Aerosol specification in single-column CAM 5

Introduction Conclusions References


Introduction
The Single Column Model (SCM) version of Community Atmospheric Model (CAM) is a very important tool efficient development of model numerics and physics.Based on observed test cases, many SCM intercomparison studies of stratocumulus and cumulus cloud-top boundary layers have been undertaken with the goal of improving physical parameterizations of clouds and cloud-related processes and their interactions.A number of SCM intercomparison studies by the Global Energy and Water Experiment (GEWEX) Cloud Systems Study (GCSS) Boundary Layer Cloud Working Group (BLCWG) have been conducted to understand common biases in climate models.For example, one of the early SCM intercomparison studies (Moeng et al., 1996) simulated nocturnal non-precipitating stratocumulus clouds and showed that the LWP decreased substantially during the initial period of the simulation, which was explained by excessive dry air entrainment.Another SCM intercomparison simulations of the Second Dynamics and Chemistry of the Marine Stratocumulus field study (DYCOMS II) research flight RF01 (DYCOMSRF01) also showed low liquid water path (LWP) despite improvement of entrainment rates in the models (Zhu et al., 2005).SCM intercomparison of drizzling stratocumulus from the DYCOMS II research flight 02 (DY-COMSRF02) by vanZanten and Stevens (2005) tested the impact of drizzle in SCMs and found that drizzle decreased LWP substantially in most of the models.Another SCM study by Wyant et al. (2007) also carried out SCM intercomparison simulations for the DYCOMSRF02 case.They found that models need improvement in drizzle, sedimentation, and sub-cloud evaporation parameterizations.A recent SCM and cloud-resolving model intercomparison study by Klein et al. (2009) simulated the mixed-phase stratocumulus cloud observed during the Atmospheric Radiation Measurement (ARM) program's Mixed-Phase Arctic Cloud Experiment (MPACE-B).They found that models generally showed ice water path (IWP) in good agreement with observations while LWP was severely under predicted.This was attributed to the interaction between liquid and ice-phase microphysics suggesting the need to improve the representation of Figures mixed-phase microphysics.Previous SCM and LES intercomparison studies were also undertaken for deep (ARM southern Great Plain (ARM SGP) site) and shallow (Rain in Cumulus over the Ocean (RICO)) convective cases.Ghan et al. (2000) performed an SCM intercomparison study for ARM SGP using eleven SCMs and found that no individual models stood out as superior, and the model ensemble showed close agreement with observations.A recent study by VanZanten et al. (2011) used twelve LES simulations to study the interplay between micro and macro physics processes in the evolution of clouds and precipitation, with a wide range of microphysical representations, during the undisturbed period of the RICO field study.Many features of their LES simulations generally agreed with observations.Similar thermodynamic and energetic behavior was produced as compared to previous studies based on SCMs.A significant fraction of the uncertainties in climate projections results from the representation of aerosol (Houghton et al., 1996;Haywood and Boucher, 2000;Forster et al., 2007).Aerosols affect climate by directly absorbing and reflecting atmospheric radiation (known as the direct effect) and by changing cloud optical properties and lifetimes (known as aerosol indirect effects).As a result, development and testing of aerosol parameterizations have been a high priority in the climate modeling community.
The inclusion of the prognostic aerosol model in CAM had been a major breakthrough in its development (Abdul-Razzak and Ghan, 2000;Liu et al., 2012).However, CAM5-SCM has not been updated appropriately to handle the addition of prognostic aerosol in CAM.In particular, it initializes aerosol mass mixing ratios to zero.As a result, the default SCM release substantially underestimates IWP and LWP of the SCM simulations for a variety of cloud regimes.
In this study we test the impact of the zero aerosol initialization problem, and we introduce fixes for this issue.To ensure representativeness, we test SCM simulations for a variety of cloud regimes.The SCM cases used for this study include summertime mid-latitude continental convection (ARM95), shallow convection (RICO), subtropical drizzling stratocumulus (DYCOMSRF02), and multi-level Arctic clouds (MPACE-B).Results are analyzed and compared to observations and previous LES results.Introduction

