Wintertime cloud cover in the troposphere over the Arctic ranges between 40 to 50% (Wyser et al., 2008), and low-level mixed phase clouds and optically thin ice clouds dominate. Thin ice clouds are difficult to detect by passive imagery and this results in a substantial underestimation of cloud cover (Curry et al., 1996; Wyser and Jones, 2005; Karlsson and Dybbroe, 2010). Grenier et al. (2009) have shown that there are two types of optically thin ice clouds. The first one is formed by a large concentration of small ice crystals and the second one is formed by larger precipitating ice crystals in smaller concentration. The latter cloud type is correlated with aerosol concentration (Grenier et al., 2009). These results suggest that there might be a relationship between the second type of thin ice clouds and large aerosol concentrations possibly of anthropogenic sources.
The Arctic is highly polluted during the cold season with high concentrations of aerosols often observed (Schnell, 1984; Yli-Tuomi et al., 2003; Law and Stohl, 2007). These aerosols, which are mainly emitted over northern European cities, China and Siberia (Shaw, 1995), are transported from the mid-latitudes to the Arctic by large-scale atmospheric circulations (Barrie et al., 1989; Shaw, 1995; Bourgeois and Bey, 2011; Fisher et al., 2011). This transport is favoured by the southward progression of the polar front during winter in the Northern Hemisphere. Anthropogenic aerosols are emitted north of the front in an environment with little precipitation, resulting in limited loss of aerosols by wet deposition. The aerosol component consists of a significant fraction of highly acidic sulphate, and previous work has shown that most of the submicron aerosol particles are coated with this highly acidic sulphate (Bigg, 1980; Cantrell et al., 1997).
Laboratory experiments and field observations suggest that acid coatings on ice nuclei (IN) can have an important effect on heterogeneous ice nucleation, which can occur either in the deposition, condensation-freezing, immersion and contact modes (Pruppacher and Klett, 1997). In the immersion mode, ice nucleation occurs on a solid particle immersed in either an aqueous solution in subsaturated air with respect to liquid water or in an activated cloud water droplet. In the condensation-freezing mode, ice nucleation occurs on a solid particle immersed in an aqueous solution and above liquid water saturation. In the deposition mode, ice nucleation occurs on a solid particle or a solid particle only partially immersed in an aqueous solution. Finally, in the contact mode, ice nucleation occurs on a solid particle in contact with a water droplet.
Archuleta et al. (2005) have shown that the decrease of ice nucleation at temperatures below − 40 °C by immersion and condensation modes due to sulphuric acid coating is variable and depends on the IN chemical composition. Other laboratory experiments performed at temperatures ranging between − 10 and − 40 °C also show that the heterogeneous freezing temperature initiated by immersion of various mineral dust particles decreases as the percentage by weight of sulphuric acid in the particle increases (Ettner et al., 2004). More recently, Eastwood et al. (2009) have shown that deposition ice nucleation on kaolinite particles is considerably altered at temperatures below 243 K, requiring an additional 30% ice supersaturation for ice nucleation to occur when compared to uncoated particles. According to Sullivan et al. (2010a), the de-activation effect of sulphuric acid on dust particle is irreversible and is still active once the acid has been neutralized with ammonia. Chernoff and Bertram (2010) have shown that other coatings such as ammonium bisulphate also increases the onset relative humidity for ice nucleation when compared to uncoated dust particles but to a lesser extent than sulphuric acid coating. Other laboratory experiments on coated and uncoated mineral dust particles have been performed by Knopf and Koop (2006), Salam et al. (2007), Cziczo et al. (2009), Niedermeier et al. (2010) and Sullivan et al. (2010b). The coating effect on ice nucleation was also indirectly observed during the Arctic Gas and Aerosol Sampling Program (AGASP) by Borys (1989). In this study, the authors showed that the IN concentration is decreased by 3 to 4 orders of magnitude during highly polluted (Arctic haze) events (Borys, 1989). While deposition ice nucleation seems strongly affected by acid coating, immersion freezing of activated cloud water droplets does not seem to be affected (Sullivan et al., 2010b).
Blanchet and Girard (1994, 1995) were the first to hypothesize that the ice nucleation inhibition effect of acid coatings on aerosols can have an important effect on cloud microstructure and on the surface energy budget during winter in the Arctic. According to their hypothesis, the decrease of the IN concentration leads to the formation of fewer but larger ice crystals. This process leads to the formation of optically thin ice clouds recently identified by Grenier et al. (2009). Acid coating on IN and the resulting larger ice crystals dehydrate the troposphere by increasing the precipitation over large areas. This results in the decrease of the greenhouse effect due to the strong effect of water vapour, primarily in the broad rotational band in the far infrared. The reduced greenhouse effect further promotes dehydration and cooling (Blanchet and Girard, 1995; Curry et al., 1995). This hypothesis is referred to as the dehydration-greenhouse feedback.
