Particle properties and water content of noctilucent clouds and their interannual variation

Authors


Abstract

[1] Noctilucent clouds (NLC) have been observed by a multicolor lidar in northern Norway (69°N, 16°E). From three backscatter coefficients we calculate the parameters of a monomodal particle size distribution. We deduce the mean of the size distribution, the width, and the average number density of the ensemble. Using the backscatter coefficients at the peak of the layer the particle size above ALOMAR is investigated by comparing the observations with model results for spherical and aspherical particles assuming either lognormal or Gaussian size distribution. From the analysis of 645 particle size soundings (142 h of measurements) we find that the average size of all NLC particles above ALOMAR from 1998 to 2005 is 47.7 ± 1 nm for cylinders with Gaussian distribution while it is 39.7 ± 1 nm for the traditional model having spherical particles with lognormal distribution. The distribution width is 16.6 ± 0.5 nm for Gaussian distributed cylinders while the particle number density is 85 ± 6 cm−3. We compare our results in detail to previously published measurements and find a satisfying agreement between the observations taking into account the limitations of previous studies and the different locations of the measurements. From the particle properties we calculate a mean surface density of (4.4 ± 0.2) × 10−8 cm2/cm3 and a mean volume density of (6.0 ± 0.2) × 10−14 cm3/cm3. The mean volume density of faint and strong clouds is 1.6 × 10−14 cm3/cm3 and 7.9 × 10−14 cm3/cm3, respectively. From the volume density we calculate the year-to-year variation of the seasonal mean cloud water content to be about 40% and the average observable NLC ice mass flux through 70°N to be about 11 kilotons.

1. Introduction

[2] Noctilucent clouds (NLC) are found in the polar summer mesopause region roughly 83 km above the surface in both hemispheres [Jesse, 1889, 1896]. These ice clouds form in the extreme environmental conditions with the lowest temperatures in the atmosphere [Hervig et al., 2001]. Since the environment of the clouds is difficult to sound with other methods, NLC are used as tracer for processes in the mesopause region. The ice particles forming NLC have grown to a size that scatters light sufficiently, while the smaller icy particles that are not visible to current optical instruments, are still detectable with radar instruments and in situ detectors [e.g., Rapp and Lübken, 2004]. Particles detected by the radar instruments are roughly a few nanometers small while the NLC particles observed by optical instruments are on the order of 50 nm.

[3] NLC observations have in common that they rely on a quantity similar to the brightness that a naked eye observer sees from the ground. This “brightness” is monitored by naked eye, ground-based active lidar instruments, rocket-borne experiments and spaceborne passive remote sensing [Gadsden and Schröder, 1989; Romejko et al., 2003; Fiedler et al., 2003; Gumbel et al., 2001; Thomas et al., 1991; DeLand et al., 2003, 2006]. For active sounding of NLC by lidar instruments, the “brightness” is quantified by the backscatter coefficient (either at the layer peak or integrated over the layer, for definition see equation (2)). The brightness of the clouds depends on the particle size to the power of five to six when observed with visual or infrared wavelengths [Witt, 1968], and even in the UV, where the particle size effect on the brightness of the clouds is smaller, the particle size is primarily determined by the brightness [e.g., Baumgarten and Thomas, 2006, Figure 1].

[4] It has recently become clear that NLC strongly influence their environment by trace gas redistribution like the freeze drying effect of the polar mesopause region [e.g., Turco et al., 1982; von Zahn and Berger, 2003], which has in fact been observed from satellite by Hervig et al. [2003]. Furthermore it has been observed that metal atoms usually present in the mesosphere/lower thermosphere (MLT) region vanish in the presence of NLC particles [von Zahn et al., 1988; Plane et al., 2004; Lübken and Höffner, 2004]. Since such heterogeneous chemistry depends on the surface area of the particles present, the size, shape, and number density of NLC particles are essential for the reaction rates of such processes [Raizada et al., 2007]. The volume density of the particle ensemble is important for the water budget of the mesopause, especially to identify anthropogenic modifications of the mesopause region, e.g., by the propellant of liquid fueled rockets like the space shuttle [Stevens et al., 2005].

[5] When interpreting the results from remote sensing of NLC it has to be taken into account that the instruments always observe an ensemble of particles. Even for instruments with a relatively high spatial (Δx = 15 m, Δz = 150 m) and temporal resolution (Δt = 1 min) the number of scatterers in the sounding volume is roughly 1014. For current satellite instruments (Δx = 50 km, Δy = 200 km, Δz = 1.5 km) the number of scatterers observed is roughly 1021. On the other hand the NLC show large brightness variations on short timescales (hours) due to the interaction of the NLC with waves in the ambient atmosphere [Witt, 1962; Jensen and Thomas, 1994; Rapp et al., 2002]. On even shorter scales (minutes or less), where NLC behave as an inert tracer, those brightness variations are presumably dominated by changes in the number density [Fritts et al., 1993]. To investigate the particle property variations leading to the observed brightness variations a high temporal resolution of a few minutes and spatial resolution of a few kilometers are needed. If high temporal and spatial resolutions are not used the shape of the particle size distribution might not be described well by microphysical models.

[6] Recent size measurements generally agree with each other and the r = 50 nm “standard” established by Thomas and McKay [1985] and Rusch et al. [1991]. Even without the restriction of the retrieval to a fixed distribution width, von Cossart et al. [1999] confirmed this standard. Most of the particle size retrievals rely on the assumption that the optical properties of NLC are well described by spheres, although the particles are not expected to be spherical from the microphysical point of view [Gadsden, 1977; Turco et al., 1982]. Recently it could be shown that nonspherical particles indeed exist in NLC [Baumgarten et al., 2002; Eremenko et al., 2005]. Taking these observations into account, the optical signatures of the cloud particles are to the first order still well described by the assumption of spherical particles [Baumgarten and Thomas, 2006]. However, for detailed analyses, like particle size retrieval or the interpretation of spectral signatures of NLC, the nonspherical nature of the cloud particles has to be taken into account [Rapp et al., 2007]. For the special case of three color measurements with a lidar using widely separated wavelengths, it was observed that the particle size retrieval is not sensitive to the particle shape. By extending the analysis to nonspherical particles a larger fraction of the observed NLC agrees to the optical model and can be used for particle size retrievals [Baumgarten et al., 2007]. With this new robust method the volume equivalent size of the particles can be retrieved from the extensive multiannual data set of NLC observations above ALOMAR to investigate the mean particle properties and also the properties of different cloud classes observed. The investigation of cloud classes gives insight into the nature of the particles forming the clouds and also into the amount of water trapped in the different cloud classes. This additionally allows reviewing the year-to-year variation of the cloud particles and the ice mass above ALOMAR.

