Spatial Distribution of Conductances and Currents Associated With a North-South Auroral Form During a Multiple-Substorm Period

Using the method of characteristics to invert ground-based data of the ground magnetic field disturbance and of the ionospheric electric field, we obtain spatial distributions of ionospheric conductances, currents, and field-aligned currents (FACs) associated with a north-south auroral form that drifts westwards over northern Scandinavia around 2200 UT on December 2, 1977. This auroral form is one in a sequence of such north-south structures observed by all-sky cameras, and appears 14 min after the last of several breakups during that extremely disturbed night. Our analysis shows that the ionospheric Hall conductance reaches values above 200 S in the center of the form, and upward flowing FACs of up to 25 μA/m2 are concentrated near its westward and equatorward edge. The strong upward flowing FACs are fed by an area of more distributed, but still very strong downward-flowing FACs northeastward of the auroral form. In contrast to the conductances, the electric field is only slightly affected by the passage of the form. We point out similarities and differences of our observations and results to previously reported observations and models of ‘auroral fingers’, ‘north-south aurora’, and ‘auroral streamers’ which are suggested to be ionospheric manifestations of bursty bulk flows in the plasma sheet.


Introduction
A number of studies has been carried out to investigate the relation between auroral forms and ionospheric electrodynamic quantities, such as conductances, currents, and ®eld-aligned currents (FACs) during substorms. They can be ordered into two groups: one of them makes use of simultaneous observations from loworbiting satellites to detect the magnetic ®eld eect of FACs, and of either ground-based all-sky cameras or high orbiting satellites that provide global auroral images (e.g., Bythrow and Potemra, 1987;Lopez et al., 1991;Homan et al., 1994). The other focuses on combining the auroral images with ground-based data from magnetometers, radars, and other instruments if available (e.g., Robinson et al., 1985, Kamide et al., 1989Untiedt and Baumjohann, 1993, and references therein), while other satellite data may be used as supplementary evidence.
A shortcoming of satellite-based studies is that to obtain a spatial picture, many dierent passes, often from dierent days, have to be combined. Therefore, the results are mainly of a statistical nature, as in the classical study of Ijima and Potemra (1978) of the FAC distribution during substorms. Fujii et al. (1994) and Homan et al. (1994) have binned a large data set of satellite passes according to several sectors of the substorm aurora, and scaled the passes from dierent days with respect to the intensity of each substorm to construct a combined pattern of FACs and ionospheric currents for the whole substorm area. Despite their very careful data treatment, it is still not clear if their combined result could represent a single event, and it is dicult to link their results to individual, instantaneously observed auroral forms. Moreover, since no information on conductances and the electric ®eld was available, the estimation of ionospheric currents from FACs contains considerable uncertainty. Another problem in obtaining FAC densities from magnetic ®eld measurements along satellite passes is that for a proper quantitative analysis, the assumption that the FAC con®guration is that a (practically) in®nite current sheet has to be made.
During and after the International Magnetospheric Study (IMS, 1976±1979), several studies were carried out that combined ground magnetic ®eld data from the arrays set up during this period with ionospheric electric ®eld data measured by coherent scatter radars (e.g., Akasofu et al., 1980;Kamide et al., 1982; see the review of Untiedt and Baumjohann, 1993, for the studies carried out with observations from Scandinavia). Many of these works addressed the spatial conductance, electric ®eld, currents, and FAC distribution associated with auroral forms, such as the westward traveling surge (WTS; e.g., Inhester et al., 1981;Opgenoorth et al., 1983a), the auroral breakup (e.g., Opgenoorth et al., 1980;Baumjohann et al., 1981), or omega bands (e.g., AndreÂ and Baumjohann, 1982;Opgenoorth et al., 1983b). The method of analysis used in these studies was the`trial and error' method, i.e., the searched ionospheric parameters were modeled until a sucient ®t between the measured quantities and their values as calculated by the models was obtained. While these studies are still standard references, later works, particularly those including data from incoherent scatter radars, have shown that their method of analysis has introduced considerable smoothing of the resulting features (compare, for the WTS case, e.g., Kirkwood et al., 1988;Aikio and Kaila, 1996). One reason for this is that`trial and error' method typically starts out with rather uniform ionospheric distributions, and gradients are only gradually introduced if they are unavoidable to obtain a ®t of the measured ground magnetic ®eld disturbance that matches a given relative error. However, due to the integrating eect of the ground magnetic ®eld disturbance, there is also a general resolution boundary of about 100 km wavelength for studies that utilize ground magnetometers.
