polar cap boundary above Mawson, Antarctica

polar cap boundary above Mawson, Antarctica

Journal of Atmospheric and Terrestrial Physics,Vol. 58, No. 16, pp. 1973-1988, 1996 Copyright © 1996 Elsevier Science Ltd Printed in Great Britain. Al...

1MB Sizes 3 Downloads 24 Views

Journal of Atmospheric and Terrestrial Physics,Vol. 58, No. 16, pp. 1973-1988, 1996 Copyright © 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved

~ ) Pergamon

PII: S0021-9169(96)00007--4

0021-9169/96 $15.00+0.00

Fabry-Perot spectrometer observations of the auroral oval/polar cap boundary above Mawson, Antarctica J. L. Innis, ] P. A. GreeP 'z and P. L. Dyson 3 1Australian Antarctic Division, Channel Highway, Kingston, Tasmania 7050, Australia 2Institute of Antarctic and Southern Ocean Studies, University of Tasmania, Hobart, Tasmania 7001, Australia 3School of Physics, La Trobe University, Bundoora Campus, Victoria 3083, Australia (Received in final form 1 February 1996; accepted 1 February 1996) Abstract--Zenith observations of the oxygen 2630 nm auroral/airglow emission (produced at an altitude of ~ 220 to ~ 250 km) were obtained with the Mawson Fabry-Perot Spectrometer (FPS) during three 'zenith direction only' observing campaigns in 1993. The data show many instances of strong (50 to 100 m s-1) upwellings in the vertical wind, when the auroral oval is located equatorward of the zenith. Our data appear consistent with the existence of a region of upwelling up to ~ 4 ° poleward of the poleward boundary of the visible auroral oval, rather than short duration, explosive heating events. The upwellings are probably the vertical component of wind shear produced by reversal of the zonal thermospheric winds, which occurs near the poleward boundary of the visible auroral oval. Zenith temperature was also seen to increase when the oval was equatorward of Mawson, showing rises of up to 300 K or more. However, this increase is at times unrelaled to the upwellings, and seems to be caused by the expansion of the warm polar cap over the observing site. On a number of nights the boundary between the polar cap and the auroral oval was observed to pass over our site several times, occasionally showing a quasi-periodic expansion and contraction. We speculate that this quasi-periodic movement may be related to periodic auroral activity that is known to generate large-scale gravity waves. Copyright © 1996 Elsevier Science Ltd INTRODUCTION The original aim of this study was to obtain F a b r y Perot Spectrometer (FPS) observations of the oxygen 2630 nm emission line profile at a sufficiently high time resolution to allow a search to be made for gravity wave signatures in the data. The presence of quasiperiodic signals in FPS vertical wind data, presumed to be due to gravity waves, has been reported by various authors (e.g. Hernandez (1982); Wardill and Jacka (!986), Siple:r et al. (1995)). Atmospheric gravity waves at high latitudes tend to be of two types, known as large and medium scale waves, following Georges (1968). Large scale waves are identified with travelling ionospheric disturbances (TIDs) seen to be propagating towards the equator from polar regions (e.g. see review by Hunsucker (1982)). Recent work by Williams et al. (1988), Williams et al. (1993) has indeed detected the generation of such waves in the auroral electrojet (due to periodic auroral activity), and their equatorward propagation. As Mawson station (geographical position 67°.6 S, 62°.9 E; invariant latitude = 70°.5 S; L = 9.0) passes under the auroral oval during nights of normal geomagnetic activity, it seemed possible that an intensive

observing campaign may detect gravity waves near their source. Campaigns of zenith-only observations using ground-based FPSs have been performed by a number of groups, including the work already mentioned. Rees et al. (1984) report oxygen 2630 nm observations of vertical (and horizontal) winds from several years from two (northern) high-latitude sites. A large diurnal variation (i.e. from 30 m s-~ downwards to 50 m s- 1upwards) was seen. While no periodic signals were detected in their data, they do report several events suggestive of a burst of 'explosive' heating, indicated by a large amplitude, short duration upwelling in the wind (no temperature data were presented), which they interpret as one mechanism for the generation of gravity waves. Crickmore et al. (1991) present FPS wind data (also of oxygen 2630 nm emission) from Halley Bay, Antarctica, finding strong downward winds (up to 50 m s 1) when the station is under the equatorward edge of the auroral oval. They do see at times a suggestion of oscillatory behaviour in the wind data which might be due to gravity waves, but they note that the periods and amplitudes indicated are close to the resolution of the data, and do not claim a detection.

1973

1974

J. L. Innis et al.

Price and Jacka (1991) report oxygen 2558 nm vertical wind and temperature data from Mawson, and found several instances of short duration upward winds (of around 30 m s-l), which they suggested was a response of the neutral atmosphere to auroral activity. They were able to distinguish two classes of events, class I showing a positive correlation between wind and intensity (upward wind associated with an increase in intensity) and/or a positive correlation between wind and inverse temperature, and class II, where there was a negative correlation between wind and intensity and/or a negative correlation between wind and inverse temperature (i.e. a positive correlation between wind and temperature). D-region radar data were also available. Price and Jacka (1991) concluded that the class I events were from heights around 110 km, where particle heating dominates, while the class II events were occuring at greater altitudes. Conde and Dyson (1995) present seven months of vertical wind data (oxygen 2630 nm emission) from Mawson from the austral winter in 1992--the year proceeding our study. A diurnal variation of low amplitude ( ~ 5 to 10 m s-1) was seen, the exact amplitude showing a dependence on geomagnetic activity. The low time resolution of these data ( ~ 40 minutes) precluded a detailed search for gravity waves or for upwelling events such as those reported by Price and Jacka (1991). Aruliah and Rees (1995) present a number of years of oxygen 2630 nm data from Kiruna, Sweden. They detected a small diurnal variation during intervals of moderate geomagnetic activity (Kp between 2 and 5), as well as a solar cycle dependence in the measured winds. They present arguments to suggest that the usual assumption of a time-averaged zero-mean vertical wind at any given site may be in error by 5 to 10 m s l, depending on geomagnetic activity. Smith and Hemandez (1995) show oxygen 2630 nm and 2558 nm wind and temperature data from the South Pole station for a 2-week interval in the austral winter of 1991. A large diurnal variation is present in the oxygen 2630 nm data, of average amplitude of 40 m s-~, depending on geomagnetic activity. Extremes in the wind observations of around 150 m s -1 were seen at times on the most active nights. Temperature variations of around 1000 K were seen in the upper thermosphere, well correlated with changes in the Ap index. Smith and Hernandez (1995) concluded that it was likely that the observed vertical winds were a consequence of divergence in the horizontal wind field. Of particular relevance to the data to be presented here is the work of Price et al. (1995), who obtained simultaneous oxygen 2630 nm and 2558 nm obser-

