GAMES, a comprehensive gas aerosol modelling evaluation system

Environmental Modelling & Software 21 (2006) 587–594 www.elsevier.com/locate/envsoft

GAMES, a comprehensive gas aerosol modelling evaluation system Marialuisa Volta*, Giovanna Finzi Department of Electronics for Automation, University of Brescia, Via Branze 38, 25123 Brescia, Italy Received 28 July 2003; received in revised form 18 June 2004; accepted 21 June 2004 Available online 12 March 2005

Abstract In this work the modelling system GAMES is described. It consists of the transport and chemical models CALGRID and TCAM, the meteorological model CALMET, the emission evaluation model POEMPM, the boundary and initial condition module (BICM) and the system evaluation tool (SET). GAMES has two configurations to estimate ozone exposure and to analyze photochemical and aerosols pollution episodes. The system has been designed, implemented and tested over a Northern Italy domain by the Environmental System Modelling and Control research group of Brescia University in the frame of EUROTRAC2SATURN project. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: Photochemical pollution; Aerosol; Multiphase modelling system; Ozone exposure; Episode simulation; Decision Support System

1. Introduction High ozone and particulate concentrations in summer season are related problems, both characterized by non-linear behaviours: (1) O3 and PM are non-linear chemical couplet, (2) the cause–effect relation between precursors on one side and O3 and PM levels on the other are non-linear (Sillman, 1999; Meng et al., 1997). Such a complex phenomenology requires adequate models for analysing and designing effective emission reduction strategies. Several air quality models have been described and applied at the urban and regional scales for seasonal and episode simulations (Russel and Robin, 2000; Hurley et al., 2005; San Jose´ et al., 2005). All of them require (1) input data (meteorological and emission fields, boundary and initial conditions), often computed and estimated in turn by models, and (2) post* Corresponding author. Tel.: C39 030 3715460; fax: C39 030 380014. E-mail address: [email protected] (M. Volta). 1364-8152/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.envsoft.2004.06.012

processing tools for the model assessment. Harmonizing the pre- and post-processing modules with the transport and chemical models means to design and to implement a modelling system. In this paper the transport, photochemical and aerosol modelling system GAMES (gas aerosol modelling evaluation system) is described. It has been designed following the two goals of the EPA’s Models-3 (Dennis et al., 1996), i.e. to provide  an effective Decision Support System for scientific and decision making purposes;  an integrated, flexible, modular framework to support the evolvement of models and modelling system. GAMES consists of the meteorological model CALMET (Scire et al., 1990), the emission model POEMPM (Finzi et al., 2002) providing present and alternative emission fields, the transport and chemical models CALGRID (Yamartino et al., 1992) and

588

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

transport chemical aerosol model (TCAM, Decanini, 2003), the boundary and initial condition module (BICM), the system evaluation tool (SET). POEMPM, TCAM, SET have been designed and implemented by the Environmental System Modelling and Control research group of Brescia University. GAMES has been tested and validated over a Northern Italy domain for long term and episode simulations in the frame of international and national projects aiming the design of a Decision Support System for Air Quality Management.

2. The modelling system GAMES has been designed following the classic scheme of the photochemical multiphase modelling systems. It includes and harmonizes meteorological, emission, transport, chemical models and post-processing tools (Fig. 1).

2.1. The meteorological pre-processor The 3D meteorological fields are provided by means of a three-step procedure. (1) The available local measurements (SYNOP reports and upper air sounding data, wind and temperature profiles) and the ECMWF fields are collected and analysed. (2) The background wind field is reconstructed adjusting ECMWF model output to local topography. (3) Finally the meteorological CALMET model (Scire et al., 1990) provides 3D wind fields merging background field with measurements and introducing local features revealed by ground-level measurements. Moreover CALMET estimates temperature fields as well as turbulence parameters.

