirely characteristic of those conditions, rather than, for example, being significantly influenced by prior intervals of differently-directed IMF.For these northward, but "intermediate" orientations of the IMF, we may thus expect on the basis of the above discussion that open flux production at the magnetopause persisted at modest rates, and indeed we show below that IMF B y -distorted twin-cell flow persisted throughout both intervals, with transpolar voltages typically in the range ∼20-40 kV.We might also anticipate the excitation of "lobe" convection cells on the dayside.However, because we have chosen to study intervals with good radar coverage of the nightside hours, coverage of the dayside flows is not extensive in these data.Thus, the presence or absence of lobes cells is not well determined by these data and does not form a major topic of discussion.
With regard to nightside phenomena, however, a significant question exists concerning the magnetospheric response to modest but steady dayside driving under these conditions.During the growth phase of substorms with the IMF directed southward, the transpolar voltage (and implied open flux input to the tail) is typically twice the above values (i.e.∼60-80 kV), and generally ∼30-40 min elapses before reconnection and plasmoid formation is excited in the near-Earth tail at substorm expansion phase onset.We may then ask whether substorms occur with reduced frequency or with reduced amplitude under conditions of reduced but continuous driving.In fact, we show in this paper that no classic substorm features were observed at all during these long intervals.Instead, large-scale bursts of flow were observed in the nightside ionosphere, having recurrence time scales of ∼1 h and lasting a few tens of minutes.These properties suggest a connection with the flow phenomena previously described by de la Beaujardière et al. (1994) and Senior et al. (2002), and thus also, possibly, with BBF/PBIs.In this paper we examine these flow observations in some detail, and discuss the implications for the balance between open flux production at the magnetopause and closure in the tail during northward IMF intervals of intermediate clock angle.

Instrumentation
In this section we provide brief details of the instrumentation employed in this study, which concerns the intervals 01:30-07:30 UT on 2 December and 21:00-03:00 UT on 14/15 December 1999.

SuperDARN radar data
The measurements of ionospheric convection velocities provided by the Super Dual Auroral Radar Network (Super-DARN) represent the central focus of our study.SuperDARN is an international array of HF coherent radars spanning the auroral regions of both the Northern and Southern Hemispheres (Greenwald et al., 1995).At the time of the intervals described here, the Northern Hemisphere network (used primarily in this study) consisted of six radars, whose fields of view extended ∼12 h of magnetic local time (MLT).This is illustrated in Fig. 1a, which shows the fields of view of the six radars at the centre times of each of the two intervals, i.e. 04:30 UT on 2 December, and 00:00 UT on 14/15 December.With increasing magnetic local time, the fields of view in each map are of the Saskatoon (T), Kapuskasing (K), Goose Bay (G), Iceland West (W), Iceland East (E), and Finland (F) radars.These are shown on geomagnetic grids extending from the pole to 60 • magnetic latitude, with noon at the top and dusk to the left.The coordinates are Altitude Adjusted Corrected Geomagnetic (AACGM) coordinates, a development of the PACE system discussed by Baker and Wing (1989).The fields of view of those radars from which line-ofsight data will be displayed below are shaded blue.These are the Saskatoon, Kapuskasing, Goose Bay, and Iceland West radars for the first interval, and the Goose Bay, Iceland West, Iceland East and Finland radars for the second.Data will be displayed from the beam directions marked by the red lines.It should be noted, however, that the fields of view of the radars changed significantly with respect to MLT over each interval, and encompassed the whole nightside ionosphere at ∼02:00 UT.Continuous observations of the whole nightside over an interval which extends on either side of this time are, therefore, not possible, a limitation which is reflected in the analysis of the flow data below.Data from the Southern Hemisphere network (which consisted of the four radars Tiger (R), Halley (H), Syowa (J), and Syowa East (N)) are also discussed below, although no detailed analysis is performed due to limited coverage.Their fields of view are, therefore, not shown here.
During both study intervals, all ten radars were operating in a standard mode in which each radar scans through 16 beams of azimuthal separation 3.24 • , with a total scan time of 2 min.Each beam is divided into 75 range gates of length 45 km, and so in each full scan the radars cover 52 • in azimuth and over 3000 km in range.Large-scale maps of the high-latitude convection have been derived from this data using the "Map Potential" model developed by Ruohoniemi and Baker (1998).The line-of-sight velocities are mapped onto a polar grid, and used to determine a solution for the electrostatic potential which is expressed in spherical harmonics.The equipotentials of the solution represent the plasma streamlines of the modelled convection pattern.Information from the statistical model of Ruohoniemi and Greenwald (1996), parameterised by concurrent IMF conditions, is used to stabilize the solution where no data are available.Here, spherical harmonic fits, which use terms up to sixth order, are employed to fit the Northern Hemisphere data, with the functions being expressed about a pole that has been shifted by 2 • towards the nightside.A lower-latitude convection limit of 67 • relative to this pole has also been imposed, as deemed appropriate for the contracted nature of the flow pattern during these two intervals (see Ruohoniemi and Baker (1998) for a detailed discussion of the Map Potential algorithm).The flow vectors which will be shown superposed on the electric equipotentials patterns are derived using the SuperDARN line-of-sight velocity measurements and the transverse velocity component from the spherical harmonic fit.

Ground magnetometer data
Figure 1b shows the locations of the ground magnetic stations employed, indicated in the figure by the black circles on the same magnetic grids and at the same times as in Fig. 1a.Data have been inspected from all the magnetograms of six Northern Hemisphere arrays which span the same 12 h of MLT as the SuperDARN radars.These are the Canadian CANOPUS array (Grant et al., 1992), the Geological Survey of Canada array, the MACCS array (Hughes and Engebretson, 1997), the Greenland magnetometer chain (Friis-Christensen et al., 1985), the Scandinavian IMAGE array (Lühr et al., 1998), and the SAMNET array (Yeoman et al., 1990).Data from selected stations will be presented in H , D, Z coordinates, where H is local magnetic north, D is local magnetic east, and Z is vertically down.These stations are indicated in Fig. 1b by the coloured circles, and their iden-tifier codes are also shown for reference.Data from those at lower latitudes and colour-coded red (i.e.PBQ, NAQ, and FAR) have been band-pass filtered (20-200 s), to provide information on substorm-related Pi2 signals.Those at latitudes ∼70 • -75 • and colour-coded blue (i.e.CON, IQA, and BJN) are distributed in longitude over the local time extent of the radar array, to give an indication of the overall magnetic activity during each interval.We will also show data from a number of sub-arrays, colour-coded yellow, which span individual features at much higher spatial resolution, thus enabling a more detailed analysis.

LANL geostationary spacecraft data
The occurrence of energetic ion and electron injections at geostationary orbit has been examined using data from the SOPA instruments carried by Los Alamos National Labo-ratory (LANL) spacecraft (Belian et al., 1992).Data from three operating spacecraft, 1994-084, 1989-046, and LANL-97A have been examined.During both intervals, LANL-97A was located in the post-midnight sector, as appropriate for monitoring electron injections.This data is, therefore, presented here, being representative of the entire data set.The red coloured triangle on each map in Fig. 1b shows the ionospheric footprint of LANL-97A, mapped from geostationary orbit using the T-89 model (Tsyganenko, 1989).The black solid line shows the footprint of a complete orbit.

Polar auroral images
Images of the aurora were provided by the VIS instrument on the Polar spacecraft (Frank et al., 1995) during the interval on 2 December.This instrument viewed the whole of the northern polar ionosphere from the beginning of the interval at 01:30 UT up to ∼03:00 UT, and provided images approximately every 5 min.These consist of 32.5 s integrations of UV photons at a wavelength of 130.4 nm.No images were available for the interval on 14/15 December.

