Morris CamX Evaluation

Evaluation of CAMx: Issues Related to Sectional Models: 

Evaluation of CAMx: Issues Related to Sectional Models Ralph Morris, Bonyoung Koo, Steve Lau and Greg Yarwood ENVIRON International Corporation Novato, CA (rmorris@environcorp.com) Chao-Jung Chien, Gail Tonnesen and Zion Wang UCR CE-CERT PM Model Performance Workshop Research Triangle Park, North Carolina February 10-11, 2004

Outline: 

Outline Development of CAMx4+ that combines CAMx4 with PMCAMx Mechanism 4 (M4) 2-Section treatment (fine/coarse) PMCAMx N-Section treatment Effects of sectional treatment on nitrate in SoCal Multi-Model Intercomparison using WUSA 1996 CMAQ, REMSAD, CAMx_M4 (2-Section) and CAMx N-Section Discussion of performance metrics Which ones most appropriate for PM modeling How to present model performance statistics Model performance for dummies

Science Options in CAMx4+: 

Science Options in CAMx4+ PM Size treatment M4 2-Section (fine/coarse) all secondary PM is fine N-Section (CMU treatment) Aerosol Dynamics ISORROPIA equilibrium (M4 must use) MADM dynamic HYBRID Aqueous-Phase Chemistry RADM bulk 1-section (M4 must use) VSRM (CMU multi-section module)

Effects of Particle Size Distribution: 

Effects of Particle Size Distribution Testing of assumptions of particle size distribution using new merged CAMx4/PMCAMX code (CAMx4+) M4 = CAMx4 2-Section plus RADM aqueous EQUI = N-Sections equilibrium + VRSM aqueous MADM = 10-Sections dynamic + VRSM aqueous RADM/EQ = 10-Sections equil. + RADM aqueous RADM/EQ4 = 4-Sections equil. + RADM aqueous October 17-18, 1995 Southern California Episode

Slide5: 

M4 EQUI 24-Hour Sulfate (g/m3) October 18, 1995 M4 peak SO4 39 g/m3 EQUI peak SO4 51 g/m3 ~ Long Beach Area Differences due to more sulfate production in CMU VRSM than RADM aqueous-phase chemistry Further downwind (Riverside) M4 produces more sulfate than EQUI

Slide6: 

24-Hour Nitrate (g/m3) October 18, 1995 M4 peak NO3 83 g/m3 EQUI peak NO3 54 g/m3 Observed NO3 peak at Riverside ~40 g/m3 Differences due to assuming all nitrate is fine vs. PM nitrate represented by 10 size sections (EQUI) M4 EQUI

Slide7: 

24-Hour Nitrate (g/m3) October 18, 1995 M4 peak NO3 83 g/m3 EQUI peak NO3 51 g/m3 EQUI 10-Section dry deposits NO3 faster due to coarse mode resulting in less NO3 in downwind Riverside area that agrees better with observations  Raises questions regarding CAMx_M4 & CMAQ assumption that all secondary PM is fine M4

1996 Regional PM Modeling of Western US: 

1996 Regional PM Modeling of Western US WRAP Section 309 SIP modeling used 1996 36 km WUSA Database 1996 MM5 Simulation (Olerud) 1996 Base Case Emissions (UNC/UCR) 1996 Base Case Modeling using CMAQ and REMSAD Old (~2001) version of CMAQ Many updates to emissions as part of Section 309 modeling Use 1996 database to evaluate updates Model updates CMAQ, REMSAD, CAMx Emission Updates

1996 Regional PM Modeling: 

1996 Regional PM Modeling Three models to Intercompare and Evaluate CMAQ Version 4.3 (August 2003) REMSAD Version 7.06 CAMx Version 4+ Develop Processors to Facilitate Intercomparison CMAQ-to-REMSAD Emissions, IC, and BC Processors CMAQ-to-CAMx Emissions, IC and BC Processors Use CMAQ plume rise estimated in 3-D emission files Substantial reduction in size of emission inputs 3-D CMAQ files to 2-D plus (i,j,k) data

Slide10: 

Science Algorithms Selected for 1996 Modeling

Notes on Science Summary of PM Models: 

Notes on Science Summary of PM Models CMAQ, CAMx_M4 & CAMx_4Sec all used RADM Aqueous-Phase Chemistry CAMx_4Sec (N-Section) can also use CMU VRSM, but more computationally demanding All models used equilibrium (ISORROPIA) approach Dynamic and Hybrid available in CAMx4+ but computationally demanding All models configured with CB4 Chemistry REMSAD uses Micro-CB4 Some changes in rates, especially Nitrate chemistry

Slide12: 

Modal vs. Sectional Size Approaches Three Modes: Ten Sections: Although can integrate modal distribution in CMAQ to get PM2.5, in practice usually assume first two modes make up PM 2.5

Old CMAQ/REMSAD SO4 PerformanceSec 309 Old CMAQ = V0301 Early 2001New CMAQ = V4.3 August 2003 + MCIP2.2: 

Old CMAQ/REMSAD SO4 Performance Sec 309 Old CMAQ = V0301 Early 2001 New CMAQ = V4.3 August 2003 + MCIP2.2 October 2003 WRAP MF Meeting Improvements in CMAQ Performance using new version

