Vertical ion motion observed with incoherent scatter radars in the polar lower ionosphere

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 110,, doi: /2004ja010705, 2005 Vertical ion motion observed with incoherent scatter radars in the polar lower ionosphere S. Oyama, 1 B. J. Watkins, 1 S. Nozawa, 2 S. Maeda, 3 and M. Conde 1,4 Received 27 July 2004; revised 13 January 2005; accepted 25 January 2005; published 16 April [1] Vertical ion velocities in excess of a few tens m s 1 have been found in the lower ionosphere ( km) at high latitudes. Statistics using data from the European Incoherent Scatter (EISCAT) Tromsø UHF radar are performed to understand general character of vertical motions in the lower ionosphere for geomagnetically quiet conditions. We estimate the probability distribution function for the real vertical ion velocity, taking into account the measurement uncertainty. This calculation suggests that while most vertical ion velocities have amplitudes smaller than 20 m s 1 at 100 km level, the magnitudes larger than 20 m s 1 are observed as statistical exception. To understand if these large vertical motions in the lower ionosphere are associated with thermospheric motions, differences between ion and neutral wind velocities are calculated using data from experiments with the Sondrestrom incoherent scatter radar on 5 and 12 September The vertical ion velocities show magnitudes larger than 20 m s 1 below 103 km where the ion velocity is considered to be equal to the neutral wind velocity during the experiments according to calculations of the velocity difference. Citation: Oyama, S., B. J. Watkins, S. Nozawa, S. Maeda, and M. Conde (2005), Vertical ion motion observed with incoherent scatter radars in the polar lower ionosphere, J. Geophys. Res., 110,, doi: /2004ja Introduction [2] The coupled thermosphere-ionosphere system at high latitudes exhibits complicated temporal and spatial variations due to, for example, ion drag and frictional heating. While our knowledge of this system has been extended during the last 2 decades, there remain various unexplained but interesting phenomena in the polar ionosphere. One such phenomenon is the vertical motion in the lower ionosphere and thermosphere ( km). [3] A limited number of researchers has reported lower thermospheric vertical winds observed with Fabry Perot Interferometers (FPI) (l = nm) [Price and Jacka, 1991; Price et al., 1995; Smith and Hernandez, 1995; Ishii et al., 1999] and incoherent scatter (IS) radar [Kofman et al., 1996]. These results indicate that there is an unexpected level of small-scale structure in the polar thermospheric dynamics with large amplitudes. Price et al. [1995] measured vertical winds at Poker Flat, Alaska, with an FPI at wavelengths of and nm, corresponding to principal emission altitudes around 110 and 240 km, respectively. On one occasion, large upward winds reaching 42 m s 1 (using the nm auroral emission) and 138 m s 1 (using the nm auroral emission) were 1 Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska, USA. 2 Solar-Terrestrial Environment Laboratory, Nagoya University, Nagoya, Japan. 3 Kyoto Women s University, Kyoto, Japan. 4 Now at La Trobe University, Victoria, Australia. Copyright 2005 by the American Geophysical Union /05/2004JA measured simultaneously in association with enhanced auroral activity. This event might lead us to an explanation where the entire thermosphere from E to F region heights is lifted up simultaneously due to thermal expansion associated with enhancement of particle and Joule heating. It must be, however, emphasized that the total number of observations supporting this hypothesis remains small; for example, vertical winds observed with an FPI simultaneously at both wavelengths show not only correlated but also anticorrelated variations [Ishii et al., 1999]. The response of the polar thermosphere against geomagnetic activities often appears chaotic, so it is difficult to capture a complete picture of the behavior using limited measurements. We also need more understanding on general and long timescale variations in the lower thermospheric/ionospheric vertical motions even for understanding the small-scale structures. [4] This paper addresses statistics of the vertical ion velocity measured with the European Incoherent Scatter (EISCAT) Tromsø UHF radar (69 N, 19 E, 66 magnetic latitude). The statistical results are supported using data from special experiments with the Sondrestrom IS radar (67 N, 309 E, 73 magnetic latitude) on 5 and 12 September The next section discusses our approach to collecting the data. Section 3 presents results from the statistics. Discussions are given in section 4, including results from the special experiments with the Sondrestrom IS radar. A summary and conclusion complete the paper in section Data and Analysis Method for Statistics [5] Data sets from the Tromsø UHF radar operating in CP-2 mode (four-position observation including geographic zenith) during geomagnetically quiet periods (daily 1of14