A considerable improvement in the analysis approach of ground-based data was introduced with thè method of characteristics' (Inhester et al., 1992;Amm, 1995Amm, , 1998. In contrast to the`trial and error' approach, this method is a forward technique in which a 2D partial dierential equation is solved along its characteristics for the Hall conductance as the primary output. Together with the measurements and an estimate of Hall to Pedersen conductance ratio, spatial distributions of the remaining electrodynamic parameters are inferred. The method of characteristics has been shown to provide substantially more detailed results than the`trial and error' method, and it is, in particular, more suited to resolve inhomogeneous ionospheric situations (Amm, 1995(Amm, , 1998. Moreover, the number of assumptions needed is reduced compare to the`trial and error' method, for which both the Hall and Pedersen conductance have to be assumed independently. In contrast to the individual conductances, their ratio has been shown to be relatively stable and assessable from the level of ground magnetic activity (see Schlegel, 1988;Lester et al., 1996). In addition, Amm (1995) has shown the in¯uence of the estimate of that ratio on the ®nal results of the method of characteristics to be small.
We analyze a north-south auroral structure that appeared on December 2, 1977, over northern Scandinavia as one in a sequence of such north-south auroral forms. The form ®rst appeared at 2155 UT, 14 min after the last of several auroral breakups during the extremely disturbed night, and drifted westward with a speed of about 1.4 km/s. We use the method of characteristics to infer spatial distributions of conductances, ionospheric horizontal and ®eld-aligned currents for the time step at 2200 UT, when the backscatter of the coherent scatter radar STARE is sucient to provide electric ®eld data over nearly its complete ®eld-of-view, and compare our results with the auroral display. We also show how the speci®c ground magnetic signature of the form is produced.
Not many studies are available in the literature about north-south auroral forms. Rostoker et al. (1987) have described such north-south structures which they call auroral ®ngers' viewed by the Viking satellite imager. They found these forms to be a facet of the auroral oval when it has a north-south width of more than 10°of latitude, which only happened during very strong substorms, and at least 10 min after the substorm onset. The north-south extent of their structures is about 450 km, and multiple forms with a longitudinal separation of 150±300 km occurred. They attributed these forms to traces of drift paths of high energetic electron populations which have been injected along the velocity shear zone separating the central plasma sheet (CPS) from the low latitude boundary layer (LLBL) during the substorm expansive phase, and subsequently drift Earthward. During that drift, the electrons experience Fermi and betatron acceleration, so that the equatormost edge of the north-south auroral form should be associated with the most energetic electrons.
This scenario has been re®ned and supported with quantitative considerations by Liu and Rostoker (1993). They suggest the Kelvin-Helmholtz instability to be the mechanism to inject cold plasma from the LLBL into the CPS. The charge separation that such a cold plasma blob undergoes due to adiabatic drift is compensated by a pair of one upward-and one downward-¯owing FAC in their model. They expect electrons with energy upward to 10 keV to be precipitating to the ionosphere in the upward FAC region.
In their comprehensive substorm aurora study, Nakamura et al. (1993) have identi®ed north-south auroral structures (they call`N-S aurora') in the central auroral bulge just east of the surge. They also found multiple, longitudinally separated north-south auroral forms that drifted westwards and expanded equatorwards. Consistent with the observations of Rostoker et al. (1987) and Liu and Rostoker (1993), the westernmost of their N-S aurora forms showed a hook-like tilt to the west, whereas the easternmost ones show such a tilt towards the east. In later substorm phases, the N-S aurora develops to diuse or pulsating aurora. Nakamura et al. (1993) explain their events in terms of a local two-cell potential at the auroral bulge, where the negative charges located in the surge and the positive ones at the northernmost arc of the auroral bulge east of the surge. The auroral evolution is then explained as traces of rapid plasma¯ows in the local two-cell potential.
