SEARCH

SEARCH BY CITATION

Keywords:

  • noctilucent clouds;
  • polar mesospheric clouds;
  • lidar;
  • mesosphere

Abstract

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[1] Noctilucent clouds (NLC) are an important tracer of temperature and dynamics of the summer mesopause region. Our site at Kühlungsborn (Germany, 54°N) is at the equatorward edge of the NLC region and therefore of special interest for the understanding of these clouds. 41 nights (63 h) of NLC are observed since 1997. They form the largest lidar data set from mid-latitudes. NLC are typically weak, with nearly 70% having a backscatter coefficient βmax,532nm<2 ⋅ 10−10 m−1 sr−1. The seasonal variation of NLC shows maximum occurrence around the temperature minimum (saturation maximum) but lower temperatures (higher saturation) at the beginning compared to the end of the season. Mean centroid altitude is 82.7 ± 0.03 km, with strong NLC being typically lower and vertically thinner compared to weak clouds. NLC occurrence was lowest in the years 2000–2002 and reached a maximum in 2009 with a rate of 19%. Overall, NLC are less frequent and dimmer compared to higher latitudes. The occurrence is highly anti-correlated with solar activity. Beside NLC, we are measuring mesospheric temperatures since 2002 by lidars, complemented by microwave observations of water vapor (since October 2009) and radar observations of mesospheric winds. NLC occurrence is found anti-correlated with ambient temperatures (r = −0.85 at 84 km), while low temperatures are necessary but not sufficient for individual events. Meridional winds at 84 km are weakly anti-correlated with NLC occurrence (r = −0.58 at 84 km). Furthermore, we find some biennial variation of NLC occurrence in part of the time series. Any additional trend has not yet been detected.

1 Introduction

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[2] Noctilucent clouds (NLC), or polar mesospheric clouds (PMC), are known since ~125 years as phenomena, where processes in the summer mesopause region become directly visible [Leslie, 1885; Jesse, 1885]. NLC mainly exist in the polar regions, while they get rare around 55° latitude. Extended data sets of NLC occurrence exist by visual observations from the ground [e.g., Fogle and Haurwitz, 1966; Gadsden, 1998; Romejko et al., 2003; Kirkwood et al., 2008] and are complemented in recent years by instrumented observations from the ground as well as from space [e.g., DeLand et al., 2007; Fiedler et al., 2011; Gumbel and Karlsson, 2011]. The Aeronomy of Ice in the Mesosphere (AIM) satellite has been launched with several instruments as a dedicated mission to examine NLC/PMC in both hemispheres [e.g., Russell et al., 2009; Chandran et al., 2010]. Because of their relation to ambient conditions of temperature and humidity, NLC are an important tracer for processes in the mesopause regions on scales of minutes to hours [Baumgarten et al., 2009].

[3] The suitability of NLC as tracer for long-term trends has been discussed controversially [Thomas, 2003; von Zahn, 2003]. Possible effects of climate change should be most apparent at the equatorward edge of the NLC region, i.e., between 60° and 50° latitude [e.g., Thomas, 1996]. In this region, mean temperatures typically do not allow ice existence, so NLC occurrence is expected to be related to atmospheric waves of all scales and to long-term variations of mean temperatures, water vapor concentrations, and winds. Model calculations by Lübken and Berger [2011] revealed an increase of NLC occurrence rates at high latitudes but no trend south of 60°N and no southward extension of NLC. Long-term observations of NLC from mid-latitudes are sparse and do not provide a conclusive picture. Visual data provide occurrence rates for regions ~3–6° poleward of the observer and show essentially no trend [e.g., Fogle and Haurwitz, 1966; Gadsden, 1998]. Satellite observations with SBUV instruments are available since summer 1979 [e.g., DeLand et al., 2007]. Their lowest latitude bin (54°N–64°N) shows a distinct trend in PMC occurrence but extends ~10° farther north and is affected by higher uncertainty than the other latitude bins [Shettle et al., 2009]. Apart from that, only few studies from lidar [e.g., Thomas et al., 1994; von Cossart et al., 1996; Alpers et al., 2000; Taylor et al., 2002; Gerding et al., 2007] or satellite data [e.g., Stevens et al., 2009] are available, covering either single NLC events or a single season. Therefore, additional observations and multi-year studies are required for a better understanding of mid-latitude NLC. In general, ground-based lidar provide the potential for the quantitative description of mean NLC properties and their long-term variation.

[4] We operate lidars at Kühlungsborn/Germany (54°N, 12°E) with regular soundings of the mesopause region by different (Doppler) resonance and Rayleigh-Mie-Raman lidars since ~1997 [von Zahn and Höffner, 1996; Gerding et al., 2000; Alpers et al., 2004]. In 2002, we started regular temperature soundings covering the whole atmosphere between the troposphere and the lower thermosphere [Alpers et al., 2004; Gerding et al., 2008]. In total, we observed NLC in 15 seasons (1997–2011). To the best of our knowledge, this is the largest data set of NLC parameters actively observed at mid-latitudes, and it provides important insights into mean NLC parameters and their seasonal and interannual variation. For this study, we limit the data set to nighttime operation (see below), to avoid issues of tidal variation of NLC parameters and of different signal-noise ratios during day and night. Overall, ~63 h of NLC data have been obtained within ~900 h of lidar operation in the summer season (1 June–4 August). This data set covers 41 nights with NLC, out of 268 nights observed in total within the 15 summers.

[5] In the next section, we describe our NLC data set and how we extract NLC parameters. The seasonal variation of NLC occurrence is described in section 3, followed by the altitude and brightness distribution of NLC examined in section 4. These parameters are averaged across the whole data set. The interannual variation of NLC occurrence is characterized in section 5 together with the variation in solar radiation, ambient temperatures, and winds.