Conclusions References
Tables Figures

Back Close
Full

SCAM5 setup
In this study we employed the SCM version of CAM5 (SCAM5), which consists of physics parameterizations driven by prescribed advective tendencies (Hack and Pedretti, 2000).
There are two types of clouds in SCAM5: stratus clouds with symmetric turbulent properties and cumulus clouds with vertically stretched shapes and asymmetric turbulence properties.We use the Morrison and Gettelman stratiform cloud microphysics scheme (Morrison and Gettelman, 2008) and the Park et al. (2014) macrophysics scheme to model stratiform clouds.Deep convection is handled by the modified Zhang-McFarlane parameterization scheme (Zhang and McFarlane, 1995), and shallow convection is parameterized by the University of Washington shallow convection parameterization scheme (Park and Bretherton, 2009).Turbulence is handled following Brethorton and Park (2009).Radiation is calculated using the Rapid Radiative Transfer Model (RRTMG) radiation scheme (Mlawer et al., 1997).
CAM5 is the first version of CAM that was designed to simulate aerosol-cloud interactions.It has a three mode simplified modal aerosol model (MAM3) (Easter et al., 2004;Ghan et al., 2012) with Accumulation, Aitken, and Coarse modes.MAM3 is capable of treating complex aerosol physical, optical, and chemical processes and simulating aerosol size, mass and number distributions.The aerosol size distribution is lognormal, and internal and external mixing between aerosol components is assumed in the model.As mentioned previously, this prognostic aerosol model in SCAM5 mode initializes the mass-mixing ratio of the different aerosol species to zero.Hence we test other fixes to solve this problem as described below.Introduction

Conclusions References
Tables Figures

Back Close
Full  3. The last method we employed is the observed aerosol case where we use observed mixing ratios and size distributions of the aerosols in MAM3.This method (hereafter named obsAero) modifies the PrescAero methodology to instead use observed mass mixing ratios of the different aerosol species for all the modes.To use this mode, observed values are needed for parameters N j , a mj , and σ j for the multimode lognormal aerosol size distribution given by the following equation (Abdul-Razzak and Ghan, 2000): where N j , a mj , and σ j are the number concentrations of the aerosol mode, the geometric mean dry radius, and the geometric standard deviation of aerosol mode j , respectively.Introduction

Conclusions References
Tables Figures

Back Close
Full

SCAM cases
In an attempt to test aerosol effects over the full range of cloud types, we tested our fixes using case studies from four different cloud regimes.We set up four SCAM5 case simulations using the Default configuration and each of the three different fixes discussed in the previous section.et al. (2007), respectively.For this paper we used an experimental configuration similar to Wyant et al. (2007).Subtropical stratocumulus are important because of all cloud types they have the biggest impact on the planetary radiation budget (Hartmann et al., 1992), and difficulty in simulating them is the leading source of uncertainty in climate sensitivity (Bony and Dufresne, 2005).
Like Wyant et al. (2007), for maintaining an approximate balance between radiative cooling and subsidence warming above the inversion, a constant divergence with a value 3.75 × 10 −6 s −1 was used to create an omega profile in the DYCOMSRF02 case, the RRMTG shortwave radiation was turned off, and we ran our simulations for 6 h.Constant surface latent and sensible heat flux values of 93 and 16 W m −2 (respectively) were imposed based on observed mean values from vanZanten and Stevens (2005).
Unlike Wyant et al. (2007), the default RRMTG longwave radiation code was used instead of applying an idealized radiation scheme.We also kept u and v for our simulations constant.Cloud processes were turned off above 700 hPa in order to Introduction

Conclusions References
Tables Figures

Back Close
Full prevent ice formation at the troposphere, which occurs due to assumptions related to subgrid relative humidity variability in CAM5.
For the FixHydro case, an observed N d value of 55 cm −3 was used as recommended by Wyant et al. (2007).The bimodal lognormal distribution (Eq. 1) was assumed to consist of sulfate aerosols with dry density 1.77 g cm −3 .The total number, mode radius, and geometric standard deviation for the aitken (125 cm −3 , 0.011 µm, 1.2) and accumulation (65 cm −3 , 0.06 µm, 1.7) modes, respectively were used.These values were chosen by Ackerman et al. (2009) to produce an in-cloud droplet concentration in their LES, which matched the observed droplet concentration value of 55 cm −3 .