vations at Poker Flat, Alaska, on 21 nights over two observing seasons. No clear periodic behaviour was seen, but on two occasions, each of around 15 to 30 minutes duration, a strong upwelling was observed simultaneously in both the upper and lower thermosphere. The peak upwelling velocities were around 140 m s-~ and 40 m s-l, with associated temperature increases of around 400 and 200 K, in the upper and lower thermosphere respectively. For both occasions, the regions of upwelling were located poleward of the auroral oval. Price et al. (1995) indicated that the observed increase of the vertical winds at these times was more likely to be due to the movement of an existing region of upwelling into the (zenith) field of view of their instrument, rather than the signature of the onset of a sudden heating event. Price et al. (1995) were able to estimate the size of this postulated upwelling region, assuming that it moved with the aurora, finding that the centre of the upwelling was some 80 to 230 km from the poleward edge of the auroral oval. Our data, from 54 nights from 3 separate observing campaigns in 1993, showed the presence of a number of upwelling events (up to ~ 100 m s-1), and occasionally a strong downwelling, on the poleward side of the auroral oval. The upwellings are very similar to those reported by Price et al. (1995) from Alaska, and also to the 'explosive' upwellings reported by Rees et al. 0984). Our FPS data, in conjunction with Mawson Auroral Imager video tapes from these nights, show that these events occur when the auroral oval is equatorward of our observing site. i.e. the events are seen poleward of the poleward boundary of the visible auroral oval, consistent with the findings of Price et al. 0995). Additionally, poleward of the oval we have often observed a hot region, with temperatures up to 1700 K and low 2630 nm intensifies, indicating we are looking into the polar cap (e.g. Hays et al., 1984) while in the oval itself temperatures were lower by 200 to 300 K. At other times we observe both upward and downward winds, generally of smaller magnitude and shorter duration than for the events mentioned above, while the auroral oval is overhead at Mawson. These winds may be related to the presence of auroral arcs passing through the field of view of the FPS. Our data set was subjected to a full spectral analysis of the wind, temperature, and intensity time series, as a search for the presence of gravity wave signatures. While evidence for periodic (or quasi-periodic) signals was found for a small number of nights, these signals are not due to gravity waves, but are from repeated movement of the polar cap/auroral oval boundary back and forth through the zenith (and hence through

Fabry-Perot spectrometer observations the field of view of the Mawson FPS in these campaigns) in a periodic (or quasi-periodic) manner. At this stage, we do not know if the detection of this periodic motion is indicative of some real physical process occuring in the magneto-tail, or whether it is only a series of coincidences. INSTRUMENTATIONAND OBSERVATIONS The Mawson FPS is a dual-etalon, separationscanned instrument, described in detail by Jacka (1984). For the night time observations reported here, the low-resolution etalon was removed from the optical path, and only the high-resolution etalon (diameter 150 mm) was used. The resolving power is approximately 2 x 105. During 1993 the FPS was used to observe the oxygen 2630 nm auroral and airglow line. In order to obtain observations with sufficiently high time resolution to search for atmospheric gravity waves, the FPS was operated in a 'zenith only' mode during three campaign intervals: February 24-April 15; July l-July 31; and September 15-October 15. One five minute observation of the sky vertically overhead was obtained approximately every eight minutes. Three minutes were required for an observation of a mercury lamp to monitor instrument drift, logging the data to a PC, and switching the light path between lamp and sky. An analysis of these data will be presented in this paper. In the previous observing season of 1992, and outside of the zenith campaigns in 1993, the Mawson FPS was used to obtain cardinal point observations. This is a cycle of measurements (taking approximately 40 minutes) in the directions of north, south, east, west (all at 60° zenith angle), and the zenith. Hence, while zenith observations were still obtained, the time interval between successive observations was around 40 minutes, which precluded a study of gravity waves (Conde and Dyson, 1995). FPS data were supplemented with data collected by other instruments operating at Mawson, including a fluxgate magnetometer and a vertical incidence ionosonde. (These instruments are of standard design (McLoughlin et al. (1989), Burns and McLoughlin (1989)).) Additionally, an Auroral Imager (AI) video system (Keo Consultants) was also in operation, which obtained one all-sky image at 2558 nm (from ~ 120 to 150 km altitude) every 6 seconds. RESULTS

FPS zenith observations were obtained on a total of 54 nights during the three campaign intervals,

