Available online at www.sciencedirect.com
Advances in Space Research 42 (2008) 1253–1258 www.elsevier.com/locate/asr
Meteor radar temperatures over Collm (51.3N, 13E) G. Stober b
a,*
, Ch. Jacobi a, K. Fro¨hlich a, J. Oberheide
b
a Institute for Meteorology, University of Leipzig, Stephanstr. 3, 04103 Leipzig, Germany Atmospheric Physics Group, Wuppertal University, Gauss Strasse 20, 42119 Wuppertal, Germany
Received 17 October 2006; received in revised form 4 October 2007; accepted 23 October 2007
Abstract The method of obtaining absolute temperature estimates by measuring the ambipolar diffusion coefficient with meteor radars in the mesopause region is basically well known. However, there is still a need to refine and adjust the background temperature gradient model which is necessary to calculate the temperature values. Therefore, a detailed comparison with independent temperature measurements is necessary to evaluate the performance of the method and to obtain more information about the temperature gradient. Recent studies provide some evidence that the impact of the gradient model on temperature estimates affects the absolute temperatures, but that it is of minor importance for wave analysis. This paper focuses on a detailed evaluation of the meteor radar temperatures by comparing them with SABER satellite and OH-emission mesopause region temperatures. The seasonal variation of the observed temperatures is well reproduced by the COMMA general circulation model. 2007 COSPAR. Published by Elsevier Ltd. All rights reserved. Keywords: Meteor radar; Mesopause temperatures; Temperature tides
1. Introduction As meteoroids enter the Earth’s atmosphere, they form a cylindrical plasma trail which reflects the transmitted radar energy and which can be used to determine atmospheric parameters from the reflected signal. Since August 2004 there is an all-sky meteor radar (Hocking et al., 2001) in operation at Collm Observatory (51.3N, 13E). The operation frequency is 36.2 MHz with a 2144 Hz pulse rate, a 2 km height resolution and an angular accuracy of approximately 2. The result is a near Gaussian height distribution of the meteor count rates with a centroid altitude near 90 km (Fig. 1). The expansion of the meteor plasma trail due to ambipolar diffusion results in an exponential decay of the reflected energy for so-called underdense meteors. The ambipolar diffusion coefficients enable us to calculate temperature. The first step, when temperatures are measured,
*
Corresponding author. E-mail address:
[email protected] (G. Stober).
is to evaluate the performance of the radar with respect to temperature analysis using other measurements that are located close-by. To this aim, GRIPS OH-emission temperatures (Offermann et al., 2006) from the Wuppertal University (51.3N, 7E) as well as SABER satellite temperatures are used. Finally, the data of a global circulation model (COMMA-LIM model, Fro¨hlich et al., 2003, 2005) are compared to the radar temperatures. The idea is to show the range of time scales, from tidal to monthly means, the meteor radar temperatures are able to cover and to point out the performance and limits.
2. Theory The analysis procedure, as described e.g. in Hocking (1999), Hocking et al. (2001) leads to the following equation for the temperature Tmeteor; T meteor
0273-1177/$34.00 2007 COSPAR. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.asr.2007.10.018
dT mg þ ¼ Sm 2 ; dz k
ð1Þ
1254
G. Stober et al. / Advances in Space Research 42 (2008) 1253–1258 110 105
altitude / km
100
Chi2 R2
= 396.09619 = 0.99561
y0 xc w A
3.23334 89.66398 9.1482 10019.10837
±3.85318 ±0.05174 ±0.11817 ±135.70787
95 90 85 80 75 70 0
100
200
300
400
500
600
700
800
900
meteor counts Fig. 1. Height distribution of meteor echoes using 2 km vertical gates and one day of data. The diagram includes approximately 5500 meteors.
which will be the basis of the further temperature analysis. In Eq. (1) the following known parameters are used: the acceleration due to gravity at 90 km (g = 9.47 m/s2), k as the Boltzmann constant and m being the mass of an average air molecule (m = 4.749 Æ 1026 kg). Applying the binning and the bias adjustment (Hocking et al., 1997) to the data allows to estimate the vertical logarithmic deviation of the ambipolar diffusion coefficient Sm. The connection between the exponential decay time s1/2 and the ambipolar diffusion coefficient Damb is given by s1=2 ¼
k2 ln2 : 16p2 Damb
ð2Þ
The parameter k is the transmitting wavelength of the radar. From Eq. (2) the vertical logarithmic deviation of the ambipolar diffusion coefficient Sm can be estimated; 1 d / lnDamb : S m dz
ð3Þ
A scatter plot of height vs. logarithmic inverse decay time and fitting the slope Sm illustrates this procedure (Fig. 2).