[7] In the following sections we will briefly describe the method for particle size retrieval and then describe the database of NLC observations with the ALOMAR RMR-lidar with respect to the particle property retrieval. Then we will present the application of this method to data collected during the summers 1998–2005. We use this data to assess year-to-year variations of the particle properties and then discuss our observations in the context of previously published particle properties.

2. Instrumentation and Data Analysis

[8] The ALOMAR Rayleigh/Mie/Raman (RMR) lidar is an active remote sensing instrument for investigation of the Arctic middle atmosphere, located at the island Andøya in northern Norway (69°N, 16°E) and is operated on a routine basis throughout the year [von Zahn et al., 2000]. The lidar measures relative density profiles and particle (aerosol) properties in the stratosphere and mesosphere. Throughout the NLC season (1 June to 15 August) the lidar is held operational 24 h a day since 1997 to take chance of even short measurements permitted by the weather conditions. The size of the back-scattering particles is derived from the measurements at three widely separated wavelengths (1064 nm, 532 nm and 355 nm) by comparison to modeled particle scattering. As emitter we use the primary, frequency doubled and tripled emissions of a Nd:YAG laser and at the receiving side a 1.8 m diameter telescope with a field of view of 180 μrad (18 m at 100 km distance) and narrow band (4–10 pm) spectral filters [von Zahn et al., 2000]. We optimized the instrument for the wavelength of 532 nm allowing detecting NLC during all local times of the Arctic summer at this wavelength [Fiedler et al., 2004]. For each of the three wavelengths used, the detectors record photons emitted by the laser and backscattered from air molecules and NLC particles as function of altitude z as well as sunlight scattered by the atmosphere. After subtraction of both solar background and thermionic emission of the detectors an altitude profile of the total backscatter signal Sλ(z) is obtained. The ratio of the measured total backscatter signal Sλ(z) and the molecular signal Sλ,M(z) yields the backscatter ratio Rλ(z) as a measure for the presence of NLC particles (Rλ(z) > 1):

equation image

where βλ,M(z) and βλ,NLC(z) are the volume backscatter coefficients for air molecules and NLC particles, respectively. Using air densities at the lidar location given by Lübken [1999] and air backscattering cross sections listed by Bucholtz [1995] we derive βλ,M(z), and finally the aerosol volume backscatter coefficient βλ,NLC(z). We calculated the measurement uncertainty by propagation of the statistical uncertainty when counting photons (Poisson-Statistics). Gaussian error propagation is applied to the process of background subtraction and normalization of the received signal Sλ(z) to the expected molecular signal Sλ,M(z). The signal is normalized in the altitude range of 45 to 50 km.

[9] The backscattering coefficient depends directly on the number density N and the differential scattering cross section (equation image(λ)). Through the differential scattering cross section the backscattering coefficient depends indirectly on the particle properties:

equation image

[10] To separate the effect of the average particle number density N and the particle properties (size, shape, distribution type, distribution width), we calculate the color ratios CR355 and CR1064 with respect to the most sensitive wavelength of the lidar λ = 532 nm:

equation image

[11] The lidar is able to measure the vertical profile of backscattering by the NLC, but to achieve the best possible data quality we limit the analysis to the layer around the altitude of maximum backscattering (zmax). We sum up the backscattering in the altitude range where β532 nm,NLC(z) >0.7 × βmax. We use βmaxβ532 nm,NLC(zmax) which we define as the brightness of the NLC. In addition, the backscattering has to be significant which we test by the 2-sigma measurement uncertainty 2 × Δβ532 nm,NLC(z) < β532 nm,NLC(z). These significant detections of NLC are analyzed for particle sizes when in addition the measurement errors of the color ratios are small ΔCR1064 nm,NLC(z) < 0.08 and ΔCR355 nm,NLC(z) < 1.0. These limits were found to give a good compromise of the precision of the single measurement and the number of measurements analyzed.

[12] For the NLC signatures identified we compare the measured color ratios CR1064 nm,NLC(z) and CR355 nm,NLC(z) with tabulated color ratios for different particle shapes, particle size distributions, distribution widths, and particle sizes. The tables of color ratios were derived from differential scattering cross sections that were calculated using the so-called T-matrix method as realized by Mishchenko and Travis [1998]. The model results for different parameter combinations were projected into the measurement space of color ratios CR1064 nm,NLC(z) and CR355 nm,NLC(z). The parameter range investigated is listed in Table 1. For color ratio combinations where multiple sets of parameter combinations are found the variance of the parameters is treated as model uncertainty and propagated together with the uncertainty of the measurement to the total uncertainty of the retrieved particle properties.

Table 1. Overview of Parameters and Parameter Ranges Used for Particle Size Retrievala
PropertyValues
  • a

    The step size of the varying parameters is given in brackets. For the axis ratio (AR) the stepping was 0.2 for AR > 1 and we used the corresponding reciprocal values for AR < 1.

Particle shape 
Needles/platesAR = 0.1. 1. 10 (0.2)
SpheroidsAR = 0.1. 1. 10 (0.2)
SpheresAR = 1
Distribution type 
Gaussians = 2. 50 nm (1 nm)
Lognormalσ = 1.0. 2.2 (0.01)
Monodisperseσ = 1.0
Particle size 
Volume-equivalentr = 1. 350 nm (1 nm)
Refractive index [Warren, 1984] 
λ = 355 nm1.324 + 3.632−9i
λ = 532 nm1.312 + 2.623−9i
λ = 1064 nm1.300 + 1.901−6i

[13] After the retrieval of mean and width of the particle size distribution from the color ratios we calculate the number density from the backscatter coefficient β532 nm,NLC(z) via equation (2). A description of the method and a detailed investigation of the sensitivity of the derived particle size to the particle shape are given by Baumgarten et al. [2007]. From the retrieved particle properties we derive the ensemble properties like ice volume density and surface density at the layer peak. Please note that the surface density is also called volumetric or specific surface area [Plane et al., 2004; Murray and Plane, 2005].

