Emission Strength Validation Using Four-Dimensional Data Assimilation: Application to Primary Aerosol and Precursors to Ozone and Secondary Aerosol

ABSTRACT

Three-dimensional air quality models (AQMs) represent the most powerful tool to follow the dynamics of air pollutants at urban and regional scales. Current AQMs can account for the complex interactions between gas-phase chemistry, aerosol growth, cloud and scavenging processes, and transport. However, errors in model applications still exist due in part to limitations in the models themselves and in part to uncertainties in model inputs. Four-dimensional data assimilation (FDDA) can be used as a top-down tool to validate several of the model inputs, including emissions inventories, based on ambient measurements. Previously, this FDDA technique was used to estimate adjustments in the strength and composition of emissions of gas-phase primary species and O₃ precursors.

In this paper, we present an extension to the FDDA technique to incorporate the analysis of particulate matter (PM) and its precursors. The FDDA approach consists of an iterative optimization procedure in which an AQM is coupled to an inverse model, and adjusting the emissions minimizes the difference between ambient measurements     

Assimilation of conventional and satellite wind observations in a mesoscale atmospheric model for studying atmospheric dispersion

A mesoscale atmospheric model PSU/NCAR MM5 is used to provide operational weather forecasts for a nuclear emergency response decision support system on the southeast coast of India. In this study the performance of the MM5 model with assimilation of conventional surface and upper-air observations along with satellite derived 2-d surface wind data from QuickSCAT sources is examined. Two numerical experiments with MM5 are conducted: one with static initialization using NCEP FNL data and second with dynamic initialization by assimilation of observations using four dimensional data assimilation (FDDA) analysis nudging for a pre-forecast period of 12 h. Dispersion simulations are conducted for a hypothetical source at Kalpakkam location with the HYSPLIT Lagrangian particle model using simulated wind field from the above experiments. The present paper brings out the differences in the atmospheric model predictions and the differences in dispersion model results from control and assimilation runs. An improvement is noted in the atmospheric fields from the assimilation experiment which has led to significant alteration in the trajectory positions, plume orientation and its distribution pattern. Sensitivity tests using different PBL and surface parameterizations indicated the simple first order closure schemes (Blackadar, MRF) coupled with the simple soil model have given better results for various atmospheric fields. The study illustrates the impact of the assimilation of the scatterometer wind and automated weather stations (AWS) observations on the meteorological model predictions and the dispersion results.     

Emission strength validation using four-dimensional data assimilation: application to primary aerosol and precursors to ozone and secondary aerosol.

Three-dimensional air quality models (AQMs) represent the most powerful tool to follow the dynamics of air pollutants at urban and regional scales. Current AQMs can account for the complex interactions between gas-phase chemistry, aerosol growth, cloud and scavenging processes, and transport. However, errors in model applications still exist due in part to limitations in the models themselves and in part to uncertainties in model inputs. Four-dimensional data assimilation (FDDA) can be used as a top-down tool to validate several of the model inputs, including emissions inventories, based on ambient measurements. Previously, this FDDA technique was used to estimate adjustments in the strength and composition of emissions of gas-phase primary species and O3 precursors. In this paper, we present an extension to the FDDA technique to incorporate the analysis of particulate matter (PM) and its precursors. The FDDA approach consists of an iterative optimization procedure in which an AQM is coupled to an inverse model, and adjusting the emissions minimizes the difference between ambient measurements and model-derived concentrations. Here, the FDDA technique was applied to two episodes, with the modeling domain covering the eastern United States, to derive emission adjustments of domainwide sources of NO., volatile organic compounds (VOCs), CO, SO2, NH3, and fine organic aerosol emissions. Ambient measurements used include gas-phase inorganic and organic species and speciated fine PM. Results for the base-case inventories used here indicate that emissions of SO2 and CO appear to be estimated reasonably well (requiring minor revisions), while emissions of NOx, VOC, NH3, and organic PM with aerodynamic diameter less than 2.5 microm (PM2.5) require more significant revision.     

