JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. D22, PAGES 28,769-28,773, NOVEMBER 27, 1998 Seasonal variation of the turbopause' One year of turbulence investigation at 69øN by the joint University of TromsoUniversity of Saskatchewan MF radar C. M. Hall Dept. of Physics, University oftromso, Tromso, Norway A. H. Manson and C. E. Meek Institute of Space and Atmospheric Studies, University of Saskatchewan, Saskatoon, Canada Abstract. Following upgrades to the University of TromsoUniversity of Saskatchewan MF radar, located in northern Norway at 69øN, 19øE, we have been able to complete a full calendar year of estimates of mesospheric turbulent intensity. The results represent the first such study using continuous measurements from this region, and temporal and height variations are satisfyingly in accordance with expectation. Since in the past there has been a degree of disagreement as to absolute intensities, we briefly compare our results with some from totally independent methods. The resulting dissipation rates, to be regarded as maxima for the turbulent dissipation, are used to identify an upper limit to the turbopause. The Arctic turbopause appears to exhibit an annual variation, being lower in the summer, in agreement with Danilov et al. [ 1979] but refuting Blum and Schuchard [1978]. 1. Introduction of signal in January and December. The signal fading times are converted into estimates of the kinetic energy dissipation rate ' The joint University of TromsoUniversity of Saskatchewan (as distinct from the true turbulent energy dissipation rate ) MF radar, located in northern Norway at 69øN, 19øE was, during according to the method described by Hall et al. [ 1998a]. Value ' the winter of 19961997, refurbished to some degree. This has may be interpreted as an upper limit to as it may inevitably meant that the data quality, particularly its reliability, and include some contributions from short-period gravity waves. accessibility for ease of analysis has been greatly improved and However, this aspect is addressed to some degree in section 2 which has in turn facilitated the acquisition of a complete calendar where we attempt to estimate errors. year (1997) of estimates of turbulent intensity in the mesosphere. Prior to this work no continuous monitoring of mesospheric During the autumn of 1997 the transmitter antenna array was turbulence had been performed at this location. Only intermittent refurbished, resulting in a slightly altered beam shape, and in situ and incoherent-scatter-based estimates existed, summarized subsequently, changes were made to the receiver software and by Hall et al. [1997] and Hall and Hoppe [1997], respectively; firmware. Nevertheless, as we shall see, the clear seasonal thus these results represent unprecedented observations 69 ø N. variation apparent in our data is unlikely to be an artefact. Table 1 Finally, we use the energy dissipation rates we derive to estimate gives the specification of the radar for the period prior to the the turbopause height, following the method by Hall et al. modifications; the transmitter antenna beam width being greater [1998b]. The resulting seasonal variation is characterized by a later in the last 3 months of the year. The salient points of Table 1, lower turbopause in summer than in winter. This result is in for the purposes of this study, are the final height and time agreement with Danilov et al. [ 1979]. resolutions of 3 km and 5 rain, respectively. Winds are determined by the spaced antenna method elegantly described by 2. Method Hocking [1997] using the full correlation analysis (FCA) described by Briggs [1984]. Indeed, Hocl 'ng [1997], comparing The radar illuminates the mesosphere and power is reflected the various methods of data analysis, suggests that FCA is less from scattering structure subsequently forming a diffraction prone to beam-broadening effects than the Doppler-broadening pattern on the ground that moves at twice the speed of the method, and so the antenna modifications are less likely to affect scatterers, as described by Hocl 'ng [1997]. This reflected power our results, thanks to our analysis strategy. The radar operates is detected by spaced antennae, the individual signals being unattended round the clock so that 288 profiles of signal fading autocorrelated and cross-correlated using the FCA. The time are available each day. Not each profile is complete, autocorrelations indicate the rate at which the scattering structures however, because of either poor signal-to-noise ratios (typical at are being dissipated, characterized by the "fading time" night in winter) or due to total reflection of the signal (during Subsequently, velocity fluctuations v', appropriate to an observer periods of high auroral activity); indeed, we see evidence for lack moving with the background wind, are computed according to Briggs [1984], at a height resolution of 3 km and time resolution Copyright 1998 by the American Geophysical Union. Paper number 1998JD200002. 0148-0227981998JD200002509.00 28,769 of 5 min via v'= Z In 2 (1)