3. Database

[14] The ALOMAR RMR lidar has been operated for approximately 2545 h during the NLC seasons from 1998 to 2005 in three-color mode. The lidar observed ∼905 h of NLC signatures and ∼175 h with βmax >13 × 10−10 m−1sr−1 which are so called strong NLC [Fiedler et al., 2003]. This data set was analyzed with a temporal resolution of 14 min and a vertical resolution of 150 m. We apply a vertical 5-point binomial filter which minimizes the smearing of small-scale structures in NLC but still allows deducing the particle properties. For a wind speed of 40 m/s the temporal resolution corresponds to a horizontal advection of about 34 km. An example of the small-scale structures is given by Baumgarten et al. [2007]. We investigate the particle properties at the peak of the layer which we define as the altitude range where the backscatter coefficients are larger than 70% of βmax. In average this selection criterion resulted in a peak extent of about Δz = 500 m. After applying the color ratio selection as described in section 2 we identified 142 h of NLC soundings of high enough quality to derive particle properties. Table 2 gives an overview of the measurements from 1998 to 2005. We observe that the number of measurements is very different from year to year. While there are more than 100 soundings per year since 2003 there are only 48 soundings in the year 1998 and less than 10 soundings per year from 1999 to 2002. Although the lidar operated in multicolor mode during the years 1999 to 2002 for 175–325 h per season there are very few particle property soundings because of two reasons: (1) year-to-year variations of NLC (there were fewer and weaker NLC in these years) and (2) the performance of the channels λ = 355 nm and λ = 1064 nm was lower than in other years. So 90% of all measurements were performed in the years 2003–2005. The NLC observations allow the particle size retrieval for measurements from the 2nd week of June to the end of the season covering nearly all local times. However, there are no particle size measurements around noon. The retrieval of particle size is possible if either the solar background light is low or a moderate to strong cloud occurs. Because of these requirements we did not measure particle sizes during noon since the brightness of NLC shows a minimum around noon and the solar background is highest. Additionally the gap in the data is fostered by the diurnal variation of the NLC occurrence as reported by Fiedler et al. [2005], where they observed that the clouds preferably occur in the morning hours and they occur least frequently around noon.

Table 2. Database of NLC Measurements 1998–2005 Used for Particle Property Retrievala
YearTotal Time, hNumber of MeasurementsNumber of EventsFirst Day of ObservationLast Day of ObservationTime of Day, LST
  • a

    Per year the total observation time was accumulated by the number of measurements listed. The number of events gives the number of independent NLC events, which are defined as independent of each other when a cloud was not observed for more than 2 h. The hours (in local solar time) during which the observations were made are listed in the last column.

199810.748814 Jul4 Aug1730–0700
1999
20001.25218 Jul3 Aug2230–0200
2001
20021.6729 Jul30 Jul0030–0300
200324.31081223 Jun11 Aug1900–0600
200467.62991619 Jun4 Aug1300–1030
200541.51771014 Jun12 Aug1500–1100
1998–2005142.46455114 Jun12 Aug1300–1100

[15] The combination of the measurements covers now nearly each day of the year throughout the NLC season and all local times. This database includes also the data of summer 1998 analyzed by von Cossart et al. [1999], but the original raw data is treated with the new analysis method. A detailed comparison of the improved analysis method in comparison to the von Cossart et al. [1999] results is given in Baumgarten et al. [2007].

4. Observations

[16] The total of 645 selected multicolor measurements is the most comprehensive data set of active NLC soundings allowing deducing robustly the characteristic particle size of NLC. We will start with the combined data set of all NLC above ALOMAR to derive the characteristic particle size, distribution width, number density and surface density and volume density of the particle ensemble. Then we will divide the data set into three cloud classes defined by the brightness of the clouds following Fiedler et al. [2003]. They found that the year-to-year variation of NLC depends on the brightness class investigated. For example the strongest clouds showed a prominent year-to-year variation while the weak and medium clouds persistently reoccurred. A detailed understanding of the particles involved in these cloud classes will allow studying the cloud water content and help to identify sources for the observed year-to-year variations.

4.1. Characteristic Particle Size

[17] In this section we present the combined observations from 1998 to 2005 to deduce the mean particle size above ALOMAR. In Figure 1 we present all measured color ratios as points in color ratio space. The rainbow colored area indicates the color ratios combinations, that can be interpreted using the model with cylinders and Gaussian particle size distribution. We discuss in detail the results using the Gaussian size distribution assuming non spherical particles (cylinders) as found to be the best model to combine expectations from NLC modeling and observations [Rapp et al., 2007]. To generate meaningful average particle properties only those observations are taken into account that can be interpreted in the 1-sigma measurement uncertainty environment. The effect of this data reduction can be seen by comparing Figure 1 (top and bottom). Since the density of measurements is higher in the colored model solution range Figure 1 is misleading and overemphasizes outliers. In Figure 2 we show the corresponding density of measurements. We do not observe a significant difference of the density of measurements before and after filtering the results. The only obvious effect of the filtering is seen in the reduced occurrence of measurements with CR1064 < 0.05, those measurements are not compatible with the optical model. This reduced backscattering of IR light has been reported previously and is attributed to the occurrence of a “strange” noctilucent cloud [Alpers et al., 2001]. The authors could not identify a technical or optical reason for the missing IR signal. We confirm the occurrence of strange NLC but Figure 2 shows that most of the clouds show a spectral signature compatible with the current optical models of NLC.

Figure 1.

(top) Color ratios measured above ALOMAR from all clouds observed in 1998–2005 indicated by black and orange symbols. Black symbols indicate measurements that can be interpreted with the model. The colored area indicates the size of NLC particles assuming cylinders with Gaussian particle size distribution. (bottom) Only measurements that can be interpreted properly using the model assuming cylinders and Gaussian distribution. The length of the crosses is defined by the measurement uncertainty. For further explanations see text.

Figure 2.

(top) Density of measured color ratios for all data from Figure 1. (bottom) Density of measurements that can be interpreted in the 1-sigma uncertainty by the model with cylinders and Gaussian size distribution. The probability density is normalized to the maximum value.

[18] From our color ratio measurements 504 NLC observations (78%) can be interpreted with the cloud model of cylinder shaped particles. The remaining 22% of the observations are not compatible with the optical model within the 1-σ measurement uncertainty. To understand the non conformance of measurement and optical modeling we investigate these mismatches closer: 14% of the measurements show an IR deficit, while 8% show either an IR or UV excess. If we assume that the IR or UV excess is solely caused by measurements falling outside the 1-σ uncertainty interval than we observe that only about 6% of our measurements show an IR deficit and are so called “strange” NLC. From the statistical point of view it can happen with a probability of 32% that the measurements, because of their measurement uncertainty, do not agree with the model. If we extend the analysis to the 2-sigma level, we find that 93% of all measurements can be interpreted using the model. We have investigated the significance of the IR deficit by limiting our analysis to extremely precise color ratio measurements where ΔCR1064 nm,NLC(z) < 0.02 and ΔCR355 nm,NLC(z) < 0.25, i.e., 1/4 of the typical measurement uncertainty. We identified 99 extremely precise measurements and observed that in the 1-sigma or 2-sigma environment either 88% or 94% of the measurements are compatible with the model. Again 6% of the measurements showed an IR deficit, even in the 2-sigma environment.