Assessment of the nested grid model estimates for driving regional visibility models in the southwestern United States

The Nested Grid Model (NGM) is a primitive-equation meteorological model that is routinely exercised over North America for forecasting purposes by the National Meteorological Center. While prognostic meteorological models are being increasingly used to drive air quality models, their use in conducting annual simulations requires significant resources. NGM estimates of wind fields and other meteorological variables provide an attractive alternative since they are typically archived and readily available for an entire year. Preliminary evaluation of NGM winds during the summer of 1992 for application to the region surrounding the Grand Canyon National Park showed serious shortcomings. The NGM winds along the borders between California, Arizona and Mexico tend to be northwesterly with a speed of about 6 m/sec, while the observed flow is predominantly southerly at about 2-5 m/sec. The mesoscale effect of a thermal low pressure area over the highly heated Southern California and western Arizona deserts does not appear to be represented by the NGM because of its coarse resolution and the use of sparse observations in that region. Tracer simulations and statistical evaluation against special high resolution observations of winds in the southwest United States clearly demonstrate the northwest bias in NGM winds and its adverse effect on predictions of an air quality model. The "enhanced" NGM winds, in which selected wind observations are incorporated in the NGM winds using a diagnostic meteorological model provide additional confirmation on the primary cause of the northwest bias. This study has demonstrated that in situations where limited resources prevent the use of prognostic meteorological models, previously archived coarse resolution wind fields in which additional observations are incorporated to correct known biases provide an attractive option.     

Assessment of the Nested Grid Model Estimates for Driving Regional Visibility Models in the Southwestern United States

ABSTRACT

The Nested Grid Model (NGM) is a primitive-equation meteorological model that is routinely exercised over North America for forecasting purposes by the National Meteorological Center. While prognostic meteorological models are being increasingly used to drive air quality models, their use in conducting annual simulations requires significant resources. NGM estimates of wind fields and other meteorological variables provide an attractive alternative since they are typically archived and readily available for an entire year. Preliminary evaluation of NGM winds during the summer of 1992 for application to the region surrounding the Grand Canyon National Park showed serious shortcomings. The NGM winds along the borders between California, Arizona and Mexico tend to be northwesterly with a speed of about 6 m/sec, while the observed flow is predominantly southerly at about 2-5 m/sec. The mesoscale effect of a thermal low pressure area over the highly heated Southern California and western Arizona deserts does not appear to be represented by the NGM because of its coarse resolution and the use of sparse observations in that region. Tracer simulations and statistical evaluation against special high resolution observations of winds in the southwest United States clearly demonstrate the northwest bias in NGM winds and its adverse effect on predictions of an air quality model. The “enhanced” NGM winds, in which selected wind observations are incorporated in the NGM winds using a diagnostic meteorological model provide additional confirmation on the primary cause of the northwest bias. This study has demonstrated that in situations where limited resources prevent the use of prognostic meteorological models, previously archived coarse resolution wind fields in which additional observations are incorporated to correct known biases provide an attractive option.     

Air Quality Modeling in the South Coast Air Basin of California: What Do the Numbers Really Mean?

Abstract

This study evaluates air quality model sensitivity to input and to model components. Simulations are performed using the California Institute of Technology (CIT) airshed model. Results show the impacts on ozone (O₃) concentration in the South Coast Air Basin (SCAB) of California because of changes in: (1) input data, including meteorological conditions (temperature, UV radiation, mixing height, and wind speed), boundary conditions, and initial conditions (ICs); and (2) model components, including advection solver and chemical mechanism. O₃ concentrations are strongly affected by meteorological conditions and, in particular, by temperature. ICs also affect O₃ concentrations, especially in the first 2 days of simulation. On the other hand, boundary conditions do not significantly affect the absolute peak O₃ concentration, although they do affect concentrations near the inflow boundaries. Moreover, predicted O₃ concentrations are impacted considerably by the chemical mechanism. In addition, dispersion of pollutants is affected by the advection routine used to calculate its transport. Comparison among CIT, California Photochemical Grid Model (CALGRID), and Urban Airshed Model air quality models suggests that differences in O₃ predictions are mainly caused by the different chemical mechanisms used. Additionally, advection solvers contribute to the differences observed among model predictions. Uncertainty in predicted peak O₃ concentration suggests that air quality evaluation should not be based solely on this single value but also on trends predicted by air quality models using a number of chemical mechanisms and with an advection solver that is mass conservative.     