For the least square fit we consider only the error in height. During our experimental stage, we also used least square fitting routines considering both errors. But the results were completely unstable. The reason for this is the problem to derive the error of the quantity ln(1/s1/2), which shows no typical distribution. This makes it difficult to estimate the real error for this quantity from statistics and a direct measurement of the inaccuracy is not available. The last unknown parameter in Eq. (1) is the temperature gradient (dT/dz), which has to be taken from a model. Singer et al. (2003) compared several measurements from LIDAR, rockets and OH-emission measurements to create a model. But an error of 1 K/km in the gradient model leads to an error of 10 K in absolute temperature. Therefore, the comparison of absolute temperatures of the meteor radar with other data is always connected with an estimate of the variability of the temperature gradient. This variability becomes even more important for the calculation of the temperature tides, because the diurnal tide with short vertical wavelength is sometimes significant. We use an updated temperature gradient model, which mirrors a typical seasonal variation (Fig. 3). The update was necessary due to the small offsets during the equinoxes in comparison to the SABER data (3–7 K). In the case of the diurnal tide, Hocking and Hocking (2002) includes in their analysis of the temperature tide an amplitude and phase correction using the vertical wavelength of the wind tide and assuming an energy (Hm) and pressure scale height (H) with Hm = H = 7 km. This method provides a significant improvement in deriving diurnal tidal amplitudes and phases if the vertical wavelength is well known. 3. Results 3.1. Collm meteor radar and SABER satellite temperatures The SABER (Sounding the Atmosphere using Broadband emission Radio-metry) instrument is an infrared radiometer with 10 channels from 1.27 to 16.9 lm. The carrier 0.50
115
altitude / km
105 100
0.25 0.00
temperature gradient / K/km
110
Linear Regression z = S m * X + c(T.z) Parameter Value Error ------------------------------c(T.z) 73.25026 0.26519 Sm 14.15889 0.22244
95 90 85 80
-0.25 -0.50 -0.75 -1.00 -1.25 -1.50 -1.75 -2.00 -2.25 -2.50 -2.75 -3.00 -3.25
75
-3.50 -3.75
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
decay time ln(1/τ1/2) Fig. 2. Height vs. logarithmic inverse decay time of one day of data.
30
60
90
120
150
180
210
240
270
300
days / d Fig. 3. The temperature gradient model used.
