Atmospheric ice nuclei in the Eyjafjallaj ökull volcanic ash plume

We have sampled atmospheric ice nuclei (IN) and aerosol in Germany and in Israel during spring 2010. IN were analyzed by the static vapor diffusion chamber FRIDGE, as well as by electron microscopy. During the Eyjafjallaj ökull volcanic eruption of April 2010 we have measured the highest ice nucleus number concentrations ( >600 l−1) in our record of 2 yr of daily IN measurements in central Germany. Even in Israel, located about 5000 km away from Iceland, IN were as high as otherwise only during desert dust storms. The fraction of aerosol activated as ice nuclei at −18C and 119 % rhice and the corresponding area density of ice-active sites per aerosol surface were considerably higher than what we observed during an intense outbreak of Saharan dust over Europe in May 2008. Pure volcanic ash accounts for at least 53–68 % of the 239 individual ice nucleating particles that we collected in aerosol samples from the event and analyzed by electron microscopy. Volcanic ash samples that had been collected close to the eruption site were aerosolized in the laboratory and measured by FRIDGE. Our analysis confirms the relatively poor ice nucleating efficiency (at −18C and 119 % ice-saturation) of such “fresh” volcanic ash, as it had recently been found by other workers. We find that both the fraction of the aerosol that is active as ice nuclei as well as the density of ice-active sites on the aerosol surface are three orders of magnitude larger in the samples collected from ambient air during the volcanic peaks than in the aerosolized samples from the ash collected close to the eruption site. From this we conclude that the ice-nucleating properties of volcanic ash may be altered substantially by aging and processing during long-range transport in the atmosphere, and that global volcanism deserves further attention as a potential source of atmospheric ice nuclei.


Introduction
The recent eruption of the Eyjafjallajökull volcano in Iceland, besides affecting aviation in Europe, raises the question about potential effects of volcanic emissions on weather and climate. The climate impact of explosive volcanic injections into the stratosphere 15 through the radiative effects of secondary H 2 SO 4 /H 2 O aerosol particles formed from the injected SO 2 , is well documented (Robock, 2000). However, only few eruptions reach the stratosphere, and the total emission of 20 Tg yr −1 of fine ash into the troposphere by small volcanic eruptions (Mather et al., 2003) exceeds the time-averaged annual return flux of volcanic sulphate from the stratosphere to the upper troposphere 20 by a factor of 10-40. Volcanic ash particles can affect the phase of supercooled tropospheric clouds by acting as ice nuclei (IN) (Durant et al., 2008). About 50% of the global cloud population at −20 • C is found to be supercooled (Choi et al., 2010), i.e. to contain metastable water that freezes upon the presence of suitable ice nuclei. Glaciation affects the phase, size distribution and colloidal stability of the cloud particles, as well as Introduction that on the planetary scale the fraction of supercooled clouds (at −20 • C) and the coincident dust aerosol frequency are negatively correlated (Choi et al., 2010), likely due to glaciation by dust. Accompanying model calculations reveal that the cloud albedo is significantly affected upon glaciation. Despite their relevance, ice nuclei remain an undersampled component of the climate 5 system. A quantitative assessment of their sources and a climatology of their atmospheric abundance are not available. Many components of atmospheric IN have been identified, such as minerals of desert dust, bacteria, pollen and plant debris (Pruppacher and Klett, 1997;Szyrmer and Zawadsky, 1997) although for many of them the concentrations and relevance to cloud processes are still unclear (Möhler et al., 10 2007;DeMott et al., 2010). In addition to these surface-derived sources, volcanism is debated as a source of atmospheric ice nuclei, with conflicting evidence from field measurements (Isono et al., 1959;Price and Pales, 1963;Hobbs et al., 1971;Schnell and Delany, 1976;Radke et al., 1976). Recent satellite observations (Gasso, 2008) show that natural degassing or weakly explosive volcanoes in the South Atlantic and 15 North Pacific affect low marine stratocumulus for up to 1300 km downwind by decreasing droplet effective radius and increasing visible brightness, and may add cloud cover in otherwise cloudless areas. While the latter observations did not consider explicitly the ice-nucleating ability of the volcanic aerosol, our measurements during the Eyjafjallajökull eruption show a significant enhancement of atmospheric IN when the dispersed 20 ash cloud reached central Europe in April 2010 and the eastern Mediterranean in May 2010.