28,770 HALL ET AL.' SEASONAL VARIATION OF THE TURBOPAUSE Table 1. MF radar system features. background wind; we have not attempted to compensate for departures from this. Parameter Value Errors in FCA analyses from individual soundings are difficult to estimate. To overcome this, we have taken advantage of the Geographic coordinates 69.58øN, 19.22øE Operating frequency 2.8 MHz large numbers of soundings in our data. For each day we perform Pulse repetition frequency Hz a spectral analysis of e' at each height and identify noise floors at Peak transmitter power N20 kw the high-frequency ends of the power spectra. To clarify this, we Transmitter antenna b earn width 17 ø detect the noise floor at high frequencies where it becomes Receiver antenna three spaced inverted-v dipales Height resolution 3 km discernible above the diminishing signal; this noise is then Postintegration time 5 min assumed to be white noise spanning the entire spectrum. From individual case studies we found this to be typically from periods of 20 min or shorter. We do not include explicit examples of these spectra here; indeed, in the routine analysis of 365 days of data, the noise floor is estimated automatically as the average power where 7, is the radar wavelength. Thereafter, we assume a total between 20-rain periods and the Nyquist frequency. The noise velocity fluctuation [Hall et al., 1998a] and use floor from each day is subsequently used to estimate the error in e' = 0.8v':r (2) e'. Note that this approach does not use variances and deem them to be experimental errors. Instead, we use the presence of a noise where TB is the Brunt-Vfiisalfi period in seconds. Note that we floor in the spectra of the data, where buoyancy subrange power recognize that the spatial scale inherent in the experiment cannot law dependence is expected. The derived errors thus also give an preclude buoyancy-scale fluctuations, because the outer scale of indication of the high-frequency gravity wave contribution in e', turbulence LB can be expected to be as little as 200 m in the i.e., that which differentiates it from e. It is importanto note the mesasphere which is why we choose to introduc e' to distinguish difference between the mere daily standard deviations and the our measurement from e. Furthermore, equation (2) implicitly noise-floor error we employ here. Figure 1 shows the two kinds of assumes them is no motion of the observer relative to the errors obtained using our data for 88 km. Recall that the standard deviation that a whole day's data includes changes in the energy dissipation induced by larger-scale dynamics right up to the diurnal tide and to a certain degree, even longer timescales. We standard deviation in 88 km E' see that the standard deviation is considerably larger than the 60... '... '... '... high-frequency noise-floor-derived error. Furthermore the standar deviation exhibits a distinct seasonal variability, perhaps 5O not surprisingly considering the seasonal differences in gravity 40 wave propagation. There is a similar signature in the noise-floor error showing that it still reflects larger-scale dynamics to some 3o extent but not to the same degree as the standardeviation. 20 i2 10 8 6 200 300 day number (noise floor) error in 88 km('... i... i... i...... i... i... i... 1 oo 200 300 doy number Figure 1. (top) Daily standard deviations of ' for 88 km as a function of season; (bottom) daily noise-floor-derived errors in g' for 88km as a function of season. A 30-day boxcar smoothing has been apphed to each time series (the spiky nature of the time series end points is a failure of the smoothing). 