1. Christian Seigneur AER San Ramon, CA
Three-Dimensional Modeling of Particulate Matter Current Performance & Future Prospects
Christian Seigneur
AER
San Ramon, CA

2. Schematic representation of a PM Eulerian model
Initial and Boundary Conditions
Meteorological
Model
Emissions
Air Quality PM Model
Transport
PM Chemistry and Physics
Droplet
Chemistry
Wet
Deposition
Dry
Deposition
Gas-phase
chemistry
Concentrations of gases and PM

3. Current model performance for three typical regional PM studies
Southern Appalachian Mountains Initiative (SAMI): URM for 9 episodes over the southeastern U.S. (Georgia Tech & TVA)
EPA/NOAA: CMAQ for 2001 annual simulation (Eder et al., this workshop)
Big Bend Regional Aerosol & Visibility Observational Study (BRAVO): MADRID 1 & REMSAD for 4 months over Texas and Mexico (AER, EPRI & CIRA)
Southern Oxidants Study of 1999 (SOS 99): CMAQ, MADRID 1, MADRID 2 & CAMx for 1 episode over the southeastern U.S. (AER)

4. Some model performance statisticsa for PM2.5 components

5. Diagnostic performance evaluation
There are many possible causes for model error
model inputs (boundary conditions, meteorology, emissions)
model formulation (transport, transformation, deposition)
Diagnostic analyses provide insights into those causes
sensitivity analyses
specific performance evaluations
spatial and temporal displays
model intercomparisons

6. Importance of boundary conditions Sulfate over the United States (Source: REMSAD, Mike Barna, CIRA)

7. Model performance for transport BRAVO tracer released from a point source 750 km northeast from the receptor
Regional models cannot reliably predict the impact of individual sources at long distances

8. Spatial display of model error can provide insights into possible causes
Sulfate error for
CMAQ-MADRID 1
in BRAVO:
Emissions?
Coastal meteorology?

9. Diagnostic analysis using fine temporal resolution
Observed and simulated (CMAQ-MADRID 2) organic mass
in SOS 99, Cornelia Fort, July 1999

10. Examples of diagnostic analyses using seasonal or monthly statistics
CMAQ 2001 annual simulation (Eder et al., this workshop)
PM2.5 performance is lowest in winter, possibly because PM2.5 is dominated by nitrate and carbonaceous components in winter
BRAVO 4-month simulation (Pun et al., 2004)
Sulfate is underestimated in July when Mexican contribution is highest and is overestimated in October when U.S. contribution is highest

11. Comparison of two PM models Importance of vertical mixing
CMAQ
CAMx
PM2.5 concentrations over the U.S. on 6 July 1999 differ
primarily because of different algorithms for vertical mixing

12. Comparison of three SOA modules (Pun et al., ES&T, 37, 3647, 2003)
The three SOA modules differ in:
the total amounts of SVOC and SOA
the gas/particle partitioning
the relative amounts of anthropogenic and biogenic SOA

13. Evaluating model response
A satisfactory operational evaluation does not imply that a model will predict the correct response to changes in precursors emissions
There is a need to conduct a diagnostic/mechanistic evaluation to ensure that the model predicts the correct chemical regimes
Indicator species can be used to evaluate the model’s ability to predict chemical regimes

14. Response of PM to changes in precursors (adapted from NARSTO, 2003)

15. Major chemical regimes
Sulfate
SO2 vs. oxidant-limited
Ammonium nitrate
NH3 vs. HNO3-limited
Organics
Primary vs. secondary
Biogenic vs. anthropogenic
Oxidants (O3 & H2O2)
NOx vs. VOC-limited

16. Example of indicator species Sensitivity of O3 formation to VOC & NOx
H2O2 / (HNO3 + Nitrate) as an indicator
High values:
NOx sensitive
Low values:
VOC sensitive O3 NO
NO2
HNO3
H2O2
HO2 OH
VOC

17. Example of indicator species Sensitivity of nitrate formation to NH3 & HNO3
Excess NH3 as an indicator
High values:
HNO3 sensitive
Low values:
NH3 sensitive
Ammonium nitrate
Ammonium sulfate
HNO3
NH3

18. Qualitative estimates of uncertainties (adapted from NARSTO, 2003)

19. Possible topics for improving PM model performance
Emission inventories (ammonia, primary PM, etc.)
Transport processes (e.g., vertical mixing, plume-in-grid)
Assimilation of cloud and precipitation data
SOA formation
Deposition velocities
Heterogeneous chemistry
Boundary conditions from global models

20. Acknowledgments
Funding for the BRAVO simulations was provided by EPRI and EPA
Funding for the SOS 99 simulations was provided by Southern Company, EPRI, MOG and CRC
Betty Pun, AER, Prakash Karamchandani, AER, Mike Barna, CIRA, Robert Griffin, University of New Hampshire, Brian Eder, EPA and Robin Dennis, EPA provided valuable inputs