METEOROLOGICAL DATA

TOPOGRAPHY AND LAND-USE ACTIVITY DATA

METEOROLOGICAL PROCESSOR EMISSION PROCESSORS TURBULENCE AND DEPOSITION VELOCITIES

CHEMICAL BOUNDARY AND INITIAL CONC.

CHEMISTRY TRANSPORT MODELS

2.2. The emission model The emission processor POEMPM has been specifically designed to produce present and alternative emission field estimates by means of an integrated top–down (disaggregating a large scale yearly inventory) and bottom–up (inventorying the polluting activities and applying emission factors) approach. POEMPM can be applied to the CORINAIR database and considers diffuse and main point sources coming from different activity sectors. Thanks to its technology and fuel-oriented formulation, this emission processor can be used to provide scenarios consistent with new fuel trades and pollutant abatement technologies. Model outputs are the results of four algorithms: the spatial disaggregation, the time modulation, the VOC and PM splitting (Fig. 2).  As for the spatial distribution, the model estimates municipality emissions starting from the province level, then distributes emissions on a grid domain. The spatial allocation procedure makes use of surrogate variables, highly correlated with emissions and defined by means of national and local statistical sources, GIS and land-use information.  As for the time modulation, the model can provide emission fields for any assigned time interval on a daily or hourly basis starting from the annual database. In accordance with EUROTRAC2/GENEMIS Project (Friedrich, 1997), fuel use, temperature, degree-days, working time, production cycle, traffic counts and road statistics are the main indicators being used for the temporal modulation of emission activities.  The total VOC amount is split into SAROAD classes of individual compounds and then lumped in agreement to the photochemical mechanism implemented in the transport model. To allow for such flexibility in chemical schemes, the FCM interface (Kumar et al., 1995) has been implemented in the model. spatial disaggregation

time modulation

VOC split

PM split

NUTS III level

year

total amount

total amount

municipality level

time interval

SADOAD classification

gridded domain

hour

CHEMICAL MODULES

SYSTEM EVALUATION MODULE

Fig. 1. Multiphase modelling system scheme.

VOC lumped classes: SAPRC90-97 COCOH97 CB IV

Fig. 2. POEMPM splitting algorithms.

• 10 size bins • 6 chemical classes

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

 The POEMPM also estimates size resolved and chemically split particulate matter emission fields. The considered chemical species are the organic carbon (OC), the elementary carbon (EC), SO2ÿ 4 , NOÿ 3 , H2O and others (including the heavy metals and the undefined compounds), while the size bins range from 0 to 11.4 mm focusing the classification on the smaller particles taking to a deeper impact on human health. At present POEMPM includes the 1998 gas emission inventory (CH4, NH3, NOX, N2O, CO, CO2, VOC and SOX) and the PM one for road traffic. In particular, the PM emissions have been estimated for diesel vehicles by COPERT III model (Ntziachristos and Samaras, 2000) and for gasoline vehicles by processing experimental emission factors. The size and chemical split average profiles have been estimated for each CORINAIR road traffic emission activity extrapolating the experimental profiles performed by several European Research Laboratories (Carnevale and Volta, 2003). 2.3. Boundary and initial condition module Initial and boundary conditions are assigned by BICM for the species solved by the photochemical model. The module algorithm follows two methodologies according to the available data (model simulations for a larger domain or measurements). 2.3.1. Nesting procedure The module estimates initial and boundary conditions interpolating available concentration fields simulated for larger domain. At present the BICM can extract ground level and vertical profiles from the runs of (1) EMEP Lagrangian photoxidant model (Jonson et al., 1999) and (2) unified EMEP model (Simpson et al., 2003). The module (1) fits EMEP co-ordinates, temporal and spatial resolutions according to the current GAMES domain and (2) maps EMEP chemical species in CALGRID and TCAM. 2.3.2. Measurement extrapolations The module classifies three different cell types (urban, rural and mountain) on the basis of the domain land use. For each cell type, the typical hourly concentration profiles at ground level (one for each gas species) are estimated from experimental measurements. The estimated vertical profile is an exponential decreasing function that gradually reduces the concentrations down to a 0.1 factor of the ground values. As for the ozone, the vertical distribution is estimated establishing the top value and defining a sigmoid profile obtained on the basis of experimental measurements. As for aerosols, the same method used for gas species is applied. Size and

