How orographic mixed-phase clouds respond to the change in
cloud condensation nuclei (CCN) and ice nucleating particles (INPs) are
highly uncertain. The main snow production mechanism in warm and cold
mixed-phase orographic clouds (referred to as WMOCs and CMOCs, respectively,
distinguished here as those having cloud tops warmer and colder than
Snowpack in the Sierra Nevada is California's largest source of fresh water. Understanding the factors contributing to snow precipitation over the mountains has important implications for predicting the hydrology and local climate of the western US. This has motivated a series of CalWater field campaigns carried out since 2009 to improve understanding of processes influencing precipitation and water supply in California (Ralph et al., 2016). Closely linked to precipitation is the distribution of cloud liquid and ice phases, which may be influenced by supercooled liquid commonly occurring in orographic clouds over the Sierra Nevada (Rosenfeld et al., 2013). In addition to precipitation, cloud radiative forcing and cloud feedback in the climate system are also highly dependent on cloud phases because of the very different radiative effect of liquid and ice particles. Hence, understanding the key processes and factors impacting cloud phases is critical, but our lack of understanding and ability to model supercooled liquid and cloud phases is limiting skillful predictions on weather and climate timescales.
Many factors, such as large-scale dynamics, solar heating, and aerosol particles, can impact cloud properties and precipitation over the Sierra Nevada (Shen et al., 2010; Rosenfeld et al., 2008). Atmospheric rivers (ARs) are one of the primary large-scale dynamical features that bring large amounts of water vapor from tropics to the US west coast, and they can create extreme rainfall and floods (Bao et al., 2006; Ralph et al., 2011; Neiman et al., 2010). Aerosols can modify cloud microphysical processes and potentially alter the location, intensity, and type of precipitation (Tao et al., 2012) by acting as cloud condensation nuclei (CCN) or ice nucleating particles (INPs). In California, anthropogenic aerosols from the densely populated coastal plains and the Central Valley may be incorporated into the frontal air mass before orographic ascent and influence precipitation in the Sierra Nevada (Rosenfeld and Givati, 2006). Long-range transported aerosols (mainly dust particles) have also been found to have a potential influence on clouds and precipitation in the winter and spring seasons (Uno et al., 2009; Ault et al., 2011; Creamean et al., 2013).
Aerosol impacts on clouds not only depend on aerosol properties such as number, size, and composition but also dynamics and thermodynamics. Rosenfeld et al. (2014) showed significantly different supercooled-water (SCW) and precipitation processes in two contrasting cloud cases with air masses containing maritime and continental aerosols, respectively. Many studies have shown that CCN can reduce warm rain precipitation from orographic clouds by reducing the efficiency of cloud droplet conversion into raindrops (e.g., Lynn et al., 2007; Rosenfeld and Givati, 2006; Jirak and Cotton, 2006), and they can also reduce snowfall precipitation due to reduced riming efficiency (Lowenthal et al., 2011; Rosenfeld et al., 2008). However, some recent studies show a possibility of increased precipitation by CCN in orographic mixed-phase clouds (Fan et al., 2014; Xiao et al., 2015). Other studies have shown that CCN may not have a significant effect on the total precipitation but rather they shift precipitation from the windward to leeward slope, a so-called “spillover effect” (Lynn et al., 2007; Saleeby et al., 2011, 2013). By acting as INPs, aerosols can enhance ice growth processes such as deposition and riming and thereby significantly increase snow precipitation (Fan et al., 2014). Both observational and modeling studies have shown that long-range transported dust and biological particles can enhance orographic precipitation in California by serving as INPs (Ault et al., 2011; Creamean et al., 2013, 2014, 2015; Fan et al., 2014).
