The interactions that occur between aerosols and a mixed-phase cloud system, and the subsequent alteration of the microphysical state of such clouds, are a problem that has yet to be well constrained. Advancing our understanding of aerosol–ice processes is necessary to determine the impact of natural and anthropogenic emissions on Earth's climate and to improve our capability to predict future climate states. This paper deals specifically with how aerosols influence ice mass production in low-level Arctic mixed-phase clouds. In this study, a 9-year record of aerosol, cloud and atmospheric state properties is used to quantify aerosol influence on ice production in mixed-phase clouds. It is found that mixed-phase clouds present in a clean aerosol state have higher ice water content (IWC) by a factor of 1.22 to 1.63 at cloud base than do similar clouds in cases with higher aerosol loading. We additionally analyze radar-derived mean Doppler velocities to better understand the drivers behind this relationship, and we conclude that aerosol induced reduction of the ice crystal nucleation rate, together with decreased riming rates in polluted clouds, are likely influences on the observed reductions in IWC.
Surface temperatures in the Arctic are rising in response to increases in radiative forcings. The rate of warming in the Arctic is significantly higher than the mean rate of temperature increase for the globe (Manabe and Stouffer, 1980; Navarro et al., 2016). This warming has consequences to the physical and ecological systems of the Arctic environment, and these impacts are expected to become more severe in the future (Stroeve et al., 2008; Swart, 2017; Jay et al., 2011; Hinzman et al., 2013; Bindoff et al., 2013). These changes to the Arctic system have implications for biological and human activity in the region.
Numerous feedback mechanisms have been proposed as drivers of the observed amplified surface warming signal in the Arctic (Serreze et al., 2009). Modeling studies have indicated that surface-albedo and temperature feedbacks are the main mechanisms responsible (Serreze and Francis, 2006; Screen and Simmonds, 2010; Taylor et al., 2013; Pithan and Mauritsen, 2014). Yet limitations of models, including the required treatment of clouds through sub-grid parameterizations, leave gaps in our understanding of the role that clouds play in regulating the Arctic surface temperature response. Clouds are a prevalent and critical contributor to these central feedback processes because of the role they play in modulating the flux of energy to the surface. The micro- and macrophysical properties of clouds influence the thermodynamic and radiative properties of the atmosphere (Curry and Ebert, 1992; Pinto 1998; Shupe et al., 2011). The net impact of a cloud on the surface energy budget is strongly dependent on, among other factors, the phase of the water of which it is composed (Shupe and Intrieri, 2004). Cloud phase also impacts precipitation characteristics and is a factor in cloud lifetime, which is another relevant parameter governing how clouds fit into the Arctic climate system. Understanding phase partitioning in clouds is therefore critical, but an incomplete view of key microphysical processes, including ice nucleation, inhibits such understanding (Prenni et al., 2007).
Ultimately, our incapacity to understand the physics driving cloud systems hampers our ability to evaluate future climate states. Global circulation models (GCMs) allow us to assess the Earth system response to a variety of climate forcing scenarios. Even though such models are routinely invoked for guiding policy and scientific understanding, they are constrained by their inability to represent certain physical processes. Limited computational power requires sub-grid parameterizations of clouds and cloud processes that often do not represent reality. The representation of clouds and cloud phase requires substantial improvement, with both temperature-dependent and prognostic phase partitioning schemes having been demonstrated to be inadequate. For example, Cesana et al. (2015) determined that, even with state-of-the-art prognostic cloud microphysics, models such as CAM5 and HadGEM still had significant biases in the representation of ice clouds. Such biases result in models having significant surface temperature errors, such as those found over Greenland's ice sheet in CAM5 (Kay et al., 2016). The impacts of these model limitations become particularly clear in sensitive parts of the world, such as the Arctic, where there is significant variability in cloud phase. Improved understanding of cloud processes can help to alleviate GCM shortcomings.
