MESOSCALE NUMERICAL PREDICTION OF AEROSOLS AND DUST

1
MESOSCALE NUMERICAL PREDICTION OF AEROSOLS AND
DUST

• William Cotton, Elizabeth Zarovy, Gustavo Carrió,
  Steve Saleeby, Hongli Jiang, David Stokowski ,
  and Susan van den Heever
• Colorado State University,
• Dept. of Atmospheric Science
• Fort Collins, Colorado

2
Aerosol Impacts on DOD Operations

• Directly by visibility obscuration in dust
  storms.
• Indirectly by visibility alterations in fogs and
  clouds.
• Indirectly by altering cloud cover.
• Indirectly by affecting precipitation processes.
• Indirectly by impacting aircraft icing conditions.

3
Microphysical Processes Represented in RAMS

• Cloud droplet nucleation in one or two modes
• Ice nucleation
• Vapor deposition growth
• Evaporation/sublimation
• Heat diffusion
• Freezing/melting
• Shedding
• Sedimentation
• Collisions between hydrometeors
• Secondary ice production

4
Hydrometeor Types

• Cloud droplets
• Rain
• Pristine ice (crystals)
• Snow
• Aggregates
• Graupel
• Hail

5
RAMS Liquid Hydrometeor Distributions
6
Unique Features of RAMS Microphysics

• Uses generalized gamma distribution basis
  functions
• where n(D) is the number of particles of
  diameter D, Nt is the total number of particles,
  ? is the shape parameter, and Dn is some
  characteristic diameter of the distribution. The
  Marshall-Palmer (exponential) and Khrgian-Mazin
  distribution functions are special cases of this
  generalized function.
• Simulations can be done with one or two moments.
  When two-moments of a hydrometeor class is
  predicted, all that is needed to completely
  specify the distribution function given by (1) is
  the specification of ?.

7

• Collection is simulated using stochastic
  collection solutions rather than continuous
  accretion approximations. Owing to the use of
  look-up tables, it became apparent that it is no
  longer necessary to constrain the system to
  constant or average collection efficiencies. Thus
  the formerly ad hoc auto-conversion formulations
  in RAMS was replaced with full stochastic
  collection solutions for self-collection among
  cloud droplets and for rain (drizzle) drop
  collection of cloud droplets. This approach is
  being extended to all hydrometeor interactions.

8

• The philosophy of bin representation of
  collection was also extended to calculations of
  drop sedimentation. Previously, bulk microphysics
  schemes have treated sedimentation of
  hydrometeors by integrating over the entire
  particle size-spectra and obtaining a
  mass-weighted fall speed. Bin sedimentation is
  simulated by dividing the gamma distribution into
  discrete bins and then building look-up tables to
  calculate how much mass and number in a given
  grid cell fall into each cell beneath a given
  level in a given time step.

9
Donor cell
Bin Computations
Sedimentation
Donor cell
10
Cloud Droplet Nucleation
Number nucleated obtained from lookup table as a
function of
CCN number concentration Vertical
velocity Temperature
Lookup table generated previously (offline) from
detailed parcel-bin model Nc1Nccn Nc2Ngccn
Sw gt 0.0
11
Ice Habits

• Pristine ice and snow are allowed to have any of
  five different habits (shapes) columns,
  needles, dendrites, hexagonal plates, and
  rosettes. The dependence of mass and of fall
  velocity on diameter are different for each
  habit.

12
Ice Crystal Nucleation
1. Deposition nucleation Condensation freezing
Ni NIFN exp 12.96 (S - So) So 0.4
T lt -5oC rv gt rsi (supersaturation with respect
to ice)
T lt -2oC rv gt rsl (supersaturation with respect
to liquid)
CCNIFN
vapor
13
Ice Crystal Nucleation -- continued
2. Contact nucleation
T lt 0oC
3. Homogeneous freezing of cloud droplets T lt
-30oC
4. Homogeneous freezing of haze Rate
depends on T, rv, amount of haze present
14
Variations in precipitation and cloud optical
properties due to changes in aerosol
concentrations

• Summer storms over Florida
• 25 variation in the total volumetric
  precipitation received at the surface
• 52 variation in optical thickness
• Winter storms
• CO 15-30 snow
• NE USA 10-30 rain 30-50 snow

15
Implementation of aerosol-activated microphysics
in mesoscale and larger-scale models

• Aerosol activation requires explicit prediction
  of cloud-scale vertical velocities.
• Moreover, collection processes require
  cloud-scale LWCs and concentrations.
• Mesoscale and larger-scale models predict only
  grid-averaged quantities that are often one two
  orders of magnitude smaller than cloud-scale
  quantities.

