Post-noon two-minute period pulsating aurora and their relationship to the dayside convection pattern

Poleward-moving auroral forms, as observed by meridian-scanning photometers, in the vicinity of the cusp region are generally assumed to be the optical signature of flux transfer events. Another class of quasi-continuous, short period (1–2 min) wave-like auroral emission has been identified, closely co-located with the convection reversal boundary in the post-noon sector, which is similar in appearance to such cusp aurora. It is suggested that these short period wave-like auroral emissions, the optical signature of boundary plasma sheet precipitation in the region 1 field-aligned current system, are associated with ULF magnetohydrodynamic wave activity, which is observed simultaneously by ground magnetometer stations. This association with ULF wave activity is strengthened by the observation of several harmonic frequencies in the pulsation spectrum, each an overtone of the fundamental standing wave resonance frequency.


Introduction
A recent study of optical and radar aurora in the vicinity of the cusp region (Milan et al., 1999) indicated the presence of quasi-periodic poleward-moving auroral forms (PMAFs) associated with enhancements of the convection velocity. Whilst bearing some similarity to the expected optical signature of¯ux transfer events (Sandholt et al., 1986;Fasel, 1995;, 1997Farrugia et al., 1998), Milan et al. (1999) suggested that the highly wave-like nature and 2 min period-icity of the PMAFs were inconsistent with the typical repetition rate of transient reconnection. The study of Milan et al. (1999) was based on the ®rst day of an eight day experimental campaign conducted between 17 and 25 December, 1995. Subsequently it was found that this short period wave-like auroral phenomenon is a common feature of dayside optical emission, being present on 5 out of the 8 campaign days. The present study employs observations of the red and green line emission by a meridian-scanning photometer located at Ny A Ê lesund, Svalbard, Norway, from 18 December 1995, to investigate these PMAFs in more detail.
Plasma drift measurements from the CUTLASS Finland coherent HF radar allow the optical observations to be placed within the context of the dayside convection pattern. Ground-based magnetometers from the IMAGE network are employed to determine the relationship between the auroral observations and ULF wave activity in the magnetic ®eld. Supporting observations are provided by a white light all-sky camera located at Longyearbyen, Svalbard, and low-altitude particle precipitation measurements from the polarorbiting DMSP F13 satellite.

Experimental arrangement
The SuperDARN coherent HF radars  employ backscatter from high-latitude ®eld-aligned ionospheric plasma density irregularities (radar aurora) as tracers of the bulk plasma motion under the in¯uence of the convection electric ®eld, and hence as a diagnostic tool for the investigation of largescale magnetosphere-ionosphere coupling. The Co-operative UK Twin Located Auroral Sounding System (CUTLASS) forms the eastern-most pair of radars of SuperDARN, of which the Finland radar, located at Hankasalmi (62. 3°N, 26.6°E), is the focus of the present study (see Fig. 1). Sixteen beams are sounded sequentially with a dwell time of 7 s, completing a 52°azimuth scan every 2 min. Seventy-®ve range gates are sampled for each beam, with a gate length of 45 km and a range to the ®rst gate of 180 km. A 7 pulse scheme is transmitted, and analysis of the auto-correlation function of the returned signals yields backscatter Doppler spectra, from which the mean Doppler velocity, an estimate of the line-of-sight plasma drift velocity, and special width can be derived.
The meridian-scanning photometer (MSP) located at Ny A Ê lesund (78.9°N, 11.9°E) Svalbard, scans roughly along the magnetic meridian, 35°west of geographic north, to 10°above the northern and southern horizons, with a scan period of 18 s. The meridian closely follows beam 9 of the CUTLASS Finland radar (see Fig. 1). The line-of-sight intensities of 630.0 nm (red line) and 557.7 nm (green line) auroral emissions are measured, corresponding to transitions from the ID and IS metastable states of atomic oxygen (OI). The red line emissions can have an altitude of peak emission intensity between 250 km and 550 km, excited by electrons of energies of order 100 eV. In the present study, a red line emission altitude of 300 km is assumed, as this gives an excellent correspondence between the location of the aurora and the radar backscatter (see Sect. 3.3). Green line emissions are generally assumed to originate from an altitude between 110 km and 150 km (Sigernes et al., 1996) However, in the present study, it is found that an emission altitude of 200 km for the green line gives an exact match between the red and green line observations; this will be discussed in Sect. 4. In addition, observations from a white light all-sky camera (ASC) located at Longyearbyen (78.1°N, 16.0°E), Svalbard, are employed to allow the azimuthal extent of auroral features to be determined. The approximate extent of the ASC ®eld-of-view is illustrated in Fig. 1.