DMSP auroral particle data
Measurements of ion and electron fluxes by the SSJ/4 instrument on board the DMSP F12, F13, and F14 spacecraft (Hardy et al., 1984) have been employed to examine the patterns of polar particle precipitation and their relationship to the plasma flow.The instrument points toward zenith at all times, and provides 1 s resolution spectra of ion and electron fluxes between 30 eV and 30 keV.

Upstream interplanetary parameters
Interplanetary conditions during each interval were monitored by the ACE spacecraft (Stone et al., 1998), located at GSM coordinates (X, Y , Z) = (222, 21, −25) R E during the 2 December interval and (X, Y , Z) = (226, 36, −15) R E during the 14/15 December interval.The variation of the spacecraft position over each interval was not significant.Solar wind and IMF data were obtained by the SWEPAM and MAG instruments, respectively (McComas et al., 1998;Smith et al., 1999).The propagation delay of field changes from ACE to the dayside ionosphere has been estimated using the algorithm of Khan and Cowley (1999).This estimate includes the propagation time in the solar wind upstream of the bow shock, the frozen-in transit time across the subsolar magnetosheath, and the Alfvénic propagation time along open field lines from the subsolar magnetopause to the cusp ionosphere.With solar wind proton number densities n p ∼15 cm −3 and speeds v p ∼330 km s −1 , we find a propagation delay of 76±3 min for the 2 December interval.For 14/15 December, n p was ∼5 cm −3 and v p ∼380 km s −1 , giving a propagation delay of 68±5 min.These times have been used to lag the ACE solar wind and IMF data displayed here.It should be noted that during the 2 December interval, no data were available from ACE/SWEPAM from ∼03:30 to ∼06:30 UT.Therefore, we have used data from the Geotail The top two panels show solar wind density and velocity data (the dashed line represents data from Geotail, which was used where data was missing from ACE).The middle three panels show IMF data in GSM coordinates, and the bottom two panels show the field magnitude and clock angle.The clock angle is defined with respect to north, such that 0 • corresponds to a transverse field purely in the +z direction (northward), ±90 • to ±y (respectively) and ±180 • to −z (southward).
CPI instrument (Frank et al., 1994) to give an indication of n p and v p during this period.A negligible propagation delay existed between Geotail and the ionosphere, due to Geotail's location at (X, Y , Z) ≈ (15, −20, 2) R E , consistent with features observed in both the ACE and Geotail data sets from before and after the ACE/SWEPAM data dropout.
3 Observations overview: Interval 1 (01:30-07:30 UT 2 December 1999) In this section we discuss the ground and space observations made during the first of the two 6-hour intervals examined here, on 2 December 1999.We begin with an overview of the interplanetary data, and subsequently look in more detail at the magnetic, flow, particle, and auroral data.G round S catter S as katoon (0) K apus kas ing (5) K apus kas ing (15) G oos e B ay (6) Iceland Wes t (8) F ig. 3

Upstream interplanetary conditions
In Fig. we present data from the ACE spacecraft, lagged by 76 min via the method described above, for the 18-h interval from 19:30 UT on 1 December to 13:30 UT on 2 December 1999.This is an expanded interval describing the upstream interplanetary conditions for 6 h on either side of the interval of specific interest, which is delimited by the vertical dashed lines.The top two panels show the solar wind proton number density (n p ) and speed (v p ) from ACE/SWEPAM.Data were unavailable for the ∼3-h period, ∼03:30-06:30 UT, as indicated above, so data from the Geotail CPI instrument has been used to fill the gap (indicated on Fig. 2 by the dashed lines).Before the data gap, the density remained relatively constant at ∼10 cm −3 and the velocity between 330-380 km s −1 .The Geotail data indicates a decrease in velocity of ∼30 km s −1 approximately coincident with the SWEPAM data dropout.The middle three panels show the GSM components of the IMF from ACE/MAG, with the magnitude and clock angle of the field shown in the bottom two panels.IMF B x was near zero or positive throughout the main interval of interest, while B y was large and negative, and B z predominantly positive (and had been near zero or positive for more than 3 h previously).These values translate to a field strength which was nearly constant at ∼10±1 nT, and a clock angle which, for the most part (∼02:30-06:30 UT), remained near ∼−60 • .The mean value over the main interval was −59 • , with a standard deviation of 17 • .

SuperDARN flow data overview
In Fig. 3 we present line-of-sight SuperDARN velocity data for the 6-h interval, from the five radar beams shown by the red lines in the left-hand map in Fig. 1a.These beams have been chosen for their good backscatter returns, as well as for their wide coverage of the nightside ionosphere.Immediately evident is the continuous flow observed over the entire interval, with speeds of ∼500-1000 m s −1 .These flows are directed away from the radar at Saskatoon (beam 0) and Kapuskasing (beam 5), indicative of sunward "return" flow in the dusk convection cell, towards the radar at Kapuskasing (beam 15) and Goose Bay (beam 6), indicative of antisunward flow over the polar cap into the auroral zone, and strong but variable (directed mainly east-west) at Iceland West (beam 8).There is also evidence of strong variability or pulsing in the flow, which is particularly marked, for example, at Goose Bay (beam 6).The vertical dashed lines indicate six such "pulsed" features in the flow, the details and timing of which will be discussed below.It is immediately evident that this is an interval of dynamic and often high speed nightside flow.
In order to investigate the nature of the flow further, we have derived a sequence of convection maps from all the available SuperDARN data (in both hemispheres), using the Map Potential model.In Fig. 4a we present four representative examples of the Northern Hemisphere flow pattern, imaged at 02:00, 04:00, 06:00 and 07:00 UT, which employ 2min resolution data from the six radars indicated in Fig. 1a.The diagram in the bottom right-hand corner of each plot indicates the direction and strength of the concurrent (lagged) IMF projected onto the y-z plane, emphasising again that al-though these flow data span 5 h of UT, the IMF conditions remained very similar.Correspondingly, although the details vary from plot to plot, the overall flow pattern also remained similar.It can be seen that the convection was consistently dominated by a large crescent-shaped "dusk" flow cell, which spanned the local time interval roughly from noon (the details being inaccessible due to the lack of post-noon dayside data), via dusk into the post-midnight hours.A clear "dawn" flow cell is present only in the map at 04:00 UT, where it is confined essentially to the dayside MLT quadrant between dawn and noon.In general, the dawn sector flows are highly structured and variable, and appear to be essentially absent, for example, at 06:00 UT.The sense of the dawn-dusk flow asymmetry is exactly that expected for Dungey-cycle flow driven in the presence of strong and consistent negative IMF B y (Jørgensen et al., 1972;Cowley, 1981;Reiff and Burch, 1985;Heppner and Maynard, 1987;Weimer, 1995;Ruohoniemi and Greenwald, 1996).In par- ticular, the predominant sense of anti-clockwise vortical flow within the polar cap, giving faster antisunward flows at dusk than at dawn, is that expected to be produced by the east-west field tension effect on open field lines, which pulls newlyreconnected field lines within the dayside cusp eastward into the dusk sector.Despite the northward direction of the IMF, there is little evidence in these maps for the concurrent presence of high-latitude "lobe" convection cells.However, as mentioned above, the lack of data coverage in the post-noon dayside quadrant makes this statement less than definitive.
If the Northern Hemisphere convection is due principally to IMF B y -asymmetric Dungey-cycle flow, we would expect to observe similar flows in the Southern Hemisphere, but with an oppositely-directed asymmetry between dawn and dusk.Examination of the Southern Hemisphere data shows that the coverage is much more restricted than in the north, and backscatter is essentially absent after ∼06:00 UT.Therefore, in Fig. 4b we show four selected maps from earlier in the interval, at 02:04, 03:00, 04:08, and 05:00 UT, chosen to optimise the coverage.In order to aid comparison, these maps are drawn as though viewed from the north, i.e. with noon at the top and dusk to the left.Although the data are restricted, they are nevertheless compatible with a flow asymmetry which is opposite to that which occurs in the north.In particular, the data suggest the continuous presence of strong antisunward and westward flows in the post-midnight and dawn sector polar cap, in a region where the Northern Hemisphere flows are more sporadic but generally directed eastwards.
As a further check on the nature and origin of the flow patterns, we have also examined all available DMSP particle data and VIS auroral imagery during the interval, in order to determine the nature of the concurrent particle precipitation patterns (though, as noted above, VIS images are available only until ∼03:00 UT).On the basis of the above interpretation, we would expect the central region of the convection patterns where the flow is directed antisunward but with an anti-clockwise vortical twist (in the Northern Hemisphere) to correspond to open field lines with cusp/mantle (on the dayside) or polar rain (on the nightside) precipitation, while the surrounding region at lower latitudes should correspond to closed field lines with trapped BPS/CPS pre-   The mean transpolar voltage for the entire interval was 35 kV, corresponding, of course, to a magnetic flux throughput in the flow system of ∼35 kWb s −1 .If this flow was associated with essentially unbalanced dayside reconnection and open flux production, the amount of open flux that would be created over the 6 h interval (across a small merging gap) would be ∼8•10 8 Wb, thus essentially tripling the amount of open flux present in the system.Clearly this does not happen.Similarly, if the flow was associated with essentially unbalanced nightside reconnection at this rate, the polar cap would contract to zero size in ∼3 h.Clearly this also does not happen.(In addition, we would not expect the IMF B yassociated flow asymmetry to persist in this case.)Rather, if we set a rough limit for the change in open flux over the interval of say, ∼1•10 8 Wb, corresponding to a change in the overall boundary position of ∼2 • , the implied difference in the dayside and nightside reconnection rates averaged over the 6 h of the interval is ∼5 kV, small compared with the overall voltages deduced from the flow of ∼35 kV.The flow maps thus imply approximate balance between dayside and nightside reconnection rates, averaged over the whole interval, to within 10-15%.However, as we will show below, this does not imply that detailed balance occurs throughout, as the pulsed nature of the nightside flows noted above already suggests.

