We report on observations of noctilucent clouds (NLCs) by a ground-based lidar located in northern Norway at 69°N, 16°E. The ALOMAR Rayleigh/Mie/Raman (RMR) lidar conducted measurements of the Arctic middle atmosphere from 1 June to 15 August during each year from 1997 to 2001. This data set contains 1122 hours of lidar observations whereof 408 hours include NLC signatures. The interannual variation of the NLC occurrence frequency shows a decrease of strong NLCs, while weak NLCs occur more frequent. The seasonal variation of the NLC occurrence shows a well pronounced core period where NLCs appeared during 43% of the time. The basic properties of NLCs are characterized by three parameters: maximum value of the volume backscatter coefficient βmax (≡brightness), centroid altitude zc, and half width δz (≡thickness). A typical NLC above ALOMAR during the 5-year period reported here owns a brightness of βmax = 9.6 × 10−10 m−1 sr−1, an altitude of zc = 83.3 km, and a thickness of δz = 1.2 km. The interannual variation of the parameters shows a decrease of the brightness, an increase of the altitude, and a nearly constant thickness, while seasonal variability is higher than these interannual changes. During the core period, the NLCs are noticeably brighter than at the beginning as well as the end of the season. Altitude and thickness of NLCs decrease during the season.
 Noctilucent clouds (NLCs) have recently attracted the interest of a widespread research community as the question has been raised whether they might be a blazing signal for changes in the middle atmosphere caused by prospering industry and growing mankind. This idea was released by their sudden and persistent appearance since 1885 [Schroeder, 1999] as observed by numberless professional and amateur observers with the naked eye since then. These ground-based observations of NLCs also point out that the occurrence frequency of NLCs is modulated by a 9–12 year oscillation [Fogle and Haurwitz, 1966; Gadsden, 1998; Romejko et al., 2003]. Since the early days in NLC research by still irreplaceable ground-based observers many questions about their nature have been answered by detailed case studies using ground-based [Jesse, 1896], rocket-borne [Hemenway et al., 1964], and satellite-borne [Donahue et al., 1972] experiments. Already in the early days of NLC research the potential of active sounding with light was seen [Jesse, 1887] and this challenging problem was addressed in the sixties [Fiocco and Grams, 1969]. For more than 10 years, it is now possible to investigate NLCs from the ground by active lidar sounding [Hansen et al., 1989] and gain more detailed information about occurrence, altitude and brightness of NLCs. From the lidar observations it is possible to monitor the temporal variation of NLCs as well as estimate the size and shape [von Cossart et al., 1999; Baumgarten et al., 2002a] of the particles forming the NLCs. In addition to detailed case studies, lidars are used to investigate the basic properties of NLCs like brightness, altitude and vertical extent on a climatological basis. This database of NLC observations covering day and night revealed for example a diurnal modulation of the NLCs in altitude and brightness [von Zahn et al., 1998; Chu et al., 2001b]. The ability to detect NLCs during the whole day is one of the key advantages for research on the occurrence of NLCs with lidars, since passive observations are limited to twilight conditions and observations by satellites are either limited by their Sun-synchronous orbits [Thomas and Olivero, 1989] or the need for illumination of the NLC region by the Sun [Gadsden, 2000]. However, the high resolved observation of NLCs with lidars is paid by the need of a considerable experimental setup. In recent years, a few of these experiments have observed NLCs in the Northern Hemisphere [Thomas et al., 1994; Langer et al., 1995; Thayer et al., 1995; Stebel et al., 2000; Alpers et al., 2001; Collins et al., 2003] and Southern Hemisphere [Gardner et al., 2001], whereof even less are capable of performing NLC observations during the whole day. Typically these experiments observe NLCs above the lidar stations and are limited to good weather conditions which modifies the statistics of the occurrence frequency of NLCs in an unknown way. If the weather conditions restrict the observations of NLCs in a random way it is however possible to estimate the true occurrence frequency of NLCs. Since it is known that the occurrence frequency of NLCs shows a seasonal variation [Fogle and Haurwitz, 1966; Gadsden, 1998] most likely because of changes in the thermal structure of the atmosphere and the water vapour concentration [Lübken, 1999; Seele and Hartogh, 1999] it is clear that a seasonal variation of the observation conditions for the lidar will falsify the estimation of the NLC occurrence frequency. Hence the deduction of the occurrence frequency using lidar measurements requires careful analysis of potential variations in the lidar performance as we will show in the following sections. The following investigations will also address the question of the difference of passive ground-based and lidar observations of NLCs by grouping the NLCs according to their brightness into different classes as well as the key question whether a significant change in the occurrence frequency is observed.