b. b.MPACE-B case
The second case is MPACE-B, which consists of mixed-phase stratocumulus over open Ocean near Barrow, AK.The MPACE-B case is based on Arctic stratocumulus observed during the Mixed-Phase Arctic Cloud Experiment period B, which was the subject of an intercomparison by Klein et al. (2009).The case focused on the period 9 October, 17:00 UTC to 10 October, 05:00 UTC, 2004.This case is useful because it is relatively simple yet includes both liquid and ice processes.The setup of this case was similar to that of Klein et al. (2009).Above 700 hPa, all variables were kept near to their initial values by nudging temperature and moisture with a time scale of 1 h.While Klein et al. (2009)  to zero at TOA.The value used for advective temperature (moisture) tendency at the surface is −4.63 × 10 −5 K s −1 (−3.47 × 10 −8 kg kg −1 s −1 ); it increases linearly to a value of −0.174 × 10 −5 k s −1 (−0.19 × 10 −8 kg kg −1 s −1 ) at 850 hPa, and stays constant above this level.
For the FixHydro case, an ice crystal concentration of 0.16 L −1 (were used as recommended by Klein et al., 2009) and a N d value of 50 cm −3 .For the ObsAero case, the aerosol mass mixing ratios of the three modes were diagnosed from the number mixing ratio and bimodal log-normal size distributions (Eq. 1) with aerosol partitioning of 70 % SO 4 and 30 % primary organic matter (POM) for the accumulation mode and 10 % SO 4 , 85 % sea salt, and 5 % dust for the coarse mode.We also used total number concentration, mode radius, and geometric standard deviation of the accumulation (72.2 cm −3 , 0.054 µm, 2.04) and coarse (1.8 cm −3 , 1.3 µm, 2.5) modes, respectively (again following Klein et al., 2009).One unique aspect of the RICO case is that radiation tendencies are included in the prescribed temperature advection tendency.As a result, we had to turn off the shortwave and longwave radiation schemes.There was no nudging applied for this case since specifications were chosen to be energetically and moisture Introduction

Conclusions References
Tables Figures

Back Close
Full  2011), piecewise linear profiles of u, v, omega, and large-scale forcings of heat and moisture were used.The u value used was −1.9 m s −1 near the surface linearly increasing to −9.9 m s −1 at the top of the boundary layer.The v value was kept constant to −3.8 m s −1 .We used a subsidence rate (w s ), which linearly increased from 0 to −0.5 cm s −1 to about 2.2 km and was constant from this level to 4 km, then decreased linearly to zero at the TOA.The large-scale heat forcing was kept constant at a value of −2.5 K day −1 , and the moisture forcing profile increased from −1 g kg −1 day −1 close to the surface to 0.3456 g kg −1 day −1 at about 3 km and was fixed at that value throughout the free troposphere.The driving conditions were created by averaging observations over 16 December 2004 to 8 January 2005.
For the FixHydro case, an observed N d value of 70 cm −3 was used (vanZanten et al., 2011).For the ObsAero case, the aerosol mass mixing ratios of the three modes were diagnosed from the number mixing ratio and two log-normal size distributions (Eq. 1) assumed to consist of SO 4 with dry density of 1.77 g cm −3 .
We also used a total number, mode radius, and geometric standard deviation 90 cm −3 , 0.03 µm, and 1.28 for the aitken mode; 150 cm −3 , 0.14 µm, and 1.75 for the accumulation mode.Coarse aerosol mass is assumed to be zero.This specification is recommended by vanZanten et al. (2011).

d. d.ARM95
The ARM95 included because it is the default case, which has long been included with CAM releases.It is also an example of continental convection, which is an important climate regime.The ARM95 case tests the deep convection scheme and to some extent the mixed-phase cloud processes.The case spans 18 July to 3 August 1995, and we used the full shortwave and longwave radiation.All the cases were run at the default time step of 1200 s and 30 vertical grid levels with 20 levels in the free troposphere.We carried out four simulations each for DYCOMSRF02, MPACE-B, and RICO and two simulations for ARM95.Results from each method and each case are discussed in the four sections below.

