Noctilucent Clouds (NLCs) are an important phenomenon of the summer mesopause region. While relatively common in high latitudes, NLCs are sparse (≤10% occurrence rate) below 60°latitude. We present the first study of diurnal variations of midlatitude NLCs based on lidar observations with full diurnal coverage at Kühlungsborn since 2010 independent from solar elevation. Overall, ∼100h of NLCs with a backscatter coefficient of βmax,532nm>0.5·10−10m−1sr−1are observed within ∼1800h. Occurrence rates decrease regularly from 12% at 5local solar time (LST) to ∼2% at 19LST. The mean NLC brightness varies between ∼1 and ∼3·10−10m−1sr−1with maxima at 4 and 18LST. The simultaneously observed temperatures show a systematic (tidal) variation, but we do not find a direct relation to NLC rates. Comparing NLCs and ambient winds, we find strong indications for the meridional wind (advection) being the main driver for NLC occurrence above our site.
 Twilight observations of Noctilucent Clouds (NLCs) have been carried out since ∼125years to learn about the properties of the mesopause region [Leslie, 1885; Jesse, 1885]. Only nowadays, a few lidars at polar latitudes are capable of observing NLC during daylight, e.g., for studies of tidal variations [e.g., Chu et al., 2003; Fiedler et al., 2011]. For midlatitudes, NLC diurnal variations have so far only been examined from the Spatial Heterodyne Imager for Mesospheric Radicals (SHIMMER) satellite data [Stevens et al., 2009, 2010] but not by ground-based observations. This was because suitable, daylight-capable lidars were virtually nonexistent at midlatitudes until this study. At our site in Kühlungsborn we perform nighttime NLC soundings since 1997 [e.g., Alpers et al., 2000; Gerding et al., 2013]. In 2010 we built a new Rayleigh-Mie-Raman (RMR) lidar with full diurnal coverage for NLC observations independent of solar elevation. For the first time, these data allow examination of the diurnal variations of NLC parameters from a ground-based system at midlatitudes. Kiliani et al.  show that NLC parameters are mainly determined by local background conditions. For our site, simultaneous temperature soundings by the same lidar and by a colocated potassium Doppler lidar as well as wind soundings by a nearby radar allow the comparison with ambient atmospheric parameters. In the following we will provide a short description of the new lidar and introduce the NLC, temperature, and wind data from four summers 2010–2013.
2 Instruments and Data Sets
 In 2010 the Leibniz-Institute of Atmospheric Physics (IAP) at Kühlungsborn (54°N, 12°E) complemented its suite of lidars for temperature and NLC soundings [e.g., von Zahn and Höffner, 1996; Gerding et al., 2008, 2013] with a new, daytime-capable Rayleigh-Mie-Raman lidar. This lidar is designed for temperature soundings up to 75km during day and 85km during night as well as NLC soundings during day and night. The lidar uses an Nd:YAG laser at 532nm wavelength with 30pulses per second and up to ∼700mJ per pulse. The f/4 telescope has ∼0.8m diameter and a field-of-view (FOV) of only 62μrad. A piezo-mounted beam-guiding mirror stabilizes the laser beam within the FOV using an automated, optical beam-stabilization system similar to the one described by Höffner and Lautenbach . The receiver comprises a narrow-band interference filter of 0.13nm bandwidth (full-width at half maximum, FWHM) and two Fabry-Perot etalons (FPE) of ∼4pm FWHM. An avalanche photo diode is used as a single photon detector. Photon count profiles are saved every 1000pulses with 15m vertical resolution. For NLC detection the profiles are integrated in 195m bins and smoothed vertically by a Hanning window (±2bins). A running average across 15,000 pulses is calculated every 33s (∼8min integration time). Temperature calculations based on hydrostatic integration of the photon count profiles are performed every 15min, integrating over 2h and 1km. Absolute temperatures are influenced by the FPE, because their spectral width is in the order of the Doppler broadening of the backscatter signal. Therefore, the transmission depends on atmospheric temperature, what needs to be acknowledged for absolute temperature calculation. We limit our analysis to temperature deviations from the mean. In these data the FPE-induced systematic error is much less than 1K, i.e., much smaller than the statistical uncertainty. The temperature profiles from the RMR lidar are complemented above NLC heights by temperature data of the daytime-capable potassium lidar [e.g., Höffner and Lautenbach, 2009] for the years 2010–2012. In 2013 the potassium lidar was not in operation. Horizontal winds are measured continuously during day and night by a meteor radar at Juliusruh (55°N, 13°E), i.e., 120km east of the lidar location [e.g., Singer et al., 2003; Hoffmann et al., 2010]. To compare with NLC observations, we limit the data set to the June/July period of the years 2010–2013.
