Accurately simulating aerosol and cloud microphysical processes is becoming increasingly important in process- scale, mesoscale, and climate-scale models. Much of our understanding of aerosol and microphysical effects on the water cycle and global climate rely on the use of such models (Solomon et al. 2007). The indirect effects of cloud and ice nucleating aerosols (Twomey 1974; Albrecht 1989), including cloud albedo, cloud lifetime, and precipitation efficiency, remain relatively poorly understood (Solomon et al. 2007). As such, improve- ments in the representation of aerosols and related cloud processes in models is necessary to improve confidence regarding aerosol impacts on climate. One such model being used for aerosol and microphysical research across multiple cloud scales is the
The RAMS model has proven to be versatile across atmospheric scales as a large-eddy-simulation model (e.g., Jiang et al. 2001; Jiang and Feingold 2006), cloud- resolving model (e.g., Saleeby et al. 2009;
RAMS incorporates a two-moment bulk microphysics package that predicts the mixing ratio and number con- centration of cloud droplets, drizzle, rain, pristine ice, snow, aggregates, graupel, and hail (Walko et al. 1995; Meyers et al. 1997; Saleeby and Cotton 2004a). Each hydrometeor species is represented by a gamma distri- bution, given as
(All variables and symbols in this paper are defined in Table 1.) The model microphysics simulates cloud nu- cleation, ice nucleation, vapor deposition, evaporation, collision--coalescence, melting, freezing, sedimentation, and secondary ice production (Verlinde et al. 1990; DeMott et al. 1994; Walko et al. 1995, 2000a; Meyers et al. 1992, 1997; Cotton et al. 2003; Saleeby and Cotton 2004a, 2008). The two-stream radiation model of Harrington (1997) accounts for attenuation and scat- tering of hydrometeors.
Saleeby and Cotton (2004a) implemented a cloud nucleation scheme in RAMS, version 4.3, for two- moment prediction of cloud and drizzle droplets. La- grangian parcel-bin simulations were performed offline with a single-column model (Heymsfield and Sabin 1989; Feingold and Heymsfield 1992) to generate lookup tables that contain the percentage of aerosols that activate over a range of vertical velocity, temper- ature, aerosol concentration, and aerosol size. Studies using this aerosol parameterization have examined the cloud droplet nucleating effects of aerosols over a wide variety of cloud systems, including orographic snowfall (Saleeby et al. 2009, 2011, 2013), shallow clouds (Cheng et al. 2009; Lee et al. 2009), deep con- vection (Lee 2012;
With use of a cloud parcel model, Flossmann et al. (1985, 1987) demonstrated the importance of repre- senting aerosol nucleation scavenging, impaction or precipitation scavenging, and aerosol regeneration upon evaporation of hydrometeors. They showed that nucleation scavenging accounts for the majority of aerosol removal, but that precipitation scavenging of aerosol can be of importance in many situations. Further, representing the regeneration of aerosols can be quite important in evaporative regions, and it can lead to a significant change in the aerosol dis- tribution size, chemistry, and hygroscopicity or solu- bility due to aerosol mixing within drops grown by collision--coalescence. More recent work has been done to include these aerosol sources and sinks into more complex aerosol--cloud microphysics models (Ekman et al. 2004, 2006) and to interface these with limited cloud dynamics models (Wang and Chang 1993) so as to provide a more complete rep- resentation of aerosols and their impacts on complex cloud systems. Such models are then capable of rep- resenting aerosol life cycles and thus more realistically simulate their impacts on clouds. Improvements in the understanding of ice nucleation mechanisms and their representation in numerical models are also of great importance in prediction of ice concentrations, radiation budgets, and frozen precipitation (Fridlind et al. 2007). As such, we have implemented a new scheme for heterogeneous ice nucleation that is based on data from numerous of field studies (DeMott et al. 2010).
The goal of the research presented in this paper is twofold. The first is to present the details of de- velopment completed within the RAMS aerosol module with a view to improving the representation of the key sources and sinks of the aerosol life cycle, how they relate to cloud condensation nuclei and ice nuclei (IN), and their links to other microphysical and dynamical processes. The second is to demonstrate the relative importance of the aerosol source and sink mechanisms across cloud scales via an examination of simulated precipitation within raining stratocumulus clouds, deep convection, and orographic snowfall. This man- uscript documents the individual modules governing the treatment of aerosols from emissions to activa- tion, scavenging, and regeneration and the radiative impacts of aerosols, as well as the importance of such changes on the surface precipitation of stratiform and convective precipitation. The inclusion of these in- dividual aerosol representations will allow for more robust studies of aerosol effects over a range of cloud systems. The aerosol modules in RAMS 6.0 are dis- cussed in the following sections.
2. Developments in the treatment of aerosols
a. Aerosol activation and cloud droplet nucleation
There are nine microphysically active aerosol species: 1) submicrometer sulfate, 2) supermicrometer sulfate, 3) submicrometer mineral dust, 4) supermicrometer min- eral dust, 5) film-mode sea salt, 6) jet drop-mode sea salt, 7) spume-mode sea salt, 8) submicrometer regenerated aerosols, and 9) supermicrometer regenerated aero- sols. Each aerosol spectrum is represented by a log- normal distribution given by
Users are allowed to input initial values of aerosol num- ber concentration and mass or median radius of the dis- tributions for any of the aerosol categories, though the regenerated aerosol category is initialized as zero and obtains mass and number as the model microphysics evolves in a given simulation. Sources of dust and sea salt aerosols are also available for use instead of user- prescribed concentration (sections 2d and 2e).