Overview of substorm indicators
Given the implied existence of nightside reconnection at an appreciable averaged rate during this interval, it is of interest to enquire whether this process is expressed in terms of substorm expansions, as is characteristic of intervals of southward-directed IMF.Data relevant to this issue are summarized in Fig. 6.The top three panels present H component magnetograms from the azimuthal magnetometer array at ∼71 • -73 • latitude (colour-coded blue in Fig. 1b), which we examine for the presence of magnetic signatures, indicative of the formation of substorm electrojets.These stations are at a higher latitude than those at which substorm expansion onsets usually occur during or directly after intervals of southward IMF (∼65 • ), but they are relevant in the present context in view of the contracted nature of the flow patterns displayed in Figs. 4 and 5.We also confirm that they are representative of all the data from all the networks surveyed in this study.These data show that the geomagnetic field was "quiet" during the whole interval, with no indications of impulsive substorm activity.Field variations were continuously present on time scales of several tens of minutes, but the amplitudes were typically only a few tens of nT.There is some evidence of a relationship between these variations and those in the flow, marked by the vertical dashed lines, as in Fig. 3, and this will be discussed later.
The lack of impulsive substorm-like activity is emphasized in the next three panels of Fig. 6, which show representative magnetic field data from lower-latitude stations (colourcoded red in Fig. 1b), which have been filtered between 20-200 s, corresponding to the Pi2 band.None of these data (including those from stations not shown here) indicate any impulsive substorm-like Pi2 features, though the data from NAQ does indicate some continuous activity of amplitude ∼1 nT during the first ∼2 h of the interval.However, there does not appear to be any detailed correspondence between this low-level activity and the times of the flow features discussed above.Finally, the bottom panel of the figure shows representative data from the geostationary energetic particle monitors, specifically two channels of energetic electron data from spacecraft LANL97A, located in the dawn sector (see Fig. 1b).Small periodic flux variations are evident over the interval, but no injection signatures are present, or any obvious relationship to the flow features, as is indeed the case in the data from all the LANL spacecraft.Thus, if particle injections from the tail did occur, they did not reach geostationary orbit.Overall, therefore, we summarize our findings by stating that no indications of substorm bays, Pi2s, or geostationary orbit particle injections were observed during the interval.The tail reconnection which occurred was, therefore, not manifest as classic near-Earth substorm expansions.Nevertheless, impulsive activity was indeed observed in the nightside flow, as noted above, and in the next section we turn to examine its behaviour in more detail.

Nightside flow bursts
We now examine more closely the variability of the nightside flow, whose presence is already indicated in the data presented in Figs. 3 and 4. In Fig. 7 we extract parameters from the Map Potential plots which provide measures of the variability of the flow.The top panel shows the total transpolar voltage in the flow maps, representing a measure of the total flux transport in the flow system.The level of confidence in this parameter in general depends on the coverage of the polar ionosphere by the radar data.Examination shows that during this interval the data generally do span the extrema in the voltage values (see, e.g.Fig. 4), so that we may have reasonable confidence overall.It can be seen that the voltage values Inspection of the flow maps shows that these enhancements are manifest as localized bursts of fast flows, generally located in the westward "dusk" cell flow spanning the midnight sector.The second panel of Fig. 7 then shows a complementary measure of the flow variations, namely the peak nightside flow speed, occurring in each 2-min Map Potential plot specifically in these westward "return" flows.In order to provide a representative value, we have averaged the vector velocities in bins which are 220 km square, corresponding to 2 • latitude and typically ∼7 • longitude, representing four "pixels" of the Map Potential algorithm.Again, the bursty nature of the flow is evident.Typical values of ∼500-800 m s −1 are interspersed with ∼10-15 min bursts, reaching up to ∼1500 m s −1 .The averaged value of the peak flow speed over the interval, 891 m s −1 , is indicated by the horizontal dotted line in this panel, emphasizing the peaks in the flow.The most prominent and long-lived of these peaks are marked by the vertical dashed lines, these being the same lines as those shown previously in Figs. 3 and 6.Comparison with the data in the upper panel of Fig. 7 shows that most of the peaks in the speed also correspond to a peak in the transpolar voltage.This indicates that an overall enhancement in the flux transport took place in association with the flow bursts, and not just a reconfiguration in the flow pattern.
The third panel of Fig. 7 shows the vector components of the flow associated with the peak speed shown in the second panel.The solid line shows the westward flow component (perforce positive, since the peak flow speed displayed corresponds specifically to the westward flow of the "dusk" return cell), while the dashed line shows the north-south flow (positive equatorward).These data show that, with the ex-ception of the burst peaking at 07:16 UT, the peak flows all occur in the westward component.Concurrent changes in the north-south component are also observed, however, with enhancements in the equatorward component often occurring in the vicinity of the peak in the westward flow.The lower two panels of the figure also show the magnetic latitude and MLT of the location of the peak flow speed.It can be seen that the latitude of the peak flow is generally located in the range 73 • -75 • , and shows little trend during the interval, consistent with our previous discussion concerning the lack of overall change in the size of the flow system.Similarly, the local time of the peak flow speed is scattered ∼2-3 h on either side of midnight, and also shows no overall preference or trend.

