JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117, A03312, doi:10.1029/2011JA017161, 2012
Statistical investigation of horizontal propagation of gravity waves in the ionosphere over Europe and South Africa J. Chum,1 R. Athieno,2,3 J. Baše,1 D. Burešová,4 F. Hruška,1 J. Laštovička,4 L. A. McKinnell,2,3 and T. Šindelářová4 Received 13 September 2011; revised 6 January 2012; accepted 9 January 2012; published 9 March 2012.
[1] A statistical investigation into the horizontal propagation of 8–30 min gravity waves (GWs) in the ionosphere over a 1 year period from June 2010 to May 2011 is presented. The GWs were observed by multipoint continuous Doppler sounding systems installed in the Czech Republic and in the Western Cape, South Africa. Measurements of GW propagation in the ionosphere over South Africa have never been presented before. Simultaneous measurements from nearby ionosondes made it possible to estimate the height of the GW observations and show that the analyzed GWs propagated at altitudes from 150 to 250 km. The analyzed waves were mainly observed after sunrise and around sunset. Our statistical study shows that the analyzed GWs propagated with horizontal velocities from 70 to 250 m/s. The average observed horizontal velocities were 100 m/s in the local summer and 125–150 m/s in the local winter. The waves propagated approximately poleward in the local summer, whereas roughly equatorward propagation was observed in the local winter. Westward propagation was rarely observed in the Czech Republic, and eastward (southeast) propagation was seldom observed in South Africa. A comparison with neutral wind velocities shows that the analyzed GWs propagated approximately against the neutral winds calculated by the HWM07 model. The estimated horizontal wavelengths of the analyzed waves were 100–300 km. Citation: Chum, J., R. Athieno, J. Baše, D. Burešová, F. Hruška, J. Laštovička, L. A. McKinnell, and T. Šindelářová (2012), Statistical investigation of horizontal propagation of gravity waves in the ionosphere over Europe and South Africa, J. Geophys. Res., 117, A03312, doi:10.1029/2011JA017161.
1. Introduction [2] Atmospheric gravity waves (GWs) transfer energy and momentum between the lower, middle, and upper atmosphere. The theoretical and experimental investigation of GWs has, therefore, attracted the attention of many atmospheric scientists since the pioneering work of Hines [1960]. [3] Investigation into propagation directions and speeds of GWs in the mesosphere and thermosphere has been based mainly on optical observations of airglow at heights of 85–100 km and 200–300 km [Dou et al., 2010; Nakamura et al., 1998; Shiokawa et al., 2009; Taylor et al., 1998]. When a GW propagates through the emitting layer, it causes perturbations of the air density and temperature and hence the 1 Department of Upper Atmosphere, Institute of Atmospheric Physics, Prague, Czech Republic. 2 South African National Space Agency Space Science, Hermanus, South Africa. 3 Department of Physics and Electronics, Rhodes University, Grahamstown, South Africa. 4 Department of Aeronomy, Institute of Atmospheric Physics, Prague, Czech Republic.
Copyright 2012 by the American Geophysical Union. 0148-0227/12/2011JA017161
variations of the airglow intensity. Shiokawa et al. [2009] reviewed the propagation of the nighttime GWs in the upper atmosphere over Japan, Indonesia and Australia. They found seasonal variations of their occurrences and propagation directions. Poleward propagation of small-scale (less than 100 km) GWs dominated in the mesopause region at most of the stations in the local summer, suggesting that mesospheric GWs might be generated by intense convective activity in the tropics. On the other hand, in the thermosphere at heights 200–300 km, Shiokawa et al. [2009] reported that medium-scale GWs (100–1000 km) propagated predominantly equatorward and westward in all seasons. A summary of the GW propagation directions in the mesopause region over the northern Colorado, United States, was recently reported by Dou et al. [2010], who found that GWs propagated preferentially poleward in the summer and equatorward in the winter. The poleward propagation of GWs over Urbana, Illinois, United States, in the summer was also reported by Hecht et al. [2001]. Medeiros et al. [2003] concluded that the GW propagation directions over Brazil are controlled by strong filtering of the waves in the middle atmosphere by stratospheric winds. Smith et al. [2009] observed stationary mesospheric waves over Argentina generated probably by 70 m/s eastward winds flowing over the Andes. Suzuki et al. [2009] observed preferentially westward propagating
A03312
1 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
waves at high latitudes over Resolute Bay, Canada, which they found consistent with the prevailing eastward winds in the mesosphere. [4] From space, Chandran et al. [2010] investigated structures in polar mesospheric clouds (PMCs) observed by the AIM satellite around the summertime mesopause region. These structures were qualitatively similar to structures seen in noctilucent clouds from ground-based photographs and are generally considered to be manifestations of upward propagating atmospheric GWs. Chandran et al. [2010] also mentioned that GWs can influence PMC brightness and occurrence, and investigated the correlation of PMC wave structures with temperature fluctuations at 83 km. [5] Despite numerous observations, a continuous global picture of the time and space distribution of GW activity and their characteristics is far from complete, especially in the upper atmosphere. A systematic investigation of GW propagation directions in the ionosphere, i.e., in the ionized part of the atmosphere, is even more rare. A review of various kinds of atmospheric waves and their sources in the ionosphere was given by Hocke and Schlegel [1996], and Laštovička [2006]. GW propagation in the ionosphere has often been studied from perturbations of the total electron content (TEC) measured by a network of GPS receivers [e.g., Lee et al., 2008]. Lay and Shao [2011] recently presented an interesting new method for probing ionospheric fluctuations in the D layer at heights 80 km using the analysis of very low and low-frequency electromagnetic waveforms produced by lightning from storms several hundred kilometers away. Laštovička [1999] and Bošková and Laštovička [2001] used the absorption of low-frequency radio waves to investigate the occurrences of GWs in the lower ionosphere. Chum et al. [2010] performed a statistical analysis of propagation velocities of short-period GWs in the ionosphere at heights of 150–250 km over the Czech Republic using continuous multipoint Doppler radar measurements and found that the observed GWs mostly propagated equatorward in the winter and poleward in the summer. This is a very similar result to many previously reported optical middle-latitude observations of GWs in the mesospause region mentioned in this section. [6] Though the relationship between the neutral and ion density fluctuations has been studied both theoretically and experimentally [e.g., Fritts and Thrane, 1990; Kirchengast et al., 1996], it is necessary to note that the coupling between the neutral atmosphere and ionosphere is still not well understood in detail [e.g., Kelley and Miller, 1997; Kelley, 2009]. Studies have been undertaken to investigate if the GWs can cause ionospheric instabilities and lead to formation of midlatitude spread F [Perkins, 1973; Miller, 1997; Miller et al., 1997; Xiao et al., 2009]. The systematic observation of GWs in the ionosphere is of great importance to better understanding of the coupling, momentum and energy transfer between the neutral and ionized part of the atmosphere and the role of GWs in ionospheric dynamics. [7] In this paper we will extend the statistical analyses of GW propagation directions in the ionosphere reported by Chum et al. [2010] by presenting the horizontal propagation velocities measured by the continuous multipoint Doppler sounding system on two opposite hemispheres: in the Czech Republic, central Europe, and in the Western Cape, South
A03312
Africa over 1 year from June 2010 to May 2011. We also are able to analyze more events by relaxing the previous requirement of observing the same S shape at all measuring points; the details are explained in section 2.
