Synergisms in the Development of the CMAQ and CAMx PM/Ozone Models

1
Synergisms in the Development of the CMAQ and
CAMx PM/Ozone Models
Ralph E. Morris, Greg Yarwood Chris Emery,
Bonyoung Koo ENVIRON International
Corporation 101 Rowland Way Novato, CA Presented
at CMAS Models-3 Users Workshop October 27-29,
2003 Research Triangle Park, NC
Presentsslides/
2
Introduction

• Numerous challenges in particulate matter
  modeling
• Multiple Components
• SO4, NO3, SOA, POC, EC, Crustal, Coarse, Other
• Multiple Processes
• Gas-, Aqueous-. Heterogeneous-, Aerosol-Phase
  Chemistry
• Rainout/washout, dry deposition of Gases and
  Particles
• Advections and Diffusion
• Clouds, Canopy, Terrain, etc.
• Numerous Uncertainties
• Chemistry (e.g., nitrate, SOA, aromatic, etc.),
  PM Size Distribution, Meteorology, Emissions,
  Measurements

3
Introduction

• CMAS Workshop Good Forum to Discuss Challenges,
  Approaches and Potential Solutions for Improving
  PM Modeling
• CMAS Workshop Theme Emphasizes the Common
  Challenges of PM Modeling
• One Atmosphere
• One Community
• One Model

4
One Atmosphere
5
One Community
6
One Model

• CMAQ

7
One Model?

• CMAQ
• MM5 RAMS WRF

8
One Model??

• CMAQ
• MM5 RAMS WRF
• SMOKE EMS EPS OPEM

9
One Model???

• CMAQ
• MM5 RAMS WRF
• SMOKE EMS EPS OPEM
• MOBILE NONROAD EDMS EMFAC AP42

10
One Model????

• CMAQ
• MM5 RAMS WRF
• SMOKE EMS EPS OPEM
• MOBILE NONROAD EDMS EMFAC AP42
• IMPROVE CASTNET STN AQS/AIRS NADP SuperSites

11
Multi-Model Intercomparisons

• Intercomparing models and alternative
  formulations is an integral part of model
  development
• Photochemical grid model development has taught
  us that much more can be learned from comparing
  different models with different formulations
  this is even more true for PM models due to more
  uncertainties in processes
• Early 1980s UAM vs. CIT
• 1990 UAM vs. CALGRID
• Early 1990s UAM-V vs. UAM vs. SAQM
• Mid 1990s UAM-V vs. CAMx vs. MAQSIP
• Early 2000s CMAQ vs. CAMx

12
Early CMAQ vs. CAMx Comparisons for Ozone

• 1991 Lake Michigan Ozone Study (LMOS) Databases
• Tesche ands co-workers (2001) (available at
  www.crcao.com as CRC Project A-25)
• MM5 and RAMS Meteorology
• No one model performing sufficiently better than
  another
• CMAQ and CAMx using MM5 more similar than CAMx
  using RAMS
• Similar ozone responses to VOC/NOx controls
• CMAQ using QSSA and SMVGEAR chemistry solvers
  takes 5 and 8 times longer to run than CAMx
• ? EPA implements faster Hertel/MEBI chemistry
  solver in CMAQ

13
Early CMAQ vs. CAMx Comparisons for Ozone

• July 1995 NARSTO-Northeast Ozone Episode
• Morris and co-workers (available at www.crcao.com
  as CRC Project A-24)
• MM5 and RAMS Meteorology
• Layer 1 KV mixing issues
• ? EPA implements 1.0 m2/s minimum KV in MCIP,
  land use specific lower layers minimum KV
  used with CAMx
• QSSA chemistry solver accuracy and stability
  issues
• ? Hertel/MEBI solver implemented in CMAQ
• Smolarkiewicz advection solver is overly
  diffusive.
• ? Smolarkiewicz removed from CAMx (not in CMAQ)