Flow burst geometry
Having investigated the overall nature of the nightside flow data, we now examine in more detail the nature of the flow variations which occur in two of the flow bursts, those centred near (and indicated by the vertical lines at) 03:00 and 05:18 UT. Figure 8a shows eight selected 2-min flow maps which span the first of these intervals, between 02:36 and 03:20 UT.The black triangles in these plots show the positions of key ground magnetometer sites whose data will be examined below.The first map, for the 2-min interval starting at 02:36 UT, is illustrative of the nature of the flow over the previous ∼20-min interval.It shows a well-formed IMF B y -asymmetric flow pattern of modest intensity, in accordance with the overview presented above, with a welldeveloped crescent-shaped "dusk" flow cell stretching from noon via dusk to the post-midnight hours, together with a weaker rounder "dawn" flow cell which is centred in the prenoon sector.The transpolar voltage is estimated to have been 35 kV.Over the next ∼15 min, the dusk cell flow then grew in magnitude, with fast equatorward flows near midnight, and fast westward "return" flows in the pre-midnight sector, which eventually extended to the edge of the radar field of view at ∼19 MLT.The transpolar voltage reached 44 kV at 02:52 UT, and remained at around that value for the next ∼20 min.Rapid equatorward and westward flows were also observed in the dawn sector after this time, which, bearing in mind the asymmetric nature of the flow pattern, could also have been connected with the enhancements observed closer to midnight.The flow magnitude fell rapidly after 03:12 UT, however, such that the transpolar voltage reduced to values of ∼35 kV by 03:20 UT, similar to that occurring before the burst.
Figure 8b shows in the same format the development of a second flow burst during the interval from 05:02 to 05:30 UT.The first map, starting at 05:02 UT, shows a weak flow pattern with a transpolar voltage of 31 kV, which is dominated by clockwise vortical "dusk" cell convection.The "return" flows in the pre-midnight sector were relatively strong, however, as had been the case for the previous ∼20 min.These flows persisted until ∼05:06 UT, when an enhancement began in the pre-midnight sector, which developed over the next ∼15 min into a predominantly azimuthal flow feature spanning some 6 h of MLT about the midnight meridian (within the fields of view of the radars), and containing flows which peaked at ∼1500 m s −1 .Peak estimated transpolar voltages of ∼40 kV occurred at ∼05:20 UT, the flows then subsequently declining to reach transpolar voltages of ∼30 kV once more after ∼05:30 UT.
Overall, the geometry of these two flow burst events are seen to be similar to each other, both involving a strong enhancement in the nightside westward dusk cell "return" flow.However, in the second case the region of rapid equatorward flow out of the polar cap appears to have shifted from near midnight into the post-midnight sector, such that the region of enhanced westward flow also moved eastward to span the midnight sector.In Fig. 8c we show a gallery of flow maps taken from the centres of each of the remaining bursts that occurred during the interval, at the times indicated by the vertical dashed lines in Fig. 7. Within the limits of the radar data coverage, each of them displays an enhancement in the westward dusk cell "return" flows in the pre-midnight-tomidnight sector, with enhanced equatorward flows out of the polar cap in the midnight-to-post-midnight sector.The burst centred at 01:44 UT also shows the presence of rapid equatorward flows in the dawn sector, with westward flows extending into the pre-midnight sector, which may be a related phenomenon.

Bursts as nightside reconnection events
The geometry of the flow surges revealed by the above analysis suggests that they are formed by bursts of enhanced re-connection in the geomagnetic tail.If so, we might expect them to be associated with overall poleward contractions in the open-closed field line boundary and in the flow pattern.It is evident from the flow maps shown in Fig. 8 that only in certain cases do the flow data consistently define a sharp flow reversal boundary whose motion can be reasonably clearly followed over the duration of the burst, and then only over a limited local time interval in the pre-midnight sector.It is primarily for this reason that we make the following analysis of just the first of the two bursts discussed above.In Fig. 9  which suggest a sudden drop in the latitude of the boundary appear to be associated with an additional, smaller scale flow feature near midnight and are, therefore, unlikely to be representative of the location of the boundary elsewhere.Despite this, and the evident scatter and the quantization, these data do suggest the occurrence of a small poleward contraction of the boundary across the event, by ∼1 The implied difference in rates is by several tens of kV averaged over the duration of the burst.In Sect. 5 below we perform a more detailed analysis and attempt to quantify the dayside and nightside reconnection rates during the burst interval within the limitations of the available data.

Burst magnetic signatures
Although the above data suggest that the observed flow bursts are due to intervals of enhanced reconnection in the tail, we have already demonstrated above that they are not associated with classic substorm features.However, small ampli-tude fluctuations do occur in the nightside magnetometer data during the interval, and we now investigate the relationship of these variations with those in the flow data.In Fig. 10 we present magnetometer data for each of the two bursts, from the high-spatial resolution array (yellow dots) shown in Fig. 1b.This array is also indicated by the black triangles on each of the flow maps in Fig. 8.For each burst we show the perturbation field superposed with the flow data from the 220 km square bin corresponding to the location of the magnetometer (such that the location of the bin corotates with the station).If the horizontal perturbations are due to overhead Hall currents, then they should correlate with the flow variations according to H ∝ H V W and D ∝ H V N , where H and D are the northward and eastward components of the magnetic perturbation, respectively (as before), V W and V N are the westward and northward components of the flow, and H is the height-integrated ionospheric Hall conductivity.Such simple relationships will not always hold, particularly in high spatially structured circumstances, due to the integrating nature of ground-based magnetic observations.However, they should be approximately valid here, where the spatial scale for variations in the flow is several plied value of H is just a few mho.Although as indicated above, it can in general be dangerous to relate local fields and flow in this way; we can confirm that only low-amplitude magnetic effects were observed over the whole array of stations examined here.This, together with the general lack of consistent correspondence between the field perturbations and the flow, suggests that the auroral precipitation during these events within the region of enhanced flow was very low, certainly by comparison with substorm intervals.Finally, panel (iii) contains the Z-component magnetic field data which show similar fluctuations to the other two components and a large negative deflection at ∼03:10 UT.This feature can also be identified in the H -and D-components, but has no clear signature in the flow data.
In Fig. 10b we show the same magnetometer data bandpass filtered between 20-200 s, as discussed above.Station RAN measured the largest amplitude activity in the H -and D-components, but this does not appear to be related to the time of the burst.At CHX, IQA, and BLC, however, there is some evidence of enhanced activity during the burst, although this is of a more continuous nature, unlike the impulsive Pi2-signatures typical of substorms.
Figures 10c and d present data in the same format for the second burst.The magnetometer and flow data in Fig. 10c show similar trends to those for the first burst, although there now appears to be a rather clearer correlation between the two data sets in this case.This is particularly evident in the H -component magnetometer and east-west flow data.Again, the amplitudes of the correlated perturbations are such that the values of H implied by these data are no more than a few mho, suggesting that, as in the previous example, no strong conductivity changes are present and hence, no strong precipitation.Figure 10d shows the filtered (20-200 s) magnetometer data for this burst.The general picture is much the same as for the first burst, with some small amplitude activity observed at certain stations.In particular, there appears to be a well-defined "onset" of low-amplitude activity observed at CHX and BLC associated with the central time of the burst, at ∼05:18 UT.