589

chemical profiles can be assigned from measurements and literature. 2.4. The photochemical models Games implements two chemical and transport models, CALGRID and TCAM. 2.4.1. CALGRID The well-known photochemical model CALGRID is an Eulerian three-dimensional model. It implements an accurate advection–diffusion scheme in terrain-following co-ordinates with vertical variable spacing. The CALGRID chemical module implements the SAPRC90 (Carter, 1990) and the CB4 (Gery et al., 1989) mechanisms. The QSSA (Quasy Steady State Approximations) algorithm solves the kinetic equations (Hessvedt et al., 1978). 2.4.2. TCAM The transport chemical aerosol model is an Eulerian photochemical three-dimensional multiphase model. It implements the advection–diffusion scheme derived by CALGRID code. The dry deposition is treated using a resistance-based algorithm which takes into account pollutant properties, local meteorology and terrain features (Smolarkiewicz, 1983). The model implements the flexible chemical mechanism (FCM) interface for mechanisms based on both lumped molecule (SAPRC90, SAPRC97, COCOH97) and lumped structure (CB-IV) approaches. The FCM plays two roles. It works as interface between the photochemical model and POEMPM: FCM estimates the chemical parameters for the photochemical model and returns the lumping parameters to the emission module. Moreover the FCM provides to the photochemical models the code routines implementing the chemical mechanism. The TCAM model also includes and harmonizes an aerosol module derived from the MAPS box model (Durlak and Baumgardner, 2000). It describes the aerosols by means of a fixed-moving approach. The particle is represented as an internal core containing all the non-volatile material. The dimension of the core is constant. On the core, a surface layer contains the volatile material. The layer dimension is time variant. Both fractions of the particle are supposed to be mixed. TCAM allows the size resolved representation of particulate matter based on discrete sections. The aerosol module is coupled to COCOH97 chemical mechanism (Pandis et al., 1992), an extended version of SAPRC97, and it describes the dynamics of 21 chemical compounds split in 10 size bins, so that the prognostic variables of the module are 210. The ÿ C inorganic species are 12 (H2O, SO2ÿ 4 , NH4 , Cl , ÿ C C NO3 , Na , H , SO2(aq), H2O2(aq), O3(aq), elemental

590

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

carbon and other), while the organic species are 9, namely a generic primary and 8 classes of secondary organic species, each of them corresponding to one of the Condensable Organic Compounds included in the gas phase chemical mechanism. TCAM models the most relevant aerosol processes: the condensation, the evaporation (Seinfeld and Pandis, 1998; Kim et al., 1993), the nucleation (Jaecker-Voirol and Mirabel, 1989) involving all the considered chemical compounds and the aqueous oxidation of SO2 (Seinfeld and Pandis, 1998). The TCAM model solves the mass balance equations by means of a splitting operator (Marchuk, 1975). It integrates, for each simulation hour, the horizontal transport, the vertical transport and the chemistry for an half time step and then performs the same calculations in the reverse order for the remaining half. The gas phase chemistry is solved implementing the IEH algorithm (Sun et al., 1994), which treats separately the slowly reacting species and the fast reacting species. The former ones are solved by means of an explicit second order method, the latter by means of the implicit scheme LSODE (Hindmarsh, 1975). 2.5. System evaluation tool GAMES includes a module performing the assessment of the system simulations by statistical and graphical methods. The indexes and the graphics can be obtained for each measurement station or for a limited number of representative stations of the domain. The procedure to select the latter stations is based on the Cluster Analysis. This approach, detecting the similarity of the patterns recorded in different stations in terms of concentration levels and daily shape, allows for each homogeneous pattern group (1) to select the most representative station or (2) to generate a virtual station averaging, hour by hour, the measurements recorded in the stations belonging to the group. The measured and simulated patterns of the selected (actual or virtual) stations constitute the database for the system evaluation. SET, processing measurements and simulated concentrations, provides:  the computation of two US-EPA statistical indexes (EPA, 1991): the Mean Normalized Bias Error and the Mean Normalized Gross Error;  the computation of the ‘‘Directive 2002/3/EC’’ indicators: the maximum percentage deviation between recorded and simulated ozone concentrations for the 1-h concentration average and the 8-h daily maximum concentration;  the correlation coefficient between simulated and measured concentrations and the Root Mean Square Error;  AOT40, AOT60 ozone exposure indexes;  the mean concentrations for the simulated period;