2 Table 1. List of Tromsø UHF Radar Data Sets Used in This Paper Together With Daily Ap, Kp, and F 10.7 Indexes Date, YYMMDD Daily Ap Kp F Ap < 10) are statistically analyzed as tabulated in Table 1. In this paper we use only data from the geographically vertical beam. The vertical measurement is conducted every 6 min. The height resolution for data used here is 3 km. Since the thermospheric dynamics depend on solar activity and season, data must be collected for the same solar activity and seasonal conditions. Here we select the near solar maximum summer condition. [6] Figure 1 shows an example of time series of the vertical ion velocity for CP-2 experiment, which was conducted on June The temporal variations from 95 to 111 km do not show notable long timescale oscillations such as tidal variations, but vertical ion velocities at 108 and 111 km seem to have positive trends for a few hours around 1600 UT. Vertical ion velocities in this height region show fluctuations at high frequencies of a few tens of minutes. While these amplitudes appear to be larger at local night than at local daytime, it should be noted that data quality at local night is usually lower than at local daytime due to lower ionization at E region heights. Statistics in this paper have taken this day/night difference into account. [7] We use data sets covering a 24-hour interval to reduce temporal dispersion of the data count and collect six successive segments and stable hourly data count. Since four of six segments are obtained during one successive run from 1 to 5 July 1999, statistical results in this paper may have a sampling bias. This data restriction may not allow us to regard the statistical results as representative for near solar maximum summer conditions. [8] We have developed a statistical analysis using a weighting function from the measurement uncertainty. The measurement uncertainty g i, being a value associated with data X i, is derived with GUISDAP (Grand Unified Incoherent Scatter Design and Analysis Package) (available at which is employed on IS spectra to derive ionospheric parameters such as the electron density, ion and electron temperatures, and line-of-sight ion velocity. This method can be expressed with equation (1) where N is the gross data number to be used for the statistics [Press et al., 1992]: 96 to 111 km. From 96 to 105 km, while the magnitude of the hourly mean velocity calculated is smaller or comparable to the hourly mean measurement uncertainty, the statistical standard deviation is significantly larger than the magnitude of the mean velocity. At 108 and 111 km, the mean velocity from 1400 to 2300 UT shows variations from upward to downward but the amplitude is smaller than the mean measurement uncertainty. The magnitudes of the mean velocity at other UTs are relatively small. The mean measurement uncertainty and the statistical standard deviation have daily variations in this height region; daytime values are slightly smaller than nighttime ones, which can be seen clearly at lower heights. [10] Figure 3 presents histograms of vertical ion velocity measured from 96 to 111 km heights using same data sets for Figure 2. The gross data count, shown in the upper right corner at each height, increases with height from 826 to 1359 because of the increase in the electron density or the signal-to-noise ratio. Histograms look symmetric to approximately 0 m s 1 and show that some measurements exceed 50 m s 1. The statistical standard deviation shown in the parentheses appears to be slightly larger at upper and lower heights, although these differences may not be significant according to the hourly mean measurement uncertainty shown in Figure 2. [11] To discuss the occurrence of relatively large vertical ion velocities, we calculate the ratio of vertical ion velocity in excess of 20 m s 1 as shown in Figure 3 (see upper left corner at each height). Below 105 km, the ratio is 0.14 with respect to the gross measurement and slightly increases with height, reaching 0.20 at 111 km. These results suggest that the Tromsø UHF radar observes upward ion velocities larger than 20 m s 1 for more than 20 hours of six 24-hour segments or 144 hours in the lower ionosphere. X ¼ X N i¼1, X XN i 1 g 2 i¼1 i g 2 i : ð1þ 3. Statistical Results Using Tromsø UHF Radar Data 3.1. Temporal Variations and Histograms Using All Available Data Sets [9] Figure 2 shows temporal variations in the vertical ion velocity from Tromsø UHF radar data over the height range Figure 1. Temporal variations in the vertical ion velocity measured with the Tromsø UHF radar on June 1993 from 95 to 111 km. Positive velocity is geographically upward. 2of14

3 [15] At local night, a large portion of the observation interval has vertical ion velocities larger than 20 m s 1,as shown in Figure 4a. These large velocities, however, may not be always significant because they are comparable or smaller than the measurement uncertainty. On the other hand, since the ratio of measurement uncertainty larger than 20 m s 1 is very low at local daytime, vertical ion velocities larger than 20 m s 1 at local daytime are considered to be reliable ionospheric phenomena. [16] We have made histograms of the vertical ion velocity using one of 24-hour data sets in order to investigate the effects of daily ionospheric variations on histograms. Histograms on 7 July 1992 (not shown here) present that the statistical standard deviation at local daytime is smaller than that at local night, and the vertical ion velocity appears to be symmetric to 0 m s 1, although the symmetry is less apparent compared with histograms shown in Figure 4a because of a smaller amount of individual measurements. Histograms on 15 June 1993 also show similar features (not shown here). This comparison suggests that statistical features shown in Figures 2, 3, Figure 2. Temporal variations in the hourly mean vertical ion velocity observed with the Tromsø UHF radar from 96 to 111 km heights (thick solid curve). The vertical bar presents two statistical standard deviations (that is, ±1s) at each 1 hour. The hourly mean measurement uncertainty is illustrated with thin solid curve (also ±1s). Positive value is geographically upward. [12] The histogram shown in Figure 3 is broadened with respect to the real histogram due to the measurement uncertainty, which has daily variations, as shown in Figure 2. In the next section, we present two kinds of histogram to investigate the day-night dependence Day-Night Dependences of the Histogram [13] Figure 4a presents histograms of the vertical ion velocity observed at local night (from 2100 to 0300 UT) and local daytime (from 0900 to 1500 UT) from 96 to 111 km using same data sets for Figures 2 and 3. The data count available for this calculation at local night