Recently, several works have related north-south auroral forms, partly called`auroral streamers', to bursty bulk¯ows (BBF) in the central plasma sheet (Sergeev et al., , 1999Angelopoulos et al., 1997;Henderson et al., 1998) which were explained by Chen and Wolf (1993) as rapidly Earthward convecting plasma bubbles with lower particle, but higher magnetic pressure than the surrounding plasma. These auroral streamers are localized auroral activations that expand equatorward from the poleward boundary of the aurora to form north-south structures, often during the expansion phase of a substorm (Elphinstone et al., 1996). They are therefore quite similar to the N-S aurora described by Nakamura et al. (1993). Chen and Wolf (1993) predicted the ionospheric current signature of the BBF to be a very localized current wedge, with the downward FACs on the eastern and the upward FACs on the western side of it.
None of the previous studies presents a self-consistent model of ionospheric currents, conductances, and FACs associated with the auroral events. We will discuss our results for these quantities in the context of the studies mentioned.

Instrumentation
For our analysis, ground magnetic data of the Scandinavian Magnetometer Array (SMA; KuÈ ppers et al., 1979) is used (Fig. 1), which is still the most extensive magnetometer array ever set up in northern Scandinavia. Except for the six magnetometers of the chain KUU to SKA which are of a digital¯uxgate type, the instruments are of a modi®ed version of the classical Gough-Reitzel type.
Ionospheric electric ®eld data is provided by the STARE coherent scatter radar (Greenwald et al., 1978). In the area of overlapping radar beams of the two radar stations (Fig. 2), vector information of the ionospheric electric ®eld can be composed from the electron drift velocities along the beam directions measured by the individual radar stations, provided that both of them receive sucient backscatter. As one of several conditions, backscatter is regularly observed if the line-of-sight (i.e., along the beam directions) electron velocity exceeds about 360 m/s which corresponds to a electric ®eld magnitude perpendicular to the beam of about 18 mV/m (e.g., Haldoupis et al., 1990). However, occasionally backscatter can also be received for line-of-sight electron velocities down to about 200 mV/m.
All sky camera (ASC) data was used for this study from the stations in Kiruna, Muonio, and Ivalo (see KIR, MUO, and IVA in Fig. 1). However, because of partial cloudiness, not all pictures of these station could be used during the period of interest.

Data description and analysis
The day of December 2, 1977, was characterized by a long-lasting and extremely strong auroral and ground magnetic activity. The Kp index took the high value of 7 from 15-24 UT, and ground magnetic disturbances of the order of 1000 nT and more were observed. At 1630 UT, the Harang discontinuity event studied by Kunkel et al. (1986) took place. This early appearance of the Harang discontinuity points to a equatorward Map of the two STARE radar stations and schematic representation of their radar beams. The area of overlapping beams corresponds to the ®eld of view of the radar in which vector electric ®eld data can be measured; from Greenwald et al. (1978) displacement of the ionospheric current systems connected with an expansion of the auroral oval that is often observed during strongly disturbed periods.
The event studied here started at 2155 UT, 14 min after one of several auroral breakups observed during that night and a corresponding renewed intensi®cation of the ground magnetic disturbance, when a north-south (N-S) auroral form showed up over the zenith at Muonio, i.e., at the eastern edge of the STARE ®eld of view (compare Fig. 2). This form moved westwards with a velocity of about 1.4 km/s, and disappeared from the western horizon of the all sky cameras and from the western edge of the STARE ®eld of view at 2207 UT. Since the pictures of the cameras at Muonio and Kiruna were partly disturbed by clouds, in Fig. 3 we show an undisturbed example of such a north-south aurora from Ivalo (east of the STARE ®eld-of-view) at 2201 UT, which is taken from the next one in the sequence of westward-moving north-south auroral forms during the period of interest.
While the initial development of the N-S auroral form under study cannot be followed in detail in the Muonio ASC data because of the clouds, an initial equatorward expansion is seen near the western horizon from the camera at Ivalo. Since the equatorward edge of the auroral form in associated with a minimum of the Z (vertical, positive downward) component of the ground magnetic disturbance as will be shown below, an estimate of the equatorward speed of this initial expansion can be obtained from the Z magnetograms of the north-south chain MAT-PEL ( Fig. 4a; SOY and OUL did not measure Z components). As indicated by the black arrows in Fig. 4a, minimum of the Z moves equatorward around 2155 UT with a velocity of about 2.4 km/s. Figure 4b shows a distinct``sawtooth-like'' appearance of the N-S auroral form in the Y (eastward) ground magnetic disturbance component. The arrows in Fig. 4b show the westward movement of the zero crossings of the Y component from positive to negative after 2155 UT. The location of these zero crossings approximately, but not exactly, corresponds to the longitudinal center of the auroral form (see the detailed analysis).
