Towards Parameterization of Atmospheric Aerosol in Regional Forecast Models

1
Towards Parameterization of Atmospheric Aerosol
in Regional Forecast Models

• David B. Mechem, and Yefim L. Kogan
• Collaborators Yi Lan, Paul Robinson, Yuri
  Shprits

The University of Oklahoma
Seminar presented at NRL, Monterey, 7 January 2005
Acknowledgements. This research was supported by
the Office of Naval Research and the U. S.
Department of Energy Atmospheric Radiation
Measurement Program.
2
Aerosol dramatically influences the radiative
characteristics of PBL clouds
First indirect effect cloud droplet radius and
concentration influences albedo gt ship tracks
3
Ship tracks in the eastern Atlantic
Photo credit Robert Wood, University of
Washington
Photo credit Robert Wood, University of
Washington
4
Aerosol affects the thermodynamic structure and
persistence of PBL clouds
Second indirect effect drizzle may lead to cloud
breakupgt Pockets of Open Cells (POCs)
5
POCs
Photo credit Robert Wood, University of
Washington
Photo credit Robert Wood, University of
Washington
6
Regional simulations of aerosol-cloud-drizzle
interactions using the COAMPS mesoscale model
coupled with the CIMMS drizzle parameterization
7
Model setup

• COAMPS v2.0.14
• 18/6/2 km grid. Vertical grid spacing stretched
  from 10 to 800 m
• 1.5-order subgrid closure (Level 2.5 Mellor and
  Yamada 1982)
• 24 h simulation. Two 12 h cycled pre-forecasts
  establish a reasonable boundary layer structure
• Bulk drizzle parameterization (Khairoutdinov and
  Kogan, 2000)
• Prognostic equations for qc, Nc, qr, Nr, and NCCN
• Initial and boundary condition CCN value of 45
  cm-3

Goal
Compare drizzling (KK) and nondrizzling (ND) runs
to evaluate the effect of drizzle on a mesoscale
forecast.
8
1800 UTC COAMPS LWP comparison of 5-moment
drizzle parameterization (KK) and the operational
(Kessler) microphysics scheme (18 km grid)
5-moment scheme (KK)
Operational microphysics
1
1
2
2
LWP g m-2
3
3
Three significant improvements of KK drizzle
scheme
Reduced entrainment from drizzle-stabilization
leads to a further northern extent of cloud wedge
1
Reduction in LWP and cloud coverage south and
east of Point Conception.
2
3
Open oceanic LWP is better match with climatology
9
1800 UTC COAMPS LWP comparison of 5-moment
drizzle parameterization (KK) and the operational
(Kessler) microphysics scheme (2 km grid)
A1-A1' cuts across banded cloud structures with
weak resolved vertical velocity. We take these
bands to represent ensembles of PBL cumulus.
Significant improvements from the KK drizzle
scheme, inferred by LES results

• More realistic cloud base structures and
  variability
• Improves ability of COAMPS to represent broken
  PBL cumulus fields
• Represents the transition from unbroken
  stratocumulus to PBL cumulus

10
Comparison of surface bulk CCN concentration
using the 5-moment drizzle parameterization (KK)
1200 UTC 25 July
12 h later
cm-3
11
In these previous COAMPS simulations, aerosol
characteristics were represented by a single
parameter NCCN.
Attempting to distill aerosol characteristics to
a single parameter often gives an incomplete and
sometimes incorrect portrayal of aerosol
properties
12
Effect of coarse mode/giant aerosols
Background sulfate only
Background sulfate plus giant aerosol

• Giant aerosol above the inversion
• Enhance drizzle production
• Attenuate PBL turbulence
• Accelerate stratocumulus breakup

When pollution above the inversion is
predominantly fine-mode, drizzle production is
suppressed.
13
Effects of surface windssea salt aerosols

• The presence of sea salt results in
• Significant drizzle formation
• Reduction in mean drop concentration
• Large variations in cloud base
• Greater variability in cloud top
• More complex internal cloud structure
• Significant differences in overall cloud geometry
  implying possible future breakup of cloud field

14
Susceptibility of cloud drop concentration to
sea-salt addition S(Nss-N)/N

• 3 LES simulations
• cleanlow background concentration
• polluted with low and high Aitken nuclei
  concentrations

Sea salt effect depends on the sulfate aerosol
concentration, N When N is low, the effect of
sea-salt is to significantly increase cloud drop
concentration. When N is high, the effect
depends on the concentration of Aitken nuclei
15

• Advanced prediction of aerosol-cloud-drizzle
  feedbacks should include 3 main aerosol
  parametersCoarse mode (giant)
  aerosolsBackground (fine mode) sulfate
  aerosolsAitken nucleiandParameterization of
  the effects of surface winds sea-salt aerosols

16
Full system of equations describing coupled
aerosol-cloud interactions

• Equations for cloud drop parameters (4 equations
  in KK approach) need to be complemented by 3
  equations for major aerosol parameters

17
Cloud microphysics formulation
18
Prediction of Aerosol Parameters
i1,2,3
Si,ccn represents (interstitial) source and sink
terms of aerosol, e.g. transformation,
sedimentation, production from DMS, sea-spray
Parameterization of cloud parameter conversion
rates (for example)
Parameterization of aerosol-aerosol,
aerosol-cloud conversion rates yet TBD
19
What components are required for an accurate
mesoscale forecast of aerosol-cloud-drizzle
system?