The thermodynamic conditions (e.g., temperature and supersaturation) available for cloud formation in the troposphere necessitate that aerosols be present for cloud development to occur. Thus, aerosols are a fundamental component of mixed-phase clouds, acting as cloud condensation nuclei (CCN) and ice-nucleating particles (INPs). In the Arctic, aerosol concentrations follow a seasonal cycle with a high number of aerosol particles transported to the region from midlatitudes in winter and spring. This phenomenon, known as Arctic haze, results from accumulation of transported particles in a thermodynamically stable environment, where precipitation and chemical reactions are both limited due to the cold Arctic night (Barrie, 1986; Shaw, 1995; Quinn et al., 2007; Law et al., 2014). Yet understanding the relevance these aerosols have to Arctic cloud processes is difficult because of our limited understanding of the aerosol composition, size and vertical distribution present. For example, scarcity of INPs (Bigg, 1996) is a significant limitation on ice mass production and is a feature of the Arctic environment used to explain the long persistence times of mixed-phase clouds (Pinto, 1998; Harrington et al., 1999). Additionally, INP concentrations have been shown to vary greatly in time and space in the Arctic environment (Fountain and Ohtake, 1985; Rogers et al., 2001), and the development of INP parameterizations based on limited observational data has proven to be challenging (DeMott et al., 2010, 2015). This inadequate understanding of INP properties has led to difficulties in modeling ice-containing clouds. Arctic aerosol composition is an equally murky problem. Quinn et al. (2002) have shown that aerosol composition varies significantly throughout the year, with sulfate-coated particles being highly prevalent in spring. Still, a proper representation of aerosol concentrations and information on composition in and around mixed-phase cloud systems is lacking.
That being said, several aerosol–cloud effects have been detected in mixed-phase cloud systems: the first and second aerosol indirect effects have been observed (Lohmann and Feichter, 2005). These two aerosol indirect effects, associated with the liquid phase of cloud, lead to further aerosol-induced implications in mixed-phase clouds. The thermodynamic indirect effect, whereby the reduced mean liquid drop diameter caused by increasing CCN makes cloud conditions less favorable to secondary ice production (e.g., rime splintering, collision fragmentation), has the effect of reducing ice water content (IWC) in mixed-phase clouds with high CCN levels (Lohmann and Feichter, 2005). Similarly, the riming indirect effect, the process in which CCN reduce the liquid drop size distribution so that the liquid drops are less efficiently collected by falling ice crystals, reduces the riming rates within a mixed-phase cloud (Borys et al., 2003). The reduced riming rate decreases ice production and lowers cloud IWCs. Finally, the glaciation indirect effect, in which an increase in aerosols (traditionally INP from black carbon) is associated with greater levels of ice nuclei, promotes greater conversion of liquid to ice within the mixed-phase cloud layer (Lohmann, 2002, 2004). Yet the specifics of how these cloud processes play out over time to determine the macroscale properties of clouds is poorly understood.
Several observational studies have found evidence for aerosol impacts on Arctic mixed-phase clouds. Using surface-based sensors at Barrow, both Garrett and Zhao (2006) and Lubin and Vogelmann (2006) showed that a reduction of droplet size associated with elevated aerosol particle concentrations results in elevated emissivity of the cloud layer, thereby significantly increasing longwave radiation at the surface and contributing to warming. Lance et al. (2011) used in situ data from Arctic clouds to show that CCN concentrations, through the first indirect effect and riming indirect effect, may have a stronger influence on ice production than do INP concentrations. These past studies suggest that further interrogation of aerosol alterations to the microphysical state of mixed-phase clouds systems is warranted.
In this paper, we aim to demonstrate that aerosol interacts with Arctic mixed-phase cloud systems in ways that control ice crystal nucleation rates, as well as ice mass growth processes. To do this, we utilize a 9-year record of radar, microwave radiometer and radiosonde measurements from the US Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Program facility in Utqiaġvik (formerly Barrow), Alaska, along with aerosol measurements made by the National Oceanic and Atmospheric Administration (NOAA) Global Monitoring Division (GMD) to evaluate relationships between cloud IWC and aerosol concentrations near the surface. In the following sections, we first provide an overview of the instruments and methods used in this study. This is followed by observational results and a discussion of these results and their impact on our understanding of cloud ice production.
A multi-sensor method is used to identify stratiform mixed-phase clouds, which are the subject of this study. These clouds are characterized by having shallow liquid layers, the tops of which are at heights less than 2 km from the surface. The clouds may or may not be precipitating to the surface. Radar and other remote-sensing tools are used to characterize ice and liquid properties of these cloud layers. Ground-based measurements of aerosol scattering coefficients are used to approximate the aerosol loading of the lower atmosphere. Finally, radiosondes, in combination with ground-based remote sensors and model output, are used to classify the thermodynamic state of the atmosphere during cloudy periods.