Interval 1 summary
We now summarize our main findings.During this interval, in which an IMF clock angle of ∼−60 • was maintained for ∼6 h, dayside reconnection continued to drive Dungey-cycle flow in the magnetosphere-ionosphere system.It is found that, over the entire interval, the size of the convection pattern remained relatively constant, implying that tail reconnection was also ongoing, at an average rate approximately equal to that on the dayside.However, the usual substorm cycle did not occur, evidenced by the lack of magnetic bay signatures at high latitudes, Pi2 signals, and geostationary orbit energetic particle injections.Instead, ∼1000 m s −1 bursts of flow were observed in the nightside ionosphere, lasting a few tens of minutes and recurring every hour or so, accompanied by weak magnetic signatures and low amplitude Pi2-band (5-50 mHz) activity.One burst studied in detail, for which there was a well-defined flow reversal boundary in the flow data, appeared to coincide with a ∼1 • -2 • contraction of the boundary.Thus, these data indicate that the bursts are associated with intervals of enhanced tail reconnection, during which the tail reconnection rate temporarily exceeds the dayside reconnection rate.
In this section we now discuss the ground and space observations made during the second of our 6-hour intervals, on 14/15 December 1999.

Upstream interplanetary conditions
In Fig. 11 we present the ACE interplanetary data in the same format as Fig. 2, lagged by 68 min, for the 18-hour interval from 15:00 UT on 14 December to 09:00 UT on 15 December 1999.The steady nature of the solar wind during the main region of interest is evidenced in the proton number density (n p ) and speed (v p ) data.The density remained relatively constant at ∼5 cm −3 and the velocity between 370-420 km s −1 .The IMF was also steady throughout the main interval.B x was near zero or negative, B y was large and negative, and B z was predominantly positive (and had been for the previous 6 h).This translates to a total field strength of ∼8±2 nT, and a clock angle with a mean value of −51 • and a standard deviation of 15 • .However, it may be noted that a small change in the field occurred near the middle of the interval (∼00:30 UT), such that the field magnitude and clock angle were modestly increased in the later period compared with the earlier.

SuperDARN flow data overview
Presented in Fig. 12 are line-of-sight SuperDARN velocity data from the five radar beams, shown by the red lines, in the right-hand map in Fig. 1a.In common with the previous interval, continuous flow was observed over the entire 6 hours, with speeds of ∼500-1000 m s −1 .These flows are predominantly directed towards the radar at Goose Bay (beam 13) and Iceland East (beam 12), indicative of westward "return" flows in an azimuthally extended dusk convection cell, and away from the radar at Iceland West (beam 3), indicative of sunward "return" flow in the same cell.At Finland (beam 4) for the early part of the interval and Iceland West (beam 8) for the later part, the flows are both towards the radar in the more poleward ranges and away from the radar in the near ranges, indicative of equatorward flow out of the polar cap with a flow reversal in the auroral zone.Again, there is also evidence of strong variability in the flow, which is particularly marked, for example, in the Iceland West data.However, this variability is of a less regular nature than that observed during the previous interval (see Fig. 3).The vertical dashed lines indicate episodes of high flow variability, which will be discussed further in Sect.4.3 below.To a first approximation, however, this is a similar interval to that discussed above, with dynamic and often high speed flows, driven during an interval of northward and B y -dominated IMF.
Inspection of data relevant to the identification of sub-storms (not shown here) also suggests that this is a similar interval to that described above.The H -component magnetometer data again shows only small field variations of a few tens of nT amplitude, indicative of a "quiet" geomagnetic field, similar to that on 2 December shown in Fig. 6.The lack of substorm activity is emphasized in the filtered data, none of which indicates any impulsive substorm-like Pi2 features.Representative data from the LANL97A SOPA instrument (located in the dawn sector, as shown in Fig. 1b) also indicate no injection signatures or indeed any significant variations in the particle fluxes.Overall, therefore, this interval appears to be as devoid of substorm activity as the first.
Nevertheless, as mentioned above, impulsive nightside activity was again observed in the flow, which we now go on to examine.

Nightside flow bursts
In Fig. 13 we present flow parameters extracted from the Map Potential plots as before.The voltage values, again shown in the top panel, lie typically between ∼20 and ∼40 kV.The average over the interval was 29 kV.This is somewhat less than the average during the first interval, possibly due to the slightly reduced IMF magnitude and clock angle.Nevertheless, the values do exhibit a similar variability.The plot of the peak nightside flow speed, however, appears to be divided into two distinct parts.The first half of the interval (up to ∼00:30 UT) appears relatively quiet, with typical values of the peak flow around ∼600-800 m s −1 (close to the averaged value of the flow speed over the full 6 h of 732 m s −1 ).Only one apparent "burst" occurred, at ∼22:40 UT.Then, after ∼00:30 UT, the peak flows resemble more those during the first interval, with regular burst-like features evident in the data, reaching speeds up to ∼1200 m s −1 .This change in flow characteristic could be associated with the modest concurrent change in IMF conditions which was noted in Sect.4.1.Those bursts which can be distinguished from these data (and from a close inspection of the flow maps) have been marked by the vertical dashed lines, corresponding to those in Fig. 12.Comparison with the data in the upper panel of Fig. 13 shows that the peaks which have been identified in the flow speed also generally correspond to peaks in the transpolar voltage.This is consistent with our conclusion about the bursts during the first interval, in that an overall enhancement in the flux transport took place in association with the flow bursts.However, it is also evident that substantial enhancements in the transpolar voltage occurred at times when no flow speed excitation is apparent in the nightside data.This generally occurred early in the interval, when the radar coverage included a significant portion of the dayside ionosphere, and, as can be inferred from the flow maps, the voltage enhancements were then significantly influenced by the dayside driven component of the flow.The vector components of the peak flow also show a similar trend to the corresponding data from the first interval.The bursts tend to peak in the westward direction (indicated by the solid line), with relatively little enhancement in the 21 00  22 00 23 00 00 00 01 00 02 00 03 00 UT MLT (hrs .)21 00  22 00 23 00 00 00 01 00 02 00 03 00 UT north-south component (dashed line).The latitude of the peak flow is generally located in the range 69 • -73 • (a few degrees lower than observed for the 2 December interval) and does not show significant trends over the interval.This is again consistent with our previous discussion concerning the lack of overall change in the size of the flow pattern.There is some scatter in the local time of the peak flow speed, although around the times of the bursts, specifically in the second half of the interval, the flows did tend to peak consistently within 1-2 h of midnight.

