1. SAPRC07T Implementation within the CMAQ model.
Bill Hutzell and Deborah Luecken
(AMDB/AMAD)
2009 CMAS Conference
October 20, 2009
9/10/2018

2. Why bother with SAPRC Mechanisms?
It provides an alternative approach for approximating photochemical reactions.
Lumped species based on emissions and reactivity
More easily expanded or de-lump, using their documentation
Historically, SAPRC mechanisms have been more comprehensively evaluated.
Evaluation used chamber studies
Peer reviewed
Many regulatory, i.e., CARB, and research groups prefer SAPRC mechanism.

3. What does the upgrade provide?
Updated reaction kinetics and branches
VOC species and reactions represents more compounds
New operators addressing completion within VOC and peroxy reactions with NO radicals
Alters production of radicals, peroxides and other oxygenated compounds
Improved chemistry of aromatics and organic acids
Optional chemistry of chlorine compounds
Possibly better mechanism for SOA precursors

4. What version of SAPRC07? SAPRC07B
BASE or first version, released in 2007
CS07A and CS07B
Simplified versions of the BASE
Mimics CB05 in # of species and reactions
SAPRC07A
Retro-version of the BASE
We are using SAPRC07T
Actually using the “B” variation

5. Why use SAPRC07T?
Explicitly represents several hazardous air pollutants.
Such as acetaldehyde, 1,3-butadiene, acrolein, and xylene isomers
Includes organic acids needed to better represent chemistry within cloud droplets.
Specifically peroxyacetic and higher organic acids
Adds compounds to chlorine chemistry
enables chlorine production from sea salt aerosols
Specifically CLNOy compounds used in the Knipping and Dabdub (2003) mechanism

6. What is the price?
Using current data structures in CMAQ photochemical mechanisms.
Mechanism would contain over 600 reactions
Algebraic manipulation can reduce the number for a compressed version
Hence, the “B” variation
Approximately 330 reactions
Code development need to implement “B” version of SAPRC07T
To describe and calculate new rate expressions used

7. Simulations January and July 2002 MM5 meteorology
Two nested domains with 14 layers
36X36 km2 over the continental United States
12X12 km2 over the eastern United States
2002 NEI version 3.0
Simulations conducted CMAQ version 4.7
Using either SAPRC99TX3 and SAPRC07T
mechanism

8. How do SAPRC07T predictions compare with its predecessor?
SAPR07T averages tend to have higher values at lower concentration. Overall, SAPRC07T predictions have a smaller range of concentrations at a site.
Averages determined from hourly observations at monitoring sites located in 12X12 km2 domain

9. O3, 8-Hour Averages: SAPRC07TB-SAPRC99TX3
However, SAPRC07T has slight more ozone at some rural location while predicting less ozone in urban areas.

10. Comparing regulatory standards between mechanisms
Regulatory measures show similar patterns. Largest differences occur in urban (high NOx areas) with SAPRC99TX3 higher.

11. Bulk Comparison to Observations
The Taylor diagram indicates that the SAPRC99TX3 and SAPRC07T have similar precisions over time and space at matching observations. The left-hand plot shows that the mechanism have similar error ozone and benzene predictions but formaldehyde and acetaldehyde error might increase for SAPRC07T.

12. Closer Look at Sites observations: Ozone Distributions
Both mechanisms produce almost identical distributions of hourly ozone predictions and fail to match the observed frequency of ozone concentrations at the lower range.

13. Hourly Ozone : Comparison to observations SAPRC99TX3 SAPRC07T
For individual sites, the mechanism produce almost identical plots for accuracy and precision measures. Errors increase for the lower range of concentration and predictions at given sites tend to under predict standard deviations in their observed ozones.

14. Ozone versus Location: Normalized Mean Bias , July 2002: SAPRC99TX3
Ozone versus Location: Normalized Mean Bias , July 2002: SAPRC99TX3 SAPRC07T
The NMB versus location does not show much difference between mechanism. Errors increase in rural areas and southern portion of the domain. Does this suggest failures in reactions involving biogenic compounds or in low NOx chemistry.

15. Ozone, July 2002: Normalized Predicted s SAPRC99TX3 SAPRC07T
However, the mechanism show some difference in the predicted range of concentrations.

16. Zoom: On region with differences SAPRC99TX3 SAPRC07T
Note that the area around Lake Michigan show difference in predicted standard deviation. Does this difference show an interplay between being up and down wind of the high NOx regions such as Chicago.

17. Why do both mechanisms fail to simulate the observed variability?
Boundary Conditions
For 36X36 km2 grid, BC’s for Ozone set to 35 ppb
Observed minimum 20 ppb lower
Photolysis rates
Temperature and Pressure effect generally neglected in their s and f values
 2/3 of rates observed to show these dependencies
Reaction schemes within mechanisms
HOx cycle: isoprene scheme needs updates
NOx cycle: several reactions uncertain and controversial such as NO2* and NO2-OH reactions
Note that the items under the last bulletin are ongoing development activities in AMAD such as Ying’s and Sarwar’s research.

18. Why use SAPRC07T instead of SAPRC99TX3?
Attempts to correct SAPRC99TX3’s shortcomings
Products from aromatic oxidation
VOC oxidation and peroxy reactions under low NOX
Chlorine chemistry for urban environments
More alkenes are explictly represented in radical cycles
HAPs such as Xylene isomers and Toulene
Biogenic compounds such as Alpha-Pinene
Better potential pathways for Secondary Aerosol Production (SOA)
Via higher organic peroxides
Prior HAPs participate in ozone chemistry
Truer to the Multi-Pollutant Concept than current version
1,3-Butadiene and Acrolein

19. Implications for HAPs: Higher Acrolein Predictions over Eastern States.

20. SAPR07T has photochemical sources besides 1,3-Butadiene
Specifically, Explicit and Lumped Alkenes Species

21. Further Work For July 2002 simulations
Analysis for individual as to setting
Examine NOy cycling
Analyze predictions from January 2002 simulations
Sensitivity Tests
Detailed investigation using IRR
Temperature and Pressure effects on photolysis rates
Emission reduction scenarios

22. Conclusions Implementation of SAPRC07T within CMAQ model possible
Step require revised how model describes and solved a photochemical mechanism
SAPRC07T gives similar ozone prediction to SAPRC99TX3
Some variability is lost in ozone predictions
Both old and new version compared similar with observations
SAPRC07T has slightly higher errors for aldehydes
Still recommend moving to SAPRC07T
corresponds current science in other photochemical mechanisms such as RACM2
Better suited future air quality applications such as SOA and Multi-Pollutant assessments

23. Slides with supplementary information

24. Comparing regulatory standards between mechanisms
Largest differences occur in urban (high NOx areas) with SAPRC99TX3 higher.

25. What development is needed for CMAQ implementation?
Need new expression for reaction rate constants
Not true constants
Sums of products, i.e., true rate constant times species concentrations
CMAQ mechanism files require a modified pre-processor programs
Additional data block contains new expressions
Solvers require modified routines that use the new rate expressions