Sampling took place at the ARM North Slope of Alaska (NSA) site, located just
to the northeast of Utqiaġvik, Alaska (71.323
Vertical profiles of radar reflectivity and retrieved IWC are based on
reflected power measured by a vertically pointing Ka-band 35 GHz millimeter
cloud radar (MMCR; Moran et al., 1998; Kollias et al., 2007; ARM, 1990a). The
MMCR product used here provides a 45 m vertical resolution and 10 s
temporal resolution. Five-minute averages of the reflectivity are used to
estimate IWC using an empirically derived power-law relationship:
We use ice crystal fall speed (
We use liquid water path (LWP) to classify the amount of liquid water in the mixed-phase cloud. This classification is done to control for environmental influence on cloud liquid water, which can interact to form ice within the cloud. That is, we are interested in aerosol effects on clouds for different cloud system types as defined by the LWP of the cloud. By comparing cloud ice properties for narrow LWP values, distinguishing between the microphysical differences that exist among clean and polluted clouds becomes possible. LWP is derived from brightness temperature measurements at 23.8 and 31.4 GHz from a microwave radiometer (Turner et al., 2007). When the physical method for retrieving LWP was not available, a variable-coefficient, bilinear statistical method was used (Liljegren et al., 2001). Respectively, these are the ARM MWRRET and MWRLOS retrievals (ARM, 1993).
Cloud top height is inferred from radar reflectivity profiles and is defined to be the height of the highest radar return of the low-level cloud, similar to the method of Moran et al. (1998). Cloud base height is defined at the bottom of the liquid-containing layer, which is determined from 905 nm Vaisala ceilometer measurements (15 m vertical resolution). In clouds devoid of liquid water or with intense precipitation, the ceilometer backscatter signal does not clearly define a cloud base height. In these cases, the discontinuity point in this ceilometer signal is used to identify cloud base (Shupe et al., 2013).
We use 1 min averaged values of scattering coefficient at 550 nm, which are measured at the surface by a TSI nephelometer deployed as part of the aerosol observing system (AOS; ARM, 1990b). These surface-based measurements are used to approximate aerosol concentrations in the cloud layer. Scattering coefficients have been used to identify atmospheric aerosol loading in past studies because cloud-relevant aerosols are often efficient at scattering 550 nm light (Garrett et al., 2004; Garrett and Zhao, 2006). The AOS did not continuously operate over the 9-year period, causing numerous periods with missing scattering coefficient data. Linear interpolation is used between scattering coefficient measurements separated by less than 24 h to infer the scattering coefficient value at 1 min intervals to match the time and resolution of the IWC profiles derived from the MMCR. If the sampling time for any given IWC profile is more than 24 h from the nearest aerosol data point, the profile is not used in this study.
We separate the data set into clean and polluted regimes to study the aerosol impact on cloud IWC. For the remainder of this paper, polluted conditions are defined to be the top 30 % of recorded scattering coefficient values, and clean condition the lowest 30 % of scattering coefficients for the set of cloud IWC profiles under study. The middle 40 % of the data are not considered.
Radiosondes were launched by both the DOE ARM program and the National
Weather Service (NWS) office in Utqiaġvik at a frequency of one to four
times per day over the course of the MMCR data record. These radiosonde
measurements are used to evaluate temperature and supersaturation with
respect to ice and liquid within the cloud layer. Because the radiosondes
were launched at 6- or 12-hourly increments, we use the DOE ARM
Merged Sounding value-added product (MERGESONDE).
value-added product to obtain information for the time periods between the
balloon flights (ARM, 1996). This product combines radiosonde, ground-based
remote sensor and forecasting model data to interpolate temperature and
humidity fields between radiosonde profiles (Troyan, 2012). We expect dry
biasing errors to be minimal in both the radiosonde and
MERGESONDE data sets because the
data used in this study come from relatively warm and humid regions of the
atmosphere (Fleming, 1998), and therefore we did not correct for these
effects. The corresponding temperature and humidity profiles for each IWC
profile are identified from the MERGESONDE data, and they are used to characterize the maximum
in-cloud relative humidity with respect to ice (RH
To observe the effects of aerosol on IWC in mixed-phase clouds, we examine
the shape of a mean IWC profile under polluted and clean aerosol conditions.
To do so, we create vertical IWC profiles that are normalized in depth. Cloud
base, the bottom of the liquid layer, is assigned a value of 0, and cloud top
a value of 1. For each IWC profile the IWC values are placed on a linear grid
between 0 and 1, proportional to their fractional height above cloud base.