16
Parameterization of Cloudy Boundary Layers using
a PDF Approach

• Represent the subgrid-scale variability of
  vertical velocity, temperature, and moisture
  using a joint probability density function (PDF)
• Use double Gaussian family of PDFs proposed by
  Larson et al. (2002). This family depends on 10
  parameters. It was tested against aircraft
  measurements and outputs from large-eddy
  simulations.
• The PDF is used to diagnose cloud fraction,
  liquid water, as well as close all higher-order
  moments.
• PDF parameters are determined from the predicted
  moments

17
A PDF-based Parameterization

• The parameterization was used to simulate a
  variety of boundary layer regimes.
• Wangara dry convective layer.
• BOMEX trade-wind cumulus clouds.
• ARM cumulus clouds over land.
• FIRE nocturnal stratocumulus clouds.
• ATEX cumulus rising into broken stratocumulus.
• Overall, the parameterization produced results
  that compared favorably with large-eddy
  simulations. This was accomplished without the
  need for any case-specific adjustments.

18
Interfacing the PDF Single Column Model in the
3-D RAMS Model
PDF 9 Prognostic Variables
Update PDF/RAMS Variables Liquid potential
temperature Cloud water content Total water
content Cloud fraction Sub-grid TKE Eddy mixing
coefficients (Ks) Cloud droplet concentration
PDF Input Variables Sensible latent heat
flux Momentum fluxes Liquid potential
temperature Exner function for pressure Cloud
water content Total water content X, Y, Z winds
Sub-grid TKE, PDF Gaussian weighted parameters
Updated PDF/RAMS variables with
sub-grid contribution ----------------------------
- Sub-grid TKE eddy mixing coefficients --------
--------------------- Updated sub-grid CCN
activation and cloud droplet nucleation from
Microphysics/PDF interface
Computation of Vertical Diffusion from PDF
sub-grid K Coefficients
RAMS 3-D Mesoscale Model
Single Column PDF Model
PDF input variables
Compute save PDF 9 Prognostic Variables
Computation of Horizontal Diffusion Coefficients a
nd tendencies
Update RAMS scalar variables with diffusion
tendencies before next timestep
19
Interface the PDF model with microphysics

• Our cloud microphysics parameterization requires
  vertical velocity (adiabatic cooling) to activate
  CCN, GCCN, and IFN.
• Using the bulk microphysics with its CCN
  activation parameterization, we performed LES
  simulations of four observed cases (ARM, ATEX,
  BOMEX, FIRE) with an initially uniform CCN
  spectra. The simulated cloud was sampled to
  obtain an average droplet concentration at
  roughly 50 m above average cloud base. Given the
  CCN spectra and the computed , we
  determined from the CCN activation scheme the
  corresponding w-scale that yielded a droplet
  concentration similar to the average droplet
  concentration at that level diagnosed from the
  LES data. PDFs of w were then extracted from the
  LES output at 40-50m above cloud base to find out
  the w-moment that will yield .
• The w-scale was found to be related to the
  root-mean-square (rms) of the second moment
  through a nearly constant coefficient. The
  expression is given by
• with C0.226 derived from analysis of the four
  LES simulations.

20
Number concentration of droplets activated from
RAMS activation scheme.
21
Latin Hypercube Sampling (LHS)

• Like a Monte Carlo model, LHS permits sampling
  only the cloudy part of a grid box which will
  provide the sought for characteristic w, LWCs, S
  needed to activate cloud droplets and ice
  crystals and drive collection processes.

