Middle atmosphere dynamical studies at Resolute Bay over a full representative year: Mean winds, tides, and special oscillations

Transcription

1 Radio Science, Volume 36, Number 6, Pages , November-December 2001 Middle atmosphere dynamical studies at Resolute Bay over a full representative year: Mean winds, tides, and special oscillations W. K. Hocking Department of Physics and Astronomy, University of Western Ontario, London, Ontario, Canada Abstract. Wind motions in the km altitude region above Resolute Bay, Nunavut, Canada (75øN, 95øW), are presented for a representative year. The measurements were made using an interferometric meteor radar during 1997, 1998, 1999, and Results presented include a comprehensive study of monthly mean winds, using all of the available data from these 4 years, as well as running tidal studies and power spectra over the time frame and studies of special persistent oscillations which appear from time to time. All studies are accompanied with a detailed consideration of the effects of the spectral windows used. A superposed epoch analysis is used to investigate a dominant 15- hour oscillation which appeared in November 1999, and evidence of period changes, associated with changes in resonance conditions, is demonstrated. Other periodic motions are also seen, especially at periods in the range 9-12 hours. 1. Introduction The polar regions of the Earth's middle atmosphere are some of the least understood areas of the planet's atmospheric environment. Previous measurements of winds and temperatures have been made in these regions, but often over short time frames, since many such studies have been made by optical methods, which are often interrupted because of cloud. Optical methods are also generally limited to wintertime studies and produce no data during the polar summers. The only instruments capable of yearlong ground-based observations are atmospheric radars, and these are somewhat rare in these regions. Longterm studies throughout all seasons are therefore still limited in number. Furthermore, the majority of studies have concentrated on the Antarctic region, rather than the Arctic. There appear to be significant differences between Arctic and Antarctic dynamics, so it is not always possible to transpose Southern Hemisphere results to the Northern Hemisphere. Fraser [1984, 1989], MacLeod and Vincent [1985], Phillips [1989], and Phillips and Vincent [1989], among others. Satellite measurements represent another possible source of data for global studies, but they have generally not been able to make direct wind measure- ments above -- 60øN[e.g., Lieberman et al., 1998]. The CIRA standard atmospheric model of wind measurements [e.g., Fleming et al., 1990] for this region is based largely on satellite temperature measurements, and the CIRA temperatures and pressures at these high latitudes are known to be somewhat unreliable [e.g., Ltibken and von Zahn, 1991; Hocking, 1999]. As a result, it can well be expected that the CIRA wind measurements are also likely to be unreliable. For all these reasons, extended radar observations at additional northern polar sites are very much needed. Up until the present, data from only four radar stations have been reported which have in- cluded long-term mesospheric studies north of 65øN, Although we will not discuss them extensively in this and these were operative in different time frames. article, Southern Hemisphere studies of monthly These radars were sited at Heiss Island (81øN), mean climatologies have been provided by Portnyagin Tromso (70øN), Kiruna (68øN), and Poker Flat [1986], Portnyagin et al. [1993], Riggin et al. [1999], (65øN), and results have been reported by Portnyagin [1986], Portnyagin et al. [1993], Manson et al. [1990a, 1990b], Manson and Meek [1991], Massebeuf et al. [1979], and Avery et al. [1989]. The objectives of the Copyright 2001 by the American Geophysical Union. work presented in the current paper include presen- Paper number 2000RS tation of climatologies for a full 12-month period of /01/2000RS observations using a radar at 75øN. The work there- 1795

2 1796 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY fore complements the existing (limited) information about polar mean winds and tides. In addition to long-term studies, shorter-term investigations of special events and processes are also required. As noted, a large number of such specialized studies exist, especially ones made using optical methods. However, these studies are limited to nonsummer conditions and are often intermittent. Some of the special optical studies include reports by Hernandez et al. [1990, 1992a, 1992b, 1993, 1996], Collins et al. [1992], Fraser et al. [1993], Walterscheid and Sivjee [1996], Sivjee and Walterscheid [1994], Espy and Witt [t996], and Fisher et al. [1999]. Other special studies include investigations of planetary-scale and 10- and 12-hour oscillations by meteor and MF radar, such as those by Fraser et al. [ 1993], Forbes et al. [ 1995, 1999a, 1999b], Palo et al. [1998], Portnyagin et al. [1997, 1998], and Hall et al. [1998]. However, despite these studies, much still remains to be learned about the nature of special oscillations near the poles, especially during the summer and equinoxes. Forbes et al. [1999b] conclude their paper by especially emphasizing the need for such studies in the northern aparan [1997], except that the antennas did not form Arctic regions. part of the larger array, but were separately installed. In order to be able to perform the studie suggested The use of the split beam optimizes the system gain in above, an observatory has been established at Reso- the four orthogonal directions, which is ideal for wind lute Bay in Nunavut, Canada (75øN, 95øW), which determinations. contains a variety of instruments for studying the The pulses of radio waves could have a variety of upper and middle atmosphere. In this paper, we use shapes and various code types, but as a rule we results determined over several years with a meteor transmitted 5-bit Barker-coded pulses [e.g., see Farradar to investigate wind motions in the km ley, 1985; Golomb and Scholtz, 1965] in which the height region and, in particular, to build a yearly subelements had pulse lengths of 2 km. The pulse climatology. In section 2, we describe the instrumentation used. We then present monthly mean winds as a function of height and season. Following this, we investigate the behavior of the tides in this region and then move on to spectral studies of the region using running power spectra. Finally, we concentrate on specific events and present detailed studies of hour and 15-hour oscillations in particular. Other, longer-period waves are noted, but discussion of these will be left to a later study. 2. Instrumentation The instrument used for these studies was a meteor radar which was a subsystem of a larger radar that had been installed for a variety of studies at the Early Polar Cap Observatory near Resolute Bay, Canada. The full system is described in detail in a companion paper [Hocking et al., this issue] and was similar in design to the Canadian (London, Ontario) VHF atmospheric radar (CLOVAR) system, which has been described by Hocking [1997a] and Hocking and Thayaparan [1997]. In this section, we will concentrate solely on a description of the meteor radar. The reader will be frequently directed to Hocking and Thayaparan [1997] for reference, since the Resolute Bay system was based heavily on that earlier design. Another variant of the same technique is also described by Hocking et al. [2001]. The radar was a 51.5 MHz interferometric radar, which used a four-way split beam for transmission and four spaced receivers for reception. Pulses of radio waves were transmitted from a cluster of 16 antennas which were phased in such a manner that they produced four broad beams (hence the term "split beam") directed approximately to the north, south, east, and west, with mean offsets from the zenith of ø. The polar diagram of this array was similar to that shown by Hocking and Thayaparan [1997, Figure 1]. The placement of transmitter antennas was also similar to that described by Hocking and Thay- repetition frequency was 750 Hz, and the duty cycle was 5%. The transmitted radio pulses are scattered from irregularities in the atmosphere, including meteor trails, and some of the backscattered signal returns to the receiver antennas. The receiver array comprised four groups of four antennas (where the clusters of four antennas are called "quartets") which were subgroups of the larger VHF radar but which could be separately accessed using computer-controlled switches. Until June 2000 the signals received using these four receiving quartets were multiplexed into a receiver for subsequent digitization and analysis. Following June 2000, the system was upgraded by the installation of four separate receivers, which could be digitized simultaneously, thereby substantially improving the signal-to-noise ratio. The recorded signal was then searched for meteors using software algorithms, and the phase differences

3 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1797 between the signals received on the various receivers were determined. These phase differences then per- mit determination of the location of the meteor in the sky. The range of the meteor can be found from the time delay between the transmitted pulse and the reception of the returned signal. The height of the meteor above the ground can be easily found from the meteor angles and range, allowing for a curved Earth. The algorithms used to search for meteors have been carefully developed to isolate primarily underdense meteors, which are characterized by a rapid signal increase followed by an exponential decay in strength in timescales of typically s. A brief discussion of the algorithms used to isolate the meteors is presented by Hocking and Thayaparan [1997], and a much more detailed summary is presented by Hocking et al. [2001]. The algorithms are over 99% successful in identifying meteors, so contamination from auroral echoes, lightning, E region reflections, turbulent scatter, and other types of reflectors is very small indeed. The system software then determines the radar lifetime of the meteor, and the radial drift speed. As a rule, the majority of meteors can be unambiguously located, but ---20% of them can have ambiguous locality determinations. In other words, there can be two positions in the sky which give the same combinations of phase differences (within experimental error) between the various receivers. In these cases, the meteor information is stored, but the data are not used for further analysis. The meteor count rate also shows an annual variation, being maximum in summer and minimum in winter. This is partly due to a natural annual cycle but is exacerbated in winter because of radio frequency interference from various other instruments. Once the meteors are located, the radial velocities are determined, and these are then used collectively to determine the mean wind over the sky in either 1- or 2-hour bins at approximately 3-km height increments. These wind measurements include both zonal and meridional components. The details of the conversion to all-sky mean winds are discussed by Hocking and Thayaparan [1997]. An important aspect of meteor studies is the number of meteors recorded per day, since this affects the quality of the data analysis. As noted, prior to June 2000, signals from the four receiving antennas were multiplexed through a single receiver on a pulse-topulse basis. During this time frame the radar successfully detected and identified typically mete- ors per day, with the largest rates being during summer. The reduced rates during winter were in part due to interference from other instrumentation housed in the same building and partly due to a series of switches which seemed to function less efficiently in winter. These rates are lower than the system capability, because the meteor mode is interlaced with a separate radar mode which is used to investigate tropospheric phenomena. This reduces the time available for meteor detection by ---20%. After June 2000 the meteor count rates increased substantially, because of the upgrade mentioned previously, and in November 2000 the problem with the switches which was mentioned previously was resolved. As a result, count rates after that time were typically over 1500 meteors per day during all of the year, and often exceeded In our subsequent presentation we will concentrate on the data recorded prior to 2000, since these have been analyzed most thoroughly. However, for the monthly means we will use all of the data available up until This is because we have found significant year-to-year variability, and we would like our mean values to be reasonably representative of longer-term means. Because our primary data come from the period prior to 2000, we will generally use 2-hour bins for studies of shorter-period oscillations and will average data into two height regimes, centered at 86- and 93-km altitude. For tidal studies we will use a "composite day" approach, as will be outlined in section 3.1, thereby permitting better height resolution. These approaches help compensate for our lower meteor count rates in the early years of the radar operation. Hence the data presented herein will include all information available up until February 2001 for calculation of monthly means. For studies of planetary waves, tides, and special oscillations we will concentrate on the periods July and August 1997, April to September (inclusive) 1998, November to December 1998, and January to December We should comment on the quality of our data. We have taken considerable effort to cross-check our system. Since the radar can be operated as a Doppler radar, we have compared tropospheric wind measurements in this mode to radiosondes. In addition, because the radar can detect polar mesosphere summer echoes (PMSE), we were also able to calculate mesospheric winds using the main beam Doppler mode, during periods of strong PMSE. Furthermore, during meteor shower conditions we were also able to