Flow burst geometry
As indicated above, the variability of the flows observed here differs somewhat from that observed during the first interval, particularly in the initial period.Although the background flow pattern is similar, being driven under the influence of large negative IMF-B y , the tail does not seem to respond with quite the same impulsive regularity as before.In Fig. 14a  "quiet" half of the interval, as described above.Although the details vary from plot to plot, the overall flow patterns remain similar to each other, and to that observed during the previous interval (see Fig. 4a).Due to the earlier UT of this interval, however, there is much better coverage of data in the post-noon sector, which shows the existence of dayside anti-sunward flows in the throat of the convection cells, indicative of dayside reconnection.There is also some evidence early on in the interval for the concurrent presence of high-latitude "lobe" convection cells.Later in the interval, in the map at 00:30 UT, for example, a "dawn" flow cell is also apparent, confined essentially to the same dayside MLT quadrant between dawn and noon as in the previous interval.In general, however, the lack of data in the dawn hours makes it difficult to draw conclusions about the nature of the flows in that sector.The extended nature of the dusk flow cell does, nevertheless, imply the existence of the expected dawndusk flow asymmetry, which is corroborated by the Southern Hemisphere data (not shown).These data again show much more restricted coverage than in the north, but are consistent with the opposite asymmetry in the flow.Examination of all available DMSP particle data for the interval provides similar corroboration of the open nature of the polar cap.A number of Southern Hemisphere overpasses by the F13 and F14 spacecraft, in particular, provided good evidence of a band of CPS/BPS precipitation at ∼75 • (observed at 21, 05, and 09 MLT), with lower energy polar cap precipitation polewards of this.Although this interval is, therefore, not identical to the first, it is important to point out that some bursts are still observed, particularly in the second half.The centre times of five enhancements in the flow are indicated in Fig. 13 by the vertical dashed lines at 22:40, 00:52, 01:30, 02:10, and 02:40 UT.As mentioned above, the first half of the interval sees only one distinctive burst (the start of which can be seen in the 22:30 UT map of Fig. 14a).In this instance, the burst takes the form of enhanced sunward return flows in the dusk convection cell.During the second half of the interval, when the flow variability was enhanced, the bursts begin to take the form of azimuthal flows in the midnight sector, more like the bursts observed during the first interval.However, they are noticeably less discrete than before, with the flows varying more randomly in magnitude.
The burst centred at ∼02:10 UT is perhaps the best example, and exhibits similar characteristics to the bursts shown in Fig. 8. Therefore, in Fig. 14b we show four flow maps which span this burst between 01:54 and 02:18 UT.The black triangles again show the positions of key ground magnetometer sites whose data will be examined below.The first map, starting at 01:54 UT, is illustrative of the nature of the flow over the ∼20-min period since the end of the previous burst.It shows an IMF B y -asymmetric flow pattern, with evidence of some enhancement of the westward flow near midnight, and a relatively low transpolar voltage of 24 kV.Over the next ∼20 min, the westward flow in the midnight sector grew in magnitude, peaking at 02:10 UT, with a corresponding peak in transpolar voltage of 38 kV.In the third map, showing this central time of the burst, the flow enhancement consists primarily of westward flows spanning magnetic midnight, similar to a number of the bursts observed during the first interval.Over the following ∼10 min the flows subsided, with the transpolar voltage falling to 24 kV by 02:18 UT.
The main conclusion we wish to draw is that, as in the first interval, continuous quasi-steady Dungey-cycle flow was driven during this 6-h interval of steady interplanetary conditions.This implies again that there must be a consistent nightside reconnection-driven element to the flow which is balancing the observed dayside driving, but which is not manifest as typical substorm phenomena.

Burst magnetic signatures
In Fig. 15 we present magnetometer data for the ∼02:10 UT burst from the high-spatial resolution array shown in the 00:00 UT map in Fig. 1b. Figure 15a presents unfiltered magnetic data along with the corresponding flow components, in the same format as in Fig. 10.Panel (i) contains the H -component magnetic field data (solid line), and the overhead westward component of the flow (dashed line).As found for the bursts during the first interval, there is no consistent correlation between the two data sets.Certain features are present in both, for example, an enhancement in the flow at the location of SKT is accompanied by an increase in the local H -component.In the D-component and the corresponding northward flow (dashed line), shown in panel (ii), there is a similar correlation.Again, however, values of ∼1000 m s −1 for the flow speed, and magnetic perturbations of at most ∼50 nT (at these or any of the stations  examined here), imply a Hall conductance of a few mho or less, indicative of a low level of auroral precipitation during these events within the region of enhanced flow.
In Fig. 15b we show the same magnetometer data band-pass filtered between 20-200 s.Station RAN measured the largest amplitude activity in the H -and D-components, but this does not appear to be related to the time of the burst.Indeed, none of the stations show discernable variability in the Pi2 band over the hour of the plot.

Interval 2 summary
The second of our two ∼6-h intervals occurred under similar interplanetary conditions to the first.The IMF clock angle remained relatively constant at ∼−50 • , somewhat smaller than for the first interval, and the field intensity remained between 6 and 8 nT, also somewhat smaller than before.Dayside reconnection again continued to drive Dungey-cycle flow, although it was observed to be weaker than during the first interval, possibly due to the reduced clock angle and field strength.As before, it was found that the size of the convection pattern remained relatively constant over the entire interval, thus implying that tail reconnection was again ongoing at an average rate approximately equal to that at the dayside.The lack of substorm phenomena was again evidenced in the magnetic and geostationary orbit energetic particle data, while a detailed investigation of the flow data revealed that similar nightside flow bursts occurred.However, only one clear burst was observed during the first half of the interval, with more frequent "bursty" activity occurring during the second half.This change in the nature of the flow coincided with a small change in the IMF conditions, in which the field magnitude was slightly increased and the clock angle slightly reduced.Nevertheless, the basic nature of the bursts appears to be similar to those observed during the first interval.

Reconnection rates during burst intervals
The analysis of the principal nightside flow bursts described above shows that the transpolar voltage is enhanced during the bursts, and also suggests (somewhat less strongly) that a modest poleward displacement takes place in the nightside open-closed field line boundary (see Fig. 9).These features indicate that the bursts are associated with enhancements in the nightside reconnection rate.In this section we analyse these data further, and attempt to quantify the reconnection rates within the limitations of the information available.The theoretical framework of our discussion is the Cowley and Lockwood (1992) paradigm of reconnection-associated flow excitation, leading to the reconnection-rate formulas derived previously by Grocott et al. (2002).
The physical situation which applies here is shown in Fig. 16 (where for simplicity the B y -associated asymmetry is not represented), in which we assume that dayside and nightside reconnection are both continuously present, but that the rate of flux closure on the nightside exceeds that on the dayside, such that overall, the polar cap contracts.In this diagram the segments of the open-closed field line boundary mapping to the reconnection sites (i.e. the "merging gaps") are shown by the dot-dashed lines, while the adiaroic segments of the boundary (which are not so connected and which thus move with the flow) are shown by the heavy solid lines.The lighter arrowed lines show the plasma streamlines.The action of the flow is to transport closed flux from the merging gap region of closure sunward to other local times, such as (in effect) to distribute the newly-closed flux around the boundary.In the Cowley-Lockwood paradigm the flux is distributed uniformly round the boundary, such that it contracts with its shape unchanged.A key feature of the diagram is that the foci of the plasma flow cells under these circumstances will be located near the ends of the nightside merging gap (see also, e.g.Siscoe and Huang, 1985).At these points the plasma flow across the boundary will reverse from being directed equatorward in the sector of the merging gap, to being poleward in the contracting adiaroic sector on either side.Thus, we expect in this case that the total transpolar voltage (as estimated from the SuperDARN radar data) will correspond to the ionospheric voltage across the nightside merging gap.By the same argument, when dayside reconnection dominates, the transpolar voltage will correspond instead to the ionospheric voltage across the dayside merging gap.These conclusions would not follow if, for example, significant voltages are also present associated with "lobe" convection cells.However, as indicated above, we have not found evidence for a substantial contribution from such flows cells in the data examined here.
Under these conditions, Grocott et al. (2002) have shown that the nightside and dayside reconnection rates averaged over some interval τ, ˙ n and ˙ d , respectively, can be estimated from  B during the event is shown on the horizontal axis.The dark-shaded area in the bottom right corner of each plot represents an area of inconsistent results where the dayside reconnection rate becomes negative.The horizontal dotted lines mark the likely range of f mg for the event (obtained from the SuperDARN maps), with the exterior region shown shaded.Also shaded is the region of unreasonable dayside reconnection rates as discussed in the text.The remaining unshaded area corresponds to the region of solutions that conforms to all of these constraints.These results are plotted for the 03:00 UT burst on 2 Dec. tive boundaries, ˙ n becomes increasingly negative as f mg and B both increase, while ˙ d drops to smaller values as f mg decreases and B increases.Eventually, ˙ d becomes negative in the dark-shaded area in the bottom right corner of the plot, this representing an area of inconsistent results (as discussed above) when (1 − f mg A becomes too negative.
We now consider the likely ranges of the controlling parameters, and the consequent likely range of values of the reconnection rates.First, examination of the flow maps during the interval indicates that the nightside merging line is relatively extended, with f mg lying in the range ∼0.35 to 0.65, corresponding to 0.15 on either side of the average value for the duration of the burst.This range is marked by the horizontal dotted lines, with the exterior region shown as shaded.Second, according to the above argument, we can reasonably expect the dayside reconnection rate to have fallen in the range ∼30 to ∼40 kV, corresponding to 5 kV on either side of the total transpolar voltage, averaged over the whole 6-h interval of observations.The region outside these limits is also shaded.The remaining unshaded area then represents the solutions which are compatible with both of these constraints.The location of this area first suggests an overall poleward boundary displacement between ∼0.5 • and ∼1 • , reasonably consistent with the ∼1 • displacement inferred in Fig. 9.It also suggests nightside reconnection rates of ∼45-60 kV, which thus exceed concurrent dayside rates by 10-20 kV.
The implication of these results is that during the burst interval examined here, the nightside reconnection rate exceeded the dayside reconnection rate by 50%.Under conditions of weak northward-directed, but B y -dominated IMF, therefore, it appears that the flow continues to be dominated by Dungey-cycle flow, as it is during intervals of southwarddirected IMF, and the "return" flow from the tail is often bursty and sporadic.However, our data also show that these bursts are not accompanied by phenomena characteristic of substorms, such as enhanced auroral zone conductivities, magnetic bays, Pi2s, or geostationary energetic particle injections.It seems likely that the impulsive phenomena concerned, while possibly being related physically to substorms, were taking place considerably further down the tail than is typical of more strongly-driven intervals.