Results and discussion
a. a.DYCOMS RF02  values illustrate the difficulty of interpreting output from sequentially-split climate models.
Cloud base and cloud top were both slightly higher than observed yet entrainment was much smaller than observed.This suggests that the subsidence we imposed may be too weak.Surface precipitation is too weak when realistic N d is used.This could be due to excessive re-evaporation of precipitation below the cloud base.This is consistent with the fact that the ObsAero and FixHydro models have the highest below-cloud base evaporation of precipitation given by 5.85 × 10 −8 and 4.45 kg kg −1 s −1 respectively, while the Default and PrescAero have lower values (3.62 × 10 −8 , and 1.33 × 10 −8 kg kg −1 s −1 , respectively).
Figure 1a shows N d profiles of the different aerosol specification cases averaged over the last two hours of the simulation period.We have also included the 10 year July average N d profile of the corresponding 3-D CAM5 run in which N d values were extracted at the closest grid point to the DYCOMSRF02 location.All SCM cases showed higher N d values at the cloud base and slightly lower values at the cloud top.This is inconsistent with observations, which tend to show constant values throughout the cloud (e.g.Martin et al., 1994).The Default model showed the lowest N d values (an average of 33 cm −3 ).This is probably due to the zero aerosol initialization; aerosol in the run increased as the simulation progressed due to emission sources.The PrescAero case showed highest N d values (an average of 139 cm −3 ) and the highest total aerosol burden, while the obseAero case showed slightly higher N d values (an average of 74 cm −3 ) as compared to the observations (an average of 55 cm −3 ), even though it had lower aerosol burden.
The 3-D model N d values are as high as the PrescAero case; however, there is a shift of the whole profile towards the surface, suggesting a collapsed boundary layer.
Even though stratocumulus are non-convective clouds, shallow convection is triggered occasionally, with higher frequency in the Default case than the other cases.Introduction

Conclusions References
Tables Figures

Back Close
Full N d of DYCOMSRF02 case correlates well with the total aerosol burden.The PrescAero case has the highest aerosol burden resulting in high values of N d , while the zero-aerosol initialized Default case has the lowest.The ObsAero case has higher aerosol burden in the accumulation and aitken modes resulting in N d values slightly higher than observed.
Figure 2 shows the temporal evolution of the LWP pre and LWP post of the DY-COMSRF02 case.There is high variability of LWP during the first few hours in all cases, with the highest variability in the Default case.During the last two hours this case performed worst and showed low LWP due to low N d that caused clouds to precipitate out.The FixHydro and ObsAero cases showed good agreement as compared to the observational ranges.The PrescAero case had higher LWP due to higher N d values.
In summary, the DYCOMSRF02 case shows strong sensitivity to aerosol specification.In the Default case, detrainment from shallow convection is a major source of N d , which limits sensitivity to aerosol burden.In other cases, higher aerosol burden translates to higher droplet concentration.Introduction