4 1798 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY use the meteor locations to determine shower radi- this stage, we simply recognize that we need to be ants [e.g., see Hocking et al., 2001]. All of these various comparisons have been described by Hocking aware of the possibilities of contamination but also note that our method is no more or less susceptible et al. [this issue]. In all cases, the comparisonshowed than other methods. good agreement, indicating that the radar is properly calibrated and that wind directions are accurate to within better than 3 ø and wind magnitudes are accurate to typically better than 10% in meteor mode. In addition, we also examined closely the behavior of the meteor winds when we upgraded the radar in June No significant changes in general behavior occurred during this time frame, indicating that although the upgrade improved our data quality, it did not change the character of our data and that the data prior to the upgrade and after it are both reliable indicators of the true atmospheri conditions. Finally, we need to be aware of potential contaminants which might affect our data. The most serious potential problem is the possibility of the meteor trails being controlled by electric fields, as modeled by Oppenheim et al. [2000]. However, these studies were performed at an altitude of 105 km, well within the E region, and were performed for equatorial conditions. Only one study has yet shown any support for such contamination [Chang et al., 1999], and that was also at the equator. Thus both such studies were for high-altitude equatorial conditions, and any similar effects at the poles are unknown. We do acknowledge that some effect might be possible, but this can only be considered as speculation at this time. We therefore largely avoid this issue by concentrating our studies on data below 98 km, with the greatest emphasis being on data recorded at heights below 94 km. We will therefore consider all of our data as this altitude regime: for example, optical methods are biased by their inability to operate in daytime, strong moonlight, and cloudy conditions. The MF spaced antenna method is known to be suspect at times at heights above 92 km [e.g., Hines et al., 1993; Hocking, 1997b; Guruburan and Rajaram, 2000]. Mesosphere- stratosphere-troposphere (MST) Doppler data are often biased by the fact that they can only record data during periods of strong turbulence, and so forth. At 3. Monthly Mean Winds and Tides 3.1. Analysis Method Our first investigations of the winds at Resolute Bay will be in regard to monthly mean winds and tidal oscillations. In order to optimize the information in this regard, we adopted the following "composite day" analysis (also called "superposed epoch" analysis). For each meteor within any particular month the meteors were binned solely according to time of day, independent of day of the month. This permits a large number of meteors to occur in each time bin. Then the all-sky wind determination discussed in section 2 is applied to these data, to give a "typical" wind variation over a day. This procedure was applied to produce wind parameters at 82-, 85-, 88-, 91-, 94-, and 98-km altitude. In each case, the winds refer to a height bin of about +_2 km around the representative height. There is thus some overlap between successive bins. Two-hour time bins are generally used, with a 15-min overlap temporally between successive time bins. The data were then fitted with mean, 24-hour, and 12-hour oscillations by using standard least squares fitting techniques. The mean offsets give the monthly mean winds, and the amplitudes and phases of the diurnal and semidiurnal oscillations describe the tidal characteristics. Each will be discussed in due course. Our data cover the period from 1997 to February 2001, with the earlier data being more accurate reflections of the true atmospheric dynamics intermittent and the later data being of a more continuous nature. and present them as such. If indeed there is contamination at the very highest levels, it is not apparent at this time. If in several years' time it becomes clear that such contamination does exist, presentation of this data will still have served a useful purpose in helping elucidate these effects. This approach is the 3.2. Monthly Mean Winds Figure 1 shows the monthly mean values of the zonal and meridional winds determined by the above procedure at Resolute Bay. Prominent features for the zonal component include a wind weakening and same as that used in all methods of measurements in reversal below 88 km in summer and a large increase in wind speed above 91 km in summer. In regard to the meridional component it is noteworthy that the winds are southward for most of the year and become especially strong in summer at km altitude. Several of these features are unexpected. Perhaps the most surprising is the fact that the meridional flow is almost always southward at these heights. Theoretical predictions suggest that the meridional flow should be northward in winter, and southward in

5 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1799 Velocity (m/s) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month of Year Velocity (m/s) b 15. :&, t:m.., -, :::- :... '*,9.'.'.-'.- ;, i.'.' ': ::::::*....': ::::::.' :.'./' :.,,.'.: :.. i,.,. ::...::... : :.: : :.:.: : :.: : :.:..:.:.:.:.:.:.:.:.: :.: : :.:..:.:.: :.:.:.: :.: :.:.: : ::.45 ' :::::.".-::::::. :::::::::::::::::::::::::: ========================= i:i:i:! i"'""'"'":':" Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month of Year... ;'i i... :::::::::::::::::::::::::::::::: Figure 1. (a) Contour plots of the nonthly mean zonal winds at Resolute Bay, for the altitude range from 82 to 98 km. Actual monthly values are also written on the figures. Positive values represent eastward winds. (b) Monthly mean meridional winds at Resolute Bay. Positive values represent northward winds. (c) Zonal monthly mean winds for Resolute Bay based on the CIRA empirical model. (d) Zonal monthly mean winds for Kiruna as measured by Massebeuf et al. [1979].(e) Meridional "meteor-level" winds for three northern polar sites, including Resolute Bay (see text for details) summer, because of the effects of gravity wave drag on the zonal flow at all latitudes [Lindzen, 1981; Holton, 1982, 1983]. This hypothesis consistent with our observations that the flow becomes more strongly southward in summer, so that flow out of the pole is increased. However, the fact that the winds do not become northward during winter is very surprising and is not consistent with this model. Nevertheless, our results are evident down to 82-km altitude, and we consider it unlikely that this is an electric field contamination effect. at all heights, with a particularly strong effect at km altitude. There is no evidence at all of a strong eastward jet in summer above 91 km. It is, at first glance, more consistent with the flow proposed by Holton [1982, 1983]. One needs to therefore ask whether the meteor winds are reliable, or whether there was perhaps an equipment malfunction? In answer to this query, we have partly addressed this issue in section 2, where we discussed the radar calibration, but we will consider these points more strongly here. We begin by noting that the meteor selection criteria which we use are very robust. We are confident that over 99% of all our recorded It is important to compare our results to previous data and physical expectations. Figure lc shows the zonal flow as presented in the CIRA model [Fleming meteors were true underdense meteor signals. We et al., 1988, 1990]. It shows a flow reversal in summer can also say with considerable confidence that our

6 1800 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY Velocity (m/s) , 91.0.g 7.s Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month of Year Velocity (m/s) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month of Year 20- Meridional Wind km. Poker Flat, 65øN, Resolute Bay, 75øN, Heiss Is., 8 IøN, I I I I I I I I I I I Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month of Year Figure 1. (continued) determinations of the meteor positions are very accurate. We know this because on several occasions we used inverse radiant mapping techniques to deter- mine the radiants of meteor showers [e.g., see Hocking et al., 2001; Jones and Brown, 1993, 1994; Jones and Morton, 1982]. We have been able to unequivocally determine the radiants of the Geminids, the Zeta Perseids, and the Quadrantid meteor showers, to within _+2 ø declination and _+4 ø in right ascension. This would not have been possible if our meteor direction determinations were wrong. The radar has also been used in tropospheric mode to compare the directions of radar-derived tropospheric winds to directions determined by balloon-borne radiosonde releases [Hocking et al., this issue]. Directions agree to within ø of azimuth. The radar is also care-

