A long-term study of cloud residuals from low-level Arctic clouds

Abstract. To constrain uncertainties in radiative forcings associated with
aerosol–cloud interactions, improved understanding of Arctic cloud formation
is required, yet long-term measurements of the relevant cloud and aerosol
properties remain sparse. We present the first long-term study of cloud
residuals, i.e. particles that were involved in cloud formation and cloud
processes, in Arctic low-level clouds measured at Zeppelin Observatory,
Svalbard. To continuously sample cloud droplets and ice crystals and separate
them from non-activated aerosol, a ground-based counter-flow virtual impactor
inlet system (GCVI) was used. A detailed evaluation of the GCVI measurements,
using concurrent cloud particle size distributions, meteorological parameters,
and aerosol measurements, is presented for both warm and cold clouds, and the
potential contribution of sampling artefacts is discussed in detail. We find
an excellent agreement of the GCVI sampling efficiency of liquid clouds using
two independent approaches. The 2-year data set of cloud residual size
distributions and number concentrations reveals that the cloud residuals
follow the typical seasonal cycle of Arctic aerosol, with a maximum
concentration in spring and summer and a minimum concentration in the late
autumn and winter months. We observed average activation diameters in the
range of 58–78 nm for updraught velocities below 1 m s−1. A
cluster analysis also revealed cloud residual size distributions that were
dominated by Aitken mode particles down to around 20–30 nm. During
the winter months, some of these small particles may be the result of ice,
snow, or ice crystal shattering artefacts in the GCVI inlet; however, cloud
residuals down to 20 nm in size were also observed during conditions
when artefacts are less likely.


11 43 (0 5) cm 3 532  .Dependence of activated fraction on ambient air temperature.The activated fraction is here defined as the ratio of cloud residual and total particle concentrations integrated above 100 nm diameter (as in Fig. 7 in Verheggen et al., 2007).The ratios are based on 30 min average concentrations.Solid and dotted lines show median and mean values, respectively, and shaded areas indicate the 25 th to 75 th percentile ranges.
Table S1.Hours of concurrent data from different instruments.Table showing the amount of simultaneous data, rounded to the nearest hour, for different combinations of instruments.The data are divided into cloudy (i.e.GCVI on and visibility < 1 km) and clear (i.e.GCVI off and visibility > 1 km).Data combinations that are not available or not used within our analysis are marked with dashes.

Figure S1 .Figure S2 .
Figure S1.Wind conditions during cloud sampling at Zeppelin Observatory.a-b show monthly averages (upper panels) and histograms (lower panels) of horizontal wind speed and updraft, respectively.For the upper panels in a-b, solid and dotted lines show median and mean values, respectively, and shaded areas indicate the 25 th to 75 th percentile ranges.c shows a histogram of the wind direction.In all panels, data are only from cloud sampling periods.

Figure S3 .
Figure S3.Comparison of different numbers of clusters in k-means.Normalised (left) and non-normalised (right) cloud residual number size distributions resulting from k-means clustering using a-b 2 clusters, c-d 3 clusters, e-f 4 clusters, g-h 5 clusters, i-j 6 clusters.Solid and dotted lines show median and mean values, respectively, and shaded areas indicate the 25 th to 75 th percentile ranges.

Figure S4 .
Figure S4.Transmission efficiency of the GCVI inlet and cloud particle size distributions.Left y-axis: The squares show the experimentally determined transmission efficiency of the GCVI from Shingler et al. (2012) (for the tubing prior to the expansion segment).The grey curve shows the transmission efficiency extrapolated to cover the fog monitor particle size range.Right y-axis: Cloud particle number size distribution (CPSD) during the sampling period shown in red.Solid and dotted lines show median and mean values, respectively, and the shaded area indicates the 25 th to 75 th percentile range (inter-quartile range, IQR).The grey shaded region covers the part of the size distribution that is smaller than the GCVI cut size, and hence is not measured by the instrument.
Figure S5.Comparison of measured and calculated visibility.a 2D histogram of visibility as measured by the GCVI visibility sensor (y-axis) versus visibility calculated from the cloud particle size distributions measured by the FM-120 fog monitor (x-axis).b Same as a, but colourcoded by temperature instead of data point density.Visibility was calculated assuming spherical particles, a refractive index of 1.33, and a wavelength of 880 nm of the visibility sensor.

Figure S6 .Figure S7 .
Figure S6.Heatmaps of wind parameters for different cloud residual and cloud particle concentrations.The figure shows heatmaps where cloud residual and cloud particle number concentrations have been cross tabulated with a wind speed b updraft and c wind direction (rounded to the nearest 10 degrees).Note that the pixels do not contain the same number of data points, please refer to Figure 2b in the main manuscript for the corresponding density plot.

Figure
Figure S8.D 50% dependence on updraft.a Ratio of mean size distributions, i.e. cloud residual concentrations divided by total particle concentrations.The coloured ratio curves have been normalised to 1 at 200 nm.The grey dotted curves represent the position of the nonnormalised ratio curves (Fig.5in main manuscript).b Interpolated D 50% for each updraft interval (i.e. the diameter at which the ratio in a is 0.5).The "error bars" represent the width of the diameter bin within which each interpolated value falls.

Figure S9 .
Figure S9.Monthly average cloud residual number size distributions.Data have been segregated based on temperature, T .The columns show data for a all T , b T > −4 • C, c T ≤ −4 • C. Solid and dotted lines show median and mean values, respectively, and shaded areas indicate the 25 th to 75 th percentile ranges.The numbers in the upper right corner of each panel indicates the 25 th to 75 th percentile ranges of the integrated cloud residual number concentrations (and cloud particle number concentrations in parentheses).The grey numbers below indicate the number of data points we have per month.

Figure S10 .Figure S11 .
Figure S10.Comparison of cloud residual number concentrations and cloud particle concentrations for the clusters in Fig. 2 in the main manuscript.Density scatter plots of cloud residual versus cloud particle number concentrations, including an orthogonal distance linear regression with coefficient of determination (grey line), for each cluster.The black dashed line represents the 1:1 line and the dotted lines represent 10:1 and 1:10 lines.The text in the lower right corner indicates the percentage of data where the two concentrations are withina factor of 2, 5, and 10 from each other (i.e. the amount of data between, e.g., the 10:1 and 1:10 lines etc.).Panels a-e show the data for clusters 1-5, respectively.The transmission efficiency of the GCVI inlet(Shingler et al., 2012) has been included in the calculation of the cloud particle number concentration in all panels.