Implementation and Testing of EQULV II in the CMAQ Modeling System Yang Zhang North Carolina State University Mark Z. Jacobson Stanford University, Stanford, CA CMAQ, September 26-29, 25
Acknowledgments NOAA: Project Sponsor Ken Schere, Prakash Bhave, and Shaocai Yu, U.S. NOAA/EPA, for providing all input files for setting up the SOS episode and Fortran script for statistical calculations. Alice Gilliland and Steve Howard, U.S. NOAA/EPA, for providing Fortran code for extracting data from observational databases and CMAQ. Eric Edgerton and Rick Saylor, ARA, Inc., for providing the SEARCH database. Ping Liu, Kai Wang, and Jianlin Hu, NCSU, for help in post-processing model results.
Presentation Outline Background and Objectives Box Model Comparison -D Testing and Evaluation Episode, Model Configuration, Available Measurements Preliminary Results Summary and Future Work
Background and Objectives Background Thermodynamic equilibrium affects gas-particle partitioning and PM formation (e.g., nitrate, ammonium, chloride); Uncertainties in experimental data (e.g., activity coefficients, DRH); Numerical difficulties in simulating a mixed-phase system; Existing modules are computationally efficient but numerically less accurate and vice versa. Objectives Improve EQULV II in terms of numerical accuracy and computational speed; Incorporate EQULV II into CMAQ and CMAQ-MADRID as an alternative thermodynamic module to RRPIA; Apply and evaluate CMAQ-EQULV II s performance for the 12-28 June 1999 episode.
Main Features of RROPIA and EQULV II RROPIA EQULV II System solved -Na-NO -SO -Cl -Na-Ca-Mg-K-NO -SO -Cl-CO Equilibrium reaction/ chemical species Solution method 15 reactions 21 species ( gases, 9 liquids, 9 solids) Iterative bisection Iterative bisection-newton for H Chemical regime Subdomains (sulfate very rich/rich/poor) Full domain Activity coefficient Binary - Kusik and Meissner, 1978 Multi-component Bromley, 197 Precalculated lookup table Temperaturedependence Solid treatment Equilibrium constants, DRHs (1) Solid formed when RH < MDRH (metastable) (2) No solid formation for all RHs (liquid only) Size-resolved equilibrium No Yes 18 reactions 26 species ( gases, 1 liquids, 9 solids) Mass Flux Iteration (MFI) and Analytical Equilibrium Iteration (AEI) Binary - Hamer and Wu, 1972; Goldberg, 1981; Bassett and Seinfeld, 198; Filippov et al., 1985; and Pitzer, 1991 Multi-component - Bromley method, 197 Precalculated lookup tables Equilibrium constants, DRHs, activity coefficients, solute and water activity coefficients down to 19 K for several species (1) Solid formed when RH < CRH (meta-stable) (2) Solid formed when RH < DRH (stable) () No solid formation for all RHs (liquid only) -D implementation Single-cell only (1)Single-cell (2)Multiple-cell (vectorization) Reference Nenes et al., 1998, 1999 Jacobson et al., 1996; Jacobson, 1999 a & b