A Lagrangian parcel bin model was run to simulate activation of a distribution of aerosol particles and sub- sequent nucleation of new cloud droplets. The saturation ratio over an aerosol surface was represented by the Koeurohler equation as
From the Koeurohler equation, the model first iteratively computes the equilibrium droplet diameters. Parcel sim- ulations are then initialized with chosen aerosol spectra and ambient conditions, and air parcels are lifted just beyond the level of maximum saturation while the droplet growth equation determines the binned sizes of nucleated droplets. The fraction of aerosols that result in newly formed cloud droplets (.2.0-mm diameter) is computed and cataloged. The activation fractions ob- tained over a range of aerosol characteristics and en- vironmental conditions were used to develop a set of four-dimensional (4D) cloud droplet nucleation lookup tables that vary with vertical velocity, temperature, and aerosol number concentration and median radius.
Given the newly added aerosol species and the po- tential for large variability in aerosol solubility, we have extended the dimension of the lookup tables to include soluble fraction «. Ward et al. (2010) and Reutter et al. (2009) represented aerosol solubility with the kappa parameter k (Petters and Kreidenweis 2007). These studies demonstrated the sensitivity of aerosol activation to the solubility. Rather than using a single « for all aerosol species, it is most reasonable to allow sea salt aerosols to be nearly fully soluble, while mineral dust may be nearly in- soluble; meanwhile, sulfate-based particles could vary over a range of solubility depending on the aerosol source and degree of sulfate coating. The user assigns the aerosol solubility for each species at the time of model initializa- tion, and the solubility remains constant in time for each aerosol category. Figure 1 displays a plot of the percentage of activated aerosols that lead to new droplet nucleation over a range of solubility « from 5% to 100% and for several values of vertical velocity and aerosol median ra- dius. The solubility has a minimal impact for a situation with 1) strong updrafts and large aerosol median radii (dotted black line) and 2) weak updrafts and small aerosol median radii (solid red line), which represent the upper and lower bounds of the nucleation percentage in the plot. Intermediate combinations of size and updraft lead to larger variations in nucleation with a change in solubility. Permitting variability with « provides a more accurate representation of aerosols with known chemis- try and allows for differentiation among aerosol species.
During the simulated cloud nucleation process, the 4D lookup tables are accessed each time step with the fraction of aerosol number to activate being determined from the five parameters of the lookup tables. For each aerosol species, if the median radius of the distribution is less (greater) than 1 mm at the time of activation, newly nucleated droplets enter the cloud water (drizzle) cate- gory. From parcel model results, larger particles tend to result in larger initial droplets that exceed the upper bound on model cloud droplet diameter (2--50 mm) and fit more closely within the drizzle range (50--100 mm).
b. Nucleation scavenging and aerosol regeneration
For a given supersaturated grid cell, each of the aerosol categories competes for potential activation, assuming that the solubility fraction is greater than zero. Relative competition among aerosol categories is based on the total surface area for the number of potentially activated aerosols in each category. The total surface area is computed for each category and then weighted against the sum total surface area of all aerosols. The amount of vapor above saturation that is available for nucleation of new cloud droplets is then divided among the aerosol categories with each given its respective weighting. From the total distribution aerosol particle mass and number concentration, we can relate the mean mass (volume) radius to the median radius of the log- normal distribution that is used for computing the nu- cleation lookup tables. This relationship is given as
A distribution width of s g 5 1.8 was assumed for the aerosol spectra used in the creation of the nucleation lookup tables (similar to Saleeby and Cotton 2004a). Given this relationship between the mean radius and median radius, the median radius for a certain mass and number of aerosols can then be computed as
The aerosol density is a free parameter to be set by the user but is initially assigned the following default values: sulfate (1.769 g cm23), submicrometer mineral dust (clay 2.500gcm23), supermicrometer mineral dust (silt 2.650 g cm23), and sea salt (2.165 g cm23). By knowing number, mass, and median radii of the various aerosol type categories, the aerosol distributions can be com- puted at the time of new cloud droplet nucleation. The lognormal distributions are partitioned into 100 bins over an optimized size range that is specific for a given median radius. The initial implementation of aerosol removal, via nucleation scavenging in Saleeby and Cotton (2004a), did so from an aerosol distribution of 200 bins that covered a single size range that included all aerosol mass and numbers for median radii from 0.01 to 0.96 mm. This representation of the distribution results in low bin resolution that can sometimes produce discrepancies when trying to match bin mass and number for aerosol removal. In the updated scheme we established discrete distribution bin size ranges for each represented me- dian radius in both the Lagrangian parcel model and the representation of the lognormal distributions used within RAMS. This improves bin resolution and allows us to precisely encompass the aerosol mass and number within this range without having wasted bin space. It also eliminates the need for the 200 size bins and specified distribution truncation initially employed by Saleeby and Cotton (2004a) to handle the discrepancy between the aerosol mass and number distributions. Upon binning the distribution, the aerosol number and mass that participate in nucleation are removed from the environment by subtracting the nucleated number concentration from the large end of the aerosol distri- butions since the larger particles are preferentially acti- vated over those at the small end of the distribution. Following this subtraction, an updated median radius is computed, which will be a smaller value than that prior to nucleation. Since this is a bin-emulating approach, the remaining aerosol number and mass are redistributed according to the lognormal basis function based on the newly computed median radius.