1975

resulting in over 3300 individual spectra of the oxygen 2630 nm emission line profile. Early and late in the season the observing night was only a few hours long, while in July almost 16 hours of data could be obtained each night. Raw spectra were processed to yield vertical winds (velocities) and temperatures. Effects of instrumental broadening were removed by deconvolving an instrument function obtained from observations of a frequency-stabilised laser, assumed to have zero Doppler width. The peaks and widths of Gaussians fitted to the processed profiles were used to derive the winds and temperatures. We cannot derive absolute values of the wind velocity, as no absolute wavelength reference for the 2630 nm line is available. We have adopted the usual approach of assuming that the mean vertical wind observed on any given night is zero. This is clearly an approximation to reality--even if the mean wind over any given 24 hour interval is zero, observations taken over part of a night are unlikely to yield a zero mean. For the spectral analysis (described below) where we were looking for periodicities within one night the value of the mean wind is of little importance. In general, however, what we conclude from our wind data is how the wind changes during the night, rather than absolute vertical velocities. For the reasons presented below we believe that we have detected regions of upwelling and downwelling motion. (See Crickmore et al. (1991) and Aruliah and Rees (1995), for further discussion regarding the mean value of the vertical wind.) Relative intensities were also determined, by a summation of the counts obtained in each profile, corrected for the photomultiplier dark count. For convenience we will refer to these data as 'intensity measurements' rather than 'relative intensity measurements'. Inspecting our data set, we found many instances of large excursions in the vertical wind, often associated with temperature enhancements (of several hundred degrees): at least one obvious 'event' in the wind and/or temperature was seen on 23 of our 54 nights of data. (Another 13 nights showed evidence of a smaller amplitude variation.) While these events had some common features, the detailed behaviour was seen to vary markedly from event to event. The events could be up to around one hour in duration, and included both upwellings and downwellings in the wind (up to around 100 m s l in extreme cases). On a few nights they were seen to repeat in a quasi-periodic manner. We have selected 4 nights from our data set to illustrate the range of events we have seen. (Typical quiet or mean conditions will not be presented, as they

1976

J. L. Innis et al. FPS data, Mawson 1993 D260

120

~

qo

.~

0

(a)

-80 -,

,

,

,

,

,

,

16

ooK

~

3o

~

20

~

.

.

.

16

t--,

10

,

,

.

i

18

I"

.

.

,

,

.

.

20

.

.

.

.

.

l

.

,

,

,

22

.

.

.

.

.

.

.

.

18

20

22

2't

18

20

22

2q



16

,

2zl

1

Universal Time (hours) Fig. 1. Time series of wind (a), temperature (b), and relative intensity (c), from Mawson FPS zenith data for DOY 260 from the oxygen 2630 nm emission.

have been discussed at length by Conde and Dyson, 1995.) The wind, temperature and intensity time series data for the selected 4 nights are shown in Figs 1-4, where the data are from the nights of Day of Year (DOY) 260, 264, 094, and 099 respectively. The error bars shown are the statistical errors from the Gaussian fit, as described above. Many instances of significant upwellings and downwellings can be seen in the wind data from the selected days. A common theme is that such events are normally associated with very low intensity measurements. This is not a consequence of poorer signal-to-noise data at these times, as the variations seen are several times the statistical uncertainty of the observations, and show consistency between adjacent measurements (e.g. the large upwellings in the winds on DOY 099 and DOY 264, shown in Fig. 4(a) and Fig. 2(a)). We will briefly outline the main features of each of the 4 selected nights of data, as plotted in Figs 1-4. We have used the Auroral Imager video tapes from these nights to follow the movement of the auroral

oval over Mawson. Images from these tapes are shown in some of the following figures. In these images, the centre of the circle is the zenith, geographic north is at the top, while geographic east is to the left.

Description o f selected nights DOY 260 (1993 September 17, Ap = 4, Fig. 1). The Ap index of 4 indicates it was a night of low activity. However, a significant event in the wind and temperature data was seen at around 22:00 Z. This event is shown on a plot with an expanded time axis in Fig. 5(a), while video images for this night are shown in Fig. 5(b). In the early part of the night, Mawson was clearly on the equatorward side of the oval (Fig. 5(b), frames 1 to 4). (The band of light on the western horizon in frame 1 is evening twilight.) The wind was generally a downwelling from the start of the data to around 1 7 : 3 0 Z. Inspection of the video revealed the presence of low intensity pulsating aurora at this time (not visible on the 'snapshot' images we present here), indicating that this emission was occuring on closed field

Fabry-Perot spectrometer observations

1977

FPS data, Mawson 1993 D264

"G"

160 ~ 120 . . . . . . 80 q 'O 0 -qO 16 18 20

(a)

200

-

8

0

,

.

,

,

,

,

i

,

,

,

,

,

,

16oo........

22 ,

,

,

.

1

24 ,

,

(b)

l qO0 k,

M

13., [-,

1200

]

1000

BOO 600

i6

.

i

,

,

,

18 ,

,

,

,

20 I

,

,

,

22 ,

,

,

,

24 ,

50

(o)

qO cI

,

30 20 lO 0

,

,

16

.

,

.

J

,

18

,

.

20 i

.

-~-~:=

.

22 ,

,

,

,

,

.

.-~

Zq

Universal Time (hours) Fig. 2. As for Fig. 1, but for DOY 264.

lines. The general downwelling on the equatorward side of the oval is in agreement with the work of Crickmore et al. (1991). There may also be structure in the wind and temperature data during this interval (such as the drop in temperature around 17:20 Z - frame 3). Between 18:00 and 19:00 Z the oval moved into the overhead position (frame 5), as indicated by the rise in intensity over this time. For the next one and a half hours the wind showed very little variation, despite the fact that on two occasions the oval appeared equatorward of the zenith: at around 20:15 Z (frame 6) and at around :21:20 Z (frame 7). The temperature appeared briefly low, by ~ 150 K, in the first case, and then higher, by about the same amount, for the second. The oval began to move poleward from low on the north-western horizon at around 21:20 Z (frame 7). By 21:40 Z it was overhead, as seen by the intensity record, and in fi'ame 8. This poleward movement reversed soon afterward, and by 22:00 Z the oval was once more on the north-western horizon (frame 9). This time however, as the intensity decreased, the wind

and the temperature both increased (Fig. 5(a)), the temperature by nearly 300 K to around 1100 K, and the wind becoming an upwelling of around 40 to 50 m s -1.

