Letter | Published:

The importance of feldspar for ice nucleation by mineral dust in mixed-phase clouds

Nature volume 498, pages 355358 (20 June 2013) | Download Citation

  • An Erratum to this article was published on 10 July 2013

Abstract

The amount of ice present in mixed-phase clouds, which contain both supercooled liquid water droplets and ice particles, affects cloud extent, lifetime, particle size and radiative properties1,2. The freezing of cloud droplets can be catalysed by the presence of aerosol particles known as ice nuclei2. One of the most important ice nuclei is thought to be mineral dust aerosol from arid regions2,3. It is generally assumed that clay minerals, which contribute approximately two-thirds of the dust mass, dominate ice nucleation by mineral dust, and many experimental studies have therefore focused on these materials1,2,4,5,6. Here we use an established droplet-freezing technique4,7 to show that feldspar minerals dominate ice nucleation by mineral dusts under mixed-phase cloud conditions, despite feldspar being a minor component of dust emitted from arid regions. We also find that clay minerals are relatively unimportant ice nuclei. Our results from a global aerosol model study suggest that feldspar ice nuclei are globally distributed and that feldspar particles may account for a large proportion of the ice nuclei in Earth’s atmosphere that contribute to freezing at temperatures below about −15 °C.

Main

Pure cloud droplets remain liquid until cooled to the homogeneous freezing threshold at around 237 K (ref. 2). At warmer temperatures, freezing can be catalysed by heterogeneous ice nuclei. Heterogeneous nucleation can occur in several ways depending upon temperature and humidity1,2. Field observations and modelling studies of mixed-phase cloud formation have shown that supercooled liquid water is usually required before significant in-cloud ice formation occurs2,8. Therefore, in this study we focus on heterogeneous ice nucleation by mineral dust particles immersed in supercooled water droplets.

Atmospheric mineral dusts are inorganic particles of rock and soil that have been lifted into the atmosphere, predominantly from arid regions such as the Sahara9. Mineral dusts from these regions are considered an important source of ice nuclei in mixed-phase clouds, owing to their nucleation efficiency7,10,11 and abundance in the atmosphere9. The importance of mineral dusts as ice nuclei is also supported by their number concentrations within atmospheric ice crystals, which are enhanced relative to the background aerosol12. At present the components in mineral dust responsible for ice nucleation are very poorly constrained. Although atmospheric dust concentrations and mineralogy vary spatially and temporally9,13 (Supplementary Fig. 1), a large fraction of observed atmospheric dust mass around the world is made up of just a few minerals. Individual minerals are classified by their crystal structure and chemistry and can be identified with diffraction techniques2. The mineralogical composition of dust sampled from the atmosphere is shown in Supplementary Table 1. The clay minerals contribute approximately two-thirds of dust mass (kaolinites, 13%; montmorillonites, 2%; chlorites, 3%; micaceous minerals, such as the illites, 44%), with quartz (16%), feldspars (sodium/calcium feldspars (Na/Ca-feldspar), 8%; potassium feldspar (K-feldspar), 3%) and calcite (3%) responsible for much of the remainder.

Previous studies have investigated the ice-nucleating behaviour of dusts sampled from arid source regions or dusts selected as proxies for natural dust1,2. Studies of ice nucleation by individual minerals of varying purity immersed in water have focused on the clay minerals1,2,4,5,6. However, minerals are rarely available in a pure state, and quantification of secondary minerals associated with a particular sample is often neglected. Such characterization is necessary because a minor component may dominate ice nucleation. In this study, we present measurements of ice nucleation by samples of individual minerals in which the impurities were quantified using X-ray diffraction.

To determine the ice-nucleating behaviour of each mineral, we used an established droplet freezing technique4,7. Hundreds of micrometre-sized droplets containing a known amount of solid material were cooled at 1 K min−1, and freezing was monitored by optical microscopy. The freezing temperatures of individual 14–16-μm-diameter droplets containing a range of different minerals are shown in Fig. 1a. In experiments with similar dust surface areas, the temperature at which 50% of droplets were frozen was 250.5 K for K-feldspar, 247 K for Na/Ca-feldspar, 242.5 K for quartz and less than 237.5 K for the clay minerals and calcite. These results suggest that it is the minerals of the feldspar group, in particular K-feldspar, that make mineral dust particles effective immersion-mode ice nuclei in the atmosphere. This contrasts with the prevailing view1,2, which is that clay minerals are the most important component of atmospheric mineral dust for ice nucleation.