This process was first evaluated by Girard et al. (2005) using a single-column model. They simulated four cold seasons over Alert (Canada) using observed aerosol composition and concentration. The acid aerosol scenario shows a tropospheric cooling ranging between 0 and 2 K when compared to an uncoated aerosol scenario. Girard and Stefanof (2007) have used a regional climate model with prognostic aerosols to evaluate the effect of acid aerosols on the Arctic surface radiative budget for February 1990. They have assumed two aerosol scenarios: (1) an acid scenario in which the reduction of the IN concentration depends on the sulphate concentration and (2) a natural aerosol scenario in which the IN concentration is unaltered. Their results show an increase of precipitation and dehydration of the troposphere and up to 3 K in cooling in the coldest part of the Arctic.
In these previous investigations, the acid aerosol scenario was treated rather subjectively and crudely because few laboratory studies on the effects of coating on ice nucleation were available. The decrease of IN concentration due to acid coating was a function of the sulphate concentration using an exponential function and was based on Borys (1989) observations of IN decrease in polluted events. The choice of an exponential function was made to produce substantial IN reduction for cases where the sulphate concentration was large. Furthermore, it was assumed that all ice nucleation modes were altered equally by the presence of acid coatings. Finally, the one-moment microphysics scheme used in the Girard and Stefanof (2007) investigation was relatively simple with no detailed representation of the ice crystal sedimentation process.
This research aims to refine the representation of the effects of acid coating on IN to get a more realistic evaluation of the potential effect of anthropogenic aerosols on arctic wintertime clouds and radiation budgets. Laboratory data from Eastwood et al. (2008, 2009) on ice nucleation properties of uncoated and sulphuric acid-coated kaolinite particles are used to develop a more physically based and refined parameterization of ice nucleation in subsaturated air with respect to liquid water. This new treatment of heterogeneous ice nucleation is implemented into an elaborate two-moment cloud microphysics scheme. In the second and third sections of this manuscript, the model is described with an emphasis on the new parameterization. In the fourth section, results of the simulations showing the effect of acid coatings on clouds and surface radiation budget for the months of January and February 2007 are described and analyzed. Section 5 presents a brief summary and a discussion of the results.
2. Model description
The limited-area version of the Global Multiscale Environmental Model (GEM-LAM) is used in this study. The numerical formulation of the model is described by Côté et al. (1998). The radiation scheme is from Li and Barker (2005) and is based on the correlation-k method with nine bands in the longwave frequencies and three bands in the shortwave frequencies. Emission and absorption of the following gaseous species are accounted for: H2O, CO2, O3, N2O, CH4, CFC11-14. The land-surface scheme ISBA (Interactions Soil-Biosphere-Atmosphere) developed by Noilhan and Planton (1989) is used to determine the lower boundary conditions for the vertical diffusion of temperature, moisture, and momentum, as well as evaluating the evolution of 10 prognostic variables: the surface temperature, the mean soil temperature, the near-surface soil moisture, the liquid and frozen bulk soil water content, the liquid water retained on the foliage of the vegetation canopy, the equivalent water content of the snow reservoir, the liquid water retained in the snow pack, the snow albedo, and the relative snow density.
The microphysics scheme used in this study is from Milbrandt and Yau (2005). Two versions of this scheme are available in GEM-LAM (single and double moments). In this study, we use the two-moment version. The scheme includes the following prognostic variables: the mixing ratio and the number concentration of cloud liquid water, cloud ice water, rain, snow, hail and graupel. The description of the various microphysical processes is available in Milbrandt and Yau (2005). In this section, the emphasis is put on the parameterization of ice crystal nucleation due to the importance of this microphysical process in this investigation.
Homogeneous freezing of cloud liquid droplets is based on the parameterization of DeMott et al. (1994) in which the ice nucleation rate is a polynomial function of temperature. The fraction of cloud droplets that freeze homogeneously gradually increases from 0 at − 30 to 1 at − 50 °C. Therefore, both homogeneous and heterogeneous freezing processes can occur simultaneously in this temperature range. Contact freezing is parameterized following Young (1974), in which the number concentration of contact IN is parameterized as a function of temperature. Immersion freezing of activated rain and cloud water droplets follows the parameterization of Bigg (1953). The representation of heterogeneous ice nucleation by water vapour deposition and condensation freezing is particularly important in this investigation. In the original version of the Milbrandt and Yau (2005) microphysics scheme, deposition and condensation freezing depends on the ice supersaturation following the empirical relationship of Meyers et al. (1992). This parameterization for deposition ice nucleation has been modified to distinguish ice nucleation on sulphuric acid-coated IN from ice nucleation on uncoated IN.