is smaller than that at local daytime because of daily variations in the signal-to-noise ratio. Of particular interest is the difference in the statistical standard deviation, larger at local night than at local daytime in this height region. Vertical ion velocities in excess of 20 m s 1 and downward ion velocities more frequently appear at local night than local daytime. The symmetry might be less apparent at local night below 105 km, but it is not apparent at daytime at 108 km. [14] Figure 4b presents histograms of the measurement uncertainty of vertical ion velocity at local night and daytime. Distributions obviously have dependences on local time; the measurement uncertainty at local night is more widely distributed than at local daytime. At local night, the ratio of measurement uncertainty larger than 20 m s 1 is higher than 0.3, although that at local daytime is negligible. Figure 3. Distribution of occurrences of the vertical ion velocity observed from 96 to 111 km with the Tromsø UHF radar. Each speed bin is 5 m s 1 wide. The number of individual data is shown at the upper-right corner at each height. The statistical standard deviation at each height is shown in the parentheses. Positive value is geographically upward. 3of14

4 Figure 4a. Same as Figure 3, but for local night (from 2100 to 0300 UT; left column) and local daytime (from 0900 to 1500 UT; right column). Vertical dotted lines show ±20 m s 1. and 4a 4b are independent, at least on the observation dates tabulated in Table Discussion 4.1. Estimate of the Real Distribution [17] In this section we estimate real distributions in the vertical ion velocity extracting information from histograms of the measured velocity and the measurement uncertainty assuming that their histograms are determined by the normal distribution function. Since the histogram at local daytime shows symmetry more clearly than that at local night, this assumption can be applied to the daytime case rather than the nighttime case. We thus show the daytime case here. [18] The normal distribution function may be written as follows: f ¼ 1 p s ffiffiffiffiffi exp 1 x m 2 ; ð2þ 2p 2 s where s is the statistical standard deviation, x is a data of velocity, and m is the mean value. If we know the normal distribution function for real velocity, f real, function for the observed velocity, f obs, may be expressed as a convolution between f real and f un (function to show data fluctuations due to the measurement uncertainty): Z 1 f obs ðþ¼ x f real ðaþf un ðx aþda: ð3þ 1 [19] We assume that s for f un is equal to the measurement uncertainty averaged over a selected time interval and that f un has a peak at a value of a. We calculate the probability distribution function (PDF) for observed data using equation (3); then the least squares method is applied to derive the best-fitted model function for observed velocity by changing s and m for f real. [20] Figure 5 shows PDF of the real and observed velocity at local daytime. For example, PDFs at 96 km have a peak at 1.2 m s 1 for both cases and have the standard deviation of and m s 1 for observed and real velocities, respectively. The larger standard deviation for the observed velocity is due to the measurement uncertainty. The PDF of the real velocity shows that a few percent of the observation interval is occupied with amplitudes of 15 m s 1 and there are almost no chances to measure velocities larger than 20 m s 1. This result suggests that amplitude smaller than 20 m s 1 can be measured as 4of14

5 Figure 4b. Same as Figure 4a except for the measurement uncertainty of vertical ion velocity. statistically general phenomena at 96 km and that amplitudes larger than 20 m s 1 are statistically exceptional cases. Note that results from PDF do not insist that amplitudes in excess of 20 m s 1 are meaningless because some largeamplitude events cannot be represented with the normal distribution function applied to the PDF calculation. [21] PDFs at other heights also show larger standard deviations for the observed case than the real case. The velocity at a peak PDF appears to shift slightly from negative to positive in association with gradual increase in the standard deviation with height. Below 105 km, amplitude larger than 20 m s 1 is considered to be the statistically exceptional case. At 108 and 111 km, PDFs of the real velocity show that a few percent of the observation period is occupied with amplitude of 20 m s 1, which is larger than the case below 105 km. These altitudinal variations may be relevant to attenuation of the E B drift with decrease of height and altitudinal dependences on the vertical amplitude of atmospheric tides. Future use of more abundant data sets to investigate seasonal or geomagnetic activity dependences may provide further understanding of these altitudinal variations in the vertical ion velocity Case Studies Using Sondrestrom IS Radar Data [22] While the statistical results show that most vertical ion velocities at 100 km level have amplitudes less than 20 m s 1, the magnitudes sometimes exceed 20 m s 1 even during geomagnetically quiet periods. Magnitudes over 20 m s 1 are considerably larger than are predicted by thermospheric models [Walterscheid et al., 1985; Sun et al., 1995; Shinagawa et al., 2003]. It is important to understand if these large vertical motions in the lower ionosphere are associated with thermospheric motions. The special experiments with the Sondrestrom IS radar on 5 and 12 September 2003 were conducted to determine the upper height where we can assume that the vertical ion velocity is equal to the vertical neutral wind velocity. This section discusses measurements of the vertical ion velocity in excess of 20 m s 1 even below the upper height Observation Mode and Geomagnetic Conditions [23] The special experiments were conducted from 1133 to 1957 UT and from 1132 to 1957 UT on 5 and 12 September 2003, respectively, with a beam-swinging method using short and long pulses, which provide the height resolution of 3 km in the E region and 42 km in the F region, respectively. Figure 6 shows a schematic drawing of the antenna position. It takes about 10 min for one sequence with 3 min integration time at individual three antenna positions. The distance between the geographic vertical beam and the beam a is a few ten kilometers at E region heights. [24] Figures 7 and 8 show temporal variations in H-component of the geomagnetic field measured with the Greenland Magnetometers of the Danish Meteorological Institute on 5 and 12 September 2003, respectively. The offset curve of the H-component is defined as the average over six geomagnetically quiet dates near 5 and 12 September 2003, namely 31 August, 2, 8, 15, 27, and 5of14