Figure 1: Experimental freezing results for the individual minerals.
Figure 1

a, Fraction of droplets, 14–16 μm in diameter and containing a range of mineral dusts, frozen as a function of temperature during cooling. An experiment in which droplets contained no solid inclusion and froze homogeneously is shown for comparison. Temperature uncertainty (not shown) is estimated at ±0.2 K. The indicated uncertainty in the measured fraction frozen is the root mean squared error (68% confidence limit) determined from the pairwise differences between the data and the fraction frozen calculated from the log-linear best fits to the ns values. b, Nucleation site densities for droplets between 9 and 19 μm in diameter collected into four 2.5-μm-wide bins. The uncertainty in ns is primarily due to droplet size measurements, and temperature uncertainty is as in a. The kaolinite parameterization is from ref. 4.

Droplet freezing temperatures are dependent on experimental parameters such as droplet volume and mineral surface area, and are therefore of limited value2. To normalize the efficiency with which a material nucleates ice, we determine the number of nucleation sites per unit surface area2,11,14, ns (Fig. 1b and Supplementary Information). This method of quantifying ice-nucleation efficiency neglects the role of time dependence in nucleation on the basis that the particle-to-particle variability of ice nuclei is more important than the time dependence of nucleation2,11,14,15. Our derived ns values for 9–19-μm-diameter droplets are shown in Fig. 1b. These data show that the feldspar minerals, in particular K-feldspar, are the most efficient mineral dust ice nuclei per unit surface area.

In airborne dusts, the abundance of clay minerals tends to be greater than the abundance of the feldspars, and it is therefore not immediately clear which minerals dominate ice nucleation in the atmosphere. The ns values presented in Fig. 1b were combined with the average mineralogical composition of atmospheric dust to estimate the temperature-dependent ice nuclei concentration (Fig. 2). We have assumed that all particles are spherical to estimate their surface area and have made two limiting calculations, one assuming that dust particles are internally mixed (that is, each particle contains all eight minerals) and the other assuming they are externally mixed (each particle is composed of an individual mineral). The mixing state of atmospheric dust is poorly constrained but falls between these two limiting cases16. Despite accounting for only 3% of atmospheric dust by mass, K-feldspar dominates the number of ice nuclei above 248 K in both the internally mixed and externally mixed cases. One potential caveat to this conclusion is that clay mineral particles may have a smaller particle size than feldspar or quartz13, and therefore may have a greater surface area per unit mass, which would increase the concentration of clay ice nuclei. However, even if the surface area of the clays were 100 times higher (probably an overestimate7), the feldspars would remain the dominant ice-nucleating minerals (Supplementary Fig. 4).

Figure 2: Concentration of ice nuclei due to various minerals, for externally and internally mixed cases.
Figure 2

Ice nuclei concentrations were estimated using the abundance of various minerals from Supplementary Table 1, taking the annually and globally averaged dust concentration (1.4 cm−3) and size distribution from the GLOMAP model together with the ns values in Fig. 1b (the K-feldspar line uses the ns parameterization in Fig. 3). Lines with specific mineral names refer to individual minerals in an externally mixed case. Thick lines denote the range of experimental data and thin lines denote extrapolations outside this range.

Because clouds glaciate over a wider range of temperatures than is achievable in the experiments presented above, it is important to quantify the nucleating efficiency of atmospheric ice nuclei over a broader temperature range. For example, an ice nuclei concentration of as few as 10−5 cm−3 in a cloud at 266 K may be able to trigger substantial glaciation through the Hallett–Mossop ice multiplication process17, and so measurements are needed at these temperatures. To extend the data for K-feldspar to this temperature regime, we performed a series of experiments with larger droplets, allowing for much larger particle surface areas per droplet and correspondingly lower values of ns to be quantified. Results for K-feldspar using droplets 1 μl in volume are shown in Fig. 3. These data extend the range of experimental data up to 268 K. Combining the two techniques allowed us to determine ns values over a range of eight orders of magnitude. The data show that the ice-nucleation efficiency of feldspar is measureable at temperatures relevant for the Hallett–Mossop process (about 265–270 K).