In addition to precipitation, aerosols may have significant impacts on cloud phase and SCW in the mixed-phase clouds, which directly change cloud radiative forcing and Earth's energy balance. Modeling studies have shown that CCN tend to increase SCW via the processes such as suppressed warm rain and/or reduced riming efficiency (Khain et al., 2009; Ilotoviz et al., 2016; Saleeby et al., 2013). A recent observational study corroborated that increasing CCN decreases the cloud glaciation temperature and thus increases the abundance of the mixed-phase regime (Zipori et al., 2015). With abundant INPs such as dust particles, clouds glaciate at a much warmer temperature (Rosenfeld et al., 2011; Zipori et al., 2015). It was found that commonly occurring supercooled water in the clouds near the coastal regions of the western US is associated with low-CCN and limited-INP conditions (Rosenfeld et al., 2013). Models generally have difficulties simulating SCW and cloud phases. For example, recent evaluation of the Community Atmosphere Model version 5 (CAM5) with satellite data showed that the model has insufficient liquid cloud and excessive ice cloud from the midlatitudes to the polar regions, and liquid deficit bias maximizes over the Southern Ocean where supercooled water is prevalent (Kay et al., 2016). For cloud model simulations with cloud-resolving models, ice nucleation parameterizations often need to be modified in order to produce the mixed-phase clouds in the Arctic region (Fan et al., 2009; Fridlind et al., 2007). Considering that many microphysical processes are sensitive to aerosol types (CCN or INP), temperature, and/or supersaturation (e.g., deposition growth), aerosol impacts on cloud phase and SCW can be complicated depending on cloud dynamics and thermodynamics. Our current understanding of cloud microphysical processes impacting SCW and cloud phase in different meteorological environments is poor. Therefore, it is important to conduct process-level studies to improve our understanding.
Fan et al. (2014) conducted a study for two mixed-phase orographic cloud
cases with different cloud temperatures and showed different significance of
the CCN and INP impacts on precipitation between the two cases, with much more
significant impacts of INPs. The two cases are 15–16 February 2011 (FEB16)
and 1–2 March 2011 (MAR02). FEB16 has a cloud top temperature as cold as
As in Fan et al. (2014), simulations are performed using WRF version 3.1.1
developed at the National Center for Atmospheric Research (NCAR) (Skamarock
et al., 2008) coupled with a
SBM model (Khain et al., 2009; Fan et al., 2012).
The SBM model is a fast version of the full SBM model described by Khain et al. (2004),
in which ice crystal and snow (aggregates) in the full SBM model are calculated
based on one size distribution with separation at 150
As discussed in Fan et al. (2014), hereafter referred to as FAN2014, the ice
nucleation parameterizations in the SBM model used for this study were
modified. A new ice nucleation parameterization of DeMott et al. (2015) (cited
as DeMott et al., 2013 in FAN2014 before the parameterization was published)
was incorporated in the SBM model to investigate the impacts of dust as INPs. The
parameterization connects nucleated ice particle concentration under a
certain atmospheric condition with aerosol particle number concentration with
a diameter larger than 0.5
In FAN2014, simulations were done for the two nested domains with a
horizontal grid spacing of 10 and 2 km. To focus on the
orographic clouds over the Sierra Nevada and provide a better
process-level understanding, we conduct new simulations using a smaller
domain of 300 km
CCN in the model are represented by a spectrum with 33 size bins with
prognostic CCN number concentrations for each bin. As stated above, dust–bio
particle number concentration serves as a proxy for INP concentration in this
region. For the purpose of this study, we conduct sensitivity tests by
varying CCN and INP proxy (i.e., dust–bio particle) concentrations over a
wide range from extremely low to extremely high concentrations, as shown
in Table 1. The initial CCN concentrations for the sensitivity simulations
are set to be 30, 100, 300, 1000, and 3000 cm
Model simulations that are run for different CCN and INP proxy
aerosol concentrations. Please note that INP proxy aerosol concentrations
denote dust–bio particle number concentrations with particle size
> 0.5
As described earlier, the CMOC case on FEB16 has cloud-top temperatures of
about
INP concentrations (L
The number concentrations (top row) and mass mixing ratios (bottom row) of droplets (first column), rain (second column), cloud ice (third column), and snow (fourth column) for the CMOC. The data are averaged over the grid points over the domain by excluding the lateral boundary grid points below the 7 km altitude and over the simulation time by excluding the first 2 h.