6 temperature at 100 and 103 km, the lower ionosphere for most of the experiment period is considered to be geomagnetically quiet. However, there are a few notable variations in these parameters. The electron density shows enhancements over the time period 1200 to 1300 UT due to hard particle precipitation. The ion temperature sometimes shows overturning, in particular after 1800 UT, in association with fluctuations in the electron density. The temperature difference between these two heights does not remain constant. The F region ion temperature, which can be used as a reference of the electric field enhancement, does not show significant enhancements during this experiment; thus no significant frictional heating is indicated. [26] Figures 9d 9f show temporal variations in the ion velocity measured with the vertical beam at 97, 100, and 103 km. The magnitudes sometimes exceed 20 m s 1, which are significantly larger than the measurement uncertainty in this height region. [27] The vertical ion velocity shows notable wavelike structures, which have large amplitudes after 1800 UT. Relationships of these wavelike structures among 97, 100, and 103 km are complicated. From 1600 to 2000 UT, these oscillations are almost in phase, but it is relatively difficult to investigate the phase relationship before 1600 UT because of small amplitudes. The ion velocity at 103 km from 1300 to 1500 UT remains approximately constant with an offset about +20 m s 1, although those at two other heights fluctuate. Amplitudes at 103 km appear to be larger than those at other heights from 1600 to 1800 UT; by contrast, Figure 5. PDF (probability distribution function) from 96 to 111 km at local daytime ( UT) for distributions of the real vertical ion velocity (solid curve with dots) and of the observed velocity (solid curve without dots). The mean value and the standard deviation are shown outside and inside of the parentheses, respectively. 18 September (daily Ap are 6, 12, 5, 4, 6, and 4, respectively). Here we show ten individual measurements along the west coast of Greenland. The Sondrestrom IS radar site is denoted as STF on the figures. The local geomagnetic activity at the Sondrestrom site appears to be quiet for most of the experiment periods, although some activity can be seen at stations distant from the radar site. Geomagnetic activity at the equatorward of Sondrestrom is also quiet. Notable enhanced geomagnetic activity is limited to the sites Qeqertarsuaq (GDH) and Uummannaq (UMQ) on 5 September. Enhancements of the latitudinal distribution of the ionospheric current on 5 September will be discussed later using the plasma convection map estimated with the Super Dual Auroral Radar Network (SuperDARN) radars Event of 5 September 2003 [25] Figure 9 shows temporal variations for the various parameters measured with the magnetometer and the IS radar at the Sondrestrom radar facility on 5 September The measurement uncertainty, which is determined through the IS spectrum fitting procedure with a theoretical function, for each IS radar parameter is plotted with ±1s or 68.27% confidence. Figure 9a shows temporal variations in the magnetic field. Since we cannot find large fluctuations in (Figure 9b) the electron density and (Figure 9c) the ion Figure 6. Schematic drawing of the antenna pattern for the Sondrestrom IS radar experiments on 5 and 12 September The antenna is directed to three positions; geographical vertical, along the magnetic field line (Azimuth: 140, Elevation: 80 ), and the other direction a (Azimuth: 140, Elevation: 70 ). The antenna stays at each position for 3 min and each data is derived every about 10 min. Values of D 1 (D 2 ) at 100, 150, and 250 km heights are 17.6, 26.4, and 44.1 (18.8, 28.2, and 46.9) km, respectively. 6of14

7 Figure 7. Temporal variations in H-component of the geomagnetic field measured on 5 September 2003 at THL (Qaanaaq; 77 N, 291 E, 85 geomagnetic latitude), UPN (Upernavik; 73 N, 304 E, 79 geomagnetic latitude), UMQ (Uummannaq; 71 N, 308 E, 77 geomagnetic latitude), GDH (Qeqertarsuaq; 69 N, 306 E, 76 geomagnetic latitude), ATU (Attu; 68 N, 306 E, 75 geomagnetic latitude), STF (Kangerlussuaq; 67 N, 309 E, 73 geomagnetic latitude), SKT (Maniitsoq; 65 N, 307 E, 72 geomagnetic latitude), GHB (Nuuk; 64 N, 308 E, 70 geomagnetic latitude), FHB (Paamiut; 62 N, 310 E, 68 geomagnetic latitude), and NAQ (Narsarsuaq; 61 N, 315 E, 66 geomagnetic latitude). Thick solid line with dots represents the special experiment interval. Kp and daily Ap values are illustrated around the horizontal axis. site. Since the relative velocity highly depends on the ion velocity, we assume that the geomagnetically meridional relative velocity does not exceed the zonal one. In this case, we can assume that the relative velocities calculated are regarded as maximum values for the meridional one. Since mean measurement uncertainty of the vertical ion velocity is 5ms 1, that of the meridional velocity might be 5 tan m s 1, taking into account the geomagnetic dip angle. Figure A1 in Appendix A shows that relative velocities