Fifteen IMAGE magnetometer stations were also employed in the study (see Fig. 1) X, Y and Z geographic component magnetic ®eld direction measurements were made with a sampling resolution of 10 s. Most stations had a resolution of 1 nT, though the three lowest latitude stations, OUJ, HAN, and NUR, have a resolution of 0.1 nT. The measurements presented in this study have been rotated such that the X mag and Y mag coordinate axes are positive towards geomagnetic (IGRF) north and east, respectively.

Observations
The bulk of observations discussed in the present study were made between 0900 and 1300 UT, 18 December 1995. During this interval the ground-based instrumentation was located in the noon and post-noon sectors, the MSP meridian progressing from 12 MLT to 16 MLT. Near winter solstice, Svalbard is in darkness even at local noon, allowing optical observations of the cusp and other dayside auroral regions to be made.

Solar wind and interplanetary magnetic ®eld observations
During the period of interest the WIND spacecraft was located upstream of the Earth's bowshock, near (X GSM , Y GSM , Z GSM ) = (49 R e , 34 R e , 0 R e ). The interplanetary magnetic ®eld (IMF) observations showed a southward turning at 0920 UT after which B z remained negative, varying between A3 and A4 nT, for the remainder of the interval, except for a brief return to f z % 0 nT at 1000 UT. The B y component varied between 0 and 2 nT throughout this interval except for a brief excursion to A3 nT at 1000 UT, and B x varied between 0 and 2 nT. The solar wind speed and density were of the order of A350 km s A1 and 4 protons cm A3 . A solar wind delay of some 12 min is expected between the spacecraft and the bowshock. A further 5 to 10 min delay is then possible to the ionospheric footprint of the cusp region.

Radar observations
The top six panels of Fig. 2 illustrate the line-of-sight Doppler velocity measurements and the backscatter spectral width from beams 2, 9 and 13 of the CUTLASS Finland radar, between 0930 UT and 1300 UT. Positive (negative) velocities represent motion towards (away from) the radar. Signi®cant radar backscatter was ®rst observed shortly after 0930 UT, possibly in response to the southward turning of the IMF observed to the WIND spacecraft at 0920 UT. Most ionospheric backscatter of interest occurred between geomagnetic latitudes of approximately 75°and 81°in the AACGM (altitude adjusted, corrected geomagnetic) coordinate system (Baker and Wing, 1989). As the dusk solar terminator swept westward across the radar ®eld-of-  Fig. 1. The location of the ®eld-of-view of the CUTLASS Finland radar; beams 2, 9, and 13 are indicated by dashed lines. The Ny A Ê lesund MSP meridian, projected to an altitude of 300 km, and the ®eld-of-view of the Longyearbyen ASC are also indicated, as are the 15 IMAGE ground magnetometer stations employed in the study. The approximate location of the dayside auroral oval for quiet conditions is superimposed for context view between 1130 UT and 1300 UT, HF propagation to far ranges was no longer supported and ionospheric backscatter was lost at 1150 UT, 1200 UT and 1230 UT in beams 13, 9 and 2, respectively. Beam 2 observes predominantly negative line-of-sight velocities, indicative of poleward and westward plasma drift, and high spectral widths, above 400 ms A1 . In beam 9, the line-of-sight drift velocity is generally slower than that observed in beam 2, though as the beam closely follows a magnetic meridian all information regarding the azimuthal component of the¯ow is lost. Between 0950 UT and 1020 UT, the observed spectral width is high, above 400 ms A1 . After 1020 UT the spectral width falls in general to below 300 ms A1 and after 1130 UT to 200 ms A1 . Beam 13 indicates the presence of a convec-tion¯ow reversal, positive and negative velocities being observed equatorward and poleward of approximately 77°CGMLAT. This corresponds to azimuthal¯ow, eastward at higher latitudes and westward at lower latitudes. The spectral width is in general low, mainly below 300 ms A1 , at all times in beam 13. Examination