UV limb-scatter spectra of noctilucent clouds consistent with mono-modal particle size distribution



[1] Limb-backscatter spectra of noctilucent clouds (NLCs) are retrieved from limb measurements made with the Scanning Imaging Absorption spectroMeter for Atmospheric CartograpHY (SCIAMACHY) on the Envisat satellite. The spectral dependence of many hundred sun-normalized NLC spectra measured during the 2005 NLC season in the northern hemisphere can be well approximated with a power-law in the 240–300 nm spectral range. In particular, no spectral signatures indicative of the presence of a second particle mode at radii exceeding 200 nm are found, in contradiction to recent studies. The results presented are consistent with a mono-modal NLC particle size distribution.

1. Introduction

[2] Noctilucent clouds (NLCs) or polar mesospheric clouds (PMCs) are optically thin H2O ice clouds occurring at about 82–85 km altitude near the polar summer mesopause, where temperatures frequently drop below 150 K. Although many NLC properties are well known, the actual size distribution of NLC particles is still not fully established. For remote sensing measurements of NLC sizes using optical spectroscopy or photometry a NLC particle size distribution has to be assumed. Commonly used size distributions are mono-modal log-normal distributions [e.g., Debrestian et al., 1997; von Cossart et al., 1999; Carbary et al., 2002; von Savigny et al., 2005] and δ-functions (i.e., a monodisperse model) [e.g., Gumbel and Witt, 1998]. The monodisperse particle size distribution is an obviously inadequate model, but is used because of its simplicity. A log-normal size distribution generally provides a good representation of the actual size distribution if the particle growth is dominated by coagulation. However, this is not the case for the crystalline NLC particles that are believed to grow by heterogeneous nucleation and subsequent water vapor deposition on the particle surface [e.g., Rapp and Thomas, 2006].

[3] Multi-color LIDAR measurements allow the determination of both free parameters r0 and σ of the log-normal size distribution, assuming that the log-normal size distribution is an appropriate model: von Cossart et al. [1999] presented 11 measurements at Andoya/Norway (69°N) and found mean values of r0 = 51 ± 21 nm and σ = 1.42 ± 0.22.

[4] Model simulations can also be used to calculate the NLC size distribution. Berger and von Zahn [2002] used the COMMA/IAP model to show that the determined size distribution is better approximated by a mono-modal normal distribution than by a log-normal distribution. More recently, Rapp and Thomas [2006] showed that simulations with the CARMA model also yield size distributions that can be well approximated with a mono-modal normal distribution. Furthermore, Rapp et al. [2007] demonstrated that such a distribution is consistent with spectrally resolved measurements by LIDAR, and satellite instruments like SME and SNOE when non-spherical ice particles are assumed.

[5] The basic assumption, that the NLC particle size distribution is generally mono-modal was recently questioned by Carbary et al. [2004], who presented sun-normalized NLC limb radiance spectra measured with the UVISI instrument on the MSX satellite that exhibited a characteristic local maximum (“hump”) at a wavelength of about 260 nm. Carbary et al. [2004] argued, that this local maximum is evidence for the existence of a second particle mode at radii exceeding 200 nm. This is in contrast to the general idea, that NLC particle radii are smaller than about 80–100 nm, supported both by previous observations [e.g., Debrestian et al., 1997; Gumbel and Witt, 1998; Carbary et al., 2002; von Cossart et al., 1999; von Savigny et al., 2004, 2005] and model simulations [e.g., Berger and von Zahn, 2002]. Carbary et al. [2004] suggested, that unknown spectral signatures in the complex refractive index of ice may be another possible reason for the spectral “hump”. In this context DeLand et al. [2005] used SBUV/2 nadir-backscatter measurements between 250 and 300 nm to check the hypothesis of a “hump” around 260 nm, but failed to identify any signature around this wavelength.

[6] In this publication we use SCIAMACHY limb scatter measurements – much more sensitive to NLCs than nadir observations – to check the presence of a spectral signature in the NLC back-scatter spectra at a wavelength of about 260 nm. Although several recent studies show that NLC particles are most likely non-spherical [Baumgarten et al., 2002; Eremenko et al., 2005; Rapp et al., 2007] for the present study Mie-simulations are used and compared to measured NLC limb spectra. This appears justified since Rapp et al. [2007] recently found that measurements obtained in forward scattering geometry are significantly less influenced by particle shape effects than measurements for scattering angles > 90°.