Conclusions References
Tables Figures

Back Close
Full Table 2 shows the observed and modeled cloud-related variables averaged during the last four hours of the MPACE-B case.The variables are N i , N d , LWP, IWP, w e , z b , z i , and surf Pr.The N i values for the Default, PrescAero, and ObsAero cases are 0.4, 0.7, and 0.6 L −1 , respectively.All of these cases overestimated the observed N i value (0.16 L −1 ).Aircraft and ground based remote sensors observed the existence of boundary layer mixed-phase clouds, which contained liquid and ice and were capped by a weak inversion with a cloud top temperature of about −15 • C (Klein et al., 2009).However, except for the FixHydro case all simulations produced not liquid.This is because ice removes all supersaturated vapor (and liquid) when crystal numbers are too high.The FixHydro case showed reasonable LWP (133 g m −2 ) and w e (12.37 mm day −1 ) due to the realistic use of N d and N i ; however, it underestimated the IWP (0.63 g m −2 ) and overestimated z b (1783 m) and surf Pr (0.5 mm day −1 ).
Figure 3 shows MPACE-B profiles of liquid water content (LWC) and ice water content (IWC) including and excluding snow mass as a function of scaled height, before and after micro-physics.The dark-shaded region, light-shaded region, and black solid line depict the median value, the inner 50 %, and the outer 50 % envelope of the high frequency observed aircraft data respectively, from Klein et al. (2009).Before microphysics, a reasonable amount of liquid water is shown by the FixHydro case, while the other cases showed shallower cloud and smaller amounts of liquid water (Fig. 3a).After the execution of microphysics, except for the FixHydro case, the microphysics physics removed all the liquid water in the other three models, resulting in complete depletion of liquid water.All cases showed good agreement of IWC as compared to aircraft observations (Fig. 3b and c), with some overestimation of IWC by the FixHydro case.The microphysics slightly removed some IWC from the Default case but did not make any change to the three other cases (Fig. 3b and c).However, IWC consistes entirely of snow Introduction

Conclusions References
Tables Figures

Back Close
Full except for the FixHydro case, which showed some cloud ice before microphysics (Fig. 3d).
Figure 4 shows the N i profiles of the different cases averaged over the last four hours of the MPACE-B period.We have also included the 10 years October 2004 average N i profile values of the 20 min timestep, 30 levels, 3-D CAM run, and values extracted at the closest grid point to the MPACE-B location.Except for the FixHydro case all the other cases overestimated N i .Despite the difference in the aerosol burden, the Default, PrescAero, and ObsAero cases showed no sensitivity to the aerosol specification except for slightly higher N i values for the ObsAero case.Similarly, except for the FixHydro case, which had N d value of 50 cm −3 , all the other cases showed N d value of zero due to the complete depletion of liquid water by the microphysics discussed above.However, all the cases simulated cloud fraction well as compared to aircraft and remote sensing observation (Fig. 5).Reasonable cloud fraction yet zero cloud condensate is possible in CAM5 because cloud fraction is computed before microphysics and is unchanged by physical processes, while cloud mass is affected by subsequent processes.
There exist large uncertainties in the representation of the ice nucleation processes in climate models.In CAM, homogeneous and heterogeneous (deposition, condensation freezing, contact freezing, and immersion freezing) ice nucleation processes in the mixed-phase regime (−40 < T < −3 • C) are represented as follows.
Deposition/condensation freezing ice nucleation process is represented by the Meyers et al. (1992) empirical formulation, which only depends on temperature and saturation vapor pressure.Similarly, immersion freezing is prescribed using the formulation of Bigg (1953) and contact freezing on dust is represented using the formulation of Young (1974).Detailed literature of ice nucleation formulation and parameterization for cirrus and mixed phase clouds can be found in Gettelman et al. ( 2012).Introduction

Conclusions References
Tables Figures

Back Close
Full In our SCM simulation of MPACEB, N i did not show any sensitivity to aerosol specification.This is due to the dominance of the Meyers et al. (1992) deposition/condensation freezing ice nucleation, which does not use explicit aerosol information but only depends on an empirical formulation using temperature and saturation vapor pressure.The other ice nucleation processes did not produce any N i .The Meyers deposition/condensational freezing depleted all the liquid to form overestimated N i regardless of the aerosol specification.As a result, activation did not produce any liquid droplets due to the total liquid water depletion.

c. c.RICO
Table 3 shows the averages of N d , SHF, LHF, Cloud Base Mass Flux (CBMF), Cloud Cover (CLC), and LWP during the last four hours of the 24 h simulation of the RICO case for the four model simulations and from vanZanten et al. ( 2011) LES results.We use LES as a proxy for truth here because this case is idealized and thus not comparable to observations from any particular time.All the model runs from this study showed similar N d , SHF, LHF, CBMF, CLC, and LWP values when compared to one another.The Default, PrescAero, and ObsAero cases showed an average N d value of 51 cm −3 , which slightly underestimated the LES value (70 cm −3 ), which is a best estimate of an average value from flight measurements using the Fast Forward Scattering Spectrometer Probe (FFSSP) during four flights, with measurements ranging from 50 to 100 cm −3 (vanZanten