The resolution of the normalized cloud grid is set so that it matches the
number of sampled points by the radar of a 1 km thick cloud,
Reflectivity profiles for the five LWP regimes for
Radar reflectivity, IWC and
We restrict this study to clouds with liquid layers less than 1 km thick. Cloud depth is an important parameter because it helps to define the scale of the interaction zone for liquid and ice particles, with deeper clouds having more opportunity to convert liquid water to ice. Additionally, deeper clouds tend to have stronger and more complex dynamics than do shallower clouds, which can obscure the view of aerosol influences on cloud ice. Cloud base height is arbitrarily limited to below 2 km to increase the likelihood of coupling between the cloud and the surface, where the aerosol measurements occur. In this study, we do not explicitly require, or attempt to identify, coupling between the surface and cloud layer.
Summary statistics for the sampled clouds in each LWP bin. The last row is the aggregate sample from all LWP bins.
Clouds are required to have a maximum relative humidity with respect to ice
(
The liquid water layer depth and liquid water density are moderators of
deposition and riming rates. To constrain the influence of liquid water on
ice formation, we designate LWP regimes within which to compare cloud ice
properties. Clouds are sorted into five LWP bins:
LWP0
In this section we present mean in-cloud profiles of reflectivity, IWC and
Figure 1 shows reflectivity in relation to normalized height in the cloud layer.
Dashed lines represent mean reflectivity profiles from the aggregation of
polluted cases, and solid lines correspond to clean cases. The line color
designates the LWP regime of the cloud. To determine if the difference
between the clean and polluted profiles is statistically significant, we
perform an unequal variance
Vertical profiles of ice water content for clean and polluted
conditions, as in Fig. 1, for
At cloud top, reflectivity values are typically small due to the presence of small cloud hydrometeors. Reflectivity increases with decreasing height in the cloud layer as ice mass growth occurs due to deposition and riming. These increases in reflectivity are more prominent for clean cases than for polluted clouds in all LWP bins, such that reflectivity is larger near cloud base for clean clouds. This is indicative of a greater rate of ice mass growth through the column for clean clouds. There is not a linear response in reflectivity to LWP: the ice-dominated clouds (LWP0) most often have the lowest reflectivity values, relative to all other LWP bins, in the bottom half of the cloud layer. The highest reflectivity values near cloud base are found in intermediate LWP cases (LWP1 and LWP2), while the highest LWP clouds have lower reflectivity values.
Radar reflectivity is a direct measurement made by the cloud radar and includes no assumptions about cloud microphysics, though it is dependent on the cloud properties. The segregation of reflectivity profiles presented in Fig. 1 is evidence for aerosol interactions within mixed-phase cloud systems. In the following subsections, we further examine cloud IWC and ice crystal fall speed profiles for insight into the details of these aerosol–ice interactions.
The observed reflectivity values are transformed to IWCs through the power-law method outlined in the Sect. 2. The mean IWC profiles are presented in Fig. 2.
In all LWP bins, IWC values are less than 1 g m
The ordering of IWC at cloud base as a function of LWP is consistent with
that of the cloud base reflectivity. The exception is for polluted clouds,
where LWP4 has a higher cloud base IWC than does LWP3, which is the reverse of
what is found for reflectivity. While the ordering of the profiles is, more
or less, consistent, the shape and relative positions of the lines vary
between reflectivity and IWC. These inconsistencies between the two variables
are caused by the seasonal nature of the IWC power-law retrieval. In later
months (late spring) LWP tends to increase, while the
Similar to the reflectivity profiles, statistical significance is determined
by a two-sample
Vertical profiles of 120 min time-averaged mean Doppler
velocity for clean and polluted conditions for
The vertical structure of mean ice crystal fall speed in the cloud layer indicates changes in the size, surface-area-to-volume ratio, crystal habit and crystal orientation. Generally, nucleation, deposition, aggregation and riming are the significant processes that change the ice mass to cross-sectional area relationship for a given cloud volume, and therefore variations in ice crystal fall speed are inherently linked to these processes. Ice crystal fall speed, in combination with IWC, allows us to infer relative information about ice crystal size and number properties at a given location in the cloud layer if we assume similar crystal habits and crystal orientations at each layer.