14
Early CMAQ vs. CAMx Comparisons for Ozone

• July 1995 NARSTO-Northeast Ozone Episode
• SAPRC97 chemistry more reactive than CB-IV
• ? Both CMAQ and CAMx implement SAPRC99 chemistry
  
• Different horizontal diffusion (KH) formulations
  in CMAQ and CAMx
• CMAQ inversely and CAMx proportional to grid
  spacing
• ? Area of future research and sensitivity tests
  (e.g., spawned BRAVO sensitivity test)
• MM5 convective activity potentially can produce
  modeling artifacts
• ? MM5 interface an area of continued research
  for CMAQ and CAMx

15
Emerging PM Model Development Issues

• Aqueous-Phase Chemistry
• High pH dependency of aqueous-phase O3SO2
  reaction
• Coarse and fine droplets may have different
  buffering and different pH effects on
  aqueous-phase sulfate formation
• Test this effect using PMCAMx sectional PM model
  that incorporates CMU VSRM aqueous-phase
  chemistry module
• October 17-19, 1995 Southern California PM
  episode
• Two aqueous-phase chemistry modules used
• CMU 1-section bulk module
• CMU 2-section VSRM module

16
Southern California Modeling Domain
17
VSRM (Multi-Section) vs. Bulk Aqueous
ChemistryPercent Increase in Sulfate ()
By second day, VRSM estimates 15-30 more
sulfate across the SoCAB with gt 50 increase
offshore and around Long Beach
18
VSRM (Multi-Section) vs. Bulk Aqueous Chemistry
VRSM can form significantly more sulfate than the
bulk 1-section
aqueous-phase chemistry module
19
Emerging PM Model Development Issues

• Conclusions on Bulk vs. Multi-Section
  Aqueous-Phase Chemistry Tests
• Multi-section aqueous-phase chemistry module
  made significantly more sulfate in the Southern
  California test case
• Due to low sulfate in Southern California,
  differences were not significant enough to
  appreciably affect sulfate model performance
• Need further testing for eastern US where higher
  sulfate concentrations occur
• Merging of CAMx4 and PMCAMx models provides
  platform for testing RADM and CMU 1-section bulk
  aqueous-phase chemistry modules against the CMU
  VSRM multi-section module
• CMU VSR multi-section module requires 5 times
  more CPU time than CMU 1-section module (Further
  optimization warranted)

20
Emerging PM Model Development Issues

• Aerosol Thermodynamics Gas/Particle Partitioning
• Gas/Particle equilibrium usually assumed
• ISORROPIA equilibrium scheme widely used
• Fast and reliable
• CMAQ, CAMx, URM, etc.
• Equilibrium assumption may not always be correct,
  especially for coarse particles
• PMCAMx sectional PM model includes three options
  for Gas/Particle partitioning
• Equilibrium (ISORROPIA)
• Dynamic (MADM)
• Hybrid (equilibrium for fine/dynamic for coarse
  particles)
• Testing using October 1995 Southern California
  Database

21
Equilibrium vs. Dynamic vs. Hybrid
22
Equilibrium vs. Dynamic vs. Hybrid
23
Emerging PM Model Development Issues

• Conclusions on use of equilibrium approach for
  gas/particle partitioning
• For Southern California application
• dynamic and hybrid modules produce nearly
  identical results
• most of the time equilibrium approach produces
  results very close to dynamic and hybrid
  approaches, but differences as high as 30 did
  occur
• dynamic (MADM) approach requires approximately 10
  times the CPU time as equilibrium approach
• Further tests of equilibrium assumption warranted
• Given sufficient accuracy, uncertainties and
  computational requirements, equilibrium approach
  appears adequate for annual modeling

24
Emerging PM Model Development Issues

• Particle Size Distribution
• Different representations of particle size
  distribution in difference models
• CMAQ modal approach using 3 modes and assumes all
  secondary PM is fine
• CAMx4, REMSAD and MADRID1 assume fine and coarse
  PM (all secondary PM is fine)
• PMCAMx, CMAQ-AIM and MADRID2 are fully sectional
  models where PM10 is divided up into N sections
  (e.g., N10)