2. Data Analysis and Locations of Measurements [8] An introduction to continuous Doppler sounding with application to the GW observations was given in a previous paper by Chum et al. [2010, and references therein]. Chum et al. [2010] also explained the method of determination of horizontal velocities and directions of GWs in detail. Here, we present the main points in a simplified way, to add clarity to the present discussions. A vertically propagating radio wave reflects at the height where its frequency matches the local plasma frequency. A gravity wave causes, via collisions between neutral and charged particles, fluctuations (movement) of the reflecting level, and hence the Doppler shift of the radio wave reflected from the ionosphere. We use an array of Doppler transmitters to obtain several spatially separated sounding radio paths. We assume that the reflection points of radio waves are located midway between the receiver and individual transmitters, when projected to the ground. The observed horizontal velocities of GWs are then computed from the time delays between the observation of corresponding signatures on different sounding paths. An assumption of plane wave propagation is used for these calculations. Transmitters form a quasi-isosceles triangle or several quasi-isosceles triangles. This ensures that the conditions for the measurements of time delays do not depend on the propagation direction of the GWs. The actual locations of the transmitters, however, also depend on the feasibility of the location with regards to the construction of the required antenna systems. 2.1. Geographical Distribution of the System [9] Geographical positions of the transmitters and receiver located in the Czech Republic are shown in Figure 1. All the transmitters are located in the western part of the Czech Republic. To calculate the horizontal velocities we use the quasi-isosceles triangles formed by transmitters Tx1-Tx2-Tx3 (small triangle) and Tx1-Tx4-Tx5 (large triangle). As for South Africa, the system is located in the southern part of the Western Cape. The geographical positions are presented in Figure 2. Two transmitters are located along the ocean coastline: one in Cape Town and the other in Arniston, close to the southernmost point of Africa, Cape Agulhas. The third transmitter is inland, near Worcester, behind the mountain ridge. The receiver is located on the coastline in Hermanus, between Cape Town and Cape Agulhas. Although the distances from geographical equator differ by 16° of latitude for the Czech Republic (50°N) and South African (34°S) observations, the geomagnetic latitudes are relatively similar (45°N for the Czech Republic, and 42°S for South Africa). [10] The frequencies of individual transmitters are shifted by 4 Hz, thus, received signals from all the transmitters can be displayed in one common Doppler shift spectrogram [Chum et al., 2010]. The transmitted frequencies f are as follows: f (Tx1) = 3.594500 MHz, f (Tx2) = 3.594504 MHz, f (Tx3) = 3.594496 MHz, f (Tx4) = 3.594508 MHz,
2 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
A03312
Figure 1. Distribution and geographical coordinates of transmitters (blue diamonds) and the receiver (red circle) in the Czech Republic. Blue crosses indicate midpoints between the transmitters and receiver. f(Tx5) = 3.594492 MHz for the Czech Republic, and f(Tx1) = 3.594480 MHz, f(Tx2) = 3.594484 MHz, f(Tx3) = 3.594476 MHz for South Africa. [11] Note that the frequency shifts between the individual transmitters are negligibly small to cause significant differences in reflection heights (typically less than 1 m). We also neglect differences between the reflection heights (up to several km) caused by different distances between transmitters and receiver and/or by large-scale horizontal gradients in the ionosphere. Horizontal gradients induced by the analyzed GWs are taken into account in the velocity calculations and error estimates [Chum et al., 2010]. [12] The reflection heights for individual events are estimated from the ionograms obtained by the digital model DPS-4 ionosondes (http://ulcar.uml.edu/digisonde_dps. html) located in Pruhonice 1 km away from the transmitter Tx2 and 7 km from the receiver in the Czech Republic, and in Hermanus, 100 m from the South African receiver. 2.2. Limitations of Measurements and Event Selection (S-Shaped Criteria) [13] The reflection heights are almost the same for all the sounding paths, so we can only analyze propagation in the horizontal plane. It is necessary to stress that the calculated horizontal velocities are the observed phase velocities in the Earth frame, i.e., they represent the superposition of wind speeds and intrinsic phase velocities of GWs. It is important
to note that the modulus of phase velocity v is related to horizontal velocity vH by relation vH = v/cos(a), where a is the angle between the horizontal plane and propagation direction. The intrinsic horizontal velocities are calculated by subtracting the neutral wind speeds from the measured velocities. The neutral wind speeds are calculated from the recent global empirical model HWM07 [Drob et al., 2008; Emmert et al., 2008] at the reflection heights and times for individual cases. [14] We use the GWs that produce S-shaped patterns in the Doppler shift spectrogram in our analysis because the S-shaped pattern can be considered as a signature of GW propagation with a significant horizontal component [Davies and Baker, 1966; Georges, 1967]. An example of an unusually disturbed ionosphere and large amplitude GWs with S-shaped patterns observed in Hermanus, South Africa, on 11 July 2010 is shown in Figure 3. The letters A and B indicate corresponding features on different sounding paths. The horizontal velocities are computed from the time differences between the midpoints of the S-shaped patterns observed on various paths (Dt12 and Dt13 in Figure 3). The velocities are, however, also computed from the time differences between the start and end of the S-shaped patterns to estimate the error of the measurement (see Chum et al. [2010] for more details). The observed period can be estimated as the time distance between the subsequent S-shaped patterns, e.g., between disturbances A and B in Figure 3. In some
