Substorms: Externally Driven Transition to Unstable State a few Minutes Before Onset

Substorms: Externally Driven Transition to Unstable State a few Minutes Before Onset L. R. Lyons 1, I. O Voronkov 2, J. M. Ruohoniemi 3, E. F. Donovan 4 1 Department of Atmospheric Sciences, University of California, Los Angeles 45 Hilgard Ave., Los Angeles, CA 995-1565 larry@atmos.ucla.edu 2 Department of Physics, University of Alberta Edmonton, Alberta T6G 2J1, Canada igor@space.ualberta.ca 3 TheJohns Hopkins University Applied Physics Laboratory 111 Johns Hopkins Rd., Laurel, MD 2723-699. mike_ruohoniemi@jhuapl.edu 4 Department of Physics and Astronomy, University of Calgary Calgary, Alberta T2N 1N4, Canada eric@phys.ucalgary.ca Abstract. Auroral breakup at substorm onset often is observed along a new arc that forms a few minutes prior to the time traditionally identified as substorm onset. The breakup arc increases in intensity approximately monotonically prior to onset and then becomes nonlinear at onset, suggesting that the magnetosphere undergoes a transition to an unstable configuration a few minutes prior to onset and that the resulting instability is responsible for the substorm expansion phase. A reduction in the strength of dayside convection is also often observed to initiate few minutes prior onset, and we suggest that this reduction causes the magnetosphere to make the transition from stability to instability. If this inference is correct, then the critical outstanding question concerning substorms is how does a convection reduction cause the transition to instability and how does this transition lead to formation of the breakup arc prior to onset and to the substorm current wedge. 1. Introduction There is general agreement that auroral breakup and current wedge formation at substorm onset initiate on magnetic field lines of the near-earth plasma sheet, which cross the equator and equatorial radial distances r ~ 6-1R E. The major outstanding question is what causes the onset of the substorm expansion phase. Theories have generally treated the expansion phase as an internal instability of the magnetosphere that develops from the storage of energy that occurs in the nightside plasma sheet and lobes when the interplanetary magnetic field (IMF) is southward. However, it has long been known that periods of steady southward IMF can persist for several hours without the occurrence of substorms. Such periods of stable, enhanced magnetospheric convection are referred to as convection bays or steady magnetospheric convection periods (SMCs). Their existence implies that prolonged enhanced convection does not drive the magnetosphere into an unstable configuration that leads to substorms. Rather the magnetosphere is driven into a stable equilibrium configuration with enhanced energy content on the nightside. It is reasonable to assume that the equilibrium energy content of the nightside magnetosphere increases with the strength of convection. This would imply that a reduction in the strength of convection should result in an equilibrium configuration for the magnetosphere with reduced energy content on the nightside. Thus if the

energy accumulated on the nightside during a period of enhanced convection exceeds that of the equilibrium configuration after a reduction in the strength of convection, the excess energy would represent free energy that should be released after the reduction of convection [Atkinson, 1991]. This energy release process would then be the substorm expansion phase. It has now been established that substorm onsets can be triggered by IMF changes that are expected to lead to a reduction in the strength of convection [Lyons et al., 1997, and references therein]. This is as expected from the above scenario. It thus seems plausible that it is the reduction in the strength of convection that causes the transition to instability that leads to the substorm expansion phase. In this paper we first discuss evidence inferred from auroral breakup observations [Lyons et al., 22a] that the transition to instability occurs ~4-5 min prior to the time normally identified as substorm onset. This discussion includes evidence that the instability behaves as a classical instability, growing monotonically prior to expansion phase onset and becoming non-linear at onset [Voronkov et al., 2]. We then discuss evidence that the transition to instability occurs at the same time as the strength of convection starts to decrease in response to an appropriate IMF change, consistent with convection reduction causing the transition to instability. 2. Formation and Growth of Breakup Arc In the classic paper describing the auroral morphology of a substorm, Akasofu [1964] described the first indication of the substorm expansion phase as a sudden brightening of one of the quiet auroral arcs that forms during the substorm. Using modern all-sky-imager (ASI) and meridian-scanning photometer (MSP) observations of substorm onsets from the Canadian CANOPUS program, Lyons et al. [22a] have recently found that auroral break-up at expansion phase onset does not generally occur along a pre-existing growth phase arc. Instead it at least often occurs along a thin new arc that forms equatorward of all growth phase arcs a few minutes prior to expansion-phase onset. For the cases examined, this breakup arc became discernible 2-8 min before onset and grew in intensity monotonically until onset. After onset, the arcs intensity grows explosively and their shape becomes distorted by the development of large-swirls. Figure 1 gives an example of ASI observations from Lyons et al. [22a] that shows the formation and evolution of the breakup arc for times surrounding a substorm onset at 459 UT on January 18, 1996. Notice that the breakup arc became discernible at ~454 and then increased in intensity. The arc began to develop swirls at 459 UT, the time identified as substorm onset by traditional indicators. 45 Growth phase arcs 451 452 453 454 455 5577 Å Log R Break-up arc 456 457 458 459 (Onset) 54 Figure 1. 5577 Å emissions from the Gillam ASI for periods surrounding an onset at 459 UT on January 18, 1996. North is to the top and east is to the right in each image. A grid of magnetic latitude and longitude at 2 intervals is overlaid on the first image, the thicker horizontal and vertical lines indicating 67 latitude and 33 longitude respectively. Intensity scales have been adjusted to emphasis the relatively low auroral intensities prior to and at onset (from Lyons et al. [22a]).