Figure 3 shows the vertical fall speed profiles of all cloud cases. At cloud top, the mean fall speed of ice crystals for polluted clouds is greater than the mean fall speed for the clean cases in all LWP scenarios except for ice clouds (LWP0 cases). Considering the equivalent IWCs at cloud top, the greater fall speeds in the polluted clouds indicate the presence of ice crystals with larger geometric mean size, which must be matched with a reduction in ice crystal number, to drive the observed reflectivity/IWC response (a more detailed discussion is offered in Sect. 4.1, 4.2). Alternatively, the fall speed variation could be the result of aerosol-induced changes in crystal habit and orientation, though we do not have evidence that these properties are influenced by INP concentrations.
Since the measured radar reflectivity scales approximately with the sixth power of hydrometer size, it is the largest hydrometers that will reflect the most radiation back to the radar. Thus, the reflectivity and in turn the fall speed signal are dominated by the largest hydrometers in the sampled volume. If a fixed amount of ice is sampled, it is not possible to determine if increases in reflectivity are due to an increase in the mean of the geometric size distribution of the ice crystals or a broadening of this distribution. However, the nonlinear response of reflectivity to ice crystal size does mean that there is an increase in the presence of large ice crystals (sizes greater than the geometric mean). This knowledge about the relative populations of large ice crystals lets us make broad claims about ice crystal number concentrations in these clouds.
The relationship between cloud layer depth and
It is important to note that, at cloud base, LWP1 and LWP2 clouds have the
highest IWC values, yet these clouds have the lowest
A notable feature of Fig. 3 is the high cloud top
In-cloud minimum temperature and mean cloud layer temperature for profiles in each LWP bin.
The temperature and humidity properties of the environment in which a cloud
forms influence the ice properties of the cloud. To define an aerosol
alteration to the cloud microphysical state, we first need to examine the
impact the environment has on the cloud ice properties. Figures 5 and 6
present statistics of minimum cloud temperature (
Given the similarity in RHi levels, and that the differences in minimum temperatures for clean and polluted clouds within each LWP bin are minor enough to not significantly alter ice crystal habit properties (Bailey and Hallet, 2009), the distribution of ice crystal habits of the nucleated ice crystals in both cases should be similar. For LWP1,2,3,4 plate and dendrite type crystals are likely to be common, while LWP0 may be more apt to produce columns (Bailey and Hallett, 2009). For a given habit type, the reflectivity differences are dominated by variations in ice crystal size (Hong, 2007), and therefore the observed variations in measured radar reflectivity likely cannot be explained by habit effects alone. More generally, the variation in temperature and supersaturation levels between clean and polluted clouds cannot explain the observed differences in IWC. This further supports the notion that aerosols are altering the microphysical state of the cloud in manners which suppress ice mass production. These mechanisms are detailed in Sect. 4.
Mean minimum cloud temperature for clean and polluted clouds for
each LWP bin. Diamond markers with solid line indicate clean clouds; circle
markers with dashed line represent polluted cases. The bars span the
20–80th percentile of the
A few other features of the temperature and RHi distributions are
interesting to note:
The mean temperature of the cloud layer increases with increasing LWP
bin. This is likely due to the ability of warmer air masses to support higher
levels of liquid water. Additionally, there may be a slight seasonal effect
as the mean sample day (Fig. 7) of all the polluted and clean cases only
varies slightly amongst the LWP bins. It is also interesting that there are few
high-LWP clouds found at relatively cold temperatures – LWP4 has few clouds with
Likewise, there are more While the RHi distributions for clean cases are fairly uniform across all
the LWP bins, there is a high amount of variation in the RHi distributions
for the polluted-LWP bins. This could be the result of greater variability in
the meteorological conditions under which polluted clouds are found.
Mean in-cloud RHi for clean and polluted clouds for each LWP
scenario. Diamond markers with solid line indicate clean clouds; circle
markers with dashed line represent polluted cases. The bars span the
20–80th percentile of the
The observations presented in Sect. 3 indicate that polluted clouds have
reduced amounts of cloud ice mass for a given amount of condensed liquid
mass. In the cloud top region of polluted clouds, the high
Mean day of sample for the distribution of sampled profile for clean and polluted clouds for each LWP scenario. Diamond markers with solid line indicate clean clouds; circle markers with dashed line represent polluted cases. The bars span the 20–80th percentile of the day-of-year distribution for each bin.