3 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
A03312
Figure 2. Distribution and geographical coordinates of transmitters (blue diamonds) and the receiver (red circle) in South Africa. Blue crosses indicate midpoints between the transmitters and receiver. cases, like in Figure 3, the determination of the period is not unambiguous because the time distances between the subsequent S-shaped patterns change with time, possibly because of the interference. The exact spectral analysis is problematic because the S-shaped pattern is not a single valued function of time. A minor spread is present on the received signals in this case. Note also a weak ground wave, characterized by zero Doppler shift, which is observed on the sounding path Tx3-Rx (bottom trace). The disturbances seen as vertical lines that regularly repeat each 15 min in the spectrogram are caused by the quick wide-band frequency sweep of the nearby ionosonde. 2.3. Estimating the Horizontal Wavelength [15] A necessary condition for the observation of the S-shaped pattern is that the radius of curvature of the disturbance is smaller than the reflection height [Davies and Baker, 1966]. In other words, three different rays reflected from different places within the disturbance have to be received simultaneously to observe the S-shaped pattern. A horizontal movement of the disturbance is then responsible for the negative and positive Doppler shift of the outer rays, whereas the middle ray experiences the intermediate Doppler shift. Note that this condition poses an upper limit on the horizontal wavelength lH of the GWs that are characterized by the S-shaped pattern. If we assume for simplicity a disturbance of the form h = h0 + h1 cos(2px/lH), where x is the horizontal distance along the direction of propagation, then this
condition can be expressed as follows [Georges, 1967; Chum et al., 2010]: pffiffiffiffiffiffiffiffiffi lH < 2p h0 h1 :
ð1Þ
It is reasonable to assume that in the ionosphere, the average reflection height h0 is much larger than the wave amplitude h1. [16] Knowing the reflection height, wave amplitude, and horizontal phase velocity vH, we can approximately calculate the horizontal wavelength of the analyzed wave or its upper limit by applying the inequality shown in equation (1). The estimate of reflection height h0 is obtained directly from the ionogram (we take the true reflection height). The wave amplitude h1 can be related to the maximum of the measured Doppler shift and horizontal phase velocity vH. Next, we will use an approximation of the mirror-like reflection of the sounding radio wave. The observed Doppler shift Df is then related to radial velocity vR of the movement of the reflecting level by Df ¼ 2f
vR cosb; c
ð2Þ
where b is the angle between the normal to the reflecting level and incident (reflected) ray, f is the sounding frequency (3.59 MHz) and c is the speed of light. Note that if the transmitter and receiver are at the same place, then b = 0. We will further consider that cosb 1 for simplicity.
4 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
A03312
Figure 3. Doppler shift spectrogram recorded in Hermanus, South Africa, on 11 July 2010. The record begins at 15:00 UT. Various transmitter-receiver paths are offset by 4 Hz. A and B indicate analyzed disturbances observed on different sounding paths. The color indicates the common logarithm of power spectral intensity of the received sounding radio wave in arbitrary units (the antennas are not calibrated). The quantities determined from the Doppler shift spectrogram are schematically drawn in white: DfPP is peak to peak value of the Doppler shift (difference between the largest and smallest Doppler shift on the S-shaped pattern), T is the observed period of GW, and Dt12 and Dt13 are the time differences between the S-shaped patterns observed on various sounding paths. This assumption introduces only a small error since one half of the distance between the transmitter and receiver is smaller than the distance from the reflecting level, and hence b is relatively small. If we for now neglect the vertical movement of the disturbance, then the radial velocity seen by the Doppler radar is given by vR ¼ vH sin d;
ð3Þ
where d is the angle between the reflecting level and horizontal plane at the point of reflection. Measuring the maximum of Doppler shift DfM and calculating the horizontal velocity vH, we can find the maximum of angle d, denoted further dΜ. After substituting the expression for vR from
equation (3) into (2) and considering cosb 1 we get for maximum values DfM and dΜ. vH sind M ¼
DfPP c; 4f
ð4Þ
where DfPP = 2jDfMj. It is advantageous to measure the peak to peak value DfPP (Figure 3) of the Doppler shift observed on the S-shaped pattern (difference between the largest and smallest Doppler shift on the S-shaped pattern), and to calculate jDfMj = DfPP/2, instead of measuring DfM directly. The main reason for that is that the analyzed GW can be superimposed on a long-period GW or on a decreasing or increasing trend of the Doppler shift during the sunset or
5 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
A03312