2 Data Base and Retrieval Procedure

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[6] The data presented here are on the basis of observations with the Rayleigh-Mie-Raman lidar (RMR lidar) and the potassium resonance lidar (K lidar), co-located at the Leibniz-Institute of Atmospheric Physics (IAP) in Kühlungsborn (Germany, 54°N, 12°E). The RMR lidar is described by Alpers et al. [1999] and has been operated between 1997 and 1999 alternatively in tropospheric or mesospheric (NLC) mode. NLC observations during that period have been triggered by the simultaneously operating K lidar, i.e., if an NLC started appearing in the K lidar profiles, the RMR lidar was switched to mesospheric mode. By this, the NLC have also been sampled by the RMR lidar, and NLC parameters have been derived at 532 nm wavelength (section 4). In 2002, additional detection channels were introduced, allowing temperature soundings, e.g., in the upper mesosphere up to ~90 km [Alpers et al., 2004; Gerding et al., 2008] and slightly improving the sensitivity for NLC. Since 2002, the instrument's setup is unchanged; i.e., the whole decreasing branch of solar cycle 23 is covered without instrumental changes.

[7] The K lidar at Kühlungsborn is mainly designed for Doppler temperature soundings [von Zahn and Höffner, 1996]. Until 1999, a mobile K lidar was used, located right beside the institute's building. For examination of seasonal and interannual changes of NLC occurrence between 1997 and 1999, we are using the data of the mobile K lidar in combination with the RMR data (sections 3 and 5). In 2000, a stationary K lidar was built up and operated since then without significant changes. The temperature data at NLC altitudes are taken from stationary K lidar and RMR lidar for the years 2003 to 2011 [cf. Gerding et al., 2008]. This RMR lidar at Kühlungsborn is in operation on a nighttime basis only, i.e., ~3.5–5 h per 24 h in summer. Recently, a second, daytime capable RMR lidar was set up at our site. This lidar performed first tests in 2009 and went into routine operation in 2010. The data are not used in this paper to keep the data set homogenous. The power of the Kühlungsborn RMR lidar compares roughly with the ALOMAR RMR lidar, being known for a long-term NLC data set from polar latitudes [Fiedler et al., 2011]. Owing to the permanent daylight during polar summer, the sensitivity of the ALOMAR RMR lidar for very weak NLC is reduced, compared to the nighttime lidar soundings described here.

[8] All dates in this publication describe the end of the night, i.e., observations during the night 1/2 July are labeled as “2 July”. RMR backscatter profiles are integrated for 4000 laser pulses (~2 min) before further processing. All NLC parameters are calculated on running averages of five profiles. NLC are identified by eyes using the 532 nm wavelength, as it has the best signal-noise ratio when compared to the simultaneously observed 355 nm and 1064 nm backscatter. To separate molecular (Rayleigh) backscatter and NLC particle backscatter, the signal from above and below the NLC is interpolated by an exponential fit. The remaining aerosol backscatter profile is used to calculate aerosol (NLC) backscatter coefficients β. Owing to the low noise level of our nighttime soundings, the detection limit is as low as β532 = 0.1 ⋅ 10−10 m−1 sr−1. From each of these brightness profiles, the maximum value is taken to represent the NLC strength (βmax). In the following, we provide values of β in units of 10−10 m−1 sr−1. We use the term “brightness” for the aerosol backscatter coefficient because β is closely related to NLC brightness as observed by, e.g., the human eye. NLC altitudes are derived as the center of mass of the β-profile, based on 200 m altitude channels. The NLC thickness is calculated as Full-Width-at-Half-Maximum (FWHM) of the β profile. The brightness is interpolated between the bins because of the low thickness of NLC (~6 altitude bins).

[9] Overall, 63.2 h (1707 profiles) of NLC have been observed within 905 h total observation time (7% average occurrence rate). An example of an NLC event during the night 1/2 July 2009 is given in Figure 1. The lidar was in operation between 21:30 and 01:33 UT. The NLC appeared at 22:39 UT and was observed until this sounding stopped due to the rising sun. The NLC maximum was observed at 01:20 UT with β ≈ 34, i.e., the highest backscatter coefficient ever observed at our site.

image

Figure 1. Temporal evolution of the backscatter coefficient at 532 nm of an NLC observed during the night 1/2 July 2009. The gray bar in the bottom denotes the lidar operation time (21:30–01:33 UT). The NLC appeared at 22:39 UT and was observed until this sounding stops because of the rising sun. The temporal resolution is 2:13 min with a running average of five profiles (~10 min).

Download figure to PowerPoint

[10] NLC occurrence rates will also be compared with ambient wind conditions. Winds are measured continuously during day and night by the IAP MF radar at Juliusruh (55°N, 13°E), i.e., ~120 km east-north-east of Kühlungsborn. The MF radar covers an altitude range of ~68 to 93 km. The system and a long-term wind data set are described in Keuer et al. [2007].

[11] Since October 2009, the microwave radiometer MISI (MIcrowave Spectrometer at IAP) measures water vapor concentrations continuously during day and night in the altitude range of ~40–85 km above Kühlungsborn from H2O emissions at the microwave frequency 22.235 GHz [cf. Seele and Hartogh, 1999]. The center of the field of view is in northwest direction at 30° elevation; i.e., the center of the observed area (at ~82 km altitude) is ~140 km northwest of Kühlungsborn. MISI observations allow calculating the water vapor saturation ratio S = pwater/pSat. The partial water vapor pressure pwater and the saturation pressure pSat are calculated following the equations from Marti and Mauersberger [1993] using temperature data from the co-located lidars. Air pressure pair is taken from NRLMSISE-00 data [Picone et al., 2002].

[12] The seasonal variation of water vapor concentration as observed by MISI is shown in Figure 2. The data are smoothed by a 7 day running mean. The water vapor concentration is generally decreasing above the stratopause. At NLC heights, concentrations of about 2–3 ppmv are observed throughout the season. Highest water vapor concentrations in the whole mesosphere are reached at the beginning of August, i.e., about one month after summer solstice. This proves that apart from photochemistry, also the vertical transport is important for water vapor abundance in the upper mesosphere. In summer, the water vapor is uplifted from the stratosphere, while the downwelling of mesospheric air in winter reduces the water vapor concentration in a given altitude.

image

Figure 2. Seasonal variation of water vapor concentration as observed by the MISI water vapor radiometer in 2010/2011. Data are averaged for 7 days.