Summary and conclusions
We have discussed the ionospheric flow and magnetic disturbance which occurred during two 6-h near-steady intervals of northward, but B y -dominated IMF.During the first interval, on 2 December 1999, the average IMF clock angle was ∼−60 • and the field strength was ∼10 nT.During the second interval, on 14/15 December 1999, the average IMF clock angle was ∼−50 • , and the field strength was ∼7 nT.Both sets of conditions gave rise to similar continuous flows in the polar ionosphere, with a highly asymmetric twin-vortex pattern being observed, consistent with directly driven Dungeycycle convection.This flow was slightly weaker during the second interval compared with the first (the mean transpolar voltage was 29 kV compared to 35 kV), possibly due to the reduced IMF clock angle and field strength.On shorter time scales, however, the flow exhibited intervals of strong variability in the nightside ionosphere, which differed slightly in nature between the two intervals.During the 2 December interval, ∼1000 m s −1 bursts of flow were observed in the mid-night sector, lasting a few tens of minutes and recurring every hour or so.They were accompanied by weak magnetic signatures (∼10 nT) and low amplitude Pi2-band (5-50 mHz) activity, although it is not clear that either were directly associated.During the 14/15 December interval, the bursts were observed to be less regular, with only one discernable event occurring during the first half of the interval.During the second half, the bursts become more frequent, but appear to be less distinct from one another than those observed during the first interval.
Over the entire 6 h of each interval, it was observed that the size of the convection pattern remained relatively constant.This implies that tail reconnection was ongoing, at an average rate approximately equal to that at the dayside magnetopause.However, the nightside flow observations then indicate that this reconnection typically was not steady, but underwent sporadic enhancements, giving rise to the flow burst intervals.This is indicated by the fact that each burst was accompanied by an enhancement in the total transpolar voltage, and in one example, where the flow reversal boundary was well observed, appeared to be coincident with a small (∼1 • ) contraction of the polar cap boundary.These bursts are, therefore, likely to be an ionospheric manifestation of convective transport away from a distant tail reconnection site.A simple model of this process, based on the Cowley and Lockwood (1992) picture of reconnection-associated flow excitation, has enabled us to make an estimate of the nightside reconnection rate associated with one of these burst intervals, based on assumed limits on the dayside rate.We estimate that, during this burst, the nightside reconnection rate was elevated typically to values ∼45-60 kV, thus exceeding the likely concurrent dayside reconnection rates of ∼30-40 kV by 50%.Bearing in mind the fact that over the full 6 h of each interval, the polar cap remains relatively constant in size, this also implies a "non-burst" nightside reconnection rate of ∼20-25 kV.While not manifest as substorms, these discrete bursts of enhanced tail reconnection seem clearly to be a related phenomenon occurring, as suggested above, much further down-tail.In effect, we have demonstrated that while under conditions of southward IMF, the Dungey cycle will usually be manifest as the substorm cycle; the two are not, in fact, synonymous.Intermediate interplanetary conditions exist, such as those considered here, in which the Dungey cycle continues to drive the system, but without evidence of classic substorms being observed (though tail activity appears still to be pulsed on ∼1-h time scales).The question remains as to exactly what range of interplanetary conditions gives rise to a tail response of this nature.We also need to investigate the corresponding picture for positive B y -dominated IMF.These questions, however, remain the subject of future work. Fig.1

Fig. 1 .
Fig. 1.(a) Fields of view of the six Northern Hemisphere SuperDARN radars at 04:30 (left) and 00:00 UT (right), the centre times of the two intervals considered in this paper.These are plotted on a geomagnetic grid from the pole to 60 • , with 12 MLT at the top and dusk to the left.Fields of view coloured blue are those of the four radars for which detailed line-of-sight measurements will be discussed, using the beams indicated in red.(b) Shows ground-based magnetometer locations and the LANL spacecraft footprint at the same times on the same grid.Red dots indicate stations whose data has been band-pass filtered to provide information on Pi2 signals.Blue dots indicate three stations which span the local time extent of the radar fields of view.Yellow dots indicate stations which span individual events which are studied in detail.The black oval represents the ionospheric footprint of geostationary orbit.

Fig. 2 .
Fig.2.Upstream interplanetary observations from the ACE spacecraft for 19:30 UT on 1 Dec to 13:30 UT on Dec 1999, lagged by 76 min to account for the propagation delay to the ionosphere.The top two panels show solar wind density and velocity data (the dashed line represents data from Geotail, which was used where data was missing from ACE).The middle three panels show IMF data in GSM coordinates, and the bottom two panels show the field magnitude and clock angle.The clock angle is defined with respect to north, such that 0 • corresponds to a transverse field purely in the

Fig. 3 .
Fig. 3. SuperDARN line-of-sight velocity measurements from the radar beams indicated in Fig. 1a (left side) are shown for the interval 01:30-07:30 UT on 2 Dec 1999.The vertical dashed lines show the centre times of 6 pulsed flow features, which are discussed in detail in the text.

