Volume 33, Issue 11
Atmospheric Science
Free Access

Latitudinal dependence of noctilucent cloud growth

B. Karlsson

B. Karlsson

Department of Meteorology, Stockholm University, Stockholm, Sweden

Search for more papers by this author
M. Rapp

M. Rapp

Leibniz Institute of Atmospheric Physics, Kühlungsborn, Germany

Also at Department of Meteorology, Stockholm University, Stockholm, Sweden.

Search for more papers by this author
First published: 10 June 2006
Citations: 24

Abstract

[1] The latitudinal variation of NLC particle sizes is studied using scattered solar radiation spectra from the OSIRIS instrument onboard the Odin satellite obtained during the austral summer period 1 January–17 February 2005. We show that the particles grow moderately larger when approaching the pole, with effective optical radii of 65 nm/76 nm at latitudes of 70°/90° S. Microphysical modeling suggests that particles grow larger as a consequence of a combined effect of the temperature decrease toward the pole, the available water vapor, and meridional transport times. Interestingly, at latitudes closest to the pole, NLC particle sizes show a bi-modal structure. This structure is suggested to arise from repeated growth cycles due to extended residence times in supersaturated air.

1. Introduction

[2] Noctilucent clouds (NLC), also known as polar mesospheric clouds (PMC), are ice clouds that occur in the extreme environment of the polar summer mesopause region at altitudes of ∼80–85 km. This area is characterized by extremely low temperatures (<150 K), very little water vapor (only few parts per million in mixing ratio) and low pressure (0.1–1 Pa) [e.g., Thomas, 1991]. Due to its difficult accessibility, observations of this region have so far been limited. It is therefore not surprising that the microphysics of NLC has remained uncertain [e.g., Rapp and Thomas, 2006].

[3] One property that has been studied with optical methods is the typical size of the ice particles in NLC [see Thomas and McKay, 1985]. Remote sensing from space allows an almost complete coverage of the geographical and temporal variability of NLC and has become a powerful complement to ground based observations [DeLand et al., 2006, and references therein].

[4] In the current paper, we investigate the latitudinal distribution of NLC particle sizes using observations from the OSIRIS spectrograph [Llewellyn et al., 2004] onboard the Odin satellite [Murtagh et al., 2002]. In section 2 we present the method used for retrieving sizes and in the following sections 3 and 4, the results are presented and discussed. Conclusions are summarized in section 5.

2. Retrieval Method

[5] The Odin satellite observes the atmosphere in limb geometry. Scanning up and down, it moves around the globe about 15 times a day in a sun-synchronous near-polar orbit. Odin is usually operated with its instruments viewing in the direction of the satellite motion, covering latitudes up to 82°. The OSIRIS instrument observes scattered solar radiation between 277 and 800 nm with a resolution of about 1 nm. We here present an overview of the NLC analysis techniques using OSIRIS limb observations. For more details, see Karlsson and Gumbel [2005].

[6] NLC leave a distinct signature throughout the scattered solar spectrum that OSIRIS covers [Karlsson and Gumbel, 2005, Figure 1]. The region below about 310 nm is particularly suitable for the study of NLC since absorption by stratospheric ozone keeps this part of the spectrum free from radiation from below [Witt, 1967]. Nevertheless, this UV part of the spectrum is highly irregular, mainly due to solar Fraunhofer lines. In addition, airglow emissions and resonant scattering by species like Mg, Mg+, OH and NO, occur close to NLC altitudes. It is thus crucial to select wavelengths that are representative only for the light scattering by NLC particles. In this study, narrow wavelength regions centered at 277, 283, 286, 289, 292, 300 and 304 nm are chosen for spectral analysis. After subtracting the Rayleigh background from the NLC limb scatter profiles, a directional albedo, defined as the ratio between the radiance scattered toward the instrument and the solar irradiance, is determined at each wavelength [Thomas and McKay, 1985]. In order to minimize effects that may arise from calibration uncertainties of the instrument, we use an instrument-specific solar spectrum calculated from measured Rayleigh spectra rather than an absolute solar spectrum taken from literature [Karlsson and Gumbel, 2005]. The directional albedo is then compared to theoretical scattering spectra calculated for various particle sizes and the scattering angle defined by the viewing geometry of each single measurement.