Coupling of the Weather Research and Forecasting Model with AERMOD for pollutant dispersion modeling. A case study for PM10 dispersion over Pune,India

The prediction of spatial variation of the concentration of a pollutant governed by various sources and sinks is a complex problem. Gaussian air pollutant dispersion models such as AERMOD of the United States Environmental Protection Agency (USEPA) can be used for this purpose. AERMOD requires steady and horizontally homogeneous hourly surface and upper air meteorological observations. However, observations with such frequency are not easily available for most locations in India. To overcome this limitation, the planetary boundary layer and surface layer parameters required by AERMOD were computed using the Weather Research and Forecasting (WRF) Model (version 2.1.1) developed by the National Center for Atmospheric Research (NCAR). We have developed a preprocessor for offline coupling of WRF with AERMOD. Using this system, the dispersion of respirable particulate matter (RSPM/PM10) over Pune, India has been simulated. Data from the emissions inventory development and field-monitoring campaign (13–17 April 2005) conducted under the Pune Air Quality Management Program of the Ministry of Environment and Forests (MoEF), India and USEPA, have been used to drive and validate AERMOD. Comparison between the simulated and observed temperature and wind fields shows that WRF is capable of generating reliable meteorological inputs for AERMOD. The comparison of observed and simulated concentrations of PM10 shows that the model generally underestimates the concentrations over the city. However, data from this single case study would not be sufficient to conclude on suitability of regionally averaged meteorological parameters for driving Gaussian models like AERMOD and additional simulations with different WRF parameterizations along with an improved pollutant source data will be required for enhancing the reliability of the WRF–AERMOD modeling system.     

Modeling air quality during the California Regional PM10/PM2.5 Air Quality Study (CPRAQS) using the UCD/CIT source-oriented air quality model – Part III. Regional source apportionment of secondary and total airborne particulate matter

A comprehensive air quality modeling project was carried out to simulate regional source contributions to secondary and total (=primary + secondary) airborne particle concentrations in California's Central Valley. A three-week stagnation episode lasting from December 15, 2000 to January 7, 2001, was chosen for study using the air quality and meteorological data collected during the California Regional PM₁₀/PM_2.5 Air Quality Study (CRPAQS). The UCD/CIT mechanistic air quality model was used with explicit decomposition of the gas phase reaction chemistry to track source contributions to secondary PM. Inert artificial tracers were used with an internal mixture representation to track source contributions to primary PM. Both primary and secondary source apportionment calculations were performed for 15 size fractions ranging from 0.01 to 10 μm particle diameters. Primary and secondary source contributions were resolved for fugitive dust, road dust, diesel engines, catalyst equipped gasoline engines, non-catalyst equipped gasoline engines, wood burning, food cooking, high sulfur fuel combustion, and other anthropogenic sources.Diesel engines were identified as the largest source of secondary nitrate in central California during the study episode, accounting for approximately 40% of the total PM_2.5 nitrate. Catalyst equipped gasoline engines were also significant, contributing approximately 20% of the total secondary PM_2.5 nitrate. Agricultural sources were the dominant source of secondary ammonium ion. Sharp gradients of PM concentrations were predicted around major urban areas. The relative source contributions to PM_2.5 from each source category in urban areas differ from those in rural areas, due to the dominance of primary OC in urban locations and secondary nitrate in the rural areas. The source contributions to ultra-fine particle mass PM_0.1 also show clear urban/rural differences. Wood smoke was found to be the major source of PM_0.1 in urban areas while motor vehicle sources were the major contributor of PM_0.1 in rural areas, reflecting the influence from two major highways that transect the Valley.     

Emission Reductions Needed to Meet the Standard for Ozone in Southern California: Effect of Boundary Conditions

Abstract

Combinations of total reactive organic gas (ROG) and nitrogen oxide (NO_x) emissions that do not exceed the National Ambient Air Quality Standard (NAAQS) for ozone for the meteorological conditions of the August 26-28, 1987 SCAQS episode, have been determined using the California Institute of Technology (CIT) photochemical air quality model. The sensitivity of these combinations to pollutant boundary conditions is examined.     

Modeling the long-term frequency distribution of regional ozone concentrations

Efficient methods are developed for modeling emissions – air quality relationships that govern ozone and NO₂ concentrations over very long periods of time. A baseline model evaluation study is conducted to assess the accuracy and speed with which the relationship between pollutant emissions and the frequency distribution of O₃ concentrations throughout the year can be computed along with annual average NO₂ values using a deterministic photochemical airshed model driven by automated objective analysis of measured meteorological parameters. Methods developed are illustrated by application to the air quality situation that exists in Southern California. Model performance statistics for O₃ are similar to the results obtained in previous short-term episodic model evaluation studies that were based on hand-crafted meteorological inputs that are supplemented by expensive field measurement campaigns. Model predictions at one of the highest NO₂ concentration sites in the US indicate that measured violation of the US annual average NO₂ air quality standard at that site occurs because other species such as HNO₃ and PAN are measured as if they were NO₂ by the chemiluminescent NO_x monitors in current use.     