Box Model Test Conditions [H2SO ], μg m- 5 2 T/TSO 2-16.5 - TNO/TSO 1.2-12. - TNaCl/TSO, 2-8,.5-2 Temperature, K 278, 288, 298, 8 278, 288, 298, 8 RH, % 1-1 1-1 Solid treatment Iteration number used in EQULV II, N Solid formed when RH < CRH; No solid for all RHs 1, 1, 1, 1 Benchmark: 1 Total number of cases 112 112 Solid formed when RH < CRH; No solid for all RHs 1, 1, 1, 1 Benchmark: 1 Total number of simulations (112 112 N) 2 solid treatments = 11,2 (112 112 N) 2 solid treatments = 11,2
Box Model Results: HNO (g) RROPIA, V1.5 EQULV II (NROT = 1) TSO = 5 TSO = 2 2 2 1 2 2 RROPIA vs. Exact HNO (g) 1 2 2 vs. Exact HNO (g) RROPIA, V1.5 2 2 1 RROPIA vs. Exact HNO (g) 1 2 2 vs. Exact 1 1 N=1 N=1 1 2 2 1 2 2 EQULV II (NROT = 1) 2 2 HNO (g) 9 5 1 - -7-11 2 1 NMBs for HNO(g) at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 ug m - RROPIA, v1.5 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for HNO(g) at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 ug m - EQULV II (NROT = 1) vs. Exact HNO (g) 2 2 1 N=1E 1 2 2 EQULV II (NROT = 1) vs. Exact HNO (g) 2 2 1 N=1E 1 2 2-1 -2 - - 278 288 298 8 278-8 T, K RROPIA, v1.5
Box Model Results: NO - TSO = 5 TSO = 2 RROPIA, V1.5 2 2 1 RROPIA vs. Exact PNO 1 2 2 RROPIA, V1.5 2 2 1 RROPIA vs. Exact PNO 1 2 2 5 5 25 15 5-5 NMBs for NO at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - RROPIA, v1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 2 1 vs. Exact PNO 1 2 2 2 2 1 N=1 vs. Exact N=1E 1 2 2 PNO EQULV II (NROT = 1) EQULV II (NROT = 1) 2 2 2 2 1 vs. Exact N=1 1 PNO 1 2 2 vs. Exact N=1E PNO 1 2 2-15 25 2 15 1 5-5 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for NO at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 278 288 298 8 278-8 T, K RROPIA, v1.5
Box Model Results: (g) RROPIA, V1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 2 1 TSO = 5 TSO = 2 RROPIA vs. Exact (g) 1 2 vs. Exact (g) 1 2 2 1 vs. Exact (g) 1 2 RROPIA, V1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 RROPIA vs. Exact (g) 1 2 2 1 2 1 vs. Exact (g) N=1 N=1 N=1E 1 2 vs. Exact (g) N=1E 1 2 15 1 11 9 7 5 1-1 - 8 6 2-2 NMBs for (g) at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - RROPIA, v1.5 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for (g) at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 278 288 298 8 278-8 T, K RROPIA, v1.5
Box Model Results: RROPIA, V1.5 2 1 TSO = 5 TSO = 2 RROPIA vs. Exact P 1 2 RROPIA, V1.5 2 1 RROPIA vs. Exact P 1 2 5 1-1 - NMBs for at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - RROPIA, v1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 vs. Exact P 1 2 2 1 vs. Exact P 1 2 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 2 1 vs. Exact P N=1 N=1 N=1E 1 2 vs. Exact P N=1E 1 2-5 2-2 - 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for at T = 278-8 K, RH =1-5%, and [TSO]=5, 2 g m - 278 288 298 8 278-8 T, K RROPIA, v1.5
Box Model Results: HCl(g) RROPIA, V1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 2 1 TSO = 5 TSO = 2 RROPIA vs. Exact HCl(g) 1 2 vs. Exact HCl(g) 1 2 2 1 vs. Exact HCl(g) 1 2 RROPIA, V1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 2 1 2 1 RROPIA vs. Exact HCl(g) 1 2 vs. Exact HCl(g) 1 2 vs. Exact HCl(g) N=1 N=1 N=1E N=1E 1 2 21 16 11 6 1 - -9-1 12 8 - NMBs for HCl(g) at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % RROPIA, v1.5 NMBs for HCl(g) at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 278 288 298 8 278-8 T, K RROPIA, v1.5