Upon nucleation of new cloud droplets, the aerosol mass that is removed from the aerosol population is transferred to a 3D scalar variable that is used to track the amount of total aerosol mass contained within cloud droplets, called aerosol in cloud.Thismassisaconglom- erate of all aerosol types consumed during nucleation. There is a different 3D scalar tracking variable associated with each hydrometeor type so as to allow transfers of aerosol mass in hydrometeors among hydrometeor cate- gories whenever transfers of condensate mass occur from one hydrometeor category to another. During the con- densate mass transfers that would occur during collision-- coalescence, freezing, or melting, the aerosol mass in hydrometeors is transferred in proportion to the amount of transferred hydrometeor mass. So, for example, if 50% of the hail mass melts in a given grid cell and becomes rain, then 50% of the aerosol mass contained within hail particles, referred to as aerosol mass in hail,willbe transferred to the aerosol mass in rain category. This method of tracking aerosol mass contained within hy- drometeors is similar to the aerosol mass ratio transfer scheme of Rutledge et al. (1986) and Hegg et al. (1986).
When grid cells containing hydrometeors undergo evaporation, a number of precipitation particles may fully evaporate and restore aerosols back to the environment. The amount of restored aerosol mass is in proportion to the amount of mass of fully evaporated hydrometeors relative to the total hydrometeor mixing ratio. If the mass of the fully evaporated hydrometeors in a given hydro- meteor category is 10% of the total hydrometeor mass in that category, then 10% of the tracked aerosol mass within that hydrometeor species is restored to one of the regenerated aerosol categories. The number of restored aerosols equals the number of fully evaporated hydro- meteors. From the restored aerosol mass and number, Eq. (5) is used to compute the median radius of the re- generated aerosols. If the median radius is less than (greater than) 1 mm, then the aerosols are returned to the submicrometer (supermicrometer) regenerated aerosol category. By not restoring aerosols back to a parent cat- egory, we can examine the spatial and temporal changes in the initial aerosol categories. It should be noted that the regenerated aerosols must be given a constant solubility fraction. A reasonable approximation could be computed as the mass-weighted solubility of the total initial aerosol distributions.
Figure 2 displays a cross section of the number con- centration of submicrometer sulfate aerosols and regen- erated aerosols through the main updraft of an idealized deep convective storm at 2 h into the test simulation (similar to Saleeby and Cotton 2004a). The model was initialized with 3D homogenous aerosol number con- centrations of 1000 cm23 and a median radius of 0.04 mm in the submicrometer sulfate category only. Other aero- sol species were initialized with zero concentration. There is a distinct area of aerosol nucleation scaveng- ing (reduced aerosol concentration) where the center of the updraft resides. Regenerated aerosols are con- centrated near and below cloud base (2--3 km) and along the edges of the updraft (up to 11 km). It is likely that the modest amount of aerosol regeneration along the region where cloud edges would exist at mid to upper levels is a result of turbulent mixing and entrainment. Engstroeurom et al. (2008) found similar regions of aerosol regeneration along cloud boundaries because of mix- ing with dry air. The greatest zone of aerosol re- generation is to the rear of the storm (left side), where the convective downdraft resides and enhanced subsidence increases evaporation of hydrometeors.
c. Heterogeneous ice nucleation
The parameterization of heterogeneous ice nucleation implemented here is from DeMott et al. (2010). Their active IN formulation is based on observations from nine field projects collected over 14 years across a wide range of locations. Observations were taken with the CSU Continuous Flow Diffusion Chamber in water-saturated conditions (relative humidity . 100%). As such, this technique captures ice particles nucleated via deposition nucleation, condensation freezing, and immersion freezing. A strong correlation was found between IN concentration and the concentration of aerosol particles with diameters greater than 0.5 mm. The DeMott formula, given as
where a 5 0.000 059 4, b 5 3.33, c 5 0.0264, and d 5 0.0033, is a power law fit to these data and represents the maximum number of IN activated down to a tem- perature of 2358C. Number concentrations in Eq. (6) are specified at standard temperature and pressure (STP). This equation takes into consideration all sampled aero- sol species, which includes aerosols with different chem- ical compositions as well as insoluble mineral dust. Figure 3 displays plots of the DeMott formula at 2 158 , 2 258 , and 2358C. The trend toward greater IN activation at colder temperatures is evident. To apply the DeMott formula in RAMS, we determine the median radius for each of the non--sea salt aerosol categories, decompose the distributions into binned lognormal spectra, and sum the total number of aerosols greater than 0.5-mmdi- ameter. The formula then determines the number of ac- tivated particles, which are then removed from the available population as ice particles are generated. Par- ticles are activated and removed from largest to small- est. New ice mass is generated up to the amount of available water mass above water saturation.
It should be noted that the DeMott formula provides the maximum number of activated IN for a given tem- perature. Once the formula is applied at a certain model grid point it cannot be directly reapplied at this location in the subsequent model time step unless conditions are colder or more saturated so as to support additional nucleation; otherwise, overnucleation of IN occurs. To prevent overnucleation, the number of activated IN are tracked within the model. The IN tracking variable is treated as a new 3D scalar variable with predictive ten- dencies computed accordingly throughout the model. By knowing the number of activated and unactivated IN at a given location, we can determine if additional ice nu- cleation should occur for the given ambient conditions.