The microphysical process rates of
Figure 2a shows the accumulated surface precipitation averaged over the domain
for the CMOC case (FEB16). Increasing the INPs generally enhance the
domain-averaged precipitation except at extremely high CCN concentration
(i.e., 3000 cm
By looking at the in-cloud microphysical properties, as shown in Fig. 3,
increasing CCN enhances snow number concentration and mass mixing ratio
(
From the process rates of the major microphysical processes shown in Fig. 4,
we see that the increase in
At very low INP concentrations (IN0.1), the riming rate is similar to the
deposition rate in this CMOC case (Fig. 4c and e). As the INPs increase, the
contribution of riming is reduced significantly because of the reduction of
supercooled droplets resulting from increased ice particles in the
mixed-phase zone. Thus, the riming process is liquid-limited in this CMOC case. As
a result of increased ice particles, deposition is enhanced significantly,
and it becomes 3–4 times larger than riming in IN10. In the observed
condition (i.e., CCN are between 30 and 300 cm
The Wegener–Bergeron–Findeisen (WBF) processes refer to ice depositional growth at the expense of liquid through evaporation in mixed-phase clouds. Therefore, the mixed-phase cloud regime where vapor pressure falls between the saturation vapor pressure over water and ice is defined as the WBF regime. As CCN increase, the WBF processes get stronger, as shown in Fig. 5a and b. The ratio of evaporation through WBF to the total evaporation is larger than 0.92 in all simulations (Fig. 5a), meaning that drop evaporation in this CMOC case occurs predominantly in the WBF regime. Generally only 50–70 % of deposition occurs in the WBF regime even when INP concentrations are in a range (IN0.1 to IN1) that is typical for this region in winter (Fig. 5b); thus, a significant portion of deposition occurs outside of the WBF regime, and the portion increases as INPs increase. Therefore, increasing the INPs generally reduce the WBF regime because of the reduced liquid due to enhanced depositional growth. In this CMOC case, the ratio of riming that occurs in the WBF regime to the total riming is small (generally around 0.2–0.4 in Fig. 5c), meaning that riming mainly occurs outside of the WBF regimes under any CCN and INP conditions. The ratio is increased by CCN but generally decreased by INPs as a result of the increase or decrease of the liquid regime, respectively (Fig. 5c).
Vertical profiles of
We see that all major microphysical processes (condensation–evaporation,
deposition–sublimation, and riming) are highly sensitive to INPs, while
generally having much lower sensitivity to CCN when CCN are below
1000 cm
The west–east cross section of
Since the results of significantly enhanced precipitation from CCN1000 to CCN3000 are unusual, aside from verifying the use of identical initial and boundary meteorological conditions in all the experiments to eliminate simulation differences arising from inadvertent factors, we also conducted sensitivity tests by restoring the ice nucleation mechanisms to the default parameterizations (i.e., Meyers et al., 1992 for condensation and/or deposition and Bigg (1953) for immersing freezing) in the SBM but this yielded a similar conclusion. Consequently, the significantly increased snow precipitation associated with elevated CCN concentrations is not the result of the particular ice-forming parameterization or the implementation approach of the parameterization.
Since the precipitation enhancement begins at 14:00 UTC, which is a couple
of hours into the simulations, we focus on the time period of
14:00–16:00 UTC and use the simulations of different CCN concentrations for
the IN1 case to examine the mechanism. By taking a close look at ice
nucleation (using model outputs every 6 min), we find that the total
nucleated ice particle number concentration increases as CCN increase and
there is a large jump from CCN1000 to CCN3000 (Fig. 6a). The increase is
caused by more cloudy points where ice nucleation (i.e., immersion
freezing) occurs (Fig. 6b) and the enhanced nucleation rate (i.e., the
nucleated ice particles per liter of air volume within 1 h) in the lower
altitudes (Fig. 6c). Considering that the major ice formation mechanism in this study is
immersion freezing, which requires the existence of drops for
primary nucleation of ice, there is much more supercooled
liquid cloud area and/or volume available for nucleation in the lower altitudes as
CCN increase (Fig. 6e). As shown in Fig. 6d, the increase in cloud water
(
Differences in
What causes the drastic increase in
The spatial distribution of wind field at about 1.7 km above the
ground for
The changes in cloud fields described above must involve dynamic and
thermodynamic changes. By examining the differences of dynamic and
thermodynamic fields between CCN3000 and CCN30 (Fig. 9), we clearly see that
a band of increased water vapor and RH from the
valley and/or foothills to the mountain at the higher altitudes (Fig. 9a–b). The
corresponding temperature is only slightly decreased (Fig. 9c), which should
not affect the saturation water pressure and ice nucleation efficiency by
much. Therefore, the increased RH is mainly caused by the increased water vapor, and