of spatial maps from the interval, of which that presented in Fig. 3 is a typical example, help clarify the¯ow pattern. The observed line-of-sight velocities are consistent with a convection reversal in the post-noon sector, near a latitude of 77°,¯ow being sunward (westward) at lower latitudes and antisunward (eastward) at higher latitudes. In this region, the spectral widths are low, less than 300 ms A1 in general. In the noon sector, the observed¯ows are consistent with westward and antisunward plasma drift and the spectral width is high, above 400 ms A1 in general. Figure 4 reproduces ion and electron spectrograms observed during an overpass of the DMSP F13 satellite from east to west through the radar ®eld-of-view between 1022 UT and 1036 UT. Along the satellite track (illustrated in Fig. 3) regions of central plasma sheet (CPS), boundary plasma sheet (BPS) and mantle precipitation can be identi®ed from the criteria of Newell et al. (1991). These regions (CPS: orange; BPS: red; mantle: blue/green) have been indicated at the top of Fig. 4 and along the satellite track of Fig. 3; gaps between these regions correspond to portions of the spectrograms that are deemed``void'' or``unclear'' by the Newell et al. (1991) criteria. The transition from BPS to mantle precipitation, i.e. the boundary between closed and open ®eld lines, is closely co-located with the convection reversal boundary observed by the CUTLASS radar in the post-noon sector. It is also suggested that the sunward¯ow region of the radar backscatter is associated with BPS precipitation, and no backscatter appears at the ionospheric footprint of the CPS region. Examination of the spectrograms from this overpass suggests that the mantle precipitation region can be further subdivided, a marked increase in the¯ux of low energy ions (indicated in green) being coincident with the noon sector region of high spectral width backscatter.
The¯ow pattern surmised from the observations is sketched schematically in Fig. 5. This pattern is found to remain highly stable throughout the interval from 0950 UT to 1200 UT or later, concordant with the relatively uniform nature of the IMF during this period. Variations in the measurements made by the radar are due principally to the rotation of the Earth under this stable convection pattern and the progression of the radar ®eld-of-view to later local times. For instance, the decrease in spectral width observed in beam 9 near 1020 UT (Fig. 2) corresponds to the edge of the noon sector high spectral width region, located near 0130 to 0200 MLT (Fig. 3).

Optical observations
Between 0900 UT and 1300 UT the MSP meridian was located between 12 MLT and 16 MLT. The auroral observations for this interval are illustrated in the bottom two panels of Fig. 2, and have been projected  Fig. 3. The line-of-sight velocity and spectral width observed by the radar during the scan starting 1026 UT, projected onto geomagnetic latitude and local time coordinate axes; backscatter with low velocity and spectral width equatorward of 72°C GMLAT is ground backscatter and of no interest to the present study. The location of the statistical auroral oval is included, as well as the 300 km projection of the Ny A Ê lesund MSP meridian. Beam 9 of the radar is indicated by dashed lines. Overlaid is the east-west pass of DMSP F13 between 1023 UT and 1034 UT to a geomagnetic latitude frame by assuming peak emission altitudes of 300 km and 200 km for the red and green lines, respectively. Prior to approximately 1030 UT the aurora are diuse and weak, the red line peaking at an intensity of some 1.5 kR, and the green line barely visible above the ambient background glow, an auroral con®guration known as the``midday gap'' (Dandekar and Pike, 1978). After 1030 UT the auroral emission intensity increases in both lines. Between 1030 UT and approximately 1130 UT, the intensities in the two lines are comparable, peaking between 3 and 6 kR. After 1130 UT, however, the intensity in the red line diminishes to below 2 kR and the green line is enhanced, peaking near 10 kR.
