1. Some Thoughts on the Current State of Emissions-Based PM Air Quality Modelling
Michael Moran
Air Quality Research Branch
Meteorological Service of Canada
Toronto, Ontario, Canada
EMEP Workshop on PM Measurement & Modeling
New Orleans, Louisiana, U.S.A April 2004

2. Talk Outline Workshop objectives and emissions-based AQ modelling:
How well are we doing?
What new insights have we gained?
What do we need to do better?
What are some “low-hanging fruit”?

3. How well are we doing?
Encouragingly well.
Models are improving, and
current models have some
predictive skill for episodic and
seasonal simulations.
AURAMS examples 

4. AURAMS Overview regional emissions processing system;
A Unified Regional Air-quality Modelling System
Episodic, Eulerian, regional, size-resolved, chemically-characterized particulate-matter (PM) modelling system intended initially for research and policy support
“unified’ in that it considers multiple air pollutants and can be applied to multiple AQ issues (PM, O3, acid deposition) for integrated AQ management
consists of three main components:
regional emissions processing system;
prognostic regional meteorological model;
regional sectional PM air-quality model
current PM resolution: 12 size bins ( m) and 8 chemical components (SO4=, NO3-, NH4+, BC, OC, CM, SS, H2O)
Slide 10.

5. Time Series, Feb. 7–14, 1998
PM2.5
PM2.5
PM2.5 O3

6. 24-Hour PM2.5 Species Scatterplots: Feb. 7-14, 1998, IMPROVE & GAViM
SO4
NH4
NO3

7. What are some new insights that we have gained to date?
Atmospheric responses to
emission reductions may vary
by season
These responses may also be
significantly nonlinear

8. 2020P - 2020B Scenario “Deltas” for SO2 and NOx Emission Reductions, July 1995 & Feb. 1998 Cases
PM2.5 SO4
PM2.5 NO3
PM2.5 NH4
July 8-18
Feb. 7-15

9. Change in PM2.5 Nitrate Due to Reductions in SO2 and NOx Emissions
July 8-18, 1995
Feb. 7-15, 1998

10. What do we need to do better?
Everything. There is room for
improvements related to:
emissions
meteorology
PM process representations
ambient measurements
numerics

11. Iterative Improvement Process
emissions,
meteorology
air-quality
models
ambient AQ
measurements

12. Treatment of Emissions
Treatment of primary PM emissions
need PM speciation profiles
need size disaggregation profiles for PM2.5
need better transport factors (AQ model?)
need wind-blown dust & wildfires
Treatment of NH3 emissions
need monthly and diurnal variation
need subgrid-scale removal

13. Some Less Certain AQ Process Representations
cloud and precipitation properties
condensable p-OC and SOA formation
PM dry deposition
PM cloud processing (how good is SO4?)
subgrid-scale vertical cloud transport
PM wet removal

14. A Few Evaluation Issues
1. Consideration of comprehensive data sets
meteorological measurements
PM mass, composition, size distribution
gaseous co-pollutants
deposition measurements
optical measurements (visibility, optical depth)
2. Frequency (1/1 vs. 1/3 or 1/6 or 2/7)

15. What are some “low-hanging fruit”?

16. Inexpensive Non-Technical Actions to Improve Source- Based PM Modelling (1)
Include a reading list in report from this workshop (list some papers and reports that are important and recent)
Organize a focussed PM modelling workshop (current mtgs such as AMS-APM, AMS-AC, NATO-CCMS, AAAR, AGU, EGU are not focussed; NARSTO? EMEP? CMAS?)

17. Inexpensive Non-Technical Actions to Improve Source- Based PM Modelling (2)
Expedite access to field study data (e.g., encourage early modelling participation)
Describe model evaluation“good practice”: identify useful data sets, techniques, metrics
Encourage model confidence-building initiatives, such as joint studies, detailed evaluations, model intercomparisons, collaborations with receptor models and data analysts, …

18. Inexpensive Non-Technical Actions to Improve Source- Based PM Modelling (3)
Develop WMO GTS exchange code (BUFR?) for transmission of air concentration measurements
Rely on evidence-based research prioritization instead of intuition-based or interest-based (e.g., model performance for different PM components)
Build data “warehouses” for input data sets