GMDD Introduction Conclusions References
Tables Figures

Back Close
Full However, the N d value is underestimated as compared to observations (70 cm −3 ).The ObsAero case showed the lowest N d (about 14 cm −3 ).At the cloud top, except for the FixHydro case, all the SCM cases showed N d values approaching the no aerosol case (black line).The no aerosol case was run without aerosols to estimate the N d values due to convective detrainment.
The N d of the RICO case at the cloud base correlates well with the different aerosol burden values in the different modes.The PrescAero and Default cases have comparable N d values due to high aerosol burden.The ObsAero case shows the lowest N d values at the cloud base due to its low aerosol burden.Hence, the activation process is dominant at the cloud base in creating the droplets.However, at the cloud top, despite the differences in the aerosol specifications, the N d values did not change.Thus activation is the dominant process at the cloud base while detrainment dominates at the cloud top.
Vertical structure of the cloud mass flux and condensate is important for studying the parameterization of clouds and precipitation.Shallow convective mass flux maximizes near the cloud base and decreases with height, consistent with observations (Siebesma et al., 2003).However, the mass flux at the cloud base is overestimated in all the cases (Fig. 8a).Unsurprisingly, the condensate profile also shows overpredicted condensate at the cloud base and decreases with height (Fig. 8b).
The total cloud fraction is also overestimated as compared to LES (Fig. 9).At cloud base the overestimation due to both shallow convective and stratiform clouds.Modeled cloud extends further into the troposphere than observed

GMDD Introduction Conclusions References
Tables Figures

Back Close
Full This case spans 3 different weather regimes.Due to the existence of a largescale stationary upper-level trough over the continental US during the first ten-day period, there existed variable cloud cover and precipitation every other day.There followed a 3 day period of high pressure and clear skies, and the final 7 days consisted of stormy weather with high cloud cover and intense precipitation.
Figure 10 shows the time series of LWP and IWP for the Default and PrescAero cases.The time series of the LWP observations are also plotted from Xu and Randall (2000).Generally, SCAM over estimated LWP at all periods.Both runs showed comparable LWP, IWP, and surface precipitation (Fig. 10) as well as N d (Fig. 11).Aerosol optical depth in the visible range was 0.163 for PrescAero and only 0.081 for the Default case, however, indicating that the ARM95 case is insensitive to aerosol specification.As noted above, this result is not surprising since CAM's convective schemes do not use aerosol information.More surprising, however, is the fact that N d for the SGP region from both SCM and GCM simulations is ∼ 25 cm −3 , a factor of 8 smaller than typically observed in this region (e.g.

GMDD Introduction Conclusions References
Tables Figures

Back Close
Full Iacobellis and Somerville, 2006).This is a major bias in cloud properties which likely has significant negative effects on climate simulations.

Summary and conclusions
In this study we identified a problem with SCAM5 in its default configuration and introduced fixes to the identified problem.We used three new aerosol specification methods in our SCM simulations.The aerosol cases considered are Default case (with prognostic aerosol, initialized to zero), PrescAero case (with monthly climatological aerosol values), ObsAerosol case (with aerosols from observations), and the FixHydro case (with fixed droplet and ice crystal concentrations).We use SCM simulations for a variety of cloud regimes.The sites used for these studies include summertime mid-latitude continental convection (ARM95), shallow convection (RICO), subtropical drizzling stratocumulus (DYCOMSRF02), and multi-level Arctic clouds (MPACE-B).The DYCOMSRF02 case shows strong sensitivity to aerosol specification.Activation dominates over convective detrainment so a number of droplets are formed when you have higher aerosol burden.Convection does occur in all runs, however, and convective detrainment is source of N d in all cases, regardless of the aerosol specification.Default aerosol treatment in DYCOMSRF02 produced greatly underestimated N d and LWP.All proposed fixes substantially improve N d and LWP.
In MPACE-B, N i was too large and was insensitive to aerosol specification in all cases except FixHydro.This is due to the dominance of the Meyers et al. (1992) deposition/condensation freezing ice nucleation, which does not use aerosol information but only depends on empirical formulation using temperature and saturation vapor pressure.The other ice nucleation processes did not produce any N i .The Meyers deposition/condensational freezing was also too strong, causing all supersaturated vapor to freeze.This resulted in zero LWP for all cases except FixHydro, which had LWP value of 30 g m −2 (in agreement with observations).Introduction