The region of main auroral emission occupies a latitudinal region which varies in width between approximately 0.5°and 1.5°, and has a general equatorward progression from 78°at 1030 UT to 72.5°at 1300 UT. In Fig. 2, the radar data for beam 9, the beam most aligned with the MSP meridian, has superimposed on it the equatorward boundary of the red line aurora as determined from the latitude at which the emission intensity falls to 1 kR, for the period 0950 UT to 1205 UT. Throughout this interval, an excellent agreement is found between the locations of the equatorward boundaries of the radar and optical aurora, con®rming the assumed emission altitude of the red line aurora of 300 km. Also, importantly, this co-location is found  Fig. 5. A schematic representation of the radar and optical observations for the interval of study. The radar ®eld-of-view and MSP meridian are indicated for 1130 UT. The pulsating aurora are observed co-located with BPS precipitation, located between the post-noon convection reversal boundary and the equatorward boundary of the radar aurora between 1000 UT and 1030 UT, when the equatorward boundary of the red line aurora is overhead the MSP zenith and hence the mapping of its location is independent of the assumed emission altitude.
Throughout the interval 1030 UT to 1300 UT, the red and green aurora take the form of poleward-moving intensity enhancements which have a duration and repetition rate near 2 min. A closer examination suggests that these poleward-moving forms are wave-like in nature and occur in wave packets. Five such packets, numbered I to V, have been indicated in Fig. 2 and are reproduced in more detail in Fig. 6. For the ®rst three intervals, both the red and green line emission intensities are illustrated; in the case of intervals IV and V, only the green line observations are shown as the red line emissions are weak and show no signi®cant structure. The correspondence between the variations in the red and green lines in examples I to III suggests that the emissions are excited by a single source of precipitating particles with a relatively broad range of energies. It also gives con®dence in the 200 km assumed emission altitude of the green line observations in this study. Each wave packet tends to be between 7 and 10 cycles in duration. To the eye, each wave packet, especially I, II, IV, and the ®rst half of V, displays a near 180°phase change in the emission intensity variation across the central latitude of the wave structure. The green line observation of wave packet II, being perhaps the most well-de®ned example, is illustrated in more detail in Fig. 7. The emission intensities measured at the 26 MSP scan positions which map to latitudes between 78.1°and 78.9°for the interval 1127 UT to 1137 UT are indicated in Fig. 7a. The self-normalised Fourier spectra of each time series, illustrated in Fig. 7b, indicate that the predominant spectral component is 12.2 mHz (82 s) at all latitudes. The relative spectral power and phase in this component are indicated as a function of latitude in Fig. 7c, d respectively, and con®rm the intensity and phase structure of the wave discussed. A similar analysis of wave packet IV shows that its more complicated structure is produced by the mixing of several frequency components, 7.1 mHz and 14.1 mHz (141 s and 70 s) being dominant. Interestingly, these two components are harmonically related, the second being twice the frequency of the ®rst. This is illustrated in Fig. 8a, a time series of the optical intensity at a latitude of 77.1°d uring wave packet IV: the double-humped peaks of the oscillation are characteristic of frequency-doubling. Again, an approximate 180°phase change over the latitude range is observed in each frequency component. Wave packets I, III, and V contain the frequency components 6/12 mHz, 10/15 mHz, and 10 mHz, respectively.
The MSP observations allow no estimation of the east-west extent of these auroral forms. This information can be gained from white light all-sky camera images from Longyearbyen, collected during the period of interest. Four typical examples, from 1222 UT and 1230, 1231 and 1232 UT (during and at the end of wave packet IV), are illustrated in Fig. 9, rotated such that geomagnetic north and east are directed towards the top and left, respectively; the last three images form a time series, with a time step of 60 s. The bright feature to the extreme SW of each image is light contamination. The times of these four images have been chosen somewhat arbitrarily, though they are from the end of the period of study when the green line emission intensity dominated over that of the red line (see Fig. 2) and hence the optical emission originated predominantly from a single altitude only. The optical features corresponding to the MSP observations are signi®cantly extended in azimuth, though are not exactly L-shell aligned, sloping slightly from SE to NW. Multiple auroral forms are present simultaneously, spaced azimuthally along approximately constant latitude, producing a``twill'' pattern. Assuming an emission altitude near 200 km, the azimuthal separation of the features (the azimuthal`w avelength'') can be estimated to be of the order of 500 km.