Figure 3: Nucleation site density for K-feldspar and natural dusts.
Figure 3

Data from Fig. 1b (picolitre experiments) extended to higher temperatures by use of microlitre-sized droplets, with a fit provided (ln(ns) = −1.038T + 275.26, valid between 248 and 268 K). Experimental K-feldspar concentrations in weight percent are provided in the key. Temperature uncertainties for microlitre experiments (not shown) were estimated at ±0.4 K and uncertainty in ns (not shown) is estimated at ±25%. We also compare with data for several natural dust samples from ref. 10 (N12) and ref. 11 (C09). The mineralogical compositions of the dusts used in those references are unknown, but feldspar mass content in natural soils typically varies between 1% and 25% (see text). Hence, we have scaled our ns values assuming K-feldspar is present at between 1% and 25% of the natural dust particles’ surface.

In Fig. 3, we compare our K-feldspar results with cloud-chamber-derived ns values for a range of natural mineral dusts sampled from arid source regions10. Because feldspar is a major component of Earth’s crust2, it is ubiquitous in soils around the globe. K-feldspar makes up as much as 24% by mass of soils throughout the Asian and African dust belt18 and is also present in airborne dust in concentrations ranging from a few per cent13 to 25% (ref. 19). We have estimated ns values for natural dust in Fig. 3 by assuming that between 1% and 25% of the surface area of the dusts tested in ref. 10 was feldspar. In the temperature range where the data sets overlap, the agreement is very good.

Results presented here may explain discrepancies in existing experimental data for ice nucleation by mineral dusts. For example, a kaolinite sample from the Clay Mineral Society4 (CMS) had much lower ns values than a sample obtained from Sigma-Aldrich2,5. In this study, we have determined the mineralogical composition of these dusts and shown that the Sigma-Aldrich kaolinite contained 4.5% K-feldspar whereas the CMS kaolinite contained no detectable feldspar (Supplementary Table 2). It is likely that many individual particles of Sigma-Aldrich kaolinite contained feldspar, as is the case in atmospheric dust particles16. We have also quantified the mineralogy of the four montmorillonite samples used in a previous study of ice nucleation6 (see Supplementary Table 2 for sample mineralogy). Three of the samples (M SWy-2, M KSF and M K-10) contain K-feldspar and had higher freezing temperatures than the M STx-1b sample, which did not contain measurable quantities of feldspar. Arizona test dust had the highest freezing temperature of any mixed-mineral dust proxy6 and also contained the most K-feldspar (20 wt%). In general, the more feldspar a sample contains, the higher the freezing temperature. We propose that the feldspar component controlled the nucleation of ice in these experiments, highlighting the need to characterize sample mineralogy in such work.

The mineralogical composition of soils in arid regions around the world varies substantially. Therefore, to quantify the global contribution of feldspars to ice nuclei concentrations it is necessary to use a global aerosol model. Global maps of dust number concentration, feldspar mass fraction and ice nuclei concentration from an aerosol and chemical transport model, GLOMAP, are shown in Fig. 4a–d. To calculate ice nuclei concentration, the mineral-resolved size distribution of dust was combined with our parameterization for the ice-nucleation efficiency of feldspar at 253 K (ice nuclei concentrations at other temperatures are shown in Supplementary Fig. 7). In addition, we also make the assumption that the minerals in mineral dust aerosol are externally mixed; this assumption produces a better match to the observational ice nuclei data at lower temperatures than does the opposing internally mixed assumption (Supplementary Fig. 8), although in reality the mixing state of atmospheric dust will lie somewhere between the internally mixed and externally mixed states.

Figure 4: Dust aerosol modelling study results.
Figure 4

a, Modelled dust number concentrations. b, Total feldspar mass fraction of dust. c, Ice nuclei concentration due to K-feldspar at 253 K, calculated using our ns parameterization and modelled particle surface areas. On the basis of observations in Supplementary Table 1, we assume 35% of feldspar mass is K-feldspar. d, Latitudinal zonal mean values of ice nuclei from c. e, Comparison of c versus ice nuclei concentrations calculated using a mineralogy-independent parameterization based on desert dust samples at 253 K (ref. 10). f, Comparison of model ice nuclei concentrations from K-feldspar mineral dust with field measurements of total ice nuclei. Annual mean modelled ice nuclei concentrations are taken at the same pressure level as the field observation, with the observation temperature used to calculate ns. Only observations between 248 and 258.15 K are shown; a comparison at higher temperatures is shown in Supplementary Fig. 9. Vertical error bars represent the maximum and minimum modelled monthly mean values. See Supplementary Table 4 for field campaign details. Parts ac and e use concentrations at an altitude where the pressure is 600 hPa.