Within 10 minutes of these changes, both the wind and temperature decreased, the temperature to just below 800 K, and the wind to a downwelling of around 60 m s-l. The oval appeared not to have moved from the horizon at this time (frame 10), although the effect of foreshortening at such a low elevation may make this statement hard to verify. However, within a few minutes it was apparent that the oval was moving poleward once more (frame 11). The wind and temperature again increased, mirroring their behaviour of some 30 minutes previously. The poleward boundary of the oval was still equatorward of Mawson at this time. Soon after 22:30 Z, the wind returned to near zero, the temperature returned to ~800 K, as the intensity increased when the oval reached the zenith (frame 12). The 'mirroring' of this event in wind and temperature as the oval moved back and forth over Mawson is suggestive of the existence of a relatively stable,

1978

J.L. Innisetal.

FPS data, Mawson 1993 D094 8O

0 ~:

-40

i

16

' 18

20

22

2~

26

(b)

1800 1600 1400 1200

lOOO 16

18

20

22

25

26

16

18

20

22

2~

26

180 150 "~ "-'

120 90 60

-=

30 0

Universal Time (hours) Fig. 3. As for Fig. l, but ~ r DOY 094.

hot region with some type of wind shear, but of limited extent, which shared the motion of the poleward edge of the auroral oval. DOY 264 (1993 April 21, Ap = 14, Fig. 2). This night shows a very large upwelling in the wind, with an associated increase in temperature. We will briefly describe the FPS data, making reference to the Auroral Imager video data (not shown). From the start of the night to approximately 20:30 Z the auroral oval was generally poleward or overhead as seen from Mawson, with the exceptions at around 17:30 (briefly), 18:30 (for about 30 minutes) and 20:00 Z (for about 15 minutes). There is evidence for upwellings in the wind and temperature increases, at or near these times. At about 20:30 Z the oval moved equatorward of Mawson, remaining equatorward until around 21:30 Z. During this hour, the wind was strongly upwelling, up to 130 m s -1, while the temperature reached 1400 K, or some 600 K above the temperature measured when in the oval. With the return of the oval to the overhead position at around 21:30 Z the temperature dropped immediately, and the wind became a significant downwelling, and

generally remained this way for the rest of the night. DOY 094 (1993 April 04, Ap = 58, Fig. 3). This day had the equal highest Ap index in our data set, and hence we may expect to see some indication of this increased activity in our observations. Selected frames from the Auroral Imager video appear in Fig. 6. (The Moon is also visible in frames 1 to 8 on this night.) We note too that both the temperature variation and the actual temperatures measured here were among the highest we observed, while the peak intensifies were the highest in our data set. On this night the poleward edge of the auroral oval made a series of passes over Mawson, moving back and forth (equatorward then poleward and back again) several times, giving rise to the observed features in the wind, temperature and intensity data, as discussed in the following paragraphs. The auroral oval is seen to be overhead in frames 1, 3, 5, 7, and 9 and equatorward of Mawson in frames 2, 4, 6, and 8 in Fig. 6. The upwellings in the wind data (Fig. 3(a)) and the intervals of elevated temperature generally coincide with the times when the oval is equatorward of Mawson (i.e. at 18, 20, 22, and 23:30 Z).

Fabry-Perot spectrometer observations

1979

FPS data, Mawson 1993 D099

80

&., 40 0 -40 .

lZt ,

.

.

.

16 .

.

.

,

18 .

.

.

,

20 .

.

.

,

22 .

.

.

,

2 z] .

.

.

,

26 .

.

.

(b)

1400 1200 1000 [.,,.,

14

16

18

20

22

2':1

80

26

(o)

60 ~0 @ 3

20 0

ILl

16

18

20

22

2q

26

Universal Time (hours) Fig. 4. As for Fig. l, but for DOY 099.

Note that the increases in wind and temperature do not necessarily oo~ur simultaneously with movement of the oval equatorward of the Mawson zenith. For example, the oval was clearly equatorward of the zenith soon after 21:30 Z, as shown by the decrease in intensity (Fig. 3(c)). By 22:00 Z (frame 6) the oval was on the north-western horizon. The wind, and to an extent the temperature, do not increase until ~ 22:15 Z. This behaviour is suggestive either of a spatial separation between the visible poleward boundary of the auroral oval and a hot upwelling region located poleward of this boundary (and which appears to move with the oval), or that this hot region was only produced around 22:15 Z, presumably in response to an increase in auroral activity. Given that we saw a similar hot upwelling region poleward of the oval earlier in the night, around 20:00 Z, we are inclined to the former interpretation. (We qualify this statement by noting that the video images were obtained at 2558 nm, while the FPS data were obtained at 2630 nm, which may be of significance.) Conversely, the upwelling at ~23:30 Z occur,,; almost as soon as the oval has cleared the zenith (frame 8), suggesting the hot region

is now located closer to the poleward edge of the oval. DOY 099 (1993 April 09, Ap = 28, Fig. 4). This day, like DOY 094, also saw the poleward edge of the auroral oval making a number of passes back and forth over Mawson, but unlike DOY 094 the temperature generally showed little variation. Selected images from the Auroral Imager video are presented in Fig. 7. (The Moon's image is also visible.) The oval was to the equatorward of Mawson several times during this night, at around 16:30, 17:30, 19:30, and 22:00 Z. For the three later times, clear upwelling events are seen in the wind data, seemingly increasing in amplitude during the night. The temperature data does follow the wind data to a small extent, but generally with an amplitude of less than 100 K. This indicates that the hot polar cap and the region of upwelling are not intrinsically linked by a single physical process, but arise from distinct causes. The oval is overhead during the last event seen in the wind (around 24:30 Z). No single obvious auroral feature could be identified with this event (frame 15), although several arcs were seen to pass through the zenith during this interval. Perhaps images taken at

1980

J. L, Innis et al. FPS data, Mawson 1993 D260 - detail

Q-.

E ,el

160 120 80 'qO 0 -qO -80 21

21.5

22

22.5

23

21

21'.5

22

22'.5

23

21

21'.5

22

22.5

23

1200 f,.,

I000 800

@..I

[...,

6OO

,i

qO 3O 20

r'!. 0

Universal 'rime (hours) Fig. 5. (a) Time-expanded plot of wind, temperature and intensity data for DOY 260.

