A Comparative Dynamic Evaluation of the AURAMS and CMAQ Air Quality Modeling Systems

1
A Comparative Dynamic Evaluation of the AURAMS
and CMAQ Air Quality Modeling Systems

• Steven Smytha,b, Michael Moranc, Weimin Jianga,
  Fuquan Yanga, Wanmin Gongc, and Paul Makarc
• aICPET, National Research Council of Canada,
  Ottawa, ON
• bNow at Greenhouse Gas Division, Environment
  Canada, Gatineau, QC
• cAir Quality Research Division, Environment
  Canada, Toronto, ON
• 7th CMAS Conference, Chapel Hill, NC 8 October
  2008

2
Talk Outline

• Study approach
• AURAMS vs. CMAQ
• Harmonized base-case set up and results
• 4 emission scenarios for dynamic evaluation
• Results of comparative dynamic evaluation
• Conclusions

2
3
Acknowledgements

• Environment Canada (EC) and U.S. EPA provided
  emissions inventories
• Carolina Environment Program provided CMAQ, MCIP,
  and SMOKE
• Natural Resources Canadas PERD program and EC
  provided funding

3
4
Approach What is a harmonized comparative
dynamic evaluation?

• Comparative evaluation
• side-by-side comparison of predictions of two or
  more models
• Harmonized comparative evaluation
• models use same grid and same inputs
• Dynamic evaluation
• evaluate models predictions of changes in air
  concentrations or deposition due to changes in
  either emissions or meteorological inputs

4
5
Modelling Systems Compared

• CMAQ v4.6
• SAPRC-99 chemical mechanism AERO4 NRC PMx
  post-processor time-invariant chemical lateral
  boundary conditions Yamo advection scheme
• AURAMS v1.3.1b
• modified ADOM-II gas-phase and ADOM aqueous-phase
  chemical mechanism HETV thermodynamic
  equilibrium
• sectional representation of PM size distribution
  (12 bins from 0.01 to 41 µm in Stokes diameter)
• nine PM chemical components (SO4, NO3, NH4, EC,
  POM, SOM, CM, SS, H2O)
• zero-gradient chemical lateral boundary
  conditions
• semi-Lagrangian advection scheme

5
6
Harmonization Aspects

• Same map projection, domain, and horizontal grid
• secant polar stereographic projection true at
  60N, North American continental domain, 150x106
  grid pts, 42-km spacing
• Same emissions inventories and emissions
  processor
• 2000 Cdn EI, 2001 U.S. EI, 1999 Mexican EI SMOKE
  v2.2
• BEIS v3.09
• Same meteorological input fields
• from ECs GEM v3.2.0 operational weather forecast
  model
• Same simulation period July 1-29, 2002
• But AMPP vs. GEM-MCIP met preprocessors
  different vertical coordinates and discretization

6
7
Measurement Data (for 2002)

• O3 - hourly measurements from the EC NAPS
  network (190 sites) and U.S. EPA AQS network
  (1087 sites)
• PM2.5 - hourly measurements from NAPS (92 sites)
  and AQS (262 sites)
• Speciated PM2.5 1-in-3-day 24-h measurements
  from NAPS (17 sites) and U.S. EPA STN network
  (205 sites)

O3 Measurement Sites
PM Measurement Sites
7
8
Base-Case O3 Performance
Statistics O3 (ppb) O3 (ppb) daily peak O3 (ppb) daily peak O3 (ppb) daily low O3 (ppb) daily low O3 (ppb)
Statistics AURAMS CMAQ AURAMS CMAQ AURAMS CMAQ
meas. mean 35.6 35.6 60.8 60.8 11.7 11.7
mod. mean 43.0 52.2 68.4 71.6 16.5 32.6
MB 7.4 16.5 7.6 10.8 4.7 20.9
NMB () 21 46 13 18 40 178
ME 16.7 19.3 17.1 15.3 11.1 21.8
NME () 47 54 28 25 94 186
r2 0.395 0.433 0.342 0.488 0.104 0.086