[19] We now analyze all measurements that are compatible with the model. Those measurements are analyzed for particle properties and the results are shown in Figure 3. In Figure 3 we also show the complementary cumulative distribution function which gives the probability of finding parameters above a certain threshold. For example the likelihood of finding particles with radii r > 50 nm is 36% while particles larger than r > 100 nm occur less than 1% of the time. On the other end of the distribution of particle sizes observed we do not find any particles with sizes smaller than r = 20 nm, which is caused by the instrument sensitivity. When the particles are small the size retrieval becomes difficult since (1) the backscatter coefficient becomes small, and hence the measurement uncertainty of the color ratios increases and (2) the wavelength dependence of the color ratios comes close to the Rayleigh limit. The distribution width s shows a monomodal shape like distribution of the observed particle sizes. We observe no particle ensembles with distribution widths of less than 6 nm and larger than 40 nm. On the other hand 90% of the measurements show distribution widths between 12 nm and 28 nm. For the particle number density we observe that 80% of the measurements show number densities of less than 200 cm−3 and 50% with N < 100 cm−3.

Figure 3.

Distribution of particle properties in 1998–2005 above ALOMAR from all clouds observed. (top) Volume equivalent mean size of particles above ALOMAR. (middle) Distribution width. (bottom) Number density. The thick black line gives the complementary cumulative distribution function (right scale).

[20] We observe that the mean particle properties are: r = (47.7 ± 1.4) nm, s = (16.61 ± 0.52) nm, N = (85 ± 6) cm−3. Here the uncertainties correspond to the error of the mean calculated from the standard deviation and the single measurement uncertainty. The standard deviations from the distribution of the measurements are: δr = 17 nm, δs = 5 nm, δn = 103 cm−3. Figure 3 shows that the distribution of number densities is not symmetric. To take this into account when deriving the mean values, we calculated the mean and the standard deviation in log(n) space. An overview of the properties can be found in Table 3. The median values of the distributions are: r = 42.6 nm, s = 15.5 nm, N = 87.4 cm−3.

Table 3. Particle Sizes for Different Cloud Brightness Classes 1998–2005a
ClassMeasurement Time, hAltitude, kmBrightness βmax, 10−10/(m sr)Radius, nmDistribution Width, nmNumber Density, 1/cm3Volume Density, 10−14 cm3/cm3Surface Density, 10−8 cm2/cm3
  • a

    For definition of the classes see text. The measurement time gives the sum of the observations that could be interpreted with the model used (cylinders/Gaussian). For each particle property we list the mean, the error of the mean, and the standard deviation (in parentheses). The volume density and the surface density were retrieved for each measurement and the mean value as well as the error of the mean were calculated afterward.

Strong69.882.320.9 ± 0.5 (8.1)50.4 ± 2 (17)17.2 ± 0.7 (6)94 ± 8 (104)7.9 ± 0.3 (3.4)5.4 ± 0.3 (3.7)
Medium35.282.98.9 ± 0.1 (1.5)43.6 ± 2 (15)15.9 ± 0.9 (4)82 ± 11 (104)4.6 ± 0.3 (2.2)3.8 ± 0.3 (3.0)
Weak10.383.05.1 ± 0.1 (0.8)44.4 ± 5 (20)15.4 ± 1.7 (5)56 ± 16 (84)2.9 ± 0.4 (1.5)2.6 ± 0.5 (2.3)
Faint2.183.82.9 ± 0.2 (0.4)41.8 ± 9 (12)15.5 ± 3.6 (3)33 ± 20 (43)1.6 ± 0.4 (0.9)1.5 ± 0.6 (1.2)
Statistic115.382.515.8 ± 0.4 (9.0)47.8 ± 1 (17)16.6 ± 0.5 (5)86 ± 6 (104)6.1 ± 0.2 (3.5)4.5 ± 0.2 (3.5)
All117.482.615.6 ± 0.4 (9.1)47.7 ± 1 (17)16.6 ± 0.5 (5)85 ± 6 (103)6.0 ± 0.2 (3.5)4.4 ± 0.2 (3.5)

4.2. Particle Size of Different Brightness Classes

[21] Following Fiedler et al. [2003] we divided the NLC observations into different brightness classes where we use the maximum backscattering of the wavelength λ = 532 nm at the peak of the layer (βmax in units of 10−10m−1sr−1) as so called cloud brightness. We will call clouds with 0 < βmax < 4 faint, those with 4 ≤ βmax < 7 are called weak while NLC with 7 ≤ βmax < 13 are called medium and βmax > 13 are strong clouds. We also show the results for combined data set of all clouds and clouds with βmax > 4 where the latter class is called “statistic” as it is used for the occurrence rate statistics presented by Fiedler et al. [2003] or Fiedler et al. [2005]. Table 3 lists the results for the combined measurements 1998–2005 for the different brightness classes. There is a good data coverage for strong and medium clouds with 69.8 and 35.2 h observation time, respectively, while there are less observations for the weak clouds. The average brightness of the strong clouds is roughly 2.3 times the brightness of the medium clouds, while the size of the particles increased by 16%. From standard small particle approximation (βr5 to βr6) and the observed brightness increase we would expect that the particles of the strong clouds are 15 to 19% larger. From medium to strong clouds the approximation is obviously well valid, where βr6 represents the observed brightness and radius change of 16% better than βr5. From medium to weak or faint clouds the brightness varies more than we would expect from the particle size, that remains more or less constant within the measurement uncertainty. The brightness variations are more affected by the changes in number density and distribution width. The distribution widths of strong clouds are larger than those of medium to faint clouds, however the mean distribution width only increased from s = 15.5 nm to s = 17.2 nm. Also the number density of the strong clouds is larger than that of the medium clouds.