below 103 km are smaller than 28 m s 1.On the other hand, those above 106 km sometimes have larger values. This result allows us to determine the upper boundary of 103 km for this observation interval Event of 12 September 2003 [30] Figure 10 shows the 12 September case. Temporal variations in the electron density and the ion temperature are analogous to those for the 5 September case. The electron densities at 100 and 103 km show enhancements around 1200 UT due to hard particle precipitation. The ion temperatures at not only those heights but also F region heights do not show notable enhancements. [31] Temporal variations in the vertical ion velocity also, on the whole, have similar features with the 5 September case. The magnitudes frequently exceed 20 m s 1 and appear more frequently after 1700 UT than before. The vertical ion velocities at 97 and 100 km after 1700 UT show large magnitudes with changing the sign frequently; by contrast the velocities at 103 km after 1700 UT oscillate at longer periods with slightly smaller amplitudes than the velocities at 97 and 100 km. While one can see various wavelike structures before 1700 UT, they seem to be out of amplitudes at 97 and 100 km are larger than those at 103 km from 1800 to 2000 UT. [28] The 15-s integrated IS spectra from 1830 to 1930 UT when large vertical ion speeds are observed have been checked. Since strange echoes were not seen during the period, we conclude that the ion velocities derived for the experiment represent real motions in the lower ionosphere. [29] In order to determine the upper boundary where the ion velocity is equal to the neutral one, we calculate the relative velocity in the y- or geomagnetically zonal component between ions and neutrals using observed data (see Appendix A). In the polar ionosphere, the geomagnetically zonal ion velocity accelerated by the E B force is usually larger than the meridional one except in the polar cap and the convection reversal region. Plasma convection maps estimated with the SuperDARN radar during the special experiment also support this relationship at the Sondrestrom Figure 8. Same as Figure 7 except for 12 September of14

8 Figure 9. Temporal variations in (a) deviation of H-component of the geomagnetic field at STF, (b) the electron density, (c) the ion temperature, and (d f) the vertical ion velocity in the lower ionosphere observed with the Sondrestrom IS radar on 5 September Solid and open circles in Figures 9b and 9c are data at 100 and 103 km, respectively. Positive velocity is geographically upward. Dotted lines are illustrated at ±20 m s 1. Vertical bars in Figures 9b 9f are the measurement uncertainty for each parameter with ±1s. phase in this height region. While the vertical ion velocity at 1400 UT at 100 km is 43 m s 1, thus downward velocity, the velocities at 97 and 103 km show considerably different values of 10 and 4 m s 1, respectively. [32] The relative velocity in the y- or geomagnetically zonal component between ions and neutrals are calculated with the same method in section 4.2. As is the same with the 5 September case, the relative velocities below 103 km are smaller than 28 m s 1. Thus we conclude that the measured ion velocity is equal to the neutral wind velocity below 103 km for this experiment interval Plausible Explanations on Generation of Large Vertical Motions in the Lower Ionosphere [33] Statistics of the vertical ion velocity measured with the Tromsø UHF radar show that magnitude of the hourly mean velocity calculated is smaller than the statistical standard deviation and the hourly mean measurement uncertainty (see Figure 2). PDF at local daytime shows that most measurements have amplitudes smaller than 20 m s 1 at 100 km level (see Figure 5), although there are some individual measurements in excess of 20 m s 1, as shown in Figure 4a. Some portions show large amplitudes in excess of 50 m s 1 for daytime when the measurement uncertainty scarcely exceeds 20 m s 1. [34] The special experiments with the Sondrestrom IS radar suggest that the thermosphere below 103 km enables it to fluctuate vertically with amplitude of a few tens m s 1. The special experiments have measured the vertical ion velocity in excess of 20 m s 1 at higher ratio than the statistical value of 0.14; for the 5 September case, 0.20, 0.20, and 0.42 at 97, 100, and 103 km, respectively, and for the 12 September case, 0.25, 0.24, and 0.36, respectively. Upward and downward velocities with amplitudes in excess of 50 m s 1 have been measured for both experiments without coincidence of notable Joule heating events. The mean velocity during the experiments is significantly larger than the statistical results; for the 5 September case, 7, 11, and 16 m s 1 at 97, 100, and 103 km, respectively, and for the 12 September case, 6, 8, and 14 m s 1, respectively. 8of14

9 Figure 10. Same as Figure 9 except for 12 September [35] What is the source for such large speeds? Here we discuss it, although we have not yet identified feature of the primary source. Furthermore, we do not have enough experimental information to decide if we should consider the source separately for statistically expected and exceptional cases. Here we list various mechanisms that may play a role in producing vertical motions Effects of the E B Drift [36] Ion motions perpendicular to the magnetic field due to the E B drift have a vertical component because the magnetic field line is off vertical at the Tromsø and Sondrestrom sites. We conclude, however, that this component has relatively small effects on the vertical ion velocity measured at 100 km level as discussed below. [37] Figure 11 shows the height profile of the electric field component that may cause vertical ion velocities of 20 and 50 m s 1 by the E B drift at the Tromsø and the Sondrestrom sites. The ratio of the ion gyrofrequency to the ion-neutral collision frequency calculated using data