Further down in the cloud layer, as ice crystals undergo growth processes,
The cloud top region is an area of the cloud where ice crystals tend to have
had little time to interact with the cloud system, and therefore this region
contains the most direct picture of heterogenous ice nucleation. As
previously stated, polluted clouds appear to have lower ice crystal
concentrations, which is evidence for a reduced ice crystal nucleation rate.
Two potential mechanisms for this reduced nucleation rate are as
follows.
Hydrophobic aerosols can be coated by hydrophilic compounds and thus limit
their effectiveness to nucleate ice (Diehl and Wurzler, 2004; Girard et al.,
2005; Kulkarni et al., 2014). Variations in the wintertime and springtime
scattering coefficient measurements used in this study are most strongly
influenced by fluctuations in sulfate aerosols (Quinn et al., 2002).
Therefore, the conditions we define as polluted are ones in which INPs are
likely to be coated by sulfates, reducing the efficiency at which they
nucleate ice. Through this lens, we would expect polluted conditions to be
associated with a reduction in ice crystal nucleation rate. We refer to this
as the INP mechanism for ice nucleation suppression. There may be a size dependence to the ability of liquid droplets to freeze,
but the literature is murky as to why this is the case. Ideas include
freezing point depression from increased solute concentrations (de Boer et
al., 2010), though at the liquid drop sizes typical of Arctic mixed-phase
clouds this seems unlikely to be causing the reduced nucleation rate. More
probable is that the greater surface area of larger drops provides a larger
interaction area for the liquid drop with their environment, such as
contacting an INP, and thus larger drops have a greater probability to
freeze. Through the first indirect aerosol effect, the increase in aerosol
concentration reduces both the mean droplet size and the width of the drop
size distribution (Chandrakar et al., 2016). The suppression of the ice
nucleation rate through a reduction in the mean diameter of liquid droplets
in mixed-phase clouds (e.g., Lance et al., 2011) could explain the observed
reduction in ice crystal number at cloud top. We refer to this as the CCN
mechanism for ice nucleation suppression.
Our observations are consistent with simulations done by Girard et al. (2005) that show increasing sulfuric acid aerosols in Arctic clouds reduces ice crystal number concentrations, while mean ice crystal size is increased. Other studies have found evidence for the ability of sulfates to suppress the onset of heterogeneous freezing (Eastwood et al., 2009), and such inhibition results in the generation of fewer but larger ice crystals (Jouan et al., 2014). However, we currently do not have the measurements needed to determine which mechanism (CCN or INP) is playing the bigger role in controlling ice production in these clouds. Observing in-cloud ice crystal size distributions with optical probes would provide insight into the size and shape variability of nucleated and grown ice crystals. The variability in nucleated ice crystal size should be linked to the in-cloud CCN or INP properties, with higher ice crystal size variability expected if INPs are the dominant control on nucleation.
The reduced nucleation rate in polluted clouds has implications for the total amount of depositional ice mass growth in the cloud layer. The depositional growth rate for an individual crystal is proportional to the inverse of the effective radius of that ice crystal for most crystal habits (Rogers and Yau, 1989). This implies that depositional growth will lead to convergence of ice crystal sizes given sufficient time for growth to occur. In the clouds under study, we believe the in-cloud residence time of an ice crystal is greater than the time it takes for this size convergence to occur – see Appendix A. This suggests that IWC gained through deposition is strongly determined by initial crystal number, not by initial crystal size. Thus, the higher ice crystal number concentrations found in clean clouds directly result in greater total amounts of depositional growth.
In addition to deposition, riming and rime splintering can contribute to
increased IWC in clean clouds. The highest LWP regime, LWP4, is the only case
in which ice crystals in clean clouds have greater
We expect the level of riming to be related to the amount of liquid water
contained within the cloud. In the LWP1,2 cases, the low amount of liquid
water makes riming relatively less efficient and perhaps non-existent.