Fig. 4a .
Fig. 4a.Streamlines and vectors of the ionospheric flow derived from the SuperDARN velocity measurements shown on geomagnetic grids, obtained from the "Map-Potential" algorithm (see Sect. 2.1).Four maps are shown from the first 6-h interval studied here for (a) the six Northern Hemisphere radars and (b) the four Southern Hemisphere radars.Maps are shown at 02:00, 04:00, 06:00, and 07:00 UT for the Northern Hemisphere, and 02:04, 03:00, 04:08, and 05:00 UT for the Southern Hemisphere.At the bottom of each map the total transpolar voltage V PC (left-hand corner), and the direction and magnitude (in the Y − Z plane) of the lagged IMF (right-hand corner) are indicated (Fig. 4 continues).
Fig. 5. (a) A map of the UV auroral intensity at 01:51:07 UT (the centre time of the 32.5 s integration interval of the UV image) projected onto a geomagnetic grid.Superposed on this image are SuperDARN "Map-Potential" streamlines and flow vectors, and the orbital track of the DMSP F12 spacecraft from 01:45-02:04 UT, colour-coded to the different regions of particle precipitation (the red triangle indicates the footprint of the spacecraft at the time of the UV image).Also indicated on the map is the derived total transpolar voltage and the direction and magnitude (in the Y − Z plane) of the lagged IMF, as in Fig. 4. The superposed SuperDARN vectors correspond to the two-minute interval which encompasses the centre time of the UV image.(b) Standard precipitating particle spectrograms from the SSJ/4 instrument on board the DMSP F12 spacecraft for the overpass shown in (a).Electron (top panel) and ion (bottom panel) data are presented for 01:45-02:04 UT.Also indicated at the bottom of the figure is the magnetic latitude (MLAT) and magnetic local time (MLT) at corresponding universal times (UT).

Fig. 6 .
Fig. 6.The upper three panels show H-component magnetic field measurements from the blue colour-coded array in Fig. 1b, for 01:30-07:30 UT on 2 Dec 1999.Panels 4-6 show filtered (20-200 s) H-component magnetic field measurements from the lowlatitude magnetometer array (red dots in Fig. 1b) for the same interval.The bottom panel shows geostationary electron fluxes in two energy channels (50-75 keV and 75-105 keV) from the SOPA instrument on the LANL97A spacecraft.The vertical dashed lines correspond to those in Fig. 3.

Figure
Figure 5a also shows the track of the DMSP F12 spacecraft as it passed over the polar regions during the interval 01:45-02:04 UT.The spacecraft position at the centre time of the VIS image (and also very nearly the centre time of the SuperDARN scan) is indicated by the red triangle.The track is colour-coded by the nature of the precipitation observed, i.e.CPS, BPS, cusp, and mantle.The data on which these classifications are based are shown in standard spectrogram format in Fig.5b.It can be seen that the regions designated as CPS/BPS lie within the region of UV auroral emission, and sunward or clockwise vortical flow.The region of "mantle" precipitation corresponds to the central region of weak emission, and antisunward and anti-clockwise flow in the post-noon sector, while the "cusp" precipitation, taken to be indicative of newly-opened dayside field lines, extends across the poleward boundary of the UV emission near noon, in the region where the Map Potential streamlines turn poleward and eastward.Overall, these data (and those obtained in other VIS images and on other DMSP passes)

Fig. 7 .
Fig. 7. Flow measurement time series obtained from the Super-DARN radar observations and the "Map-Potential" algorithm.The top panel shows the total transpolar voltage, V PC .The second panel shows the peak nightside ionospheric flow speed (the horizontal dotted line is the average over the interval), and the third, the north-south (dotted line) and westward (solid line) components of the peak flow.The bottom two panels show the location of this peak flow in magnetic latitude (fourth panel) and MLT (bottom panel).The vertical dashed lines indicate six peaks in the flow speed, corresponding to those in Figs. 3 and 6.

Fig. 8a .
Fig. 8a.Flow maps illustrating the flow "bursts" indicated by the vertical dashed lines in Fig. 7, shown in the same format as Fig. 4. Eight selected maps are shown for each of the two burst intervals (a) 02:36-03:20 UT and (b) 05:02-05:30 UT.The exact time of each 2-min map is indicated in the top left-hand corner.Maps corresponding to the centre times of each of the other four bursts are shown in (c).Also indicated on the maps by the black triangles are the locations of the magnetometer stations colour-coded yellow in Fig. 1b (Fig. 8 continues).

Fig. 11 .
Fig. 11.Upstream interplanetary observations as in Fig. for 15:00 UT on 14 Dec to 09:00 UT on 15 Dec 1999, lagged by min to account for the propagation delay to the ionosphere.The top two panels show solar wind density and velocity data, the middle three panels show IMF data in GSM coordinates, and the bottom two panels show the field magnitude and clock angle.

Fig. 14a .
Fig. 14a.SuperDARN flow maps in a similar format to those shown in Fig. 4. Four maps are shown from the 6-h interval for (a) the "quiet" half of the interval for which maps are shown on the half hour for 21:30-00:30 UT and (b) the 02:10 UT burst from the second half of the interval, spanning 01:54-02:18 UT.The exact time of each 2-min map is indicated in the top left-hand corner.Also indicated on the maps (in b) by the black triangles are the locations of the magnetometer stations colour-coded yellow in Fig. 1b.(Fig. 14 continues).

FFig. 15 .
Fig. 15.The H -, D-and Z-components of the ground magnetic field for the interval 01:40-02:40 UT on 14/15 Dec 1999, measured by the stations colour-coded yellow in Fig. 1b, and indicated in Fig. 14b by the black triangles.Also shown are the east-west (in (i)) and north-south (in (ii)) SuperDARN flow components corresponding to the locations of the magnetometers.In (b) the magnetic field data has been bandpass filtered (20-200 s) to highlight Pi2-band activity.The scales to the left of each plot indicate the range between the baselines (horizontal dotted lines) and the vertical dashed lines delimit the ∼30 min burst duration.

F ig. 16 Fig. 16 .
Fig. 16.Schematic of the flow situation appropriate to the burst intervals examined here, in which dayside and nightside reconnection are both continuously present, but with the rate of flux closure on the nightside exceeding the rate of open flux creation at the dayside, such that overall the polar cap contracts.The segments of the open-closed field line boundary mapping to the reconnection sites (i.e. the "merging gaps") are shown by the dot-dashed lines, while the adiaroic segments of the boundary (which are not so connected and which thus move with the flow) are shown by the heavy solid lines.The lighter arrowed lines again show the plasma streamlines, and the large arrows show the implied boundary motion.

Fig. 17 .
Fig. 17.Reconnection rate contours plotted on the f mg − B plane, where the fraction of the open-closed field line boundary occupied by the nightside merging gap, f mg , is shown on the vertical axis, and the poleward displacement of the magnetic latitude of the nightside flow reversal boundaryB during the event is shown on the horizontal axis.The dark-shaded area in the bottom right corner of each plot represents an area of inconsistent results where the dayside reconnection rate becomes negative.The horizontal dotted lines mark the likely range of f mg for the event (obtained from the SuperDARN maps), with the exterior region shown shaded.Also shaded is the region of unreasonable dayside reconnection rates as discussed in the text.The remaining unshaded area corresponds to the region of solutions that conforms to all of these constraints.These results are plotted for the 03:00 UT burst on 2 Dec.
• -2 • , from ∼76 • to ∼78 • .Extrapolated around the boundary of the polar cap, such a displacement implies a net closure of open flux of ∼1±10 8 Wb, due to enhanced tail reconnection exceeding the concurrent dayside reconnection rate during the burst interval.