2. Instrumentation and Analysis Method
 The ALOMAR Rayleigh/Mie/Raman (RMR) lidar is a twin lidar system, using two independent power lasers and two tiltable receiving telescopes for detailed case studies of the Arctic middle atmosphere at altitudes between 15 and 105 km as well as exhaustive climatological investigations. The lidar is a highly automated system [Fiedler and von Cossart, 1999; von Zahn et al., 2000] and part of the international ALOMAR facility (69°N, 16°E) near Andenes, Norway. For the NLC observations reported here, we used the frequency-doubled emission of a Nd:YAG laser at 532 nm wavelength. The backscattered light from the atmosphere was collected by a zenith-pointing telescope of 1.8 m diameter using a field-of-view of 180 μrad (equal to 18 m/100 km) which is just wide enough to receive the backscattered light of the laser beam having a full divergence of less than 100 μrad. For further suppression of sunlight scattered in the atmosphere a double Fabry-Perot interferometer with an effective spectral bandwidth of about 4 pm (FWHM) was installed in front of the photon-counting detector. Usually the lidar is set to operate with an altitude resolution of 150 m.
 The lidar records single photons backscattered from air molecules and NLC particles. The raw data of photon counts versus altitude z is converted into an altitude profile of the total backscatter signal S(z) by subtracting the sum of solar background and thermionic emission of the detector which are continuously measured during operation. As a simple measure for the presence of aerosol particles we use the backscatter ratio R(z), which is defined as the ratio of the measured total signal to the molecular signal SM(z):
where βM and βNLC are the volume backscatter coefficients for air molecules and NLC particles, respectively. The volume backscatter coefficient is defined as:
with n as number density and dσ(180°)/dΩ as effective backscatter cross section. The equation (2) applies for scattering on molecules and NLC particles in a similar way.
 To calculate the volume backscatter coefficient of the NLC particles we use the backscatter ratio:
The molecular backscatter coefficient is derived from air densities representing mean values over 10 years at the lidar location [Lübken, 1999]. The total backscatter signal is normalized to the molecular backscattering at 55 km altitude. As a measure of the NLC brightness we use the volume backscatter coefficient (βNLC) rather than the backscatter ratio (R) since the latter one depends on the molecular number density (R(z) ∝1/nM(z) if βNLC ≫ βM) and thus on the altitude of the NLC even if the total amount and size of the particles are identical at different altitudes.
 The lidar data were recorded with a time resolution of 167 s, corresponding to 5000 transmitted laser pulses. While this time resolution gives a sufficient data quality to study bright NLCs we use a ∼14 min mean (5 × 167 s) as a compromise of improved data quality and blurring of NLC properties by variable NLC structures [Baumgarten et al., 2002b]. The data quality of each record was judged by the strength of the molecular scattering signal at an altitude of 55 km. If the quality was found to be insufficient the record was not included in the further analysis. In the following sections, we call this data set “Total Measurements.” The set of “Total Measurements” was searched for NLCs by looking for significant backscatter coefficients defined by βNLC(z) > ΔβNLC(z), where ΔβNLC(z) is the corresponding measurement error. The resulting subset is called “NLC Measurements.” For each single record included in this dataset we have extracted the maximum value βmax of βNLC(z) in the altitude range from 78 to 90 km, the centroid altitude, and the thickness (full width at half maximum). The centroid altitude is defined as:
and the half maximum limits where found by comparing βNLC(z) with βmax while moving from the lower and upper altitude limits (78 and 90 km) towards the altitude of maximum backscattering.