• AURAMS has lower bias
• Similar levels of error and correlation
• CMAQ over prediction mainly due to inability in
  predicting daily lows

8
9
Base-Case Total PM2.5 Performance
statistics total PM2.5 (µg m-3) total PM2.5 (µg m-3) daily peak PM2.5 (µg m-3) daily peak PM2.5 (µg m-3)
statistics AURAMS CMAQ AURAMS CMAQ
meas. mean 14.4 14.4 28.3 28.3
mod. mean 12.9 5.0 22.7 9.1
MB -1.5 -9.4 -5.6 -19.2
NMB - 10 - 65 - 20 - 68
ME 9.8 10.2 16.4 19.6
NME 68 71 58 69
r2 0.073 0.152 0.038 0.081

• AURAMS has lower bias
• Similar levels of error and correlation

9
10
Four Emissions Scenarios

• Increase NOx emissions by 50 (1.5NOx)
• Decrease NOx emissions by 50 (0.5NOx)
• Decrease VOC emissions by 50 (0.5VOC)
• Decrease NOx and VOC emissions by 50
  (0.5NOx0.5VOC)

10
11
(Base Case - 0.5NOx) NMD Field for O3 (left)
and PM2.5 (right)
AURAMS
CMAQ
12
(Base Case - 0.5VOC) NMD Field for O3 (left)
and PM2.5 (right)
AURAMS
CMAQ
13
Comparison of AURAMS and CMAQ Mean O3 and PM2.5
Concentrations (units of ppbV or µg m-3) for
Base and Sensitivity Cases at Measurement Site
Locations Only. NMD Values are Percentages.
Scenario\ Species Base 1.5NOx 1.5NOx 1.5NOx 0.5NOx 0.5NOx 0.5NOx 0.5VOC 0.5VOC 0.5VOC 0.5NOx 0.5VOC 0.5NOx 0.5VOC 0.5NOx 0.5VOC
Scenario\ Species MC MC MD NMD MC MD NMD MC MD NMD MC MD NMD
AURAMS
hourly O3 43.0 45.3 2.3 5.3 36.4 -6.6 -15.3 36.4 -6.6 -15.3 33.1 -9.9 -23.0
max. O3 68.4 74.2 5.8 8.4 55.5 -12.9 -18.8 57.3 -11.1 -16.2 49.6 -18.8 -27.5
min. O3 16.5 15.8 -0.7 -4.2 15.8 -0.7 -4.2 14.2 -2.3 -13.9 14.8 -1.7 -10.3
total PM2.5 12.9 13.6 0.7 5.4 11.8 -1.1 -8.5 10.9 -2.0 -15.5 10.1 -2.8 -21.7
CMAQ
hourly O3 52.2 57.2 5.0 9.6 43.1 -9.1 -17.4 49.4 -7.8 -14.9 43.2 -9.0 -17.2
max. O3 71.6 81.4 9.8 13.7 55.7 -15.9 -22.2 66.7 -4.9 -6.8 55.0 -16.6 -23.2
min. O3 32.6 33.3 0.7 2.1 29.8 -2.8 -8.6 31.6 -1.0 -3.1 30.4 -2.2 -6.7
total PM2.5 5.0 5.2 0.2 4.0 4.8 -0.2 -4.0 5.1 0.1 2.0 4.9 -0.1 -2.0

• AURAMS is more VOC-sensitive for O3 and more
  NOx- and VOC-sensitive for PM2.5
• CMAQ is more NOx-sensitive for O3
• signs of predicted response are different for
  one species in two scenarios