2630 nm, rather than at 2558 nm, are required in this case. We note that the wave-like nature of the wind data (Fig. 4(a)) initially suggested to us that we were seeing the signature of a long period gravity wave. However, as the main features of the wind data are correlated, empirically at least, with the movement of the poleward edge of the auroral oval back and forth over Mawson, we do not believe that we have observed a gravity wave. Our search for evidence for gravity waves in our data is described in the next section. Spectral analysis--the search for gravity waves

We performed a spectral analysis for each of our 54 nights of zenith observations. Our approach, following that of Hoegy et al. (1979), assumed that the presence of a gravity wave in the observing volume will affect all of the wind, temperature and intensity measurements, although not necessarily with zero phase lag between each quantity. Hence, for each day, the frequency spectra for the wind, temperature, and intensity time series were calculated. As well, the cross-

spectra between each of the wind and temperature, the wind and intensity, and the temperature and intensity time series pairs were found. We calculated the frequency spectra and cross-spectra using subroutines from the Numerical Algorithms Group (NAG) library, and also checked these results with a Maximum Entropy Method (MEM) routine based on that presented by Dyson and Hoegy (1978). The results from these two approaches were almost indistinguishable for the ten nights chosen for the comparison. A standard statistical test was applied to investigate the significance of apparent peaks in the cross-spectra. In calculating the cross spectra, we used the same amount of smoothing in the Fourier domain for all of the calculations. This in turn determined the number of degrees of freedom (equal to 7) in the final spectra. The squared coherency estimates then have an F(2,7) distribution. The N A G routine returns the 95% confidence value of the coherency (for which our calculations always equals 0.69). The null hypothesis that the two time series are not related at any given fre-

Fig. 5. (b) Mawson Auroral Imager data (2558 nm) for selected times from DOY 260.

O

7

O t~r

O

O

1982

J.L. Innis et al.

Fig. 6. Mawson Auroral Imager data for selected times for DOY 094.

Fig. 7. As for Fig. 6, but for DOY 099.

Fabry-Perot spectrometer observations quency can be rejected at this level of confidence if the squared coherency at this frequency is greater than this value of 0.69. From the total data set, only seven nights were found in which a statistically significant signal (greater than 95 % confidence level) was seen in all of the wind, temperature, and intensity spectra, as well as in all of the cross-spectra (i.e wind xtemperature, wind× intensity, and temperature x intensity). If we relaxed our criteria, and only required detections in the individual wind, temperature and intensity spectra, but not in all of the cross-spectra as well, we found evidence for periodicities on a further two nights. We tested the validity of these criteria by running a series of trials on simulated wind and temperature data sets, of representative length, and of various signal-to-noise ratios, and then determining how well we were able to recover the input parameters. Our raw data indicated that the average observational errors in an individual deten'nination of wind and temperature were typically 15 m s -~ and 30 K respectively. We therefore construc~ted data sets made up of sinusoidal signals of amplitudes between 0.5 and 1.0 times these observational errors, with added Gaussian (white) noise of standard deviation equal to the the respective wind and temperal;ure error. The trials indicated that our detection criteria were reasonable, and that, for measurements with signal to noise ratios (as defined above) approaching unity, secure detections could be made. The appearance of the raw time series for the nine nights which satisfied our criteria were often suggestive of a wave (particularly in the wind data, e.g Fig. 4(a)). However, our investigation of these data showed that, in each case, the observed variations were largely due to repeated crossings of the poleward boundary of the auroral oval over Mawson. As noted, the movement of the oval resulted in the FPS alternating between observations of the cool, bright, sometimes downwelling auroral oval, and the hot, dark, often upwelling region located poleward of the poleward boundary of the oval. We conclude that the spectral analysis of the FPS data has not identified any features that we could interpret as gravity waves. We also performed a spectral analysis of fluxgate magnetometer data obtained at Mawson on the nights when significant periodicities were seen in the FPS data. For the two nights with the highest Ap indices, DOY 094 and 099, the same period (around 2.5 h) was also seen in the magnetometer data (horizontal component). The vertical component of the magnetic field exhibited evidence of a correlation with the wind data, where an interval of upward wind was likely to be associated with an interval of increased upward

1983

field strength. As the vertical component of the field is an indication as to whether the auroral electrojet currents are poleward or equatorward of the site, and as the increased upward field indicates the electrojet was equatorward of Mawson, these magnetometer data are consistent with our identification of an upwelling region poleward of the auroral oval. A point that may be of interest from our study is that the periods we have found from the FPS data, apparently of statistical significance, are indicative of a quasi-periodic expansion and contraction of the auroral oval. The periods found are generally around 1 hour, which is near the observed peak of the spectrum of long period gravity waves (e.g. De Deuge et al., 1994). As long period gravity waves can be generated in the auroral oval by periodic auroral activity (Williams et al., 1988, 1993; Millward, 1994), it is tempting to suggest that the quasi-periodic motion of the oval we have seen may be related to the periodic auroral activity that generates the waves. In order for the movement of the oval to appear periodic from one site, the site would need to be near the equilibrium position of the oscillation. Our analysis, which has searched for periodicities, may therefore have selected nights when Mawson was near the equilibrium position, and rejected other nights, with possible periodic oval movement, when the equilibrium position was elsewhere. The point is worth further investigation.