[7] As opposed to liquid droplets, NLC particles are unlikely to be perfect spheres. The shape of the OH vibrational band at 3300 cm−1 as well as polarization measurements with lidar indicate that the particles are slightly non-spherical [Baumgarten et al., 2002; Eremenko et al., 2005]. However, there is no a priori reason to anticipate any specific shape. Therefore we assume that the particles are randomly oriented compact grains whose light-scattering properties may to first approximation be described by Mie-theory.

[8] At relevant NLC particle sizes the Mie scattering cross section is highly dependent on the radius, that is, the larger particles contribute most to the signal. Certainly, the true NLC particle population will contain many different sizes. However, in order to minimize the number of assumptions, a single size parameter is fitted to the measured OSIRIS spectra. This means that directional albedo-spectra, each calculated for an ice sphere of a monodisperse radius 10 < r < 250 nm, are least square fitted to the OSIRIS observations. Accordingly, from each NLC observation, one effective optical radius is determined. This effective radius, which is a measure of the optically dominating particle size in the observed cloud volume, is defined as
equation image
where r is the radius, f(r) is the particle size distribution, and σ(r, λ, ϑ) is the differential Mie scattering cross section depending on wavelength λ and scattering angle ϑ.

[9] We note that this approach is equivalent to analyzing the observed directional albedo-values in terms of a lognormal particle size distribution with fixed distribution width σ [e.g., von Savigny et al., 2005]. For example, for our range of λ and ϑ-values, the effective radius reff and the median radius of a lognormal size distribution rmed with fixed σ = 1.4 are related by reff = 20 nm + 1.25 rmed for reff < 100 nm within an accuracy of <10%. This relationship is derived empirically from testing multiple values of rmed.

[10] It is also important to note that the effective optical radius retrieved from a given NLC population is via σ(r, λ, ϑ) dependent on λ and ϑ (see equation (1)). This is a basic caveat when comparing sizes retrieved from very different observation conditions. For the present study, this is not critical since the range of ϑ-values is restricted to the narrow range of 73°–84°.

[11] The retrieval is illustrated in Figure 1, where OSIRIS' directional albedo observations (circles) are plotted together with calculated spectra for particle radii between 10 and 220 nm. In this case, the best fit yields reff = 60 nm. This figure also shows that small particles cannot be distinguished from particles larger than 100 nm (solid lines) based on the wavelength range between 277 and 310 nm alone. However, the Mie spectra for the smaller and larger particles diverge substantially at wavelengths >∼400 nm. Thus by taking advantage of OSIRIS' measurements at longer wavelengths it is possible to discriminate the smaller particles from the larger particles in the fit (black diamond at 683 nm in Figure 1). Since this point is influenced by radiation from below, it is not included in the least square fitting but serves only as a divider between the two extremes of the particle size range.

Details are in the caption following the image
Theoretical spectra (lines) for several particle radii are plotted together with an OSIRIS measurement of directional albedo (circles). (See text for further details.) The curves and the directional albedos are normalized at 304 nm. The uncertainty in the directional albedo is less than 3% in this case and the error bars are within the size of the symbol.

3. Effective Radii Distributions and Variation with Latitude

[12] The method described in the section 2 was applied to the observations obtained during the NLC season in the southern summer 2004/2005. During the period 1 January to 16 February 2005, Odin was pointing off-track and toward the pole as it flew over the southern hemisphere, thereby extending the latitude coverage all the way up to 90°. In addition, this mode of observation reduced the range of scattering angles to vary only from 73° to 84°.

[13] In Figure 2, retrieved effective radii are plotted against latitude. The solid line represents the mean value taken over 5° latitude bands, from 70° to 90° S. This picture shows that the mean radii actually show very little variation with latitude, that is, they only increase from ∼65 nm at 70° to 76 nm at the latitudes closest to the pole.