 During the summers of 2010–2013 the lidar was in operation whenever weather allowed. The NLC profiles have been selected after visual inspection of the background-corrected raw data. We limit the data set to all NLCs with βmax>0.5·10−10m−1sr−1, because these NLCs can clearly be identified independent of solar elevation and signal quality. (In the following we give β≡βmax in units of 10−10m−1sr−1). Overall 100h of such NLCs (complemented by 26h of NLCs with β≤0.5) have been observed within ∼1800h lidar operation time in the period between 1 June and 4 August of each year. Table 1 gives an overview of the annual data coverage and NLC rate. Figure 1 shows an example of a photon count profile around 14:20local solar time (LST) at 55° solar elevation (24 June 2010). The NLC is clearly visible even before background subtraction. This NLC was observed from the beginning of operation at 13:13LST until 15:47LST, when the NLC vanishes, while the sounding continued until midnight. The NLC brightness and altitude are modulated by short-period gravity waves, but the average altitude remains roughly constant. The temporal evolution of the NLC is shown in the right part of Figure 1. At our site local solar time is ∼45min ahead of universal time.
Table 1. Data Coverage With Daylight-Capable RMR-Lidar and NLC (β>0.5) at Kühlungsborn
Lidar Observation (h)
NLC rate (%)
3 Diurnal Variations of NLC Parameters and Ambient Atmosphere
 Lidar soundings have been performed during all local times. The diurnal coverage after four seasons is roughly homogeneous with ∼60–90h per particular hour-of-day. The average occurrence rate for β>0.5 is ∼5–6%. The time resolved rates vary between ∼2 and 12% (Figure 2). The highest rates are observed in the early morning hours (∼5LST) and the lowest in the early afternoon (∼19LST). A secondary maximum appears around 14LST. A harmonic fit reveals the amplitude of the semidiurnal variation being ∼2/3 of the diurnal amplitude (Table 2). The semidiurnal component is also prominent in the diurnal variation of bright NLCs (β>4, see Figure 2). The first maximum is observed around 4LST (i.e., slightly earlier than for dimmer NLCs) and the second, broad maximum is visible in the afternoon. Similar to the whole data set, rates of bright NLCs are lowest in the late morning hours and in the late evening.
Table 2. Fit Amplitudes (A) and Phases (φ/hLST) for NLC Occurrence Rates (O/%), Brightness (β/10−10m−1sr−1), and Altitudes (z/km)a
Correlations are given with respect to the raw data set.
 The “large” rate of bright clouds around 4LST and in the afternoon is also shown up in the diurnal variation of the maximum backscatter coefficient (brightness). The mean brightness varies in the course of the day between β≈1 and β≈3 (Figure 3). The brightest clouds are typically observed around 4LST and 18LST, while the clouds are fainter on average in the late morning (∼10LST) and evening (∼23LST). Figure 3 also shows the diurnal variation of the NLC centroid altitude. The mean centroid altitude decreases during the morning and afternoon hours by ∼500m. Later it increases by >1km within 3h after 17LST, followed by some oscillations. The standard error of the mean is in the order of only ±300m, but the variability of the data is large (standard deviation ∼1.5km), because NLCs often ascend or descend by a few kilometers due to gravity wave perturbations. Therefore, we assume that the observed altitude variation is partly caused by the variability of NLCs and still by incomplete sampling. A larger data set is required to resolve the typical diurnal altitude variation of NLCs.