from the Mass Spectrometer Incoherent Scatter (MSIS) model [Hedin, 1991] for near solar maximum summer condition has been used to develop this figure. Slight differences in these sites are due to the local magnetic field and the ionneutral collision frequency. This calculation suggests that electric fields about 100 and 300 mv m 1 are necessary for vertical ion velocities of 20 and 50 m s 1 at 100 km, respectively. These electric fields might be measured in the polar ionosphere sporadically and/or locally, but such large values are generally not observed for a long interval of order of hour, especially during geomagnetically quiet periods Effects of AGW [38] Wavelike structures in the vertical ion velocity can be seen in Figures 9 and 10, as mentioned in section 4.2. If these wavelike structures are relevant to upward propagating atmospheric gravity waves (AGWs), the amplitude should increase with height in association with the decrease in the atmospheric density although specific effects depend on the vertical wavelength. For example, since oscillations from 1600 to 2000 UT on 5 September 2003 at 97, 100, and 103 km are in phase, the vertical wavelength must be significantly longer than 6 km (the difference between 97 and 103 km), although this estimate is not readily justified because of the limited number of altitude measurements. If these values are real, we can understand the reason why amplitudes at 100 km are slightly larger than those at 97 km; however, it is not clear why amplitudes at 103 km suddenly attenuate compared with those at 100 km. [39] Figure 12 shows temporal variations in the ion velocity measured at 100 km on 5 September 2003 with three individual beams, as illustrated in Figure 6. The ion 9of14

10 Figure 11. Height profile of geomagnetically zonal component of the electric field to generate vertical ion velocity of 20 (thin curve) and 50 (thick curve) m s 1 due to the E B drift at the Tromsø (dashed curve) and the Sondrestrom (solid curve) sites. velocity from the beam a presents a relatively long timescale oscillation at period of approximately 6 hours with positive and negative peaks around 1700 and 1400 UT, respectively. The magnetic field-aligned ion velocity also presents similar long timescale oscillations but smaller amplitudes. The vertical ion velocity, however, does not show such long timescale oscillations. Since horizontal motions of neutral particles do not drag ions vertically through collisions, these comparisons suggest that long timescale oscillations can be seen in the horizontal neutral wind but not the vertical one. Short timescale oscillations appear less correlated among three individual velocities, which may relate to the spatial scale of the short timescale oscillations. [40] The combination of two different kinds of waves at short and long oscillation periods may contribute to complicated structures in the vertical ion velocity. This hypothesis has a foundation in previous results with optical instruments and numerical calculations in the upper mesosphere [Taylor and Hapgood, 1990; Fritts et al., 1997; Hecht et al., 1997, 2000; Yamada et al., 2001]. Simulations of wave breaking with a three-dimensional model suggest that instability structures produced by the combination of two different kinds of gravity waves at short (from 25 to 50 km scale) and long (from 500 to 1500 km) wavelengths account best for small-scale (less than 10 km) structures in AGWs [Fritts et al., 1997]. These studies do not, however, refer to the lower thermospheric features. While AGWs propagating upward decelerate the background neutral wind in the upper mesosphere [Lindzen, 1981], we may also have to consider the combination in the lower thermosphere, if a portion of AGWs can propagate upward into the lower thermosphere. [41] Vertical motion should affect the chemistry through vertical transfer of thermospheric and ionospheric materials and temperature changes due to adiabatic cooling and heating. The temperature decreases due to upward motion of 3 and 6 km, corresponding to one and two IS radar gates, are 29 and 58 K at the adiabatic lapse rate of K m 1. The ion temperature shown in Figure 9c presents the overturning in association with fluctuations in the electron density after 1800 UT when the vertical ion velocity has large amplitudes, although the temperature differences may not be significant because they are smaller than the measurement uncertainty. These ionospheric phenomena may be results from coupling between chemistry and dynamics in the lower ionosphere Effects of the Solar Terminator [42] Since the ground-sunset time is 2237 and 2208 UT on 5 and 12 September 2003, respectively, the solar terminator does not affect the lower thermospheric dynamics directly for the Sondrestrom experiments. It is, however, interesting to take its effect into account because there are experimental reasons to believe that AGWs and traveling ionospheric disturbances (TIDs) are generated in association with the solar terminator [Beley et al., 1995; Galushko et al., 1998]. Since the previous results are derived using data in the F region, we cannot simply regard the solar terminator as the potential source for large amplitudes of the lower ionospheric vertical motion Effects of Thermal Expansion Due to Joule and Particle Heating [43] Another candidate is atmospheric thermal expansion in association with geomagnetic activity. It might be inconsistent with predictions by thermospheric models [Walterscheid et al., 1985; Sun et al., 1995; Shinagawa et al., 2003]; but Kofman et al. [1996] estimate vertical neutral winds in excess of 50 m s 1 at E region heights during geomagnetically disturbed periods. Figure 12. Temporal variations in the ion velocity measured at 100 km on 5 September 2003 with three different beams; vertical, magnetic field-aligned, and other oblique beams illustrated in Figure 6. Positive value is upward along each line-of-sight. Vertical bar is the measurement uncertainty with ±1s. 10 of 14