The new parameterization for ice nucleation in subsaturated air with respect to liquid water is based on the classical theory of heterogeneous ice nucleation of Fletcher (1962). The new parameterization can represent both deposition nucleation on uncoated IN and immersion freezing of haze droplets in subsaturated air with respect to liquid water, which will be referred to as the deposition-immersion nucleation mode in this study. It is assumed that the surface of the IN is energetically uniform for ice nucleation. The only additional unknown parameter is the contact angle (Θ) between the ice embryo and the IN. Following the single contact angle approach (Hung et al., 2003; Chen et al., 2008; Eastwood et al., 2008, 2009; Fornea et al., 2009; Chernoff and Bertram, 2010), the contact angle has been derived using the results of the laboratory experiments of Eastwood et al. (2008, 2009) for uncoated kaolinite particles (Θuncoated = 12°) and for kaolinite particles coated with sulphuric acid (Θcoated = 27°). The following equation is then used to determine the number of ice crystals (Nice) nucleated in a given time step (Δt):
where Akaolinite is the surface area of the kaolinite particles, Nkaolinite is the total concentration of kaolinite particles and J is the nucleation rate of ice embryo per unit area of the particle and is defined as:
where B is the pre-exponential factor (Pruppacher and Klett, 1998), ΔG* is the critical Gibbs free energy for the formation of an ice embryo, k is the Boltzman constant, σiv is the surface tension between ice and water vapour, ρi is the bulk ice density, Rv is the gas constant for water vapour, T is the temperature and Si is the saturation ratio with respect to ice. f(cos Θ) is a function that depends on the contact angle as defined by Pruppacher and Klett (1998) for an infinite plane surface.
Recent studies have shown that more refined models than the single contact angle model used in this study better reproduce the laboratory experiments for dust particles of different sizes (Marcolli et al., 2007; Niedermeier et al., 2010; Vali, 2010; Wheeler and Bertram, 2011). Contact angle probability density function and active sites models (both stochastic and deterministic) have been proposed (Marcolli et al., 2007; Connolly et al., 2009; Lüönd et al., 2010). In our investigation, the simpler single contact angle model is used. This model was recently evaluated against in situ measurements taken during the Surface Heat and Energy Budget of the Arctic (SHEBA) field experiment. Du et al. (2011) have compared the deposition-immersion ice nucleation parameterization used in this study with the empirical approach of Meyers et al. (1992). They have shown that, as opposed to the empirical approach of Meyers et al. (1992), the new parameterization of deposition-immersion ice nucleation described above reproduces quite well the annual cycle of the cloud thermodynamic phase, the downwelling longwave radiation and the cloud radiative forcing (CRF) at the surface during the cold part of the year. These results suggest that the model is able to simulate reasonably well wintertime cloud and radiation processes in an Arctic environment, which is of prime importance in this investigation. Further work will be needed to test the sensitivity of the results to other models to parameterize laboratory results.
3. Design of the experiment
The months of January and February 2007 are simulated. This choice was motivated by the fact that the transport of aerosols was particularly effective during this time period with the presence of a series of uncommon strong extra tropical storms over the North Atlantic and Northern Europe (Fink et al., 2009). These storms have contributed to enhance the transport of pollution emitted over Northern Europe and Siberia to the Arctic. Two aerosol scenarios are considered. In the first scenario (hereafter aerosol scenario A), it is assumed that dust particles are uncoated. In the second aerosol scenario (hereafter aerosol scenario B), dust particles are coated with sulphuric acid. The appropriate contact angle for deposition-immersion ice nucleation (see previous section) is used in each aerosol scenario. Note that the same parameterizations for contact freezing and immersion freezing of activated cloud water droplets are used for both aerosol scenarios. Since the model does not simulate explicitly the aerosol composition and concentration, it is assumed that the concentration of dust particles is constant in time and space with a value of 0.38 cm−3. This value is based on observations taken during field experiments in the Arctic, which shows variable dust mass concentrations ranging between 50 and 3000 ng m−3 (Winchester et al., 1984; Franzén et al., 1994) depending on the air mass origin. The assumed number concentration of dust particles (Nkaolinite) and surface area of kaolinite particles (Akaolinite) in our simulations corresponds to a mass concentration of 500 ng m−3 and a diameter of 0.5 µm. The chosen concentration is representative of the dust concentration over the Arctic during winter (Quinn et al.1996). The same experiment is repeated using a dust concentration reduced by a factor 2 to test the sensitivity of the results to the prescribed dust concentration. Aerosol scenarios A2 and B2 will refer to aerosol scenarios A and B, respectively, with a reduced dust concentration (results presented in Section 4.4.).