22
An Aerosol Source model

• Direct emission sources include desert dust, sea
  salt, natural and anthropogenic sources of SO2
  and sulfates, certain strains of bacteria, forest
  fires and biomass burning.
• Dust emissions are based on surface topographic
  features, near-surface winds, erodible area, and
  soil moisture.
• Sea salt emissions are based on near-surface
  winds.

23
Model Development

• After implementing the aerosol source functions
  the model will be tested using satellite and
  ground-based measurements for comparisons.
• After testing, parameterizations will be
  implemented converting the aerosol distributions
  to CCN, GCCN and IFN distributions for input in
  the microphysics package.

24
CCN

• Parameterizations for conversion of SO2 and
  sulfate concentrations to CCN are being
  constructed. This includes oceanic DMS production
  and conversion to CCN, and smoke conversions to
  CCN.
• At this time chemistry will be parameterized by a
  conversion factor from SO2/DMS to sulfate rather
  than assuming background oxidant concentrations
  and explicitly modeling chemistry.

25
GCCN

• The primary sources of GCCN are dust, sea salt,
  industrial (like paper pulp mills), diesel
  engines, and perhaps biogenic material.
• Parameterizations of conversion from these
  sources are being constructed.

26
IFN

• Primarily sources of IFN are inorganic soil
  particles like clays (correlated with dust),
  biogenic materials, and industrial sources like
  heavy metal industries.
• Conversion rates from dust to IFN has to be
  derived.
• Biogenic sources include decay of plant litter
  over land and blooms of phytoplankton over the
  oceans.

27
IFN sources and Climate Zones

• A-zone tropical (extensive vegetation)
• B-zone dry
• C-zone humid mesothermal (extensive vegetation)
• D-zone microthermal (extensive vegetation)
• E-zone polar
• H-zone undifferentiated highlands
• M-zone marine

28
From Schnell and Vali, 1976.
29
Leaf-derived nuclei

• Koppen Climate Zones
• Extensive vegetation A-tropical C-mild mid-lat
  D-cold mid-lat
• Minimal vegetation B-dry E-polar H-highlands

30
IFN and climate zones

• A-type low
• B-type somewhat larger than A
• C-type highest
• M-type variable depending on T and
  upwelling/mixinghigh values near Australian and
  Antarctic coasts are regions of upwelling and
  continental shelf where plankton are highest in
  the world.

31
CCN/GCCN/IFN Retrieval Scheme

• Objective
• Examine the feasibility of retrieving
  cloud-nucleating aerosol concentrations using the
  cloud resolving version of RAMS in an ensemble
  Kalman filter assimilation system.

32
Brief Algorithm Description

• Identify locations where the CCN, IFN and GCCN
  concentrations wish to be estimated.
• Do backward trajectory analyses using RAMS (or
  another mesoscale model), in regions where
  boundary layer clouds are prevalent.
• Produce ensembles of CRM forward integrations
  along those trajectories in which we apply
  ensemble Kalman-filter assimilation procedures.
• Cloud-nucleating aerosol concentrations are
  retrieved at the end of the forward integrations.

33
Brief Algorithm Description (continued)

• Forward integrations are performed with a CRM
  version of RAMS that uses the explicit cloud
  nucleating microphysical module.
• The trajectories generated by the mesoscale model
  define time-evolving boundary conditions for the
  CRM.
• Even when the CRM domain is 2-D, only horizontal
  averages are involved in the system state vector.
• Satellite radiances will be periodically
  assimilated into the CRM, minimizing the cost
  function between simulated and observed cloud
  radiances.

34
Summary and Conclusions

• A Case has been made for the importance of
  aerosol to DoD operations for predictions of
  visibility, laser-guided systems, and indirectly
  via their impacts on cloud coverage, cloud
  optical thicknesses, aircraft icing and
  precipitation.

35
Summary and Conclusions (cont)

• A modeling system has been described which can
  forecast both direct and indirect effects of
  aerosols.
• To initialize aerosols in models an aerosol
  source and transport/ removal model is under
  construction.
• In addition, a cloud-nucleating aerosol satellite
  retrieval model is under development.
• I predict that the next-generation mesoscale
  forecast models will include explicit prediction
  of aerosols and their direct and indirect effects.