Download figure to PowerPoint

3 Seasonal Variation of NLC Occurrence

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[13] Figure 3 shows the occurrence rate of NLC above Kühlungsborn as a function of season. The data have been binned for 10 day. For each 10 day period (e.g., days 175–184), the total NLC occurrence time as a sum of 1997–2011 is divided by the total lidar sounding time within these years. NLC appear mainly in June and July. The distribution is not changing symmetrically, but with a faster increase than decay. Taking all NLC into account, the highest occurrence rate is observed around day 180 (29 June). The first NLC has been observed at day 156 (06 June). After the maximum, the occurrence decreases slowly. The last NLC occurred at day 216 (04 August). In Figure 3, the overall NLC occurrence rate is compared to the rate of strong NLC (βmax>4). Occurrence of strong NLC is highest around day 170 and decays slowly until beginning of August when they vanish completely.

image

Figure 3. Upper panel: Seasonal variation of NLC occurrence for all NLC (gray histogram) and only NLC with βmax>4 ⋅ 10−10 m−1 sr−1 (red histogram). Data are derived from K lidar observation times (1997–1999) and RMR lidar observation times (2000–2011). The black line shows the total sounding time. Middle panel: Mean temperatures for different altitudes calculated by a harmonic fit to the data, updated from Gerding et al. [2008] and covering the period June 2002 until December 2011. The data points show the individual nightly means (83 km) during the NLC season. The dashed line describes the frost point temperature at 83 km using co-located MISI H2O observations. Lower panel: Water vapor saturation ratios (blue/red line) from MISI H2O observations (2010–2011) and the fitted temperature variation as well as mean meridional wind (1997–2011) in 5 day bins from MF radar soundings (light blue/orange squares). Dates given at the upper abscissa axis are in day/month format.

Download figure to PowerPoint

[14] The seasonal variation of NLC occurrence shows a slight shift compared to the temperature variation in this altitude range. We include average temperatures for our location at 83–87 km in Figure 3, updated from Gerding et al. [2008] for the period June 2002–December 2011. Lowest temperatures at NLC heights are typically observed around day 175, i.e., near the NLC maximum, depending on altitude. Comparing start and end of the season (days 160 and 220, respectively), temperatures at the beginning are ~4 K (~13 K) lower at 83 km (87 km). Even though there is a lot of natural temperature variability in the individual nightly mean temperatures (points in Fig 3), the error of the mean is ~1 K. Therefore, the temperature difference between beginning and end of the NLC season is significant.

[15] NLC existence depends not only on temperatures but also on water vapor concentration. The frost point temperature is calculated from the observed water vapor concentration at 83 km and plotted in the middle panel of Figure 3. Average temperatures are always a few Kelvin above the frost point, and only few individual nightly mean temperatures are below the frost point. Obviously, NLC existence above our site requires additional gravity waves and tides, pushing the temperature below the frost point (see below). Nevertheless, the seasonal change of water vapor saturation ratio S calculated from typical temperatures and water vapor concentrations gives some indication for the changing probability of NLC observation. The saturation ratios S for 83 and 85 km altitude are plotted in the lower panel of Figure 3. The highest saturations are reached a few days after the temperature minimum, because water vapor concentration is increasing slightly during the NLC season (see above). The highest NLC occurrence coincides with the period of highest saturation, as expected. But at the end of the season, the saturation is reduced by 90% compared to the beginning (days 220 and 160, respectively). This is because the increasing water vapor concentration cannot compensate for the increasing temperatures. In summary, if we assume a linear dependence of NLC occurrence and water vapor saturation, changing saturation should have reduced NLC occurrence rates to 1/3 or 1/9 (days 210 and 220, respectively) compared to the beginning of the NLC period (day 160). Nevertheless, our soundings show rates of ~3/5 or ~1/3, i.e., rates are higher than expected from saturation.

[16] As mentioned above, NLC above our site are limited to the existence of long-period gravity waves and larger scale waves, pushing temperatures below the frost point temperature for a sufficiently long time. Therefore, an asymmetry in the temperature variability might produce the observed asymmetry in NLC occurrence. At least for short-period gravity waves, being expected to reduce NLC probability [Rapp et al., 2002], Rauthe et al. [2008] found no significant asymmetry in the seasonal variation. We conclude that the local temperature variability due to short-period gravity waves does not directly influence the seasonal variation of NLC occurrence.

[17] Therefore, we suggest that the NLC variation (with βmax>0) does not resemble the temperature and water vapor variation at mid-latitudes but partially mimic polar conditions, where temperatures are ~5 K colder at the beginning of August compared to the beginning of June [see, e.g., Lübken, 1999, Table 7]. The dependence on polar conditions agrees with earlier findings that NLC above Kühlungsborn are essentially limited to conditions of southward directed meridional winds, advecting NLC from polar latitudes [Gerding et al., 2007]. Further proof is given by the seasonal variation of mean meridional winds as observed by the nearby MF radar at Juliusruh, plotted in the lower panel of Figure 3 in bins of 5 days. Southward wind is prevailing throughout the season, decreasing slightly till the beginning of August, when also the NLC probability gets low. Focussing on strong NLC only (βmax>4), the occurrence rate decreases faster towards the end of the season compared to all NLC. This at least suggests that strong NLC are more related to local temperature conditions than weak NLC, even if the significance of this result is limited owing to the low number of strong events. In summary, southward directed meridional winds advect the NLC to our site, and local temperatures and water vapor concentrations influence the particle growth shortly before observation of the NLC. Model results by LIMA/ICE confirm that the very few hours before reaching maximum brightness are essential for the NLC properties [J. Kiliani et al., JASTP, 2012, submitted].