3. Time Distribution of the Measurements
 The NLC season at 69°N lasts roughly from 1 June to 15 August, corresponding to day of year 152–227 (to be increased by 1 in leap years). During the NLC seasons from 1997 to 2001 a total of 1122 hours of observations were performed with the ALOMAR RMR lidar (“Total Measurements”). As outlined in the introduction it is important to cover the NLC season with a sufficient amount of measurements to ensure that a representative value of the desired physical parameter is observed. To obtain as many measurements as possible the lidar was operational 24 hours a day during the whole NLC season year by year. Nevertheless lidar observations are limited to approximately 30 out of 76 days per season by the weather conditions. Therefore we have decided to combine the data sets of the last 5 seasons resulting in a coverage of 73 out of 76 days throughout the season. Figure 1 points out that the sampling shows no pronounced seasonal variation and furthermore that the coverage of observations per day shows no significant breakdown. The diurnal variation of the “Total Measurements” is an expression of the sensitivity of the lidar, which will be addressed in the next section.
 To study the seasonal variation of NLC measurements the season was subdivided into five periods (≡PD) of 15 days each (the last period contains 16 days). Thus each period covers approximately 2 weeks, giving the following intervals: where the day number is given relative to summer solstice. The temporal coverage with measurements for each year and period is given in Table 1, which lists:
measurement time [hours]
diurnal coverage with measurements [%]
number of days with measurements
By turning the attention to the details in the data set it becomes obvious that the measurements are well balanced both among the years and throughout the season. This guarantees that the average seasonal behavior is dominated neither by a 1-year runaway nor by insufficient seasonal or local time coverage.
1– 15 June
−20 to −6
−5 to 9
31 July to 15 August
Table 1. Yearly and Seasonal Coverage With Lidar Dataa
Each cell shows from top to bottom: measurement time [hours], diurnal coverage with measurements [%], and numbers of days with measurements.
4. NLC Detection Efficiency
 The question “How often have we been able to detect a NLC of a given brightness?” leads to an investigation of the lidar efficiency when searching for NLCs. This topic is the most sensitive part of the task to derive the NLC occurrence frequency. A similar question is often raised in discussions about an increase in NLC occurrence frequency as observed by passive ground-based observers (it is often speculated that even if a constant number of observers is hunting for NLCs their growing feasibility in detecting NLCs by self training might explain an increase in NLC observations). There are mainly three processes that determine the sensitivity of a lidar:
instrumental effects like transmitted laser energy, efficiency of the detection system, efficiency in suppressing the solar background
solar background light, scattered in the atmosphere which the backscattered laser light has to compete with. The solar background is primarily depending on the solar elevation, which (obviously) shows a diurnal and seasonal variation where the daily values (during the NLC season) vary from −6.7° to +44.0°. A minor influence on the solar background is given by tropospheric clouds, which increase the scattered light if sunlit.
tropospheric transmission depending on the weather conditions, for example clouds and haze.
 These effects lead to a variable efficiency for the detection of weak NLCs.
 From the lidar data recorded it is not possible to treat these three effects separately and therefore we have decided to analyze each record of the data set “Total Measurements” regarding the capability to detect a NLC of a certain minimum brightness (βthresh) during the lidar measurements. This is the case if the 1 − σ error (ΔβNLC(z)) in the altitude range of the NLC is smaller than the brightness threshold chosen. Tests have shown that per record the variation of the ΔβNLC in the altitude range from 80 to 90 km is less than 10% which allowed us to simplify the analysis and use one averaged error per record instead. We define the number of all cases where ΔβNLC < βthresh divided by the number of records of the data set “Total Measurements” as NLC detection efficiency of the lidar measurements.
Figure 2 shows the average NLC detection efficiency per season for different brightness thresholds as a function of years. The encouraging result is that averaged over day and nighttime, good and bad weather conditions, etc. the lidar was able to detect NLCs with β > 4 × 10−10 m−1 sr−1 during more than 90% of the measurement time. This value of the NLC detection efficiency also applies to most of the periods PD1–PD5 for the different years. Only in four out of the 25 periods the efficiency was slightly lower but still higher than 80%.