2. Brief Instrumental Description

[7] SCIAMACHY, the SCanning Imaging Absorption spectroMeter for Atmospheric CartograpHY [Bovensmann et al., 1999], is one of ten scientific instruments aboard the European Space Agency's Envisat spacecraft. Envisat was launched on March 1, 2002 from Kourou (French Guiana) into a polar, sun-synchronous orbit with a 10:00 LST (local solar time) descending node. SCIAMACHY measures solar radiation scattered by and transmitted through the atmosphere in nadir, solar/lunar occultation and limb mode. For this study only limb-scatter observations – fully calibrated Level 1 data with Level-0-to-1 processor version 5.04 [Brooker et al., 2002] – are employed. Limb-scatter observations are particularly suited to study optically thin phenomena such as NLCs due to the long slant paths through the atmosphere. In limb observation mode SCIAMACHY scans the Earth's limb from the surface up to about 92 km in steps of 3.3 km for the data used here. The geometrical field of view (FOV) is 2.6 km in the vertical direction and 110 km in the horizontal direction. Azimuthal scanning at every tangent height leads to an effective spatial smearing over a distance of 960 km perpendicular to the viewing direction. In viewing direction the spatial smearing corresponds to about 500 km.

[8] Apart from limb-scatter observations SCIAMACHY also provides measurements of the solar irradiance spectrum [Skupin et al., 2005]. These solar irradiance measurements are used in this study. For the spectral range considered here the SCIAMACHY solar irradiance measurements agree to within a few percent with the standard Kurucz solar spectrum and with solar spectra measured with the SOLSPEC, SOLSTICE and SUSIM satellite instruments (for details we refer to Skupin et al. [2005]).

3. Sun-Normalized NLC Limb-Scatter Spectra

[9] In order to test whether the “hump” in the sun-normalized NLC limb-scatter spectra is also present in the SCIAMACHY limb observations we use SCIAMACHY channel 1 data in the spectral range between 240 nm and 300 nm. The use of wavelengths smaller than 300 nm has the advantage that the multiple scattering contribution to the measured limb signal is negligible [Witt et al., 1976] – a consequence of the massive absorption of solar photons in the Hartley-bands of O3.

[10] In single-scattering approximation the Rayleigh-corrected NLC backscatter spectrum INLC is directly proportional to the solar irradiance spectrum I0(λ), the NLC scattering cross-section σ(λ), and the NLC scattering phase function P(Θ), where Θ is the scattering angle:

equation image

[11] The ratio INLC(λ)/I0(λ) is then proportional to the differential scattering cross section, i.e., the product of scattering cross section and phase function. The differential scattering cross section is – for limited spectral intervals – often well approximated by a power law, i.e.: σ(λ) × P(Θ) ∝ λα with the spectral or Ångstrøm exponent α.

[12] Figure 1a shows a sample background-corrected SCIAMACHY NLC limb-scatter spectrum together with the solar irradiance spectrum. The sun-normalized NLC spectrum is shown in Figure 1b. No obvious local maximum can be seen around 260 nm. However, this is only an individual example. The measured NLC backscatter spectra also exhibit signatures of several different atmospheric emissions, most prominently the NO-γ bands, but also, e.g., the weaker O(1S–3P) emission at 297.1 nm, and some Mg and Fe emissions. For really bright NLCs – as for the measurement shown in Figure 1 – these emissions are too weak to significantly affect the NLC particle size estimates. In order to eliminate any potential impact we only use emission-free spectral sub-windows for the determination of the NLC spectral exponents. Furthermore, since small wavelength misalignments lead to strong oscillations in the sun-normalized spectra if the solar spectrum exhibits strong Fraunhofer structures, e.g., for the Mg line at 280 nm, we also omitted these spectral sub-windows for the spectral binning and the subsequent determination of the Ångstrøm exponents.

Figure 1.

(a) Sample background-corrected NLC backscatter spectrum (left ordinate), and the Skupin et al. [2005] solar irradiance spectrum measured with SCIAMACHY (right ordinate). (b) Sun-normalized NLC backscatter spectrum derived from the spectra in the Figure 1a. The shaded spectral windows contain terrestrial airglow emissions or strong Fraunhofer lines and are not considered for the determination of the Ångstrøm exponents.