4 Brightness and Altitude Distribution of NLC

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[18] Figure 4 shows the brightness distribution for all NLC observed between 1997 and 2011 (RMR lidar, 532 nm). The occurrence rate decays roughly exponentially from very dim clouds to bright clouds. Approximately 35% of all NLC are in the two lowest bins (βmax<0.5) and called “very weak”. Approximately 33% of the NLC are “weak” with 0.5<βmax<2, ~14% are of medium strength (2<βmax<4), and ~19% are “strong” with βmax>4. The mean brightness for our mid-latitude NLC is inline image = 2.5 with FWHM of 0.9, but the strong asymmetry of the distribution should be noted. NLC at our site are generally much dimmer than, e.g., at 69°N/ALOMAR [Fiedler et al., 2011]. At polar latitudes, the weakest clouds above the long-term limit have βmax>4 and “strong” NLC need βmax > 13, occurring with a rate of 55% or 16%, respectively. The exponential decay of the brightness distribution is also observed at ALOMAR [Fiedler et al., 2009] and comparable to the g-distribution of satellite-based albedo measurements [Thomas, 1995; DeLand et al., 2003].

image

Figure 4. Histogram of maximum backscatter coefficients at 532 nm from individual NLC profiles (five profiles running mean) of all NLC. The particular fractions are plotted in bins of 0.25 ⋅ 10−10 m−1 sr−1. The fraction is given in percent on a logarithmic scale.

Download figure to PowerPoint

[19] We derive NLC centroid altitudes from individual brightness profiles. All NLC centroid altitudes occur between zmin = 79.3 km and zmax = 86.0 km. The altitude distribution is roughly Gaussian. In Figure 5, the individual NLC heights are sorted in bins of 1 km width (gray histogram). The mean centroid altitude of all NLC is 82.7 ± 0.03 km with a standard deviation of 1.3 km. Brighter NLC are observed lower in the atmosphere. NLC with βmax>4 are more frequent in the lower altitude bins and have never been observed above 84.3 km. Overall, the mean altitude of strong NLC is 82.3±0.06 km (standard deviation 1.1 km), i.e., 0.4 km lower than the total centroid altitude. These altitudes are ~700 m lower than the values published by Fiedler et al. [2009] for the ALOMAR observations at 69°N. They found mean altitudes of 83.2 km for all NLC, 83.0 km for βmax>4, and 82.5 km for βmax>13. NLC above Søndre Strømfjord (67°N) are on average observed at 83.0 km, with observations being focussed on the end of the season and βmax>2 [Thayer et al., 2003]. At even higher latitudes, NLC altitudes are again slightly increasing, reaching 83.7 km at 78°N [Lübken et al., 2008]. It has been shown from observations and LIMA model simulations that at polar latitudes the average NLC altitude compares well with the 145 K isotherm that is changing by only 250 m from the pole to 60°N [Lübken et al., 2008]. This relation fails for the NLC above Kühlungsborn, where such low temperatures are hardly reached, not even at the mesopause. Again, advection from (colder) polar latitudes is important for mid-latitude NLC observations. Concordantly, LIMA/ICE predicts NLC being only ~500 m lower at Kühlungsborn compared to ALOMAR (on the basis of year 2001 data) [Lübken et al., 2008], which is close to the measured multi-year average distance. Recently, Chu et al. [2011] predicted a mean NLC altitude at 54°N at ~82.6 km on the basis of a postulated linear decrease with latitude, which is close to the observed NLC altitudes at our site.

image

Figure 5. Histogram of centroid altitudes calculated from individual NLC (five profiles running mean). The gray histogram describes all NLC, the red histogram only NLC with βmax>4 ⋅ 10−10 m−1 sr−1. The dashed lines denote the centroid altitudes for both data sets.

Download figure to PowerPoint

[20] In Figure 6, the FWHM of the NLC is plotted against NLC altitude. Typically, the FWHM of NLC above our site is between 0.4 and 0.8 km. The NLC width generally decreases with decreasing altitude. While the width of the NLC decreases its brightness increases (Figure 7). NLC thicker than 2 km FWHM are only observed for weak and very weak NLC, while the strong NLC are generally thinner than 1.3 km. From NLC statistics, an unambiguous explanation for the relation of NLC width, brightness, and altitude cannot be given. Most simply, the relation can be understood as a sedimentation process. Assuming first a homogenous NLC layer, the speed of sedimentation is a function of ambient (i.e., exponentially changing) air density; the higher particles sediment faster than lower particles. This results in a shrinking of the layer width during sedimentation [e.g., Pruppacher and Klett, 1997; Gerding et al., 2003] and an increasing ice mass density, therefore in a higher backscatter coefficient (larger particles or higher particle density). In reality, at least two other mechanisms must be taken into account: First, during sedimentation in supersaturated air, the ice particles grow due to the uptake of water vapor molecules. This growth typically occurs at the bottom of the layer, resulting in layering with respect to particles sizes: the particle radius increases with decreasing altitude. By this, the NLC gets brighter and the FWHM decreases, because the particles mainly grow at the bottom of the layer. A non-homogenous NLC with layered larger (heavier) and smaller (lighter) particles certainly, as second mechanisms, sediments different compared to a homogenous layer. Since all these effects are not significant with a single, local observation and only appear on a statistical average, a more detailed examination is beyond the scope of this paper.

image

Figure 6. NLC thickness (FWHM) depending on centroid altitude from individual NLC (five profiles running mean). The color code shows how many NLC profiles are observed in a given FWHM-altitude bin (width 0.1 km for FWHM, 1 km for altitude). For FWHM smaller than 2 km, the FWHM generally increases with altitude, while thicker NLC do not show a distinct altitude dependance.

Download figure to PowerPoint

image

Figure 7. NLC thickness (FWHM) depending on NLC brightness βmax from individual NLC (five profiles running mean). The color code shows how many NLC profiles are observed in a given FWHM-β bin (width 0.1 km for FWHM, 1 ⋅ 10−10 m−1 sr−1 for backscatter coefficient). Few NLC with βmax>30 ⋅ 10−10 m−1 sr−1 are omitted. The FWHM generally decreases with NLC brightness, and only very dim clouds show FWHM larger than 2 km.