Ground-based magnetometer observations
Ground-based magnetometer observations from the interval of interest indicate the presence of quasicontinuous, though low amplitude, wave activity. Figure 10 illustrates the X and Y component magnetograms from a high-latitude magnetometer station close to the location of the optical aurora, HOR, and a mid-latitude station, NUR. Wave activity is most evident in the X component of the mid-latitude station. This wave activity is perhaps most clear between 1200 UT and 1230 UT, around the time of wave packet IV in the optical observations (see Figs. 8 and 10), and this interval has been selected for more detailed analysis. Figure 11a and c illustrates the X and Y component magnetograms from the latitudinal IMAGE chain between 1200 UT and 1230 UT. The data have been high-pass ®ltered with a cut-o period of 600 s, to suppress frequencies below approximately 2 mHz. Self-normalised spectra for each station and component are illustrated in Fig. 11b, d. The spectra indicate that at the higher latitude stations, especially LYR and HOR, two frequency components near 3.5 to 4 mHz and 7 mHz (indicated by vertical dotted lines) are present in both X mag and Y mag . A spectral peak is also observed near 2 mHz, though its proximity to the cut-o frequency of the applied ®lter places its signi®cance in doubt; the only time at which such a frequency (equivalent to a period of 7 to 8 min) is clearly apparent in the magnetograms is after 1210 UT (as seen in the Y component of HOR, Fig. 10). The lower frequency components also dominate the magnetograms over a broad range of latitudes. At the lowest latitude stations, OUJ, HAN, and NUR, the 7 mHz frequency component is again observed, predominantly in the X mag magnetograms. At both higher (NAL, LYR, HOR) and lower (OUJ, HAN, NUR) latitudes, the latitudinal phase variation of the two frequency components, especially the 7 mHz component, is in the same sense as the phase variation observed in the optical emission, with phase lag increasing with increasing latitude.

Discussion
The observations presented can be summarised as follows. During an interval of relatively stable IMF f z`0 nT and f y J0 nT conditions, HF radar observations indicate that the dayside convection pattern consisted of a convection¯ow reversal near a latitude of 76°CGMLAT in the post-noon sector, and poleward and westward¯ow in the local noon sector. In the noon sector, the radar backscatter had high spectral width which, together with simultaneous large poleward velocities, has previously been interpreted as the ionospheric footprint of the cusp region (Baker et al., 1990. Prior to 1030 UT, when the Ny A Ê lesund MSP was located underneath this region,``midday gap'' aurora were observed (Dandekar and Pike, 1978). Such an auroral con®guration is generally attributed to precipitation of magnetosheath origin, consistent with the cusp interpretation of the radar measurements. Moreover, low-altitude particle measurements indicated that this region corresponded to mantle precipitation with a signi®cant¯ux of low energy ions, indicative of recently reconnected ®eld lines. In the post-noon sector, the lowaltitude particle observations indicated that the radar backscatter equatorward of the convection reversal boundary, in the sunward return¯ow region, was colocated with BPS precipitation on closed ®eld lines. Poleward of this, in the antisunward¯ow region, mantle precipitation was observed, indicating open ®eld lines. No radar backscatter was observed in the CPS precipitation region. A continuous (though pulsating) optical auroral feature was observed in both red and green lines in the post-noon sector, co-located with the convection return¯ow backscatter region and BPS precipitation; this optical feature was present between 1030 UT and 1300 UT (when observations ceased), corresponding to between 1330 MLT and 1600 MLT. No signi®cant optical emission was observed in the CPS or mantle precipitation regions. The signi®cance of boundaries in HF radar backscatter has been discussed previously. Milan et al. (1998) reported that in the dayside auroral zone both the equatorward and poleward edges of radar backscatter are in general, independent of instrumental limitations such as propagation mode, and hence must correspond to