 the 25th, 50th, 75th and 95th percentile values for AOTx, 1-h and 8-h daily maximum concentrations;  the assessment of the over threshold concentrations correctly simulated. Moreover the evaluation module allows (1) to analyze the sensitivity of the system to chemical modules, numerical solver, meteorological parameters and (2) to evaluate the impact of emission scenarios.

3. Modelling system configurations and time computing performances The system is available in two configurations. The first one consists of the meteorological model CALMET, the emission model POEMPM for gas species and the CALGRID photochemical model. In this configuration the system can provide long-term simulations for the homogeneous gas phase. The second configuration implements the transport chemical aerosol model and allows to perform accurate episode simulations for multiphase (gas and aerosols) processes. The two configurations require different efforts (1) in collecting emission, I.C. and B.C. input data sets and (2) in term of computing resources. In order to evaluate the time computing (including meteorological, emission, chemical and transport simulations) required by different system configurations (for gas and multiphase) an exercise has been set up. The system has been implemented and tested over a Northern Italy domain (240 ! 232 km2) (Fig. 3) including the whole Lombardia Region and some portions of the neighboring provinces. The area has been subdivided according to a grid system having 60 per 58 horizontal cells, with 4 km step size. Table 1 compares the time computing for three test 24-h simulations (equipment: Pentium 4, 2.40 GHz, RAM 512 MB). The first two runs consider the homogeneous gas phase (implementing CALGRID and TCAM, respectively), the latter one describes the multiphase, performing TCAM. The data underline that (1) the high time computing required by the system for the multiphase allows just episode simulations, (2) CALGRID, having shorter run time than TCAM due to the simpler solver (QSSA vs. IEH), can perform yearly-seasonal simulations.

4. Modelling system applications The central Po Valley (Northern Italy) is characterized by complex terrain, high urban and industrial emissions, a dense road network and it is often affected by severe photochemical and PM pollution episodes. This domain, including the Milan metropolitan area, has been deeply investigated by means of experimental

591

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

3500.00 3000.00

SO

2500.00 VA

CO

LC

2000.00 BG

1500.00

BS

MI

VR

1000.00

LO PV

CR

MN

500.00 0.00

Fig. 3. The simulation domain: location and orography (m a.s.l.).

campaigns (Vecchi and Valli, 1999; Neftel et al., 2002; Dommen et al., 2002) and modelling assessments (Silibello et al., 1998; Finzi et al., 2000; Martilli et al., 2002; Louka et al., 2003). GAMES has been tested and validated on this area for long term and episode simulations. In order to (1) show the system capability to correctly simulate chemical and transport phenomena and (2) to validate the input database (emissions, B.C. and I.C.), the following sections briefly describe the system applications to the 1996 summer season and to the 1–5 June 1998 photochemical smog episode. 4.1. Homogeneous gas phase long term simulation The simulations have been performed by POEMPM– CALMET–CALGRID system configuration over six months, from April 1 to September 30, 1996 (Pirovano et al., 2004; Gabusi et al., 2003). The initial and boundary conditions have been obtained by means of a nesting procedure from the EMEP Lagrangian photoxidant model. To evaluate the system performances in simulating temporal and spatial features of ozone concentrations, statistical indexes have been calculated in correspondence of a selection of monitoring stations located in the area with different orographic and emissive features. The reference monitoring stations have been classified into two groups (High Emission Density Area and Low Emission Density Area) taking into account (1) the ozone concentration pattern and trend and (2) the precursor emissions estimated in the Table 1 Time computing for different configurations (minutes for 24 simulated hours) Configuration