It should be noted here that, while the DeMott for- mula is not active in water subsaturated conditions, the microphysics model simulates contact nucleation, ho- mogeneous freezing of cloud droplets, and homogeneous freezing of deliquesced, but unactivated, haze particles (Walko et al. 1995). Furthermore, the user retains the option to use the Meyers et al. (1992) formula, which allows water subsaturated activation of IN.
d. Sea salt model
A model for the prediction of sea salt aerosols has been implemented (Carrio and Cotton 2011). Emission of sea salt particles from the ocean surface is a function of wind-driven processes. O'Dowd et al. (1997, 1999) provide empirical relationships between near-surface wind speed over the ocean and the concentration of sea salt aerosols. They divided the sea salt aerosol distri- bution into three naturally occurring modes: 1) a spume drop mode of ultragiant particles with mode radius of rg5 6.0 mm that is generated from the shearing of wave crests, 2) a jet drop mode (rg 5 1.0 mm) that is formed from small jets that are emitted from bursting bubbles in the foam top of white caps, and 3) a film drop mode (rg 5 0.1 mm) that forms from the film of the bursting bubbles that generate the jet drops. The numerical relationships for the number concentration of the three modes, based on the 10-m wind speed, are given (in mks units) as
O'Dowd et al. (1999) compiled these relationships for wind speeds from 2 to 17 m s21, Fan and Toon (2011) presented sea salt emissions for wind speeds up to 20 m s21, and Smith et al. (1993) provided spume emis- sions for winds speeds up to 32 m s21. As such, we apply an upper wind speed limit of 20 m s2l for the film and jet modes and 32 m s2l for the spume mode so as to prevent overproduction of sea salt in high wind conditions. When implementing Eq. (7) for sea salt number concentration, we apply these values to the lowest model level above ground. If mixing or deposition reduces the salt concen- trations at the surface, we apply a time weighted tendency function, acting on a time scale of Dt/tsalt, to bring the concentrations back to the predicted values.
Figure 4 displays a vertical cross section of the number concentration of particles in the film sea salt mode result- ing from emissions due to strong surface winds associated with convective downdraft outflow over an ocean surface (the same test simulation as used in Fig. 2, but here with an ocean surface). These results occurred during the mature phase of the convection when near-surface hori- zontal winds were strong. The top (bottom) panel dis- plays the concentrations that result without (with) the maximum wind speed limit of 20 m s21. There is a maxi- mum source region of sea salt particles near the peak in the surface wind speed, and some of the emitted particles are drawn into the updraft. Without the wind speed limit, number concentrations become twice as large.
e. Dust aerosol source model
A model for dust lofting from an arid surface based on soil moisture, wind speed, and soil clay percentage has been implemented in RAMS (Smith 2007; Seigel and
Note that there is an inverse relationship between par- ticle size and the threshold friction velocity. Smaller particles require a larger wind speed to overcome cohesion forces. Alfaro and Gomes (2001) and Shao (2001) have taken a similar approach, while recent work by Kok (2011) suggests that size-specific dust emission does not vary with threshold friction velocity. However, as model wind speeds are increased beyond the range of the threshold friction velocity [;(0.5--2.5)m s21]thechange in the emitted particle size distribution decreases, and the differences in these parameterizations are minimized.
Soil cohesion and the threshold friction velocity are also a function of soil type and soil moisture. Water in the soil increases capillary forces that bond particles together and enhance cohesion (F^ecan et al. 1999). However, the adsorption capacity of the soil determines when the water capillary forces become strong enough to enhance co- hesion. The soil has to reach adsorption capacity before cohesion increases, and subsequently stronger wind speeds are needed for lofting. From F^ecan et al. (1999), the volumetric soil moisture at maximum ad- sorption is a function of clay content, given as
where ''% clay'' refers to the fraction of clay composition within a given soil type, expressed as a percent. The wet soil threshold friction wind velocity (in cgs units) is
Soils with higher clay content can be lofted more easily for a given wind speed because of the enhanced ability to ad- sorb water and limit interparticle cohesion from water capillary tension. Table 2 displays the soil types in the Land Ecosystem--Atmosphere Feedback, version 3 (LEAF-3), land surface model [the former version, LEAF-2, is docu- mented in Walko et al. (2000b)] along with the maximum volumetric soil moisture
We represent lofting of dust particles across seven particle radius bins (0.15, 0.26, 0.47, 0.83, 1.50, 2.65, and 4.71 mm) since there may be variability in lofting with size (Tegen and Fung 1994; Tegen and Lacis 1996; Ginoux et al. 2001). Dust mass flux Fp bin follows Ginoux et al. (2001) (in cgs units) as
Greater roughness length reduces the erodible fraction and the amount of dust lofting. Table 3 displays the vegetation land surface classes that may act as dust lofting sources as well as the vegetation roughness length, veg- etation height, and total leaf area index (TAI). (TAI values are representative of the summer growing season. The roughness length for certain vegetation classes will vary as the TAI varies during seasonal changes.) Dust lofting is not active for surface patches covered by snow or for the vegetation classes of ocean, inland water, ice, marshes, wetland trees, or highly urban.