Figure 4. Doppler shift spectrogram recorded in Prague, Czech Republic, on 3 April 2011. The record begins at 05:00 UT. A indicates the analyzed disturbance observed on different sounding paths. The color indicates the common logarithm of power spectral intensity of the received sounding radio wave in arbitrary units (the antennas are not calibrated). sunrise when the reflection level moves up or down, respectively. Thus, computing DfM from the DfPP value partially eliminates the influence of a long-period wave, trend, or vertical component of propagation of the analyzed wave. [17] At the same time, dΜ is also given by the maximum slant of the disturbance and tan(dΜ) can be determined from the maximum derivative of the disturbance. For the sinusoidal disturbance of the form h1cos(2px/lH) we get j tanðd M Þj ¼
2ph1 ; lH
ð5Þ
The term on the right-hand side of equation (5) can also be expressed using the inequality (1). Combining equations (1), (4), and (5), we finally get DfPP c lH ≤ 2ph0 tan sin1 ; 4f vH
ð6Þ
2.4. Extension of the Analysis (Relaxing Strict S-Shaped Criteria) [18] Several reasons could cause the S-shaped signature to be observed on only one or several paths: The GWs never propagate strictly horizontally as follows from the dispersion relation. At the same time, the group velocity of short-wavelength GWs is more vertical than horizontal. That can lead to low cross correlation between signals received from distant reflection points for narrow ray bundles. Next, as follows from equation (1), the condition for observing the S-shaped signature depends on the wave amplitude and height of the reflection; thus, wave attenuation and different reflection heights for different sounding paths might also contribute to not observing the S-shaped signature on all the sounding paths. [19] To enlarge the number of measurements, we included into our statistical analysis (both in South Africa and Czech
6 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
Republic) cases when the S-shaped pattern was observed only on one or two signal paths, and a corresponding disturbance(s) could be recognized on the other path(s). An example of such an observation is shown in Figure 4, which presents a measurement from the Czech Republic recorded on 3 April 2011. The letter A again indicates the corresponding disturbances observed on different sounding paths. The GWs have moderate amplitudes in this case. A strong ground wave is observed on the path Tx2-Rx because of the proximity of Tx2 to Rx. An indistinct S-shaped pattern is observed on the Tx1-Rx trace, but on Tx2-Rx, Tx3-Rx, Tx4-Rx and Tx5-Rx paths, the corresponding disturbances are represented by almost vertical slopes on the signal traces, so the corresponding feature can be found when the derivative has the maximum positive value. Therefore, for all such observations, we take the time of the maximum derivative of the analyzed disturbance to calculate the time delays between the detection of corresponding features on different sounding paths. The maximum derivative corresponds to the change from negative Doppler shift to the positive when the crest of the quasi-horizontally propagating wave passes over the particular sounding path of the Doppler radar. If the curvature radius of the wave crest (reflecting level) decreases, but is still larger than the reflection height, then this change becomes more rapid (positive derivative increases). When the curvature radius of the reflecting level becomes less than the reflection height, three rays start to be received simultaneously and S-shaped signature appears [Georges, 1967]. Thus, the analysis based on maximum derivative remains consistent with the analysis based only on the S-shaped patterns. We check the consistency by comparing the velocities and azimuths determined by this method from the small and large triangle in Figure 4. The horizontal velocity calculated from the small triangle (Tx1-Tx2-Tx3) is 95 30 m/s with an azimuth of propagation of 112°. The velocity calculated from the large triangle (Tx1-Tx4-Tx5) is 91.5 7.3 m/s and its azimuth is 110°. The results of statistical investigation (presented in Figures 7a and 8a) show that the accuracy of velocity calculations from the large triangle (cyan) is generally better than the accuracy of velocities obtained from the small triangle (magenta), since the errors in the measurement of time differences are similar for both triangles, but the distances over which the velocities are computed differ. [20] Using our selection criteria, 158 events were analyzed in the Czech Republic and 71 in South Africa. There are two reasons for the larger number of measurements in the Czech Republic. First, most of the GW velocities in the Czech Republic were computed from the small triangle, for which the distances between the reflection points are relatively small and the received signals on different signal paths are well correlated. On the other hand, it was often difficult to find corresponding features on the individual paths for the large triangle. The argument for the low cross correlation on larger distances is the same as in the case of the S-shaped pattern observed just on one sounding path, namely, the narrow ray bundle. In total, we analyzed 158 events in the Czech Republic from the small triangle, but only 73 of those events could also be analyzed from the large one. The distances in the South African triangle are similar to the distances in the large Czech Republic triangle rather than to distances in the small triangle. Second, there were significant
A03312
data gaps owing to technical problems in South Africa from the beginning of December 2010 to the mid of February 2011. We cannot, however, exclude that the lower number of measurements in South Africa is partially caused by different geographical locations, and thus by different conditions for wave generation and propagation.