[22] Before studying the volume density and surface density of the clouds we like to point out that both quantities are calculated for each single measurement and are averaged afterward. We calculated the errors of the volume densities and surface densities not by Gaussian error propagation of the uncertainties of the particle properties, but using the same method as used for the particle property retrieval. Using this method the error calculation is not biased by the correlation of the properties (e.g., r and N). The volume density of the cloud is frequently called cloud water content and gives the volume mixing ratio of H2O in the clouds. Since all of the parameters describing the particle ensemble increase from faint to strong clouds also the volume density and the surface density in the clouds increases. We observe that the strong clouds carry 70% more water than the medium clouds, and weak clouds carry only 63% of the water in medium clouds. Faint clouds carry only 20% of the water of strong clouds. In previous studies the assumption was often made that the cloud water content is well described by the small particle approximation and the assumption of constant number densities. To investigate the quality of this approximated cloud water content we also apply this approximation to our data but compare it to the measured cloud water content. Using the simplified approximation we calculate that the observed brightness increase from medium to strong clouds corresponds to a volume density increase of 33% and that faint clouds carry 52% of the water of strong clouds. By comparing the observed volume density change of 20% from faint to strong clouds, with the result from the simple approximation (52%) we find that the simple approximation under estimates the different cloud water content by a factor of two. This is caused by the fact that the different cloud classes also show differences in the distribution widths and the number densities.

[23] The surface area available for heterogeneous chemical reactions is 42% larger for strong clouds than for medium clouds while weak and faint clouds only offer 68% and 39% of the surface area of medium clouds, respectively. The particle properties of all clouds are dominated by strong and medium clouds and hence the values fall between those of strong and medium cloud types. The values derived for those clouds that are used for statistics of the multiannual data set above ALOMAR [Fiedler et al., 2003] are within the measurement uncertainty of the mean properties of all clouds. To check the year-to-year variations of the cloud water content of the different clouds and investigate the robustness of the mean value we have investigated the 2 years 2004 and 2005 separately. For 2004 the values are: 7.8, 4.8, 3.2, 1.9 × 10−14 cm3/cm3 for strong, medium, weak, and faint clouds, respectively. For 2005 the values for the different cloud classes are: 8.5, 4.7, 2.8, 1.4 × 10−14 cm3/cm3, respectively. The comparison shows that the cloud classes show a similar cloud water content within the measurement uncertainty and well within the standard deviations of the different cloud events in each cloud class. The observed year-to-year variations of the cloud water content are 10% or less for weak medium and strong clouds from 2004 to 2005.

4.3. Year-to-Year Variation of Particle Size

[24] To investigate the year-to-year variations we have listed the results of the particle properties for each year in Table 4. For the years 2000 and 2002 there are only a few measurements and it is likely that the clouds investigated are biased toward strong clouds, that give the best signal to noise ratio. The small number of measurements is also the underlying reason for the larger measurement uncertainty of the values in these years. Nevertheless, we observe that the variation of the particle size from 1998 to 2005 is less than Δr = 17 nm with the minimum annual mean particle size of 44.7 nm in the year 2005 and the maximum of about 62 nm in 1998 and 2002. In the years 2003–2005 the mean particle size is between 48 and 44 nm. The observed variation of the annual mean values is slightly smaller than the standard deviation per year, which also comprise the seasonal variations of the particle size. However the observed variations are larger than the error of the mean which is less than 6 nm for most of the years. The distribution width of the particle ensemble remains constant within the error bars where the year 2004 shows the largest distribution width of s = 17.7 nm and the minimum is found in the year 2005 with s = 15.1 nm. When neglecting those years with an uncertainty of the number density larger than 30 cm−3 we observe that the particle number density has increased from the value of 55 cm−3 in 1998 to 72–105 cm−3 in the years 2003–2005. The volume density decreased from 6.8 × 10−14 cm3/cm3 in 1998 to 5.1 × 10−14 cm3/cm3 in 2003 when neglecting years with uncertainties of the volume density of more than 2 × 10−14 cm3/cm3. The year-to-year variation of the cloud water content is smaller than the standard deviation per season, and for all years (except 2003) also smaller than the errors of the mean. The year-to-year variation of the surface density is smaller than the variation throughout the seasons and also smaller than the error of the mean. In summary we observe that the volume density and the surface density remains constant over the years with variations of less than ±15%.

Table 4. Year-to-Year Variation of Particle Properties of All Cloudsa
YearMeasurement Time, hAltitude, kmBrightness βmax, 10−10/(m sr)Radius, nmDistribution Width, nmNumber Density, 1/cm3Volume Density, 10−14 cm3/cm3Surface Density, 10−8 cm2/cm3
  • a

    The measurement time gives the sum of the observations that could be interpreted with the model used (cylinders/Gaussian). For each particle property we list the mean, the error of the mean, and the standard deviation (in parentheses). The volume density and the surface density were retrieved for each measurement and the mean value as well as the error of the mean were calculated afterward. See also Figure 5.

19989.182.523.7 ± 2.1 (13.3)61.3 ± 6 (22)16.6 ± 2 (5.6)55 ± 13 (66)6.8 ± 0.8 (3.7)4.2 ± 0.6 (3.1)
20000.981.921.3 ± 2.5 (5.0)53.5 ± 16 (11)16.1 ± 6 (5.8)81 ± 50 (60)7.5 ± 2.0 (1.9)4.6 ± 1.7 (1.8)
20021.482.320.2 ± 3.9 (9.6)62.4 ± 16 (22)16.9 ± 6 (4.4)46 ± 36 (77)5.9 ± 2.2 (4.6)3.5 ± 1.7 (3.6)
200320.183.113.2 ± 0.7 (6.3)47.8 ± 3 (16)16.5 ± 1 (4.7)72 ± 12 (87)5.1 ± 0.4 (3.0)3.8 ± 0.4 (3.1)
200451.882.314.9 ± 0.5 (7.9)46.7 ± 2 (16)17.7 ± 1 (5.6)86 ± 9 (104)6.0 ± 0.3 (3.3)4.5 ± 0.3 (3.5)
200534.182.715.6 ± 0.8 (9.5)44.7 ± 2 (13)15.1 ± 1 (4.2)105 ± 14 (124)6.2 ± 0.5 (4.1)4.8 ± 0.5 (4.0)

