1. A NEW OPERATIONAL AIR QUALITY PREDICTION SYSTEM OVER ITALY Maria Chiara Metallo 1 *, Attilio A. Poli 2, Fabrizio De Fausti 1, Luca Delle Monache 3, Pierluca Di Giovandomenico 1, Cristina Faricelli 1, Margherita Moreno 1, Alessandra Scifo 1 1 Environmental System Analysis srl, Bracciano, Rome, Italy 2 Take Air srl, Bracciano, Rome, Italy 3 National Center for Atmospheric Research, Research Applications Laboratory, Boulder, CO 8 th Annual CMAS Conference, Chapel Hill, NC, October 19-21, 2009 CONTACTS: Chiara Metallo (c.metallo@takeair.info), Luca Delle Monache (lucadm@ucar.edu)

2. Outline LaMiaAria modeling system Nested domains Model configurations The Dust model Emissions Operational timelines Preliminary Results Future work LaMiaAria modeling system Nested domains Model configurations The Dust model Emissions Operational timelines Preliminary Results Future work

3. Meteorology (MM5) Meteorology-Chemistry Interface Processor (MCIP) Emission Processing System (SMOKE) Chemistry Transport Model (CMAQ) Post-Processing LaMiaAria Modeling System Structure

5. Nested Domains Domain CoverageSpatial resolution Domain1Europe + North Africa54 km (77 X 111) Domain2intermediate18 km (84 X 78) Domain3Italy6 km ( 177 X 213)

6. CMAQ CONFIGURATION Current operational CMAQ forecast still uses static profile lateral boundary condition (LBC). The initial conditions (IC) for CMAQ are set from the previous forecast cycle. Current operational CMAQ forecast still uses static profile lateral boundary condition (LBC). The initial conditions (IC) for CMAQ are set from the previous forecast cycle. ADOPTED SCHEMES:  Yamartino global mass-conserving scheme to calculate horizontal and vertical advection  diffusion coefficient based on local wind deformation  calculate vertical diffusion using the Asymmetric Convective Model version 2  deactivate plume in grid model  2nd generation CMAQ aerosol deposition velocity routine  RADM-based cloud processor that uses the asymmetric convective model to compute convective mixing  Aerosol module : the 3rd generation modal CMAQ aerosol model (AERO 3) ADOPTED SCHEMES:  Yamartino global mass-conserving scheme to calculate horizontal and vertical advection  diffusion coefficient based on local wind deformation  calculate vertical diffusion using the Asymmetric Convective Model version 2  deactivate plume in grid model  2nd generation CMAQ aerosol deposition velocity routine  RADM-based cloud processor that uses the asymmetric convective model to compute convective mixing  Aerosol module : the 3rd generation modal CMAQ aerosol model (AERO 3)

7. ADOPTED SCHEMES:  NCEP /GFS data  No grid analysis nudging  No observation nudging  Reisner mixed phase  Kain-fritsch cumulus parameterization (54 and 18 km grid)  MRF pbl (Troen and Mahrt, 1986)  Atmospheric radiation scheme: CLOUD (Dudhia)  Shallow convective scheme  Multi-layer soil model ADOPTED SCHEMES:  NCEP /GFS data  No grid analysis nudging  No observation nudging  Reisner mixed phase  Kain-fritsch cumulus parameterization (54 and 18 km grid)  MRF pbl (Troen and Mahrt, 1986)  Atmospheric radiation scheme: CLOUD (Dudhia)  Shallow convective scheme  Multi-layer soil model MM5 CONFIGURATION Nested Domains D5480 x 114 D1894 x 91 D6193 x 216 Vertical Layers 29 sigma pressure Nested Domains D5480 x 114 D1894 x 91 D6193 x 216 Vertical Layers 29 sigma pressure

8. THE DUST MODEL S Erodibility factor (to reveal Hot Spots)[Ginoux, 2001] T Tunable Factor A m Bare soil fraction[Zender, 2003] α = f(soil texture)= 100exp[(13.4 M clay -6.0)ln10] Mobility Efficiency Q = Q(u *,u *t ) =const · u *t 3 [1- (u *t / u * ) 2 ] [1+ u *t / u * ] Horizontal Flux u * = (  / ρ) 1/2 Friction Velocity u *t = f(D, Re *t, ρ p ) · F c Threshold Friction Velocity [Iversen & White, 1982] S Erodibility factor (to reveal Hot Spots)[Ginoux, 2001] T Tunable Factor A m Bare soil fraction[Zender, 2003] α = f(soil texture)= 100exp[(13.4 M clay -6.0)ln10] Mobility Efficiency Q = Q(u *,u *t ) =const · u *t 3 [1- (u *t / u * ) 2 ] [1+ u *t / u * ] Horizontal Flux u * = (  / ρ) 1/2 Friction Velocity u *t = f(D, Re *t, ρ p ) · F c Threshold Friction Velocity [Iversen & White, 1982] F = α Q(u *, u *t ) · A m · T · S The algorithm used to assess surface dust flux is based on the Dust Entrainment and Deposition model (DEAD, Zender, 2002). The flux of dust, expressed in Kg/m 2 s, released in the atmosphere and than transported by CMAQ (in 2 bins fine/coarse fractioned following D’Almeida [1987] size distribution) is given by: The algorithm used to assess surface dust flux is based on the Dust Entrainment and Deposition model (DEAD, Zender, 2002). The flux of dust, expressed in Kg/m 2 s, released in the atmosphere and than transported by CMAQ (in 2 bins fine/coarse fractioned following D’Almeida [1987] size distribution) is given by:

9. Saharan dust Outbreak over Sicily 15-16/05/09

10. For the 54 and 18 km grid, the contributions of the anthropogenic sources (road transport, non road transport, industry, agricultural sources, etc.) are implemented using the last available version of:  European Monitoring and Evaluation Programme (EMEP) emission database;  Emission Database for Global Atmospheric Research (EDGAR), excluded particulate matter, for north African areas;  European Pollutant Emission Register (EPER) for industrial point sources. For the 54 and 18 km grid, the contributions of the anthropogenic sources (road transport, non road transport, industry, agricultural sources, etc.) are implemented using the last available version of:  European Monitoring and Evaluation Programme (EMEP) emission database;  Emission Database for Global Atmospheric Research (EDGAR), excluded particulate matter, for north African areas;  European Pollutant Emission Register (EPER) for industrial point sources. The spatial disaggregation is evaluated according to the methodology of the surrogate variables, using geographic data in a GIS platform (primary traffic, CORINE land cover by European Environment Agency) related to the emissions sources. 54 and 18 km GRID EMISSIONS

11. The inventory of emissions for the Italian national territory (6 km grid) is carried out using the National Emission Inventory provided by the Institute for Environmental Protection and Research (ISPRA), available according to the CORINAIR classification The municipal spatial disaggregation is carried out from the emissive data on a provincial base according to the methodology of the proxy (or surrogate) variables. The inventory of emissions for the Italian national territory (6 km grid) is carried out using the National Emission Inventory provided by the Institute for Environmental Protection and Research (ISPRA), available according to the CORINAIR classification The municipal spatial disaggregation is carried out from the emissive data on a provincial base according to the methodology of the proxy (or surrogate) variables. 6km GRID EMISSIONS

12. Operational Timelines

13. Preliminary results: hourly values 50 40 30 20 10 0 PM10 Modeled (blue) VS Observed (red) values NO 2 Modeled (blue) VS Observed (red) values O 3 Modeled (blue) VS Observed (red) values O 3 Modeled (blue) VS Observed (red) values 140 120 100 80 60 40 20 0 35 30 25 20 15 10 5 0 Ponzone (Al, 14 May 2009 ) for PM10, Cremona (15 June 2009) for NO2, Alessandria (20 June 2009) and Acqui Terme (8 May 2009) for O3. µg/m 3 140 120 100 80 60 40 20 0

14. Fractional Bias Preliminary Results (114 stations)(83 stations)(53 stations)

15. Results European directive for modeling uncertainty LM: European AQ Limit value (Target value for O3) EVA with values exceeding the regulatory target (50%) depicted in red

16. Scatter plots 24-h Mean for PM10 in 53 stations. 24-h Max for O 3 and NO 2 in 83 and 114 stations respectively. Analyzed Period: 1 July- 30 September Summer O 3 daily-max modeled values are 96.4% inside the range ±30% and 86.7% inside ±20% Results

17. AQ Forecast samples High summer ozone in Italy

18. Tomorrow NO 2 and O 3 forecast NO 2 O3O3

19. Tomorrow SO 2 and PM10 forecast SO2PM10

20. LaMiaAria web site: www.lamiaaria.it

21. Region maps

22. Future works Short-term (1 year): * WRF *AERO4 *CB-V * Emission improvements road transport and natural sources Medium-term (1-3 years): * Postprocessing, i.e., bias correction (e.g., KF-based algorithm) * Global CTM BC Long-term (3-5 years): * Probabilistic prediction system based on ensemble data assimilation Short-term (1 year): * WRF *AERO4 *CB-V * Emission improvements road transport and natural sources Medium-term (1-3 years): * Postprocessing, i.e., bias correction (e.g., KF-based algorithm) * Global CTM BC Long-term (3-5 years): * Probabilistic prediction system based on ensemble data assimilation

23. A NEW OPERATIONAL AIR QUALITY PREDICTION SYSTEM OVER ITALY Thank you for your attention Luca Delle Monache (lucadm@ucar.edu) National Center for Atmospheric Research, Research Applications Laboratory Boulder, CO 8 th Annual CMAS Conference, Chapel Hill, NC, October 19-21, 2009 CONTACTS: Chiara Metallo (c.metallo@takeair.info), Luca Delle Monache (lucadm@ucar.edu)