14
Comparison of AURAMS and CMAQ Mean Concentrations
of Total PM2.5 and Various Major Species (all in
units of µg m-3) for Base and Sensitivity Cases
at Measurement Site Locations Only. NMD Values
Are Percentages.
PM Species base 1.5NOx 1.5NOx 1.5NOx 0.5NOx 0.5NOx 0.5NOx 0.5VOC 0.5VOC 0.5VOC 0.5NOx 0.5VOC 0.5NOx 0.5VOC 0.5NOx 0.5VOC
PM Species MC MC MD NMD MC MD NMD MC MD NMD MC MD NMD
AURAMS
total PM2.5 12.9 13.6 0.7 5 11.8 -1.1 -9 10.9 -2.0 -16 10.1 -2.8 -22
PM2.5 SO4 5.5 5.4 -0.1 -2 5.3 -0.2 -4 5.2 -0.3 -5 5.2 -0.3 -5
PM2.5 NO3 1.9 2.6 0.7 36 0.92 -0.98 -52 1.9 0.0 0 1.1 -0.8 -42
PM2.5 NH4 1.6 1.8 0.2 12 1.4 -0.2 -12 1.6 0.0 0 1.4 -0.2 -12
PM2.5 EC 0.28 0.28 0.0 0 0.28 0.0 0 0.28 0.0 0 0.28 0.0 0
PM2.5 TOA 4.8 4.8 0.0 0 4.5 -0.3 -6 2.4 -2.4 -50 2.4 -2.4 -50
CMAQ
total PM2.5 5.0 5.2 0.20 4 4.8 -0.20 -4 5.1 0.10 2 4.9 -0.10 -2
PM2.5 SO4 2.4 2.5 0.10 4 2.2 -0.20 -8 2.7 0.30 12 2.5 0.10 4
PM2.5 NO3 0.23 0.35 0.12 52 0.09 -0.14 -61 0.28 0.05 22 0.13 -0.10 -43
PM2.5 NH4 0.80 0.85 0.05 6 0.71 -0.09 -11 0.90 0.10 11 0.81 0.01 1
PM2.5 EC 0.32 0.31 -0.01 -3 0.33 0.01 3 0.33 0.01 3 0.34 0.02 6
PM2.5 TOA 1.0 0.98 -0.02 -2 0.98 -0.02 -2 0.71 -0.29 -29 0.72 -0.28 -28
15
Gas- and Particle-Phase Coupling

• Stockwell et al. (1988) found that in low-NOx
  areas,
• NOx emission reductions decrease oxidant levels
  and hence gas-phase SO2 oxidation and p-SO4
  formation
• VOC emission reductions increase oxidant levels
  and hence gas-phase SO2 oxidation and p-SO4
  formation
• Pun et al. (2008) have suggested that the above
  changes in oxidant levels will affect p-TOA (via
  SOA) formation in the same direction as p-SO4

15
16
Summary and Conclusions (1)

• O3, total PM2.5, and PM2.5 major species
  concentration changes resulting from four sets of
  NOx and VOC emissions scenarios were analyzed for
  July 2002 paired simulations with harmonized
  set-ups of the AURAMS and CMAQ AQ modeling
  systems
• Such a harmonized comparative dynamic evaluation
  provides a measure of the uncertainty in the
  predictions of two important pollutants for
  policy applications of these AQ modeling systems
• AURAMS was found to be more VOC-sensitive for O3
  whereas CMAQ was more NOx-sensitive
• AURAMS was found to be more NOx-sensitive and
  VOC-sensitive for PM2.5
• Differences were also evident in the spatial
  distributions of the predicted changes in O3 and
  PM2.5

16
17
Summary and Conclusions (2)

• NOx emission changes affect p-NO3 but also p-SO4
  and p-TOA VOC emission changes affect p-TOA but
  also p-SO4 and p-NO3
• While the magnitudes of the predicted changes O3
  and total PM2.5 varied considerably, the signs of
  the predicted changes were consistent except for
  daily minimum O3 for the 1.5NOx scenario and
  for total PM2.5 for the 0.5VOC scenario
• For the PM2.5 major species, however, differences
  in the signs of the predicted changes were more
  common, and these contributed to the total PM2.5
  change sign difference for the 0.5VOC scenario

17
18
Thank you!