11 Figure 13. Plasma convection map estimated with the SuperDARN radar in the northern polar region from 1450 to 1540 UT on 5 September Solid dot represents location of the Sondrestrom site. The potential extrema are shown as cross symbols (negative) and plus symbols (positive). 11 of 14

12 [44] Since thermospheric disturbances can propagate horizontally as AGWs by a few hundred kilometers [Oyama et al., 2001, and references therein], we may have to study the effects of the thermospheric disturbances even though the local ionosphere observed is geomagnetically quiet. During the Sondrestrom IS radar experiment on 5 September 2003, regional enhancements in the electric field have been measured at about 400 km poleward from the Sondrestrom site around 1500 UT, as shown in Figure 13. This enhancement corresponds to negative deviations of the H-component of the geomagnetic field at UMQ and GDH from 1500 to 1600 UT (see Figure 7). Wavelike structures in the magnetic field-aligned ion velocity analyzed by Oyama et al. [2001] show typical values for the medium-scale AGWs according to the classification by Hunsucker [1982]. This classification summarizes that the oscillation period of medium-scale AGWs is from 15 to 60 min and the horizontal propagation speed is from 1000 to 250 m s 1.If medium-scale AGWs are generated by enhancements of the electric field at 1500 UT, the wave front should reach the observation point approximately from 1507 to 1527 UT. The vertical ion velocities at 97 and 103 km after 1500 UT seem to fluctuate with larger amplitudes and shorter periods than those before 1500 UT. The changes occurred close to the expected timing. The ion velocity at 100 km, however, does not show notable change in the wave characteristics, which suggests that above explanation is insufficient to perfectly explain features in the observed oscillations. 5. Summary and Conclusions [45] In this paper we present statistical results of the lower ionospheric vertical velocity observed with the Tromsø UHF radar for geomagnetically quiet summer conditions. While the statistical results show that most vertical ion velocities at 100 km level have amplitudes less than 20 m s 1, the magnitudes sometimes exceed 20 m s 1 even during geomagnetically quiet periods. Vertical ion velocities measured with the Sondrestrom IS radar on 5 and 12 September 2003 also show large magnitudes below 103 km where we have concluded that the ion velocity is equal to the neutral wind velocity. We summarize findings below: [46] 1. Statistics have been used with the vertical ion velocity data from the Tromsø UHF radar in the CP-2 mode for geomagnetically quiet summer conditions. The magnitude of the hourly mean velocity is a few m s 1 in the lower ionosphere, although the actual mean value has not been determined precisely due to large measurement uncertainties and statistical standard deviations. [47] 2. Histograms of the vertical ion velocity at local daytime suggests that some portions of observation periods have amplitudes larger than 20 m s 1, which is significantly larger than the measurement uncertainty at each height. [48] 3. The statistical standard deviation at local daytime is smaller than that at night, although the reason was not specified because of the comparable measurement uncertainties. [49] 4. We estimate the probability distribution function (PDF) of real vertical ion velocity. This calculation suggests that most measurements have amplitudes smaller than 20 m s 1 around 100 km, although amplitudes larger than 20 m s 1 can be measured as statistical exception. [50] 5. The vertical ion velocities observed with the Sondrestrom IS radar at 100 km level on 5 and 12 September 2003 show amplitudes larger than a few tens m s 1 during geomagnetically quiet periods. Calculations of the relative velocity between ions and neutrals using observed data suggest that neutral particles below 103 km also have fluctuated with such large amplitudes. [51] 6. We discuss a hypothesis for the generation of such large vertical motions in the lower ionosphere. While we cannot yet determined what the primary source is, energy input in association with geomagnetic activity seems not to be the unique source. The coupling of neutral dynamics in the mesosphere and lower thermosphere should be an important component in understanding the physical mechanism. Appendix A: Estimates of the Vertical Neutral Wind for the Sondrestrom IS Radar Experiments on 5 and 12 September 2003 [52] In this section we explain how to estimate the upper boundary height where the ion velocity is equal to the neutral wind velocity. Since one antenna cycle takes about 10 min, and there is distance of a few tens kilometers between the vertical beam and the beam a in the E region, we have to assume that the ionosphere is uniform during one sequence and in this distance to estimate the boundary. This assumption is appropriate because of geomagnetically quiet conditions, as shown in Figures [53] We calculate differences in the y-component between ion and neutral wind velocities. The y-component of neutral wind velocity can be written from the ion momentum equation as follows: V y ¼ U y þ W i E y þ V z B x V x B z Bn in ða1þ Since the magnetic field vector can be written as equation (A2) in the coordinate shown in Figure 6, equation (A1) can be written as equation (A3), B ¼ ð cos I; 0; sin IÞB; ða2þ V y ¼ U y þ W i E y =B V z cos I þ V x sin I ; ða3þ n in where I is the magnetic dip angle, which is 80 at the Sondrestrom site. The V x can be written simply using observed velocities by vertical and a beams or V 1 and V 3, respectively: V x ¼ V 3 cos a V 1 tan a; ða4þ where a is the elevation angle of beam a or 70. The y-component of electric field, E y, can be estimated using data measured in the F region: E y ¼ V F z BF x þ V F x BF z ; ða5þ 12 of 14