[13] Figure 2 shows several examples of sun-normalized NLC spectra for an arbitrarily chosen orbit on July 12, 2005. The spectra are binned in 5 nm intervals similar to Figure 3 in Carbary et al. [2004]. The spectral feature present in the MSX limb spectra has a width of about 20 nm and a relative magnitude of 25–30% [Carbary et al., 2004]. However, there are no indications for a spectral feature of this magnitude at or around 260 nm in the SCIAMACHY sun-normalized spectra shown in Figure 2. There are small spectral signatures below 260 nm – particularly for the two weakest NLCs shown – which are most likely a combination of poorer signal-to-noise ratios and larger errors introduced by the Rayleigh-background correction for weak NLCs. We investigated more than 1800 sun-normalized SCIAMACHY NLC spectra measured during the 2005 NLC season in the northern hemisphere, and no indications for any “hump”-like spectral feature with the size present in the MSX spectra around 260 nm were found. Figure 3 shows mean sun-normalized NLC spectra for different relative brightness intervals (Relative brightness refers the ratio of the maximum Rayleigh-corrected NLC limb radiance and the Rayleigh background). There are again deviations between the power-law fit and the measured sun-normalized spectra below about 255 nm, but the relative differences never exceed 3%, which is well within the calibration uncertainties of both the SCIAMACHY solar irradiance and limb radiance spectrum measurements.

Figure 2.

A set of sun-normalized NLC limb-scatter spectra derived from measurements during Envisat orbit 17593 on July 12, 2005, and binned into 5 nm wavelength bins. Also shown are least-squares fits of power-law functions and the derived Ångstrøm exponents.

Figure 3.

Mean sun-normalized NLC backscatter spectra for different relative brightness intervals averaged over NLCs detected in the northern hemisphere in July 2005. Apparently no indications for systematic deviations from a power-law dependence are present. Also given are the derived spectral exponents for the 240 to 300 nm spectral range and the number of measurements N.

[14] Inspecting Figure 3, we cannot confirm the Carbary et al. [2004] finding that a characteristic spectral structure at a wavelength of about 260 nm is present in the majority of sun-normalized NLC spectra. This is in-line with the results published by DeLand et al. [2005], who also failed to identify an unusual spectral structure in NLC-nadir spectra measured with SBUV. The range of derived Ångstrøm exponents shown in Figure 3 corresponds to NLC radii of about 30–50 nm, if a log-normal particle size distribution with σ = 1.4 is assumed. For a Gaussian particle size distribution with σ = 12 nm the derived radii are about 40–90 nm [von Savigny and Burrows, 2007].

4. Discussion

[15] It has to be mentioned that here we only show SCIAMACHY NLC spectra measured in the northern hemisphere, where the scattering angles are between about 35° and 65°. The scattering angles of the UVISI/MSX measurements that showed the hump at around 260 nm, were between 125° and 145° [Carbary et al., 2004]. In principle SCIAMACHY also performs limb measurements in the southern hemisphere where scattering angles range from about 140°–160° at polar latitudes. Due to the large scattering angles the NLC signatures are much weaker than in the northern hemisphere. Because of the low signal-to-noise ratios we found that the SCIAMACHY spectra measured in the southern hemisphere were not usable in the entire 240 nm–300 nm spectral range. However, the SCIAMACHY limb measurements in the northern hemisphere – with scattering angles between 35° and 65° – are more sensitive to the presence of large particles, because of the forward scattering peak of the phase function. Figure 4 (left) shows synthetic differential Mie-scattering efficiencies for a bi-modal particle size distribution with discrete modes at r1 = 50 nm and r2 = 220 nm, a relative weighting of the modes as given by Carbary et al. [2004], and for scattering angles Θ of 60° and 130°. For Θ = 60° the spectra are obviously entirely dominated by the mode at r2, and cannot be approximated by a power-law at all. Therefore, a second mode of the particle size distribution at 220 nm should be clearly visible in the SCIAMACHY NLC spectra.

Figure 4.