3. Propagation Variability of GWs Over Czech Republic and South Africa 3.1. Time Distribution of the Analyzed Events [21] In this section, we present the results obtained over a 1 year period from June 2010 to May 2011. The analyzed GWs were mainly observed around sunrise and sunset since the occurrence of the S-shaped patterns is highest at these times. Around noon, especially in the summer, the radio waves reflect from low altitudes where the amplitude of GWs is lower, so the observed Doppler shift is usually very small, and the detection of corresponding features is impossible. On the other hand, Doppler shift frequencies often form thick (spread) traces in Doppler shift spectrogram around midnight, which makes the determination of time differences also impossible or unreliable. These thick traces occur when a spread F is observed in the ionograms. Also, GWs with long periods or large-scale horizontal wavelengths that should not produce the S-shaped patterns prevailed around midnight. The diurnal and seasonal distribution of the analyzed events is shown in the form of scatterplots in Figure 5 for the Czech Republic and in Figure 6 for South Africa. Dashed lines represent the times of sunrise and sunset on the ground, and the solid lines indicate the times of sunrise and sunset at an altitude of 200 km. Figure 6 also shows that we analyzed relatively few sunrise events in South Africa. The GW activity was actually lower around sunrise than around sunset in South Africa for many days. It should be noted that the critical frequency of the ionosphere was often lower than the sounding frequency (3.59 MHz) during the night and/or before sunrise. A rapid ionization and decrease of reflection height was observed at sunrise. This could also reduce the number of detected GWs. 3.2. Horizontal Velocities and Propagation Direction of the Analyzed Waves [22] The seasonal variation of the horizontal velocities of GWs over the Czech Republic with error bars is shown in Figure 7a. Figure 7b shows the Czech Republic seasonal variation of azimuthal measurements. Magenta points correspond to measurements from the small triangle, and cyan circles represent the values obtained from the large triangle. Figure 8 shows the diurnal variation of the observed horizontal velocities (Figure 8a) and their azimuths (Figure 8b) over the Czech Republic. Figures 9 and 10 show similar measurements for South Africa. Southward or southeast propagations (180° azimuth), dominated over the Czech Republic during local winter (days 1–50 and 300–365), whereas approximately northward propagations (0°) were observed for local summer (days 100–250). The accuracy of measurements is better during the local summer rather than during the local winter. As for the diurnal dependence, the southward and southeast propagation was observed between 08:00 and 15:00 UT, whereas the northward propagation was only observed in the early morning or late evening
7 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
Figure 5. Scatterplot of diurnal (vertical axis) and seasonal (horizontal axis) distribution of the events analyzed in the Czech Republic. The solid lines show the times of sunrise and sunset at the altitude of 200 km. The dashed lines indicate the times of sunrise and sunset on the ground.
Figure 6. Scatterplot of diurnal (vertical axis) and seasonal (horizontal axis) distribution of the events analyzed in South Africa. The solid lines show the times of sunrise and sunset at the altitude of 200 km. The dashed lines indicate the times of sunrise and sunset on the ground. 8 of 13
A03312
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
Figure 7. (a) Observed horizontal velocities of GWs in the Czech Republic with error bars and (b) their azimuths as a function of day of year. Magenta is used for quantities calculated from the small triangle, and cyan shows velocities and azimuths obtained from the large triangle.
Figure 8. (a) Observed horizontal velocities of GWs in the Czech Republic with error bars and (b) their azimuths as a function of universal time (UT). Magenta is used for quantities calculated from the small triangle, and cyan shows velocities and azimuths obtained from the large triangle. 9 of 13
A03312
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
Figure 9. (a) Observed horizontal velocities of gravity waves (GWs) in South Africa with error bars and (b) their azimuths as a function of day of year.
Figure 10. (a) Observed horizontal velocities of GWs in South Africa with error bars and (b) their azimuths as a function of UT. 10 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
A03312
A03312
Table 1. Statistical Properties of Horizontal Velocities (m/s)a AZ (deg)
sAZ
vI (m/s)
DAZ (deg)
Small Triangle 29.8 28.4 23.5
x 138 5
x 33 32
173.3 172.4 175.5
149 143 165
115.8 117.2 98.0
Large Triangle 30.1 30.1 21.2
x 142 2
x 33 33
171.2 170.7 169.6
156 153 163
130.9 110.3 146.0
South Africa 43.0 30.9 45.5
x 212 21
x 27 35
165.7 155.7 182.4
143 137 149
Mean (m/s)
Median (m/s)
CZ, whole year CZ, Oct–Feb CZ, Apr–Aug
122.2 128.9 106.0
121.2 126.8 103.2
CZ, whole year CZ, Oct–Feb CZ, Apr–Aug
117.7 122.5 102.0
SA, whole year SA, Oct–Feb SA, Apr–Aug
134.9 114.0 153.6
svO (m/s)
a
See section 3.2 for description of given parameters. CZ, the Czech Republic; SA, South Africa.
(Figure 8). Figures 7b and 8b reveal that westward propagation is rare in the Czech Republic. In South Africa, the northeast propagation was mainly observed in the local winter (days 150 and 200), whereas the southwest propagation dominated for the rest of the year (Figure 9). Eastward and southeast propagation was seldom observed over South Africa. For both locations, approximately poleward propagation dominated the observations in the local summer, whereas roughly equatorward propagation was most frequent in the local winter. It should be noted that it is difficult to distinguish between the diurnal and seasonal dependence at both locations because the analyzed waves were observed around sunrise and sunset. Thus sunset or sunrise propagations prevail in the local summer, whereas equatorward propagation directions are typical for daytime observations and local winter. [23] The statistical quantities of the horizontal velocities and their directions are summarized in Table 1, which presents the mean and median values of the observed horizontal velocities, the standard deviations of observed velocities svO, mean azimuths of propagation AZ, standard deviations of these azimuths sAZ, mean values of intrinsic velocities vI, and mean differences between the azimuths of GW propagations and azimuths of the neutral winds (DAZ). The statistical properties were calculated for three periods separately, for the whole year considered, April to August, and October to November. The mean azimuths of propagations were not calculated for the whole year, because of the opposite propagation directions in the summer and winter. [24] The following conclusions can be drawn from Table 1. First, the intrinsic velocities are higher than the observed velocities, which indicates that the GWs propagate against the neutral winds rather than parallel with them. The antiparallel (quasi-antiparallel) propagation can also be seen from the average azimuth difference DAZ between the observed horizontal velocities and neutral winds obtained by the HWM07 model. This is consistent with previous theoretical studies [e.g., Cowling et al., 1971; Sun et al., 2007; Vadas, 2007]. For the Czech republic, the propagation against the winds is more remarkable in the local summer than in the local winter. Second, the observed horizontal velocities are lower in the local summer than in the local winter. This seasonal difference is, however, less than (or comparable with) the standard deviation of the distributions. The difference might be partially attributed to seasonal