DISCUSSION The events we see in our zenith wind and temperature data are very similar to the two upwelling events reported by Price et al. (1995), both in general morphology, and in the amplitude of the wind and temperature variations. The upwellings we detected are also very similar to several events observed in vertical winds reported by Rees et aL (1984)~the upwelling precedes a large increase in intensity as an intense auroral arc (if not the auroral oval itself) moves polewards from the horizon towards the zenith of the observing site. Rees et aL (1984) suggest that the short lived upwellings are the (short-duration) result of an episode of explosive heating during the expansion phase of a geomagnetic disturbance. Price et al. (1995) believe that a better explanation for their data is the movement of an existing region of upwelling into the overhead position, rather than being from a sudden burst of heating. Our data, where we have seen upwellings poleward of the auroral oval regardless of which way it is moving at the time (i.e. both expanding and contracting), appear consistent with

1984

J. L. Innis et al.

the suggestion by Price et al. (1995) (e.g. the multiple upwellings on DOY 094 and DOY 099; the mirror pattern of the event on DOY 260, and so on). The observation of a region of a strong wind field near the poleward boundary of the auroral oval has been well documented by a number of authors, from data obtained with the Dynamics Explorer 2 (DE 2) satellite. For example, Spencer et al. (1982) saw a region of strong upwelling vertical winds at auroral latitudes, correlated with the latitude of a temperature increase visible over most of the pole. These data would appear to be the earliest reported observations of a 'hot upwelling event' of the type described here and elsewhere. Hays et al. (1984) obtained data on the neutral winds at high latitudes, concluding that the winds were primarily driven by momentum transfer between ions and neutrals. A similar conclusion was reached by Roble et al. (1984). Killeen et al. (1988) obtained simultaneous observations of aurora and neutral winds. They found the neutral wind field was highly spatially correlated with the auroral oval, and that both the equatorward and poleward boundaries of the oval were regions where the zonal wind reversed direction. These gradients in the wind field followed the expansion and contraction of the oval. We suggest that our ground-based FPS data are observations of the vertical component of this wind pattern. The detection of upwards winds of over 100 m s-1 suggest a non-negligibleamount of momentum can be present in the vertical direction. Temperatures measured when the auroral oval is equatorward of Mawson can be significantly higher (sometimes by several hundred degrees) than when the oval is overhead. The observations are consistent with data obtained from the DE 2 satellite reported by Hays et al. (1984), who found that the polar cap was the warmest part of the polar thermosphere, and also noted a significant decrease in temperature (by approximately 300 K) near the evening auroral oval crossing. The hot polar cap is also apparent in the DE-2 data presented by Spencer et al. (1982, their Fig. 6(b), (c), and (d)), where an increase of several hundred degrees was measured, although significant orbit to orbit changes can occur in the temperature structure at high latitudes. McCormac et al. (1988) reported a diurnal temperature variation of the polar cap (of around 100 K), and suggested that the warm cap may be accounted for by the transport of preheated parcels of air to high latitudes. There will be some difference in temperature between the low intensity and high intensity data as a consequence of the higher energy of the incident electrons during intense aurora. (Higher energy leads to a lower emission height and therefore a lower

thermospheric temperature.) We can estimate the height of the high and low intensity emission for DOY 094 from ionosonde data obtained on this night. (The ionosonde was operated on most nights during the year, but generally the data obtained on nights with moderate to strong auroral activity have long duration intervals where absorption results in non-scalable ionograms.) Figure 8 shows for DOY 094 the height of the F2 peak (plotted as squares) scaled from the M(3000)F2 parameter using the formalism of Dudeney (1974), along with the wind data (crosses) obtained on this night (from Fig. 2(a)). The close correspondence between the observed wind and the altitude of the F2 peak indicates the emission during the intervals of upwelling was from around 60 to 70 km higher than for the observations taken in the oval. Reference to the CIRA models for the thermosphere (Rees, 1988) for the approximate geomagnetic conditions on this night suggests that the expected temperature increase for this height change would be at most 100 K, which accounts for about one third of the apparent variation in temperature. The remaining 200 K or more is due to the intrinsically higher temperature of the polar cap, as noted above. We also note that the very good agreement seen near the middle of the data run between the observationally independent data sets (ionosonde and FPS) shown in Fig. 8 gives us confidence that we are measuring real events in the vertical wind. We do not always see upwellings or downwellings in the wind, nor large increases in temperature, every time the auroral oval moves to the equatorward of Mawson. (e.g. DOY 260, at 20:15 and 21:20 Z, Fig. 3 and Fig. 5(b)) The fact that we do not may relate to both conditions in the thermosphere at the boundary between the polar cap and the auroral oval, and possibly also to the limited time resolution of the data. It seems likely that the region of high vertical wind is of limited latitudinal extent only. It is also clear that the oval can move very rapidly, both equatorwards and polewards, so that the region of high vertical wind may pass through our field of view too quickly to be measureable with our 8 minute time resolution. However, we also have instances such as on DOY 264 from ~ 18:30 to 19:00 Z (Fig. 2), where we have 4 or 5 data points taken poleward of the oval. There was very little variation in the vertical wind, implying there was no upwelling region over Mawson at this time. At times there appears to be a strong overlap between the region of strong vertical wind and the low latitude edge of the polar cap, as shown by the data for DOY 264 where we have a large upwelling simultaneously with a large increase in temperature. However, the data for DOY 099 are suggestive of a

Fabry-Perot spectrometer observations

60

1985

'

qO 20

-21

, 900

1000

I100

1200

s"

,'

tJ

0 0

960

', ,

oO i

', 0

.so

i:' ~Do

lObO

[] (b)1 0

', [] I~EI.O"

0

200

1500

[] O

250

l'qO0

0

300

N r..

1300

11O0

12b0

13b0

14b0

1500

UT ( m i n u t e s ) Fig. 8. (a) FPS wind data for DOY 094. These data have been averaged to givea resolution of ~ 15 minutes to allow for comparison with the ionosonde data. (b) Mawson ionosonde data for the height of the F2 peak for this day. There is close agreement at times between these data and those shown in (a).

