Development and Evaluation of a New Canadian Operational Air Quality

1. Development and Evaluation of a New Canadian Operational Air Quality Forecast Model: GEM-MACH15 Michael Moran1, Louis-Philippe Crevier2, Sylvain Ménard2, Donald Talbot2, Paul Makar1, Sylvie Gravel1, Wanmin Gong1, Alexander Kallaur3, Hugo Landry2, Ping Huang4, and Sunling Gong1 1Air Quality Research Division, Environment Canada, Toronto, ON 2Air Quality Modelling Applications Section, Environment Canada, Montreal, QC 3Air Quality Research Division, Environment Canada, Montreal, QC 4Independent consultant, Toronto, Ontario 7th CMAS Conference, Chapel Hill, NC 7 October 2008

2. Talk Outline • What is GEM? • Environment Canada’s operational global/medium-range and regional/short-range weather forecast model • What is GEM-MACH? • GEM with in-line chemistry from AURAMS • What is GEM-MACH15? • proposed operational limited-area configuration of GEM­MACH with 15-km grid spacing • Preliminary GEM-MACH15 evaluation results • Next steps

3. “GEM” stands for modèleGlobal Environnemental Multi-échelle et / and Global Environmental Multiscale model

4. "Multiscale" Examples:Three GEM Grid Configurations regional global limited area

5. GEM Characteristics • rotated latitude-longitude horizontal grid • three different grid configurations • global uniform model (used for 10-day forecasts) • global variable model (used for 2-day regional forecasts) • limited area model (LAM: selected for GEM-MACH15) • hybrid vertical coordinate • semi-Lagrangian advection • hydrostatic and non-hydrostatic options • four-dimensional data assimilation • extensive physics library

6. “GEM-MACH” stands for Global Environnemental Multi-échelle -Modélisation de la qualité de l'Air et de la CHimie et / and Global Environmental Multiscale – Modelling Air quality and CHemistry

7. GEM-MACH Family Tree (Colours Denote Chemistry Mechanisms, Heavy Box Outlines Denote In-Line Models, Years Refer to Foundation Publication) ADOM (1988) CHRONOS (1997) MC2-AQ (2001) AURAMS (2006) GEM-AQ (2006) GRAHM (2004) GEM-Strato- Chem GEM-MACH GEM-BACH

8. Pros and Cons of In-line Treatment of Chemistry in GEM • allows chemistry calculationson GEM grid and time step, provides access to all GEM fields, and avoids need for interpolation to another model grid • provides better framework for implementation of chemical data assimilation • provides potential (in future) for chemical feedback to meteorology (e.g., via radiation, clouds) • provides framework for both global and regional chemistry models via GEM’s various grid configurations • increases complexity of in-line model but arguably simplifies modelling system (e.g., no met preprocessor) • provides easy parallelization route as GEM is already parallelized (using both MPI and OpenMP)

9. AQ Processes Now Implemented in GEM-MACH (1) • acquisition of all required meteorological fields • initialization of chemical species (including cycling of chemical tracers from run to run) • chemical lateral boundary conditions (for LAM) • chemical upper boundary conditions • input and injection of anthropogenic surface emissions and elevated emissions (incl. plume rise) • calculation and injection of biogenic emissions • vertical diffusion of chemical species

10. AQ Processes Now Implemented in GEM-MACH (2) • gas-phase chemistry (ADOM-II mechanism) • secondary organic aerosol formation (IAY scheme) • aqueous-phase chemistry (ADOM scheme) • heterogeneous inorganic chemistry (HETV scheme) • variable number of PM size distribution “bins” • dry deposition of gaseous and particle species • wet deposition of gaseous and particle species • other PM size-dependent processes: sea-salt emissions, nucleation, condensation, coagulation, activation, sedimentation

11. GEM-MACH Data Flow (2005) CAN-USA-MEX Emission inventories SMOKE off-line system GenBiogenX Biogenic info (Beld3, emis factor) Area Major points Climatology Geophys Meteorological Analysis RPN standard file format RPN standard file format GEM-MACH Emissions pre-processor GEM Meteorological pre-processor on-line system Emissions binary files for each CPU tile Meteorological binary files for each CPU tile GEM-MACH Main model PM2.5, O3, NO2, ... Temperature, pressure …