The mass flux in each size bin is converted into number concentrations of newly lofted dust within the first model layer above ground in a given time step (in cgs units) as
The four (three) smallest (largest) size bins are summed to comprise the submicrometer (supermicrometer) dust category with median radius of 0.69 (2.95) mm. Figure 5 displays the surface dust flux as it varies with soil mois- ture, clay percentage, and 10-m wind speed. Dust flux increases with a decrease in soil moisture, increase in soil clay percentage, and increase in wind speed. The greatest variability occurs with wind speed, then with soil mois- ture, and least with clay percentage. Seigel and
f. Aerosol wet deposition (precipitation scavenging)
An aerosol wet deposition (precipitation scavenging) scheme was initially implemented by Smith (2007) in RAMS 4.3 that accounts for scavenging by rain drops in the subcloud layer based on the surface rain rate. It has since been integrated into RAMS 6.0 and is applied to each of the nine aerosol species. It has been extended to additionally consider precipitation scavenging within clouds by all hydrometeor species with the use of 3D precipitation rates. Precipitation particles falling through the atmosphere collide with aerosol particles and collect a portion of them. Higher precipitation rates lead to greater collection and removal of aerosol particles. The rainfall scavenging coefficient for a monodisperse pop- ulation of raindrops, presented by Seinfeld and Pandis (2006) and Wang et al. (2010) and based on Slinn (1983) is computed (in mks units) as
The aerosol diameter used here is the wet or deliquesced aerosol diameter computed from the dry diameter as a function of relative humidity, according to Fitzgerald (1975).
The size specific collection efficiency E(dp , Dp )be- tween aerosols and raindrops is a function of three primary collection mechanisms: Brownian diffusion, in- terception, and inertial impaction (Slinn 1983; Seinfeld and Pandis 2006; Berthet et al. 2010; Wang et al. 2010). Collection by Brownian diffusion is a function of random motions that bring aerosols in contact with falling drops, and as such, decreases with increasing aerosol particle size; it is most effective for aerosols with diameter less than 0.2 mm. Interception occurs primarily for inter- mediate size particles (0.2 , dp , 1.0 mm). This method occurs as aerosols follow the airflow streamlines around a drop. These aerosols are of small enough mass that they tend not to cross flow-field streamlines; only particles that come within a half diameter distance of the drop will collide. This method is the most inefficient of the three. Inertial impaction occurs for larger aero- sols (dp . 1.0 mm). The flow of these aerosols is more a function of the mass rather than the diameter; aero- sols with larger mass have a greater inertia and can cross flow-field streamlines as falling drops scavenge out a volume of air. Given the greater dependence on mass, higher density aerosols are scavenged more efficiently at smaller diameters than lower density particles. The collection efficiencies are computed (in mks units) as
E(dp , Dp ) [collection efficiency (0 , E , 1)] 5
(Brownian diffusion contribution)
(inertial impaction contribution), 16
This equation set [Eq. (16)] computes a collection efficiency value based on single aerosol and raindrop sizes; however, in the absence of a bin model, the RAMS bulk microphysics model uses hydrometeor gamma distributions and aerosol lognormal distributions. To apply the collection efficiency and scavenging coefficient equations to RAMS, we must use the mean raindrop diameter for Dp and the median aerosol diameter for dp as reasonable approximations to represent these bulk quantities in the size-specific scavenging equations. Power-law relationships for raindrop mass and fall speed are used for computing collection efficiencies. The ap- plication of Eq. (16) to Eq. (15) provides the scavenging coefficients for a range of aerosol particle sizes, particle densities, and raindrop sizes (Fig. 6). We extend this scavenging approach to ice hydrometeors by assuming them to be spheres for Eq. (16), although power-law co- efficients for mass and fall speed are ice habit specific. Seinfeld and Pandis (2006) suggest that the aerosol col- lection efficiencies for ice hydrometeors with complex shapes may be higher than for spheres. Therefore, our implementation for scavenging by ice particles may rep- resent a lower bound on scavenging in cold and mixed- phase clouds.
For verification, Fig. 6 displays scavenging rates for aerosol particle collection at a rainfall rate of 1 mm h2l. These are in good agreement with Wang et al. (2010) and Seinfeld and Pandis (2006). A demonstration of scavenging is shown in Fig. 7 for the collection of 3-mm diameter dust aerosols by precipitation hydrometeors along a transect through the main updraft within our deep convection simulation. We have isolated the pre- cipitation scavenging process by not allowing dust to be removed by nucleation scavenging. Figure 7a (Fig. 7b) reveals dust number concentrations near the main up- draft of the storm without (with) scavenging. The initial dust concentration vertical profile can be seen in the lower levels of the undisturbed region to the right of the updraft. Without scavenging, these large dust particles are transported from the boundary layer into the updraft and diverge into the anvil region. With active scaveng- ing, precipitation in the updraft removes the dust, thereby preventing dust from reaching high concentra- tions at upper levels. Though not shown, plots of the scavenging of smaller particles with lower scavenging rates showed less aerosol removal and more transport to the anvil. It is important to also note that aerosols subject to precipitation scavenging are tracked within hydrome- teor species and can be regenerated by hydrometeor evaporation.
g. Aerosol dry deposition (gravitational settling)
The fall speed of settling aerosol particles results from a balance between gravitational and atmospheric drag forces. First, we compute the wet particle size from the dry particle size, aerosol solubility fraction, and relative humidity according to Fitzgerald (1975). From Baron and Willeke (2001) and Seinfeld and Pandis (2006), the gravitational settling velocity through atmospheric layers above the first model layer is expressed for the Stokes regime (particle Reynolds numbers ,0.1), as
Here dp is the aerosol wet particle diameter. The aerosol distribution mean fall velocity is based on the bulk lognormal distribution median diameter and is applied equally to the settling of aerosol mass and number down to the second model layer above ground. The dry de- position of particles in the first model layer is additionally subject to surface characteristics and surface layer tur- bulence effects.