7 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1801 fully calibrated for range calculations. Therefore we feel confident that our system is performing properly and that the features described are real. We therefore turn to examination of previous studies on monthly mean winds Monthly Mean Winds: Comparisons With Other Studies Certainly, the CIRA model does not show the strong eastward jet above 91 km in summer. Portnyagin et al. [1993], using a meteor radar at Heiss Island (81øN), did not show evidence for such behavior either (e.g., see their Figure 4b). The radar used for those data had only limited height information, which may have hidden an upper level jet. However, as seen in Figure ld, Massebeufet al. [1979] did indeed see a similar jet during a 1-year study in Portnyagin [1986] shows weak evidence for a small upper level jet above 94 km in summer (data provided by Avery et al. [1983, 1989]). Manson et al. [1990a, p. 285, Table 1] (Poker Flat) also show a weak maximum in eastward winds above 95 km during summer. Manson et al. [1990b, Figure la] and Manson and Meek [1991, Figure 2a] also show a weak jet at Tromso (70øN) above km. These measurements are close to the E region, where the MF technique may be less reliable than at lower heights [e.g., see Hocking, 1997b], but the indications are certainly suggestive. The wind speeds in the data presented by Manson et al. [1990b] and Manson and Meek [1991] are not as strong as those demonstrated by Massebeuf et al. [1979] for Kiruna (68øN), but at least a jet is present. In addition, evidence for this jet has also been reported using the European Incoherent Scatter (EIS- CAT) UHF radar. rdi and Williams [1993] found evidence for a strong eastward jet which occurred in the height regime between 95 and 115 km, with wind speeds approaching 50 ms -1. With this considerable amount of other evidence for such a jet we therefore feel that it is entirely appropriate that we report it here as well. There is undoubtedly a discrepancy with the CIRA model, but we feel that our observations are real. It is true that our data could be contaminated by electric field effects, but it is equally true that the CIRA model could be an inaccurate representation of the true wind field. The CIRA model is based on satellite temperature gradients, which can be unreliable at high latitudes. Other measurements of upper level winds in the polar regions are provided in large part by MF methods, and these can be somewhat unreli- able (often providing quite weak winds) at heights above km, as already discussed. With regard to the summer zonal wind reversal at 68øN, Figure l d shows that it is much weaker, and confined to lower heights, than the CIRA model would suggest. Portnyagin et al. [1993] also shows that the reversal does take place at Poker Flat, but again not with the intensity shown in Figure l c. We surmise that the reversal takes place, but at lower altitudes than indicated in Figure lc, and possibly the height of the reversal decreases as the latitude increases. We expect that most of the reversal takes place at Resolute Bay at altitudes below 80 km. We therefore conclude that our zonal winds, shown in Figure la, do indeed properly represent the monthly mean winds at Resolute Bay and that the CIRA model is not a good representation of the behavior, especially in summer. This is consistent with earlier statements which we made about the inaccuracy of the CIRA for describing temperatures, pressures, and densities over the polar regions. With regard to validation of our meridional data there are even fewer data to check against. CIRA does not produce a meridional model. We have therefore decided to compare our results to the meteor data presented by Portnyagin [1986] and Portnyagin et al. [1993]. However, those data were acquired using a meteor radar which had no height discrimination, so that all the data were ascribed by the authors to a single height of 95 km, this being the height at which the authors claim the meteor count rate peaks. This estimate of 95 km for the peak height is, in fact, in error. By operating several radars which do have height discrimination we have been able to determine the following about the altitude of the peak height of meteor count rates [see Hocking et al., 2001; Hocking, 1999; Hocking et al., 1997]. For upper level temperatures of around 200 K and more, the position of the peak count rate is at an altitude of km for a 51.5 MHz radar (as at Resolute Bay), km for a radar working in the range MHz, and 96 km for a radar working at MHz. However, during northern latitude summer, when the temperatures become as cold as 120 K, the height of peak count rate increases by km. For example, at Resolute Bay it increases from 86 km to 88 km. At 30 MHz the peak count rate occurs at km. Using this information, we have therefore averaged our own meridional winds to produce the "effective" winds which a radar without height resolution should produce. During nonsummer months we averaged

8 1802 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY Zonal --- Meridional _ \ /3,,\ 0 ;n F;b r A r y ;n ;I A;g S;p ct v D;c ;n F;b r r y ;n ;I A;gS;p O;t v D;c 24- o. n F b r r y n onth 1 AugS p a vd c n Feb r r yjun Jul AugSep Oct v Dec onth Figure 2. Amplitudes and phases of the (left) diurnal and (right) semidiurnal tides as measured at Resolute Bay in 1998 and Data are shown for altitudes of 85, 88, 91, and 94 km and were calculated by vectorially averaging the data from the two successive years. Zonal data are designated by solid lines, and meridional data are designated by dashed lines. The different heights are not distinguished in this figure: The figure is designed to show the general behavior of the gross features, rather than individual idiosyncrasies at any particular height. The approximately 90 ø phase shifts between the zonal and meridional phases (hours of maximum) are particularly apparent. our data over heights of 88, 91, and 94 km, while during the months of May to August we averaged the data over the height range km. The results can shown as "error bars" in Figure le for the Heiss Island January and July data. Therefore we consider that our results are broadly consistent with the few be seen in Figure le as the large solid circles. These other data which exist in the literature and are data are compared to measurements by Portnyagin et confidenthat our results are a true representation of al. [1993], from their Figure 4a, using all their data the winds over Resolute Bay. from 1965 to We have also applied a similar averaging process to the data for Poker Flat, as 3.4. Tidal Results presented in Figure 2 of that paper, and these are shown as the shaded squares. Clearly, both the data from Heiss Island and Resolute Bay show predominantly southward flow throughouthe year, and in- The results of our harmonic fits and tidal analyses are shown in Figures 2-4. These results are concentrated on the period from 1998 to 1999, inclusive. In Figure 2 we show amplitudes and phases throughout deed, agreement is excellent in all months except the year, but we have not separately indicated each April, May, and September. This sustained southward height, since we only wish to produce an overview of flow is not apparent at Poker Flat, which is a more the general behavior. We should also emphasize that southern site. Portnyagin [1986, Figure 4a] also shows although these results are averages over 2 years, we that there is considerable interannual variability and shows that the monthly mean in January can vary have found them fairly consistent between the two years, and the average is quite representative of both anywhere between -10 and 10 ms -1 while the years. Further studies over a longer time frame will be monthly mean in July can vary between -12 and 2 ms -1 depending on the year. These ranges are left to future publications. With regard to the data currently under discussion,

9 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1803 eø t Amplitudes km km '... A' t' ' Jan F bmar Apr May Jun Jul ugs po Nov Dec ' Ampl#udes krn // ' ' 94km _ L _.. km ;" ' ;,... n F b M r Apr May,Jun Jul Aug Sap Oct Nov Dec O :z: a 3O 85km 18 Phases (Local Time) Jan i Feb i Mar I Apr I May i Jun i j I Aug i Sap i Oct i Nov i Dec I Month 3O :-. m r' Phases (Local Time) 0 / i, i i i i i i i i i I I Jan Feb Mar Apr May Jun Jul Aug Sap Oct Nov Dec Month 2O l 15 Amplitudes km lo,...,,n?m 82km I I I I I I I I I I I I Jan Feb Mar Apr May Jun Jul Aug Sap Oct Nov Dec 2O 15 0 / Ampl#udes km I I I i i i I i i I I I Jan Feb Mar Apr May Jun Jul Aug Sap Oct Nov Dec 18 ß km 0' Phases (Local Time) Jan i Feb i Mar I Apr I May i Jun i j I Aug i Sap i Oct I Nov i Dec i Month km lkm o' Ph es (Local Time) i i i i I i I I I I I n Faber r y n I Aug p Oct vdec h Figure 3. Amplitudes and phases of the measured tides compared to the global scale wave model (GSWM) [Hagan et al., 1993, 1995] for the zonal component, in the regions km and km altitude, for both the (a) diurnal and (b) semidiurnal components. The GSWM values are shown by the large grey circles. The meridional components have not been plotted. important results include the fact that the diurnal tide has amplitudes in the range 5-10 ms - throughout the year and that there is a clear phase quadrature between the zonal and meridional components of the diurnal tide. The meridional component maximizes to the north 6 hours before the eastward component maximizes, indicating a clockwise rotation of the tidal wind vector with time. With regard to the semidiurnal component there are maxima in activity in January and February and in September. Again, the zonal and meridional components are in phase quadrature, with clockwise rotation of the tidal vector. The phases (times of local maxima) for both the diurnal and semidiurnal tides are relatively invariant throughout the months of April to October, although there is some extra variability when the tides are weakest (as may be expected). During December to February the zonal semidiurnal phases seem somewhat displaced relative to those from April to October, suggesting evidence of a "bimodal character" [cf., e.g., Manson et al., 1999]. In Figure 3 we have compared our zonal tidal results with the predictions of the global scale wave model [Hagan et al., 1993, 1995, 1999]. We have also extracted some of this information from Manson et al. [1999] and include comparisons at around 82 km and

10 ß HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY,Diurnal, April 98 M Semidiurnal, August 98 M z M Z M g M Z M M Z 82, Semidiurnal, September / / /0 Local Time of Max. b Local Time of Max. ½ Local Time of Max. Diurnal Tide 1 O0.... ' *"i ;c ' o... :::":":':'7.:."...:' '"::... : '"..:.'. ::. '.' so '....: , ' /./ \ 0 \ 0 /'_,/..'\..... / \,, - -5o '-'"'-'...'... 57::::v... "? ' "':':' ' ' -"-.:-,v... d -1 O0... :: '... ' Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec loo I I I I "--' I I I I I I I I Semidiurnal Tide Month Meridional Zonal '" 1 Zonal (model) - so ß -50 _ -100 i " [... i... ]... [... I i [ t Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month Figure 4. Examples of the progression of the tidal phases (hour of maximum) as a function of altitude, for both the zonal (Z) and meridional (M) components, for selected months: (a) a typical diurnal case, (b) a semidiurnal case where no determination of vertical wavelength was possible, and (c) an example of a semidiurnal tide where a vertical wavelength determination was possible. (d) Vertical wavelengths of the diurnal tides at Resolute Bay as a function of month. Wavelengths determined from the GSWM [Hagan et al., 1993, 1995] are not shown, but generally they are predicted to be very large, approaching infinite (at least for the zonal component [e.g., see Manson et al., 1999]. Negative wavelengths refer to cases where the phase increases with increasing height, which could indicate upward phase progression and downward group propagation, or could result from interference between differentidal modes. (e) Vertical wavelengths of the semidiurnal tides at Resolute Bay as a function of month. Wavelengths determined from the GSWM [Hagan et al., 1993, 1995] are also shown for the zonal component (shaded squares).