separation between the region of high vertical wind and the hot polar cap, as on this night we see strong upwellings poleward of the oval, but the temperature data show little variation. We suggest that had the oval moved further towards the equator on this night the temperature may have been observed to have increased to a greater extent, along with a diminution of the strong upwelling. Hays et al. (1984) concluded that the neutral wind field was primarily controlled by ion/neutral friction, while Joule heating and lowenergy particle precipitation produced the heating in the polar thermosphere. Hence a (small) spatial separation between the region of high winds (near the poleward boundm~ of the oval) and the high temperature of the hot cap may not be surprising. It may be relewmt to note that Price et al. (1995) calculated that the centre of the upwelling region they detected was 80 to 230 km from the poleward edge of the auroral oval. If the upwelling region is twice this size, it indicates a latitudinal extent of around 1.5 to 4 °. For our data, we can estimate the size of the upwelling region from the greatest separation observed between the poleward limit of the upwelling and the poleward edge of the visible oval. On DOY 077, at ~21:30 Z (Fig. 6(a) and (b)),. upwelling is seen in the wind data when the oval has moVed equatorward, out of the field

of view of the Auroral Imager. This implies a width of around 4° in latitude, in good agreement with the estimate of Price et al. (1995). (Similar circumstances are seen on DOY 094 at ~22:00 Z, and on DOY 099 at ~21:30 Z, for example.) There is evidence to suggest that the boundary between open and closed field lines is some 5 to 10° poleward of the visible auroral oval poleward boundary (e.g. Elphinstone et al., 1991). The rough agreement between the size of the upwelling region as estimated from our data and their results may indicate that the upwelling is confined between the boundary between the open and closed field lines and the poleward edge of the visible oval. We also have seen a few instances of strong vertical winds (both up and down) when the oval is overhead. An experimental study of the winds and temperatures near a quiet arc was performed by Eastes et al. (1992), again using data from the DE 2 satellite. Measurements through the arc showed a temperature drop of around 100 K, and both upward and downward winds (up to ~ 25 m s-l) in the near vicinity. Our observed winds are somewhat greater, perhaps as much as a factor of two, which could be explained by a more active arc in our case. Eastes et al. (1992) compared their observations with three high-resolution, two-

J. L. lnnis et al.

1986

CONCLUSIONS

A spectral analysis of fifty four nights of zenith wind, temperature and intensity data revealed nine nights when significant periodicities were seen. In each case, we can relate most of the variations in the time series to the movement of the poleward boundary of the auroral oval back and forth over Mawson, suggesting a quasi-periodic oscillation of the position of the oval. The periods we detected were generally around 1 h, which is near the peak of the observed spectrum of long period gravity waves, known to be generated by periodic auroral activity. We note the possibilty that the quasi-periodic movement of the oval we have seen may be related to the same periodic auroral activity that can generate gravity waves.

We have detected a number of instances of significant upwellings (50 to 100 m s -1) in FPS vertical wind data obtained at 2630 nm on or poleward of the poleward edge of the auroral oval, often but not always associated with a temperature increase (up to 300 K). We suggest these upwellings are the vertical component of the wind shear, seen in the reversal of the zonal wind near the poleward edge of the oval reported by Killeen et aL (1988). The temperature increases are largely a result of viewing the warmer polar cap (Hays et al., 1984) when the auroral oval has expanded to be equatorward of Mawson. The upwelling region was not always present, but was seen to occur in the area immediately poleward to around 4 ° poleward of the poleward boundary of the visible auroral oval. The size of this region is thus comparable to the separation between the poleward edge of the visible aurora and the boundary between open and closed field lines. This latter boundary may also mark the transition from the cool oval to the warm polar cap.

Funding for this work came from the Australian Antarctic Division (AAD), and through Australian Research Council (ARC) and Antarctic Science Advisory Committee. Pene Greet is supported by an ARC postdoctoral fellowship. The Mawson Fabry l~rot spectrometer is on permanent loan to La Trobe University from the University of Adelaide, and is operated as a collaborative project between La Trobe University and the AAD. Technical support for the FPS at Mawson in 1993 was provided by AAD Atmospheric and Space Physics (ASP) engineer Martin Tait. We thank Dr Ray Morris, ASP program manager, for the opportunity to carry out this work and for his support for the program. The Mawson fluxgate magnetometer is operated by the Australian Geophysical Survey Organisation, and in 1993 was run by Anton Rada. The Mawson ionosonde is operated by IPS Radio and Space Services (IPS RSS), and in 1993 was run by John Innis and Martin Tait on behalf of IPS RSS. We thank Peter Davies of IPS RSS for providing the scaled ionogram values used in this study, and Dr Phil Wilkinson (IPS RSS) for discussions regarding these data. We also acknowledge a number of useful discussions with Dr Mark Conde. John Innis wishes to thank the members of the 1993 Australian National Antarctic Research Expeditions who wintered at Mawson station for their assistance and support of all aspects of the AAD Atmospheric and Space Physics program.

dimensional models of auroral arcs, but found that none provided a completely satisfactory description of the data. A revised model by Walterschied and Lyons (1992) resulted in better agreement, but still could not reproduce the observed temperature drop. Unfortunately, due to the dynamic nature of the aurora during our observations, together with our reasonably limited time resolution of ~ 8 minutes for our FPS data, we are not easily able to identify observations from a single arc, and thus are unable to add to the experimental work of Eastes et al. (1992).

Acknowledgements

REFERENCES

Aruliah A. L. and Rees D.

1995

Conde M. and Dyson P. L.

1995

Crickmore R. I., Dudeney J. R. and Rodger A. S.

1991

De Deuge M. A., Greet P. A. and Jacka F.

1994

Eastes R. W., Killeen T. L., Wu Q., Winningham J. D., Hoegy W. R., Wharton L. E. and Carignan G. R.

1992

Elphinstone R. D., Hearn D., Murphee J. S. and Cogger L. L.

1991

Georges T. M.

1968

The trouble with Thermospheric Vertical Winds: Geomagnetic, Seasonal and Solar Cycle Dependence and High Latitudes. J. atmos, terr. Phys. 57, 597~09. Thermospheric Vertical winds above Mawson, Antarctica. J. atmos, terr. Phys. 57, 589-596. Vertical thermospheric winds at the equatorward edge of the auroral oval. J. atmos, terr. Phys. 53, 485-492. Optical Observations of gravity waves in the auroral zone. J. atmos, terr. Phys. 56, 617~29. An Experimental Investigation of Thermospheric Structure Near an Auroral Arc. J. Geophys. Res. 97, 10539-10549. Mapping using the Tsyganenko Long Magnetospheric Model and its relationship to Viking Auroral Images. J. Geophys. Res. 96, 1467-1480. HF Doppler studies of travelling ionospheric disturbances. J. atmos, terr. Phys. 30, 735-746.