Slinn and Slinn (1980) addressed dry deposition onto a water surface. For calm winds the deposition velocity equals the gravitational settling velocity. For nonzero winds in the surface layer, the deposition velocity can be expressed (in mks units) as
The dry deposition of aerosol particles onto a vegetation or bare soil surface is parameterized according to Slinn (1982) and Zhang et al. (2000) (in mks units) as
and «0 is a constant of 3.0. The vegetation surface col- lection efficiencies (Ex) are
where g, A, a, and b are empirical constants that vary with surface types given by Zhang et al. (2000). The vegetation Stokes number is used for vegetation sur- faces only and reduces to the Stokes parameter in Eq. (21) for nonvegetative surfaces, Rb reduces to 1.0 for water, and EIN reduces to 0.0 for nonvegetative surfaces (desert, tundra, water, and ice).
Figure 8 displays the aerosol dry deposition velocities from gravitational settling and deposition onto water and vegetation for a range of aerosol sizes. This plot agrees well with the results from Slinn and Slinn (1980) for deposition onto water, and from Slinn (1982) and Zhang et al. (2000) for deposition onto vegetation. Note that the gravitational settling velocity of submicrometer- diameter particles in the free atmospheric is substantially lower than surface deposition. The increase in surface dry deposition of particles .;1 mm in diameter is from the increased effects of inertial impaction.
h. Aerosol direct radiation effects
An aerosol radiative transfer scheme, implemented by Stokowski (2005) into RAMS 4.3, has been applied to the nine aerosol species in RAMS 6.0. This scheme runs interactively with the hydrometeor-sensitive two-stream radiation model of Harrington (1997) that computes the absorption and scattering of primary atmospheric gases (Ritter and Geleyn 1992) and hydrometeors (Mie 1908) across eight radiation bands. Mie theory is also applied to the aerosol distributions to compute their impact on the optical depth (tp , total extinguished radiation), the single-scatter albedo (vp , fraction of extinguished radi- ation that is scattered), and the asymmetry parameter (Gp , direction of scattered radiation). Mishchenko et al. (1997) concluded that the optical parameters differ very little between a spherical and nonspherical aerosol as- sumption. Under a spherical aerosol assumption, the Mie solution for aerosols only requires input of the wave- length of incident radiation, the wavelength-dependent complex index of refraction of the aerosol species (d'
At each model radiation time step, the aerosol distribution-total Mie coefficients are accessed from the lookup tables and optical parameters are computed for each aerosol species a by summing the size and number of concentration-weighted bin-resolved coefficients over the distributions in 17 (aerobins) size bins i (in mks units) as
The optical parameters for all aerosol species are com- bined similarly to hydrometeor extinction and scattering following Liou et al. (1978) and Slingo and Schrecker (1982):
Summation of the absorption and scattering by aerosol particles, hydrometeors, and atmospheric gases across all radiation bands allows for computation of total at- mospheric net radiation fluxes as well as heating rate profiles.
An idealized simulation was run to assess the aerosol radiation scheme in a controlled environment. Dust layer concentrations for the submicrometer and super- micrometer modes (Fig. 9a) were approximated from the vertical profiles of aerosols shown in DeMott et al. (2003). The previously used thermodynamic sounding was ap- plied, though initial winds were set to zero to minimize short-term mixing of the aerosol layer. The solar angle was set at 658, and incoming solar radiation at the top of the aerosol layer (4 km) is ;1100 W m22.Figure9b displays the radiative impact of dust particles in a non- cloudy atmosphere after 1 h. The radiative impacts of dust lead to an increase in the heating rate in the dust layer of ;0.2Kh21. This falls within the range of heating rates shown by Carlson and Benjamin (1980) when integrated over a full solar day.
3. Testing of new modules
Recent developments described herein have been tested in simulated cases of precipitating stratocumulus clouds, deep convection, and winter orographic precipi- tation. In these cases, we examined the aerosol sensitivity with regard to 1) aerosol solubility, 2) aerosol regenera- tion, 3) precipitation scavenging, 4) nucleation scaveng- ing, and 5) the DeMott ice nucleation scheme.
The stratocumulus simulations (referred to as ATEX) were initialized horizontally homogeneous with a sound- ing from the Atlantic Trade Wind Experiment (ATEX) experiment (Stevens et al. 2001). Simulations were run in 2D for 24 h with periodic boundary conditions, 200-m horizontal grid spacing, 50-m vertical grid spacing, and an ocean surface with SST of 298 K. A sample cross section of the precipitating stratocumulus field is shown in Fig. 10a. Simulations of deep convection (referred to as STORM) were initialized horizontally homogeneous with a high CAPE sounding suitable for producing a su- percell thunderstorm; this is the same set of initial con- ditions from Saleeby and Cotton (2004a). The storm was simulated for 2 h with 1-km horizontal grid spacing, 100-m vertical grid spacing at the surface stretched to 1000 m aloft, and convection was initiated with a 2-K warm bubble. A sample cross section through the de- veloped supercell is shown in Fig. 10b. Last, we ran the winter orographic snowfall experiments (referred to as TOPO) for 42 h over the central mountains of
For each simulation, aerosols were initialized with a vertically decreasing profile of ammonium sulfate aerosols with 0.04 micrometers median radius [similar to Saleeby et al. (2013)]. Initial maximum aerosol concentrations were 1000 cm23 for the ATEX and STORM simulations and 1500 cm23 for TOPO. The higher concentration for the TOPO simulations was set to that used in the sim- ulations by Saleeby et al. (2013). For direct comparison between the Meyers and DeMott IN schemes, a common IN concentration profile was initialized that scales with the decrease in density with height with a maximum concentration of In the control simulations aerosols are 90% soluble, removed via nucleation and precipitation scavenging, and regenerated via hydro- meteor evaporation. Heterogeneous ice nucleation is represented by the Meyers scheme.