5. Occurrence Frequency of NLCs
 To derive the occurrence frequency of NLCs above ALOMAR we calculated the ratio of accumulated time of NLC observations (“NLC Measurements”) to total time (“Total Measurements”) for each year in the report period. The results are shown in the upper graph of Figure 3. To study the dependency of the occurrence frequency on the NLC detection efficiency we have plotted the values including all NLCs seen by the lidar (βthresh = 0) and additionally the occurrence frequency of NLCs brighter than βthresh = 4 × 10−10 m−1 sr−1. The difference of these two curves represents the occurrence frequency of NLCs less bright than βthresh = 4 × 10−10 m−1 sr−1 which ranges between 6% and 22%. The variation of this subset of NLCs is driven by changes of the true NLC occurrence frequency and variations of the NLC detection efficiency. Since we cannot separate both effects we concentrate in the following discussion on NLCs brighter than βthresh = 4 × 10−10 m−1 sr−1, unless otherwise stated.
 The interannual variation of the NLC occurrence frequency is dominated by a significant decrease from 1997 to 1999, while in 2000 the value rises up to 32.8%, which is nearly twice as much as in the preceding year. In 2001 the NLC occurrence frequency is lower again but remains slightly higher than in 1999.
 For a more detailed investigation the NLC measurements are divided into 4 different brightness classes βclass [10−10 m−1 sr−1]. For the further discussion we call the brightest clouds with βclass > 13 strong and the least bright clouds with βclass = 4–7 weak, while clouds with βclass = 7–13 are called medium NLCs. The corresponding curves are shown in the lower panel of Figure 3. They show that averaging all brightness classes hides some underlying features since the occurrence frequency of strong NLCs is decreasing while that of weak NLCs is increasing.
 To study the seasonal variation of the NLC occurrence frequency, Figure 4 shows the 5-year average occurrence frequency as function of the defined periods. Clearly visible are the beginning and the end of the NLC season at PD1 and PD5, respectively, with distinctly lower values for the occurrence frequency of NLCs compared to the core period (PD2–PD4). During the 5 years reported here, we observed the earliest NLC on 29 May 2000, which was a single and very weak cloud while the latest NLC was observed on 16 August 2000. Taking into account all NLCs seen by the lidar, i.e., βthresh = 0, NLCs appeared during 42.8% of the core period and during 36.3% of the whole season. When limiting the data set to NLCs brighter than βthresh = 4 × 10−10 m−1 sr−1 the corresponding numbers are 31.1% and 25.4%.
 To allow a more detailed look at the data addressing the seasonal and interannual variation, Figure 5 shows the analysis separately for each period of the seasons investigated. During the early season (PD1) there are no strong NLCs. For weak and medium NLCs the occurrence frequency looks bathtub-like.
 The beginning of the core period (PD2) reflects a similar picture as seen during the whole season. The occurrence frequency of all NLCs observed drops from around 70% down to 13% with a pronounced peak during 2000 which is caused by the large number of weak clouds. During PD2 strong clouds suffer a fast interannual decrease and disappeared in 2001.
 A less clear behavior can be seen during the early July period (PD3). However, at least an anticorrelation of the occurrence frequency between weak and strong clouds is obvious from 1999 to 2001.
 The late July period (PD4) is dominated by a large number of strong NLCs in 1999, reaching an occurrence frequency of 38%, contrary to values of only 7–12% during the other years. The weak clouds increased in the course of the years until 2000 but became rare in 2001.
 During the late season (PD5) the NLC occurrence frequency rose from zero in 1997 to a maximum in 2000, which is also the case for medium NLCs, while the occurrence frequency of strong clouds was low until 1999 but increased since then.
 Generally speaking, the variability of the occurrence frequency of weak as well as strong NLCs mostly contributes to the total seasonal and interannual variations while medium NLCs have little impact on the overall variations.