Time computing

CALGRID TCAM (COCOH97): gas phase TCAM (COCOH97): gas and aerosol phase

32 72 158

neighbors of the measurement sites. The MNBE and MNGE statistical indexes, evaluated for 1-h and 8-h daily maximum, agree with EPA recommendations and the correlation values are satisfactory (Table 2). The long-term impact of ozone concentrations has been assessed, according to the WHO and European guidelines, estimating the AOT40 index for crops and forest protection and AOT60 index for health protection. The simulations confirm that Lombardia region is exposed to high ozone concentrations for all summer season both in urban and rural areas. AOT40 and AOT60 values are very high over the whole domain, largely exceeding UE recommendations. Highest values are estimated outside urban areas in correspondence of the foothills, as the breezes favor the accumulation of secondary pollutants produced in the plain urbanized areas (see, for example, Fig. 4 showing AOT40 forest). 4.2. Multiphase episode simulation The second configuration modelling system, implementing TCAM, has been used to simulate the gas and aerosol production and transport (Decanini et al., 2003; Decanini and Volta, 2003a,b) during the 1–5 June 1998 episode, monitored by the PIPAPO campaign (Neftel et al., 2002) in the frame of EUROTRAC2 project. During the selected period, the strong photo-oxidant activity has been stressed by the stagnating meteorological conditions associated with high solar radiation. First gas phase and then the multiphase have been simulated by GAMES. The performances have been Table 2 Performance indicators for high and low emission monitoring stations 1-h daily max

8-h daily max

MNBE MNGE Correlation MNBE MNGE Correlation (%) (%) (%) (%) HEDA ÿ3.1 LEDA 8.7

24.0 20.9

0.70 0.74

1.6 11.7

23.4 22.2

0.75 0.74

592

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

5150

[ppb*h] 90000 80000 VARENNA

UTM [km]

5100

5050

SO

70000

ISPRA

60000

VIMERCAT LEGN_SMA LIMBIATE

50000

LC

CO

VA

MI_JUVA

40000

BG

30000

BS

MI

5000

VR

20000 LO

10000 PARMA

PV

0

4950 500

550

600

MN

CR

650

200.00 190.00 180.00 170.00 160.00 150.00 140.00 130.00 120.00 110.00 100.00 90.00 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00

UTM [km]

Fig. 4. AOT40 forest estimated by GAMES for summer 1996. Fig. 6. Ground level O3 (ppb) at 4 p.m. of 4th June 1998 (SAPRC97 simulation).

assessed comparing computed and measured data. In the Fig. 5 the ozone simulation results are compared with measured concentrations in four relevant urban and rural monitoring stations. The ozone patterns computed implementing SAPRC97 and CB4 mechanisms agree with the experimental data even if some discrepancies can be noted: (1) the vertical turbulence parameterization during night-time seems to be not correctly computed; (2) the model underestimates the peak concentrations in sub-urban sites and (3) is not able to reproduce local effects. As it has also been monitored by the experimental campaign, the simulated ozone fields show two ‘‘hot spot’’ areas located at north

and south of Milan metropolitan area (Fig. 6), as confirmed by the long term simulation test. GAMES well describes the phenomenon explained by meteorological conditions which occur during photochemical episodes (the wind directions are biased by mountain– valley breeze regimes; so the Milano plume is driven both north-eastwards forced by Alps, in particular, by Valtellina valley, both southwards forced by Appennini). As far as PM10 is concerned, the simulation results show that the time modulation of the estimated concentrations meets the behaviour of the experimental

O3: Erba, June 1-5 1998

160 140 120 100 80 60 40 20 0

conc. (ppb)

conc. (ppb)