In our single time step analysis, we focus of the instant at 2200 UT when the auroral form under study is in the western part of STARE ®eld-of-view, because then the radar receives backscatter in nearly its whole measurement area, whereas earlier when the form was more in the center of the STARE area there were considerable holes in the electric ®eld vector data.
In the period between the breakup and our event, the horizontal ground magnetic disturbancef qYh in the SMA array was dominated by a relatively uniform, at 500 nT already quite strong, westward electrojet. About between 2155 and 2207 UT a short disturbance occurs which is characterized at its climax at 2200 UT by three signatures (Fig. 5a): ®rst, a clear intensi®cation of the westward electroject can be seen, particularly between 67 and 69°of latitude and 18 to 24°of longitude, with a magnitude of the horizontal disturbance of about 720 nT. Second, about between 20 and 25°of longitude, a distinct northward deviation of the by 90°rotated ground magnetic disturbance is observed. Finally, in the western part of the SMA area at the southern border of the westward electrojet, a curl-like structure is visible, the center of which can roughly be estimated at 63°of latitude and 18°of longitude. Like the auroral form, these magnetic signatures are traveling westwards with a speed of about 1.4 km/s. Although the signatures as such remain stable during our event, the ground magnetic disturbancef q is far from being stationary. After about 2207 UT, the magnetic disturbance decreases, and the discrete aurora fades or evolves into diuse aurora.
Compared to the ground magnetic disturbance, the electric ®eldĩ as measured by the STARE radar is much less structured (Fig. 5b). Except for local small-scale inhomogenities, it shows mainly southwestward-directed vectors east of 20 degrees of longitude, and mainly southward-directed ones west of this. The magnitude of i is around 20 mV/m, with little variability. This general picture ofĩ shows little dierence to the one observed a few minutes earlier to later, except for regions of missing backscatter before the time step analyzed here. However, also during instances of missing backscatter, the remaining vectors show the same features as described.
No westward motion is visible in the electric ®eld data. From these observations alone it can already be presumed that the main eect of the ground magnetic disturbance must originate in the ionospheric conductances. Figure 5c shows the north-south auroral form under study in the western part of the ®eld of view of the ASC camera at Muonio. The bright north-south aligned discrete arc shows a crescent shape towards the east at its southernmost edge. Note that the dark spots at the western horizon are clouds, and also around the zenith some thin clouds disturb the picture. Eastward of the bright arc, an area with diuse aurora can be seen, with decreasing intensity towards the east. The longitudinal location of the center of the bright aurora is estimated to about 16 0.5°, supported from the ASC camera at Kiruna (data not shown here, since the general pattern is the same as seen from Muonio, but Muonio has a slightly better image quality). Although the mapping of auroral forms ASC pictures to geographical locations exhibits some uncertainties if the forms are far from the zenith, we feel suciently sure about our estimate since it is con®rmed by the movement of the magnetic signatures that show an analogous spatial development to the auroral forms. At the eastern horizon, another north-south aligned aurora is seen that probably belongs to the subsequent north-south aligned aurora as shown in Fig. 3. Figure 6a shows the resulting Hall conductance R r obtained using the method of characteristics. We note that this method gives an upper and lower estimate for R r in areas where the solution is not unique. This may happen because for relatively small analysis areas and small values of jdivĩj, as in the case of our study, the boundary values for solving the dierential equation for R r along the individual characteristics may still have some in¯uence on the solution. However, typically this in¯uence is decreasing exponentially with the path length of each characteristic from the boundary (compare Amm, 1998, for details). It is important to understand that this uncertainty is not caused by the special method of solving for R r , but by the physical setup of solving the ionospheric electrodynamic equations inside a bounded domain of space. In the case of Fig. 6a, the dierence between the upper and lower values for R r is everywhere marginal except for the region east of 24°of longitude, and for the local increase in R r at the southern boundary between 22 and 24°of longitude. These regions are therefore excluded from the following analysis. Since all ionospheric distributions are dominated by the features associated with the auroral form in the west, this is not a signi®cant drawback. In areas of non-unique solution for R r , the medium between the upper and lower estimates is displayed in Fig. 6a.