 During the individual periods the development of the interannual occurrence frequency of NLCs is nonuniform, partly contrary to each other. Looking at the mean seasonal behavior, the decline of strong NLCs is dominated by the NLC observations in late June (PD2) while in contrast to this the increasing number of weak NLCs is caused by at least three periods (PD2, PD4, and PD5).
6. Properties of NLC Layers
 To describe the basic properties of NLCs as observed by the lidar during the last 5 years we have chosen three parameters: (1) maximum value of the volume backscatter coefficient βmax, (2) centroid altitude zc, and (3) half width δz. For this analysis we have taken into account all observations of NLCs regardless of the brightness threshold βthresh. The interannual variation of the mean and median value of these parameters for the whole NLC season is shown in Figure 6. Additionally the seasonal variability of these parameters described by the standard deviation of the data set is shown. Especially the brightness of the clouds shows a large seasonal variability but also the other parameters show a seasonal variability significantly larger than the interannual variation. The previous studies of the seasonal coverage however allow us to conclude that the remaining year-to-year changes in the properties can be interpreted as the average seasonal behavior and are worth a closer look.
 The mean brightness of the clouds shows a significant decrease by a factor of more than 2 from βmax[10−10 m−1 sr−1] = 14 in 1997 to 6.7 in 2000 and increased again in 2001. The difference of the mean and the median brightness of the clouds is caused by the fact that the distribution is asymmetric and shifted to less bright NLCs.
 The mean cloud altitude rises between 1997 and 2001 continuously from 82.8 to 83.7 km except for the NLCs in 2000. The latter year is the only one for which a decrease in brightness is not connected with an increase in altitude. The congruence of mean and median value is given by the symmetry of the altitude distribution.
 Despite of the remarkable changes of NLC brightness and altitude the mean thickness of the clouds remains amazingly constant at 1.2 km. The difference of mean and median half width shows that it is more likely that NLCs are thinner than 1.2 km.
 To asses the seasonal variation of the NLCs the mean and median NLC properties are plotted in Figure 7, derived from the accumulated 5-year data set.
 The brightness of the NLCs shows a considerable seasonal variation similar to the occurrence frequency of NLC. The NLC brightness increases in the early season while it is unchanged during the core period and decays in the late season. This behavior is less prominent for the median values.
 During the NLC season, the mean altitude decreases constantly with an average slope of 16 m d−1 starting at approximately 84 km in the early season. A notable drop of the altitude is observed in the early season from PD1 to PD2. The agreement of mean and median cloud altitude reveals the symmetry of the altitude distribution. The mean cloud thickness shows an obvious seasonal decrease from about 1.3 km in early June to 1.0 km at the end of the season.
 In Table 2, we give a summary of the NLC properties extracted from our data set. Due to asymmetric distributions four values are used, namely (1a) mean and (1b) standard deviation, (2) median, and (3) value of maximum occurrence (mode). We emphasize the large differences between mean and mode values.
Table 2. Basic NLC Properties Above ALOMAR Derived From Observations Between 1997 and 2001
βmax[10−10 m−1 sr−1]
9.6 ± 9.1
83.3 ± 1.2
1.2 ± 0.6
7. Discussion and Conclusions
 With the data reduction method outlined above we have extracted data not influenced by instrumental effects and hence the remaining variations are an intrinsic feature of the NLCs above ALOMAR. Despite the fact that our laser beam illuminates an area of only ∼10 m diameter at NLC altitude we claim that most of the time we observe non local features of NLCs because the NLC formation time of several hours in conjunction with the horizontal motion of the atmosphere of roughly 200 km h−1 at NLC altitude will result in an extensive NLC area observed. Furthermore the lidar is located well north of the southernmost edge of NLCs around 60°N [Gadsden, 1998] and hence well inside the polar NLC area.