Box Model Results: Cl - RROPIA, V1.5 2 1 TSO = 5 TSO = 2 RROPIA vs. Exact PCl 1 2 RROPIA, V1.5 2 1 RROPIA vs. Exact PCl 1 2 1 9 5 1 - NMBs for Cl at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - RROPIA, v1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 2 1 vs. Exact PCl 1 2 vs. Exact PCl 1 2 EQULV II (NROT = 1) EQULV II (NROT = 1) 2 1 2 1 vs. Exact PCl N=1 N=1 N=1E 1 2 vs. Exact PCl N=1E 1 2-7 6 - -6 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for Cl at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - RROPIA, v1.5 278 288 298 8 278-8 T, K
EQULV II Differences in Equilibrium Reactions 1. 2.. ( g) ( g) Cl( s) H HNO ( g) Cl NO Gives lower HNO (g) / ; higher NO - / (g)/cl -. NO ( s) NO 5. HNO ( g) HNO ( l) 6. H 2 SO ( l) RROPIA H HSO 7. 8. 9. ( g) H NO 2 Cl( s) O ( s) ( g) HCl( g) OH ( g) HNO ( g) Gives higher HNO (g)/ / HCl(g); lower NO - / (g)
Box Model Results: H RROPIA, V1.5.8.6..2 TSO = 5 TSO = 2 RROPIA vs. Exact H.2..6.8 RROPIA, V1.5.8.6..2 RROPIA vs. Exact H.2..6.8 NMBs for H at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 99 99 299 199 99 2.6E16 RROPIA, v1.5 EQULV II (NROT = 1) vs. Exact.8 N=1.6..2 H.2..6.8 EQULV II (NROT = 1).8.6..2 vs. Exact N=1 H.2..6.8-1 2 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for H at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 1 EQULV II (NROT = 1) vs. Exact.8 N=1E.6..2 H.2..6.8 EQULV II (NROT = 1).8.6..2 vs. Exact N=1E.2..6.8 H -1-2 2.2E1 278 288 298 8 278-8 T, K RROPIA, v1.5 7.E1
Predicted [H ] at T = 278 K, TSO = 2 μg m - ug m - (k) T/TSO=1., TNO/TSO=1, TNaCl/TSO=.5 2.E 1.5E 1.E E E EQULV II, NROT1 EQULV II, NROT1 RROPIA_v1.5 1 2 5 6 7 8 9 1 Relative Humidity, % ug m - (l)t/tso=1.5, TNO/TSO=1, TNaCl/TSO=.5 1.E1 8.E12 6.E12.E12 2.E12.E EQULV II, NROT1 EQULV II, NROT1 RROPIA_v1.5 1 2 5 6 7 8 9 1 Relative Humidity, % ug m - 1.E1 8.E12 6.E12.E12 2.E12.E (o) T/TSO=1.5, TNO/TSO=., TNaCl/TSO=.5 EQULV II, NROT1 EQULV II, NROT1 RROPIA_v1.5 2 6 8 1 Relative Humidity, % ug m - 1.E1 8.E12 6.E12.E12 2.E12.E (q)t/tso=1.5, TNO/TSO=, TNaCl/TSO=.5 EQULV II, NROT1 EQULV II, NROT1 RROPIA_v1.5 2 6 8 1 Relative Humidity, %
Box Model Results: H 2 O TSO = 5 TSO = 2 RROPIA, V1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 8.E5 6.E5.E5 2.E5 RROPIA vs. Exact.E.E 2.E5.E5 6.E5 8.E5 8.E5 6.E5.E5 2.E5 8.E5 6.E5.E5 2.E5 vs. Exact N=1 vs. Exact PH2O PH2O.E.E 2.E5.E5 6.E5 8.E5 N=1E.E.E 2.E5.E5 6.E5 8.E5 PH2O RROPIA, V1.5 EQULV II (NROT = 1) EQULV II (NROT = 1) 8.E5 6.E5.E5 2.E5 RROPIA vs. Exact PH2O.E.E 2.E5.E5 6.E5 8.E5 8.E5 6.E5.E5 2.E5 vs. Exact PH2O.E.E 2.E5.E5 6.E5 8.E5 8.E5 6.E5.E5 2.E5 N=1 vs. Exact N=1E PH2O.E.E 2.E5.E5 6.E5 8.E5 6 5 2 1-1 598 98 98 298 198 98-2 NMBs for H2O at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - RROPIA, v1.5 1_5 1_2 5 6 7 8_95 99 1 1_1 RH, % NMBs for H2O at T = 278-8 K, RH =1-1%, and [TSO]=5, 2 g m - 278 288 298 8 278-8 T, K RROPIA, v1.5