The most striking feature of Fig. 6a is a strip of strongly enhanced Hall conductance centered at about 16°longitude, with a maximum of 225 S at 69°latitude. This strip is well colocated with the discrete part of the north-south auroral form (see Fig. 5c). Hall conduc- component from a east-west chain from PEL to OKS; after 2155 UT, a characteristic``sawtooth-like'' structure in Y is seen. The Y = 0 crossing approximately corresponds to the center of the N-S auroral form which moves westward over the SMA array tances in that range are not often found in the literature, but on the other hand they are not uncommon in incoherent scatter radar measurements. Aikio and Kaila (1996) observed R r values above 200 S with the EISCAT radar near an intense auroral form during the expansion phase of a substorm. Such high Hall conductances are believed to be caused by high energetic electrons precipitating into the ionosphere within the auroral structure. Northward and southward of the maximum, R r takes values around 140 S in the strip which seems to show a slight tilt in northwest-southeast direction. At the southern boundary of the area under study, it drops to values around 50 S. To the east of the strip, R r decreases gradually until it reaches about 30 S at 23°longitude. This behavior agrees well with the gradual decrease in intensity of the diuse aurora located eastward of the discrete north-south auroral form seen in Fig. 5c.
The resulting ionospheric currentst (Fig. 6b) show magnitudes near 2500 mA/m in the area of enhanced conductance, reaching even values above 3500 mA/m at the R r maximum.t points nearly purely westward in the northern part of the strip of enhanced conductance, whereas in its southern part and southwest of it,t shows a substantial southward component. Eastward of the auroral form and the enhanced conductance strip, jt j decreases to about 1000 mA/m, and the direction of the currents is mainly westwards, except for a region around 71°latitude and 22°longitude whereĩ locally deviates to the west, and thereforet deviates to the north. Note that the clear northward de¯ection of the by 90°c lockwise rotatedf qYh eastward of the auroral form is Fig. 5a±c. Measurements used in the single time step analysis of the north-south auroral form on December 2, 1977, at 2200 UT. a By 90°c lockwise rotated horizontal ground magnetic ®eld disturbance, observed by the SMA. b Electric ®eld vectors observed by the STARE radar. c All sky camera picture from Muonio, with the analyzed north-south auroral form in the western (left) part (picture modi®ed for enhanced contrast); the north-south aurora at the eastern horizon is probably related to the subsequent form as shown in Fig. 3; note some clouds at the western horizon as well as thin clouds near the zenith (geographic north points to the top, east to the right) not visible, and their southward de¯ection at the western boundary of the analysis area is much less visible int .
The structure of the FACs (Fig. 6c) is also completely dominated by the conductance enhancement in the area of the discrete auroral form: north of 68°latitude, eastward of the enhanced R r strip, a north-south aligned sheet of downward¯owing FACs with amplitudes around 15 A/km 2 is visible that slightly widens towards the north. These already very strong FACs are still surpassed by the upward¯owing FACs in the southern and western part of the analysis area which form a northwest-southeast aligned layer and reach amplitudes up to about 25 A/km 2 . This upward FAC layer closely follows the shape of the westward side of the equatorward edge of the north-south auroral form (compare Fig. 5c). It is also interesting to observe that this FAC layer seems to be roughly colocated with the brightest part of the north-south aurora at its western ank. It is likely that the upward FAC sheet continues westwards of the analysis area towards the northwest, since the decrease of R r to the west is not fully contained in our analysis region. The FACs at the eastern boundary in Fig. 6c are located in the area of badly determined conductances and are likely to be arti®cial. The cross-structure size of the FAC current system can roughly be estimated to 6°longitude or 250 km on the ionospheric height. Note that, as can be shown by an order-of-magnitude calculation, for the FAC densities obtained still many fewer charge carriers are required to carry the current than are available in the auroral ionosphere (e.g., Kirkwood et al., 1988;Robinson and Vondrak, 1990).