An enhanced PM2.5 air quality forecast model based on nonlinear regression and back-trajectory concentrations

An enhanced PM_2.5 air quality forecast model based on nonlinear regression (NLR) and back-trajectory concentrations has been developed for use in the Louisville, Kentucky metropolitan area. The PM_2.5 air quality forecast model is designed for use in the warm season, from May through September, when PM_2.5 air quality is more likely to be critical for human health. The enhanced PM_2.5 model consists of a basic NLR model, developed for use with an automated air quality forecast system, and an additional parameter based on upwind PM_2.5 concentration, called PM24. The PM24 parameter is designed to be determined manually, by synthesizing backward air trajectory and regional air quality information to compute 24-h back-trajectory concentrations. The PM24 parameter may be used by air quality forecasters to adjust the forecast provided by the automated forecast system. In this study of the 2007 and 2008 forecast seasons, the enhanced model performed well using forecasted meteorological data and PM24 as input. The enhanced PM_2.5 model was compared with three alternative models, including the basic NLR model, the basic NLR model with a persistence parameter added, and the NLR model with persistence and PM24. The two models that included PM24 were of comparable accuracy. The two models incorporating back-trajectory concentrations had lower mean absolute errors and higher rates of detecting unhealthy PM2.5 concentrations compared to the other models.     

Behavior of particulate matter during high concentration episodes in Seoul

The behavior of particulate matter (PM) during high-concentration episodes was investigated using monitoring data from Guui station, a comprehensive air monitoring station in Seoul, Korea, from January 2008 to March 2010. Five non-Asian dust (ND) episodes and two Asian dust (AD) episodes of high PM concentrations were selected for the study. During the ND episode, primary air pollutants accumulated due to low wind speeds, and PM_2.5 increased along with most other air pollutants. Particles larger than PM_2.5 were also high since these particles were generated by vehicular traffic rather than wind erosion. During strong AD episodes, PM_10–2.5 primarily increased and gaseous primary air pollutants decreased under high wind speeds. However, even during the AD episode, PM_2.5 and gaseous primary air pollutants increased when the effects of AD were weak and wind speeds were low. This study corroborates that accumulation of air pollutants due to a drop in surface wind speed plays an important role in short-term high-concentration occurrences. However, low wind speeds could not be directly linked to local emissions because a significant portion of accumulated air pollutants resulted from long-range transport.     

Development and implementation of a remote-sensing and in situ data-assimilating version of CMAQ for operational PM2.5 forecasting. Part 1: MODIS aerosol optical depth (AOD) data-assimilation design and testing

This two-part paper reports on the development, implementation, and improvement of a version of the Community Multi-Scale Air Quality (CMAQ) model that assimilates real-time remotely-sensed aerosol optical depth (AOD) information and ground-based PM_2.5 monitor data in routine prognostic application. The model is being used by operational air quality forecasters to help guide their daily issuance of state or local-agency-based air quality alerts (e.g. action days, health advisories). Part 1 describes the development and testing of the initial assimilation capability, which was implemented offline in partnership with NASA and the Visibility Improvement State and Tribal Association of the Southeast (VISTAS) Regional Planning Organization (RPO). In the initial effort, MODIS-derived aerosol optical depth (AOD) data are input into a variational data-assimilation scheme using both the traditional Dark Target and relatively new “Deep Blue” retrieval methods. Evaluation of the developmental offline version, reported in Part 1 here, showed sufficient promise to implement the capability within the online, prognostic operational model described in Part 2. In Part 2, the addition of real-time surface PM_2.5 monitoring data to improve the assimilation and an initial evaluation of the prognostic modeling system across the continental United States (CONUS) is presented.

Implications: Air quality forecasts are now routinely used to understand when air pollution may reach unhealthy levels. For the first time, an operational air quality forecast model that includes the assimilation of remotely-sensed aerosol optical depth and ground based PM_2.5 observations is being used. The assimilation enables quantifiable improvements in model forecast skill, which improves confidence in the accuracy of the officially-issued forecasts. This helps air quality stakeholders be more effective in taking mitigating actions (reducing power consumption, ride-sharing, etc.) and avoiding exposures that could otherwise result in more serious air quality episodes or more deleterious health effects.     