Box Model Results: Predictions of NO - Solid Liquid vs. Liquid only SolidLiquid TSO = 5 TSO = 2 Liquid SolidLiquid Liquid RROPIA vs. Exact RROPIA vs. Exact RROPIA vs. Exact RROPIA vs. Exact RROPIA, V1.5 2 2 1 PNO 1 2 2 RROPIA, V1.5 2 2 1 PNO 1 2 2 RROPIA, V1.5 2 2 1 PNO 1 2 2 RROPIA, V1.5 2 2 1 PNO 1 15. 2 2 EQULV II (NROT = 1) vs. Exact 2 2 1 PNO 1 2 2 EQULV II (NROT = 1) vs. Exact PNO 2 2 1 1 2 2 EQULV II (NROT = 1) vs. Exact 2 2 1 PNO 1 2 2 Top: RROPIA vs. Exact Bottom: EQULV II, NROT = 1 vs. Exact EQULV II (NROT = 1) vs. Exact PNO 2 2 1 1 15. 2 2
Box Model: Timing Test 28 Cases with T=298 K, RH=1-1%, TSO= 2 μg m - Iteration Numbers MAXIT=1 NSWEEPI= RROPIA EQULV II CPU Time, s NROT CPU Time, s (Single-cell) CPU Time, s (9-cell) 5.77 1-5 1 2.97 1-5 1. 1-5 5.6 1-5 1.5 1-5 5 6.9 1-5 2. 1-5 1 8.6 1-5. 1-5 2 1. 1 -.51 1-5 1.21 1 -.1 1-5 1.5 1 -.9 1-5 5 1.51 1-1 -5 1 2.2 1-6.9 1-5 1 6.7 1-2.5 1-1. 1-1.89 1 - CPU Time, sec 1.E-2 1.E- 1.E- 1.E-5 CPU time vs. Iteration number RROPIA EQULV II (1-cell) EQULV II (9-cell) 1 1 1 1 1 Total Number of Iterations used in EQULV II
-D Test with CMAQ for the June 1999 SOS Episode Simulation Period 12-28 June 1999 Grid Resolution 2 km, 178 12 grid cells, 19 layers Meteorology MM5/FDDA Emissions EPA s NEI 99 inventory Initial/Boundary cond. 2-day spinup, CMAQ default ICs/BCs CMAQ V., SAPRC99/EBI or ROS, modified to include ternary nucleation Datasets for Evaluation CASTNet: hourly O, weekly SO 2-, NO -, at 8 sites IMPROVE: -day SO 2-, NO -,, EC, OC; 2-hr PM 2.5 at 15 sites STN: -day SO 2-, NO -,, EC, OC, PM 2.5 at 19 sites AIRS: hourly O at 1161 sites SEARCH: Hourly O ; 2-hr PM 2.5 at JST, YRK, BHM, CTR, OLF, PNS, OAK, GFP
Observed vs. Predicted PM 2.5 at SEARCH Sites 6 Observation RROPIA JST -Jefferson Street, Atlanta, GA 6 Observation RROPIA YRK - Yorkville, GA PM 2.5 (ug m - ) 2 PM2.5 (ug m - ) 2 6/1/99 6/16/99 6/18/99 6/2/99 6/22/99 6/2/99 6/26/99 6/28/99 Local Time (EDT) 6/1/99 6/16/99 6/18/99 6/2/99 6/22/99 6/2/99 6/26/99 6/28/99 Local Time (EDT) 6 Observation RROPIA CTR - Centreville, AL 6 Observation RROPIA OAK - Oak Grove, MS PM 2.5 (ug m - ) 2 PM 2.5 (ug m - ) 2 6/1/99 6/16/99 6/18/99 6/2/99 6/22/99 6/2/99 6/26/99 6/28/99 Local Time (CDT) 6/1/99 6/16/99 6/18/99 6/2/99 6/22/99 6/2/99 6/26/99 6/28/99 Local Time (CDT)
Observed vs. Predicted PM 2.5 and Its Composition at JST and YRK, GA 2 hr-avg. PM2.5 at JST, June 17, 1999 2 hr-avg. PM2.5 at YRK, June 17, 1999 Mass conc8 ug m - 5 25 2 15 1 NO 1 observation RROPIA EQULV II, N=2 EC 1 Mass conc8 ug m - 5 25 2 15 1 NO 1 observation RROPIA EQULV II, N=2 EC 1 5 5 PM2.5 NO- SO= EC OM PM Species PM2.5 NO- SO= EC OM PM Species Mass conc8 ug m - 5 25 2 15 1 2 hr-avg. PM2.5 at JST,June 18, 1999 NO 1 observation RROPIA EQULV II, N=2 EC 1 Mass conc8 ug m - 5 25 2 15 1 2 hr-avg. PM2.5 at YRK, June 18, 1999 NO 1 observation RROPIA EQULV II, N=2 EC 1 5 5 PM2.5 NO- SO= EC OM PM Species PM2.5 NO- SO= EC OM PM Species