330
360
G. Stober et al. / Advances in Space Research 42 (2008) 1253–1258
for the instrument is the TIMED (Thermosphere, Ionosphere, Mesosphere Energetics and Dynamics) satellite with a 625 km circular orbit inclined 74 to the equator. The instrument measures temperatures and chemical species. The routine temperature retrieval provides profiles from 10 to 105 km with 2 km height resolution and an along track resolution of 400 km. The aim here is to compare daily mean radar temperatures with SABER (mean of ascending and descending orbit node data) zonal mean temperatures at 50N (the northernmost long term available data), which leads to a general problem. The zonal mean temperature smoothes out all waves along a certain latitude. This results in a relative root mean square error of 10–12 K, which is of the same order of magnitude as the expected radar error. The second problem is that the satellite shows an aliasing depending to the TIMED yawcycle. The satellite changes its turning points every two months. There are a northward looking case (83N to 52S) and a southward looking case (52N to 83S). During the southward looking yaw-cycle this difference becomes small, which leads to an aliasing of 5–7 K, due to tidal activities. In the northward looking yaw-cycle the aliasing is smaller (1–3 K), because of a local time difference of 8– 9 h between ascending and descending nodes. However, the satellite should be more accurate in absolute temperature than the radar. As Fig. 4 illustrates, both temperatures are of the same absolute value range and obviously there is no systematic offset, after our temperature gradient adjustment in April and September in the order of 0.3 to 0.7 K/ km (Fig. 3). It should be noted that stratospheric warmings in winter can cause large errors of the radar temperatures, if there is a shift in temperature gradient, which is not reflected by our model. 230
1255
Finally we calculated the root mean square errors for the daily radar data set and SABER with 8.4 K and for the 60 day-smoothed radar temperatures and the satellite with 6 K. In both cases the systematic offset was 1 K, which is negligible under consideration of the errors. The larger difference for the daily mean radar temperatures are explainable by two thinks maybe simply the radar error due to the use of an empirical temperature gradient or the planetary wave activity, which is detected by the radar but smoothed by the SABER zonal mean temperatures. 3.2. Collm meteor radar and OH-emission The OH temperatures in Fig. 4 represent a weighted average over a layer centered around 87 km and with a FWHM (Full Width at Half Maxima) of 8 km, which is comparable to the radar with a peak height of 90 km and a FWHM of 9 km. A difficulty in comparing meteor radar temperatures with OH-emission data is that only nighttime values from GRIPS (Ground Based Infrared P-Branch Spectrometer) are available. As Fig. 4 shows, the OH temperatures during winter are significantly higher (15–20 K) than the daily mean radar temperatures. This difference can be partly explained by a bias of 7 K to the TIMED/ SABER data (Oberheide et al., 2006). Subtracting this bias reduces the difference to 8–13 K. Finally it has to be considered that the OH-emission refers to a level 3 km below that of the radar average temperature. Together with our temperature gradient model the difference should be in the range of 8–11 K for the winter and almost zero in summer. We conclude from this that our gradient model is in an overall agreement with the OH data that in turn were not used for the model adjustment. Finally, the tidal
225
Radar (51.3°N, 13°E) radar 60-days smoothed SABER (zonal mean 50°N) 90 km COMMA (zonal mean 52°N) 90 km
220
GRIPS OH (night measurements) 87 km
215 210
temperature / K
205 200 195 190 185 180 175 170 165 160 155 0
25
50
75
100
125
150
175
200
225
250
275
300
325
350
time in days 1=01.01.2005 Fig. 4. Comparison of 2005 data from OH-emission night values, SABER zonal mean 50N data, the COMMA-LIM model for 52N and the meteor radar measured daily mean temperatures.
1256
G. Stober et al. / Advances in Space Research 42 (2008) 1253–1258
activity may explain the remaining difference of 2–5 K. Hence, data measured during the same time interval during the course of one day have been compared, to minimize these tidal effects. Both instruments should detect the same tidal activity because of their equal latitude and only 7 difference in longitude. The latter introduces a tidal phase difference of 30 min only, which is negligible because time intervals of at least 4 h are compared. The high meteor flux during the night allows a good time resolution and the remaining gradient problem was minimized taking the daily mean temperature gradient. The latter may result in a meteor radar temperature bias such that only the root mean square errors will be increased compared to the satellite. The root mean square for a seasonal comparison following this procedure was 13.7 K. In general, the error is larger during summer (14 K) than during winter (11–12 K). From this we conclude that the radar maybe is capable to detect shorter period variations during high meteor flux times (night). The weaker errors during the winter months may be explained by a better statistics, due to the fact that the GRIPS dataset has only a few measurements longer than 4 h during the summer, but almost all measurements in winter extend this time span.