this increase can be up to 8 % in RH (e.g., from RH of 70 to
78 %). The large increase in
Results for the two simulations without ice-related microphysics,
i.e., CCN30IN1_noice and CCN3000IN1_noice,
which are based on CCN30IN1 and CCN3000IN1, respectively, for the WMOC:
The changes in winds are only significant on the slope of the mountains and occur only after 2 h of the simulations (Fig. 10a), suggesting that they stem from more latent heat release as a result of more clouds over the valley and foothills (feedbacks of radiation and precipitation take a much longer time, especially considering the 2 h time, 04:00–06:00 LST). The clouds at the valley and/or foothill locations are generally shallow. Many literature studies, including both observations and model simulations, have shown that CCN enhance shallow cloud formation and deepen shallow clouds (e.g., Chen et al., 2015; Yuan et al., 2011; Pincus and Baker, 1994; Koren et al., 2014), which can be due to various reasons such as cloud lifetime effect, enhanced turbulent convection by larger entrainment rates as a result of stronger evaporation, and greater latent heat release due to larger drop surface area for stronger condensation. We find that condensation is indeed much enhanced over the valley and/or foothills from CCN30 to CCN3000 under IN1 (Fig. 9f), which results in much reduced supersaturation with respect to water (supersaturation around the cloud base in CCN30 at 13:00 UTC is about 0.28 %, while only 0.04 % in CCN3000). The enhanced condensation as well as the cloud lifetime effect (i.e., conversion of smaller droplets into rain is slow and clouds can be sustained for a longer time) contributes to more shallow clouds in the valley or foothills. The more latent heat resulting from enhanced condensation leads to the change in local circulation, which transports more moisture to the windward slope of the mountain, resulting in more active mixed-phase clouds and snow precipitation through enhanced deposition and riming. In addition, over the mountains more supercooled liquid would be lifted to the higher altitudes in the polluted condition, forming ice and snow more efficiently through immersion freezing at the colder temperature, which contributes to more snow precipitation as well.
It should be noted that the mixed-phase clouds over the mountains are the key
to the enhanced precipitation by CCN. This is confirmed by sensitivity tests
based on the WMOC case where ice-related microphysics were turned off in
CCN30IN1 and CCN3000IN1. We chose the WMOC for this sensitivity test because
the similar mechanism is present and the WMOC has less mixed-phase regime
compared with the CMOC; thus, the factor would have a more significant role in the
CMOC if it plays a role in the WMOC. As shown in Fig. 11a, precipitation is
dramatically suppressed from CCN of 30 to 3000 cm
In summary, increasing CCN forms more clouds in the valley and foothills (generally shallow) through much-enhanced condensation, which induces a local circulation change due to more latent heat release that enhances the zonal transport of moisture. This leads to the invigoration of the orographic mixed-phase clouds and drastically increased snow precipitation in this CMOC case. Therefore, aerosol impacts on orographic mixed-phase clouds can be extraordinary in extremely polluted conditions, especially under the influence of atmospheric rivers. In addition to the key role of ice processes for leading to greatly enhanced precipitation, orographic dynamics is another important factor since we do not see such impacts in the sensitivity tests where the terrain height is set to be 600 m for the locations with a terrain height > 600 m (precipitation becomes very small in those sensitivity tests and the increase from CCN30 to CCN3000 is small as well).
The increases of
The liquid mass fraction vs. temperature for the
The fraction of the liquid phase (left), ice phase (middle), and
mixed phase (right) for the
By changing the microphysical process rates, CCN and INP impact the cloud
phases and SCW content. Figure 12 shows that INPs have
the most striking impact on SCW. Increasing INPs enhance ice particle
formation and then facilitate the deposition and riming processes in this
CMOC case, as discussed in Sect. 3.1.1. The enhanced deposition in the WBF regime,
along with riming, leads to a faster conversion of liquid to ice in the
mixed-phase and glaciates the clouds faster. Therefore, SCW is substantially
reduced as INPs increase (Fig. 12a). For example, in the case of CCN300, a
significant amount of liquid mass fraction (0.1) exists at the temperature of
Same as Fig. 2, except for the WMOC.
Compared with the effects of INPs, the magnitude of CCN effects on SCW and cloud phases is much smaller but still significant (the lines with the same color but different line styles in Fig. 12). Moreover, the sign is opposite. Increasing CCN generally increases SCW slightly (Fig. 12a). The impact of CCN on cloud phases is generally small, except when INPs are very low, i.e., IN0.1 (Fig. 13a). In this low-INP case, increasing CCN increases ice phase fractions and reduces the mixed-phase fraction when CCN are relatively low. This is because liquid clouds are dominant, making such clouds sensitive to the CCN-enhanced ice nucleation as discussed in Sect. 3.1.2.