As a check of our analysis results, we recalculated the ground magnetic disturbance that follows from the horizontal ionospheric currents and from the FAC as shown in Fig. 6b, c and compared it with the measured ground magnetic disturbance. Since the magnetic ®eld of a line current decreases as the inverse of the distance from the axis of current¯ow, one normally cannot expect to obtain the complete ground magnetic distur- Fig. 6a±c. Results obtained with the method of characteristics from the input data as shown in Fig. 5a, b. a Hall conductance in S. b Horizontal ionospheric currents, with some Hall conductance iso contours. c Field-aligned current density (positive-downwards), with some Hall conductance iso contours. The shaded areas correspond to the estimated geographical location of the discrete N-S aurora bance in a certain domain from the currents and FACs that¯ow overhead that same domain only. However, in our special case the large conductance enhancement that is con®ned in our analysis region leads to the conjecture that here most of the ground magnetic eect might be caused inside the analysis region itself. Then it should be possible to reconstruct, at least, the essential structure of the ground magnetic disturbance from our results. Figure 7 shows the measured and the recalculated by 90°rotated horizontal ground magnetic disturbance as well as the vertical magnetic disturbance (positive downwards). The former is interpolated on the same grid as the analysis results. For the by 90°rotated horizontal disturbance (Fig. 7a, b), a very good general agreement between the two distributions is seen, especially in the center of the analysis region. In particular, the northward deviation of the vectors between 20 and 22°longitude is well represented. Also the curl-like structure in the southwest is present, although it is weaker in the recalculated ®eld than in the measured one, probably mostly because of missing upward¯owing FACs beyond the western border of our analysis region. Similarly, the dierence between the northward deviated measured vectors at the northern boundary and slightly southward de¯ected vectors in the recalculated ®eld can be explained by missing downward¯owing FACs north of that boundary. It is interesting to note that the quite inhomogeneous distributions oft and j k in Fig. 6b, c are able to reproduce the somewhat smoother ground magnetic disturbance distributions.
Also regarding the vertical component of the ground magnetic disturbance (Fig 7c, d), we obtain a generally good correspondence between the measured and the recalculated ®eld, especially in the central and southern part of the analysis area. Near the western part of the northern boundary, however, the recalculation yields a clearly higher vertical disturbance than was measured. It Fig. 7a±d. Comparison of measured ground magnetic ®eld disturbance interpolated on a grid with the ground magnetic disturbance as recalculated from the three-dimensional current system resulting from our analysis (Fig. 6b, c). a Measured by 90°clockwise rotated horizontal disturbance. b Recalculated by 90°clockwise rotated horizontal disturbance. c Measured vertical disturbance (positivedownwards). d Recalculated vertical disturbance is very likely that this dierence is caused by missing strong westward-¯owing currents that¯ow northwestward of the analysis region in the continuation of the strip of enhanced conductances and discrete aurora. They would cause a negative vertical disturbance in the northwestern analysis region whereas the horizontal magnetic disturbance component inside this region would be less aected. The fact that a similar eect is not visible at the southern boundary suggests that the real ionospheric currents south of the region under study are weak.
To summarize the recalculation of the ground magnetic disturbance, we can say that the agreement of the recalculated and the measured disturbance is at least as good as for a typical`trial and error' study. The reason why the latter method never yields such strong gradients as appear in our results lies in its more diusive nature.
Finally, we study the way that the current system produces the characteristic ground magnetic signature of the passage of the north-south auroral form. For that, a decomposition of the total horizontal currentst into those currents which are caused by divergences inside the analysis area (t f ), those which are caused by curls inside that area (t df ), and those which are curl-and divergence-free inside our region (t lp ) is carried out (e.g., Amm, 1996).t lp has a potential that satis®es Laplace's equation inside the analysis region.