11 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY km. We compare the measured data at 82 and 85 km to the predictions of the model at approximately 82 km, and we compare the measured data at 88, 91, and 94 km to the model data at -91 km. The agreement is, in general, very good, with the phases of the diurnal tide showing excellent agreement and the amplitudes showing very similar behavior between the model and experiment. With regard to the semidiurnal tide the nonsummer phases show good agreement, but there are discrepancies in the summertime phases. This may be because the model does not properly parameterize the mean winds in summer, since we have already seen that the true winds are different from the CIRA winds (Figure 1) in those months. The amplitudes of the semidiurnal tide are in general agreement, except that the peak in amplitude at 94 km in winter occurs in February and March, whereas the model predicts a peak in January and February. The model also does not produce a local maximum in tidal activity in September, which it should do if it is to match the experimental data. In Figure 4 we have used the information about phases of the tides to determine their vertical wavelengths. Figures 4a, 4b, and 4c show some examples of the phase (hour of maximum) as a function of height, to demonstrate the different types of situations which can occur. As a rule, the phase progression is moderately linear, so that vertical wavelengths can be deduced (as determined by the straight lines fitted to the plots). Vertical wavelengths deduced in this way are shown in Figures 4d and 4e. In cases like Figure 4b, however, the phase variation is not linear as a function of height, and it is not possible to determine a unique vertical wavelength. Negative wavelengths refer to cases where the phase increases with increasing height. This could indicate upward phase progression and downward group propagation of the wave, or it could also arise as an interference effect between different tidal modes. Thus negative wavelengths can be taken either as indications of downward group propagation or complex tidal structure. The predictions of the GSWM are also shown for the zonal semidiurnal component. Agreement between experiment and the model is very reasonable during January to April but not good during other months. For the diurnal component, experimental results show long vertical wavelengths of the order of km, in broad agreement with the model's prediction of evanescent behavior. 4. General Spectral Studies 4.1. Spectral Techniques We now turn to a general examination of the data for special oscillations. Because the meteor count rates are only modest (because of the use of a single receiver prior to June 2001 and also because of interference at certain times, especially during winter), we group the data into only two heights. We combine the data from 80 to 89.5 km as one group and that from 89.5 km upward as a second group. The first data set corresponds to a weighted mean height of 86 km, while the second is assigned to a weighted mean height of 93 km. We then produce 2-hourly means of these data sets and perform all our subsequent analyses on these two groups of data. Figure 5b shows the results of a running power spectrum over a full year of data, using a sliding window of length 20 days. This procedure is also called a "short time Fourier transform" [e.g., Walter- scheid and Sivjee, 1996]. The procedure involves taking a length of data W, where W is a window length (say 20 days) out of a longer time series, and Fourier transforming this shorter data set. Then the window is moved forward in time along the larger data set (typically by, say, 1 day), and the process is then repeated. Spectra produced from each successive calculation are then laid side by side and used to create a contour plot, such as the one shown in Figure 5. Within each spectral computation the mean wind is removed, and a suitable windowing function is applied. In our case, we display the mean winds at each step in Figure 5a. We frequently use a window which is flat over the inner 80% of the data but tapers down at the ends of the data set in a sinusoidal manner. The end 10% of data cover a half-cycle of a sine-waveshaped taper, with the weighting being unity nearest the center of the window and zero at the edges. Since the wind comprises both a zonal and a meridional component, we treat these as real and imaginary components, respectively, and Fourier transform this complex time series. The result produces a so-called "rotary spectrum," in which both positive and negative frequencies are represented. Positive frequencies correspond to anticlockwise rotation with increasing time, and negative frequencies correspond to clockwise rotation. After the spectrum has been performed, we then perform a boxcar running average of five points in length across the spectrum, in order to improve the statistical reliability of our spectra. A five-point running mean produces confidence limits

12 1806 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY

13 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1807 O : (s.no po.uod I I (.s m) p Is pu!a

14 1808 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY on a spectral power density estimation S of approximately 0.35S and 2.1S at the 95% two-sided level. In other words, if we measure a spectral density S, then some sort of taper on the ends, like the cosine taper described above. Other options can be used, like Hamming and Hanning windows. We then Fourier there is a 95% probability that the true spectral transform this function and find the power spectrum. density actually lies between 0.35S and 2.1S. Figure 6 shows a running power spectrum produced in the same way for the 93-km data. Following this, we then apply another running mean to the spectrum itself, using a function R(n) which is generally a boxcar of length typically three or five We mentioned above that Figure 5 was produced points. Thus our final spectrum S(n) can be written as from a full year of data. However, at the time of preparation of this graph, data had only been thoroughly analyzed to November 1999, so we have used the data from December 1998 to complete the "year." S(n) = {[F(n) Wl(n) W2(n)]* x [F(n) W1 (n) W2(n)]} R(n), This could have lead to discontinuities at the Novem- where the asterisk represents the complex conjugate, ber-december interface, but no serious problems n is the frequency, W1 is the Fourier transform of seem to have arisen. Our raw data are regularly spaced at 2-hour intervals, and our acceptance rate at 86 km is 83%. Because of missing data some authors have used special algorithms designed to compensate for irregw l(t), W2 is the Fourier transform of w 2 (t), and the circled "times symbol" represents the process of a convolution (or, equivalently, a running mean, since the windows are usually symmetric). To a rough approximation, we may consider the ular data spacing (e.g., Lomb [1976], among others). measured spectrum as a convolution between F*F However, like Walterscheid and Sivjee [1996] and (i.e., the true spectrum) and a function R' (n), where Zhan et al. [1996], we have found that for our cases R' (n) is given by the fast Fourier transform approach produces almost identical results and is much faster computationally. Where there are single or pairs of missing points surrounded on either side by good values, we inter- Figure 7. (opposite)(a) Data acceptance window for 86 km for the time period covered in Figure 5. A "1" indicates polate across the regime. Where there are more than that a useful wind measurement was made (requiring two missing points, we set the data value to zero. The detection of, and measurement of the radial velocity of, at "window function" for our data then becomes a least seven meteors in the appropriate height-time bin), while a "0" indicates that insufficient meteors were detected sequence of "ones" and "zeroes," where unity indicates data acceptance and zero indicates no valid to perform a suitable estimate of the wind at that time. In all, the acceptance rate at 86 km is 83%. This includes times when the meteor mode was turned off so that the radar data. Data acceptance requires at least seven useful detections of meteors in any height-time bin. (Unfor- could be used for other dedicated experiments. (b) Running spectrum of the data "acceptance series" shown in Figure tunately, this criterion was not always met during 7a. The data set has been processed in an identical manner winter prior to the year 2000, because of serious to that which was used to process the real data (see Figure interference (as discussed), which accounts for our 5). This spectrum therefore serves as an indicator of the periods of missing data.) We can consider our time likely sidelobe contamination which might appear in Figure series as the true time series multiplied with this 5 due to the window. The largest "sidelobes" occurred window function. We can obtain an idea of the during February and March, and November, but they were always below 8% of the maximum and were within 0.5 spectral contamination due to data gaps in the fol- cycles per day of the maximum. During the rest of the year lowing way. all sidelobes were less than 2% of peak power. (c) As in Our process can be thought of as follows. Suppose Figure 7b, but for the data at 93 km. The data acceptance the true time series is f(t), where f is a complex time window is not shown. Overall acceptance rates were -50% over the whole year but were worse in winter and better in series. The Fourier transform of f is F(n), where we use the symbol n to represent frequency. The function f is not the function we measure, but the one we would like to measure, i.e., the "true" time series. We then multiply by a window function w (t), where this function comprises a series of ones and zeroes, as described above. We then multiply by a further window w2(t), where this is a window function with summer (see text for reasons). For the months of May to September, the sidelobes were always below 4% of the maximum. For late March and April the sidelobes were below 8% of the maximum. For February and early March, and November, the sidelobes reached as high as 25%. In the latter case, sidelobes can be seen at periods of 24 hours, so it is necessary to be very cautious for these months about any spectral lines present in Figure 6 which have regular spacings of 1.0 cycles d -.

15 ß ß HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1809 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (a) Month and Year O o =1: : I= 4.o 12.0,, : o -8.0 ' 1....o :...::'": ': ':-::=:=...':::': =::':" :'-... :.--'-:'.':':' : ': -::::-::-:- -- ß : Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Contour levels 3 db steps (b) Month and Year O :' // /- ;---'---.. _-" t_ =i.= " - " ': ;". "'-- -' '- ' " '' " '"' '"' ' " t" i I i i :-12.0 ß ß -4.0 I... I... i... i... I I I Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Contour levels 3 db steps ": :' (c) Month and Year