96/1/18 Gillam 5577 Å Growth-phase arcs Break-up arc 96511 Gillam 5577 Å Growth-phase arcs Break-up arc UT UT 42 455 416 45 412 445 44 48 435 44 43 equatorward poleward Figure 2. Gillam MSP observations of 5577 Å intensities versus elevation angle from the most equatorward to the most poleward looking direction. Lines are shown for observations taken every 3 s from 43 to 51:3 UT on January 18, 1996 and are stacked vertically with increasing time (from Lyons et al. [22a]). Figure 2 shows line plots of the Gillam highresolution MSP observations of 5577 Å emissions. Emission intensities are plotted as a function of elevation from the most equatorward looking direction to the most poleward looking direction. Thin dashed lines are drawn through intensity peaks to indicate the location of the growth-phase arcs. These arcs appear to form and fade away over a time scale of ~1-15 min. The breakup arc is identified with a heavy dashed line. This arc can be seen to be a new arc that forms equatorward of all growth phase arcs, consistent with what is seen in the ASI images. Line plots of the MSP observations from a pseudobreakup with onset at 421 UT on May 11, 1996 are shown in Figure 3. As with the previous example, the data show that the breakup arc formed equatorward of the growth arcs. It was 4 equatorward poleward Figure 3. Same as Figure 2, except observations are shown for 4 to 423 UT on May 11, 1996 (from Lyons et al. [22a]). first discernible at ~416 UT and then increased in intensity before onset. Figure 4 shows the peak intensity of the breakup arcs in Figures 2 and 3 as a function of time from when the arc became discernible to the time of onset. The background 5577 Å intensity in the region of the breakup up arcs prior to arc formation has been subtracted. This figure shows that the increase in arc intensity prior to onset was essentially monotonic. These observations imply that the processes responsible for expansion-phase onset often initiate on breakup field lines at least ~4-5 min prior to the time normally identified as onset, and they are consistent with a classical instability that grows monotonically prior to onset and then becomes nonlinear [e.g., Voronkov et al., 2]. Thus the transition of the magnetosphere to the unstable configuration that leads to the substorm expansion phase must occur ~4-5 min prior to onset.