For each event type (ATEX, STORM, and TOPO), five sensitivity tests were performed in which a single aerosol effect was modified relative to the control sim- ulation: 1) aerosol solubility was reduced from 90% to 5%, 2) aerosol regeneration was turned off, 3) aerosol precipitation scavenging was turned off, 4) aerosol nu- cleation scavenging was turned off (aerosol nucleation was supersaturation limited such that droplet number concentration cannot exceed aerosol number concen- tration), and 5) the IN scheme was changed from Meyers to DeMott. For these five sensitivity experiments, we will focus the discussion on their impact on precipitation rates. Figure 11 reveals histograms of the spatial and temporal summation of the counts of grid cells that fall within bins of light, moderate, and heavy precipitation rate (mm h21). Note from the panel labels that cases vary by row and precipitation rate varies by column; further, the scales of gridcell count vary in each panel for visualization purposes. The percentage change relative to the control simulation is displayed above each sensi- tivity test histogram.
One of the most noticeable characteristics drawn from these histograms is that the various aerosol schemes are of differing importance among simulated cloud systems. In the ATEX simulations, both the low solubility and no-regeneration simulations lead to substantial percent- age increases in all precipitate rate bins. Both of these experiments lead to production of fewer but larger nucleated cloud droplets, relative to the control, that produce an efficient warm rain process and greater pre- cipitation production. In the test with no precipitation scavenging, more aerosols remain available for nucle- ation. This leads to more numerous cloud droplets and slight precipitation suppression, though the suppression is small. It should be noted that the median radius of the initial aerosol distribution is in the size range where precipitation scavenging is limited. Had we initialized with supermicrometer-sized aerosols or much smaller aerosols, this scavenging effect would likely have been larger, but the impact of solubility and nucleation scav- enging would be substantially altered. In the test with no nucleation scavenging, there is a large increase in lightly precipitating grid cells and a decrease in heavy precipitation. Without nucleation scavenging, the model tends to have higher cloud droplet concentration over a greater area, thus leading to suppression of heavier precipitation and an increase in the frequency of light precipitation. The DeMott IN scheme is not tested in the ATEX case since all clouds are below the freezing level.
In the STORM simulations, only the regeneration and nucleation scavenging tests produce precipitation rate changes greater than a few percent. Without aerosol re- generation, the environment remains cleaner and leads to a more efficient warm rain process. This results in fewer lightly and moderately precipitating grid cells and a greater number of heavily precipitating grid cells. Without nucleation scavenging, there is the potential for enhanced cloud droplet production, which leads to droplets of smaller size. This response leads to sup- pression of precipitation, with the greatest impact on heavy precipitation. The DeMott IN test shows very little change in this case since heterogeneous ice nucleation produces few ice crystals in comparison with homoge- neous freezing of cloud droplets lofted above the 2408C level by the updraft. The impact of these aerosol parameterizations in the STORM case is small rela- tive to the percentage impact in the ATEX case. The STORM case is strongly dynamically forced, thus making these particular aerosol parameterizations of secondary importance.
In the TOPO simulations, the model responds most noticeably to the choice of IN parameterization com- pared to the other aerosol tests. In a winter orographic precipitation environment, cloud nucleating aerosols primarily affect precipitation by modifying the riming efficiency of snow falling through clouds of supercooled water. This may have a substantial localized effect and potentially induce changes in the spatial distribution of precipitation (Saleeby et al. 2009, 2013). When switching from the Meyers to the DeMott IN scheme, there is a substantial decrease in heavy precipitation. Figure 12 displays plan views of the total accumulated precipitation across the domain as well as the difference in total pre- cipitation. Along the ridgeline and windward slope of the
4. Summary and conclusions
Recent developments in the CSU-RAMS model, version 6.0, provide for a more comprehensive aerosol model that extends beyond the capabilities of previous aerosol--microphysics--related research (e.g., Saleeby et al. 2009, 2013;
Sensitivity tests of the influence of aerosol solubility, nucleation scavenging, precipitation scavenging, and regeneration were performed for simulations of strato- cumulus clouds, deep convection, and winter orographic precipitation. The Meyers and DeMott heterogeneous ice nucleation schemes were also compared. Results indi- cate that in cloud systems with active warm rain processes, the representations of aerosol nucleation scavenging and regeneration are most influential among these five tested aerosol parameterizations, though, in weakly forced clouds, aerosol solubility can be important. In deep con- vection, the dynamical influence overwhelms these sec- ondary aerosol effects and the degree of impact is less than in shallow clouds. The impact of the DeMott ice nucleation scheme is limited in deep convection since other ice nucleation processes are dominant. However, in winter orographic precipitation, the DeMott scheme reduces orographic precipitation and appears to bring snowfall accumulations more in line with obser- vations. In future work, we will continue to refine the treatment of aerosols and work to improve the simu- lation of their physical impacts on cloud and climate systems.