Details are in the caption following the image
Retrieved effective radii (black dots) plotted versus latitude. The bars denote retrieval uncertainties, which are determined from the uncertainty in the directional albedo. The gray line marks the average size within latitude bins of 5°, corresponding confidence intervals (at the 99% confidence level) are smaller than ±2.2 nm.

[14] For the study presented here, retrieved radii with relative errors exceeding 25% are excluded. The errors increase with decreasing signal strength [von Savigny et al., 2005]. In general, the cloud peak radiance decreases by a factor of about 3 between 90° and 70° latitude. In addition, the cloud occurrence frequency decreases toward lower latitudes. To avoid the results from being biased toward larger particles by excluding weaker signals at lower latitudes, we restrict the further discussion to latitudes ≥70° S where 85% of all observations allow the retrieval of particle radii within the above stated accuracy.

[15] Figure 3 shows histograms of retrieved effective radii taken over the same latitude bands of 5° as for the average in Figure 2, that is from 70° to 90° S. Clearly, effective radii between 70 and 100 nm are more frequent pole-ward of 80°S. Mean radii and standard deviations of these histograms are given in the insert of each panel. In addition, we have used a student's t-test to determine the confidence intervals of the mean radii to check whether the claimed increase toward the pole is significant [Taubenheim, 1965]. This analysis gives confidence intervals of <±2.2 nm. Hence, the observed increase of the mean from 65 nm to 76 nm is clearly significant. Finally, it is also interesting to notice the development of a bi-modal structure when approaching the pole. We note that a similar bi-modal behavior is not obvious in the brightness distribution of the clouds.

Details are in the caption following the image
Effective optical radii distributions for latitude bands between 90° and 70° S. In each panel we have indicated the number of observations N, and the mean radius 〈r〉, its standard deviation sd〈r〉, and the confidence interval c〈r〉.

4. Discussion

[16] In this section we address the question whether we may understand the observed latitudinal variation of effective NLC radii and what it might teach us about the atmosphere and NLC microphysics. We investigate the remarkable fact that the observed NLC radii only show very little variation over the considered latitude range. In order to identify the underlying physical processes, we use the Community Aerosol and Radiation Model for Atmospheres, CARMA [see Rapp and Thomas, 2006]. In order to study the latitudinal distribution of NLC parameters with the 1D version of CARMA, we have performed separate 48 hours simulations for a set of latitudes. At each latitude the model was initialized with atmospheric background parameters (temperatures, vertical winds, and water vapor) taken from Berger and von Zahn [2002, Figure 1]. We note that this approach does not allow us to study transport effects, neither may we expect to find a perfect match between model results and observations because used input-profiles from the model by Berger and von Zahn were derived for northern latitudes while our observations were performed in the southern hemisphere which appears to be slightly warmer than its northern counterpart [Hervig and Siskind, 2006].

[17] Still, our approach allows us to study how the local microphysical processes change with these changing background parameters in general. In order to keep the effect of neglected transport within limit, we restrict our analysis to latitudes above 70° where the mean meridional wind is <5 m/s.

[18] Figure 4a shows temperature profiles used along with corresponding frostpoint temperatures that were determined from the input water vapor profiles using the expression given by Murphy and Koop [2005, equation (8)]. For each of these pairs of profiles we have also calculated the water vapour column mass density available in the altitude range with supersaturation, that is, the altitude range where the local temperature is less than the frostpoint temperature. The corresponding column densities are given in the insert. As it turns out, the total available water vapor is almost constant poleward of 75°, whereas it shows a rapid decrease equatorward of the same latitude. Figure 4b shows CARMA results of mean effective optical radii (determined for λ = 290 nm and ϑ = 80°) and particle number densities as a function of latitude.