13 Figure A1. Temporal variations in y- or geomagnetically zonal component of the relative velocity between ions and neutrals from 97 to 121 km on 5 September Vertical bar is the estimated error with ±1s. where the superscript F stands for values in the F region. Then the y-component of the electric field can be written as E y ¼ B F ðcos I þ tan a sin IÞV1 F sin I cos a V F 3 : ða6þ Substituting equations (A4) and (A6) into equation (A3), the difference between V y and U y becomes V y U y ¼ W i n in ðcos I þ tan a sin IÞ BF sin I cos a B F B V 3 F V 3 B V F 1 V 1 : ða7þ [54] Figure A1 shows temporal variations in the difference between V y and U y estimated by equation (A7) from 97 to 121 km using data from the special experiment on 5 September Temporal variations for the 12 September case are analogous to those for the 5 September case (not shown here). [55] Acknowledgments. We express our appreciation to Mary McCready, SRI International, for checking the IS spectra. The Sondrestrom facility is operated by SRI International under cooperative agreement ATM with the US National Science Foundation and in cooperation with the Danish Meteorological Institute. We are indebted to the director and staff of EISCAT for operating the facility and supplying the data. EISCAT is an international association supported by Finland (SA), France (CNRS), the Federal Republic of Germany (MPG), Japan (NIPR), Norway (NFR), Sweden (VR) and the United Kingdom (PPARC). SuperDARN research at the University of Alaska is supported National Science Foundation grant OPP from the Arctic Natural Sciences program of the Office of Polar Programs. Geomagnetic data were provided by the Danish Meteorological Institute. [56] Arthur Richmond thanks the reviewers for their assistance in evaluating this paper. References Beley, V. S., V. G. Galushko, and Y. M. Yampolski (1995), Traveling ionospheric disturbances diagnostic using HF signal trajectory parameter variations, Radio Sci., 30(6), Fritts, D. C., J. R. Isler, J. H. Hecht, R. L. Walterscheid, and Ø. Andreassen (1997), Wave breaking signatures in sodium densities and OH nightglow: 2. Simulation of wave and instability structures, J. Geophys. Res., 102(D6), Galushko, V. G., V. V. Paznukhov, Y. M. Yampolski, and J. C. Foster (1998), Incoherent scatter radar observations of AGW/TID events generated by the moving solar terminator, Ann. Geophys., 16, Hecht, J. H., R. L. Walterscheid, D. C. Fritts, J. R. Isler, D. C. Senft, C. S. Gardner, and S. J. Franke (1997), Wave breaking signatures in OH airglow and sodium densities and temperatures 1. Airglow imaging, Na lidar, and MF radar observations, J. Geophys. Res., 102(D6), Hecht, J. H., C. Fricke-Begemann, R. L. Walterscheid, and J. Höffner (2000), Observations of the breakdown of an atmospheric gravity wave near the cold summer mesopause at 54N, Geophys. Res. Lett., 27, Hedin, A. E. (1991), Extension of the MSIS thermosphere model into the middle and lower atmosphere, J. Geophys. Res., 96(A2), Hunsucker, R. D. (1982), Atmospheric gravity waves generated in the highlatitude ionosphere, Rev. Geophys., 20(2), Ishii, M., S. Oyama, S. Nozawa, R. Fujii, E. Sagawa, S. Watari, and H. Shinagawa (1999), Dynamics of neutral wind in the polar region observed with two Fabry-Perot Interferometers, Earth Planets Space, 51, Kofman, W., C. Lathuillere, and B. Pibaret (1996), Neutral dynamics of the high latitude E region from EISCAT measurements: A new approach, J. Atmos. Terr. Phys., 58(1 4), Lindzen, R. S. (1981), Turbulence and stress owing to gravity wave and tidal breakdown, J. Geophys. Res., 86(C10), of 14

14 Oyama, S., M. Ishii, Y. Murayama, H. Shinagawa, S. Nozawa, S. C. Buchert, R. Fujii, and W. Kofman (2001), Generation of atmospheric gravity waves associated with auroral activity in the polar F region, J. Geophys. Res., 106(A9), 18,543 18,554. Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery (1992), Numerical Recipes in C, Cambridge Univ. Press, New York. Price, G. D., and F. Jacka (1991), The influence of geomagnetic activity on the upper mesosphere/lower thermosphere in the auroral zone. I. Vertical winds, J. Atmos. Terr. Phys., 53(10), Price, G. D., R. W. Smith, and G. Hernandez (1995), Simultaneous measurements of large vertical winds in the upper and lower thermosphere, J. Atmos. Terr. Phys., 57(6), Shinagawa, H., S. Oyama, S. Nozawa, S. C. Buchert, R. Fujii, and M. Ishii (2003), Thermospheric and ionospheric dynamics in the auroral region, Adv. Space Res., 31(4), Smith, R., and G. Hernandez (1995), Vertical winds in the thermospheric within the polar cap, J. Atmos. Terr. Phys., 57(6), Sun, Z. P., R. P. Turco, R. L. Walterscheid, S. V. Venkateswaran, and P. W. Jones (1995), Thermospheric response to morningside diffuse aurora: High-resolution three-dimensional simulations, J. Geophys. Res., 100(A12), 23,779 23,793. Taylor, M. J., and M. A. Hapgood (1990), On the origin of ripple-type structure in the OH nightglow emission, Planet. Space Sci., 38, Walterscheid, R. L., L. R. Lyons, and K. E. Taylor (1985), The perturbed neutral circulation in the vicinity of a symmetric stable auroral arc, J. Geophys. Res., 90(A12), 12,235 12,248. Yamada, Y., H. Fukunishi, T. Nakamura, and T. Tsuda (2001), Breaking of small-scale gravity wave and transition to turbulence observed in OH airglow, Geophys. Res. Lett., 28(11), M. Conde, La Trobe University, Victoria 3086, Australia. (m.conde@ latrobe.edu.au) S. Maeda, Kyoto Women s University, 35 Kita-Hiyoshi Imakumano, Higashiyama, Kyoto , Japan. (smaeda@kyoto-wu.ac.jp) S. Nozawa, Solar-Terrestrial Environment Laboratory, Nagoya University, Furo-cho, Chikusa-ku, Nagoya , Japan. (nozawa@stelab. nagoya-u.ac.jp) S. Oyama and B. J. Watkins, Geophysical Institute, University of Alaska Fairbanks, 903 Koyukuk Drive, P.O. Box , Fairbanks, AK , USA. (soyama@gi.alaska.edu; bjw4072@mailblocks.com) 14 of 14