Acknowledgments. This work was supported by the
Albrecht, B., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227--1230.
Alfaro, S. C., and
Baron, P. A., and
Bernardet, L., and
Bohren, C. F., and
Carlson, T. N., and
Carrio, G. G., and
Castro, C. L.,
Cheng, W. Y. Y.,
Cotton, W. R., and Coauthors, 2003: RAMS 2001: Current status and future directions. Meteor.
Darby, L. S.,
DeMott, P. J.,
_____, and Coauthors, 2010: Predicting global atmospheric ice nu- clei distributions and their impacts on climate. Proc. Natl. Acad. Sci.
Ekman, A. M. L.,
_____, _____, J. Stroeurom, and
Engstroeurom, A., A. M. L. Ekman,
Fan, T., and
Feingold, G., and
Fitzgerald, J. W., 1975: Approximation formulas for the equilib- rium size of an aerosol particle as a function of its dry size and composition and the ambient relative humidity.
Flossmann, A. I.,
_____, H. R. Pruppacher, and
Fridlind, A. M.,
Harrington, J. Y., 1997: The effects of radiative and microphysical processes on simulated warm and transition season Arctic stratus. Ph.D. dissertation,
Haywood, J., and Coauthors, 2003: Radiative properties and direct radiative effect of Sharan dust measured by the C-130 aircraft during SHADE:1. Solar spectrum. J. Geophys. Res., 108, 8577, doi:10.1029/2002JD002687.
Hegg, D. A.,
Heymsfield, A. J., and
Jiang, H., and G. Feingold, 2006: Effect of aerosol on warm con- vective clouds: Aerosol-cloud-surface flux feedbacks in a new coupled large eddy model. J. Geophys. Res., 111, D01202, doi:10.1029/2005JD006138.
Kok, J. F., 2011: Does the size distribution of mineral dust aerosols depend on the wind speed at emission?
Lee, S. S., 2012: Effect of aerosol on circulations and precipitation in deep convective clouds. J.
Liou, K. N.,
Lu, L., and
Marticorena, B., and G. Bergametti, 1995: Modeling the atmo- spheric dust cycle: 1. Design of a soil-derived dust emission scheme. J. Geophys. Res., 100, 16 415--16 430.
Meyers, M. P.,
Mie, G., 1908: Beitraeuroge zur Optik trub? er Medien, spieziell kolloidaler Metalloeurosungen (Contributions to the optics of turbid me- dia, particularly of colloidal solutions). Ann. Phys., 330, 377--445, doi:10.1002/andp.19083300302.
Mishchenko, M. I.,
O'Dowd, C. D.,
Olsson, P., and
Petters, M. D., and
Pierre, C., G. Bergametti,
Reutter, P., and Coauthors, 2009: Aerosol- and updraft-limited regimes of cloud droplet formation: Influence of particle num- ber, size and hygroscopicity on the activation of cloud conden- sation nuclei (CCN).
Ritter, B., and J.-F. Geleyn, 1992: A comprehensive radiation scheme for numerical weather prediction models with potential application in climate situations. Mon. Wea. Rev., 120, 303--325.
Rutledge, S. A.,
Saleeby, S. M., and
_____, and _____, 2004b: Simulations of the North American mon- soon system. Part I: Model analysis of the 1993 monsoon season. J. Climate, 17, 1997--2018.
_____, and _____, 2008: A binned approach to cloud-droplet riming implemented in a bulk microphysics model.
_____, _____, and
Seigel, R. B., and
Seinfeld, J. H., and
Shao, Y., 2001: A model for mineral dust emission. J. Geophys. Res., 106, 20 239--20 254.
Slingo, A., and
Slinn, S. A., and
Slinn, W. G. N., 1982: Predictions for particle deposition to vege- tative canopies.
_____, 1983: Precipitation scavenging. Atmospheric Sciences and Power Production,
Smith, M. A., 2007: Evaluation of mesoscale simulations of dust sources, sinks, and transport over the
Smith, M. H.,
Stevens, B., and Coauthors, 2001: Simulations of trade wind cumuli under a strong inversion. J.
Stokowski, D., 2005: The addition of the direct radiative effect of atmospheric aerosols into the Regional Atmospheric Model- ing System (RAMS). M.S. thesis,
Storer, R. L., and
Tegen, I., and I. Fung, 1994: Modeling of mineral dust in the at- mosphere: Sources, transport and optical thickness.
Twomey, S., 1974: Pollution and the planetary albedo.
Walko, R. L.,
_____, and Coauthors, 2000b: Coupled atmosphere--biophysics-- hydrology models for environmental modeling.
Wang, C., and
Ward, D. S., T. Eidhammer,
Zhang, L., S. Gong,
STEPHEN M. SALEEBY AND SUSAN C. VAN DEN HEEVER
Corresponding author address:
Most Popular Stories
- Apple, HP, Intel May Take a Hit from Slowdown in Smartphone Sales Growth
- FDIC Files Lawsuit on Behalf of Banks Allegedly Hurt by Libor Scandal
- Some California Cities Seeking Water Independence
- Motley Crue's Nikki Sixx Marries Model Courtney Bingham
- Chinese e-Commerce Giant Alibaba Gears for IPO in U.S.
- Will Missing Malaysian Jet Prompt Aviation System Change?
- Jack Daniel's Resists Changes to Tenn. Whiskey Law