(left) Differential scattering efficiency for a bi-modal particle size distribution with two discrete modes for 60° and 130° scattering angle. (right) The sensitivity to the relative weighing of the two modes.

[16] In order to test how robust the spectral features originating from the mode of large particles are with respect to the radii and relative weighting of the two modes of a hypothetical bi-modal particle size distribution, we determined differential Mie-scattering efficiencies. The results are shown in Figures 4 and 5. Figure 5 (left) clearly shows that the wavelength, where the spectral peak occurs, depends strongly on the radius of the large particle mode. Furthermore, Figure 5 (right) indicates that the location of the spectral maximum also depends sensitively on the scattering angle. Considering the high sensitivity of the differential scattering efficiency on the relative weighting of the two modes, on the scattering angle and also on the radii it seems extremely unlikely that measurements of NLC scattering spectra yield similarly shaped spectra – with a peak at a constant wavelength – for a range of different conditions.

Figure 5.

(left) Dependence of differential scattering efficiency on the radius of the large mode and (right) on scattering angle.

[17] Unknown spectral signatures in the imaginary part of the refractive index of ice [Carbary et al., 2004] appear to be an unlikely explanation for the hump in the UVISI/MSX measurements, since they should also be present in the SCIAMACHY spectra.

[18] We have no reason to believe that the SCIAMACHY measurements of either the solar irradiance spectrum or the limb-scattering spectra are affected by calibration errors that would dilute an apparent spectral structure at around 260 nm. Furthermore, even if a spectral signature originating from a second mode of the NLC particle size distribution with large particle radii were generally present, it would be extremely unlikely that SCIAMACHY calibration errors would lead to the observed near-perfect power-law behavior of the sun-normalized backscatter spectra.

[19] We also note that there is a significant difference in the horizontal footprints of SCIAMACHY and UVISI limb measurements. SCIAMACHY averages over distances of 960 km across and about 500 km along the line of sight. In contrast, UVISI averages over 3 km across and 230 km along the line of sight [Carbary et al., 2002]. While the smaller UVISI footprint may make the detection of a single NLC – possibly with a bi-modal particle size distribution – more likely, this cannot explain why the spectral feature at 260 nm is always present in the UVISI spectra [Carbary et al., 2004].

[20] The analysis presented here does not allow the conclusion that a mono-modal normal size distribution is the actual size distribution. The Ångstrøm exponents derived from SCIAMACHY limb measurements in the 265–300 nm spectral range can be modeled not only with a normal particle size distribution, but also with a monodisperse model [von Savigny et al., 2004] as well as with a mono-modal log-normal distribution. Measurements within a limited spectral range do not allow to distinguish different hypothetical particle size distributions. However, we can make the statement that our measurements do not require a bi-modal NLC size distribution with a mode at radii exceeding 200 nm to explain the spectral dependence of the sun-normalized NLC backscatter spectra. The sun-normalized NLC spectra presented here are consistent with mono-modal NLC size distributions with mean radii on the order of 40–90 nm for a normal size distribution with σ = 12 nm or with radii between 30–50 nm for a log-normal distribution with σ = 1.4 (results not shown here) [von Savigny and Burrows, 2007]. This is in line with most of the previously reported NLC size measurements.

5. Conclusions

[21] Sun-normalized NLC limb-spectra measured with SCIAMACHY in the 240–300 nm spectral range do not exhibit characteristic spectral signatures indicative of a bi-modal NLC size distribution with a second mode at radii of about 220 nm apart from the typical 50 nm mode. This result contradicts recent findings by Carbary et al. [2004], but is in line with DeLand et al. [2005]. Synthetic differential NLC scattering efficiencies for a hypothetical bi-modal particle size distribution are shown to depend very sensitively on the relative weighting of the modes, the mode radii, and also on scattering angle. This indicates, that the spectral signature in the UVISI/MSX NLC spectra at 260 nm are most likely not caused by the large mode of a bi-modal NLC particle size distribution.


[22] We thank A. Kokhanovsky and G. Thomas for helpful discussions. This work was supported by the German Ministry of Education and Research (BMBF) and the German Aerospace Center (DLR) under grant EE0027 as well as by the University of Bremen. We are indebted to ESA for providing SCIAMACHY data. SCIAMACHY is jointly funded by Germany, the Netherlands, and Belgium.