Same as Fig. 3, except for the WMOC.
Same as Fig. 4, except for the WMOC.
For this warm mixed-phase cloud case, the surface accumulated precipitation
is suppressed by increasing CCN when CCN are lower than 1000 cm
Same as Fig. 5, except for the WMOC.
The in-cloud microphysical properties also show similar results to the CMOC:
the steep increases of the snow mass and cloud water mixing ratios from
CCN1000 to CCN3000 (Fig. 15). We have done the same investigation as in
Sect. 3.1.1 and found that the mechanism causing the increased cloud water
and snow production is similar to that in the CMOC, that is, increasing
CCN forms more shallow clouds in
the large area of the valley and foothills, which significantly induces a
change in local circulation through more latent heat release, which in turn
increases the zonal transport of moisture to the windward slope of the
mountains. Additionally, more abundant warm rain is present in the wide
valley area in this case when CCN are low (30 cm
Different from the CMOC case, riming is a more efficient ice growth process for forming snow than deposition in this case, except when INP concentrations are extremely high (IN100) where both riming and deposition contribute at a similar magnitude (Fig. 16). In addition, the riming rate increases as INP concentrations increase, which is opposite to that of the CMOC. This is because the WMOC is ice-limited and there are not enough ice particles to collide with liquid particles when INP numbers are low; therefore, increasing the INPs boosts ice particles and allows more riming to occur. In contrast, the CMOC case is liquid-limited; thus, increasing the INPs reduces liquid particles available for riming due to ice depositional growth. We also see that condensation and evaporation rates are generally more than 2 times larger in this case compared with the CMOC, and both rates increase more significantly with CCN concentration in this WMOC. This is related to the dominance of liquid clouds in the WMOC. The more significant increase in condensation from increasing the CCN compared with the CMOC case is likely a result of the more significant change in the local circulation that is associated with a larger number of shallow clouds forming in the valley. Increasing the INP number concentrations reduces evaporation simply because of the reduction of liquid due to increased deposition and riming.
Similar to the CMOC, increasing CCN enhances the WBF process for this
WMOC since more droplet evaporation and ice deposition occur (Fig. 17a and b).
With the increase in CCN, the domain-mean riming rate is not changed much
until CCN of 1000 cm
Results regarding the CCN and INP impact on supercooled water content in the WMOC case are similar to those in the CMOC case: increasing the INPs dramatically reduces SCW and increases cloud glaciation temperature, while increasing CCN has the opposite effect with much smaller significance (Fig. 12b). Compared with the CMOC, the effects of INPs on SCW are a little smaller but CCN effects are a little larger. The liquid-phase fraction (number fraction of cloudy grid points for which the liquid represents 99 % or more of the condensate mass) decreases significantly as INPs increase (Fig. 13b). Correspondingly, the fractions of the mixed-phase and ice-phase cloud volumes increase due to increased ice nucleation. Similar to the increased riming as INPs increase, the mixed-phase fraction increases in the WMOC as well, which is opposite to the case for the CMOC, resulting from the ice-limited conditions in the WMOC versus the liquid-limited conditions in the CMOC. Note that INP effects are more significant at higher INP concentrations in this case, while in the CMOC the sensitivity decreases as INP increases, suggesting that the optimal INP concentration for the maximum INP impact is higher in warmer clouds than colder clouds because ice formation at warmer cloud temperatures is less efficient. The CCN impacts on cloud phase are more significant in this WMOC compared with those in the CMOC. The decreased liquid cloud fraction with the increase in CCN is a consequence of the large increase in ice-phase fraction resulting from more active cold–cloud processes since the total cloud fraction sums up to 1 (Fig. 13b).