Download figure to PowerPoint

5 Interannual Variation of NLC Parameters

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[21] The actual NLC data set at Kühlungsborn comprises 15 summers since 1997, i.e., about the whole solar cycle 23. It allows studying the interannual variation of NLC occurrence and the dependance on, e.g., solar shortwave radiation and ambient conditions, mainly temperatures and winds. As mentioned above, the RMR lidar at Kühlungsborn was alternatively operated in the troposphere and mesosphere between 1997 and 1999, while the K lidar made continuous soundings of the mesosphere/lower thermosphere region (MLT). Therefore, the occurrence rates of NLC for 1997–1999 are taken from the K lidar data and for 2000–2011 from the RMR lidar data. The K lidar is less sensitive than the RMR lidar for detecting NLC. Even if there were additional searches for weak NLC by the RMR lidar (without additional detections), an underestimation of NLC occurrence in the years 1997–1999 cannot be excluded. Nevertheless, the main results about the interannual variation are still valid. The available data are listed in Table 1.

Table 1. Data Coverage for NLC Soundings at Kühlungsborn
YearNumber of NightsSounding Period [h]a
  • a

    1997–1999 mainly K lidar soundings are used. 2000–2002 additional care is taken on nights with visual NLC (not to miss an NLC even if number of soundings is small).

19972358.2
19981521.4
19993370.1
200069.7
200189.5
2002721.9
20031267.7
200422107.1
20052396.9
200627125.7
20071349.5
200825109.7
20092685.1
201027116.5
20112487.0

[22] During the first part of the period (1997–1999), the NLC occurrence decreased strongly from more than 11% to less than 3% (Figure 8). In the following years, NLC continued being extremely rare, with occurrence rates reaching hardly 3%. It should be noted that the small rates base on one single NLC event per year in the years 1998, 1999, 2001, and 2003, while in 2000 and 2002, no NLC have been observed at all. Starting in 2004, NLC appear more frequently above our site, reaching rates of up to ~20%. Besides this increase of NLC occurrence, a strong interannual variation is obvious, showing a 2 year period between 2004 and 2010. Because of the low number of NLC and the limitation to nighttime soundings, at least part of the variation might result from the incomplete sampling. To overcome this issue of interannual variation, we applied a ±2 year Hanning smooth to the data (red line in Figure 8). In the following, we will first concentrate on this smoothed data set.

image

Figure 8. Interannual variation of NLC occurrence rate above Kühlungsborn. The rates are plotted year-wise (red circles) and smoothed by a ±2 year Hanning filter (red line). The Lyman-α flux averaged for the particular summers (days 152–219) is plotted in orange. The correlation coefficient r and significance s are given in the plot. Lyman-α data are taken from http://lasp.colorado.edu/lisird/lya/.

Download figure to PowerPoint

[23] The yellow line in Figure 8 describes the Lyman-α flux averaged for the individual June/July periods (data provided by http://lasp.colorado.edu/lisird/lya/). Obviously, there is a distinct anti-correlation between smoothed NLC occurrence and Lyman-α flux. The correlation coefficient is r = − 0.95 which agrees with the general knowledge of water vapor photolysis and increasing temperatures (i.e., decreasing saturation) by shortwave solar radiation. We note that the direct relation of Ly-α flux and water vapor mixing ratio gets more complicated in the presence of NLC because of the redistribution of water vapor by freeze drying [e.g. von Zahn et al., 2004]. For the unsmoothed data set, the anti-correlation of Ly-α flux and NLC occurrence is strongly reduced (r = − 0.66), revealing that additional mechanisms dominate the occurrence rate on a year-to-year basis. For example, while Ly-α flux was similar in 2008 and 2009, the first year shows a very low NLC occurrence rate but the next year the highest NLC rate observed ever.

[24] Since summer 2002, lidar temperature soundings are performed at our site by combination of K resonance lidar and RMR lidar [Alpers et al., 2004]. In Figure 3, we presented the temperature variation based on a seasonal fit to all data as published by Gerding et al. [2008]. Here we want to compare the MLT temperatures with the NLC occurrence on a year-to-year basis. Mean summer temperatures (June/July) have been calculated on the basis of individual temperature profiles of the particular years. Figure 9 comprises all temperature data obtained so far. For clarity, we only show data from three relevant altitudes, namely, 83 km, 84 km, and 87 km. NLC occurrence rates are repeated from Figure 8. The data are smoothed by a ±2 year Hanning window. Between 2003 and 2007/2008, temperatures decrease remarkably by ~8 K at 84 and 87 km, and ~5 K at 83 km, with a standard deviation of the nightly means of ~6 K. In the most recent years, temperatures raised again. The temperatures are clearly anti-correlated with NLC occurrence rates with correlation coefficients of r ≈ − 0.8, showing in general the higher probability of NLC observations if temperatures are lower. Nevertheless, Gerding et al. [2007] showed that low temperatures are necessary but not sufficient for NLC observations.

image

Figure 9. Interannual variation of NLC occurrence (red line; cf. Figure 8) and summer mean temperatures at different altitudes. Both data sets are smoothed by a ±2 year Hanning filter. The correlation coefficients r and significance s are given in the plot. The standard deviation of all nightly mean temperatures of summers 2003–2011 is 1.8 K at 83 and 84 km, and 3.8 K at 87 km.