variation of neutral wind velocities since the seasonal difference of intrinsic velocities is much smaller, namely, for the Czech Republic. The situation is actually more complicated because there is also diurnal fluctuation of neutral wind velocities. In addition, it should be noted that the HWM07 model does not provide the error estimates. We can obtain some error estimates indirectly, e.g., by changing the altitudes and times for which the neutral winds are calculated. Changing the altitudes of observations by 10 km (maximum error of altitude determination, typically less than 5 km) and times by 15 min (sampling period of ionosonde and typical wave period), we obtain wind velocities that typically differ by 10–15 m/s from their original values. Third, the poleward propagation dominated during the local summer, whereas the equatorward propagation was observed in most cases in the local winter. Fourth, the observed velocities in the Czech Republic and South Africa are similar with typical observed values in the range of 100–150 m/s. 3.3. Observed Periods and Reflection Heights [25] The observed periods of the analyzed waves ranged from 8 to 30 min for both locations. The reflection heights of the 3.59 MHz radio waves estimated from nearby ionosondes were from 150 to 300 km at both locations, most of the reflections were below 250 km. The reflection heights showed diurnal dependence, the lowest reflection heights corresponded to measurements around noon. We did not observe any clear dependence of the observed period on reflection height, however, the occurrence of the longer periods slightly increased with reflection height. 3.4. Horizontal Wavelengths of the Analyzed Waves [26] Next, we will investigate the limit for horizontal wavelength lH for the examples presented in Figures 3 and 4 and for the statistical case. The example in Figure 3 represents extremely strong wave activity, whereas the example in Figure 4 is for low wave activity. For the events A and B in Figure 3, we obtain DfPP 1.9 Hz, h0195 km. We calculated vH 205 m/s for the event A, and vH 175 m/s for the event B. Using equation (6), we obtain lH < 242 km, and lH < 285 km for these values. In the case of Figure 4, we have DfPP 0.5 Hz, h0185 km, and vH 95 m/s we obtain lH < 128 km. The typical numbers obtained in our statistical analysis are DfPP 1 Hz, h0200 km, and vH 130 m/s giving lH < 204 km. From the above analysis, we estimate
11 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
that the horizontal wavelengths of the analyzed waves range from 100 km to 300 km, typically being shorter than 200 km.
4. Discussion and Comparison With Previous Reports [27] The present paper builds on the work of a previous report on horizontal velocities of GWs in the Czech Republic for the period from May 2007 to May 2009 by Chum et al. [2010]. The current paper describes the results obtained over the Czech Republic and South Africa from June 2010 to May 2011. As for the Czech Republic, the currently presented results are fully consistent with the previous report. The analysis of GW horizontal velocities in the ionosphere over South Africa is new. Unlike the previous report, we also extended our analysis by including cases, when the S-shaped signature was only observed on one or two sounding paths, whereas on the other paths, the corresponding disturbances were observed as steep positive slopes (see section 2.4). In turn, we refined the discussion related to the upper limit for the horizontal wavelength to observe the S-shaped pattern by deriving equation (6). We have estimated that the analyzed GWs had horizontal wavelengths of 100–300 km, typically less than 200 km. [28] Our results for GW propagation in the ionosphere are very similar with the GW directions found in the neutral atmosphere in the middle latitude mesopause region by optical measurements [Hecht et al., 2001; Shiokawa et al., 2009; Dou et al., 2010]. These studies also report the poleward preference of propagation in the local summers and prevailing equatorward propagation during the local winters. It should be emphasized that these studies are nighttime optical observations, whereas most of our measurements were performed close to sunset or sunrise, especially during the local summer. During the winter, some of our measurements also correspond to the daytime. The poleward propagation in the local summer is usually associated with the intense convective activity in the equatorial troposphere. On the other hand, in the thermosphere at heights of 200– 300 km, Shiokawa et al. [2009] reported equatorward and westward propagation for all seasons. The periods of reported waves were, however, mostly larger than 30 min, and the horizontal wavelengths were up to 1000 km in several cases. We should also stress that 630 nm emissions used in their studies represent the integral value of the ion (electron) density between 200–300 km [Shiokawa et al., 2009], whereas in the case of Doppler sounding, we get the fluctuations at a specific height for each individual event. In addition, the observations were made in Japan, Indonesia and Australia, not over Europe or South Africa. [29] The similarity between seasonal variation of GW propagation directions reported from optical observations in mesopause region and our results indicates that it is likely that we observed GWs that propagated upward from the lower atmosphere. If so, the neutral winds in the stratosphere and mesosphere can influence the propagation and occurrence rate of the analyzed GWs. For example, if the phase velocity of the wave matches the wind velocity, the frequency of the wave is Doppler shifted to zero, and the wave vanishes. This might be the reason for that we observed only few westward propagating waves in the Czech Republic, and
A03312
southeast propagating waves in South Africa. It should be noted that secondary waves could be generated from wavewave interactions and wave braking [Fritts et al., 2009]. We cannot also exclude that some of the analyzed waves were generated in the thermosphere by passage of the solar terminator. [30] Further measurements, theoretical studies and modeling are necessary to obtain a more accurate picture of GW propagation in the neutral atmosphere and ionosphere at various heights and locations around the globe. We expect that future simultaneous optical and radar measurements at common locations could answer some of the open questions on the propagation of GWs from the mesopause region upward, and on the coupling between the neutral and ionized part of the atmosphere. Simultaneous Doppler radar measurements at different frequencies could provide information on the vertical component of propagation in the ionosphere. We plan to install another collocated Doppler system operating at different frequency in the near future. The vertical propagation of GWs can also be studied from electron density profiles obtained by ionosonde sounding with sufficiently high sampling rates during campaigns [Šauli et al., 2006].