12. GEM-MACH vs GEM-MACH15 • GEM-MACH is the chemical model itself and can be run in different configurations by selecting: • grid configuration • physics parameterizations • chemistry parameterizations • GEM-MACH15 is just one of these possible configurations. It has been developed with EC’s operational AQ forecast needs in mind: • LAM configuration for continental North America • 15-km grid coincident with part of GEM15 uniform core • 2-bin sectional representation of PM size distribution (i.e., 0-2.5 and 2.5-10 μm) with 9 chemical components

13. GEM-MACH15 vs CHRONOS Modelling Domains CHRONOS GEM-MACH15 • Resolution: • 348 x 465 gridpoints, rotated LL • 15 km horizontally • 58 vertical levels up to ~ 30 km • Δt = 450 s • Resolution: • 350 x 250 gridpoints, secant PS • 21 km horizontally • 24 vertical levels up to ~ 6 km • Δt = 3600 s

14. Main Differences in Chemical Processes Between CHRONOS and GEM-MACH15

15. GEM-MACH15 Development Status • GEM-MACH15 is now being evaluated in experimental mode for ozone forecasts only against measurements and CHRONOS (July-Oct. 2008). Implementation of PM forecasting capabilities is being finalized with evaluation to follow. • The ozone field is essentially de-coupled from PM chemistry. Only small indirect feedbacks exist between the two. • Tests with AURAMS have shown that impact on ozone field of adding PM chemistry is of order of 0.1% bias, r2 of 0.999 • The GEM-MACH15 modelling setup should not change when PM processes are turned on: that is, • same configuration, same input fields • extra chemical processes activated by namelist keys

16. Model Evaluation: Methodology • Initial evaluation of GEM-MACH15 ozone forecasting capability was completed before experimental run began • Two forecast series were run for summer and winter conditions (24 cases each, initialized every 36 h) • Summer: June 8th, 2007 to July 13th, 2007 • Winter: February 2nd, 2008 to March 8th, 2008 • Two-week spin-up period precedes each period • Results were compared to observations over the entire North American continent for day 1 and day 2 forecasts • Will be repeated this fall for PM2.5 forecasts

17. Model Evaluation: Ozone (O3) • Results are compared to observations over the whole North American continent for day 1 and day 2 forecasts • 1,124 stations in summer • 568 stations in winter (some stations shut down in winter) • Observation network coverage (summer)

18. Ozone Forecast Comparison For Summer 2007 CHRONOS GEM-MACH15 Average ozone pattern of all 12 18-h forecasts from 00 UTC model runs from 2007-06-08 to 2007-07-13

19. Ozone Verification For Summer 2007 GEM-MACH15 24 48h forecasts (00 and 12 UTC) (2007-06-08 to 2007-07-13) CHRONOS

20. Ozone Verification: Diurnal cycle (summer 2007 00 UTC runs)

21. Ozone Forecast Comparison For Winter 2008 CHRONOS GEM-MACH15 Average ozone pattern of all 12 18-h forecasts from 00 UTC model runs from 2008-02-02 to 2008-03-08

22. Ozone Verification For Winter 2008 GEM-MACH15 24 48h forecasts (00 and 12 UTC) (2008-02-02 to 2008-03-08) CHRONOS

23. Ozone Verification: Diurnal cycle (winter 2008 00 UTC runs)

24. Example of Day 2 Hourly O3 Comparison

25. Example of Day 2 Hourly NO2 Comparison

26. Summer 2008 O3 and NO2 Monitor Locations

27. Ozone Verification For August 2008 GEM-MACH15 62 48-h forecasts (00 and 12 UTC) Units of ppbV CHRONOS

28. NO2 Verification For August 2008 GEM-MACH15 62 48-h forecasts (00 and 12 UTC) Units of ppbV CHRONOS

29. Regional Verification For August 2008 Western Canada Eastern Canada GEM-MACH15 62 48-h forecasts (00 & 12 UTC) Units of ppbV CHRONOS

30. Next Steps • Complete summer 2008 experimental run • Prepare for parallel run of full model this fall • Continue optimization of model set up • Link GEM-MACH15 with UMOS-AQ • Submit request for operational implementation →Objective is to be ready to implement operationally before summer 2009

31. Thank you for your attention!