average of every day. The planetary wave activity is smoothed, but still included for periods longer than 7 days. Using a Fourier transform and a filter for all waves longer than 3 days gives a first indication about the tidal temperature amplitudes (Fig. 5) and phases (Fig. 6). These figures show amplitudes of 2–7 K for the diurnal tide. We observe the minimum in autumn and a slightly increased activity during winter months. This is similar to the planetary wave activity, which slowly increases from the autumnal equinox to the winter. The detailed investigation of possible interaction of planetary waves and the diurnal tide is an interesting topic for future research. In spring, a small diurnal tidal minimum appears again before the highest amplitudes of about 6–7 K are reached in summer. The semidiurnal tide dominates during the year and exhibits a different seasonal behavior. Minimum amplitudes of 3 K are reached during equinoxes in March and
16 15
diurnal tide semidiurnal tide terdiurnal tide
14 13 12
3.4. Temperature tide estimate with the Collm meteor radar Measuring hourly mean temperatures to detect tides is not possible, because of the too small overall number of measurements during 1 h. In addition, the diurnal oscillation of the flux leads to very low meteor count rates in the late afternoon. Assuming a constant tidal amplitude and phase during a period of at least 7 days, however, allows to reconstruct a time series of hourly means for each month. Therefore, all meteors of the same hour of each day within this 7 day period are used to calculate an hourly mean temperature. This procedure results in a running
amplitude / K
The COMMA-LIM model is a 3-dimensional global circulation model, solving the atmospheric equations in flux form on a sphere. The model has a horizontal resolution of 5 · 5.625. The vertical structure includes 48 layers extending from the ground to 150 km altitude. The model parameterizations include gravity wave parameterization, a detailed radiation scheme, diffusion processes and basic ionospheric activities, like ion drag and Lorentz deflection. The model is structured by 36 · 64 grid points in latitude and longitude extending in the vertical from ground up to 150 km divided in 48 layers. A more detailed description of the model is given by Fro¨hlich et al. (2003, 2005) and Jacobi et al. (2006). A qualitatively good coincidence of the model data to the measurements can be noted. However, during the transition time in spring and autumn COMMA-LIM shows 5 K higher values than the radar measures which still needs investigation, especially under the aspect of the temperature gradient model.
11 10 9 8 7 6 5 4 3 2 1 0 1
2
3
4
5
6
7
8
9
10
11
12
time 2005 / month Fig. 5. Monthly mean temperature amplitudes taken from a Fourier transform from a hourly resolved time series with a 7 day composite.
16
diurnal tide semidiurnal tide terdiurnal tide
14 12 10
phase / h (LT)
3.3. Collm meteor radar and COMMA-LIM model
8 6 4 2 0 -2 -4 -6 1
2
3
4
5
6
7
8
9
10
time 2005 / month Fig. 6. Temperature tidal phases, given in LT.
11
12
G. Stober et al. / Advances in Space Research 42 (2008) 1253–1258
September. During the rest of the year the tidal amplitude is close to 7–10 K. The phase of the tide stays at around 7:00 LT during the whole year and shows a tendency towards slightly earlier times in June. It was also investigated how the semidiurnal temperature tide is related to the meridional wind tide. This was done to get an estimate of how the temperature gradient is affected by a non-dissipating tide, following Hocking and Hocking (2002). They show that a vertical wavelength of 25 km will result in a possible phase difference of 0.75 h. Even for vertical wavelengths of about 50 km the phase difference is less than 1.5 h. The meridional wind phase, which was also derived from the radar data, differs by 0–1.5 h. Considering Hocking and Hocking (2002) and following the arguments leads to an estimate of variability for the temperature gradient of 0.1–0.2 K/km. Figs. 7 and 8 illustrate the seasonal oscillation of the semidiurnal wind amplitudes. In summer, the wind phase shifts; from 6:30 LT to approximately 5:00 LT.
40
diurnal tide semidiurnal tide terdiurnal tide
35
amplitude / m/s
30 25 20 15 10 5 0 1
2
3
4
5
6
7
8
9
10
11
12
time 2005 / month Fig. 7. Meridional wind monthly mean amplitudes calculated with a Fourier transform from a hourly wind measurement of the meteor radar in 91 km altitude using a 3 km height gate.