O3: Milano Juvara, June 1-5 1998

1

0

24

48

72

96

160 140 120 100 80 60 40 20 0

120

3

0

24

hour

conc. (ppb)

conc. (ppb)

2

48

72

hour

72

96

120

O3: Verzago, June 1-5 1998

O3: Bresso, June 1-5 1998

24

48

hour

160 140 120 100 80 60 40 20 0 0

Measured SAPRC97 CBIV

1

96

120

160 140 120 100 80 60 40 20 0

1

2

Measured SAPRC97 CBIV

0

24

48

72

96

120

hour

Fig. 5. Computed vs. measured O3 concentrations (ppb) in Milano-Juvara (urban site), Erba (rural site), Bresso and Verzago (sub-urban sites).

593

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

Erba (CO) PM 10, June 1-5 1998

µg/m3

Milano, Marche. PM 10, June 1-5 1998 140

140

120

120

100

100

80

80

60

60

40

40

20

20

0

Measured Computed

0 0

24

48

72

96

120

0

24

hour

48

72

96

120

hour

Fig. 7. Computed vs. measured PM concentrations (mg/m3) in Milano-Marche (urban site) and Erba (rural site).

5. Conclusions GAMES modelling system has been designed to support the Air Quality Authorities in selecting effective

SO

VA

LC

CO

BG BS

MI

VR LO

PV

CR

MN

220.00 210.00 200.00 190.00 180.00 170.00 160.00 150.00 140.00 130.00 120.00 110.00 100.00 90.00 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00

Fig. 8. Ground level PM10 (mg/m3) at 4 p.m. of 4th June 1998.

daily mean concentrations estimated in Bresso 60 50

[µg/m3]

data (Fig. 7). Nevertheless the system estimates high concentrations during the night, stressing the hypothesis that the vertical turbulence parameterization during night-time is not correctly estimated, as noticed for ozone concentration. The simulation results evidence that the PM levels are spread all over the central and the southern regions of the domain (see, for example, Fig. 8 showing simulated PM10 at 4 p.m. of June 4th). The highest concentrations are estimated in areas surrounding the main urban emission. The PM10 spatial distribution suggests that secondary aerosol components play a relevant role in particulate production and accumulation processes in the domain. The same indication can be read in Fig. 9. The daily mean concentrations estimated for Bresso show that nitrates increase day by day due to the gas phase production of nitric acid condensing on the particles and taking part in the neutralization equilibria with ammonia.

NH4 + NO3SO4 = EC Other OC

40 30 20 10 0 1 st day 2 nd day 3 rd day 4 th day 5 th day

Fig. 9. PM10 daily mean concentrations (mg/m3) estimated in Bresso.

strategies to control and reduce the photochemical smog in polluted areas. The system integrates the meteorological processor CALMET, the emission model POEMPM and the initial and boundary condition module providing the inputs to the photochemical models (CALGRID and TCAM). The system evaluation tool processes the simulated concentration fields, assessing the agreement with the measured patterns. GAMES is highly flexible and can perform long-term and shortterm mesoscale simulations, allowing to evaluate ozone exposure and multiphase photochemical processes. The system has been performed and evaluated for long term and episode simulations revealing photochemical characteristics and regimes of central Po Valley (Northern Italy).

Acknowledgements The research has been partially supported by the Italian University and Research Ministry (MIUR), Regione Lombardia and Provincia di Brescia in the European frame of EUROTRAC2-SATURN project. The authors acknowledge Edoardo Decanini, Veronica Gabusi, Claudio Carnevale (University of Brescia), Guido Pirovano, Maurizio Riva (CESI, Milano), Giuseppe Brusasca, Giuseppe Calori, Camillo Silibello (AriaNet, Milano) for their kind and valuable cooperation.