Download figure to PowerPoint

[25] Another factor adding to the low NLC occurrence rates in the years 2000–2002 is the meridional wind. Typically, the mean meridional wind at NLC altitudes in summer is a few meters per second in southward (negative) direction. This allows advection of NLC from polar latitudes to our site. In the summers 2001 and 2002, meridional winds are very weak and do on average not endorse advection of NLC (Figure 10). (For year 2000, no data are available.) In the following years, mean meridional winds increase and reach ~5–6 m/s. Overall, meridional winds are roughly anti-correlated with NLC occurrence (r ≈ − 0.58). This is compatible with earlier findings that southward winds are necessary but not sufficient for NLC above our site [Gerding et al., 2007]. On the other hand, zonal winds are highly correlated with NLC occurrence rates (r ≈ 0.82). Westward winds are strongest around the year 2001 when the NLC occurrence is low. Since 2005, weak mean (westward) winds are observed, while the NLC occurrence is high. Nevertheless, a direct causative relation between zonal winds and NLC occurrence remains open. The high correlation of zonal wind and NLC occurrence might be a secondary effect of the solar cycle. Schmidt et al. [2006] presented a study on solar cycle variations of the mesopause region. They found an increase of westward wind during solar maximum in the lower thermosphere due to increasing ion drag. The authors claim that this effect is potentially even underestimated, i.e., might propagate farther down. The interannual zonal wind variation differs slightly from the data presented by Hoffmann et al. [2011] for July observations. The reason is the strong wind gradient below the mesopause in summer, with the altitude of wind reversal decreasing throughout the summer. Nevertheless, the data shown here are most representative for the NLC season.

image

Figure 10. Interannual variation of NLC occurrence (red line; cf. Figure 8) and summer mean winds at 84 km (blue dots: meridional wind (v); blue line: zonal wind (u)). All data sets are smoothed by a ±2 year Hanning filter. The correlation coefficients r and significance s are given in the plot. For year 2000, no wind data are available.

Download figure to PowerPoint

[26] Overall, the general increase of NLC occurrence rates since 2002 is related with the decrease of Lyman-α-radiation received from the sun and with decreasing temperatures. The large interannual variability observed between 2004 and 2010 (cf. unsmoothed time series in Figure 8) is examined in the following section.

[27] Furthermore, the interannual variation of NLC altitudes and NLC brightness has been examined (not shown). NLC altitudes do not show a clear interannual variation. Annual mean NLC altitudes vary typically between ~82 and ~83 km, which is within the standard deviation of NLC altitudes of individual years (typically 1.0–1.5 km). In fact, the standard deviation is typically increasing with the number of NLC, because the night-to-night variation is often larger than the altitude variation of a single event. The mean backscatter coefficient does also show only slight interannual variation. Typically, the mean backscatter coefficient is near β ≈ 2, with lower values in 2001/2003 (β ≈ 0.5) and higher values in 2009/2011 (β ≈ 3.5). Again, the night-to-night variation is typically much larger.

6 Discussion

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[28] The described lidar soundings are obtained during nighttime only. Therefore, they are only representative for the period of ~21:00–01:00 UT. Our soundings cover roughly 20 nights per summer with some irregular distribution across the whole season, depending on weather conditions. NLC occurrence rates above Kühlungsborn are only on the order of 10% and vary strongly depending on the individual year. The incomplete sampling provides some additional uncertainty for the derived quantities. In the Appendix, we show that for an unsmoothed data set, an occurrence rate variation by ~4% can be expected owing to limited sampling. If smoothing by a ±2 year Hanning filter is applied, the variation due to limited sampling is reduced to ~2%; i.e., the interannual variation of the smoothed data set as shown, e.g., in Figure 8 is not an artifact.

[29] In Figure 9, we presented a general anti-correlation between NLC occurrence and summer mean temperatures, after smoothing by ±2 years. In Figure 11, we show the unsmoothed time series of NLC occurrence rates and temperatures. NLC show a distinct interannual variation with 2 years period between 2004 and 2010 (high rates 2005, 2007, and 2009). In a statistical study, we have demonstrated that this periodicity is most probable a real feature of the NLC distribution above Kühlungsborn and not observed by chance, even if the particular occurrence rates show an uncertainty of ~4% (see Appendix). A similar period is found in the summer mean temperatures, showing an anti-correlation to the NLC occurrence rates for the period 2006–2010. We have shown in an earlier publication based on individual nights that low temperatures are not sufficient for NLC observations [Gerding et al., 2007]. Nevertheless, lower average temperatures certainly favor the existence of ice in the mesopause region. Therefore, it is suggested that the biennially varying summer mean temperatures at least contribute to the biennial variation of NLC occurrence rates, while additional mechanisms are needed to explain the NLC variation in the 2004/2005. Gravity wave activity is also derived from our lidar observations. Nightly mean profiles of wave activity are calculated as described by Rauthe et al. [2008]. In summer, the nightly temperature variation mainly represents the short-period (~2 h) gravity wave activity, because observations are limited to ~4 h. Figure 11b shows the nightly mean wave activity in the upper mesosphere averaged for the particular June/July season. There is a general anti-correlation of wave activity at NLC altitudes (83/84 km) and NLC occurrence rates except for the year 2003, where the NLC occurrence was low owing to high solar activity. The anti-correlation can at least quantitatively be understood from studies by Rapp et al. [2002]. On the basis of CARMA simulations, they showed that waves with periods below ~6 h, like in our observations, more likely destroy NLC, while longer period waves may strengthen NLC. Later, Thayer et al. [2003] published observations of short-period gravity waves (2–3 h) reducing NLC brightness. This is explained by the faster sublimation of ice particles compared to their growth. The fact that the relation of NLC occurrence, mean temperatures, and wave amplitudes becomes obvious in our data since 2004 only might be related to the higher mean temperatures and higher solar activity before this year: Before 2004, average conditions did not favor NLC existence at all and the modulation of NLC occurrence rates by temperatures and wave activity did not come into play. We need to point out that the summer-averaged wave activity should be taken as an indicator for the general wave activity in the mid-latitude region but not as a direct driver of individual NLC. The year-to-year differences are well within the intra-seasonal variation. Additionally, the wave activity in NLC nights does not vary significantly from non-NLC nights [Gerding et al., 2007].

image

Figure 11. Interannual variation of mean summer temperatures (a) and mean summer gravity wave activity (b). For the absolute temperatures, all June/July temperature profiles are averaged. For the wave activity, the particular (absolute) deviations from the nightly mean profile are averaged, and these mean temperature deviations are averaged for all June/July observations. Note that the data cover only waves with periods of up to ~2 h, due to the typical sounding time in summer. In the lower panel, we repeat the unsmoothed NLC occurrence rates shown in Figure 8.