geophysical boundaries, though the nature of these boundaries was not elucidated. Furthermore, Rodger et al. (1995) and Milan et al. (1999) demonstrated the co-location of dayside red line optical aurora and the equatorward portion of radar backscatter, as shown in the present study, again indicating a concrete relationship between radar aurora and dayside precipitation regions. The present study suggests that the equatorward boundary of post-noon optical emission and radar backscatter corresponds to the transition from CPS to BPS precipitation. Radar aurora are also observed in regions of mantle precipitation, and it is interesting to note that the observed spectral width of the backscatter appears to increase with the¯ux of low energy ions. In general, red line optical emissions are excited by precipitation of low energy (of order 100 eV) electrons, consistent with the energy characteristics of unaccelerated magnetosheath plasma of cusp/cleft origin on newly-reconnected ®eld lines; hence in the noon sector the equatorward boundary of the red line aurora has been identi®ed as the open/closed ®eld line boundary (e.g. . Away from noon, however, the aurora typically takes the form of a continuous arc, as observed in the present case. Sandholt et al. (1998) termed this optical feature``type 7 aurora'' in the postnoon sector, and concluded that, if located in a region of sunward convection, it was the optical manifestation of BPS precipitation. This is consistent with our observations. Hence, the open/closed ®eld line boundary occurs at the poleward edge of the aurora, in our case observed as the transition between BPS and mantle precipitation, closely co-located with the convection reversal boundary. A similar correspondence between the red line aurora and the nightside polar cap boundary has been reported by Blanchard et al. (1995). Potemra (1994) identi®ed LLBL/BPS precipitation as the region 1 Birkeland current system, which in the postnoon sector comprises an upward ®eld-aligned current (FAC), carried by precipitating electrons. This FAC is an integral component of the magnetospheric/ionospheric convection process, balancing the non-zero divergence of the convection electric ®eld in the vicinity of the post-noon convection reversal. The fundamental ®nding of the present study is the highly periodic modulation of the optical emission intensity associated with this current system. This modulation has a period near 1 to 2 min, within the Pc4 ULF magnetohydrodynamic wave period range. The latitudinal phase variation of the modulation gives the optical aurora the appearance of poleward-moving auroral forms (PMAFs) when viewed with a meridian-scanning photometer. These features are similar in nature to observations of Pc5 ULF waves in VHF radar measurements (e.g. Walker et al., 1979), where the latitudinal phase variation, a 180°phase change across the latitude of the pulsation amplitude maximum, is found to be consistent with ®eld line resonance (FLR) structures (e.g. Orr and Hanson, 1981). The PMAFs in the present study appear to be between 20 km and 50 km in latitudinal extent and progress between 50 km and 150 km along the MSP meridian. The auroral forms are extended approximately 1000 km to 1500 km in the east-west direction, similar to previous observations of systems of multiple discrete arcs located within the region 1 FAC system (Moen et al., 1994(Moen et al., , 1995, and slope slightly from SE to NW. Multiple auroral forms have an azimuthal separation of approximately 500 km, giving a``twill'' appearance. For consistency with their apparent poleward motion along the meridian, these forms must also have an eastward (antisunward) phase motion; this motion is indicated by lettering in the last three panels of Fig. 9. For a pulsation with a period of 120 s the velocity of this phase motion would be approximately 4 km s A1 ; when observed with insucient temporal resolution, this phase motion will not be apparent. Overall, then, the observed phase structure of the aurora is consistent with an FLR with azimuthal wave number m % 20 to 30. The relationship between the optical observations and ULF FLR events is further strengthened by the presence of multiple harmonic frequency components within the Fourier spectrum of the modulation of the emission intensity.