The result of such an analysis is shown in Fig. 8. Despite of some numerical roughnesses,t lp (Fig. 8a) shows a quite homogeneous westward electrojet with an average magnitude of about 1630 mA/m. It produces an average southward magnetic disturbance of 530 nT on the ground which corresponds well with the amplitude of the relatively homogeneous westward jet that is present before and immediately after the intensi®cation related to our event. Hence, this current system can be regarded as the`background system' in front of which the event which is characterized by thet f andt df current systems takes place. However, as far as it is caused by sources outside of the analysis region, the intensi®cation of the westward electrojet during the event is contained int lp , too.t f (Fig. 8b) is dominated by currents¯owing southwestward between the regions of strong downward-and upward-¯owing FACs in the central north and in the southwest of the area under study. It reaches magnitudes up to 1500 mA/m. These Fig. 8a±c. Decomposition of the total resulting ionospheric current system into its: a curl-and divergence-free part inside the analysis region. b Part that is caused by divergences inside the analysis region. c Part that is caused by curls inside the analysis region currents are therefore much stronger than those oft df which only reach magnitudes of about 800 mA/m (Fig. 8c). Sincet f together with the FAC connected to it does not produce any magnetic ®eld disturbance below the ionosphere (Fukushima, 1976;Amm, 1997), most of the current related to the north-south auroral form is thus magnetically invisible from the ground. On the other hand, in the part current systemt df , the typical ground magnetic signatures are represented which are added on the westward electrojet by the event (compare Fig. 5a): The north de¯ection of the by 90°rotated horizontal magnetic ®eld between 20 and 24°longitude is clearly visible, and also the curl-like structure in the southwestern part of the analysis area is represented.

Discussion
The results of our analysis of ground-based data during a passage of a north-south auroral form over northern Scandinavia are, in most respects, compatible with the observations and results of Rostoker et al. (1987) and Liu and Rostoker (1993) on auroral ®ngers. Their and our observations agree in that the north-south forms appear during an extremely disturbed period more than 10 min after an auroral breakup and the passage of a surge. Since global auroral images are not available for our event, we do not know if the width of the auroral bulge exceeds 10°latitude as found as a condition for the appearance of such forms by Rostoker et al. (1987), but regarding the intensity of the ground magnetic disturbance, a large bulge evolution seems to be likely. Moreover, the original formation of the surge and of north-south forms occurs eastward of the SMA and STARE ®elds of view so that we cannot account for a possible time dierence between the passing of the structures and their initial evolution. The latitudinal extent of our forms is about 500 km, in agreement with their ®ndings. The longitudinal separation between the two subsequent forms can roughly be estimated as 500 km which is somewhat larger than the values found by them.
Regarding the scenario that Rostoker et al. (1987) and Liu and Rostoker (1993) proposed for the origin of the auroral ®ngers, our results show an excellent agreement in that the strongest FACs, colocated with the brightest aurora of the north-south form, are seen at the equatorward edge of the form. According to Liu and Rostoker (1993), this location is a trace of Earthward edge of the electrons' drift path in the central plasma sheet where they are scattered into the loss cone. Moreover, the bipolar FAC geometry that the authors propose in their model is represented in our results, too (compare Fig. 6c). Although they did not precisely specify the expected location of the FACs with respect to the auroral form, it seems reasonable that in our results the upward-¯owing FACs are located inside the form since the discrete arc emission should be caused by the precipitating high energy electrons, whereas the downward-¯owing FACs are located eastwards and northeastward of it in a region that partly shows diuse aurora. However, there is one essential dierence between the model of Liu and Rostoker (1993) and our observations: since the auroral forms are drifting westwards inside the westward electrojet domain with southward to southwestward electric ®elds (compare Fig. 5b), theĩ Âf drift cannot be the reason of the motion of the form for our event, as it was assumed by the previous authors for their auroral ®ngers which mainly occurred in the eastward electrojet domain. However, the drift velocity was mainly inferred from the appearance of the forms in the evening sector, and no supporting magnetic or electric ®eld data was shown by Liu and Rostoker (1993).
That this westward drift motion inside the westward electrojet domain is not an uncommon feature of north-south auroral forms is con®rmed by the observations of Nakamura et al. (1993) who located their westward-drifting N-S aurora in the same domain using ground magnetic data. Also with respect to all other observational facts, the forms we described agree with those discussed by Nakamura et al. (1993), i.e., their appearance eastward of the surge in a welldeveloped auroral bulge, their latitudinal extent, the possibility of successive forms, and the crescent-like shape of the forms with a tilt to the east, which agree with their easternmost observed N-S auroras. Moreover, the southeastward¯ow direction corresponding to the southwestward electric ®eld would be compatible to their suggestion of a local two-cell potential associated with the auroral surge (see their Fig. 11). However, the STARE ®eld of view is too small to really prove or disprove the existence of such a potential structure. Using a single station magnetometer record, Nakamura et al. (1993) described the magnetic eect of the N-S aurora as having a similar con®guration like that of the Harang discontinuity. This conclusion is not in agreement with our spatial observations.