Conclusions References
Tables Figures

Back Close
Full The RICO case did not show sensitivity to aerosol specification except at the cloud base where activation dominates and more droplets are formed as the aerosol burden increases.At the cloud top, convective detrainment is the dominant source of droplets, and regardless of the aerosol burden the number of droplets is similar.Detrainment seems to be too strong near cloud base, resulting in profile with too much cloud near cloud base and too little above.
The deep-convection ARM95 case also did not show any sensitivity to aerosol specification.Droplet number for both SCM and GCM runs at ARM95 were consistently 25 cm −3 , which is much lower than expected over land.This indicates a problem with aerosol specification in this region.
In summary, stratiform cloud cases (DYCOMS RF02 and MPACE-B) were found to have a strong dependence on aerosol concentration, while convective cases (RICO and ARM95) were relatively insensitive to aerosol specification.This is perhaps expected because convective schemes in CAM5 do not currently use aerosol information.Adequate liquid water content in the MPACE-B case was only maintained when ice crystal number concentration was specified because the Meyers et al. (1992) deposition/condensation ice nucleation scheme used by CAM5 greatly overpredicts ice nucleation rates, causing clouds to rapidly glaciate.Surprisingly, predicted droplet concentrations for the ARM95 region in both SCM and global runs were around 25 cm −3 , which is much lower than observed.This finding suggests that CAM5 has problems capturing aerosol effects in this region.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full  Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | nudged u and v below 700 hPa, values u and v were kept constant at the observed values of −13 m s −1 and −3 m s −1 (respectively) in our study.Surface latent and sensible heat flux values of 107.7 and 136.5 W m −2 , respectively, were used and were kept constant throughout the simulation period.Klein et al. (2009) specified a vertical velocity pressure (omega) value greater then zero at the top of the atmosphere (TOA), which causes huge advective heating from the top of the model, causing the model to crash in a few time steps.For this study we replaced the omega values from Klein et al. (2009) above 500 hPa with values that exponentially decrease Discussion Paper | Discussion Paper | Discussion Paper | c. c.RICO caseShallow Convection is another important climatological regime.The RICO experiment was conducted on the upwind side of the Islands of Antigua and Barbuda during the winter of 2004 when trade winds cover the northwestern Atlantic Ocean(Rauber et al., 2007).Unlike previous experiments, namely the Atlantic Trade Wind Experiment (ATEX) and Oceanographic and Meteorological Experiment (BOMEX), which did little to measure clouds and precipitation, RICO has extensive cloud-related measurements, which make it an important study for shallow cumulus clouds and precipitation.For this case we tried to set up our case similar to vanZanten et al. (2011).The assumptions made for the RICO case are discussed below.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | balanced.Like vanZanten et al. ( Advective forcing was generated by the State University of New York (SUNY) objective analysis method and the surface fluxes were estimated using Doran et al. (1998) surface analysis technique by the Simple Biosphere (SiB2) model (Ghan et al., Discussion Paper | Discussion Paper | Discussion Paper | 2000).For this case we only simulated the Default and PrescAero cases because N d /N i , and aerosol concentration are unknown.
depletion acting over an inappropriately long time step.We also include cloud base z b computed by interpolating the level where cloud fraction first rises about 0.5 and cloud top height z i computed by interpolation the highest level where the total water mixing ratio drops below 8 g kg −1 .Cloud top entrainment velocity w e = δz i /δt − w s was also computed.The Default case underestimated the observed N d (which was 55 cm −3 ), while ObsAero and particularly PrescAero overestimated N d .As expected, runs with higher N d tend to precipitate less and as a result have higher LWP.LWP computed before microphysics is too high except for the Default case.Values after microphysics show more variability, with the Default case being too low and the FixHydro and PrescAero being too high.Difference between pre-and post-microphysics Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |Detrainment from this convection is a major source of N d in some simulations.This occurs because CAM5 detrains droplet numbers according to a fixed droplet mean volume radius assumption rather than considering the actual droplet or aerosol availability.As a result, the convective detrainment from cloud top increased the in-cloud N d values.In order to separate the number of droplets generated by activation and convective detrainment we conducted another set of sensitivity experiments where vapor rather than condensate is detrained from convection.N d profiles from these experiments are shown in Fig.1b.This figure reveals that almost all the droplets in the Default case are created by convective detrainment due to zero aerosol initialization.In the PrescAero and ObsAero cases activation dominates, though detrainment increases the total N d in all cases, especially near the cloud top.
Discussion Paper | Discussion Paper | Discussion Paper | b. b.MPACE-B Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | el al., 2011; Brenguir et al., 1998).On average all the runs overestimated the LHF (12.7 W m −2 ), LHF (207.9W m −2 ), and CBMF (0.06 m s −1 ) as compared to the LES value (8.5 W m −2 , 158 W m −2 , 0.026 m s −1 ), respectively.All the models simulated CLC (0.18), and LWP (19.4 g m −2 ) very well as compared to LES, (0.19) and (19 g m −2 ), respectively.The time series of the LWP shown in Fig. 6 also depicts high variability during the spin-up period and good agreement with LES after 15:00 UTC for all models.
Figure7shows the N d profiles of all the cases of this study averaged over the last 4 h of the simulation period.We have also included the 10 years July average N d profile values of the 20 min timestep and 30 levels 3-D CAM extracted at the closest grid point to the RICO location.The PrescAero, the Default and 3-D has similar values of N d at the cloud base where the statiform cloud was present.
Discussion Paper | Discussion Paper | Discussion Paper |because deep convection is being triggered and the runs showed deeper clouds due to the deep convective cloud fraction (Fig.9).Non of the runs show sensitivity of mass flux, condensate, and cloud fraction to aerosol specification.In summary, the RICO runs did not show sensitivity to aerosol specification except at the cloud base where activation dominates and more droplets are formed as the aerosol burden increases.At the cloud top, detrainment is dominant and regardless of the aerosol burden the N d profiles are similar.d.d.ARM95The last case is based on ARM SGP site and spans 17 days starting 18 July to 4 August 1995.It was chosen because it is the default SCM case distributed with CAM5.This case is the basis of theGhan et al. (2000) SCM intercomparison.Only the Default and the PrescAero cases are simulated due to lack of observed N d , N i and aerosol data.