16 1810 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY R' (n) = {[Wl(n) W2(n)]*[Wl(n) W2(n)]} R(n). We emphasize that this is only approximately true, but it is adequate for our work here. It is quite accurate in the vicinity of large spectral spikes, which will be the most common situation in which we are interested in sidelobe effects. We can therefore derive this function R' (n) by repeating the sequence of steps used to derive the original power spectrum, but using our data series as w l(t) in place of f(t), and not removing the mean. We usually add 4 times as many zeros as there are data points on the end of w when we do this, to improve the frequency resolution by a factor of 5 for our determination of this "window spectrum." Once this is done, we can then produce a function which gives us a good idea of the likely contamination of the spectra due to sidelobes. Figure 7b shows an example of such calculations, where the original window function w 1 is shown in Figure 7a. A similar window spectrum is shown in Figure 7c for the 93-km data. These spectra tell us how severely the spectra are contaminated because of the missing data and the finite length of the window. Typical levels of contamination are indicated in the caption. Only the data at 93 km in February and March are seriously contaminated by the effects of data gaps. Throughout this text we will present window spectra of this type, in order to give the reader an indication of the likely spectral contamination due to any data gaps Spectral Results Several features are immediately apparent from Figure 5b. First, tidal signatures are evident at frequencies of 1 and 2 cycles per day. They tend to be much more common at negative frequencies (-1 and -2 cycles per day), indicating predominantly clockwise rotation with time. This is consistent with our earlier tidal studies (section 3). It should be noted that a spectral line corresponds to a tide only if its frequency is very close to _+1 or _+2 cycles per day: Spectral peaks which deviate from these values are not tides. Examples of spectral peaks which are close to tidal periods, but are not tides, can be seen just below -2.0 cycles per day along the vertical line denoted by "A" and just below -2 cycles per day along the line "F." Examples of periods where the tides are dominant occur along the vertical lines "C" and "D": horizontal grey lines occur at exactly -1 and -2 cycles per day. The semidiurnal tide is especially (2) persistent during late August and September, and this is consistent with Figures 2 and 3. It is also clear from Figure 5 that there is more intense wave activity in winter than in summer, with a large amount of planetary wave activity at periods of a day or more. During summer the tidal activity is the predominant activity. We cannot examine all these motions, but we will concentrate on a few special cases, all with periods less than 24 hours. These have been called normal modes of the atmosphere [e.g., see Longuet-Higgins, 1968; Oznovich et al., 1997; Sivjee and Walterscheid, 1994; Sivjee et al., 1994]. Examination of longer-period motions will be left until later publications Waves With Periods of hours We begin by examining the oscillation centered at a frequency of approximately -2.4 cycles per day in Figure 5, during late January and early February. This is indicated as the lowest darkened band along the line "A" in the figure. The mean period is around 10 hours, but we will see that there is some variability around this value. We begin our investigation by repeating the running power spectrum, but in this case we use a window length of only 4 days. The result is shown in Figure 8. It is clear that many of the features seen in Figure 5 were in fact due to short-term, intermittent processes, including the tidal activity. The cause of the spectral peak near 2.4 cycles per day was in fact two separate bursts of coherent activity, both of which are labeled in the figure as "spectral peaks of interest." The first covers the approximate period from January 30 to February 2, 1999, and the wave has a period of hours (frequency centered on 2.33 cycles per day). The second covers the period February 7-12 and has a period a little less than 10 hours (frequency is centered on 2.5 cycles per day, with a corresponding mean period of 9.6 hours). In the spectrum produced using a 20-day running power spectrum these features merged into one apparent spectral line with a peak centered around 2.4 cycles per day. We have therefore concentrated on these two time intervals. Figure 9a shows the spectrum calculated for this time period, together with the data selection window (Figure 9c) and the window function's spectrum. The first sidelobe of the window spectrum is 14 db below the peak. A graph of the window spectral function is also shown as the thin dashed line in Figure 9a. Spectral lines like those seen in Figure 9a

17 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1811 Spectra] peaks of interest Contour levels 6.0.: - r. _.... : : ---,... : g, :--, ,t..,-h ½... ; '-. : ½-: :...;: ; " R-: :.,:: 3 db steps. (m2s '2 day) ß.:.:.: :...,. :.-.'.:..:: '. -..'.' -s.o -6.0 :. t:.?.... :..:.:, : :, :x../ : : t : ' ) :.:;.. : - g - - :... : '.- -: : : : ½: : :.: ::..:.::: - ; * :. T :.::....:: : :... :..:. :½(½ :' :.:.:: ;. -..: " ' '" ' :..:.....' :...?... : W:{' "- "'t'... ". ' ':' L :: 4 o ' '""' J uau Febma Time (days) & O... ' "' ]. O0.0 ' '": ':': SO. DO l, SO B. 15 '"->' ' 5.00 Figure 8. Running power spectrum of the 86-km data for the period January 15 to February 15, In this case, the window width was 4 days, and the window was stepped at intervals of 0.5 days. The window was tapered at the ends using a cosine weighting on the first and final 10% of the data and a further three-point spectral smooth was applied to the spectrum in the frequency domain. Spectral peaks of particular interest are indicated. obey a chi-square distribution with 6 degrees of freedom, since the boxcar running mean applied to the spectrum was performed over a length of three points. The peak power of each peak is only an approximation to the true value, but we can place confidence limits on the likely range of values which the "true" peak height should have. The faint horizontal dashed line labeled "95%" shows the smallest height which a true peak would need to have to ensure that there was a 95% probability that the measured peak would lie above the faint horizontal dashed line labeled "noise level" in the figure. We have used a one-sided chi-square test. Peaks below this line have at least a 5 % probability that they occur simply by chance. Thus there are only two peaks in Figure 9a which are significant at the 95% level: the one at cycles per day and the one at about +0.4 cycles per day. A smaller peak at about -1 cycle per day may be real but is not significant at the 95 % level. The first peak is the one we are interested in, and it is clearly offset from a frequency of -2 cycles per day. It has a period of just over 10 hours, with possible uncertainty of hour. We will shortly show the time series for the wave in the above time interval in more detail, but first we also want to examine the time interval between February 7 and 12, Figure 10b shows the spectrum, and Figure 10a shows the actual time series. Figures 10d and 10c show the window selection function and the corresponding window spectrum. There are three statistically significant peaks, with frequencies of -2.5 (9-10-hour period), +2.0 (12- hour period), and just below -1.0 cycles d -1. The source of these waves is uncertain: They may have been propagated from below or above or have arisen as a local resonant response. However, even if they propagated into the region, they probably need to have been amplified locally, in order to be domi- nant in the spectrum. It is therefore possible that the peak at -2.5 Hz is of similar origin to that producing the frequency of cycles per day in Figure 9 but that the "resonance conditions" have changed in the interceding interval (that is, the background winds, temperatures, and gradients have changed to produce a slightly different resonant frequency). In Figure 11 we show an example of a true tidal oscillation, in order to confirm that the displacements in Figures 9 and 10 do truly represent nontidal oscillations. We have chosen some data from August 26 to 30 as our reference. This is a somewhat unusual example, because the tidal motion is not circular, but rather linear, with the zonal and meridional oscilla-

18 1812 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 2.50 day Data ' I I t I ' Power... Sp!ctrum.,, _ p,: ¾ I' -,, _9_s_ø/_ø g X, 1 I * ' " ' ',?;: : :.. '-:: is -I 0.0, '..., ' (a) Rotation Clockwise Frequency (cycles/day) Anticlockwise Rotation -] Window ' 0.0 o,.. o= o (b) Frequency (cycles/day), 30 Jan. 31 Jan. 1 Feb. 2 Feb. 3 Feb. i (c) Figure 9. (a) Spectrum of the 86-km data for the time period January 30 to February 3, The window used was a cosine weighting on the first and last 10% of the data. A further three-point running mean was then applied to the spectrum in the frequency domain. The 95% confidence level for the hypothesis that any spectral peak lies above the noise level is also shown. The mean noise level is also shown. (b) Power spectrum of the window function used to prepare Figure 9a. Note that a logarithmic ordinate scaling has been used. The faint dashed line shows the window spectrum corresponding to the data selection function shown in Figure 9c, with a cosine bell weighting applied to the first and last 10% of the window. This graph is presented with 5 times the spectral resolution of the data spectrum shown in Figure 9a, to emphasize the effects of the sidelobes. The thick solid line shows the window function after application of a five-point running average (to return the graph to a resolution equivalent to that used for the data analysis) and then a three-point running mean across the spectrum (to duplicate the process used in forming Figure 9a). The largest sidelobe is 14 db below the maximum (i.e., ---4%). The thin dashed line in Figure 9a near 0 cycles d-1 shows the window function from Figure 9b on a linear power scale. (c) Data selection function. tions moving in phase. Approximate maxima of the zonal component are shown in Figure 11a by the vertical arrows. The key point to recognize here is that the spectral peaks are indeed very close to the expected semidiurnal frequency, and certainly more so than the peaks in Figures 9 and 10. We take this as further confirmation that the oscillations in Figures 9 and 10 are not tidal oscillations.

19 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1813 Zonal... Meridional ' ' 20 - (a) o Day in February 1999 (UT) 2.50 day '1 T = 9.6h 300'0 Data I ' I /l", ' ' _ Power Spectrum. ' ' :[', ',... _95% I ,...,..., '' Clockwise Frequency (cycles/day) Anticlockwise Rotation Rotation (c) Day in February 1999 (UT) - Window Power Spectrum (d) -40.0,,,... i...!. ß Frequency (cycles/day) Figure 10. (a) Time series of zonal and meridional components for the period February 7-12, The mean winds over the period have been subtracted. (b-d) Similar to Figures 9a, 9c, and 9b, but for the period February The thin dashed line in Figure 10b near 0 cycles d -1 shows the window spectral function from Figure 10d on a linear power scale. The 95% confidence level for the hypothesis that any spectral peak lies above the noise level is shown. The mean noise level is also shown. The oscillation in the interval January 30 to February 3 is moderately "pure," in the sense that there is only one significant peak at periods shorter than ---2 days. It should therefore be possible to see the oscillation visually in the time domain. In contrast, the oscillation from February 7 to 12 contains a mixture of at least three periods, and so the motion would appear less "organized," and we will not attempt to see this motion visually. We have therefore plotted the wind vectors for the period from January 29 to February 2 in Figure 12. It is clear that within periods of just under 12 hours the wind vector traces out fairly well defined ellipses. The time to complete an ellipse is typically in the range 8-12 hours. We cannot deduce the periods exactly, because of our coarse averaging interval of 2 hours, but the elliptical motion is apparent. The ellipses themselves show orientation changes over the 3 days, due to longer-