7.1. Basic NLC Properties
 As described above, the NLC parameters brightness and thickness have to be compared carefully because of their asymmetric distributions. Therefore we use for the comparison with other lidar observations of NLCs, performed at the South Pole [Chu et al., 2001a, 2001b, 2003] and Greenland (ARCLITE) [Thayer et al., 2003], both mean and mode value of the parameters. Some minor adjustments have been made to standardize the NLC parameters as far as possible: The South Pole results reported for the austral summers 1999/2000 and 2000/2001 Chu et al.  where combined by taking the hours on NLC observation into account. The brightness seen at a wavelength of 374 nm was rescaled to 532 nm via
where the conversion factor was calculated from scattering of light by a monomodal lognormal NLC particle size distribution (rmed = 50 nm, σ = 1.4) as observed at ALOMAR [von Cossart et al., 1999]. Further on the thickness values were converted from the original RMS width to δz (full width at half maximum) via
assuming a Gaussian profile of the NLC [Thayer et al., 2003]. For the ARCLITE results we used the given histograms to roughly estimate the mean values.
 The comparison of basic NLC parameters determined at the different lidar stations is shown in Table 3. While the NLCs at ALOMAR and ARCLITE show about the same brightness the lidar at the South Pole observes brighter NLCs. Additionally the distribution of the brightness is more symmetric at the South Pole, identifiable by the smaller difference between mean and mode values.
Mean (a) and mode (b) values of the parameters are displayed. The data sets do not cover the same years. For standardized conditions, some parameters have been recalculated (for details, see text).
βmax [10−10 m−1 sr−1]
(a) 9.6 ± 9.1
(a) 20.8 ± 10.4
(a) 83.3 ± 1.2
(a) 85.0 ± 1.0
(a) 1.2 ± 0.6
(a) 1.8 ± 0.7
no seasonal trend
−46 m d−1 (PD1 and PD2)
40 m d−1 (PD1 and PD2)
−16 m d−1 (average)
−62 m d−1 (PD3–PD5)
negative and positive
 The altitudes of NLCs at the Arctic sites are nearly identical and the slightly lower NLCs at ARCLITE could be caused by the different seasons included in both data sets. Namely the years 1994–1996 are not included in the ALOMAR data set and the 2001 season is not included in the ARCLITE report. On the other hand NLCs above the South Pole observed in the austral summers 1999/2000 and 2000/2001 are significantly higher, about 1.5 km compared to the average NLC altitude at ALOMAR from 1999 to 2001. Such a difference is not seen on a hemispheric scale by satellite observations of NLCs during austral summer 1997/1998 and Arctic summer 1999 resulting into nearly equal values for Northern Hemisphere as well as Southern Hemisphere between 82.6 and 83.2 km [Carbary et al., 2001]. Furthermore from measurements of the thermal structure of the atmosphere performed at ALOMAR and at its nearly colatitudinal counterpart (Rothera, 67°S) [Lübken et al., 1999], which show an unexpected similarity in the temperature at 82 km altitude, it seems more likely that the difference in NLC altitudes at ALOMAR and the South Pole is caused by the difference in latitude. Sorting the NLC altitude by the latitude of the stations reveals a correlation of the NLC altitude and the latitude. Drawing the attention to the NLC thickness one realizes that the distribution at ALOMAR is more extended to large values compared to ARCLITE whereas the mode values of both stations match perfectly. In comparison to the thickness at the South Pole the lower latitudinal NLCs are thinner.
 The relationship between brightness and altitude of NLCs appears to be quite different between the different latitudes (see Table 3). While at the lower latitudinal Arctic sites the brighter NLCs tend to appear at lower altitudes, the South Pole station shows no pronounced correlation between altitude and brightness even if the seasonal variation of the NLC altitude dz/dday is taken into account [Chu et al., 2003].