1996 Revised Evaluation – Western USA: 

1996 Revised Evaluation – Western USA January/July Comparisons w/ 4 Models CMAQ V4.3 REMSAD CAMx_M4 CAMx_4Sec (F) = All coarse mode PM in CM (C) = All Secondary PM is Fine IMPROVE Network (~50 Sites in WUSA & 1996) Only PM2.5 is speciated CAMx-4Sec (F) & (C) comparisons can address secondary PM coarse mode issues

Revised WRAP 1996 CMAQ Modeling: 

Revised WRAP 1996 CMAQ Modeling 95 x 85 36 km Grids EPA 1996 MM5 Simulation 18 Vertical Layers MCIP2.2, MM5REMSAD and MM5CAMx Processing of MM5 CB4 Chemistry

SO4 IMPROVE – January 1996  CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue: 

SO4 IMPROVE – January 1996 CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

SO4 IMPROVE – January 1996  CMAQ Red CMAQ Red CAMx_4Sec (F) CAMx_4Sec (C): 

SO4 IMPROVE – January 1996 CMAQ Red CMAQ Red CAMx_4Sec (F) CAMx_4Sec (C)

SO4 IMPROVE – July 1996 CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

SO4 IMPROVE – July 1996 CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

SO4 IMPROVE – July 1996  CMAQ Red CMAQ Red CAMx_4Sec (F) CAMx_4Sec (C): 

SO4 IMPROVE – July 1996 CMAQ Red CMAQ Red CAMx_4Sec (F) CAMx_4Sec (C)

Slide20: 

SO4 Time Series Grand Canyon NP CMAQ & REMSAD OBS = Red 1996 Annual CMAQ & CAMx_M4 Models Exhibit Similar Behavior, e.g., Miss Observed High SO4 in Mid-June

NO3 IMPROVE – January 1996  CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

NO3 IMPROVE – January 1996 CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

NO3 IMPROVE – January 1996  CMAQ Red CMAQ RedCAMx_4Sec (F) CAMx_4Sec (C): 

NO3 IMPROVE – January 1996 CMAQ Red CMAQ Red CAMx_4Sec (F) CAMx_4Sec (C)

NO3 IMPROVE – July 1996  CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

NO3 IMPROVE – July 1996 CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

NO3 IMPROVE – July 1996  CMAQ Red CMAQ RedCAMx_4Sec (F) CAMx_4Sec (C): 

NO3 IMPROVE – July 1996 CMAQ Red CMAQ Red CAMx_4Sec (F) CAMx_4Sec (C)

Slide25: 

NO3 Time Series Grand Canyon NP CMAQ & REMSAD OBS = Red 1996 Annual CMAQ & CAMx_M4 Models Exhibit Similar Behavior, e.g., Underestimate Summer Observed NO3

OC IMPROVE – 1996 Annual  CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

OC IMPROVE – 1996 Annual CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

EC IMPROVE – 1996 Annual  CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

EC IMPROVE – 1996 Annual CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

Soil IMPROVE – 1996 Annual  CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

Soil IMPROVE – 1996 Annual CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

CM IMPROVE – 1996 Annual  CMAQ Red CMAQ RedREMSAD Blue CAMx_M4 Blue: 

CM IMPROVE – 1996 Annual CMAQ Red CMAQ Red REMSAD Blue CAMx_M4 Blue

Presentation of PM Model Performance: 

Presentation of PM Model Performance Need to evaluate on a PM component basis (total PM mass or extinction doesn’t cut it) Many networks using different instrumentation and species definitions so should not mix networks Many statistical measures available that often give conflicting signals, which ones should we stress? Subregional model performance needed Interested in low as well as high values No real PM benchmarks available Results in tables of of numbers that are difficult to interpret and impossible to read

Presentation of PM Model Performance: 

Presentation of PM Model Performance Example Summary Model Performance plots for 1996 WUSA PM Model Intercomparison Plot Bias versus Gross Error (borrowed from TCEQ) Compare with Each Other and with Performance “Benchmarks” – Example “Benchmarks” used: 15% Bias and 35% Error (15%/35%) [borrowed from ozone modeling] 50% Bias and 75% Error (50%/75%) Not suggesting these be the benchmarks, used for example purposes only

Presentation of PM Model Performance: 

Presentation of PM Model Performance Mean Normalized Bias Error (MNBE) Mean Fractional Bias Error (MFBE) Normalized Mean Bias Error (NMBE) NMBE = Absolute Bias/Average Observed

SO4 IMPROVE January 1996: 

SO4 IMPROVE January 1996

SO4 IMPROVE July 1996: 

SO4 IMPROVE July 1996

NO3 IMPROVE January 1996: 

NO3 IMPROVE January 1996

NO3 IMPROVE July 1996: 

NO3 IMPROVE July 1996

Organic Carbon (OC) IMPROVE: 

Organic Carbon (OC) IMPROVE January 1996 July 1996

Slide38: 

Elemental Carbon (EC) IMPROVE January 1996 July 1996

Preliminary Conclusions 1996 WUSA Modeling: 

Preliminary Conclusions 1996 WUSA Modeling Although models exhibit variations in model performance, no one model is clearly performing better than the others across all species and periods Model performance in revised 1996 Base Case simulations much improved over previous runs Improved MM5 processing (e.g., MCIP2.2) Improved model formulations (e.g., CMAQ V4.3, CAMx3+, REMSAD V7) Model performance still less than stellar and varies by species and time period 1996 MM5 simulation has issues CAMx_4Sec run without NaCl estimate approximately 10% secondary PM is coarse (e.g., 12% SO4 across the WUSA IMPROVE network)