20 1814 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY (a) o /-' - -40, {,,,,{..,..,{...., Day in August 1999 (UT)... Meridional I mona, 120 / (b) 40 I... I... I Frequency (cycles/day).... I... I... I ,:...,.., (c) (d) "' " :... :--'r ;"'" ' " '½... F'"'r'"r' '""r'""' '... ';"r" ':'"" '" ' i "'r'",' 'vr'"' "." '"'.... 'i Day in August 1999 Window 0 Power Spectrum [0... 6'.0 Frequency (cycles/day) Figure 11. Same as Figure 10, but for the period August 26-30, Vertical arrows in Figure 11a highlight the approximate positions of the maxima in the zonal winds. Again, as for Figure 10b, the thin dashed line in Figure lib near 0 cycles d -1 shows the window spectral function from Figure lid on a linear power scale. The 95% confidence level for the hypothesis that any spectral peak lies above the noise level is shown. The mean noise level is also shown. period oscillations. Within the time interval shown, 4.4. A 15-hour Wave over 87% of the data showed this organized elliptical The other event which we wish to especially highclockwise rotation, and so this is not a statistical light is the spectral peak in Figure 5b during late artifact. This regular periodic elliptical motion at November with a period close to 15 hours. This is periods of 8-12 hours reinforces our earlier conclu- shown along line "E" in the figure as the secon dark sion that there is a periodic motion in the data, with a period which alters slightly with time as the local resonance conditions change. region from the top of the figure. We have spectrally analyzed the periods prior to, during, and "after" the time interval from November

21 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY Jan m/s - 30 Jan 0700 T1 30/030.0_ /0500 ' ]- -- ' o/o7oo - ( hi/s---_>. C_... J '29/ m/s 0/0100 [29/ Jan Jan 2100 T2 I [ I t -50 m/s 50 m/s 30/1700t1 t' ',, T 10-12h -50 nvs T 6-8h -50 m/s 3 q3oa, -01Feb Jan T i, 31/15øø Jan T n s -31Jan 1900! I [ 31/0901 ø I t I I 31/2 " "' ns_ t- 31/1900 ( / I 31/1900 ' // I; ß 7 T--.,10-12h -50 nvs T---,10-12h 40 nvs 01 Feb n s - 02 F /030 0_ I o oo - 0nVs " ',: " 0m/s 02 Feb Feb 1500 T6-50 m/s 0, m/s /'t I.,' so m/ ' 02/0900 T 8-10h s T 8h -50 s Figure 12. Time sequences of the wind vector in the period January 29 to February 2, 1999, at 86-km altitude. Wind vectors are aligned from the origin to the solid circles in each figure. Clockwis elliptical vector rotation, with periods in the range of typically 8-12 hours, is evident. This organized sense of rotation was evident during 87% of the entire period from January 29 to February 2 (inclusive). Times are given in UT. 16 to 30, The results of this spectral processing performed our spectral analysis on the period prior to are shown in Figure 13. Note that as discussed in November 16, we actually calculated spectra sepasection 4.1, the "December" data are really from rately for the periods November 1-4 and and Unfortunately, there was a data gap from averaged them. As a result, the resolution is some- November 5 to November 10, so when we have what coarser than for the other two spectra (Novem-

22 1816 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY h 5.0h 7.5h 10h -36h 18h?54jh h( lh.} h = = 200 -,,/1,, ] 111f' 6-30'1999'/ { IJ.*I"[* [ Dec 1-10, > 00 - } Noise level Frequency (cycles/day) Figure 13. Three sets of spectra for the period November 1-30, 1999, and December 1-10, The solid line without shading is plotted at low resolution and is the average spectrum from November 1 to 4 and 10 to 14. This was done due to a data gap from November 5 to 10. The other two spectra (November and December 1-10) are plotted at higher resolution. Bands of enhanced spectral activity are indicated. A particularly strong peak at 15 hours is evident between November 16 and 30. The 95% confidence level for the hypothesis that any spectral peak in the higher-resolution data lies above the noise level is shown. The mean noise level is also shown. ber 16, 1999, and December 1-10, 1998). However, it is clear that the "15-hour" oscillation is confined to the period between November 16 and 30, 1999, and we will concentrate on this interval. (A later check of the data for December 1999 showed that the oscilla- tion did not continue into that month.) Various spectral bands are highlighted in Figure 5b, and there are several which are statistically significant. However, the dominant oscillation is clearly one with a period of very close to 15 hours. There are also peaks at negative frequencies with periods of 7.5 and 5 hours. Although not statistically significant at the 95% level, they do both stand out above the noise and may well be real. We emphasize that they are not an artifact of the analysis, since the sidelobes of the window spectrum are much lower than the heights of these spectralines at displacements this far from the main peak. Thus, if they are indeed sidelobes, they are produced by physical distortions of the 15-hour oscillations and not due to artifacts of the analysis. Because the 15-hour oscillation is so dominant, it affords us the opportunity to study the event in more detail. We can even investigate the structure of the oscillation as a function of height. We do this in section Superposed Epoch Analysis of the 15-hour Wave Because the 15-hour wave is well defined, we can study it using a superposed epoch approach. In order to do this, we adopt the following technique. For each meteor we determine the number of hours N from the start of the year at which it occurred and then divide by the proposed period T of our wave: in this case, T = 15 hours. We then take the first whole multiple of T which is less than N and subtract this multiple from N. Thus we obtain N mudulo T. We then ascribe this hour to this meteor and change the effective time of occurrence accordingly. We then have all meteors ascribed to an hour in the range 1-T. We thus refer to this as a "superposed epoch," since all data have been superposed onto the same time interval (or epoch). This procedure has been used many times in the past by various authors. We then analyze the data as described in section 3.1, applying our standard all-sky wind determinations to

23 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY O lo -lo -30 2O lo o o ---" ;*',, 98 km o ]...'./">x. 94 km. -30 I''' I lo, o 91. -lo o,, I ß o 1./.. v 88 - o o o - 20 '1 o lo - o Time (hours) Time (hours) Figure 14. Results of a "superposed epoch" analysis on the meteor data for the period November 16-30, Data have been binned according to a 15.1-hour cycle and analyzed in a manner similar to that which was applied for tidal studies in Figures 2-4. Note that the ordinates use different scalings at each height, in order to best illustrate the temporal variation. Maximum westward and southward winds (i.e., the most negative values) for the first harmonic are shown as shaded squares, and the grey lines transgressing the graphs shows best fit lines to these squares. The vertical wavelengths may be deduced from these lines. However, such calculations assume that the oscillation has the same frequency at all heights and all times. This may not be true (see Figure 15). each hourly segment, at heights of 82, 85, 88, 91, 94, and 98 km. We may therefore determine a "typical wind variation" over this period. If the true period of the wave exactly matches the assumed period T, we should see a fairly strong sinusoidal oscillation with period T, and we may then apply standard harmonic fitting processes to determine the amplitude and phase of this wave. In reality, we apply the above process using our timing to an accuracy of a second, and times are stored as real numbers, not integers. In this way, we can allow the period T to be a noninteger. Results of this process for an assumed period of 15.1 hours are shown in Figure 14. There are clear sinusoidal oscillations at heights of 85, 88, and 98 km in the zonal component and 85 km in the meridional component.

24 1818 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY The results of fitting a first plus a second harmonic are shown as the dashed lines, and the shaded squares show the position of the minima of the first harmonic in each case. (We have used the minima because they occur in the middle of the graphs, whereas the maxima are close to the abscissa and so are more which means that it has a real second-order harmonic. Thus the rejection of data on the basis of a large second harmonic must be applied cautiously. We begin by discussing the data for 85-km altitude. Both the zonal and meridional components peak in the shaded region between 14.7 and 15.2 hours. The zonal maximum seems to be at hours, and the meridional maximum seems more diffuse, being spread out between and 15.2 hours. In both cases, the signal is also quite pure in the shaded awkward to display.) The solid grey lines passing upward between the graphs show lines which have been determined by a least squares fit through these shaded squares and therefore show the phase progression with height. If we were to accept these values, then we see that the wave has rather slow range, as can be seen by the fact that the symbols are solid, rather than open. Shaded regions correspond to periods where the first harmonic is dominant in either the zonal or meridional component. At 82 km the zonal component shows a broad phase progression with height, being almost evanes- maximum in the range hours, and the cent for the meridional component. The vertical wavelength for the zonal component is over 50 km. signal is also moderately pure. The meridional component shows no maximum of note and very often has However, we need to consider the period of the a substantial second harmonic. At 88 km the meridiwave more carefully. Figure 14 assumed a period of 15.1 hours, which was a somewhat arbitrary choice. We simply chose a period close to 15 hours for illustrative purposes. However, this may not be the true period. We therefore decided to repeat the above process for a variety of periods from 14.4 to onal component is quite strong at hours, although the second harmonic is quite strong too. However, further investigations have shown that these data only just failed the criterion for purity, and in fact, the second harmonic is only slightly above 25% of the first harmonic in power. Therefore we 15.5 hours, in steps of 0.05 hours. For each calculation consider that the band of meridional maxima at we then applied a harmonic fit and found the ampli- periods of hours is real. No clear maxima tudes and phases of both the first- and second-order are evident at 91 and 94 km, but at 98 km there seems harmonics. Figure 15 shows the amplitudes of the first harmonic produced by this process, as a function of time, for the six heights involved. Bear in mind that to be enhanced amplitudes in the region hours in both components. Thus, although the tests are limited by the presence the mean noise level increases with height, because of of other oscillations which contribute to the noise the increase of wave activity with height. Our purpose is to find maxima in wave activity by searching for maxima in the amplitude of the first harmonic. However, we also prefer that the signal be moderately "clean," with weak second harmonics. We have distinguished cases where the second harmonic has less than half the amplitude (25% of the power) of the first harmonic by using solid symbols. Therefore, in Figure 15, squares refer to zonal data, with solid level, there seems to be evidence in these superposed epoch fits of an oscillation with a period of close to 15 hours but which varies slightly in period with height. The fact that the maxima are not precisely defined also suggests that the true period may show some temporal variability in the time between November 16 and 30, Detailed studies of the running power spectra, using shorter window lengths (not shown), do indeed show suggestions of small variations in period squares being cases where the signal is moderately pure and open squares being cases where the second as a function of time. harmonic has over 25% of the power of the first 4.6. The 15-hour Wave: Discussion harmonic. Circles are used to represent the meridi- Periods at which there seems to be evidence of onal component, and again, solid circles refer to cases strong 15-hour activity in either the zonal or meridiwhere the first harmonic exceeds the second by over a factor of 4. It should be borne in mind, however, that the existence of a second harmonic cannot always be taken as cause for rejection, because if the wave is nonlinear, it may, in fact, have a nonsinusoidal form onal component are shaded in Figure 15. At 82 km it is the zonal component which seems to contain this activity, at 85 km both component seem to show it, and at 88 km the activity is mainly in the meridional component. Then there seems to be a gap, and a new occurrence of the wave at 98-km altitude. The cause of the wave is uncertain, and it may have propagated from below or above, or have been locally