• Specification of aerosol field (initial and
  boundary conditions)
• Size characteristics
• Spatial distribution

• Specification of sources and sinks
• Urban sources
• Sea salt
• Heterogeneous chemistry
• Transformation rates (fine?coarse mode)

• Transport
• Advection
• Sedimentation
• Turbulent mixing (entrainment)

• Cloud processing
• Activation
• Coagulation, rainout, diffusiophoresis
• Regeneration

20
Where we are now?

• Specification of aerosol field
• Observations and data assimilation necessary for
  all 3 aerosol parameters

• Specification of sources and sinks
• Simple parameterizations exist for sea-spray
  aerosol source
• Parameterizations of aerosol transformation rates
  have yet to be developed

• Transport
• As accurate as the models advection scheme
• Depends on how accurately the model SGS
  represents entrainment

• Cloud processing
• Processing via coagulation represented in some
  cloud physics schemes
• Recent activation parameterizations not yet
  linked to model SGS energetics

More understanding is needed of the relative role
and importance of these various processes,
sources, and sinks.
21
Example of parameterization/formulation of cloud
processing

• Control experiments with different initial CCN
  concentrations
• Sensitivity runs with various CCN source
  mechanisms and magnitudes

22
Model setup idealized

• ?x ?y 2 km ?z 25 m ?t 10 s
• Domain size 100?100?1.5 km
• Periodic horizontal boundary conditions
• Imposed large scale divergence 5.0?10-6 s-1
• Sensible and latent heat fluxes (10 and 25 Wm-2)
• Longwave only
• KK bulk drizzle parameterization
• Activation by Martin et al. (1994) and ODowd et
  al. (1996)
• Thermodynamic initial conditions from ASTEX A209
• Various initial CCN profiles and magnitudes

23
Time-height representation of qc and Nt
24
Statistics for different initial CCN
concentrations

• Smaller values of CCN result in
• Reduced entrainment and lower mean cloud top
  height
• Higher mean cloud base
• Reduced mean LWP
• Larger drizzle rates
• Increased variability

25
COAMPS aerosol budget
PBL aerosol budget is calculated in terms of
total particle concentration (CCN droplet)

• Calculate entrainment term from change of
  inversion height and magnitude of imposed
  divergence
• Any additional source/sink terms are known (i.e.
  imposed)
• ? We can back-out the cloud processing rate

26
Cloud processing for two different CCN
concentrations
27
Sensitivity experiments Entrainment source

• Assume NCCN 200 cm-3 for z lt zi, but various
  concentrations of free-tropospheric CCN at z gt zi
  .
• As PBL entrains free tropospheric air, this CCN
  is mixed down into the boundary layer

• When entrained into the PBL, free tropospheric
  CCN can
• Suppress drizzle
• Counteract depletion via cloud processing
• Increase PBL Nt, given sufficient entrainment and
  free tropospheric CCN concentration

28
Validation and parameterization of cloud
processing

• Cloud processing (depletion) is correlated to
  drizzle rates by simple power laws and largely
  independent of initial conditions
• Depletion can also be related to other model
  parameters (e.g. Nc, not shown)
• These relationships might serve as nexus of
  aerosol-cloud interactions in large-scale models

• Validation of COAMPS cloud processing
• Results from LES show similar behavior
• Hoell et al. (2000) give larger cloud processing
  for given drizzle rates
• Albrechts estimate (1989) is for a
  strongly-drizzle, highly-depleting example

29
Summary of COAMPS cloud processing results

• Results respond predictably to changes in initial
  CCN
• Idealized COAMPS runs gauge the relative
  importance of various components of a mesoscale
  aerosol forecast
• Magnitude of the entrainment source is greater
  than any reasonable values of in-situ or surface
  sources yet we know that sea-spray can play a
  vital role in PBL clouds
• Specification of vertical aerosol profile and
  species may be more vital than detailed knowledge
  of in-situ source rates Importance of remote
  sensing.

30
Conclusions

• The general requirements of how to treat
  aerosol-cloud-drizzle interactions are becoming
  clear
• Absolute magnitudes of sources/sinks are poorly
  constrained
• Major effort to estimate these quantities and
  develop parameterizations, either from
  observations or process models (LES, CRM)
• Aerosol-cloud parameterization could be
  implemented gracefully (?) into the COAMPS
  aerosol-tracer module

31
What components are required are required for an
accurate mesoscale forecast of cloud-aerosol
system?

• Specification of aerosol field (initial and
  boundary conditions)
• Size characteristics
• Spatial distribution

• Specification of sources and sinks
• Urban sources
• Sea salt
• Heterogeneous chemistry
• Transformation rates (fine?coarse mode)

• Transport
• Advection
• Turbulent mixing (entrainment)

• Cloud processing
• Activation
• Coagulation, rainout, diffusiophoresis
• Regeneration

32
(No Transcript)
33
Improvement of cloud physics parameterization in
NWP Parameterization of Sub Grid Scale (SGS)
processes

• Closing the scale gap
• Cloud physics processes scale 100 m
• NWP model grid 1-10 km

34
Sub-grid scale condensation in COAMPS
Top GOES-9 visible imagery of San Francisco Bay
region Middle Control simulation (no SGS
parameterization) Bottom Forecast with the SGS
condensation parameterization. Satellite
imagery shows nearly complete clearing by 17 UTC.
The control simulation remains cloudy until
2000 UTC. The Bay area is nearly cleared of
cloud by 1800 UTC in the SGS condensation
simulation
1
1
2
2
3
3