Increase in 5577Å Intensity (linear scale) -6 * 96/1/18 (459 UT onset) + 96/5/11 (421 UT onset) * * * * * * * * * + + + + + + + + + + * + -4-2 Min from onset Figure 4. Peak intensity of breakup arc between the time of arc formation and onset. The background auroral intensity prior to arc formation has been subtracted 3. Electric Field Reduction If it is the reduction in the strength of convection that causes the transition to instability that leads to the substorm expansion phase, then the reduction in convection should be imparted to the magnetosphere ~4-5 min prior to substorm onset. This reduction should be seen on the dayside if the reduction is indeed a reduction in the strength of large-scale convection imparted to the magnetosphere from the solar wind. While the convection reduction should also be seen on the nightside, such changes may be harder to identify on the nightside because of possible electric field changes associated with nightside auroral phenomena. A number of studies of convection as obtained from radar observations of ionospheric flows have shown that reductions in convection are associated with onsets [Nielsen and Greenwald, 1978; Opgenoorth et al., 1983; Greenwald et al., 1996; Lyons et al., 1998, 21]. Of these, only those of Greenwald et al. [1996] and Lyons et al., [21] included measurements of convection on the dayside. Greenwald et al. [1996] showed that afternoon convection decreased in association with substorm onset, but they did not look at the precise time of convection reduction relative to the time of onset. Lyons et al. [21] took advantage of excellent SuperDARN coverage of convection during a period that included two substorm onsets on November 24, 1996, a small onset at 2227 UT and a large onset at 2233 UT. They found a reduction in the strength of global * convection in association with both onsets, the reduction being significantly less for the small onset than for the large onset. Figure 5 shows line-of-sight ionospheric flows from two SuperDARN radar beams for 22-23 UT on November 24, 1996. The particular radar beams shown in Figure 5 were selected because they had relatively continuous echoes throughout this 1-hour interval. The MLT of the central time series at 22 UT is given in the stack plot for that time series, these being 1526 Vel (m/s) - 25-25 1526 MLT 22 22:3 23 mlat,mlon (deg) 1834 MLT 22 22:3 23 UT (hr) mlat,mlon (deg) (7.1, -22 (69.8, -23 (69.6,-24 (69.3,-25 (69.1,-26 (68.8,-27 (68.5,-28 (68.3,-29) (68..-3) (72.4, 2 (71.9, 2 (71.5, 2 (71.1, 2 (7.7, 2 (7.3, 2 (69.8, 2 (69.4, 2) (68.9, 21) Figure 5. Line-of-sight velocities for two SuperDARN radar beams for 22-23 UT on November 24, 1996. Time series in each plot are for an individual radar beam and are ordered by distance from the observing radar, the largest distance being at the top. Locations of each time series in magnetic coordinates are given along the right hand axes, and the magnetic local time (MLT) given in each plot is the MLT at 22 UT for the central time series in that plot. Negative velocities indicate velocities directed toward the radar (from Lyons et al. [21]).

and 1834 MLT for the top and bottom plots, respectively. The velocities for these beams were primarily negative, which indicates velocities directed toward the radar. Also, each data point was taken at a time during a 2-min interval that began at the time at which the point is plotted. The time series in Figure 5 allow us to determine the times of convection changes. Here we are interested in using the radar data to identify global changes in convection, which we define to be changes seen at approximately the same time in many individual time series. Figure 5 shows a number of transient, localized changes that are not of interest in this context. The vertical dashed and solid lines in Figure 5 identify times when a change in slope of the time series was observed at the same time in many of the time series shown in Figure 5. These lines identify the initiation of global changes in convection. Note that the precise times the changes initiated is not known to better than the 2-min resolution of the data. Most of the time series in Figure 5 show a gradual increase in line-of-sight speeds from 22 to 2224 UT, indicating a general enhancement of convection speed. The solid vertical lines in Figure 5 demarcate a time that was followed by a distinct and prolonged decrease in flow speeds. It is identified at the 223 UT data points, which indicates a large decrease in convection strength initiating within ~2 min of 223 UT. This is ~3 min prior to the time that the second, and larger, of the two expansion onsets was observed. The dashed vertical lines in Figure 5 demarcate a smaller, but still clearly identifiable reduction in global convection that initiated within 2 min of 2224 UT, which is few min prior to the first expansion onset. The observations on November 24, 1996 indicate that the reduction in global convection does indeed initiate a few minutes prior to expansion phase onset, as required if such a reduction is responsible for the transition to instability that leads to the substorm expansion phase. However, it is necessary to see whether a reduction in convection is often seen in association with substorm onsets and if so, does the reduction generally initiate a few minutes prior to ground onset. Lyons et al [22b] have identified a number of isolated substorm onsets for which there is good SuperDARN radar echo coverage on the dayside. They have found that a reduction in the strength of large-scale convection is indeed identifiable on the dayside in association with most of the identified substorm onsets and that the reduction initiated a few minutes before onset. In Figure 6, we show SuperDARN observations for two of these onsets that clearly show the timing of the convection reduction. Line-of