Fabry-Perot spectrometer observations Hays P. B., Killeen T. L., Spencer N.W., Wharton L. E., Roble R. G., Emery B. A., Fuller-Rowell T. J., Rees D., Frank L. A. and Craven J. D. Hernandez G.

1987

1984

Observations of the Dynamics of the Polar Thermosphere. J. geophys. Res. 89, 5597-5612.

1982

Vertical motions of the neutral thermosphere at midlatitudes. Geophys. Res. Let. 9, 555-557. Neutral Atmospheric Waves from Atmospheric Explorer Measurements. Geophys. Res. Let. 6, 187190. Atmospheric Gravity Waves generated in the HighLatitude Ionosphere: A review. Rev. Geophys. Space. Res. 20, 293-315. Application of Fabry-Perot spectrometers for measurement of upper atmosphere temperatures and winds. In: Vincent, R.A. (Ed.). Handbook for M A P 13, 19 40. On the Relationship Between Dynamics of the Polar Thermosphere and Morphology of the Aurora: Global-Scale Observations From Dynamics Explorers 1 and 2. J. geophys. Res. 93, 26762692. Polar Cap Diurnal Temperature Variations: Observations and Modelling. J. geophys. Res. 93, 74667477. A resonance effect in AGWs created by periodic recurrent bursts in the auroral electric field. Ann. Geophys. 12, 94-96. The influence of geomagnetic activity on the upper mesosphere/lower thermosphere in the auroral zone. I. Vertical winds. J. atmos, terr. Phys. 53, 909922. Simultaneous measurements of large vertical winds in the upper and lower thermosphere. J. atmos, terr. Phys. 57, 631~543. Cospar International Reference Atmosphere: 1986. Part I: Thermosphere Models. Adv. Space Res. 8, 1471. The Generation of Vertical Thermospheric Winds and Gravity Waves at Auroral Latitudes--I. Observations of Vertical Winds. Planet. Space Sci. 38, 667684. Thermospheric circulation, Temperature, and Compositional Structure of the Southern Hemisphere Polar Cap During October-November 1981. J. geophys. Res. 89, 9057-9068. Vertical winds in the midlatitude thermosphere from Fabry-Perot Interferometer measurements. J. atmos. terr. Phys. 57, 621~529. Vertical winds in the thermosphere within the polar cap. J. atmos, terr. Phys. 57, 611-620. Thermosphere Zonal Winds, vertical motions and temperatures as measured from Dynamics Explorer. Geophys. Res. Let. 9, 953-956. Vertical motions in the thermosphere over Mawson, Antarctica. J. atrnos, terr. Phys. 48, 289-292. The Neutral Circulation in the Vicinity of a Stable Auroral Arc. J. geophys. Res. 97, 1948919499. The generation and propagation of atmospheric gravity waves observed during the Worldwide Atmospheric Gravity-wave Study (WAGS). J. atmos, terr. Phys. 50, 323-338.

Hoegy W. R., Dyson P. L., Wharton L. E. and Spencer N. W.

1979

HunsuckerR. D.

1982

Jacka F.

1984

Killeen T. L., Craven J. D., Frank L. A., Ponthieu J.-J., Spencer N. W., Heelis R. A., Brace L. H., Roble R. G., Hays P. B. and Carignan G. R.

1988

McCormac F. G., Killeen T. L., Burns A. G. and Meriwether J. W.

1988

MillwardG. H.

1994

Price G. D. and Jacka F.

1991

Price G. D., Smith R. W. and Hernandez G.

1995

Rees D. (ed.)

1988

Rees D., Smith R. W., Charleton P. J., McCormac F. G., Lloyd N. and Steen/~.

1984

Roble R. G., Emery 13. A., Dickinson R. E., Ridley E. C., Killeen T. L., Hays P. B., Carignan G. R. and Spencer N. W.

1984

Sipler D. P., Biondi M. A. and Zipf M. E.

1995

Smith R. W. and Hernandez G.

1995

Spencer N. W., Whaiton L. E., Carigan G. R. and Maurer J. C.

1982

Wardill P. and Jacka F.

1986

Walterschied R. L. and Lyons L. R.

1992

Williams P. J. S., Crowley G., Schelegel K., Virdi T. S., McCrea L, Watkins G., Wade N., Hargreaves J. K., Lachlan-Cope T., Muller H., Baldwin J. E., Warner P., Van Eyken A. P., Hapgood M. A. and Rodger A. S. Williams P. J. S., Virdi T. S., Lewis R. V., Lester M., Rodger A. S., McC.rea I. W. and Freeman K. S. C.

1988

1993

Worldwide atmospheric gravity-wave study in the European sector 1985-1990. J. atmos, terr. Phys. 55, 683~696.

1988

J.L. Innis et al.

Reference is also made to the following unpublished material Burns G. B. and McLoughlin R. 1989

Dudeney J. R.

1974

Dyson P. L. and Hoegy W. R.

1978

McLoughlin R., Burns C. B., Grant I. F. and De Deuge M.

1989

Routine Observatory operations by the Australian Antarctic Division. In: ANARE Res. Notes. No. 69, 189-192, Conde, M., and Beggs, H. (eds.) Antarctic Division, Australia. A simple empirical method for estimating the height and semi-thickness of the F2-1ayer at the Argentine Islands, Graham Land. British Antarctic Survey Scientific Reports No. 88, National Environmental Research Council, London. Spectral analysis of atmospheric waves using the Maximum Entropy Method. Department of Physics, La Trobe University, Scientific Report. Projects operated by the Australian Antarctic Division's Upper Atmosphere Physics section at ANARE stations. In: ANARE Res. Notes. No. 69, 174-188, Conde, M., and Beggs, H. (eds.) Antarctic Division, Australia.