Extending the previous study of Fan et al. (2014), we conducted new
simulations at a higher resolution and further sensitivity studies based on the
same two cases of mixed-phase orographic clouds forming on the Sierra Nevada
barrier under the influence of atmospheric rivers during the CalWater 2011
field campaign to quantify the response of precipitation to changes in CCN
and INPs and to examine CCN and INP impacts on SCW and cloud phases. The two
mixed-phase cloud cases have contrasting thermodynamics and dynamics: FEB16
has cold cloud temperatures and northwesterly wind flow at lower levels
(i.e., the CMOC), while MAR02 has cloud temperatures that are about
10
It is found that in the CMOC case deposition contributes more significantly
to snow production than riming because deposition process is efficient at the
cold cloud temperatures (from
We find that increasing the INP concentrations enhances snow precipitation on the windward slope of the Sierra Nevada in both the CMOC and WMOC cases. With the increase in INPs, the increased ice nucleation via immersion freezing enhances snow formation by intensifying depositional growth of ice in the CMOC, while both deposition and riming contribute in the WMOC. Increasing the INPs reduces riming in the CMOC because of the liquid-limited condition in which more efficient depositional growth at higher INP number concentrations glaciates clouds and reduces liquid particles available for riming. However, in the ice-limited conditions of the WMOC, increasing the INPs boosts ice particle concentrations so that more riming can occur in a liquid-rich condition. For the same reason, increasing the INPs suppresses the WBF processes due to reduced liquid particles.
The CCN impacts on precipitation are complicated, depending on cloud
temperature and concentrations of CCN and INPs. When CCN are lower than
1000 cm
Increasing INP concentrations dramatically reduces supercooled water content and increases cloud glaciation temperature, while increasing CCN has the opposite effect but with much smaller significance. As expected, the fraction of liquid-phase clouds is decreased and the ice-phase fraction is increased by increasing the INPs in both cases. However, we see a decreased fraction of mixed-phase clouds from INP in the CMOC but increased in the WMOC, relating to the liquid-limited condition in the former where increasing ice formation enhances cloud glaciation. Conversely, in the ice-limited condition in the latter, more liquid clouds are converted to mixed-phase clouds as INPs increase. Compared with the effects of INPs, the magnitude of CCN effects on SCW and cloud phases is much smaller and the signs are opposite. Increasing CCN generally enhances SCW in both cases. The relative fractions of cloud phases are not much impacted by CCN in the CMOC, except when INPs are very low (i.e., IN0.1). However, in the WMOC, increasing CCN evidently decreases liquid cloud fraction but increases ice-phase fraction. Thus, cloud phases in the WMOC have a large sensitivity to CCN compared with CMOC.
This study provides a better understanding of the CCN and INP effects on
orographic mixed-phase cloud properties and precipitation. The result that
high CCN dramatically increase snow precipitation over the mountains (l000 cm
The mechanism leading to the enhanced precipitation over the windward slope
by increasing CCN is seen in the two cases with very different cloud
temperature, wind direction, and RH. However, the efficiency of the mechanism
could depend on dynamics (wind speed) and thermodynamics (RH). As examined in
Lynn et al. (2007), aerosol impact on orographic precipitation is reduced
when RH is very high and increased as wind speed is reduced. Over the region
of Sierra Nevada, CCN of above 1000 cm
It should be noted that our results for CCN and INP impacts on precipitation and supercooled water content may represent an upper limit since the major ice nucleation in the simulations is through immersion freezing that converts the largest liquid drops into ice or snow directly when ice nucleation occurs. This leads to very efficient conversion of liquid to ice and/or snow and then strong ice growth processes to form snow.
In our study, we do not see significant spillover effect of snowfall (i.e., decrease on the windward slope and increase on the leeward side slope by increasing CCN) as found in Lynn et al. (2007) and Saleeby et al. (2011). Precipitation mainly forms on the windward slope of the Sierra Nevada and the increase in the snow precipitation is more significant on the windward slope than on the leeward side in both cases. The differing results between our study and Saleeby et al. (2011) could be related to different locations of the clouds over the mountain and/or different mountain topography or the presence of a low-level barrier jet in the atmospheric river environment that reduces the cross barrier flow.
All the model simulation data in this paper are deposited in the PNNL Institutional Computing resources and can be accessed by contacting jiwen.fan@pnnl.gov.
This study was supported by the California Energy Commission (CEC) and the Office of Science of the US Department of Energy as part of the Regional and Global Climate Modeling program. PNNL is operated for DOE by Battelle Memorial Institute under Contract DE-AC06-76RLO1830. Paul DeMott additionally acknowledges partial support from the US Department of Energy's Atmospheric System Research, an Office of Science, Office of Biological and Environmental Research program, under Grant no. DE-SC0014354. Edited by: R. Krejci Reviewed by: A. Khain and two anonymous referees