Two main assumptions in this investigation can maximize the effects that sulfuric acid coating on dust particles can have on the Arctic clouds and radiation. Our parameterization of deposition-immersion ice nucleation is based on kaolinite particles, which is one of the most effective IN in the atmosphere. Although similar results were obtained for other mineral dust particles coated with sulphuric acid (Cziczo et al., 2009; Eastwood et al., 2009; Chernoff and Bertram, 2010), there are still several IN of different chemical compositions that have not been tested in laboratory yet. These IN of different chemical compositions could very well behave differently than kaolinite particles when they are coated with sulphuric acid. It should also be noted that the assumed constant dust concentration in time and space is a simplification that was made to avoid unjustified spatial and temporal distribution of dust in our simulation. Dust and other aerosols are often observed in thin discrete layers in the low, mid and upper troposphere and can vary in the horizontal according to the large-scale atmospheric circulation. It is therefore hazardous to assume a given spatial distribution of dust particles without either coupling GEM with an aerosol transport model or perform a comprehensive sensitivity study on different aerosol spatial distributions. Therefore, the effect of acid coating on dust particles on the Arctic clouds and radiation budget obtained in this study should be viewed as an upper limit given the above-mentioned assumptions.
The model internal variability for the Arctic climate is relatively high during winter (Rinke et al., 2004; Girard and Bekcic, 2005). Hence, a large number of simulations are required to distinguish the investigated climate signal from the model variability. Each simulation within an ensemble is initialized with different conditions following Rinke and Dethloff (2000). In our experiments, 10 simulations of each aerosol scenario were necessary to get reasonable statistics over a large area of the Arctic. Unless specified otherwise, results presented in this paper always represent the January and February (JF) ensemble mean of either aerosol scenario A or B and shadowed areas indicate that results are statistically significant with a confidence level of 95%. The statistical test is described in Girard and Bekcic (2005).
The integration domain is centred over the Arctic and covers all areas north of 50°N, which include most of Europe, North Asia, Northern Canada, Siberia, and the North Atlantic and Pacific Oceans. The simulation domain has 364 by 304 grid points with a horizontal resolution of 0.25°. There are 53 vertical levels with the highest resolution in the lower troposphere. Initial and boundary conditions for atmospheric fields are provided by the European Centre for Medium-range Weather Forecast (ECMWF) ERA-Interim reanalysis on a 2.5° by 2.5° longitude/latitude grid. Monthly mean sea surface temperature and sea ice concentration are from the Atmospheric Model Intercomparison Project (AMIP II) (Hurrell et al., 2008) reanalysis on a 1° by 1° grid. These values are interpolated on the grid used for our simulations.
Figure 1 shows the averaged mean sea level pressure for January and February 2007 from the ERA-Interim reanalysis. The typical atmospheric circulation associated with a positive North Atlantic Oscillation index with an intense low-pressure system located in the North Atlantic region and an anticyclone over Siberia characterizes this period. In January and February 2007, several strong storms have passed over Europe causing strong winds and a lot of precipitation. The large-scale circulation pattern was therefore very favourable to the transport of anthropogenic sulphate emitted over Eurasia and dust originating from East Asian deserts to the Central Arctic.
Figure 2 shows the JF mean temperature at 850 hPa and the geopotential height at 500 hPa from ERA-Interim reanalysis and model simulations. Simulations used for this comparison are the ensemble averaged of the aerosol scenario B (acid-coated aerosol scenario). The model reproduces reasonably well the geopotential height at 500 hPa with differences between ERA-Interim and the simulation varying between − 4 and + 4 decameters (dam). A larger difference of + 6 to + 8 dam occurs over Siberia. Temperature at 850 hPa is also generally well reproduced by the model with errors of less than 2 K except over Siberia with an overestimation between 2 and 4 K, which is consistent with the geopotential height bias over the same area. These errors are acceptable in the context of this sensitivity study and compare advantageously with other limited-area model simulations of the Arctic climate (Rinke et al., 2006; Wyser et al., 2008).
4.1. Cloud microstructure
Figure 3 shows the JF mean liquid and ice water path anomaly (an anomaly is defined as the difference between aerosol scenarios B and A). Over the Arctic, the ice water path is generally slightly smaller in aerosol scenario B compared to aerosol scenario A except for small isolated areas where it is larger. Over the mid-latitudes and sub-Arctic regions, the ice water path anomalies are also negative and larger (absolute values). The liquid water path difference between aerosol scenarios B and A is positive over the entire domain. The vertical profile of the ice water content and liquid water content spatially averaged over a sub-domain delimited by sea ice and open water boundary and temporally averaged over JF is shown in Figure 4. The increase of the liquid water content is the largest in the lower troposphere at about 850 hPa. This is the height corresponding to the highest occurrence of mixed-phase clouds (discussed at the end of this section). However, the increase of the liquid water path is also positive higher in the troposphere up to 600 hPa. The decrease of the ice water content in aerosol scenario B spreads over the whole troposphere from the surface to 300 hPa with a maximum at 500 hPa.
Figure 5 shows the spatial and temporal average of the vertical profile of the ice crystal number concentration and mean diameter. Below 500 hPa, the ice crystal number concentration is smaller and the ice crystal mean diameter is larger in aerosol scenario B. This result is consistent with the hypothesis formulated in the previous section. However, above 500 hPa, the ice crystal number concentration increases and the mean diameter of ice crystals decreases in aerosol scenario B.