To draw a more direct link between ULF wave activity and the pulsating aurora, ground-based magnetometer measurements were compared with the optical observations. In general, it was found that clear pulsations were not apparent in the magnetograms, especially at auroral latitudes. This is perhaps to be expected as ground magnetometers integrate over a large area of the sky, of order 100 km in spatial extent (Hughes and Southwood, 1976), comparable with the latitudinal width of the pulsating aurora, and hence considerable phase-mixing will occur. The MSP observations do not suer from this drawback, and hence the latitudinal phase and amplitude variation of the wave is clear. The clearest example of simultaneous magnetic and optical modulation occurred in the case of wave packet IV, between approximately 1205 UT and 1225 UT (see Fig. 8). At this time, the frequency components 7.1 mHz and 14.1 mHz were present in the aurora, and 3.5 to 4 mHz and 7 mHz (and possibly 2 mHz) were observed on the ground at the latitude of the aurora. Clearly, a close harmonic relationships exists between the optical and magnetometer observations. The optical frequency components appear to be close to twice the frequency of the magnetometer frequency components. It is also possible that the higher frequency component of the magnetic variations is a harmonic of the lower frequency component. This will be discussed in more detail later, but ®rst it is interesting to compare the magnetometer estimate of m with that determined from the optical observations. The stations HOR and HOP are separated in longitude by approximately 10°, and the phase measurements of the 7 mHz frequency component indicate that HOP leads HOR by some 90°in the Y component. The usual interpretation of this phase lag would be westward phase propagation with m % 9, which is not consistent with the auroral observations of eastward propagation with m % 20 to 30. Conversely, The fundamental eigenperiod of FLRs, essentially standing waves along closed quasi-dipolar magnetic ®eld lines, is twice the AlfveÂ n travel time between conjugate ionospheres. At the latitude of, say, HOR the fundamental period is calculated to be of the order of several minutes (Orr and Hanson, 1981) and is observed to be between 5 and 10 min (Poulter et al. 1984). Hence, the 3.5 to 4 mHz frequency component of the magnetogram oscillation could be consistent with either the fundamental mode or second harmonic mode. In the latter case, the fundamental component would be expected near 1.7 mHz (which possibly corresponds to the 2 mHz component observed in Fig. 11) and the higher frequency component would represent the fourth harmonic. If the observed frequency components are indeed the even harmonic modes of the ®eld line resonance, this has implications for the potential optical emission modulation mechanism, as will be discussed later. Unfortunately, the present measurements do not allow the exact harmonic structure of the observed ULF wave to be determined. Davidson (1990) stated that ULF wave activity associated with the pulsating aurora is driven by the modi®cation of the local ionosphere in response to the modulated precipitation. This is not consistent with the present observations, in which the aurora are modulated at harmonics of the local FLR fundamental frequency, i.e. appear to be driven by the wave. It is conjectured that a narrow-band source with a compressional component, such as a cavity mode, is responsible for the excitation of the FLR pulsation. This is suggested as at mid-latitudes, near NUR, HAN and OUJ, where the standing wave fundamental is approximately 7 mHz (Orr and Hanson, 1981;Orr, 1984), close to the higher harmonic component of the high-latitude wave, a secondary maximum in the amplitude of this frequency component is observed. At these latitudes also, the latitudinal phase variation and the dominance of the modulation in the X component are characteristic of FLR structures. The cavity mode, then, couples to shear AlfveÂ n waves (FLRs) at the latitudes where the cavity resonance frequency matches the standing wave fundamental (at mid-latitudes) and a harmonic of the standing wave fundamental (at high-latitudes). This situation is similar to the mechanism proposed by Xu et al. (1993), in which a cavity mode couples both to a mid-latitude FLR and an FLR at the cavity boundary at highlatitudes. The cavity resonance itself could be driven by several sources, including Kelvin-Helmholtz waves at the magnetopause or impulsive recon®gurations of the dayside magnetosphere, for instance through the occurrence of magnetopause¯ux transfer events (FTEs). That some of the optical wave packets, especially packets I, III, and IV (see Fig. 6), have the appearance of damped harmonic oscillations in re- sponse to impulsive forcing lends credence to the latter mechanism. In further support for this mechanism, the enhancement in the optical emission intensity wave amplitude at 1211 UT during wave packet IV (Figs. 6 and 8), corresponds to an increase in the un®ltered X component of the magnetic ®eld at both high-and mid-latitudes (Fig. 8), possibly indicating a slight recon-®guration of the magnetosphere. Following this, a 7 to 8 min period¯uctuation is observed in some highlatitude magnetograms (see HOR Y component in Fig. 10); this period is close to the typical repetition period of FTEs. It has been shown previously that the occurrence of FTEs is often accompanied by Pc5 ULF pulsations on closed ®eld lines, Walker et al. (1986Walker et al. ( ), éieroset et al. (1996Walker et al. ( , 1997, and Lee et al. (1988) proposed a mechanism by which these could be excited. The mechanism through which the optical emission