Finally, we note that nearly the same ground magnetic situation (1 min earlier) as analyzed in our study was discussed by Untiedt and Baumjohann (1993) in their review on SMA observations, and classi®ed as one of the most inhomogeneous situations observed by this array. However, the authors did not carry out a combined analysis with radar and ASC data. They proposed that the ground magnetic signature observed is related to an auroral breakup. Our resulting current system in fact shows some similarities to the geometry of the breakup current system as shown in Baumjohann et al. (1981), in that a locally enhanced westward electrojet is diverged by concentrated upward¯owing FACs. On the other hand, there are also important dierences between their current system and ours. Whereas in the breakup model of Baumjohann et al. (1981), the enhanced westward jet is fed by distant and more distributed downward-¯owing FACs, in our case this feeding occurs locally at the northeastern edge of the auroral form (compare Fig. 8b). Moreover, whereas in the breakup model nearly the total electrojet is diverged in the surge, in our case only the additional currents related to the auroral form are diverged, and a substantial portion of the jet continues to¯ow westward of it.
On the other hand, this picture of a localized, additional current wedge is exactly what Chen and Wolf (1993) expected to be the ionospheric current signature of the bursty bulk¯ows (BBF) in the ionosphere. Our results agree with their predictions in that the downward FAC is located at the eastern side and the upward one at the western side of the wedge. However, besides this longitudinal separation there is also a latitudinal one present in Fig. 6c. The crossstructure size of about 6°longitude that we estimated from the FAC distribution is slightly smaller, but close to the previous estimate of about 0.5 MLT by Kauristie et al. (1996). From the theory of Chen and Wolf (1993), one would expect a westward-directed electric ®eld in the magnetosphere due to polarization by the inertia drift inside the bubble that contains the BBF. Although the STARE electric ®eld (Fig. 5b) includes some westward vectors, they are not colocated with the auroral structure and do not move together with it.
Since no magnetospheric in situ data is available for our event, we cannot draw a ®nal conclusion about the magnetospheric process that causes the north-south auroral forms. However, the existing similarities of the current system associated with the form studied here and a breakup type, but more localized current system may indicate that these forms can be regarded as periodic, breakup-like but local reintensi®cations that occur during extremely disturbed periods inside an already developed auroral bulge.

Summary and conclusions
From the analysis of the ground magnetic disturbance, the ionospheric electric ®eld, and the optical all-sky camera data during a passage of a north-south aligned auroral form on December 2, 1977, we can draw the following conclusions: 1. The auroral form is associated with a strong Hall conductance enhancement, with a maximum value of about 225 S.
2. In contrast to the conductances, the ionospheric electric ®eld is only slightly aected by the passage of the auroral form.
3. Near the equatorward edge of the discrete auroral form, concentrated upward-¯owing FACs of magnitudes up to 25 A/km 2 are present. These upward FAC are colocated with the brightest part of the aurora. The current that is diverged by the upward FAC is led into the ionosphere by a sheet of more distributed downward owing FACs eastward and northeastward of the discrete auroral form. The downward FACs reach a magnitude up to 15 A/km 2 , and are located in an area that partly shows diuse aurora.
4. Most of the three-dimensional ionospheric current system that is associated with the auroral form is curlfree and therefore magnetically invisible below the ionosphere. 5. Our observations and analysis results are generally in good accordance with those of Rostoker et al. (1987) and Liu and Rostoker (1993) on auroral ®ngers, and with the observations of Nakamura et al. (1993) on N-S aurora. An essential dierence to the model of Liu and Rostoker (1993) on the origin of auroral ®ngers is that for our event, the westward-drift motion of the auroral forms cannot be explained by theĩ Âf drift. Furthermore, we cannot support the conclusion of Nakamura et al. (1993) that the magnetic eect of N-S auroras has a similar con®guration than that of the Harang discontinuity. Our resulting current system also ®ts well to the predictions of the ionospheric manifestation of bursty bulk¯ows by Chen and Wolf (1993). However, the predicted westward magnetospheric electric ®eld associated with the BBF which would be expected to map in the ionosphere in their model is not visible in the STARE measurements.