The different effects of acid coating on cloud microstructure above and below 500 hPa is related to temperature. At the upper levels, above 500 hPa, temperatures are generally below − 40 °C. At these temperatures, the homogeneous freezing rate of cloud water droplets is relatively large. In these conditions, the decrease of ice nucleation in the deposition-immersion mode in aerosol scenario B leads to the formation of more water droplets. A large fraction of them freeze homogeneously as soon as they are nucleated. This results in an increase of the ice crystal concentration and a smaller mean ice crystal diameter in aerosol scenario B. In aerosol scenario A, the nucleation rate of ice crystals in the deposition mode prevents the saturation ratio to reach the saturation point with respect to liquid water as often as in aerosol scenario B. The concentration of ice crystals remains smaller when compared to aerosol scenario B as the contribution of homogeneous freezing of water droplets is much smaller in this scenario.
Below 500 hPa, the JF mean air temperature remains mostly above − 40 °C over much of the Central Arctic. The homogeneous freezing of water droplets is not dominant at these temperatures. The Central Arctic is characterized by the presence of either anticyclones or cold decaying low pressure systems in which very weak ascents or subsidence of air prevail. In such an environment, deposition ice nucleation can be significant as the air mass slowly cools by infrared radiation and thus can stay long periods of time oversaturated with respect to ice but subsaturated with respect to liquid water. In aerosol scenario A, the nucleation rate of ice crystals by water vapour deposition is larger than in aerosol scenario B. Ice crystals are therefore smaller compared to aerosol scenario B as more ice crystals can absorb the available water vapour.
In aerosol scenario B, the heterogeneous ice nucleation rate is much smaller when the atmosphere is subsaturated with respect to liquid water. This results in a smaller concentration of larger ice crystals as shown in Figure 5. The reduced concentration of ice crystals in this scenario has also a consequence on the mean relative humidity with respect to ice (RHi). The total flux of water vapour onto the existing ice crystals is smaller, thus allowing the relative humidity to increase. Figure 6 shows the JF mean relative humidity with respect to ice (RHi) anomaly at 850 hPa and the spatially and temporally averaged RHi vertical profile within the sea ice boundary mask. The RHi increase in aerosol scenario B is the largest over the Arctic, Siberia and Northern Canada where the temperatures are the coldest. The JF RHi anomaly at 850 hPa reaches values up to 14% in the Canadian and Central Arctic. It is noteworthy to mention that the positive RHi anomaly covers more or less the Arctic air mass, characterized by the persistence of cold decaying lows and anticyclones. This suggests that weak cooling rate associated to either weak air ascent and/or infrared radiative cooling is a necessary condition for acid coating on IN to have an effect on cloud microstructure through RHi increase. At lower latitudes, synoptic systems are more active and air ascent velocity is larger. In these conditions, the air cooling rate is much larger when compared to the Arctic air mass. Consequently, ice nucleation in subsaturated air with respect to liquid water is not as important since the ice crystal concentrations are typically not large enough to deplete the available water vapour. Therefore, in both aerosol scenarios, large RHi values are reached and the cloud microstructure is similar. This explains why the RHi anomalies are close to 0 south of the Arctic front.
Figure 7 shows the JF mean frequency of mixed-phase clouds for both aerosol scenarios at 850 hPa and the difference between scenarios B and A (the anomaly). In both aerosol scenarios, the frequency of mixed-phase clouds is very large over Northern Europe and Northeastern Asia, where the Icelandic and Aleutian lows respectively are predominant. These baroclinic zones are characterized by the development of low-pressure systems with strong vertical ascents. Therefore, the effect of deposition ice nucleation is negligible. This is why the differences between both aerosol scenarios over these two areas are relatively small. However, differences are much larger in the Arctic air mass, which corresponds to the region where the RHi increase in aerosol scenario B is the largest. In aerosol scenario A, the frequency of mixed-phase clouds in the Arctic air mass varies between 0 and 20% compared to a frequency ranging between 30 and 50% for aerosol scenario B. The mixed-phase cloud frequency anomaly is positive at all levels between 600 hPa and the surface (not shown).
Figure 8 shows the January–February mean precipitation rate relative anomaly. The relative anomaly is defined as the ratio of the difference of precipitation in aerosol scenarios B and A on the precipitation rate of aerosol scenario A. Precipitation is increased over much of the Central Arctic in aerosol scenario B with relative anomalies ranging between 20 and 60%. Negative relative anomalies within the Arctic air mass are mostly not statistically significant. Bigger ice crystals have larger terminal velocity and precipitate more efficiently. The increase of precipitation can be explained by the increased frequency of mixed-phase clouds in aerosol scenario B. Relatively few ice crystals are nucleated in the thin liquid layer and rapidly grow by the Wegener-Bergeron-Findeisen effect. Precipitating ice crystals also come from upper layers. In aerosol scenario B, larger ice crystals precipitate more efficiently and can seed the mixed-phase clouds near the surface, thus favouring precipitation to the surface.
