1. A Comparison Study of CMAQ Aerosol Prediction by Two Thermodynamic Modules: UHAERO V.S. ISORROPIA Case study for January 2002 episode Fang-Yi Cheng 1, Andrey V. Martynenko 2, Daewon Byun 1 and Jiwen He 2 1 Department of Geosciences 2 Department of Mathematics IMAQS/University of Houston October 03, 2007 6th annual CMAS conference, October 1~3, 2007, Chapel Hill

2. Background Atmospheric aerosols have direct impact on earth’s radiation balance, air pollution, fog formation, visibility and human health. Inorganic particles typically consist of ammonium, sulfate, nitrate, sodium, chloride, calcium, etc. Phase state of aerosols at given T and RH are determined by thermodynamic equilibrium. A variety of thermodynamic models have been developed to predict partition of inorganic aerosols between liquid, solid and gas phases. Predicting gas/aerosol partitioning of semi-volatile inorganic aerosol is challenging task because multiple solid, liquid and gas phases could exist. Computational codes can be very complex and for application in 3-D air quality models, an efficient numerical algorithm must be used.

3. Applications in 3-D air quality model 3-D air quality models usually use pre-calculated information of phase behavior to facilitate computation, ex. ISORROPIA UHAERO-PSC is essentially similar to AIM model (Wexler and Clegg, 2002) with same physical and chemical setup but differs in numerical algorithm Air quality model (CMAQ) is performed with UHAERO module and benchmarked with simulation using ISORROPIA (currently used in CMAQ). Goal is to provide thermodynamic module that is physically general and numerically efficient for 3-D air quality applications. A new inorganic thermodynamic module UHAERO is recently developed (Amundson et al., 2006).

4. Differences between UHAERO and ISORROPIA Difference in activity coefficient methods --- ISORROPIA uses Bromley’s model (Bromley, 1973) for multicompoment activity coefficient and K-M method (Kusik and Messner, 1978) for binary activity coefficient --- UHAERO uses PSC activity coefficient model ( Pitzer et al. 1986; Clegg et al. 1992; Wexler and Clegg, 2002) which is mole fraction based and considered as the state of science model Difference in predictions of aerosol water content --- ISORROPIA uses empirical ZSR relation (Stokes et al. 1966) to calculate water content --- UHAERO directly computes water content based on water activity Difference in numerical solution methods --- ISORROPIA incorporates pre-determined equation approach and pre-calculated tables (Nenes et al., 1999) --- UHAERO uses numerical technique (primal-dual active-set algorithm) without any prior assumption to determine equilibrium state (Amundson et al. 2006).

5. Configuration of the study episode CMAQv4.6, saprc99, AERO4 Resolution 36-km continental domain; (x,y,z) = (148, 112, 14) Two simulations ( CMAQ-ISORROPIA V.S CMAQ-UHAERO ) are conducted Metastable (only liquid in aerosol phase) assumption is used Episode: January 1 ~ 23, 2002 MM5, analysis nudging, KF2, RRTM, Reisner, PX PBL/LSM NEI 2001 inventory were used for point source, biogenics were processed using BEIS 3.13, onroad mobile emissions were computed using MOBILE6 Information is provided from Bhave Prakash (EPA)

6. CASTNET ( weekly sampling), IMPROVE (Two 24-hour samples are collected each week, on Wednesday and Saturday from midnight to midnight local time) From http://www.epa.gov/castnet/site.htmlcc Observational datasets Pittsburgh Supersite, Pennsylvania (sampling time 2 hour )

7. IMPROVE Comparison is very similar between two simulations Sulfate is fairly predicted ISORROPIA UHAERO Nitrate is over-predicted

8. ammonium CASTNet Jan. 8 ~15, 2002 nitrate Model (ISORROPIA) (ring) shows over-prediction, (observation is in the core). Sites with high bias are located in eastern region

9. When X approaches 1, aerosol is NH 4 + rich When Y approaches 1, aerosol is SO 4 2- rich and NO 3 - poor XY diagram for CASTNet datasets (NH 4 ) 2 SO 4 H 2 SO 4 HNO 3 NH 4 NO 3 X Y 0 1 1

10. XY diagram for CASTNet Generally, both models over-predict NH 4 + and NO 3 - ISORROPIA predicts slightly higher NH 4 + than UHAERO The tendency of the CASTNet data distributes toward off-diagonal line OBS Model

11. RH in range 50% to 100% Temp in range –8 to +8 degree C Sulfate is fairly predicted Total NH 4 + is over- predicted Pittsburgh Excess NH 4 + reacts with HNO 3 over-produce NO 3 - Same bias is also observed by other scientist and attributing error to meteorological and emission uncertainty (Shaocai et al., 2005).

12. RH TEMP Low RH (20~50 %) is over western U.S. and central continental area High RH (>85 %) in northern part of domain, south eastern Texas, Louisiana state and ocean Temperature is below freezing in northern part, and above 285 K in southern part of domain. RH and TEMP

13. Nitrate ISORROPIA UHAERODiff. RHTEMP UHAERO shows less NO 3 - in low RH region. HNO 3

14. NH 4 + NH 3 UHAERO shows less NH 4 +, more NH 3 than ISORROPIA in low RH region ISORROPIA UHAERODiff.

15. Low RH (50~ 60 %), 637 grid points XY diagram from model (corresponding to previous one snap shot) RH (75 ~ 95 %), 4268 grid points UHAERO ISORROPIA When RH increase, the points moves further in direction UHAERO shows less NH 4 + than ISORROPIA ISORROPIAUHAERO At low RH region, UHAERO moves toward direction comparing to ISORROPIA 1 1 0 X Y

16. ISORROPIAUHAERO 1 1 0 X Y UHAERO shows less NH 4 +, more NH 3 than ISORROPIA Low RH 50% Box model for NH 3 At high RH region, the difference is small between two modules High RH 90%

17. Conclusions and future work Sulfate is fairly predicted, nitrate and ammonium are over-predicted CMAQ-UHAERO takes ~30% more computer running time than CMAQ- ISORROPIA Differences of nitrate and ammonium partitioning are mostly in low RH region XY diagram comparison for CASTnet site shows over-prediction of NH 4 + from both models with slightly better agreement in UHAERO Box model predicts less NH 4 + in UHAERO than ISORROPIA at low RH region, that is consistent with our 3-D simulation result Future work will focus on (1) deliquescence branch in which the solid phases can form; (2) collecting high time resolution data to evaluate model

18.  Thanks for Bhave Prakash (EPA) on providing the required information for model simulation, as well as the guidance and suggestions for model comparison. Acknowledgement

19. Deep PBL height  low NH 4 + Shallow PBL  high NH 4 + Emission and Meteorological Uncertainty Too high NH 3 emission ??? Alice et al. (2006) mentioned NH 3 should be increased during summer and decreasing during winter