 To allow further comparison and discussion about this topic the latter behavior has been treated in more detail for the ALOMAR observations. Figure 8 shows the correlation of NLC brightness and altitude for each year separately. With respect to the nonlinear connection of these two parameters the data set was subdivided into two brightness classes, where the separation was chosen to be the lower limit of our so-called strong NLCs. Obviously dβ/dz < 0 holds for each year and brightness class and supports the growth-sedimentation scenario where the bigger particles progress downward while they grow. Average values dβ/dz in units of 10−10 m−1 sr−1 km−1 over all years are −32.9 for the brighter and −7.4 for the less bright clouds. The higher value of dβ/dz for bright clouds agrees to the fact that (1) the sedimentation speed increases just proportional to the particle size while the backscatter coefficient increases by the fifth to the sixth power of the size and (2) the sedimentation velocity decreases at lower altitudes because of the increasing air density.
7.2. NLC Occurrence Frequency
 The seasonal variation of NLC occurrence frequency as shown in Figure 4 is no doubt affected by the anticorrelated variation in upper mesosphere temperatures [Lübken, 1999] and water vapor mixing ratio [Seele and Hartogh, 1999]. The slight seasonal asymmetry in NLC occurrence frequency, which is also seen in the occurrence frequency of PMSE [Bremer et al., 2003], could reflect such asymmetry of the water vapor abundance. Further on the seasonal variation of the NLC occurrence frequency above ALOMAR agrees with the data set collected by passive ground-based observers in northwest Europe [Gadsden, 1998] especially regarding the location of the core period.
 Concerning absolute numbers a comparison of our NLC occurrence frequencies with other observations is difficult because of e. g. interannual variations of NLC parameters and different NLC detection thresholds of lidars, satellite instruments, and ground-based observers. Our observations show in average over the 5 years a NLC occurrence frequency of 36% when taking into account all NLCs seen by the lidar and 25% when limiting to NLCs brighter than βthresh = 4 × 10−10 m−1 sr−1. The lidar observations at the South Pole show the NLCs to be present with 67% nearly twice that often [Chu et al., 2003]. This difference is most likely caused by the different latitude of both stations since satellite-borne observations of NLC show an occurrence frequency of 23–25% at 70°N, but around 60% at the poles for a data set from 1981 to 1985 [Thomas and Olivero, 1989].
 Ground-based observations during the recent 4 decades have revealed a 9–12 year oscillation of the NLC occurrence frequency [Gadsden, 1998; Romejko et al., 2003]. While it is clear that a 5-year data set is insufficient to study the secular change of NLC occurrence the well defined and constant instrumental setup allows at least a comparison to the predictions given by [Gadsden, 1998]. Following that, the NLC occurrence frequency should decrease until 2002. Our observations however show decreasing NLC occurrence from 1997 to 1999, a sudden increase in 2000 followed by again lower occurrence in 2001. If we limit our data set to strong NLCs (solid line in Figure 3b) the occurrence frequency agrees better to the prediction. This feature points to the possible problem of the difficult to judge NLC detection efficiency of passive ground-based observers which is still in discussion but believed to be significantly lower than those of the lidar.
 The more detailed look on the year-to-year changes of the NLC occurrence frequency during the season suggests a shift of the NLC occurrence to the late season. These seasonal variations taken little care of in the analysis of NLC observations so far, might also be the explanation for unexpected differences when comparing different data sets.
 From our point of view, this kind of analysis needs further extension and will result in a more detailed understanding of NLCs and the background atmosphere. Especially the improved experimental feasibilities for studies of NLCs and the dynamics of the background atmosphere, like a new lidar experiment at 78°N (J. Höffner, private communication) and the new ALOMAR MF radar (W. Singer, private communication), will allow a more detailed study and may show the underlying differences in the atmospheric dynamics.
 We are indebted to K. Bekkelund, A. Jensen, R. Lyngra, R. Eixmann, C. Fricke-Begemann, T. Köpnick, A. Schöch, U. von Zahn, and U. Blum for operating the lidar. Special thanks go to K. H. Fricke and Ulf von Zahn for sharing their wisdom not only about lidars and again to Ulf von Zahn for guiding the experiment from the very first ideas to its actual capabilities. The ALOMAR RMR lidar is a team effort of the Leibniz-Institut für Atmosphärenphysik, Germany, the Physikalisches Institut der Universität Bonn, Germany, the Service d'Aéronomie du CNRS, France, and the Hovemere Ltd., England.