sight velocities have been filtered and gridded as described by Ruohoniemi and Baker [1998]. Polar plots of radar line-of-sight flows at times before and after onsets at 913 UT on February 19, 1998 and at 1457 on October 22, 1998 are shown in Figures 6a and 6b, respectively. These plots show the data coverage on the dayside and give an overall view of the change in convection from before to after the onsets. When looking at these plots, it must be remembered that only one component of the total plasma flow velocity is given by each arrow. It is expected, however, that a change in speed that is seen in a large number of vectors indicates an overall change in large-scale convection. A reduction in speed after onset is seen in Figure 6a at nearly all measurement locations for the February 19 event, indicating a reduction in the strength of large-scale convection. The reduction in convection is not quite as clear following the October 22 onset because, despite the reduction, convection remained strong after the second onset. However reduction in the magnitude of many of the line-ofsight vectors can be discerned in Figure 6b. To look at the timing of the convection for these two onsets, we show time series of dayside, highlatitude line-of-sight velocities at fixed geomagnetic coordinates in Figure 7. For the February 19 example, all line plots at latitudes >7 that had good coverage of the onset are shown. Essentially all these time series show a prolonged reduction in convection speed initiating ~4-6 min prior to onset. For the October 22 onset, dayside time series at latitudes >7 that best illustrate the timing of the convection change are shown. (There are too many time series for all to be shown.) We do not expect a reduction in the overall strength of convection to be reflected as a reduction in all observed line of sight speeds,

9:5 UT Feb. 19, 1998 12 MLT Ha 6 o Py 7 o 14:53 UT 9 o St Py October 22, 1998 12 MLT 6 o 7 o o 18 MLT 9:17 UT 12 MLT Ha 8 o 3 o Py 6 MLT m/s 6 o 6 o 7 o Ha 8 o 18 MLT 6 MLT 15:1 UT 12 MLT m/s 6 o 7 o St Py o 18 MLT 8 o 3 o 6 MLT Figure 6a. Polar plots (MLT versus magnetic latitude) of line-of-sight velocities observed by SuperDARN radars for times before and after substorm onset at 913 UT on February 19, 1998. Vectors point toward or away from location of the radar that made the measurements. Locations of radar with echoes are indicated by heavy dots (Py, Pykkvibaer; St, Stokkseyri). Hankasalmi is at a lower latitude than shown; its longitude is indicated by Ha (from Lyons et al. [2b]). remembering that line-of-sight velocities are only one component of the ionospheric flows and that movement of the convection reversal boundary can affect the magnitudes of flow speeds at some locations. Thus some time series may show speed increases or speed reversals, particularly at locations near the convection reversal boundary. For the October 22 example a few time series show an increase in speed or a reversal in the direction of flow. However, the majority of the time series show a reduction in the line of sight speeds associated with the onset, indicating a reduction in the overall strength of convection. Also, nearly all the time series show that this change in convection initiated ~4-8 min prior to onset. Thus, as with the November 24, 1996 onsets, the ionospheric flows are consistent with a reduction in large-scale convection being responsible for the transition to instability. Ha 8 o 18 MLT 6 MLT Figure 6b. Same format as Figure 6a (from Lyons et al. [2b]). 4. Summary and Conclusion CANOPUS ASI and MSP observations show that auroral breakup at substorm onset at least often occurs along a new arc that forms a few (~4-5) minutes prior to the time traditionally identified as substorm onset. The breakup arc increases in intensity approximately monotonically prior to onset, and then becomes non-linear at onset. This behavior suggests that the magnetosphere undergoes a transition to an unstable configuration a few minutes prior to onset and that the resulting instability is responsible for the substorm expansion phase. Also, a reduction in large-scale convection is often discernable in association with substorm onset from SuperDARN observations of dayside ionospheric convection. The reduction initiates a few minutes prior to onset. That the reduction in convection proceeds onset and occurs at about the same time relative to onset as does the formation of the breakup arc suggests that the reduction in the strength of convection causes the magnetosphere to make the transition from stability to instability. It is not possible at this time to determine whether all substorms onsets are preceded by a reduction in the strength of

6 2 8 Feb. 19, 1998 Convection change V -1 m/s Onset Pykkvibar V uncertain Pykkvibar 4 6 12 7 6 12 4 1 6 1 Velocity (m/s) 8 1 4 1 12 4 1 1 4 7 4 6 12 4 848 9 912 924 936 848 9 912 924 936 UT -4 Oct. 22, 1998 V -1 m/s V 1 m/s reversal Pykkvibar 12 Stokkseyri -2-6 -2-9 -6-12 -6-4 4 6 8 4 1 1 Velocity (m/s) 2-2 2-2 2 1 6 6 Convection change Onset Stokkseyri 1 4 1 1 2 1 2 1436 1448 1 1512 1524 1436 1448 1 1512 1524 UT Figure 7. Line-of-sight velocities from the SuperDARN radars for geomagnetic longitudes and latitudes within parentheses in each panel. Each data point was taken at a time during a 2-min interval that began at the time at which the point is plotted. Black vertical lines identify the beginning of the 2-min interval in which onset occurred. Thicker, gray vertical lines indicate the data point in each panel where a change in convection speed can be identified in association with the onsets. Decreases, increases, and reversals (sign change) in convection speed of magnitude >1 m/s seen in at least four consecutive data points are indicated by line patterns indicated above each set of line plots (from Lyons et al. [22b]).