 To understand the diurnal variations of NLC parameters we need to examine the ambient atmospheric parameters like temperatures and winds. Due to the limitation of NLCs to supersaturated regions of the atmosphere, temperature is obviously an important parameter for NLCs. At midlatitudes also horizontal winds become important, potentially advecting clouds from higher latitudes [e.g., Gerding et al., 2007, 2013]. As described above, temperatures are simultaneously observed by the same RMR lidar and complemented by a potassium resonance lidar. For temperature retrieval, only soundings of at least 6h duration are used. In total we got 94 days (∼1400h) of temperature data within the June/July period 2010–2013. Here we concentrate on temporal temperature variations only. We calculate the mean temperature variation by averaging the temperature deviations from the daily mean profiles for each particular local time. Figure 4 shows the temperature variation with altitude and local time. Overall, a combination of diurnal and semidiurnal (tidal) variation dominates. In the mesopause region, temperatures are lowest around noon. The cold phase progresses downward with time and is observed around midnight in the mid-mesosphere and at ∼6LST in 55 km altitude. The amplitude of temperature variation is about ±5K at 86km. The standard error of the hourly binned data is typically around 1K at these altitudes. Unfortunately, for most of the time there are no temperature data exactly at NLC altitudes, mainly due to low signal-noise-ratio of both lidars in this altitude region.
 Obviously, there is no clear relation between NLC occurrence rates and temperature variations. For example, the highest temperatures are observed in the early morning, when also NLC rates are high. This is also true if only days with NLCs are averaged (not shown). Hence, tidal temperature variation is not the main factor for the appearance of NLC. But mean temperatures are different during days with and without NLCs. While Gerding et al.  found mean temperatures at 80–85km ∼5K lower during NLC nights compared to non-NLC nights, our new data set reveals an average temperature difference by up to ∼10K (not shown). In the daily mean temperature profiles gravity and tidal waves are suppressed and larger-scale waves (planetary waves) dominate the day-to-day variation. Therefore, NLC occurrence above Kühlungsborn is most likely related to temperature variation by planetary waves but is less influenced by tidal temperature variation.
 The mean diurnal wind variation is calculated like the temperature variation. In Figure 5 the zonal and meridional winds are displayed. Both wind components show a dominating semidiurnal variation. The zonal wind maximizes at ∼8LST and ∼20LST, with the phase changing by ∼1h between 82 and 91km. The temporal variation is superimposed by an altitudinal variation with a wind reversal around 87km, as expected for summer conditions. The meridional wind has a phase shift compared to the zonal wind by ∼3h, as typical for a circular polarized semidiurnal tide. Maxima are observed near 6 and 17LST and minima near 11 and 23LST. In other words, the wind is strongest southward shortly before noon and midnight.
 There is no direct correlation between NLC occurrence rates (Figure 2) and winds at NLC altitudes (82–85km). The zonal wind is strongest (eastward) shortly after the morning NLC maximum and during the evening NLC minimum. The meridional wind is strongest (northward) near the morning NLC maximum and before the evening minimum. Correlations coefficients are only −0.4 between NLC occurrence and zonal wind, and 0.3 between NLC and meridional wind (85km). On the other hand, the meridional wind is strongest southward when NLC rates are increasing, i.e., in the night and around noon. We will discuss the relevance of advection in the next section.
4 Discussion and Conclusions
 The statistical significance of our results is limited by the low occurrence rate of NLCs at midlatitudes and the incomplete sampling by lidar (during cloud-free conditions only). For hourly averaged NLC brightness and altitude we provided in Figure 3 the standard error of the mean, showing that the observed diurnal variation is significant, even if the standard deviation of the parameters is large. For NLC occurrence rates we take another approach. Annual mean occurrences differ strongly between years as summarized in Table 1. Nevertheless, we find a similar diurnal variation within the individual years (not shown). The occurrence rate is highest in the early morning and lowest in the evening. The particular diurnal variations in a given year are well correlated with the complete data set, each smoothed by ±1h. Correlation coefficients are 0.66, 0.86, and 0.89 for 2010, 2011, and 2012, respectively (all significant). In 2013 the correlation coefficient is only 0.40 but in that year only few NLCs have been observed.