Therefore, we speculate that ice mass gains in these low-LWP clouds are
mainly occurring through depositional growth. These clouds also tend to have
cold temperatures, promoting the growth of dendritic crystal habits. Dendrite
fall speeds are slow relative to other crystal types with similar mass
(Kajikawa, 1974), and therefore these ice crystals have long in-cloud
residence times, enhancing depositional growth. Such depositional ice mass
growth is consistent with the observed high cloud base IWC and low
A 9-year record of ground-based observations of stratiform mixed-phase clouds
from Utqiaġvik, Alaska, was used in conjunction with surface measurements
of aerosol scattering coefficient to quantify the influence of aerosols on
ice production in these clouds. Profiles of reflectivity, IWC and
Regarding nucleation processes, our observations are consistent with two views of aerosol suppression of ice nucleation. First, our measure of surface aerosols is strongly correlated with the presence of sulfates in the atmosphere. The large amount of sulfate found under polluted conditions may be interacting with potential INPs, diminishing the efficiency at which they nucleate ice crystals. Second, aerosols may be reducing the liquid drop size distribution, which inhibits the ability of the drop to interact with the environment and freeze. Determining which of these two mechanisms is responsible for the suppression of ice nucleation would require knowledge of the size distributions of the nucleated ice crystals and liquid drops in addition to better measurements of in-cloud aerosol composition. Additionally, both mechanisms might be at play in these clouds, further complicating the picture.
This paper then identifies how aerosols interact within the cloud system to affect rates of deposition, riming and rime splintering. We believe that, for depositional processes, suppression of ice nucleation in polluted clouds reduces the ice crystal number concentration and in turn the ice surface area available for deposition, and therefore the total amount of depositional ice mass is reduced. Riming, on the other hand, is dependent on the number of ice crystals in addition to the liquid drop size distribution. Increasing CCN in polluted clouds reduces the effective radius of the liquid drops, which reduces riming efficiency and in turn decreases ice mass growth. The higher number of ice crystals and larger liquid drops prevalent in clean clouds result in an environment that is more favorable to riming processes, particularly when LWP is high. Likewise, rime splintering, and the associated rate of new ice crystal formation, is directly related to the riming rate.
It is important to note that our analysis does not rely on direct knowledge of INP or CCN populations, and we make no assumptions about how the scattering coefficient measurements represent INP/CCN levels. Instead, we have treated the problem of ice mass growth in mixed-phase clouds in relation to the general aerosol population as defined by the surface measurements. In doing so, we have shown that ice mass growth is sensitive to the variations in the surface-measured aerosol population. This study supports the hypothesis that the ice properties of a cloud are influenced by CCN and liquid phase processes. Having said this, to truly understand the relative roles of INPs and alterations in the liquid properties of the cloud on ice nucleation and growth processes, a more advanced understanding of INPs present in these mixed-phase cloud systems is needed.
Advances in this area will be required to truly constrain how a mixed-phase cloud interacts with the greater Arctic and global climate system. This coupling is largely tied to cloud phase, which is inherently linked to ice production mechanisms. The rate at which ice is produced in a mixed-phase cloud has direct consequences for the cloud macroscale properties, such as the cloud net radiative effect, lifetime and precipitation characteristics. Further work will enable a more detailed understanding of how aerosols alter in-cloud microphysics and the subsequent macrophysical properties of these clouds – necessary research for a complete view of the broader climate system.
Data used in this study are publicly
available from the DOE ARM data archive
(
In Sect. 3.6 we argued that deposition will cause nucleated ice crystals of varying size to converge to the same size in a time that is typically less than the residence time of an ice crystal in the cloud layer.
An approximation of the deposition rate is given by Rogers and Yao (1989):
For a plate type ice crystal,
Equation (A1) understates the depositional rate for small ice crystals at
warmer temperatures (
The time for nucleated ice crystals to converge to within 90%
of an ice crystal growing in the same environment but with an initial
nucleated size of 1 mm,
Computed cloud residence times for clean and polluted clouds for three cloud depths.
The other consideration is the ice crystal residence time in the cloud,
Consistently,
MSN led the study and worked together with GdB and MDS to develop and write the manuscript. MDS generated the IWC retrievals. MSN, GdB and MDS analyzed and interpreted the data.
The authors declare that they have no conflict of interest.
This research was conducted primarily under support from the US Department of Energy (DOE) Atmospheric System Research Program under grant DE-SC0013306. Additionally, support was provided by DOE grants DE-SC0011918 and DE-SC0008794 and by the National Science Foundation under grant ARC 1203902. Cloud and atmosphere data products were obtained from the Atmospheric Radiation Measurement (ARM) Climate Research Facility, a US Department of Energy Office of Science User Facility sponsored by the Office of Biological and Environmental Research. Aerosol measurements were obtained from the National Oceanic and Atmospheric Administration's Earth System Research Laboratory – Global Monitoring Division. We also thank Jessie Creamean for useful discussions and support in the early stages of this of this project. Edited by: Xiaohong Liu Reviewed by: two anonymous referees