Figure 1 .Figure 2 .
Figure 1.Profiles of in-cloud droplet number concentrations (N d ) for DYCOMSRF02.3-D CAM values are 10 years July average global CAM extracted at the location of DYCOMSRF02.(a) Convective detrainment turned on (b) convective detrainment turned off.

Figure 4 .Figure 6 .Figure 10 .
Figure 4. Profiles of in-cloud N i values for MPACE-B case.3-D CAM values are 10 years July average global CAM extracted at the location of MPACE-B.Note: N i values (3-D CAM N i are divided by 10 to fit in the plot).
microphysics call.We then set N d and N i tendencies inside the microphysics to zero, which keeps the value of N d and N i to their corresponding prescribed values.2.The second method (hereafter called PrescAero) uses the new prescribed aerosol capability included in Community Earth System Model (CESM) version 1.2.

Table 1
Gettelman et al. (2014)led cloud-related variables averaged during the last two hours of the six hour DYCOMS RF02 simulations.In addition to N d , and surface precipitation (Pr), we include the liquid water path both before and after microphysics was called (LWP pre and LWP post , respectively).These values are different because CAM5 sequentially updates the model state after each parameterization is applied.As described inGettelman et al. (2014), LWP pre is often much bigger than LWP post because sequential updating leaves microphysical