5. Discussion

[25] The results listed in the previous section were derived using a state of the art model of the optical properties of NLC [Baumgarten et al., 2007]. In contrast to this, previous observations were often analyzed using a spherical particle model with a lognormal distribution. To compare our results to previously published values we have also analyzed the data set using the spherical particle, lognormal size distribution model. The results are shown in Figure 4. We observe that the mean particle size is r = 39.7 ± 1 nm while the mean distribution width is σ = 1.47 ± 0.01 and the number density is N = 183 ± 16 cm−3. A close look at the distributions in Figure 4 shows that for the particle size and the number density the distributions are asymmetric, and hence the mean value is not describing the distributions very well. Since previous publications typically report only mean properties (or single measurements) we follow the procedure here and discuss only the mean values. Table 5 lists the results from the analysis of the different cloud classes. When comparing the results of the different models used we observe that the mean particle size decreases by Δr = 8 nm when using the spherical particle model with lognormal distribution, while the number density increased by about a factor of two. For the set of our measurements and our instrument configuration (backscattering geometry and three wavelengths used) we observe that the distribution width of s = 17 nm of the cylinders and Gaussian distribution equals a width of σ = 1.47 for spheres and lognormal distribution. When using the even more simplified model of monodisperse spheres, used by 6 out of 15 publications, we observe that the particle properties are: r = 68 ± 1 nm, N = 31 ± 6 cm−3 and the standard deviations are 17 nm and 33 cm−3, respectively. The number of measurements with the ALOMAR RMR-lidar that can be interpreted using this simplified model decreases to 53 h so only 50% of the measurements that are compatible to the other models can be interpreted with this simplified model. The simplified model overestimates the true particle size by Δr = 20 nm.

Figure 4.

Distribution of particle properties in 1998–2005 above ALOMAR using the model with spheres and lognormal distribution. (top) Volume equivalent mean size of particles above ALOMAR. (middle) Distribution width. (bottom) Number density. The thick black line gives the complementary cumulative distribution function (right scale).

Table 5. Particle Sizes for Different Cloud Brightness Classes 1998–2005 Assuming Spheres and Lognormal Distributiona
ClassMeasurement Time, hAltitude, kmBrightness βmax, 10−10/(m sr)Radius, nmDistribution Width, nmNumber Density, 1/cm3Volume Density, 10−14 cm3/cm3Surface Density, 10−8 cm2/cm3
  • a

    For definition of the classes see text for Table 3. The measurement time gives the sum of the observations that could be interpreted with the model used (spheres/lognormal). For each particle property we list the mean, the error of the mean, and the standard deviation (in parentheses). The volume density and the surface density were retrieved for each measurement and the mean value as well as the error of the mean were calculated afterward.

Strong69.382.321.1 ± 0.5 (8.1)42.9 ± 1.4 (17)1.45 ± 0.01 (0.16)189 ± 19 (252)7.5 ± 0.3 (3.2)4.2 ± 0.2 (2.7)
Medium34.582.88.9 ± 0.1 (1.5)35.3 ± 2.0 (14)1.49 ± 0.03 (0.13)190 ± 31 (250)4.3 ± 0.3 (1.9)2.9 ± 0.3 (2.0)
Weak9.683.15.1 ± 0.1 (0.8)34.8 ± 4.2 (16)1.50 ± 0.05 (0.13)144 ± 52 (231)2.8 ± 0.4 (1.4)1.9 ± 0.4 (1.6)
Faint2.183.82.9 ± 0.2 (0.4)31.6 ± 8.6 (12)1.54 ± 0.12 (0.10)106 ± 82 (149)1.6 ± 0.4 (0.9)1.2 ± 0.5 (0.9)
Statistic113.482.516.0 ± 0.4 (9.0)39.9 ± 1.1 (17)1.47 ± 0.01 (0.15)185 ± 16 (251)5.8 ± 0.2 (3.2)3.5 ± 0.2 (2.6)
All115.582.615.8 ± 0.4 (9.1)39.7 ± 1.1 (17)1.47 ± 0.01 (0.15)183 ± 16 (249)5.7 ± 0.2 (3.3)3.5 ± 0.2 (2.6)

[26] There are reports on particle size measurements where bimodal distributions could not be excluded from the measurements, but these reports are limited to satellite observations [Carbary et al., 2004; Debrestian et al., 1997]. One explanation of the authors involved different growth modes in the vertical structure of the NLC. Following the growth-sedimentation model it was speculated that there exists a mode of large particles with sizes of 200–220 nm in the lower part of the NLC while there is a second mode with particles of ∼50 nm higher up in the cloud layer. From the satellite measurements the authors could not distinguish between these two growth modes because of the coarse vertical resolution of the experiments used, but the ALOMAR RMR-lidar could clearly resolve a two layer structure. So far we have not investigated the clouds with respect to a double layer structure, but only analyzed the layer with the strongest backscattering. We observed that only 1% of the measurements show particles larger than r = 100 nm while 90% of our observations show particle sizes of 20–80 nm (Figure 3). If we follow the argument that the bimodal structure of NLC is caused by a double layer structure of NLC which cannot be resolved by a (satellite) instrument with a coarse vertical resolution, we can attribute the layer we investigate to the large growth mode. The maximum particle size we have derived at the peak of the layer is 136 nm which is much less than the speculated large growth mode. On the other hand we must acknowledge that not all measurements are compatible with the model within the 1-σ uncertainty limit. Following Rapp et al. [2007] we can directly compare our color ratio measurements to results proposed by Carbary et al. [2004]. They find that the color ratio 1064/532 nm can be as large as CR1064nm = 0.47 for the large growth mode. The maximum color ratio 1064/532 nm observed above ALOMAR is CR1064nm = 0.34. So we conclude that we have not observed clouds of the type reported by Carbary et al. [2004] or Debrestian et al. [1997]. Please note that other authors could not find bimodal clouds even from satellite observations [von Savigny et al., 2007].

[27] In Table 6 we have compiled the results of all particle size measurements with monomodal or monodisperse size distribution published in literature that can be attributed to noctilucent clouds occurring in the summer mesopause region. We have not included results from reanalyses of listed measurements [e.g., Mishchenko, 1991]. We have also ignored the different analyses of the observations reported by Grishin [1955] as the analyses show a large uncertainty. For example Deirmendjian and Vestine [1959] discuss particle sizes in the range of 100 nm to 400 nm. The retrieval of particle size can easily be misinterpreted when using a model of spherical particles without taking the distribution width into account, since the symmetry of the spherical particles produces strong resonances that can lead to two possible solutions, one on the large particle side of the resonance and one on the small particle side. Taking the distribution of particle sizes into account smears out these resonances. Additionally the distribution of particle shapes in a scattering volume will lead to further suppression of the resonances. Since the observations were performed at different locations, in different years and on clouds of different brightnesses an exact comparison is difficult. When investigating the particle sizes of different cloud brightness classes we observed that the size of strong clouds is 10 nm larger than that of faint clouds. At the same time the year-to-year changes of the particle sizes are about 15 nm. Hence differences in particle sizes of about 20 nm can be attributed to cloud variability. This is especially true when comparing to those published particle sizes that were deduced from a single cloud observations. Furthermore those results listed using the simplified model of monodisperse particle sizes comprise an overestimation of the particle size of about 20 nm. Taking these uncertainties into account we observe a good agreement to the results published by von Savigny and Burrows [2007], Karlsson and Rapp [2006] and von Savigny et al. [2005]. This agreement is unexpectedly good, especially when taking the different locations, cloud classes and years of the observations into account. We conclude that the variability of the ambient background atmosphere that the NLC particles experienced before being detected above ALOMAR is comparable to latitudinal or hemispheric differences of the atmospheric background, most important the temperature. If we force our particle size retrieval to narrow distributions we obtain also particle sizes close to the values given by Carbary et al. [2002] (e.g., 68 nm for σ = 1.0). Even without forcing, the results agree within the variability discussed above.