4.2. Radiation and temperature
The energy budget at the surface and at the top of the atmosphere (TOA) is mostly driven by terrestrial radiation owing to the quasi-absence of solar radiation during January and February. The net infrared radiation at the surface or at the TOA strongly depends on the presence of clouds and their microphysical properties. According to Shupe and Intrieri (2004), the downward longwave radiation at the surface can increase by as much as 40 W m−2 if liquid water is present in low-level arctic clouds compared to an ice cloud. Modelling studies have also shown a large sensitivity of infrared radiation fluxes to cloud microstructure (Du et al., 2011, Simjanovski et al., 2011). It is therefore expected that changes induced by acid coating on the cloud microphysical properties in aerosol scenario B will have an impact on the energy budget both at the surface and at the TOA.
To estimate the effect of clouds on the radiation budget, CRF was proposed by Ramanathan et al. (1989) to characterize the cloud effect on the net radiation either at the surface or at the TOA. CRF is defined as the difference between the net radiative flux in the presence of clouds and the net radiative flux without the presence of clouds. It can also be separated into its longwave and shortwave components. Figure 9 shows the JF mean CRF anomaly at the TOA. In our analysis, only the longwave component of the CRF is discussed since the shortwave radiation is quasi-absent during winter. In both aerosol scenarios, the CRF is positive with smaller values varying between 0 and 20 W m−2 over the Arctic and higher values of up to 60 W m−2 south of the Arctic air mass (not shown). The CRF anomaly at the TOA is negative over most of the Arctic with values ranging between 0 and − 6 W m−2. Relative to the CRF absolute values of about 20 W m−2, the CRF anomalies represent a substantial reduction of the CRF at the TOA.
Mixed-phase clouds are mostly located in the lower troposphere. Their temperatures are warmer than the surface skin temperature. Therefore, the presence of these clouds leads to an increase of upward longwave radiation emission when compared to the surface. Assuming the same amount of mid and upper level ice clouds, the increase of low-level liquid (or mixed) phase clouds will enhance the CRF at the TOA. Indeed, increasing the mixed-phase cloud emissivity changes the thickness over which the mid and upper ice clouds absorb this energy. Therefore, the amount of longwave radiation going out to space increases because of the fact that the upper ice cloud layer acting as blackbody thickens and thus is warmer. Figure 10(a) shows the variation of the JF mean CRF at the TOA and the cloud liquid water path for aerosol scenarios A and B over a sub-domain delimited by sea ice. As expected, the CRF at the TOA increases as the cloud liquid water path increases in aerosol scenario A. In aerosol scenario B, this is not the case since adding more cloud liquid water does not affect much the low-level mixed-phase cloud emissivity. Clouds with liquid water path values above 10 to 15 g m−2 emit as a blackbody (Shupe and Intrieri, 2004; Du et al., 2011). Increasing the cloud liquid water path above these values does not increase significantly cloud emissivity. As a result, the CRF at the TOA does not depend anymore of the cloud liquid water path. This can be seen in Figure 10(a) for aerosol scenario B in which cloud liquid path values are above 12 g m−2.
Figure 10(b) shows that the CRF at the TOA increases as the cloud ice water path increases in both aerosol scenarios. This result was expected since as the optical thickness of mid and upper ice clouds increases, less infrared radiation can escape to space. Indeed, increasing cloud ice water path reduces the thickness of blackbody layers and therefore decreases the amount of energy that escapes out to space.
In aerosol scenario B, the increased frequency and optical thickness of mixed-phase clouds combined to the reduced cloud optical thickness in the mid and high troposphere contribute to increase the upward infrared radiation fluxes at the TOA. Figure 10(c) shows that the CRF anomaly at the TOA decreases when the cloud ice water path anomaly decreases. Such a relationship between the CRF at the TOA anomaly and the cloud liquid water path anomaly is not seen because the saturation point in mixed-phase cloud emissivity is reached in aerosol scenario B (Figure 10(d)). Therefore, at these cloud liquid water path values, whatever the increase of the low-level cloud liquid water path, the CRF at the TOA will essentially depends on the mid and upper cloud ice water path.
Figure 11 shows the JF mean air temperature anomaly at 1000, 850 and 500 hPa. The Central Arctic is colder in aerosol scenario B with a decrease of the temperature by up to − 3 K near the surface. This cooling, which spreads over much of the Arctic air mass, is also obtained higher in the troposphere with values ranging from − 2 to − 4 K at 850 and 500 hPa. Enhanced atmospheric cooling in aerosol scenario B also leads to a decrease of the vertically integrated water vapour (not shown) owing to the decrease of the saturated water vapour pressure with temperature. This also contributes to the cooling of the surface by a decrease of the water vapour greenhouse effect.