3. Basic Results Using the method referred to above, we have derived an estimate of the energy dissipation rate as a function of both month and height, and this is shown in Figure 2. Because of low photoionization in the mesasphere during January and December, the data quality is fairly poor at the very beginning and end of the plot. Nevertheless, we note the following features: (1) low turbulent levels below 80 km in the summer, giving an annual variation in the mid-mesosphere; (2) low turbulent levels around the equinoxes in the upper mesasphere (winter)lower thermasphere (summer) giving a semiannual variation. The recent statistics presented by both Hall and Hoppe [ 1997] and Hall et al. [1997] show similar tendencies and give considerable credibility to the estimation of from MF signal fading times. In the figure, a degree of smoothing has been applied to improve clarity. In order to relate these results to the horizontal dynamics that determines the stability of the upwardly propagatingravity wave field and hence turbulence production, we show the corresponding zonal (Figure 3) and meridional (Figure 4) winds selected at 90 km. These parameters are also products of the FCA. We see little seasonal variation in the meridional component. The zonal wind's autumn reversal coincides with the demise of the summer turbulent levels. In the spring, however, the upper mesasphere turbulence dies out at the end of February, over a month before the spring wind reversal. To further illustrate this, we have computed the total horizontal wind speed and plotted it on the
HALL ET AL.' SEASONAL VARIATION OF THE TURBOPAUSE 28,771 week-overacje meridional (nss + N) at 88kin ' I ' ' ' I I I I 40 North 85 20 75-2O -4O _ South 70 Figure 2. Energy dissipation rate s' as a function of season and height. The month numbers indicate the month endings. i I 2 4 6 8 10 12 month (1 : 51st JonL:ory etc) Figure 4. Meridional wind for 90 km as a function of season and corresponding to Figure 2. Positive values are southward. same axes as the energy dissipation rate in Figure 5. Here we see present results from the same geographic location as the ME radar the energy dissipation rate' s semiannual variation at 90 km (solid but use vertical velocities measured by the European Incoherent line) and a similar, though less pronounced variation in the total Scatter Radar (EISCAT) and thus a completely independent wind (pluses). Winds and turbulence have been investigated instrument and methodology. Characteristic wavenumbers are earlier by these methods. For example, Manson and Meek [ 1991 ] derived for the height range of the measurement and, considenng have compiled climatologies for this latitude and for 52øN. the poorer data yield in the lower half of this range, may be Comparisons with other instruments and models have been made thought representative of 85 km altitude. It has been suggested, [e.g., Manson et al., 1991]. Smaller scales have been addressed however, that such a major contribution to energy dissipation by, for example, Manson and Meek [ 1987]. comes from the top of the measurement regime (M. Ishii, private The absolute turbulent intensities in the mesapause have long communication, 1997) that the s are representative of 90 kin. Both been a question of debate, not least between exponents of in situ the magnitudes and the semiannual signature support this. methods and those using ground-based instruments. Danilar and Considering the assumptions implicit in each method (e.g., an Kalgin [1996] have discussed this aspecto some degree. We assumed representative Richardso number, etc.), the agreement prefer not to launch into an exhaustive review of such is remarkable and may be described as either very good or measurements here, although we do indeed note that the levels of coincidental depending on one's scepticism. Other s turbulenc estimated by our method are generally higher than determinations have been recently summarized by Hall and those determined by rocket soundings. Hall and Hoppe [1997] Hoppe [1998] who include in situ data fromltibken [1997]. 4O week-average zonal (ms, -t-e)at aak I I I ' ' ' I I 0 ' I ' ' ' I I I I._. total wind at 90 kr'n (ms) ½' at 90 km {mwkg) 2O 10-40 West 1 ', I,,, I,,, I,,, I,,, I,,, I 2 4 6 8 10 12 month (1 = 51st January etc) 2 4 6 8 10 2 month (1 = 51s January etc) Figure 5. Again, on the same time axis as the preceding figures, s' (solid line) and total horizontal wind (pluses), both for 90 km. Figure 3. Zonal wind for 90 km as a function of season and Error bars are shown for s', these being obtained from the daily corresponding to Figure 2. Positive values are eastward. noise floor.