Download figure to PowerPoint

[30] Figure 11 reveals a biennial variation between 2006 (partly 2004) and 2010 in temperature, wave activity, and subsequently in NLC occurrence rates. The reason for this biennial variation cannot be given from our lidar observations alone. Espy et al. [2011] found a biennial variation in July temperature data of the mesopause region derived from OH airglow measurements at 60°N. They claimed the stratospheric Quasi-Biennial-Oscillation (QBO) responsible, affecting the winter stratosphere in the southern hemisphere via the Holton-Tan mechanism [Holton and Tan, 1980] and coupling into the northern summer mesopause region [e.g., Becker et al., 2004; Karlsson et al., 2007]. While a detailed analysis of this topic is outside the scope of this paper, we note that the mentioned publications describe more polar latitudes. A direct phase relation of our NLC and temperature data with the Singapore wind data above ~50 hPa (acting as QBO proxy) cannot be given from our limited data set. Additionally, the model study of Mayr et al. [2009] predicts strong phase changes in the mesopause region between 50°N and 60°N. Future observations will show whether or not the biennial variation is a true feature in our data set.

[31] Only few publications are available which deal with time-resolved, multi-year observations of NLC at mid-latitudes. The most recent publication by Siskind et al. [2011] describes SHIMMER observation between 2007 and 2009. They found a remarkable low occurrence rate in 2007. This is in contrast to the high NLC occurrence observed above our site but may be explained by the different local times of the soundings. In June 2007 and 2008, SHIMMER sampling was in the evening and early morning, i.e., best comparable to our lidar sounding times. In June 2007, SHIMMER observed higher rates compared to the seasonal average of this year, while they were much reduced in June 2008. These June averages from SHIMMER compare well with the lidar-observed seasonal mean rates. The higher NLC sensitivity of our lidar cannot explain the differences, since the interannual variation is a prominent feature also in the bright NLC occurrence rates.

[32] Shettle et al. [2009] presented NLC observations from SBUV satellite between 54° and 64°N starting already in 1979. The interannual variation since 1997 is in good qualitative agreement with our observations. Visual observations from Great Britain and Denmark published by Kirkwood et al. [2008] show also a minimum of NLC occurrence starting around year 2000. But the increase was delayed until ~2005. Both data sets cover the latitude of our site, but extend much farther north into regions with more NLC, possibly affecting NLC occurrence rates. The high sensitivity of NLC existence at mid-latitudes on ambient conditions of temperature and water vapor raises the question whether the injection of water vapor by space shuttle launches contributes to the NLC variation at our site. Some case studies referred NLC (or PMC) observations at high latitudes to shuttle launches a few days before [e.g., Stevens et al., 2003; Kelley et al., 2010]. We have detected six NLCs within 16 days of a space shuttle launch. We have examined the characteristics of these NLCs and find that these NLCs vary from weak to strong and their characteristics appear typical of the NLCs observed at our site. From this analysis, we conclude that there is no significant contribution of shuttle launches to the observed NLC variability.

[33] We note that the limitation to nighttime data may introduce some uncertainty for the interpretation of the data. Some NLC were observed already at the beginning of the sounding, or observations are disrupted because of sunrise. Therefore, we cannot estimate the NLC duration since it may be biased by the limited sounding period. Occurrence rates are possibly affected by tidal variations if only nighttime NLC are examined. Fiedler et al. [2011] found that the NLC occurrence rates at 69°N vary with local time due to tides. Gerding et al. [2007] mentioned southward directed winds as a necessary condition for mid-latitude NLC. Wind directions change regularly during the day due to tidal oscillations. From radar soundings, we found a phase shift of the most important semidiurnal tide by less than 1 h since 2004 (not shown here), i.e., much shorter than the typical sounding time. Additionally, Fiedler et al. [2011] found only a slight phase shift in the tidal components of NLC occurrence rates at 69°N. Therefore, tidal phase variations cannot be responsible for the average variation of occurrence rates at our site or for the biennial variation since 2004. To examine possible diurnal variations of NLC parameters at mid-latitudes, we built up a daytime capable lidar and started regular NLC soundings during day and night in 2010.

7 Conclusions and Summary

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[34] In this paper, we report average parameters and variation of noctilucent clouds (NLC) or polar mesospheric clouds (PMC), i.e., phenomena that are quite rare at our mid-latitude site (54°N, 12°E). In a data set covering the years 1997–2011, we observe NLC with a typical annual occurrence rate of ~0–12%, up to 19% in extreme years. We limit our data set to nighttime conditions and yield a very high NLC detection sensitivity with a limit of β ≈ 0.1. 33% of the NLC are weak (0.5<βmax<2), 14% are medium (2<βmax<4), and only 19% are called strong clouds (βmax>4). Compared to higher latitudes [e.g., Fiedler et al., 2009], NLC at our site are more infrequent and on average much weaker. NLC appear most frequently in the period between 15 June and 15 July (days 166 and 195, respectively), similar to the lowest temperatures and highest water vapor saturations. After that period, the NLC season continues until beginning of August (~ day 220), resembling the mean temperature variation at polar latitudes, where NLC are typically seeded. This underlines the importance of the polar middle atmosphere also for mid-latitude NLC, supporting recent model simulations [J. Kiliani et al., JASTP, 2012, submitted).

[35] The altitudes of NLC nicely follow a Gaussian distribution with the mean altitude at 82.7 ± 0.03 km and a standard deviation of 1.3 km. Strong NLC are on average observed at lower altitudes, e.g., NLC with β>4 have a mean altitude of 82.3 ± 0.06 km. Compared to polar latitudes [Fiedler et al., 2009], NLC above our site are typically a few hundreds of meters lower but with similar distribution width. The descent of NLC with decreasing latitude is 45–50 m/deg and comparable to the results given by Lübken et al. [2008] and Chu et al. [2011], being related to the mean temperature profiles and horizontal advection of NLC from polar to mid-latitudes.