Figure 12 shows the downward longwave radiation anomaly at the surface. Not surprisingly, the anomaly is negative over much of the cooling area in aerosol scenario B with values ranging between 0 and − 4 W m−2. However, it is important to note that the anomaly of the downwelling infrared radiation at the surface is negative despite the increase of mixed-phase cloud frequency, which produces a positive anomaly of the CRF at the surface of up to 10 W m−2 (not shown). Therefore, the cooling near the surface is not directly related to changes in cloud microstructure in aerosol scenario B but rather to the atmospheric dehydration and cooling of the mid and upper atmosphere, which leads to a decrease of the water vapour greenhouse effect.
4.3. Atmospheric circulation
Figure 13 shows the mean-sea level pressure (MSLP) anomaly. The Icelandic low is deeper in aerosol scenario B and the Siberian anticyclone is strengthened. The pressure gradient is thus significantly increased by 4 to 5 hPa over the Northern part of Eurasia and Canada. The strengthening of the atmospheric circulation is due to the enhanced meridional temperature gradient in aerosol scenario B. This favours the formation of more intense synoptic storms, which is reflected in the JF MSLP anomaly. The strengthening of the pressure gradient over Eurasia has also important consequences for aerosol transport from the mid-latitudes to the Arctic. Indeed, Eurasia is the main pathway by which aerosols are transported to the Arctic. In our simulations, aerosols are not explicitly represented. Therefore, the increase in the aerosol concentration over the Arctic and possible feedbacks with clouds cannot be taken into account.
4.4. Sensitivity of the results to the concentration of dust particles
Dust concentrations are highly variable and depend on the large-scale atmospheric circulation. In this section, results obtained with a prescribed dust concentration reduced by a factor 2 are presented (hereafter aerosol scenarios A2 and B2).
Table I shows a comparison between the reference and reduced dust concentration for the JF averaged absolute values (aerosol scenarios B and B2) and anomalies of temperature and relative humidity with respect to ice at 1000, 850 and 500 hPa, liquid and ice water path, and CRF. These variables are spatially averaged over a mask delimited by sea ice. In general, temperatures, relative humidity with respect to ice, CRF and cloud water path are very similar in both acid-coated aerosol scenarios (B and B2). Reducing the dust concentration has implications on the available IN concentration at a given temperature and ice supersaturation. Aerosol scenarios A and A2 (the reference scenarios from which anomalies are calculated) give different results (not shown) since according to Equation (1), the nucleation rate of ice crystals is proportional to the dust concentration. At the same time, the acid coating effect on ice crystal nucleation rate is also proportional to the dust concentration. From Equation (1), we have:
where Δ refers to the difference between the coated aerosol scenario and uncoated aerosol scenario, NIC is the number of ice crystals nucleated per unit time, Ndust is the total aerosol concentration and κ is the fraction of dust particles nucleating ice crystals per unit time. The absolute values of the examined variables in aerosol scenarios B and B2 do not differ much as κ is relatively small in both aerosol scenarios. In aerosol scenarios A and A2, κ is larger and therefore, differences in the examined variables are much larger between these aerosol scenarios. Consequently, the anomalies of these variables are different for both aerosol scenarios, aerosol scenario B having the largest anomalies. The magnitude of cloud and radiative changes caused by acid coating on dust particles largely depends on the dust concentration as shown in Table I. The radiative effects of acid coatings on dust increases as the dust concentration increases.
Table I. Spatial (sea ice mask) and temporal averages of some variables and the respective anomalies in aerosol scenarios B and B2
Anomalies Scenario B2
Anomalies Scenario B
Liquid water path (kg m−2)
Ice water path (kg m−2)
Relative humidity w.r.t. ice at 1000 hPa (%)
Relative humidity w.r.t. ice at 850 hPa (%)
Relative humidity w.r.t. ice at 500 hPa (%)
Cloud radiative forcing at TOA (W m−2)
Temperature at 1000 hPa (K)
Temperature at 850 hPa (K)
Temperature at 500 hPa (K)
One can quantify the CRF at the TOA as a function of the dust vertical path using an acid forcing factor (Af) normalized by the dust vertical path as follows:
where F is the cloud forcing at the TOA and D the dust vertical path. Af and b depend on the dust vertical path. With the dust vertical path used in our simulations (1.75 µg m−2 in aerosol scenarios A2 and B2 and 3.50 µg m−2 in aerosol scenarios A and B), it is then possible to derive its variation with D assuming a linear relationship as follows:
By using these values, one can get an approximated value of F for any dust vertical path between 0 and 3.50 µg m−2.
5. Summary and discussion
In this study, the effects of acid-coated IN on the Arctic cloud microstructure and radiation is investigated for January and February 2007. A new parameterization for heterogeneous ice nucleation is implemented into the two-moment microphysics scheme of the Canadian Global Environmental Multiscale (GEM) model. The main objective of this study is to assess the impact of the de-activation effect of IN on wintertime Arctic clouds and energy budget.