594

M. Volta, Giovanna Finzi / Environmental Modelling & Software 21 (2006) 587–594

References Carnevale, C., Volta, M., 2003. Chemical and size characterization of PM road traffic emission: profile estimation for multiphase models. Proceedings of the European Aerosol Conference. Journal of Aerosol Science 2, S1349–S1350. Carter, W., 1990. A detailed mechanism for the gas-phase atmospheric reactions of organic compounds. Atmospheric Environment 24, 481–518. Decanini, E., 2003. Development and application of photochemical models for the evaluation of urban air quality. PhD thesis, Scuola Normale Superiore di Pisa. Decanini, E., Pola, M., Polla Mattiot, F., Volta, M., 2003. Application of REMSAD and GAMES modelling systems on a PM and ozone episode in Milan metropolitan area. International Journal of Environment and Pollution 20, 230–241. Decanini, E., Volta, M., 2003a. Application to Northern Italy of a new modelling system for air quality planning: a comparison between different chemical mechanisms. International Journal of Environment and Pollution 20, 85–95. Decanini, E., Volta, M., 2003b. The evaluation of secondary aerosol in Milan metropolitan area by means of GAMES modelling system. Proceedings of the European Aerosol Conference 2003. Journal of Aerosol Science 2, S1013–S1014. Dennis, R., Byun, D., Novak, J., Galluppi, K., Coats, C.J., 1996. The next generation of integrated air quality modeling: EPA’s MODELS-3. Atmospheric Environment 30 (12), 1925–1938. Dommen, J., Pre´vot, A., Neininger, B., Baumle, M., 2002. Characterization of the photoxidant formation in the metropolitan area of Milan from aircraft measurements. Journal of Geophysical Research 107 (D22). Durlak, S.K., Baumgardner, D., 2000. Examination of the evolution of urban plumes using MAPS (Model for Aerosol Process Studies). In: Proceedings of the 11th Joint Conference on the Applications of Air Pollution Meteorology with the Air and Waste Management Association. EPA, 1991. Guideline for regulatory application of the Urban Airshed Model. Tech. Rep. 450/4-91-013, EPA, Research Triangle Park, NC 27711. Finzi, G., Silibello, C., Volta, M., 2000. Evaluation of urban pollution abatement strategies by a photochemical dispersion model. International Journal of Environment and Pollution 14, 1–6. Finzi, G., Volta, M., Carnevale, C., Ragni, C., 2002. POEMPM: an emission model for size and chemical description of particulate matter. In: Midgley, P., Reuther, M. (Eds.), Proceedings of the EUROTRAC-II Symposium. Margraf Verlag, Weikersheim. Friedrich, R., 1997. Generation of European emission data for episodes. In: Tropospheric Modelling and Emission Estimation. Springer, pp. 180–214. Gabusi, V., Finzi, G., Pertot, C., 2003. Performance assessment of a long-term photochemical modelling system. International Journal of Environment and Pollution 20, 64–74. Gery, M., Whitten, G., Killus, J., 1989. A photochemical mechanism for urban and regional-scale computer modeling. Journal of Geophysical Research 94, 12925–12956. Hessvedt, E., Hov, O., Isaksen, I., 1978. Quasi steady state approximation in air pollution modeling. International Journal of Chemical Kinetics 10, 971–994. Hindmarsh, A., 1975. LSODE and LSODEI, two new initial value ordinary differential equation solvers. ACM-SIGNUM Newsletter 15 (4), 10–11. Hurley, P., Physick, W., Luhar, A., 2005. TAPM: a practical approach to prognostic meteorological and air pollution modelling. Environment Modelling & Software 20, 737–752. Jaecker-Voirol, A., Mirabel, P., 1989. Heteromolecular nucleation in the sulfuric acid–water system. Atmospheric Environment 23, 2053–2057.