The model data show larger ice nuclei concentrations close to the major dust sources in North Africa and Asia, which results in Northern Hemisphere dust ice nuclei concentrations around one to two orders of magnitude larger than in the Southern Hemisphere (comparing similar latitudes; Fig. 4c, d). Stratiform clouds in the Southern Hemisphere typically glaciate at lower temperatures20,21,22, consistent with a lower concentration of ice nuclei than in the Northern Hemisphere.

To investigate the importance of dust mineralogy for modelling ice nuclei concentrations, we compared ice nuclei concentration in Fig. 4c with that predicted by a parameterization for natural dusts sampled from arid source regions10. Figure 4e shows that the two parameterizations are in agreement close to dust sources. However, the natural dust parameterization predicts ice nuclei concentrations up to 70% higher in regions remote from sources. This higher prediction arises because feldspar is more common in the larger-particle-size fractions (>2 μm) and therefore sediments out more rapidly than the minerals in the small-size fractions13 (Fig. 4b). Hence, atmospheric mineral dust becomes less efficient at nucleating ice during transport through a non-chemical ageing process. Our results may also help to explain the chemical ageing process of dust ice nuclei, which is known to reduce the ice-nucleating efficiency of dust23. Feldspars are susceptible to emissions of acid gases such as SO2, which can convert the surface of feldspar grains to clay minerals24. This may block ice-nucleation sites and reduce the efficiency of feldspar as an ice nucleus, which provides an explanation for the observed sensitivity of mineral dust ice nuclei to acid processing23.

We also compare GLOMAP mineral dust ice nuclei concentrations with field measurements of ice nuclei concentrations (where the aerosol-processing temperature was ≤258 K) from around the world in Fig. 4f. The data are scattered around the 1:1 line, indicating that feldspar is one of the most important ice nuclei in Earth’s atmosphere. The model tends to over-predict ice nuclei concentrations at temperatures below 249 K. It is important to note that we do not include nucleation scavenging, where dust is removed when it serves as ice nuclei or cloud condensation nuclei, and we may therefore over-predict dust concentrations in regions remote from the source. Also, the cluster of data at 258 K, from a ship-borne study around southern Australasia (Supplementary Table 4), is consistently below the 1:1 line. This may indicate that in addition to mineral dust, other ice nuclei sources were also important in this region. At temperatures higher than 258 K, feldspar mineral dust is much less important as ice nuclei and cannot account for the observed ice nuclei concentrations (Supplementary Fig. 9). At these warmer temperatures, other types of ice nuclei, possibly of biogenic origin25, may become increasingly important.

Finally, recent work suggests that human activity has led to a substantial increase in atmospheric dust concentrations and that the sources of this dust have changed9,26. Because potential dust sources around the world have very different feldspar contents18, changes in the location of dust sources may have consequences for the concentration of ice nuclei in the atmosphere and the associated aerosol indirect effect.

Methods Summary

Experimental method

Picolitre experiments (Fig. 1) were performed using a freezing assay of micrometre-sized droplets in a manner similar to previously described4,7. The main variation between this and the previous method is the development of a new cold stage configuration to improve the thermal stability and control of the system, with improvements to temperature measurements reducing uncertainty to ±0.2 K. The microlitre experiments were performed with a separate system in which cooling, temperature measurement and control were provided by a Stirling engine-powered flat-plate chiller (Grant-Asymptote EF600). Droplets were then deposited onto a large, siliconized glass slide using a pipette, and freezing was monitored using a digital camera without magnification. The analysis of the data from the picolitre and microlitre experiments was identical. Mineral dust characterization methods have been described previously7.