 Our observations are consistent with previous ground-based and space-based observations of NLCs. Out of these, only the data of the SHIMMER instrument cover midlatitudes. Stevens et al.  report maxima in polar mesospheric clouds (PMC) rates around 5LST and 18LST. In a subsequent paper, the NOGAPS/ALPHA (Navy Operational Global Atmospheric Prediction System - Advanced Level Physics and High Altitude) reproduced the morning maximum, but failed for the 18LST maximum [Stevens et al., 2010]. NLC variations at high latitudes of the northern or southern hemisphere are similar to our observations. The highest rates have always been found between 3 and 8LST, depending on instrument and site, and the lowest rates around 20LST [e.g., Chu et al., 2006; DeLand et al., 2011; Fiedler et al., 2011]. Results on the brightness variations are less consistent. For latitudes poleward of 70°N, DeLand et al.  observed brightest clouds between 8–10LST from space-based data, but Fiedler et al.  report brightness maxima around 6LST and 19LST from their ground-based data at 69°N. Potentially, local time variations in combination with orbit precession can explain part of these differences [Fiedler et al., 2011]. Several publications from higher latitudes report an anticorrelation of brightness and NLC altitude [Thayer et al., 2003; Chu et al., 2006; Fiedler et al., 2011]. This can not be confirmed from our midlatitude data.
 Some authors explain their observed variation of NLC parameters with the ambient temperatures and/or winds. Similar to reports from Gerding et al.  for Kühlungsborn, observations from Davis Station in Antarctica (69°S) show NLCs being limited to equatorward wind [Innis et al., 2008]. In contrast to these, Stevens et al.  observed a short period of poleward wind right before the PMC maximum. They raise the question whether poleward advection might be important for ice observation. Comparable to our data, high PMC rates in Stevens et al.  are not related to temperature minima.
 Our study shows that maxima in the equatorward wind do not coincide with the largest occurrence rates. Instead, we found the highest occurrence rates at the end of southward wind periods, i.e., the longer the air comes from the pole, the higher the chance for advection of NLCs. During northward wind periods there is a higher chance to advect ice-free air due to generally higher temperatures south of our location. Radar observations of ice particles (so called Mesospheric Summer Echoes) support this assumption [Zeller et al., 2009]. The stars and circles in the upper part of Figure 2 denote possible advection of NLCs using the following criteria: southward wind periods (v85km<−3m/s), supporting advection of NLCs, are marked with a star; northward wind and weak wind periods (v85km>−3 m/s), when advection from north is inhibited, are marked with a circle. The black symbols show where the wind direction agrees to an increase (southward wind, star) or decrease (northward or weak wind, circle) of NLC occurrence rates. The grey symbols show a disagreement. Obviously, wind direction typically fits to NLC increase or decrease. Most disagreements are found after periods of weak or northward wind. This indicates that the northward/weak wind periods shift the NLCs to the north, and it needs some time before advection from polar latitudes becomes effective, again. As a final note we need to point out that the suggested mechanism is based on averaged data. Nevertheless, it is in agreement with model results by Kiliani et al.  and case studies by Gerding et al. , showing that NLC are observed during southward wind only, i.e., advected.
 In conclusion, diurnal variations of NLC occurrence rates and brightness at Kühlungsborn (54°N, 12°E) are dominated by the ambient wind conditions and advection from the north. During southward wind periods occurrence rates are generally increasing, while they decrease during northward/weak wind. Tidal temperature variations seem to play a minor role for NLC existence, but our data indicate that planetary waves are needed to provide sufficiently low temperatures.
 We acknowledge the support in lidar operation and maintenance of Torsten Köpnick and Michael Priester. Sebastian Mitreiter and Karl-Georg Eller are representative for numerous students helping with continuous lidar soundings. Part of this work was supported by the Deutsche Forschungsgemeinschaft (DFG) under grant GE 1625/2-1.
 The Editor thanks Xinzhao Chu and an anonymous reviewer for their assistance in evaluating this paper.