Modeling wintertime particulate matter formation in central California

A wintertime episode during the 2000 California Regional PM Air Quality Study (CRPAQS) was simulated with the air quality model CMAQ–MADRID. Model performance was evaluated with 24-h average measurements available from CRPAQS. Modeled organic matter (OM) was dominated by emissions, which were probably significantly under-represented, especially in urban areas. In one urban area, modeled daytime nitrate concentrations were low and evening concentrations were high. This diurnal profile was not explained by the partition of nitrate between the gas and particle phases, because gaseous nitric acid concentrations were low compared to PM nitrate. Both measured and simulated nitrate concentrations aloft were lower than at the surface at two tower locations during this episode. Heterogeneous reactions involving NO₃ and N₂O₅ accounted for significant nitrate production in the model, resulting in a nighttime peak. The sensitivity of PM nitrate to precursor emissions varied with time and space. Nitrate formation was on average sensitive to NO_x emissions. However, for some periods at urban locations, reductions in NO_x caused the contrary response of nitrate increases. Nitrate was only weakly sensitive to reductions in anthropogenic VOC emissions. Nitrate formation tended to be insensitive to the availability of ammonia at locations with high nitrate, although the spatial extent of the nitrate plume was reduced when ammonia was reduced. Reductions in PM emissions caused OM to decrease, but had no effect on nitrate despite the role of heterogeneous reactions. A control strategy that focuses on NO_x and PM emissions would be effective on average, but reductions in VOC and NH₃ emissions would also be beneficial for certain times and locations.     

A gas/aerosol air pollutants study over the urban area of Rome using a comprehensive chemical transport model

The aim of this study was to identify areas of potential relevant exposure to pollutants within Rome's urban core. To meet this goal, intensive field campaigns were conducted and simulations were performed, using the flexible air quality regional model (FARM), to study winter and summer pollution episodes. The simulations were performed using a complete emission inventory that included traffic flow model results of the Roman street network to better describe, with respect to the available diffuse national emission inventory, the hourly variation of traffic emissions in the city. The meteorological reconstruction was performed by means of both prognostic and diagnostic models by using experimental data collected during the field campaigns. To evaluate the capability of the FARM model to capture the main features of the selected episodes, a comparison of modelled results against observed air quality data for different pollutants was performed at urban and rural sites. FARM performed well in predicting ozone (O₃) and nitrogen dioxide (NO₂) concentrations, showing a good reproduction of both daily peaks and their diurnal variations. The model also showed a good capability to reproduce the magnitude of volatile alkane, aromatic and carbonyl compound concentrations. PM₁₀ model results revealed the tendency to under-predict the observed values. PM composition model results were compared with observed data, evidencing good results for elemental carbon (EC), nitrate (NO₃⁻) and ammonium (NH₄⁺), underestimation for sulphate (SO₄²⁻) and poor performance for organic matter (OM). The soil components of PM were found to be significantly under-predicted by the model, especially during Saharan dust episodes. Overall, the study results show large areas of high O₃ and PM₁₀ concentrations where levels of pollutants should be carefully monitored and population exposure evaluated.     

Simulating Urban-Scale Air Pollutants and Their Predicting Capabilities over the Seoul Metropolitan Area

Abstract

Urban-scale air pollutants for sulfur dioxide, nitrogen dioxide, particulate matter with aerodynamic diameter >10 μm, and ozone (O₃) were simulated over the Seoul metropolitan area, Korea, during the period of July 2-11, 2002, and their predicting capabilities were discussed. The Air Pollution Model (TAPM) and the highly disaggregated anthropogenic and the biogenic gridded emissions (1 km × 1 km) recently prepared by the Korean Ministry of Environment were applied. Wind fields with observational nudging in the prognostic meteorological model TAPM are optionally adopted to comparatively examine the meteorological impact on the prediction capabilities of urban-scale air pollutants. The result shows that the simulated concentrations of secondary air pollutant largely agree with observed levels with an index of agreement (IOA) of >0.6, whereas IOAs of ～0.4 are found for most primary pollutants in the major cities, reflecting the quality of emission data in the urban area. The observationally nudged wind fields with higher IOAs have little effect on the prediction for both primary and secondary air pollutants, implying that the detailed wind field does not consistently improve the urban air pollution model performance if emissions are not well specified. However, the robust highest concentrations are better described toward observations by imposing observational nudging, suggesting the importance of wind fields for the predictions of extreme concentrations such as robust highest concentrations, maximum levels, and >90th percentiles of concentrations for both primary and secondary urban-scale air pollutants.     

Air quality modeling in the South Coast Air Basin of California: what do the numbers really mean?