...,.,, 28,772 HALL ET AL.' SEASONAL VARIATION OF THE TURBOPAUSE 4. Derived Results approximation adjustments of virtual to real height have thus been applied to the profiles presented in this paper. Perhaps the best Hall et al. [1998b] investigated the feasibility of using the estimates of the turbopause height (at least with sufficient energy dissipation rate profiles to estimate the turbopause altitude continuity to examine any seasonal variation) were published as hr. In the initial work, only a limited dataset was available, but early as 1979 for Heiss Island, a little over 0 km to the now we are able to present a whole season and, furthermore, northeast of our system and at 80øN [Danilar et al., 1979]. We including error estimates. We identify an energy dissipation rate have taken the tabulated data from these authors and produced the corresponding to viscous effects alone, gmin = vcob:0.3, where v is monthly averages included in Figure 7. Here, as described by the the kinematic viscosity and cob is the Brunt-ViiisCtlii frequency authors, is an indication of high winter low summer turbopause [Hall et al., 1998b]. Defining the turbopause as the intersection of similar to our observation (note that there are actually two the profiles of turbulent and molecular energy dissipation rates independent midsummer soundings in the Danilar et al. [1979] [again, Hall et al., 1998b], we see in Figure 6 how ht has been statistics, each giving the same turbopause height). The Heiss identified for nine selected days distributed throughout the year. Island measurements from the 1970s disagree with ours somewhat Doing this for all available days and applying a 1-week-wide in exact height. Although this may be attributable to inappropriate boxcar smoothing, we arrive at Figure 7. Again, we see a seasonal assumptions in our method, there can be interannual andor trend, this time annual (the shorter-period wave is most likely the geographic differences. More important, however, is that Danilar quasi-16 day Rossby mode; we are focusing on seasonal et al. [1979] cited Blum and Schuchard [1978] (although variability in this paper however). Such determinations of the Schuchard and Blum [1978] might have been more appropriate) turbopause are somewhat unusual, most results coming from in whose modeling work suggested quite the opposite, at least for situ measurements. This may be partly due to the problems of 60ON. obtaining partial reflections from heights above 90 km. If ionization is strong, the radio wave may be totally reflected; 5. Discussion and Conclusions however, if it is indeed partially reflected, the group retardation causes the gated height to correspond to a somewhat higher true We have presented a whole year's investigation of mesaspheric altitude. This problem, well known from the earliest ionosonde turbulence. This demonstrates the usefulness of the joint soundings, was readdressed by Hall [1998], and first University of TramsoUniversity of Saskatchewan MF radar as an end of January end of February end of March end of April!! IO0 90,-'.., ""'...,- Wkg 1 end of May end of June end of July end of Aucjust 12o...,....,. 1 oo 90,,,.,.,,- Wkg I 80 1 1 end of September end of October end of November end of December 90 z IO0 90 80 1 1 1 Figure 6. Profiles of s' (solid lines) for selected days of the year. Also included are profiles of the energy dissipation attributable to molecular diffusivity (dashed lines). Intersections of these profiles indicate the turbopause hr.
HALL ET AL.: SEASONAL VARIATION OF 2t-IE TURBOPAUSE 28,773 130 4_17 estimates of the turbopause height, extending the work by Hall et a!., [1998b]. If e' is considered an upper limit for s, the derived turbopause heights are then similarly upper limits. Nevertheless, there is an indication of a lowering of the turbopause in the summer months which is in agreement with observations from Heiss Island (80øN) by Danilov et al. [1979] and, at the same time, casting doubt on the semiempirical model of Blum and Schuchard [ 1978]..+4 +2 Acknowledgments. The authors are indebted to the technical staff of the Auroral Observatory, Tromso, for both regular maintenance of and refurbishing work on the MF radar. Thanks go to the reviewers of this paper who suggested important improvements to the original manuscript. I I I I 4 6 8 10 month (1 = 31s Jonuory etc) References Blum, P.W., and K.G.H. Schuchard, The role of eddy turbulence for long period variations of upper atmosphere density, Space Res. XVIIL 18, 191-194, 1978. Figure 7. Value ht as a function of season. Again the month axis Briggs, B.H., The analysis of spaced sensor records by correlation annotation indicates month endings. Error bars correspond to the techniques, Handb. MA?, 13, 166-186, 1984. highest and lowest intersections obtained using the errors in s', as Danilov, A.D., and U.A. Kalgin, Eddy diffusion studies in the lower illustrated by Figure 6. The pluses indicate monthly averages of ht thermosphere, Adv. Space Res., 17, (11)17-(11)24, 1996. from Danilov et al. [1979] corresponding to our estimates, but for Danilov, A.D., U. Kalgin, and A. Pokhunkov, Variation of the turbopause level in the polar regions, Space Res. XIX, 83, 173-176, Heiss Island in the 1970s. The digits to the right of each plus 1979. indicate the numbers of soundings used in the averaging. Fritts, D.C., Gravity wave saturation in the middle atmosphere: A review of theory and observations, Rev. Geophys., 22, 275-308, 1984. Hall, C.M., Virtual to true height correction for high latitude MF radar, Ann. Geophys., 16, 277-279, 1998. Hall, C.M. and U.