Table 6. Properties of NLC Particles From Published Measurements Sorted by Date of Measurementsa
Particle Size, nmSize Distribution WidthNumber Density, cm−3Altitude, kmBrightness, arb.Scattering Angle, degWavelength, nmDateLocationReference
  • a

    The size is the median (or mean) radius or volume equivalent sphere radius of the distribution [Baumgarten et al., 2007]. The distribution width is given for the lognormal distribution, unless otherwise stated (s is the width of the Gaussian distribution). The brightness is the quantity used in the corresponding publications. β532nm stands for the backscatter coefficient at a wavelength of 532nm in units 10−10 m−1 sr−1. RB300nm is a relative brightness R532nm the backscatter ratio, τ448nm the slant optical depth, RE the Rayleigh-equivalent altitude, and TB410nm the total brightness.

  • b

    4/5 July 2003, 21 December 2003 to 4 January 2004.

  • c

    Baumgarten [2001].

  • d

    10−12 W cm−2 sr−1Å−1.

  • e

    Witt [1962].

41 ± 31.43 ± 0.03226 ± 4682.316.6, β532 nm180355, 532, 106414 Jun to 25 Jul 200569°N, 16°Elidar [Baumgarten et al., 2007]
(41 ± 4)(s = 22 ± 2 nm)(97 ± 19)82.316.6, β532 nm180355, 532, 106414 Jun to 25 Jul 200569°N, 16°Elidar [Baumgarten et al., 2007]
30–501.4, assumed37–65265–3001–31 Jul 200558–82°Nsatellite [von Savigny and Burrows, 2007]
(60–80)(s = 12 nm), assumed37–65265–3001–31 Jul 200558–82°Nsatellite [von Savigny and Burrows, 2007]
56–761.0, assumed73–84277–304, 6831 Jan to 15 Feb 200555–90°Ssatellite [Karlsson and Rapp, 2006]
10–501.4, assumed80.6–85.720–73, RB300 nm85–95290–3102003, 2003/2004b70–82°N, 75–82°Ssatellite [von Savigny et al., 2005]
65.2 ± 2.21.15 ± 0.04125–145200–31513 Jul/5 Aug 199968–84°Nsatellite [Carbary et al., 2002]
10–421.0, assumed200–300079.5–85.50.3–15, cβ532 nm40–1152245 Jul/14 Jul 199969°N, 16°Erocket [Gumbel et al., 2001]
20.2–27.51.5–1.6260–61081.320–50, R532 nm180355–77013/14 Aug 199854°N, 12°Elidar [Alpers et al., 2000]
51 ± 211.42 ± 0.2282 ± 5282.5 ± 111–38, β532 nm180355, 532, 106427 Jul to 4 Aug 199869°N, 16°Elidar [von Cossart et al., 1999]
16–441.2–1.6160–278082.214, β532 nm180308, 53213 Aug 199569°N, 16°Elidar [von Cossart et al., 1997]
20–701.413, assumed82–830.04.0.09, τ448nm0352–10605–12 Dec 199463–66°Ssatellite [Debrestian et al., 1997]
44–741.0, assumed14–36083.070 ± 2 km, RE105–110300–5002 Aug 199369°N, 21°Erocket [Gumbel and Witt, 1998]
20–801.4, assumed265, 296, 3931982–1985>60°Ssatellite/SME [Rusch et al., 1991]
80–1201.0, assumed1–383.5 ± 0.52.9–31,dTB410 nm74–84410, 54031 Jul 1971, 1 Aug 197369°N, 21°Erocket [Tozer and Beeson, 1974]
<501.0, assumed0.385.5–8979–81256, 5368 Jun 196869°N, 21°Erocket [Witt, 1976]
120–1301.0, assumed83.5 ± 2e20–60490, 61010 Aug 195863°N, 15°Ephotos [Witt, 1960]

[28] When we draw our attention to the size distribution width, we observe that only 5 out of 15 previous reports have investigated the particle size and distribution width at the same time, and only four out of five also retrieved the number density. This combined and common volume measurement of all parameters describing the particle distribution has up to now only been retrieved with multicolor lidar measurements. As these active remote sensing measurements require an extensive instrumental setup there are only two locations in the Northern Hemisphere where such particle property observations have been performed so far. The good agreement of the measurements performed with different methods indicates that the optical modeling of the scattering of light on NLC particles fits to the observations and is comparable to the instrumental uncertainties, otherwise we would have expected to see discrepancies between measurements performed at different wavelengths and with different scattering angles.

[29] However, it should be clear that all methods listed in Table 6 have in common that they are optical soundings which depend on the refractive index used for particle size retrieval. If for example the ice particles forming the NLC do not grow to a defined crystal structure but more to a fluffy conglomerate the refractive index is reduced, and the particle sizes listed are underestimating the particle extension. Fortunately, the volume equivalent radius will still represent the volume of particles when compressed to a compact body [Alpers et al., 2001].

[30] In Figure 5 we summarize the results of the year-to-year changes of the particle properties and combine particle information with results from NLC statistics [Fiedler et al., 2003]. We have shown in Table 3 that there are large differences in the actual volume density of the NLC particle ensembles for different cloud classes. As the year-to-year occurrence of the clouds depends on the cloud class we include this information to derive a seasonal mean cloud water content (CWC):

equation image
Figure 5.

Year-to-year changes of particle properties at the peak of the NLC layer. Shown are seasonal mean values and error of the mean calculated from the variability of the measurements. From top to bottom we show the volume equivalent sphere radius, the width of the Gaussian size distribution, the number density, cloud volume density, surface density, the occurrence rate of different cloud classes and the cloud water content. For occurrence rates and CWC we show bars, where the contribution of faint to strong clouds is stacked on top of each other (blue indicates faint and red indicates strong). Grey symbols show the CWC deduced from the total cloud occurrence rate, neglecting the dependence of the volume density on the cloud class.