5. Conclusions [31] Most frequently observed velocities of GWs in the ionosphere over the Czech Republic and South Africa are 100–150 m/s. The observed average velocities observed in the local summer are about 25–40 m/s lower than those observed in the local winter. It is necessary to stress that this seasonal difference is less than (or comparable with) the standard deviation of the distributions, and is negligible (Czech Republic) or a bit smaller (South Africa) for average intrinsic velocities obtained from the HWM07 neutral wind model. The GWs propagated roughly poleward at both locations during the local summer. Mainly southeast propagation was observed in the Czech Republic during the local winter, whereas northeast propagation dominated in South Africa during the local winter. It is however necessary to stress that it is difficult to distinguish between the diurnal and seasonal variation of GW propagation directions. Westward propagation was rarely observed in the Czech Republic, whereas eastward to southeast propagation was rare in South Africa. At both locations, the analyzed waves propagated approximately antiparallel to the background neutral winds obtained by the HWM07 model. The averaged velocities and directions are summarized in Table 1. [32] Typical reflection heights of the observed waves measured by nearby ionosondes were 150–250 km, with typical observed periods of GWs ranging from 8 to 30 min. Using the simplifying assumption of the quasihorizontal propagation of the plane wave we estimate that the horizontal wavelengths of the analyzed waves were 100–300 km. [33] Acknowledgments. The support from grants 205/09/1253 and 205/07/1367 from the Grant Agency of the Czech Republic is acknowledged. The South African project was supported by the National Research Foundation (NRF) of South Africa through a Blue Skies Program Grant UID 69868. We also acknowledge the NASA National Space Science Data Center for providing the source code of the HWM07/DWM07 wind model and the NOAA National Geophysical Data Center for access to values of the Ap magnetic index needed for the DWM07 extension of the wind
12 of 13
A03312
CHUM ET AL.: GWS OVER EUROPE AND SOUTH AFRICA
model. We thank Lindsay Magnus and Kessie Govender for help with the installation of the Doppler sounding system in South Africa and for the help with its operation. [34] Robert Lysak thanks the reviewers for their assistance in evaluating this paper.
References Bošková, J., and J. Laštovička (2001), Seasonal variation of gravity wave activity in the lower ionosphere in central Europe, Stud. Geophys. Geod., 45, 85–92, doi:10.1023/A:1021706615038. Chandran, A., D. W. Rusch, A. W. Merkel, S. E. Palo, G. E. Thomas, M. J. Taylor, S. M. Bailey, and J. M. Russell (2010), Polar mesospheric cloud structures observed from the cloud imaging and particle size experiment on the Aeronomy of Ice in the Mesosphere spacecraft: Atmospheric gravity waves as drivers for longitudinal variability in polar mesospheric cloud occurrence, J. Geophys. Res., 115, D13102, doi:10.1029/2009JD013185. Chum, J., T. Šindelářová, J. Laštovička, F. Hruška, D. Burešová, and J. Baše (2010), Horizontal velocities and propagation directions of gravity waves in the ionosphere over the Czech Republic, J. Geophys. Res., 115, A11322, doi:10.1029/2010JA015821. Cowling, D. H., H. D. Webb, and K. C. Yeh (1971), Group rays of internal gravity waves in a wind-stratified atmosphere, J. Geophys. Res., 76(1), 213–220, doi:10.1029/JA076i001p00213. Davies, K., and D. M. Baker (1966), On frequency variations of ionospherically propagated HF radio signals, Radio Sci., 1, 545–556. Dou, X., T. Li, Y. Tang, J. Yue, T. Nakamura, X. Xue, B. P. Williams, and C. Y. She (2010), Variability of gravity wave occurrence frequency and propagation direction in the upper mesosphere observed by the OH imager in northern Colorado, J. Atmos. Sol. Terr. Phys., 72, 457–462, doi:10.1016/j.jastp.2010.01.002. Drob, D. P., et al. (2008), An empirical model of the Earth’s horizontal wind fields: HWM07, J. Geophys. Res., 113, A12304, doi:10.1029/ 2008JA013668. Emmert, J. T., D. P. Drob, G. G. Shepherd, G. Hernandez, M. J. Jarvis, J. W. Meriwether, R. J. Niciejewski, D. P. Sipler, and C. A. Tepley (2008), DWM07 global empirical model of upper thermospheric storm– induced disturbance winds, J. Geophys. Res., 113, A11319, doi:10.1029/ 2008JA013541. Fritts, D. C., and E. V. Thrane (1990), Computation of the ion/neutral density ratio in the presence of wave and chemical effects, J. Atmos. Terr. Phys., 52, 827–834, doi:10.1016/0021-9169(90)90019-J. Fritts, D. C., L. Wang, J. Werne, T. Lund, and K. Wan (2009), Gravity wave instability dynamics at high Reynolds numbers. Part I: Wave field evolution at large amplitudes and high frequencies, J. Atmos. Sci., 66, 1126–1148, doi:10.1175/2008JAS2726.1. Georges, T. M. (1967), Ionospheric effects of atmospheric waves, ESSA Tech. Rep. IER 57–ITSA 54, Inst. for Telecommun. Sci. and Aeron., Boulder, Colo. Hecht, J., R. Walterscheid, M. Hickey, and S. Franke (2001), Climatology and modeling of quasi-monochromatic atmospheric gravity waves observed over Urbana, Illinois, J. Geophys. Res., 106(D6), 5181–5195, doi:10.1029/2000JD900722. Hines, C. O. (1960), Internal atmospheric gravity waves at ionospheric heights, Can. J. Phys., 38, 1441–1481, doi:10.1139/p60-150. Hocke, K., and K. Schlegel (1996), A review of atmospheric gravity waves and travelling ionospheric disturbances: 1982–1995, Ann. Geophys., 14, 917–940. Kelley, M. C. (2009), The Earth’s Ionosphere: Plasma Physics and Electrodynamics, 2nd ed., Elsevier, Amsterdam. Kelley, M. C., and C. A. Miller (1997), Electrodynamics of midlatitude spread F: 3. Electrohydrodynamic waves? A new look at the role of electric fields in thermospheric wave dynamics, J. Geophys. Res., 102(A6), 11,539–11,547, doi:10.1029/96JA03841. Kirchengast, G., K. Hocke, and K. Schlegel (1996), The gravity wave-TID relationship: Insight via theoretical model-EISCAT data comparison, J. Atmos. Terr. Phys., 58, 233–243, doi:10.1016/0021-9169(95)00032-1.