Results show that acid coatings on IN have an important effect on ice nucleation with the coatings modifying both ice and mixed-phase cloud microstructures. The primary effect of acid coating on IN is to significantly decrease the nucleation rate at a given ice supersaturation. The consequences of this change are a function of temperature as illustrated in Figure 14.
In the upper part of the troposphere the temperature is often below − 40 °C and homogeneous nucleation is the dominant freezing mechanism. In this case, the inhibition effect of acid coatings on deposition ice nucleation leads to an increased concentration of water droplets, which freeze homogenously. In the uncoated aerosol scenario, larger heterogeneous ice crystal nucleation rates in subsaturated air with respect to liquid water prevents liquid water saturation more often than in the coated aerosol scenario. As a result, the ice crystal concentration is lower and their size is larger when compared to the coated aerosol scenario. Since a very high ice supersaturation is needed to nucleate ice crystals in the acid-coated aerosol scenario, the JF mean ice water content is lower than the uncoated aerosol scenario at these levels.
In the mid and lower part of the atmosphere, heterogeneous ice nucleation dominates over homogeneous freezing. In the uncoated aerosol scenario, larger ice crystal nucleation rates by deposition nucleation more often prevents the atmosphere from becoming saturated with respect to liquid water compared to the coated aerosol scenario. This leads to an increased frequency of mixed-phase clouds in the coated-aerosol scenario with an increase of the liquid water content and a decrease of the ice water content when compared to the uncoated aerosol scenario. This effect, associated with warmer temperatures, peaks at 850 hPa.
These two different effects have a common impact on the infrared radiative budget at the top of the atmosphere. Optically thinner mid to upper ice clouds in the coated aerosol scenario increase the atmospheric transmissivity of terrestrial radiation. At the same time, the upward infrared radiation flux is increased in the coated aerosol scenario due to optically thicker and more frequent mixed-phase clouds when compared to the uncoated aerosol scenario. The end result is a decrease of the CRF at the top of the atmosphere ranging between 0 and − 6 W m−2. This leads to an atmospheric cooling that varies between 0 and − 4 K. The atmospheric cooling further promotes the formation of clouds in the coated aerosol scenario leading to a decrease in the water vapour greenhouse effect and precipitation for January and February. Results show that this Arctic cooling is large enough to strengthen the large-scale tropospheric circulation associated with the polar front through the intensification of the baroclinic zone.
The results obtained in this study show that ice nucleation plays an important role for both mid and upper ice clouds and low-level mixed-phase clouds in the Arctic, which in turn has an effect on radiation and atmospheric circulation. Quite interestingly, acid coatings have little effect on cloud and radiation south of the arctic air mass. This suggests that deposition ice nucleation is important mainly in stable air masses that cool slowly, thus preventing the relative humidity from reaching liquid water saturation rapidly.
Some assumptions made in this study could either amplify or reduce the tropospheric cooling resulting from acid coatings on IN. Firstly, the dust concentration is assumed to be constant both in time and space. Observations show that the aerosol concentration is highly variable in the Arctic during winter (Shaw, 1995). Results are sensitive to the dust concentration. Therefore, this assumption is likely to give an upper-limit value of the tropospheric cooling for a given dust concentration. It has also been assumed that the de-activation effect applies to all atmospheric IN. Although many other dust particles are equally affected by acidic coating, other IN of biological origin are not affected (Chernoff and Bertram, 2010). This assumption may lead to a slight overestimation of the de-activation effect. Finally, ice nucleation in the contact mode is assumed to be un-altered by acid coating. This assumption could contribute to underestimate the cooling obtained in this study to some extent. Hoose et al. (2008) and Storelvmo et al. (2008) in their modeling studies have assumed that acid coating on dust particles inhibits contact ice nucleation. Such an alteration of contact nucleation by acid coating can only enhance the cloud microstructure changes obtained in our study, that is an increase of the cloud liquid water and a decrease of the ice crystal number concentration. The extent to which an alteration of contact ice nucleation could contribute to cloud microstructure changes depends on its relative contribution to the total ice nucleation rate. This aspect remains to be investigated.
Using a refined treatment of the IN de-activation effect based on laboratory experiment, this research has confirmed the results obtained in previous modeling investigations (Girard et al., 2005; Girard and Stefanof, 2007) on the effect of acid coating on the Arctic clouds and radiative budget during winter. This indirect effect of acid aerosols on arctic clouds and the resulting atmospheric cooling could explain, at least in part, the unexpected observed cooling trend of surface air temperature over the Arctic Ocean during winter in the period 1979–1999 as observed by boys (Rigor et al., 2000) and by satellite remote sensing (Wang and Key, 2003).
The authors would like to thank the Canadian Foundation for Climate and Atmospheric Sciences (CFCAS), the Natural Sciences and Engineering Research Council of Canada (NSERC) and the Fonds Québécois de la Recherche sur la Nature et la Technologie (FQRNT) for support funding.