[31] Here ORclass is the occurrence rate and Vclass is the volume density of a certain cloud class. ORclass is calculated from the total of 905 h of NLC signatures. The CWC variations are mainly induced by the year-to-year changes of the occurrence frequency while we have assumed that the actual cloud volume density of the different cloud classes is constant from year to year. This is supported by our observation that the actual cloud volume density varies only within the uncertainty limits of the measurements.

[32] We state that the CWC comprises only those clouds observable by the lidar and that there is a large number of clouds not observable by the lidar because of to small particle sizes. Such clouds are observable as so called PMSE by radar instruments. We have no direct information about the CWC of these clouds, but it has been observed that strong reduction of metal abundances only occurred in the vicinity of NLC visible to the lidar [Lübken and Höffner, 2004]. This implies that the surface density of PMSE is much less than that of NLC. Additionally, as long as there is virtually no year-to-year variation of the PMSE occurrence also the derived CWC will be dominated by the cloud classes with large volume densities, observable by the lidar [Bremer et al., 2006]. From our observations we find that there is a large variability (∼40%) in the CWC of the observable clouds. This variability is mainly driven by the NLC occurrence and appears to show a quasi-biannual variation. A more detailed investigation of this topic is ongoing and beyond the scope of this paper. Combining these observations of nearly invariant annual mean volume densities and the fact that the CWC shows a larger year-to-year variation we can speculate about the meaning of this observation: We come to the conclusion that once the NLC growth has started, it reaches an end state defined by the microphysical laws, while year-to-year changes of the ambient atmosphere modulate the likelihood for trapping the water vapor in the clouds.

[33] To compare our results with previous reports we calculate the so called NLC ice mass by scaling the CWC with the density of water ρice at the temperatures of the typical NLC altitude (T = 152 K) [Lübken, 1999; Pruppacher and Klett, 1997]:

equation image

[34] Furthermore we have to assume a vertical and horizontal extent of the clouds. We estimate the vertical extent of the cloud from the observed FWHM of the layer of 1.2 km under the assumption of a Gaussian altitude profile [Fiedler et al., 2003; Thayer et al., 2003]. If we derive the volume density profile from the so derived brightness profile (βr5 to βr6) we find that the vertical profile of the cloud contains 3.2 to 3.8 times more water than we observed at the peak of the layer. For the horizontal extent we use the area between 65°N and 75°N following previous publications [Stevens et al., 2005]. We calculate that we have observed an ice mass of MNLC = 358–385 tons ice to be present on seasonal and daily average. The year-to-year variation of the ice mass is about 107 tons. If we compare this to the results from different satellite instruments we observe about 3 times the mass than reported from HALOE (90 ± 17) tons and SBUV (136 ± 27) tons measurements [Stevens et al., 2005]. It should be taken into account that the satellite observations did not take the effect of different water content for different clouds classes into account. If we also neglect the cloud class effect and calculate the ice mass from the mean particle properties and the mean occurrence rate we yield an ice mass of MNLC = 511–588 tons.

[35] The fact that the ice mass observed by the lidar is larger than that observed by the satellite instruments SBUV and HALOE is likely to be caused by a higher sensitivity of the lidar to detect NLC. This fact was also observed by Stevens et al. [2007] who found that the SNOE satellite observes more than tree times the ice mass than SBUV. If we assume that the ice mass in the mentioned latitude range is replaced because of meridional advection we can estimate the seasonal mean ice advection through the 70°N latitude circle. For the meridional transport we use wind speed of about 5 m/s as observed above ALOMAR [Singer et al., 2005]. This meridional wind speed leads to a replacement of ice mass between 65–75°N every 2.5 days. We find that per season (75 days) the total seasonal ice mass in the latitude range of 65–75°N is about MNLC(total) = 11 kilotons. When we take the horizontal extent from the pole to about 50°N into account we find that the total hemispheric NLC ice mass is about a factor of four to eight larger [Olivero and Thomas, 1986]. Taking the uncertainties in the latitudinal occurrence of the different cloud classes into account we would summarize that the seasonal integrated NLC mass is about several 10 kilotons.

[36] We should note that this is only an estimation of the lower limit of ice mass present, since the lidar cannot detect the ice when the particles are too small and/or too rare. However we expect to see a large fraction of the NLC ice mass as we have shown the volume density of fainter cloud classes to be only about 1/5 of that of strong clouds.

6. Conclusion

[37] Reviewing the observations of particle sizes in the recent years combined with the extensive set of particle size measurements above ALOMAR makes the improved knowledge about the particle size very clear especially in the context of Gadsden and Schröder [1989, chap. 5, p. 74]. The authors noted, “Perhaps the conclusion to be drawn at this stage concerning the characteristic size of the scatterers, larger or smaller, is to be expressed in the two words that can be the verdict of a jury in Scotland: ‘Not Proven!’” Today, we come to the conclusion that the characteristic particle size of NLC particles is well investigated, but the harmonization of measurements and microphysical models on medium and short scales is still ongoing. Especially the shape of the size distribution needs to be investigated and the distribution widths need to be reconciled. For the interpretation of high-resolution lidar measurements the application of a Gaussian shaped size distribution seems to be appropriate from a microphysical point of view, but for coarse resolution instruments, with a temporal resolution of several days the microphysical simulation could be improved by inducing wave modulations of the atmospheric background fields. From the analysis of 645 particle size soundings we find that the average size of all NLC particles above ALOMAR from 1998 to 2005 is 47.7 ± 1 nm the distribution width is s = 16.6 ± 0.5 nm while the particle number density is 85 ± 6 cm−3. From the particle properties we calculate a surface density of (4.4 ± 0.2) × 10−8 cm2/cm3 and a volume density of (6.0 ± 0.2) × 10−14 cm3/cm3. We observe that faint clouds contain only about 1/5 of the ice volume found in strong clouds.

[38] We have estimated the year-to-year variations of the volume density to be only about 10% while the variation of the seasonal mean cloud water content is about 40%. The total seasonal NLC ice mass transport through the latitude circle 70°N is found to be about 11 kilotons.

Acknowledgments

[39] We gratefully acknowledge the support of the ALOMAR staff in helping to accumulate the extensive data set of NLC observations. The support by more than 10 voluntary lidar operators is also acknowledged. Furthermore, we are thankful to M. I. Mishchenko for providing the T-matrix code. This project received research funding from the European Community's 6th Framework Program under the project “ALOMAR eARI” (RITA-CT-2003-506208).

Ancillary

Advertisement