[36] The presented NLC data set covers 15 years of regular observations with the RMR lidar and the K resonance lidar at Kühlungsborn. The interannual variations of NLC occurrence rates are anti-correlated with the solar Lyman-α flux, consistent with the increasing photolysis of water vapor and increasing temperatures (i.e., decreasing saturation) by increasing shortwave radiation. Minimum rates are observed between 2000 and 2002, with the Ly-α maximum occurring around year 2001. The anti-correlation of NLC occurrence and solar radiation is much higher than, e.g., in the long-term data set of the ALOMAR RMR lidar [Fiedler et al., 2009]. At mid-latitudes, NLC are more sensitive to changing ambient conditions. A decrease in the water vapor concentration due to changing solar flux results in a lower frost point temperature. Ice conditions are more seldom fulfilled, because average temperatures are above the frost point, and (stronger) wave perturbations are needed for saturation. Subsequently, NLC occurrence rates decrease. If average temperatures are already below the frost point like at polar latitudes, a (small) change in frost point temperature does not change the NLC occurrence rate. A similar argument has been published with respect to the influence of gravity wave activity on southern hemisphere NLC, with the larger influence being observed at lower latitudes [Chu et al., 2009].

[37] Temperature profiles of the mesosphere and lower thermosphere are regularly observed since 2002/2003. NLC occurrence rates are anti-correlated with mean summer temperatures at NLC height. Again, this high sensitivity on ambient conditions can be explained by the fact that our observations take place at the edge of the NLC existence region. Even if some of the variability in NLC occurrence rates must be attributed to the limited data coverage, the (unsmoothed) interannual variation partly shows a 2 year periodicity. This period is also observed in the average temperatures and wave activity. It underlines the general dependance of NLC occurrence on low temperatures and low activity of short-period gravity waves. The 2 year period is only observed after year 2004. Before, the NLC occurrence rates are too low, and the changing solar activity is the most prominent feature in the interannual variation. If there is any additional trend, it is much smaller than the variation due to solar cycle.

[38] In summary, we have shown that our observations of NLC are in good agreement with more extended observations at polar latitudes but show a stronger dependance on ambient conditions of water vapor concentration, temperatures, wave activity, and winds. Therefore, observations from mid-latitudes are a sensitive tool to observe long-term changes in the mesopause region. We will continue our soundings for the examination of changes on decadal time scales. Observations during night and day will allow us to build up a more sophisticated data base and to examine tidal variations.

Appendix A: Statistical Significance of Inter-annual Variation

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[39] We have checked how much variability in NLC occurrence is expected because of statistical limitations of our data set. In other words, we have calculated which occurrence rates would be observed with a true NLC occurrence rate of 10% and an arbitrary distribution of NLC and lidar observations. For the test, we computed a set of 12 NLC randomly distributed in a 60 day period. This simulated NLC have between 1 and 3.5 h duration, “observed” in nights of 4 h duration. Longer NLC observations are extremely rare due to the limitation to nighttime soundings. The NLC constructed this way are typical for our location and result in NLC occurrence rates of ~10% averaged over the period of 60 * 4 h. In a simulation run, we have produced an arbitrary distribution of these 12 NLC across the 60 day period. Then we have calculated an arbitrary distribution of 22 lidar observations within the 60 day, which is again typical for the real observations. From the distribution of NLC and “measurements”, we have calculated the NLC occurrence rate for this simulation. Figure 12 shows the histogram of NLC occurrence rates after 10,000 different simulations. The red histogram shows the occurrence rates after smoothing by a ±2 simulations Hanning filter, i.e., comparable to the lidar observations shown above. The mean value is 9.8% as of course expected from the data fed into the simulation. The standard deviation is 2% (in terms of occurrence rate). This simulation clearly shows that the (smoothed) lidar-observed interannual variation of NLC occurrence with rates between ~1 and 12% is not an observational artifact.

image

Figure 12. Histogram of simulated NLC observation rates using a 60 nights summer period (4 h each) with an arbitrary distribution of 12 NLC (23 h), sampled by 22 lidar measurements. Data are averaged for 10,000 simulation runs. Black: unsmoothed, red: smoothed by ±2 simulations Hanning filter.

Download figure to PowerPoint

[40] The gray histogram in Figure 12 is calculated from the unsmoothed set of simulations. Here the standard deviation is <4%. Therefore, at least part of the year-to-year variability of the unsmoothed, observed data must be attributed to the incomplete sampling of the true NLC distribution. Nevertheless, looking at the apparent 2 year period observed between 2004 and 2010, it is most probable not produced by chance. We have checked for the occurrence of a similar sequence in several runs of 10,000 simulations and found between 400 and 500 series, i.e., a probability of <5% to create such a series by chance. If we demand on at least 5% change of NLC occurrence rates in subsequent simulations, we find typically even less than 20 sequences, i.e., 0.2%. We conclude that there is a substantial experimental evidence for a true biennial variation not caused by chance.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References

[41] We acknowledge the support in lidar operation and maintenance of Torsten Köpnick and Michael Priester. Matthias Alpers was responsible for the first years of RMR lidar operation. Sebastian Mitreiter, Roman Rachholz, and Karl-Georg Eller are representative for numerous students helping with nighttime lidar soundings. We thank Paul Hartogh and Kristofer Hallgren for their support in MISI installation. QBO data taken from University of Berlin, http://www.geo.fu-berlin.de/met/ag/strat/produkte/qbo/index.html. Parts of these works were supported by the Deutsche Forschungsgemeinschaft (DFG) under grant GE 1625/1-1. We thank three anonymous reviewers for their help improving the manuscript.

References

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data Base and Retrieval Procedure
  5. 3 Seasonal Variation of NLC Occurrence
  6. 4 Brightness and Altitude Distribution of NLC
  7. 5 Interannual Variation of NLC Parameters
  8. 6 Discussion
  9. 7 Conclusions and Summary
  10. Appendix A: Statistical Significance of Inter-annual Variation
  11. Acknowledgments
  12. References