intensity is modulated by the ULF wave is unclear. Several possibilities exist, including pitch angle scattering of trapped particles (e.g. Coroniti and Kennel, 1970), or ®eld-aligned acceleration of electrons by electron inertia waves or kinetic AlfveÂ n waves, produced through mode-conversion from the shear AlfveÂ n standing waves (Hasegawa, 1976;Goertz, 1984). The mechanism of Coroniti and Kennel (1970) relies on the ULF waves being even harmonic modes, which is indeed a possibility (see earlier). Interestingly, the optical emission appears to be modulated at frequencies twice those present in the magnetic variations. Several workers have advocated ULF waves as the driving mechanism for auroral arc formation (e.g. Samson et al., 1991;Xu et al., 1993;Moen et al., 1994). In the present case, the auroral arc in question is the optical manifestation of the region 1 current system, and hence its generating mechanism is of great importance with respect to the excitation of magnetospheric/ionospheric convection. The radar scan employed during the present study did not have sucient time resolution to resolve the ionospheric plasma drift response to the ULF wave activity. However, Milan et al. (1999) reported on similar observations, from 17 December, 1995, the previous day to the present study, in the post-noon sector with a radar scan that allowed 14 s time resolution on beam 9. Enhancements of the convection return¯ow plasma drift velocity were co-located with 2 min quasi-periodic enhancements of the auroral emission intensity (see their Fig. 6), indicating a close relationship between this wave activity and convection processes. On 5 days of the 8-day experimental campaign from which the measurements presented in this study were taken, under a variety of IMF and solar wind conditions, the post-noon region 1 optical arc appeared modulated in a quasi-periodic manner, con-®rming the ubiquity and importance of this phenomenon. The observation of short-period quasi-periodic modulations of the convection velocity will perhaps in future be of use in determining the locations of important magnetospheric boundaries, such as the open/closed ®eld line boundary.
Two ®nal points must be addressed regarding the present observations. The ®rst concerns the altitude of emission of the red and green line optical intensity. The assumed emission altitude of the red line, 300 km, is consistent with previous studies and achieves excellent agreement between the MSP and radar observations (Fig. 2). To provide a ®t between the red and green line observations, and again an excellent correspondence can be achieved (see Fig. 6), a green line emission altitude of 200 km must be employed, somewhat higher that has previously been assumed (see for instance  in which the emission altitude is assumed to be below 130 km). However, this somewhat elevated emission altitude appears consistent in producing correspondence between red and green line dayside optical features throughout the eight day campaign from which the present observations were taken. This indicates that care must be exercised in studies of this nature, certainly in studies in which the relative locations of the red and green line emission are attributed geophysical signi®cance, as is the case in , as the assumed emission altitude is of critical importance in determining the mapped location of the observations. In the present study, an assumed emission altitude of 120 km would place the green line observations poleward of the red line features, by an amount dependent on the true location of the aurora relative to Ny A Ê lesund, but by over 1°of latitude in the case of wave packet V. That the red and green optical emissions are truly co-located can be surmised from periods when the aurora are overhead the MSP (78.9°N), for instance during wave packet I (see Fig. 6), and the mapping is independent of the assumed emission altitudes.
The second point is with respect to confusion that could arise in studies of magnetopause¯ux transfer events. The optical signature of an FTE at the ionospheric footprint of the cusp region is a poleward-moving auroral form (see Farrugia et al. (1998), andFasel (1995) and references therein), very similar in nature to the MSP observations of the present study. However, the occurrence of¯ux transfer events is not expected to be periodic in nature, and is generally accepted to have a mean repetition rate of approximately 8 min (though Lockwood and Wild (1993) reported that the distribution of intervals between magnetopause signatures of FTEs was highly skewed, with a mode near 3 min). It is possible that the super®cial similarity between the optical signatures of the two phenomena and the proximity of the cusp and region 1 current systems has lead to some confusion, and statistical studies of FTE occurrence based on optical observations could be contaminated by short-period ULF wave events.

Conclusions
The emission intensity of the post-noon dayside aurora associated with the region 1 FAC system is commonly modulated in a quasi-periodic manner with periods near 2 min, on ®eld lines just equatorward of the open/closed ®eld line boundary. This modulation appears related to ULF magnetohydrodynamic wave activity; several harmonics of the fundamental ®eld line resonance eigenfrequency can be detected by ground magnetometer stations located underneath the aurora. The modulation of the auroral emission intensity has frequency components equal to twice that of the ULF wave harmonics. This wave activity also causes modulation of the convection return¯ow plasma drift velocity. The importance of this wave activity in the excitation of magnetospheric/ionospheric convection must not be over-looked.