- Observed vs. Predicted HNO, NO - and PM 2.5 Composition at YRK, GA HNO (ppb) 15 12 9 6 YRK - Yorkville, GA observation RROPIA Mass conc ug m - 5 2 1 2 hr-avg. PM2.5 at YRK, June 21, 1999 NO 1 observation RROPIA EQULV II, N=2 EC 1 6/1/99 6/16/99 6/18/99 6/2/99 6/22/99 6/2/99 6/26/99 6/28/99 Local Time (EDT) PM2.5 NO- SO= EC OM PM Species 6 YRK - Yorkville, GA 2 hr-avg. PM2.5 at YRK, June 22, 1999 (ug m - ) NO 5 2 1 RROPIA EQUSOLV II, N=1 Mass conc ug m - 5 2 1 NO 1 observation RROPIA EQULV II, N=2 EC 1 6/1/99 6/16/99 6/18/99 6/2/99 6/22/99 6/2/99 6/26/99 6/28/99 Local Time (EDT) PM2.5 NO- SO= EC OM PM Species
RROPIA vs. EQULV: Performance Statistics Normalized Mean Error and Bias (Fraction) Normalized Mean Error Factor and Bias Factor (Fraction) Module CASTNet IMPROVE SEARCH NME % NMB % NME % NMB % NME% NMB % PM 2.5 EQUI 8 9-26 -29 2 2 - -5 Sulfate 2.5 EQUI 29 1 15 9 9 18 19 5 5 22 2 Nitrate 2.5 EQUI 89 75-16 -2 96 89-2 -9 76 67-71 -51 Ammonium 2.5 EQUI 28 1 5-7 6 2-22 7 6-2 -6 BC 2.5 EQUI 7-55 -5 7-5 5-51 OC 2.5 EQUI -25-25 1 1 - -
Daily Average Changes in HNO /NO - and / Due to Major Processes 6 2-2 - -6 Daily-average change in HNO 6/12 6/1 6/1 6/15 6/16 6/17 6/18 6/19 6/2 6/21 6/22 6/2 6/2 6/25 6/26 6/27 6/28 Time 1..5 -.5 Daily-average change in NO- 1 Daily-average change in 2.5 2. 1.5 1..5 -.5-1. -1.5-2. -2.5 Daily-average change in Change in conc. (ug m - ) 5-5 Change in mixing ratios (ppm) Change in HNO mixing ratios (ppm) Change in NO- conc. (ug m - ) -1. 6/12 6/1 6/1 6/15 6/16 6/17 6/18 6/19 6/2 6/21 6/22 6/2 6/2 6/25 6/26 6/27 6/28 Time -1 6/12 6/1 6/1 6/15 6/16 6/17 6/18 6/19 6/2 6/21 6/22 6/2 6/2 6/25 6/26 6/27 6/28 6/12 6/1 6/1 6/15 6/16 6/17 6/18 6/19 6/2 6/21 6/22 6/2 6/2 6/25 6/26 6/27 6/28 Time Time Horizontal Transport Vertical Transport Emissions Dry Deposition Gas Chemistry PM Processes Aqueous Processes Mass Balance Adjustment Horizontal Transport Vertical Transport Emissions Dry Deposition PM Processes Aqueous Processes Mass Balance Adjustment
Summary EQULV II (N 1) gives lower biases (< 1%) for HNO, NO -, HCl, Cl -, SO 2-, H 2 O, H, and (except for RH 99) and moderate biases (~27%) for. RROPIA gives lower biases for HNO, NO -, SO 2-, H 2 O (except RH = 1), H (except RH 5), and ; moderate biases (~-6%) for HCl, Cl - and abnormally high H 2 O and H biases (> E-2.6E1%) under some conditions. Larger biases occur for RH or 99 for most species. A thermodynamically metastable system (solid liquid) High sulfate conditions: RROPIA overestimates HNO (g) and underestimates NO - ; EQULV II (N=1) gives opposite trends. A supersaturated system (no solid) RROPIA gives some abnormally high or low HNO (g) and NO - ; EQULV II with N 1 gives slightly worse results than the metastable system (solidliquid). CMAQ with EQULV II and RROPIA performs similarly except for NO - and. EQULV II (solid and liquid) gives better NO - at SEARCH sites, RROPIA (no solid) gives slightly better NO - for IMPROVE and CASTnet sites, but it gives [HNO ] much higher than the observed one at some sites. Differences in NO - and predictions can be explained by different reactions and the high NO - predicted by RROPIA with liquid only. Thermodynamics only may not explain the observed gas/particle partitioning. Future work include multiple-cell implementation, size-resolved equilibrium, and application of CMAQ-EQULV II for regions with crustal species and sea-salt.