Details are in the caption following the image
(a) Input temperature (solid lines) and frostpoint-profiles (dashed lines) for different latitudes. Numbers indicate the column mass density of water vapor available within the supersaturated regions. (b) CARMA model results on the effective optical radius (for wavelength 290 nm and scattering angle 80°) and NLC particle number density (both along with their standard deviations over the 48h-model runs.) as a function of latitude.

[19] reff is close to constant at a value of 58–60 nm poleward of 75° but shows a moderate decay from 58 nm to 52 nm between 75° and 70°. In addition, for all these calculations the number density is near constant at ∼200 cm−3. (This number refers to the entire particle population and not the number of particles with reff.) We note that this is at apparent variance with previous model studies [e.g., Rapp and Thomas, 2006] which predict larger nucleation rates and hence the formation of more ice particles in the presence of lower temperatures near the pole. Indeed, also in our current model runs lower temperatures lead to the initial production of more particles. However, we did not only change temperatures from latitude to latitude but also the vertical winds. Since lower temperatures go together with stronger vertical upwelling, the enhanced particle production at lower temperatures is compensated by the enhanced loss of small ice particles by vertical transport in the initial phase of ice particle growth until the particles are too heavy to be blown away by the mean vertical wind. We also note that our results are generally consistent with the latitudinal structure of mean NLC particle radii and number densities derived with the 3d-model by von Zahn and Berger [2003, Figure 13].

[20] Our results suggest that the latitudinal structure of reff basically reflects the latitudinal variation of the total available water vapor in the vertical column at these latitudes (Figure 4a). Indeed, calculations of the ice column mass density on the basis of our model results show that poleward of 70° close to 100% of the available water vapour occurs in the ice phase. Since, at the same time, the number densities are constant (Figure 4b) the reff -increase with latitude basically follows the corresponding increase in available water vapor. We also note, however, that CARMA underestimates the actually observed radius increase with increasing latitude. A possible clue to understanding this difference is the bi-modal structure of the observed NLC distribution shown in Figure 3. This figure implies that there seems to be one persistent mode of NLCs with effective radii at ∼60 nm which appears at all latitudes. However, in addition to this first mode of clouds, a second mode develops toward the pole at ∼85 nm and becomes as prominent as the one at 60 nm at latitudes >85°. While the first mode at 60 nm might reflect clouds which have undergone only one lifecycle, the second mode might provide evidence for particles which after their first lifecycle shrunk down to subvisible sizes but then experienced a second lifecycle which ultimately led to bigger particle radii. This idea of an extended NLC life cycle is also supported by the fact that the meridional wind close to the pole is much smaller than at lower latitudes and hence allows ice particles to reside much longer in a latitude band close to the pole. A confirmation for this idea can ultimately only be provided by a coupled three-dimensional dynamical-microphysical model like the one by von Zahn and Berger [2003].

5. Summary

[21] NLC scattering properties in the southern hemispheric summer of 2004/05 have been studied using scattered sunlight measurements made by the OSIRIS instrument onboard the Odin satellite. An effective optical radius, which for given observation conditions is a measure of the optically dominating particle size in the cloud, was used for analysis. The results presented in this paper show that this effective radius grows (moderately) larger at higher latitudes. We suggest that this is a consequence of a combined effect of the temperature decrease toward the pole, the water vapor available in the polar region and transport times due to the meridional wind. The distribution of effective radii shows a bi-modal structure at latitudes >75°. This structure is tentatively suggested to arise from a dual-cycle growth process.

[22] In the future we intend to extend our study to several other seasons for which OSIRIS measurements are available as well as to an investigation of potential differences between southern and northern latitudes.

Acknowledgments

[23] We thank G. Witt, G. Thomas, and G. Baumgarten for sharing their expertise on the NLC subject, and J. Stegman for helpful comments. A special acknowledge to J. Gumbel for contributing with important ideas to this work. This work was supported by the Canadian Space Agency and the Natural Sciences and Engineering Research Council (Canada). Odin is a Swedish-led satellite project funded jointly by Sweden (SNSB), Canada (CSA), France (CNES) and Finland (Tekes).