Modelling with GLOMAP

GLOMAP is a size- and composition-resolving two-moment microphysical aerosol scheme27, run within the TOMCAT chemical transport model28. GLOMAP has previously been used to study atmospheric processing of mineral dust29. The model is driven by reanalysis meteorology for the year 2000. GLOMAP was extended to represent eight mineral types, as specified by the surface mineralogy map of ref. 18. Dust is represented in 12 size bins, ranging in diameter from 0.1 to >20.0 μm. Dust emissions are prescribed from AEROCOM recommendations for the year 200030. The experimentally derived ns values were combined with modelled atmospheric particle sizes, composition and concentrations (annual global averages at the altitude corresponding to a pressure of 600 hPa) to estimate ice nuclei concentrations (Fig. 2). Two particle mixing states were considered. In the internally mixed case, all particles contain the considered minerals in the same ratios, whereas in the externally mixed case individual particles are composed of single minerals with the overall population composition controlled by the dust mineralogy.

References

  1. 1.

    & Heterogeneous ice nucleation on atmospheric aerosols: a review of results from laboratory experiments. Atmos. Chem. Phys. 12, 9817–9854 (2012)

  2. 2.

    , , & Ice nucleation by particles immersed in supercooled cloud droplets. Chem. Soc. Rev. 41, 6519–6554 (2012)

  3. 3.

    et al. African dust aerosols as atmospheric ice nuclei. Geophys. Res. Lett. 30, 1732 (2003)

  4. 4.

    , , , & Heterogeneous freezing of water droplets containing kaolinite particles. Atmos. Chem. Phys. 11, 4191–4207 (2011)

  5. 5.

    , , & Experimental study on the ice nucleation ability of size-selected kaolinite particles in the immersion mode. J. Geophys. Res. 115, D14201 (2010)

  6. 6.

    , , , & Ice nucleation efficiency of clay minerals in the immersion mode. Atmos. Chem. Phys. 12, 5859–5878 (2012)

  7. 7.

    et al. Immersion mode heterogeneous ice nucleation by an illite rich powder representative of atmospheric mineral dust. Atmos. Chem. Phys. 12, 287–307 (2012)

  8. 8.

    , , & Evidence of liquid dependent ice nucleation in high-latitude stratiform clouds from surface remote sensors. Geophys. Res. Lett. 38, L01803 (2011)

  9. 9.

    , , , & Global-scale attribution of anthropogenic and natural dust sources and their emission rates based on MODIS Deep Blue aerosol products. Rev. Geophys. 50, RG3005 (2012)

  10. 10.

    et al. A particle-surface-area-based parameterization of immersion freezing on desert dust particles. J. Atmos. Sci. 69, 3077–3092 (2012)

  11. 11.

    et al. Studies of heterogeneous freezing by three different desert dust samples. Atmos. Chem. Phys. 9, 2805–2824 (2009)

  12. 12.

    et al. In situ detection of biological particles in cloud ice-crystals. Nat. Geosci. 2, 398–401 (2009)

  13. 13.

    & Saharan aerosols over the tropical north-Atlantic: mineralogy. Mar. Geol. 37, 295–321 (1980)

  14. 14.

    Quantitative evaluation of experimental results an the heterogeneous freezing nucleation of supercooled liquids. J. Atmos. Sci. 28, 402–409 (1971)

  15. 15.

    Quantitative descriptions of ice formation mechanisms of silver iodide-type aerosols. Atmos. Res. 38, 63–99 (1995)

  16. 16.

    Bulk and single-particle mineralogy of Asian dust and a comparison with its source soils. J. Geophys. Res. 113, D02208 (2008)

  17. 17.

    et al. Ice formation and development in aged, wintertime cumulus over the UK: observations and modelling. Atmos. Chem. Phys. 12, 4963–4985 (2012)

  18. 18.

    , , , & High-resolution mineralogical database of dust-productive soils for atmospheric dust modeling. Atmos. Chem. Phys. 12, 845–855 (2012)

  19. 19.

    et al. Ground-based off-line aerosol measurements at Praia, Cape Verde, during the Saharan Mineral Dust Experiment: microphysical properties and mineralogy. Tellus 63B, 459–474 (2011)

  20. 20.

    , , & Space observations of cold-cloud phase change. Proc. Natl Acad. Sci. USA 107, 11211–11216 (2010)

  21. 21.

    et al. Contrasting the impact of aerosols at northern and southern midlatitudes on heterogeneous ice formation. Geophys. Res. Lett. 38, L17802 (2011)

  22. 22.