Methods
The number concentration of atmospheric ice nuclei near the surface is monitored at the Taunus  for atmospheric research located 25 km north of Frankfurt (825 m a.s.l.). The Tel Aviv University sampling site is on the university campus, about 2.5 km from the Mediterranean Sea shore. Aerosols were collected on 47 mm diameter silicon substrates using a specially designed electrostatic precipitator (Klein et al., 2010a). All the samples were collected 5 at a flow rate of 2 liters min −1 with 10 liters and 5 liters sampled at TO and Tel Aviv, respectively. Substrates were analyzed in the isothermal static vapor diffusion chamber FRIDGE (Klein et al., 2010a;Bundke et al., 2008). Ice nuclei concentrations were measured at three different temperatures between −18 • C to −8 • C and RH ICE between 103% and 119%. The ice crystals were observed by a CCD camera and were counted 10 automatically. The regular sampling frequency usually was 1 per day, but was higher during periods of interest.
In those samples that were collected during the volcanically affected days of 16-17 April and 16-17 May the chemical composition and morphology of 150 individual particles that had been previously identified as ice nuclei by FRIDGE were determined by 15 environmental scanning electron microscopy (ESEM) combined with energy-dispersive X-ray microanalysis (EDX).
The unambiguous identification of the analyzed particles as ice nuclei was enabled by a high precision laser engraved coordinate system on the substrates. The positions on the substrates, where ice nucleation was observed by the CCD camera of FRIDGE, 20 can be recovered in the ESEM with a lateral resolution of approximately 5 µm.
The aerosol particle surface area was derived from measurement by an Electrical Aerosol Spectrometer (Tammet et al., 2002) using a particle shape factor of 1.2 and density of 2.6 g cm −3 . In the chemical and morphological analysis of ice nuclei by ESEM four groups of particles were observed: volcanic particles, soot, sea-salt, and biological particles. The volcanic particles were classified on basis of the main and minor elements of these alumosilicates (Mg, Al, Si, K, Ca, Ti and Fe) and/or their glassy morphology. This group can also contain up to 5% (relative abundance) of soil particles, if the monthly average 5 IN concentration in April of previous years is assumed to be present in our volcanic ash samples. The classification criteria for the other groups can be found elsewhere . For the samples collected at TO during the peaks of the volcanic ash event on 17 April and 16-17 May the relative number abundance of the different particle groups among the IN are shown in Table 1 (all samples) and Fig. 3 (mean values for each ash event). Confidence intervals (95%) were calculated assuming a multinomial distribution (for details see Weinbruch et al., 2002). The analysis confirms that the volcanic contribution to these particles is at least 63%. As small inclusions of volcanic material were detected within individual sea salt and soot particles the volcanic ash contribution to IN might be even higher. 15 The signal of volcanic ash in our record of atmospheric IN at TO after the air masses had travelled for at least 2200 km is surprisingly pronounced, in particular when seen in the light of the so far inconclusive and conflicting observational evidence that had been available on volcanic ice nuclei. Isono et al. (1959) found IN at Tokyo, Japan, (at −20 • C) to increase above the 5-20 IN L −1 background to up to 50 IN L −1 in air 20 that had been affected by Japanese volcanoes located 110-1180 km upwind, whereas others found no evidence for volcanic IN in the effluents of Hawaiian volcanoes (Price and Pales, 1963;Hobbs et al., 1971), of St. Augustine, Alaska (Schnell and Delany, 1976), and of Mt. Baker, Washington (Radke et al., 1976). Some of the latter results were debated recently because the situations sampled were more representative for 25 passively degassing volcanoes rather than for large explosive eruptions and because of a postulated underestimation of IN by the technique applied (Durant et al., 2008;Pruppacher and Klett, 1997 A significant volcanic contribution to IN was also recorded a month later at Tel Aviv, more than 5000 km away from the source. Tel Aviv is frequently affected by dust storms, due to the proximity to Arabian and African deserts. During such conditions (defined by a daily average PM 10 concentration higher than 100 µg m −3 ; 37 cases), the concentrations of PM 10 (Fig. 2) suggest the advection of air from the north eastern Atlantic around Iceland towards the eastern Mediterranean. While the PM 10 measurements at TAU during these days indicate 10 "clean" conditions with PM 10 values well below the 100 µg m −3 level (defined as a dust event), the IN peak at around 100 L −1 (see Fig. 3) and the average of 66.9 ± 37.5 IN L −1 (5 samples) is significantly higher than the averages for both normal "clean" days and for some "dusty" days. Furthermore, the elemental composition of individual particles that were randomly selected from the samples collected during the passage of 15 the volcanic plume over Israel was analyzed using ESEM-EDX. The analysis shows that the volcanic particles had similar composition to mineral dust, except that sea salt and some sulphate were also present. All these observations clearly point to a strong contribution of volcanic ash to the total IN population over Israel during this event.

Results and discussion
Since ice nucleation is considered a surface phenomenon, the specific density of 20 IN number per aerosol surface area is of interest and is important for parameterization used in numerical models of clouds and climate. From our measurements of IN number concentration and aerosol particle size distribution in the 0.3-20 µm diameter size range at TO during the volcanic event we derive a mean IN surface area density of 2 × 10 9 IN m −2 of aerosol at −18 • C and water saturation. This is at least a factor of 25 two higher than the IN surface densities that were observed under the same conditions in various tropospheric environments, including air masses affected by mineral dust (Klein et al., 2010c;Phillips et al., 2008). In our samples, the first ice formed at −8 • C and water saturation. This is at the upper end of the range of freezing temperatures reported so far for volcanic aerosol (Durant et al., 2008). Many of the previous data, however, were obtained for the immersion and contact freezing modes, whereas our measurements address the deposition and condensation-freezing modes. At a cloud glaciation temperature as high as −8 • C 5 (equivalent to 3.5 km altitude in the Standard Atmosphere) the Eyjafjallajökull volcanic event may have affected even lower tropospheric clouds by glaciogenic seeding. While mid-level stratus normally is considered to play a neutral role in the Earth's radiation budget because it reflects about as much shortwave solar radiation as it absorbs terrestrial radiation (Cotton, 2009), the quantitative effects of massive cloud seeding on both budget entries remains an unresolved question. As was pointed out above, the fraction of supercooled cloud particles within mixed-phase clouds has recently been found to be negatively correlated to the frequencies of dust aerosols (at the −20 • C isotherm) and the glaciation of such supercooled clouds to decrease cloud albedo (Choi et al., 2010). Considering the high effectiveness of the Eyjafjallajökull particles as IN and 15 the large areal extent of marine stratus and altostratus with possible supercooling in the high northern latitudes (Bretherton and Hartmann, 2009;Hartmann, 1994), a large scale natural cloud seeding event may have taken place, with possible consequences for cloud radiative forcing. It is also probable that high concentrations of these particles at higher levels have increased cirrus cloud cover, resulting in changes of radiation 20 fluxes at these levels.  535-536, 1976. Schumann, U., Weinzierl, B., Reitebuch, O., Schlager, H., Minikin, A., Forster, C., Baumann, R