This study evaluates air quality model sensitivity to input and to model components. Simulations are performed using the California Institute of Technology (CIT) airshed model. Results show the impacts on ozone (O3) concentration in the South Coast Air Basin (SCAB) of California because of changes in: (1) input data, including meteorological conditions (temperature, UV radiation, mixing height, and wind speed), boundary conditions, and initial conditions (ICs); and (2) model components, including advection solver and chemical mechanism. O3 concentrations are strongly affected by meteorological conditions and, in particular, by temperature. ICs also affect O3 concentrations, especially in the first 2 days of simulation. On the other hand, boundary conditions do not significantly affect the absolute peak O3 concentration, although they do affect concentrations near the inflow boundaries. Moreover, predicted O3 concentrations are impacted considerably by the chemical mechanism. In addition, dispersion of pollutants is affected by the advection routine used to calculate its transport. Comparison among CIT, California Photochemical Grid Model (CALGRID), and Urban Airshed Model air quality models suggests that differences in O3 predictions are mainly caused by the different chemical mechanisms used. Additionally, advection solvers contribute to the differences observed among model predictions. Uncertainty in predicted peak O3 concentration suggests that air quality evaluation should not be based solely on this single value but also on trends predicted by air quality models using a number of chemical mechanisms and with an advection solver that is mass conservative.     

Revised estimates of construction activity and emissions: Effects on ozone and elemental carbon concentrations in southern California

Emissions from diesel-powered construction equipment are an important source of nitrogen oxides (NO_x) and particulate matter (PM). A new emission inventory for construction equipment emissions is developed based on surveys of diesel fuel use; the revised inventory is compared to current emission inventories. California's OFFROAD model estimates are 4.5 and 3.1 times greater, for NO_x and PM respectively, than the fuel-based estimates developed here. The most relevant uncertainties are the overall amount of construction activity/diesel fuel use, exhaust emission factors for PM and NO_x, and the spatial allocation of emissions to county level and finer spatial scales. Construction permit data were used in this study to estimate spatial distributions of emissions; the resulting distribution is well correlated with population growth. An air quality model was used to assess the impacts of revised emission estimates. Increases of up to 15 ppb in predicted peak ozone concentrations were found in southern California. Elemental carbon and fine particle mass concentrations were in better agreement with observations using revised emission estimates, whereas negative bias in predictions of ambient NO_x concentrations increased.     

Characterization of Particulate Matter in California

ABSTRACT

The size, composition, and concentration of particulate matter (PM) vary with location and time. Several monitoring/sampling programs are operated in California to characterize PM less than 2.5 and 10 µm in aerodynamic diameter (PM_2.5 and PM₁₀). This paper presents a broad summary of the spatial and temporal variations observed in ambient PM_2.5 and PM₁₀ concentrations in California. Many areas that have high PM₁₀ concentrations also have relatively high PM_2.5 concentrations, and data indicate that a significant portion of the PM₁₀ air quality problem is caused by PM_2.5. To develop effective plans for attaining the ambient PM standards, improved understanding of these unique problems is needed. Since 1989, pollution control efforts—whether specifically targeted for particulate matter or indirectly via controls on gaseous emissions—have caused annual average PM_2.5 and PM₁₀ concentrations to decline at most sites in California.     

Episode selection for ozone modelling and control strategies analysis on the Swiss Plateau

An episode selection procedure was developed and applied to select sets of days representing characteristic meteorological conditions leading to high ozone episodes over the Swiss Plateau. The selection procedure was applied to data extending from January 1991 through December 1998, and is comprised of two steps: First, days were classified according to observed air quality and meteorological characteristics using classification and regression trees analysis (CART). Second, the CART results were used in conjunction with observed air quality data to identify sets of days characteristic of those leading to elevated ozone. These sets of days were selected to optimise how well a limited number of days represented seasonal air quality, and that formed longer episodes for use in the air quality modelling. CART analysis was performed for three zones of the Swiss Plateau that have different air quality and meteorological characteristics. The results for two zones were used together in the episode selection procedure in order to identify days representative for the whole Plateau. Meteorological analysis for a third zone suggested that it would be strongly impacted by pollutants transported in from outside the country. One thousand and eight hundred optimisation runs were performed to minimise the likelihood that the set of days was a local optimum, increasing the robustness for use in air quality modelling analysis. Fifteen days, grouped in four episodes ranging from 3 to 5 days were selected along with their calculated representativeness (or weight) to recreate a seasonal metric. The variety of local as well as regional meteorological characteristics showed that the episode selection procedure chose days representing a diverse set of meteorological situations which are associated with elevated ozone. This set of episodes can now be used to test air quality strategies.