25 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY i!... ' " i... i i... >-'- ' :,..- ::--:.o -="-'-:... [--'-'-- ½... i... i---:- '-' i! km 201!...? q... :...?... :... :;-'-'-F-:-...?... 'r... :... i... i--> i... ' 0 ' : o :ø 0 : : : O' :", :... '%, "... ;-'- " }... -"0:... t-..q... ;--> ' -- -, >/--:,--... ' I:i:}....,' :......,. _-,.....,....I ,:._._,-._ ::_,,:... I , i i i t i I i I i i km i : : '"" '-"" -'"' _x_., : : : : i i ß..-' a.--- [1' - '...-- :--4 -':---: ; : i :- i... '"'-' ' o.!" ' ' ' ',*'- ' ' ',"--o...- '",* 't! = [] -. _ i...- :' -...i,... 1,:-'-': ':::-0' x_ [... '. -';- '- "..."...0,... :-"-'-...' :"-'" :... I... ; ' ::':'_- '"-2,1t-... '...,,,, -o..-. '.... -< o. '<"'g.,--. i i i i i i i i i i i, km i 11iiiii::iii!: :' ii i::i::i::i::i ii:i:i!!:' ::-- ':: : :..!ii!i!::i::: i::f&.:: i:'--:----"--' :: ::!i ili: ::ii: ii!..-.'-"..i..-' i iiii... i :-:- -.-":...- i ii: -'ii::i::ii-:.:.--:. iii ::iii::?:il i i... :..: i?. i.3 :: :.. i....:..:.. :?.. : :?..: :. :...5 :..*:? : :.: :.?...` g i. : } :.: *: :: i: ii:... :... :... i... i... ]... ]... O',.- - '--,,... ß =::f.ii:.4.:i{{..::: -:."ilf:.:.."::-l..'-: :.:- {! iiii::i'! i-'-:-':.'"::":' O..' :..? :: :: :! :.? ' -'- -' ': :. : :.*.:. ::z.': " :'"' - '- ' "' - '-'- '- - - iii...:iii :...i {i::./;...5..i iiiiiii:...:ii `iiii iiiii.:...:ii i ::::i ii:::!i :::::.`. ' 5.':i?.': :.0 :.,.?..:-'<. :. 0:.{: : '- il' -./<' -/+-- i [] :... --[11""'"" : "= :"' " :,-, :..""'"' ""' '"' : ' :... i :... * [],..,. = --- :... "" %'*":':'-:? ' :----"g! *' '" '-:--.'-'"!'" - i!... "... [...?...,---::::: :-:i;: --:--- *... *..'...! km... i!... {... i... ""'-... i ß : - =-o,... _e..,:'-<... :.;..:.- i :-: :,...:,:,. ': ' : i 'i : ;: :' :- -:: -:: -: :: ::::7:-' :: -: :. ' ' : :: ::;: : : :: : :- ': ': : -.:,: :?...- e½: =: '.:*... -,,::---=.i L... :: ::; : :. i{ { : :i :...: :..::. :. i:ii :.:... 5 :, : x ---:--::s:..--::::..:s::...`. ::::.. :... :: :::...s :. : :...: : ` ::: *:.`.: : :?.?. : : ;...: :......:.:: : :.`: : : :.`::...::...: { :s. :: : * : :s. ' ' ,82 km,! o -:-!I,...--'---.., -'!,--. :,...:.?.:_..._ ' --.o.... '0 -...e '", :... la-...,... i... ; :... i i... i : :... i..., o, --: i...?::-ii:':;" :::-::' -::-2¾::: 'v"?"';" o... '... : ::: - : '".. ;:- 11:.-'8:::-'"" -:-:- :.:::- : -':... i... I T - '...?...,'"'... i' 'i, I " i i i i ' Period (hours) Figure 15. Results of applying superposed epoch analyses like that described in Figure 14, but varying the length of the epoch from 14.4 to 15.5 hours. Harmonic fits were then applied. The ordinateshow the velocity amplitudes of the first harmonic. Squares refer to the zonal component, and circles refer to the meridional component. Solid symbols with dark lines joining them refer to cases where the first harmonic contained more than 4 times the power of the second harmonic, while open symbols and fainter lines are used in cases which did not satisfy this condition. Bands where the fitting process produced maximum amplitudes in the first harmonic in at least one of the zonal or meridional components are shaded. Note that the noise level is larger at the higher altitudes, due to increased general activity at all periods at the upper heights.

26 1820 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY generated. Its strong dominance in the spectrum suggests that it has been at least amplified by local conditions in some manner. This local amplification was probably due to resonance in appropriate "cavities" defined by the background wind and temperature. Because the background conditions are continually changing, the "resonance" conditions adjust slightly both in time and in height, accounting for the small variability in wave period as a function of height, and possibly as a function of time. There seem to be two cavity regions: one between 82 and 88 km and one around 98 km. Presumably, conditions between these two regimes are not suitable to induce amplification and/or resonance. The wave also seems to contain some degree of nonlinearity, as evidenced by the existence of sidelobes at 7.5 and 5 hours in Figure 13 (second and third harmonics). It is interesting that while the main harmonic displays mainly anticlockwise rotation (Figure 13), the second and third harmonics 5. Conclusions seem to show clockwise rotation. Results of studies of the winds in the "meteor region" over Resolute Bay have been presented. Key findings include the detection of significant discrepancies with the CIRA model, particularly with regard to the summertime reversal and the occurrence of a large upper level wind maximum in summer. The zonal reversal is weaker than the model predicts, but we expect it to be much stronger below 80 km (i.e., at lower heights than the model predicts). The meridional component shows only a weakening, and the full reversal is expected below 80-kin altitude. The upper level jet is not predicted by the CIRA model but is quite strong in the measurements. Evidence that this wind speed enhancement occurs at other northern latitudes has also been presented. With regard to tidal studies we have found generally good agreement with the GSWM, except for the semidiurnal tidal phases in summer and the occurrence of a local maximum in activity in September which is not seen in the model. Several other dominant wavelike oscillations occur throughout the year, particularly at periods of 9-12 and 15 hours, and these may be due to modes being amplified or resonantly excited at these heights. Some of the characteristics of these waves show evidence of changes with height and time, as variations in the background conditions alter the resonance conditions. Acknowledgments. Special thanks go to Anna Hocking for help with data analysis and graphical presentation. Discussions with R. Walterscheid were also very useful. The radar was built with assistance from grants from the Natural Sciences and Engineering Research Council of Canada. References Avery, S. K., A. R. Riddle, and B. B. Balsley, The Poker Flat, Alaska, MST radar as a meteor radar, Radio Sci., 18, , Avery, S. K., R. A. Vincent, A. Phillips, A. H. Manson, and G. J. Fraser, High-latitude tidal behaviour in the meso- sphere and lower thermosphere, J. Atmos. Terr. Phys., 51, , Chang, J. L., S. K. Avery, and R. A. Vincent, New narrowbeam meteor radar results at Christmas Island: Implications for diurnal wind estimation, Radio Sci., 34, , Collins, R. L., D.C. Senft, and C. S. Gardner, Observations of a 12 h wave in the mesopause region at the South Pole, Geophys. Res. Lett., 19, 57-60, Espy, P. J., and G. Witt, Observation of a quasi 16-day oscillation in the polar summer mesospheric temperature, Geophys. Res. Lett., 23, , Farley, D. T., On-line data processing techniques for MST radars, Radio Sci., 20, , Fisher, G. M., T. L. Killeen, Q. Wu, and P. B. Hays, Tidal variability of the geomagnetic polar cap mesopause above Resolute Bay, Geophys. Res. Lett., 26, , Fleming, E. L., S. Chandra, M. R. Schoeberl, and J. J. Barnett, Monthly mean global climatology of temperature, wind, geopotential height and pressure for km, NASA Tech. Memo., , Fleming, E. L., S. Chandra, J. J. Barnett, and M. Corney, Zonal mean temperature, pressure, zonal wind and geopotential height as functions of latitude, Adv. Space Res., 10(12), 11-62, Forbes, J. M., N. A. Makarov, and Y. I. Portnyagin, First results from the meteor radar at South Pole: A large 12-hour oscillation with zonal wavenumber one, Geophys. Res. Lett., 22, , Forbes, J. M., S. E. Palo, X. Zhang, Y. I. Portnyagin, N. A. Makarov, and E.G. Merzlyakov, Lamb waves in the lower thermosphere: Observational evidence and global consequences, J. Geophys. Res., 104, 17,107-17,116, 1999a. Forbes, J. M., Y. I. Portnyagin, N. A. Makarov, S. E. Palo, E.G. Merzlyakov, and X. Zhang, Dynamics of the lower thermosphere over South Pole from meteor radar wind measurements, Earth Planets Space, 51, , 1999b. Fraser, G. J., Summer circulation in the Antarctic middle atmosphere, J. Atmos. Terr. Phys., 46, , Fraser, G. J., Monthly mean winds in the mesosphere at 44S and 78S, Pure Appl. Geophys., 130, , Fraser, G. J., G. Hernandez, and R. W. Smith, Eastward-