18
diurnal tide semidiurnal tide terdiurnal tide
16 14
1257
The terdiurnal tide is detectable and stays above the 2r noise level in the temperature and the wind field, but a more detailed analysis (not shown) indicates that its amplitude may be overestimated. 4. Summary and conclusions The Collm meteor radar is able to measure day-to-day temperature variation as well as tidal activity. The empirical temperature gradient model used leads to daily mean mesopause temperatures that are consistent with the temperatures from SABER and from a ground-based OH instrument (GRIPS). A comparison of shorter variations like tides is much more difficult. The OH instrument can detect short fluctuations in the temperature field, but it can not resolve the tides. The spatial sampling of the satellite instrument makes it difficult to compare the wave activity at a certain geographic position. The COMMA-model agrees very well with the measurements for longer time periods. Shorter period variations are only of qualitative agreement with the radar. We investigated the capability of the meteor radar technique to detect the temperature tidal activity in the mesopause region. Our results let us conclude that for the semidiurnal tide temperatures are strongly connected to meridional winds in the northern hemisphere. The measured phase of the meridional wind and the temperature tide shows a delay of approximately 1–2 h, which is in the same range as reported by Hocking and Hocking (2002). The diurnal tide is visible in our analysis, but the measurements are less accurate due to the shorter vertical wavelength and the strong day-to-day variability. However, the derived monthly mean temperature amplitudes show a similar behavior like those of the meridional wind. The phase, however, does not reflect this connection. The terdiurnal tide is permanently visible in the spectra and stays above the noise level of the 2 sigma interval. The amplitudes remain close to 1-3.5 K. A detailed comparison to other measurements is, at present, not possible. The data base available for the temperature terdiurnal tide is very limited and a further analysis is not yet meaningful. However, the continuous measurement of monthly mean temperature tides will provide a better climatology in the future.
phase / h (LT)
12 10
Acknowledgements
8
This study was supported by DFG under Grant JA 830/ 21-1 (GW-Code) within the DFG SPP 1176 (CAWSES). We are grateful for the very helpful discussions with Dr. Singer from IAP Ku¨hlungsborn.
6 4 2 0 -2
References 1
2
3
4
5
6
7
8
9
10
time 2005 / month Fig. 8. Meridional wind phase, given in LT.
11
12
Fro¨hlich, K., Pogoreltsev, A., Jacobi, C.h. Numerical simulation of tides, Rossby and Kelvin waves with the COMMA-LIM model. Adv. Space Res. 32, 863–868, 2003.
1258
G. Stober et al. / Advances in Space Research 42 (2008) 1253–1258
Fro¨hlich, K., Jacobi, C.h., Pogoreltsev, A.I. Planetary wave transience effects on the zonal mean flow. Adv. Space Res. 35, 1900–1904, 2005. Jacobi, C.h., Fro¨hlich, K., Pogoreltsev, A. Quasi two-day-wave modulation of gravity wave flux and consequences for the planetary wave propagation in a simple circulation model. J. Atmos. Sol. Terr. Phys. 68, 283–292, 2006. Hocking, W.K., Thayaparan, T., Jones, J. Meteor decay times and their use in determining a diagnostic mesopheric temperature-pressure parameter: methodology and one year of data. Geophys. Res. Lett. 23, 2977–2980, 1997. Hocking, W.K. Temperatures using radar-meteor decay times. Geophys. Res. Lett. 26, 3297–3300, 1999. Hocking, W.K., Fuller, B., Vandepeer, B. Real-time determination of meteor-related parameters utilizing modern digital technology. J. Atmos. Sol. Terr. Phys. 63, 155–169, 2001.
Hocking, W.K., Hocking, A. Temperature tides determined with meteor radar. Annales Geophysicae 20, 1447–1467, 2002. Oberheide, J., Offermann, D., Russell III, J.M., Mlynczak, M.G. Intercomparison of kinetic temperature from 15 m CO2 limb emissions and OH*(3,1) rotational temperature in nearly coincident air masses: SABER, GRIPS. Geophys. Res. Lett. 33, L14311, 2006. Offermann, D., Jarisch, M., Donner, M., Steinbrecht, W., Semenov, A.I. OH temperature re-analysis forced by recent variance increases. J. Atmos. Sol. Terr. Phys. 68, 1924–1933, 2006. Singer, W., Bremer, J., Hocking, W.K., Weiss, J., Latteck, R., Zecha, M. Temperature and wind tides around the summer mesopause at middle and arctic latitudes. Adv. Space Res. 31, 2055–2060, 2003.