large-scale convection. However, observations imply that at least most of those with a welldefined onset time do so. It is reasonable that most other onsets may also occur in this way; however it is not currently possible to determine for sure that this is the case. The instability that results from a reduction in convection would be expected to occur when the equilibrium energy corresponding to the lowered rate of convection is below that which has accumulated during the previous period of enhanced convection. This accumulated energy would be the equilibrium energy if the period of enhanced convection were sufficiently long to be an SMC. Shorter periods of enhanced convection, without sufficient time for equilibrium to be reached, would constitute the typical substorm growth phase and accumulated nightside energy would be below equilibrium. The instability will then reduce the energy stored on the nightside to the equilibrium level appropriate for the lowered strength of convection, and the release in energy would constitute the substorm expansion phase. If this is indeed the case, then the amount of energy released (i.e., substorm size) should increase with amount of convection reduction for a given amount of energy storage prior to the convection reduction. It is also plausible that if the convection strength were to decrease, and then increase back to near or greater than the initial strength within a short period of time (less than ~1 min?), the instability would be terminated before leading to significant energy loss leading to a pseudo-breakup. The inferences in the above paragraph should be viewed as speculative. However, the inference that substorm onset at least often results from a reduction in the strength of convection that is imparted to the magnetosphere a few minutes before onset is strongly supported by the data. If this inference is correct, then the critical outstanding question is how does a convection reduction cause the transition to instability and how does this transition lead to formation of the breakup arc prior to onset and to the substorm current wedge after onset. Acknowledgments This work was supported at UCLA in part by NSF grant OPP-136139 and NASA grant NAG5-7962. CANOPUS data have been obtained with support of the Canadian Space Agency. We thank the CANOPUS PI, John Samson, for his strong support. Support for SuperDARN was provided by national funding agencies in Canada, France, the United Kingdom, and the United States. References Akasofu, S.-I., The development of the auroral substorm, Planet. Space Sci., 12, 273, 1964. Atkinson, G., A magnetosphere wags the tail model of substorms, in Magnetospheric Substorms, eds. J. R. Kan, T. A. Potemra, S. Kokubun, and T. Iijima, p. 191, American Geophysical Union, Washington, 1991. Greenwald, R. A., et al., Mesoscale dayside convection vortices and their relation to substorm phase, J. Geophys. Res., 11, 21,697, 1996. Lyons, L. R., et al., Coordinated observations demonstrating external substorm triggering, J. Geophys. Res., 12, 27,39, 1997. Lyons, L. R., G. T. Blanchard, and K. B. Baker, Substorm onset: The result of IMF-driven reductions in large-scale convection, in Substorms-4, ed. by S. Kokubun and Y. Kamide, p. 265, Kluwer Acad., Norwell, Mass., 1998. Lyons, L. R., J. M. Ruohoniemi, and Gang Lu, Substorm-associated changes in large-scale convection during the November 24, 1996 Geospace Environment Modeling Event, J. Geophys. Res., 16, 397, 21. Lyons, L. R., I. O. Voronkov, E. Donovan, and E. Zesta, Relation of substorm breakup arc to other growthphase auroral arcs, J. Geophys. Res., 22a (in press). Lyons, L. R., J. M. Ruohoniemi, S. Liu, S. I. Solovyev, J. C. Samson, Observations of dayside convection reduction leading to substorm onset, J. Geophys. Res., 22b (submitted). Nielsen, E., and R. A. Greenwald, Variations in ionospheric currents and electric fields in association with absorption spikes during the substorm expansion phase, J. Geophys. Res., 83, 5645, 1978. Opgenoorth, et al., Three-dimensional current flow and particle precipitation in a westward traveling surge (observed during the Barium-GEOS rocket experiment), J. Geophys. Res., 88, 3138, 1983. Ruohoniemi, J.M., and K.B. Baker, Large-scale imaging of high-latitude convection with Super Dual Auroral Radar Network HF radar observations, J. Geophys. Res., 13, 2797, 1998. Voronkov, I., E. F. Donovan, B. J. Jackel, and J. C. Samson, Large-scale vortex dynamics in the evening and midnight auroral zone: Observations and simulations J. Geophys. Res., 15, 18,55, 2.