A03312
Laštovička, J. (1999), Solar activity effects on gravity wave activity inferred from radio wave absorption, Stud. Geophys. Geod., 43, 107–128, doi:10.1023/A:1023318326855. Laštovička, J. (2006), Forcing of the ionosphere by waves from below, J. Atmos. Sol. Terr. Phys., 68, 479–497, doi:10.1016/j.jastp.2005.01.018. Lay, E. H., and X.-M. Shao (2011), High temporal and spatial-resolution detection of D-layer fluctuations by using time-domain lightning waveforms, J. Geophys. Res., 116, A01317, doi:10.1029/2010JA016018. Lee, C. C., Y. A. Liou, Y. Otsuka, F. D. Chu, T. K. Yeh, K. Hoshinoo, and K. Matunaga (2008), Nighttime medium-scale traveling ionospheric disturbances detected by network GPS receivers in Taiwan, J. Geophys. Res., 113, A12316, doi:10.1029/2008JA013250. Medeiros, A. F., M. J. Taylor, H. Takahashi, P. P. Batista, and D. Gobbi (2003), An investigation of gravity wave activity in the low-latitude upper mesosphere: Propagation direction and wind filtering, J. Geophys. Res., 108(D14), 4411, doi:10.1029/2002JD002593. Miller, C. A. (1997), Electrodynamics of midlatitude spread F: 2. A new theory of gravity wave electric fields, J. Geophys. Res., 102(A6), 11,533–11,538, doi:10.1029/96JA03840. Miller, C. A., W. E. Swartz, M. C. Kelley, M. Mendillo, D. Nottingham, J. Scali, and B. Reinisch (1997), Electrodynamics of midlatitude spread F: 1. Observations of unstable, gravity wave–induced ionospheric electric fields at tropical latitudes, J. Geophys. Res., 102(A6), 11,521–11,532, doi:10.1029/96JA03839. Nakamura, T., T. Tsuda, H. Miyagawa, Y. Matsushita, H. Fukunishi, Y. Takahashi, and Y. Yamada (1998), Propagation directions of gravity wave patterns observed in OH CCD images during the SEEK Campaign, Geophys. Res. Lett., 25(11), 1793–1796, doi:10.1029/98GL01064. Perkins, F. (1973), Spread F and ionospheric currents, J. Geophys. Res., 78(1), 218–226, doi:10.1029/JA078i001p00218. Šauli, P., P. Abry, J. Boška, and L. Duchayne (2006), Wavelet characterisation of ionospheric acoustic and gravity waves occurring during the solar eclipse of August 11, 1999, J. Atmos. Sol. Terr. Phys., 68, 586–598, doi:10.1016/j.jastp.2005.03.024. Shiokawa, K., Y. Otsuka, and T. Ogawa (2009), Propagation characteristics of nighttime mesospheric and thermospheric waves observed by optical mesosphere thermosphere imagers at middle and low latitudes, Earth Planets Space, 61, 479–491. Smith, S., J. Baumgardner, and M. Mendillo (2009), Evidence of mesospheric gravity-waves generated by orographic forcing in the troposphere, Geophys. Res. Lett., 36, L08807, doi:10.1029/2008GL036936. Sun, L., W. Wan, F. Ding, and T. Mao (2007), Gravity wave propagation in the realistic atmosphere based on a three-dimensional transfer function model, Ann. Geophys., 25, 1979–1986, doi:10.5194/angeo-25-1979-2007. Suzuki, S., K. Shiokawa, K. Hosokawa, K. Nakamura, and W. K. Hocking (2009), Statistical characteristics of polar cap mesospheric gravity waves observed by an all-sky airglow imager at Resolute Bay, Canada, J. Geophys. Res., 114, A01311, doi:10.1029/2008JA013652. Taylor, M. J., J. M. Jahn, S. Fukao, and A. Saito (1998), Possible evidence of gravity wave coupling into the mid-latitude F region ionosphere during the SEEK Campaign, Geophys. Res. Lett., 25(11), 1801–1804, doi:10.1029/97GL03448. Vadas, S. L. (2007), Horizontal and vertical propagation and dissipation of gravity waves in the thermosphere from lower atmospheric and thermospheric sources, J. Geophys. Res., 112, A06305, doi:10.1029/ 2006JA011845. Xiao, S.-G., Z. Xiao, J.-K. Shi, D.-H. Zhang, X.-S. Feng, Y.-Q. Hao, and W.-Q. Huang (2009), Observational facts in revealing a close relation between acoustic-gravity waves and midlatitude spread F, J. Geophys. Res., 114, A01303, doi:10.1029/2008JA013747. R. Athieno and L. A. McKinnell, South African National Space Agency Space Science, PO Box 32, Hermanus, 7200 South Africa. J. Baše, J. Chum, and F. Hruška, Department of Upper Atmosphere, Institute of Atmospheric Physics, Bocni II/1401, 14131 Prague 4, Czech Republic. (
[email protected]) D. Burešová, J. Laštovička, and T. Šindelářová, Department of Aeronomy, Institute of Atmospheric Physics, Bocni II/1401, 14131 Prague 4, Czech Republic.
13 of 13