Jonson, J., Mylona, S., Simpson, D., Tarrason, L., Schang, J., Solberg, S., Reis, S., Sundet, J., 1999. Transboundary photo-oxidant in Europe. Tech. Rep. 2/99, EMEP. Kim, Y., Seinfeld, J., Saxena, P., 1993. Atmospheric gas aerosol equilibrium I: thermodynamic model. Aerosol Science and Technology 19, 157–187. Kumar, N., Lurmann, F., Carter, W., 1995. Development of the flexible chemical mechanism version of the urban airshed model. Tech. Rep. STI-94470-1508-FR, California Air Resources Board, Sacramento, CA. Louka, P., Finzi, G., Volta, M., Colbeck, I., 2003. Photochemical smog in South European cities. In: Moussiopoulos, N. (Ed.), Air Quality in Cities. Springer, pp. 185–222. Marchuk, G., 1975. Methods of Numerical Mathematics. Springer, New York. Martilli, A., Neftel, A., Favaro, G., Kirchner, F., Sillman, S., Clappier, A., 2002. Simulation of the ozone formation in the northern part of the Po Valley. Journal of Geophysical Research 107 (D22). Meng, Z., Dabdub, D., Seinfeld, J., 1997. Chemical coupling between atmospheric ozone and particulate matter. Science 277, 116–119. Neftel, A., Spirig, C., Pre´vot, A., Furger, M., Stuz, J., Vogel, B., Hjorth, J., 2002. Sensitivity of photooxidant production in the Milan basin: an overview of results from a EUROTRAC-2 limitation of oxidant production field experiment. Journal of Geophysical Research 107 (D22). Ntziachristos, L., Samaras, Z., 2000. COPERT III computer programme to calculate emissions from road transport – methodology and emission factors. Tech. Rep. 49, EEA. Pandis, S., Harley, R., Cass, G., Seinfeld, J., 1992. Secondary organic aerosol formation and transport. Atmospheric Environment 26A, 2269–2282. Pirovano, G., Pertot, C., Gabusi, V., Volta, M., 2004. Evaluating seasonal model simulations of ozone in Northern Italy. In: Borrego, C., Incecik, S. (Eds.), Air Pollution Modeling and Its Application XVI. Kluwer, New York, pp. 171–178. Russel, A., Robin, D., 2000. NARSTO critical review of photochemical models and modeling. Atmospheric Environment 34, 2283–2324. San Jose´, R., Stohl, A., Karatzas, K., Bohler, T., James, P., Pe´rez, J., 2005. A modelling study of an extraordinary night time ozone epsiode over Madrid domain. Environmental Modelling & Software 20, 587–593. Scire, J., Insley, E., Yamartino, R., 1990. Model formulation and user’s guide for the CALMET meteorological model. Tech. Rep. A025-1. California Air Resources Board, Sacramento, CA. Seinfeld, J., Pandis, S., 1998. Atmospheric chemistry and physics. John Wiley & Sons. Silibello, C., Calori, G., Brusasca, G., Catenacci, G., Finzi, G., 1998. Application of a photochemical grid model to Milan metropolitan area. Atmospheric Environment 32 (11), 2025–2038. Sillman, S., 1999. The relation between Ozone, NOx and hydrocarbons in urban and polluted rural environments. Atmospheric Environment 33, 1821–1845. Simpson, D., Fagerli, H., Jonson, J., Tsyro, S., Wind, P., Tuovinen, J., 2003. Transboundary acidification, eutrophication and ground level ozone in Europe – Part I: unified EMEP Model description. Tech. Rep. 1/2003, EMEP. Smolarkiewicz, P.K., 1983. A simple positive definite advection scheme with small implicit diffusion. Monthly Weather Review 111, 479–486. Sun, P., Chock, D., Winkler, S., 1994. An implicit–explicit hybrid solver for a system of stiff kinetic equations. Journal of Computational Physics 115, 515–523. Vecchi, R., Valli, G., 1999. Ozone assessment in the southern part of the Alps. Atmospheric Environment 33, 97–109. Yamartino, R., Scire, J., Carmichael, G., Chang, Y., 1992. The CALGRID mesoscale photochemical model – I. Model formulation. Atmospheric Environment 26A, 1493–1512.