27 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY 1821 moving 2-4 day waves in the winter Antarctic mesosphere, Geophys. Res. Lett., 20, , Golomb, S. W., and R. A. Scholtz, Generalized Barker sequences, IEEE Trans. Inf Theory, IT-11, , Gurubaran, S., and R. Rajaram, Signatures of equatorial electrojet in the mesospheric partial reflection drifts over magnetic equator, Geophys. Res. Lett., 27, , Hagan, M. E., J. M. Forbes, and F. Vial, A numerical investigation of the propagation of the quasi 2-day wave into the lower thermosphere, J. Geophys. Res., 98, 23,193-23,205, Hagan, M. E., J. M. Forbes, and F. Vial, On modeling migrating solar tides, Geophys. Res. Lett., 22, , Hagan, M. E., D. M. Burrage, J. M. Forbes, J. Hackney, W. J. Randel, and X. Zhang, GSWM-98: Results for migrating solar tides, J. Geophys. Res., 104, , Hall, C. M., A. H. Manson, and C. E. Meek, Spectral characteristics of spring Arctic mesosphere dynamics, Ann. Geophys., 16, , Hernandez, G., R. W. Smith, R. G. Roble, J. Gress, and K. C. Clark, Thermospheric dynamics at the South Pole, Geophys. Res. Lett., 17, , Hernandez, G., R. W. Smith, and J. F. Conner, Neutral wind and temperature in the upper mesosphere above South Pole, Antarctica, Geophys. Res. Lett., 19, 53-56, 1992a. Hernandez, G., R. W. Smith, G. J. Fraser, and W. L. Jones, Large-scale waves in the upper-mesosphere at Antarctic high-latitudes, Geophys. Res. Lett., 19, , 1992b. Hernandez, G., G. J. Fraser, and R. W. Smith, Mesospheric 12-hour oscillation near South Pole, Antarctica, Geophys. Res. Lett., 20, , Hernandez, G., J. M. Forbes, R. W. Smith, Y. Portnyagin, J. F. Booth, and N. Makarov, Simultaneous mesospheric wind measurements near South Pole by optical and meteor radar methods, Geophys. Res. Lett., 23, , Hines, C. O., G. W. Adams, J. W. Brosnahan, F. T. Djuth, M.P. Sulzer, C. A. Tepley, and J. S. Van Baelen, Multi-instrument observations of mesospheric motions over Arecibo: Comparisons and interpretations, J. Atmos. Terr. Phys., 55, , Hocking, W. K., System design, signal processing procedures and preliminary results for the Canadian (London, Ontario) VHF atmospheric radar, Radio Sci., 32, , 1997a. Hocking, W. K., Strengths and limitations for MST radar measurements of middle atmosphere winds, Ann. Geophys., 15, , 1997b. Hocking, W. K., Temperatures using radar-meteor decay times, Geophys. Res. Lett., 26, , Hocking, W. K., and T. Thayaparan, Simultaneous and colocated observation of winds and tides by MF and meteor radars over London, Canada (43øN, 81øW), during , Radio Sci., 32, , Hocking, W. K., T. Thayaparan, and J. Jones, Meteor decay times and their use in determining a diagnostic mesospheric temperature-pressure parameter: Methodology and one year of data, Geophys. Res. Lett., 24, , Hocking, W. K., B. Fuller, and B. Vandepeer, Real-time determination of meteor-related parameters ultilzing modern digital technology, J. Atmos. Sol. Terr. Phys., 63, , Hocking, W. K., M. C. Kelley, R. Rogers, W. Brown, D. Moorcroft, and J.-P. St. Maurice, Resolute Bay VHF radar: A multipurpose tool for studies of tropospheric motions, middle atmosphere dynamics, meteor physics, and ionospheric physics, Radio Sci., this issue. Holton, J. R., The role of gravity wave induced drag and diffusion in the momentum budget of the mesosphere, J. Atmos. Sci., 39, , Holton, J. R., The influence of gravity wave breaking on the general circulation of the middle atmosphere, J. Atmos. Sci., 40, , Jones, J., and P. Brown, Sporadic meteor radiant distributions: Orbital survey results, Mon. Not. R. Astron. Soc., 265, , Jones, J., and P. Brown, The radiant distribution of sporadic meteors, Planet. Space Sci., 42, , Jones, J., and J. D. Morton, A method for imaging radio meteor radiant distributions, Mon. Not. R. Astron. Soc., 200, 281, Lieberman, R. S., et al., HRDI observations of mean meridional winds at solstice, J. Atmos. Sci., 55, , Lindzen, R. S., Turbulence and stress owing to gravity wave and tidal breakdown, J. Geophys. Res., 86, , Lomb, N. R., Least squares frequency analysis of unequally spaced data, Astrophys. Space Phys., 39, , Longuet-Higgins, M. S., The eigenfunctions of Laplace's tidal equations over a sphere, Philos. Trans. R. Soc. London, 262, , Liibken, F.-J., and U. von Zahn, Thermal structure of the mesopause region at polar latitudes, J. Geophys. Res., 96, 20,841-20,857, MacLeod, R., and R. A. Vincent, Observations in the Antarctic summer mesosphere using the spaced antenna technique, J. Atmos. Terr. Phys., 47, , Manson, A. H., and C. E. Meek, Climatologies of mean winds and tides observed by medium frequency radars at Tromso (70øN) and Saskatoon (52øN) during , Can. J. Phys., 69, , Manson, A. H., et al., Description and presentation of reference atmosphere radar winds ( km), Adv. Space Res., 10(12), , 1990a. Manson, A. H., C. E. Meek, T. L. Hansen, and T. Trondsen,

28 1822 HOCKING: MIDDLE ATMOSPHERE DYNAMICS AT RESOLUTE BAY Dynamics of the upper middle atmosphere ( km) at Tromso (70øN) and Saskatoon (52øN), June-December 1987, using the Tromso and Saskatoon MF radars, J. Atmos. Terr. Phys., 52, , 1990b. Manson, A. H., et al., Seasonal variations of the semi- Portnyagin, Y. I., J. M. Forbes, and N. A. Makarov, Unusual characteristics of lower thermosphere prevailing winds at South Pole, Geophys. Res. Lett., 24, 81-84, Portnyagin, Y. I., J. M. Forbes, N. A. Makarov, E.G. Merzlyakov, and S. E. Palo, The summertime 12-hour diurnal and diurnal tides in the MLT: Multi-year MF wind oscillation with zonal wavenumber s = 1 in the radar observations from 2 to 70øN, and the GSWM tidal model, J. Atmos. Sol. Terr. Phys., 61, , Massebeuf, M., R. Bernard, J. L. Fellous, and M. Glass, The mean zonal circulation in the meteor zone above Garchy (France) and Kiruna (Sweden), J. Atmos. Terr. Phys., 41, , Oppenheim, M. M., A. F. vom Endt, and P. Dyrud, Electrodynamics of meteor trail evolution in the equatorial E-region ionosphere, Geophys. Res. Lett., 27, , Oznovich, I., D. J. McEwan, G. G. Sivjee, and R. L. Walterscheid, Tidal oscillations of the Arctic upper mesosphere and lower thermosphere in winter, J. Geophys. Res., 102, , Palo, S. E., Y. I. Portnyagin, J. M. Forbes, N. A. Makarov, and E.G. Merzlyakov, Transient eastward-propagating long-period waves observed over South Pole, Ann. Geophys., 16, , Phillips, A., Simultaneous observation of the quasi 2-day wave at Mawson, Antarctica, and Adelaide, South Australia, J. Atmos. Terr. Phys., 51, , Phillips, A., and R. A. Vincent, Radar observations of prevailing winds and waves in the Southern Hemisphere mesosphere and lower thermosphere, Pure Appl. Geolower thermosphere over the South Pole, Ann. Geophys., 16, , Riggin, D. M., D.C. Fritts, M. J. Jarvis, and G. O. L. Jones, Spatial structure of the 12-hour wave in the Antarctic as observed by radar, Earth Planets Space, 51, , Sivjee, G. G., and R. L. Walterscheid, Six-hour zonally symmetric tidal oscillations of the winter mesopause over the South Pole Station, Planet. Space Sci., 42, , Sivjee, G. G., R. L. Walterscheid, and D. J. McEwan, Planetary wave disturbances in the Arctic winter mesopause over Eureka (80øN), Planet. Space Sci., 42, , Virdi, T. S., and P. J. S. Williams, Altitude variations in the amplitude and phase of tidal oscillations at high latitudes, J. Atmos. Terr. Phys., 55, , Walterscheid, R. L., and G. G. Sivjee, Very high frequency tides observed in the airglow over Eureka (80øN), Geophys. Res. Lett., 23, , Zhan, Q., A. H. Manson, and C. E. Meek, The impact of gaps and spectral methods on the spectral slope of the middle atmosphere wind, J. Atmos. Sol. Terr. Phys., 58, , phys., 130, , Portnyagin, Y. I., The climatic wind regime in the lower thermosphere from meteor radar observations, J. Atmos. Terr. Phys., 48, , Portnyagin, Y. I., J. M. Forbes, G. J. Fraser, R. A. Vincent, S. K. Avery, I. A. Lysenko, and N. A. Makarov, Dynamics of the Antarctic and Arctic mesosphere and lower thermosphere regions, I, The prevailing wind, J. Atmos. Terr. Phys., 55, , W. K. Hocking, Department of Physics and Astronomy, University of Western Ontario, Richmond Street NTH, London, Ontario, Canada N6A 3K7. (whocking@ julian.uwo.ca) (Received March 8, 2000; revised May 4, 2001; accepted May 8, 2001.)