25
Emerging PM Model Development Issues

• Particle Size Distribution
• Testing of assumptions of particle size
  distribution using new merged CAMx4/PMCAMX code
• M4 CAMx4 2 section plus RADM aqueous
• EQUI N sections equilibrium VRSM aqueous
• MADM 10 sections dynamic VRSM aqueous
• RADM/EQ 10 sections equil. RADM aqueous
• RADM/EQ4 4 sections equil. RADM aqueous
• October 17-18, 1995 Southern California Episode

26

• 24-Hour Sulfate (?g/m3)
• October 18, 1995
• M4 peak SO4 39 ?g/m3
• EQUI peak SO4 51 ?g/m3
• Long Beach Area
• Differences due to more sulfate production in
  CMU VRSM than RADM aqueous-phase chemistry
• Further downwind (Riverside) M4 produces more
  sulfate than EQUI

M4
EQUI
27

• 24-Hour Nitrate (?g/m3)
• October 18, 1995
• M4 peak NO3 83 ?g/m3
• EQUI peak NO3 54 ?g/m3
• Observed NO3 peak at Riverside 40 ?g/m3
• Differences partly due to assuming all nitrate
  is fine vs. PM nitrate represented by 10 size
  sections (EQUI)
• Differences in M4 RADM and EQU VSRM also
  contribute

M4
EQUI
28

• 24-Hour Nitrate (?g/m3)
• October 18, 1995
• M4 peak NO3 83 ?g/m3
• EQUI peak NO3 54 ?g/m3
• EQUI 10-Section grows PM NO3 into coarser
  sections where it dry deposits faster than M4 NO3
  that is assumed to be fine
• Result is less NO3 in downwind Riverside area
  that agrees better with observations

M4
29
Sensitivity to Number of Size Sections (10 vs. 4)
_at_ (34,16)
30
Computational Efficiency Model Configurations
31
Emerging PM Model Development Issues

• Nighttime Nitrate Chemistry
• September 2003 CMAQ release
• Zero N2O5H2O gas-phase reaction rate
• 0.02 and 0.002 probability for heterogeneous rate
• April 2003 CAMx4 release
• Keep gas-phase N2O5H2O reaction rate
• German smog tests provide upper bound rate, but
  is real gas-phase reaction
• Current research suggests part of overestimation
  tendency may be due in part to assuming all
  nitrate is fine
• More updates in future

32
Emerging PM Model Development Issues

• Interface with Meteorological Model (MM5/RAMS)
• Mass Conservations and Mass Consistency
• Clouds and Precipitation (resolved and
  unresolved)
• Instantaneous meteorological data (convective
  down bursts)
• MM5 PBL heights what to do when collapsed from
  clouds/snow

33
Conclusions on Model Development Synergisms

• CMAQ and CAMx offer two completely different
  platforms to test alternative PM modules and
  formulations
• provides an independent test of the assumptions
• identifies potential for introducing compensatory
  errors
• Numerous common challenges in PM modeling, the
  more ways of looking at the problem the better
• nitrate formation, size sections and deposition
• aqueous-phase chemistry
• PM size distribution
• meteorology
• computational efficiency

34
Toola to Facilitate Model Intercomparisons

• MM5 Interface Software
• MCIP 2.2
• MM5CAMx kvpatch
• CMAQ-to-CAMx conversion software
• Emissions
• IC/BC
• CAMx-to-CMAQ conversion software
• Emissions
• IC/BC

35
Current CMAQ/CAMx Comparisons

• 1996 Western USA
• WRAP and CRC
• Jan 2002, July 2001, July 1991Eastern USA
• VISTAS
• August September 1997 Southern CalEfornia
• CRC
• Midwest US/Supersites
• MRPO