    , , & A classical-theory-based parameterization of heterogeneous ice nucleation by mineral dust, soot, and biological particles in a global climate model. J. Atmos. Sci. 67, 2483–2503 (2010)

  23. 23.

    et al. Irreversible loss of ice nucleation active sites in mineral dust particles caused by sulphuric acid condensation. Atmos. Chem. Phys. 10, 11471–11487 (2010)

  24. 24.

    , , & Naturally weathered feldspar surfaces in the Navajo Sandstone aquifer, Black Mesa, Arizona: Electron microscopic characterization. Geochim. Cosmochim. Acta 70, 4600–4616 (2006)

  25. 25.

    , , & Ice nuclei in marine air: biogenic particles or dust? Atmos. Chem. Phys. 13, 245–267 (2013)

  26. 26.

    et al. Observed 20th century desert dust variability: impact on climate and biogeochemistry. Atmos. Chem. Phys. 10, 10875–10893 (2010)

  27. 27.

    , , , & A global off-line model of size-resolved aerosol microphysics: I. Model development and prediction of aerosol properties. Atmos. Chem. Phys. 5, 2227–2252 (2005)

  28. 28.

    , & A three-dimensional model study of the effect of new temperature-dependent quantum yields for acetone photolysis. J. Geophys. Res. 110, D22305 (2005)

  29. 29.

    et al. Minor effect of physical size sorting on iron solubility of transported mineral dust. Atmos. Chem. Phys. 11, 8459–8469 (2011)

  30. 30.

    et al. Emissions of primary aerosol and precursor gases in the years 2000 and 1750 prescribed data-sets for AeroCom. Atmos. Chem. Phys. 6, 4321–4344 (2006)

Download references

Acknowledgements

We thank J. Cuadros, L. Neve and R. Finch for help sourcing mineral samples, P. DeMott for providing ice nuclei observational data, and T. W. Wilson for discussions. We acknowledge the European Research Council (FP7, 240449 ICE) and the Natural Environment Research Council (NE/I013466/1, NE/I020059/1, NE/I019057/1) for funding. K.S.C. is a Royal Society Wolfson Research Merit Award holder.

Author information

Author notes

    • Matthew T. Woodhouse

    Present address: Commonwealth Scientific and Industrial Research Organisation (CSIRO) Marine and Atmospheric Research, PMB 1, Aspendale, Victoria 3195, Australia

Affiliations

  1. Institute for Climate and Atmospheric Science, School of Earth and Environment, University of Leeds, Leeds LS2 9JT, UK

    • James D. Atkinson
    • , Benjamin J. Murray
    • , Matthew T. Woodhouse
    • , Thomas F. Whale
    • , Kelly J. Baustian
    • , Kenneth S. Carslaw
    • , Steven Dobbie
    • , Daniel O’Sullivan
    •  & Tamsin L. Malkin

Authors

  1. Search for James D. Atkinson in:

  2. Search for Benjamin J. Murray in:

  3. Search for Matthew T. Woodhouse in:

  4. Search for Thomas F. Whale in:

  5. Search for Kelly J. Baustian in:

  6. Search for Kenneth S. Carslaw in:

  7. Search for Steven Dobbie in:

  8. Search for Daniel O’Sullivan in:

  9. Search for Tamsin L. Malkin in:

Contributions

J.D.A. conducted the picolitre experiments, analysed the data and wrote the paper, and T.F.W. performed and analysed the microlitre experiments. K.J.B. and D.O. contributed to the experimental study, and S.D. helped draft the manuscript. M.T.W. led the global modelling study in collaboration with K.S.C. T.L.M. did the X-ray analysis of the mineral samples. B.J.M. oversaw the project and helped to write the manuscript.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Benjamin J. Murray.

Supplementary information

PDF files

  1. 1.

    Supplementary Information

    This file contains Supplementary Figures 1-9, Supplementary Methods, Supplementary Tables 1-4, Supplementary Discussion and Supplementary References. Section 1 contains nine experimental and modelling figures. Section 2 expands on the experimental and modelling methodologies. Section 3 lists the mineralogy observational data used, the analysis of mineral samples and the field IN data. Section 4 is a discussion of the model outputs and their comparison with observations.

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/nature12278

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.