-P. Hoppe, Characteristic vertical wavenumbers for the instrument for monitoring the small-scale dynamics of this part of polar mesosphere, proc. 13th ESA sympostum., 499-504, 1997. the atmosphere and furthermore points to a future of investigation Hall, C.M., T.A. Blix, E.V. Thrane, and F.-J. Ltibken, Seasonal variation into interannual variation. of mesospheric turbulent kinetic energy dissipation rates at 69øN, The features highlighted by the investigation of turbulent Proc. ESA Symp., 13, 505-509, 1997. intensities are (1) semiannual variation in the upper Hall, C.M., and U.-P. Hoppe, Estimates of turbulent energy dissipation rates from determinations of characteristic vertical wavenumber by mesospherelower thermosphere and (2) annual variation in the EISCAT, Geophys. Res. Lett., in press 1998. mid-mesosphere. We have seen the clear signature of the Hall, C.M., C.E. Meek, and A.H. Manson, Turbulent energy dissipation semiannual oscillation (SAO) in the total wind, although we did rates from the University of Tromso University of Saskatchewan MF not present data from other heights (turbulence observations being radar, d. Atmos. Solar-Terr. Phys., 60, 437-440, 1998a. Hall, C.M., A.H. Manson, and C.E. Meek, Measurements of the arctic the subject of this study.) The interaction between the horizontal turbopause, Ann. Geophys., 16, 342-345, 1998b. wind and the upwardly propagating gravity wave field is Hocking, W.K., Strengths and limitations of MST radar measurements of described by Lindzen [1981]. Essentially, the background wind middle-atmosphere winds, Ann. Geophys., 1,5, 1111-1122, 1997. through the troposphere and stratosphere has already determined Lindzen, R.S., Turbulence and stress owing to gravity wave and tidal the gravity wave flux arriving at the heights of our observations. breakdown, d. Geophys. Res., 86, 9707-9714, 1981. Lfibken, F.-J., Seasonal variation of turbulent energy dissipation rates at Of course, the mean background wind must be combined with the high latitudes as determined by in situ measurements of neutral density tidal component to test for gravity wave stability, but fluctuations, d. Geophys. Res., 102, 13,441-13,456, 1997. nevertheless, we have combined the overall zonal and meridional Manson, A.H., and C.E. Meek, Small-scale features in the middle components from Figures 2 and 3 to produce the total (i.e., atmosphere wind field at Saskatoon, Canada (52øN, 107øW): An analysis of MF radar data with rocket comparisons, d. Atmos. Sci., 44, absolute value) wind in Figure 5. In the mesosphere, around the 3661-3672, 1987. reversals in the zonal wind, the gravity wave horizontal phase Manson, A.H., and C.E. Meek, Climatologies of mean winds and tides velocities have a better possibility to exceed the total (Eulerian) observed by medium frequency radars at Tromso (70øN) and parcel velocities (the background wind being near zero), and Saskatoon (52øN) during 1987-1989, Can. d. Phys., 69, 966-975, 1991. therefore propagation can continue without convective instability Manson, A.H., et al., Comparisons between satellite-derived gradient winds and radar derived winds from the CIRA-86, d. Atmos. Sci., 48, [Fritts, 1984]. Thus s' exhibits a semiannual oscillation (SAO) 411-428, 1991. with minima around the zonal wind reversals or alternatively the Schuchard, K.G.H., and P.W. Blum, Global thermospheric models of corresponding total wind minima. neutral density, exospheric temperature and turbopause height, Space Inclusion of error estimates in these kinds of measurements is Res.,¾1III, 18, 195-198, 1978. somewhat unusual and our method of using the spectral noise floor seems a more appropriate approach than using the overall C. M. Hall, Dept. of Physics, University of Tromso, 9037 Tromso, standardeviation. When employing daily or even, for example, Norway. (e-mail: chris.hall phys.uit. no) 6-hourly standar deviations, it is impossible to separate the error H. Manson and C. E. Meek, Institute of Space and Atmospheric from the variability induced by gravity wave, tidal, and planetary Studies, University of Saskatchewan, Saskatoon, Saskatchewan, Canada. wave dynamics. (e-mail: manson dansas.usask. ca; meek dansas.usask. ca.) Subsequently, using the profiles of energy dissipation rate and assuming them to be representative of, we are able to make (Received August 16, 1998; accepted August 17, 1998.)