Operational model evaluation for particulate matter in Europe and North America in the context of AQMEII
Introduction
Particulate matter (PM) is a worldwide environmental concern that threatens both human health and ecosystems (Manders et al., 2009; Aan de Brugh et al., 2011). Human exposure to high PM concentrations is associated with respiratory disease and shortened life expectancy (Amann et al., 2005). PM also contributes to acid rain, visibility degradation, and modification of the Earth's surface energy balance, and thus contributes to short-term climate forcing (Forster et al., 2007; Mebust et al., 2003; Appel et al., 2008; Smyth et al., 2009; Boylan and Russell, 2006). Recent studies have suggested that long-term changes in aerosol concentrations, primarily due to decreasing use of coal for energy production, have significantly influenced regional warming rates (Vautard et al., 2009; Philipona et al., 2009). Although major efforts are being made in Europe (EU) and North America (NA) (United States and Canada) to reduce anthropogenic emissions of primary PM and PM precursors, PM levels remain problematic and their adverse effects are expected to persist (Klimont et al., 2009). The characterisation of PM sources is an area of active research as many gaps in our knowledge of the chemical speciation of PM sources, the spatial and temporal distribution of airborne particles, and the physical and chemical transformations need to be filled. This is particularly true for regional air quality (AQ) models, which incorporate a wide range of PM physics and chemistry and consider a large variety of PM emissions sources. The problem is especially difficult when simulating long temporal periods and large spatial scales due to the variety of sources involved and the chemical and physical transformations of some species that can occur over long time periods (e.g., Mathur et al., 2008).
PM is a conglomerate of many different types of chemical components (i.e., elemental and organic carbon, ammonium, nitrates, sulphates, mineral dust, trace elements, and water) with varying physical and chemical properties. Particles are either emitted directly from a source or formed from the chemical and/or physical transformation of precursor species, which depend, among other factors, on particle size. Furthermore, given its composite nature, high PM concentrations might be observed at any time during the year and under a large variety of atmospheric conditions (unlike elevated ozone mixing ratios, which are typically associated with hot and stagnant conditions). A widely accepted classification of PM is based on particle size, with PM10 indicating those particles with an aerodynamic diameter between 0 and 10 μm, while PM2.5 indicates particles with an aerodynamic diameter less than 2.5 μm (note that PM10 includes PM2.5). This classification is dictated by the fact that the mechanisms for the generation, transformation, removal, chemical composition, and optical properties of the two classes of particles are notably different. They also behave differently in the human respiratory track, with PM2.5 penetrating deeper into the lungs (see, e.g., Seinfeld and Pandis (2006) for a detailed description of particle properties). In the past decade, PM2.5 has attracted considerably more attention than PM10 due to its greater potential to cause adverse effects on public health. As a result, AQ model development for PM has focused primarily on modeling PM2.5. This development is assisted by the availability of comprehensive PM2.5 measurements, which allows model performance to be evaluated for individual PM chemical components as well as total mass, which in turn allows deductions to be made about different aspects of model performance (e.g., the relationships between emissions, dispersion, chemistry, and deposition) (e.g., Yu et al., 2007, 2008).
The analysis presented in this paper is part of the Air Quality Model Evaluation International Initiative (AQMEII). The main objective of the project is to assess the state-of-science in current regional-scale AQ models in order to improve the ability of models to accurately characterize the spatial and temporal features embedded in air quality observations (Rao et al., 2011). Within AQMEII, standardised modeling outputs have been shared on the web-distributed ENSEMBLE system, allowing statistical and ensemble analyses to be performed (Bianconi et al., 2004; Galmarini et al., 2012). A cooperative exercise was launched for modeling groups to use their AQ models retrospectively to simulate the entire year 2006 for the continents of EU and NA. The primary goal of AQMEII is to evaluate the ability of regional AQ models to reconstruct (i.e., hindcast) atmospheric pollutant concentrations and not to forecast air quality.
While there have been a number of other model inter-comparison studies for PM (McKeen et al., 2007; Smyth et al., 2009; Stern et al., 2008; Vautard et al., 2009; Hayami et al., 2008), the scale of the model evaluation and inter-comparison presented in this study is unprecedented given the number of models evaluated, the spatial extent of the model domains, and the amount of observational data collected over the two continents.
In this paper we focus on the evaluation of PM in EU and NA using simulated results from ten different state-of-the-science regional AQ models run by 15 independent groups from both continents. In addition to the model results, observational PM data have also been gathered and made available on the ENSEMBLE system. A companion paper by Solazzo et al. (2012) has reported on a related AQMEII multi-model evaluation for ozone.
Section snippets
AQ models and monitoring
Predictions from the regional AQ models participating in the AQMEII exercise are compared to observations over the full year of 2006. Modeling groups have provided gridded surface concentrations of PM10, PM2.5, and selected gaseous compounds (e.g., SO2 and NO2) for two continental areas: 15°W to 35°E and 35°N to 70°N for EU; and 130°W to 58.5°W and 23.5°N to 59.5°N for NA. In addition to the gridded surface concentrations, the modeling groups also provided interpolated model concentrations at
PM10 evaluation and model cross-comparison
In the following sections, the different AQ model simulations are denoted by the labels Mod1 to Mod10 for EU and Mod12 to Mod18 for NA. In some cases the same model, but with a different configuration, was run for both continents. Such is the case for Mod3 and Mod18, Mod4 and Mod13, and Mod10 and Mod17. Note that no direct correspondence exists between these model labels and the list of models provided in Table 1 so as to preserve anonymity.
In this section, model simulations and observations
Time series and statistics
Time series of monthly, continental mean PM2.5 surface concentrations, based on stations providing daily measurements, are shown in Fig. 9 for EU and NA. Compared to PM10 (Fig. 4), model bias is much lower for both continents, demonstrating an enhanced capability of the AQ models to simulate PM2.5. For EU, the models underestimate the monthly mean PM2.5 surface concentrations for all sub-regions, with several exceptions. In particular, the models that overestimated PM10 concentrations (Mod1,
Two episodes with elevated PM concentrations
Two episodes with elevated PM concentrations, one in EU and one in NA, have been selected for a more detailed investigation of the models' performance. It is important to determine that the AQ models not only capture the average PM concentrations correctly, but that they also reproduce the peak values as well.
For EU, a period of 16 days between 13 and 28 April 2006 was chosen for detailed analysis. During this period, elevated PM2.5 concentrations were observed at several stations in central
Conclusions
Annual AQ model simulations in the context of AQMEII have been inter-compared and evaluated. The focus has been put on surface concentrations of PM, both fine and coarse. Given the scale of the project - involving ten AQ models that were run over two continents (EU and NA) for one entire year, 2006 – the available model results allow for a comprehensive analysis.
We have analysed predictions of PM10 and PM2.5 in several sub-regions of the continental domains, quantifying bias and model
Acknowledgments
The work carried out with the DEHM model was supported by The Danish Strategic Research Program on Sustainable Energy under contract no 2104-06-0027 (CEEH). Homepage: www.ceeh.dk. The RSE contribution to this work has been partially financed by the Research Fund for the Italian Electrical System under the Contract Agreement between RSE and the Italian Ministry of Economic Development (Decree of March 19th, 2009).
References (91)
- et al.
Modal aerosol dynamics model for Europe: development and first applications
Atmospheric Environment
(1998) - et al.
Aerosol modeling with CHIMERE: preliminary evaluation at the continental scale
Atmospheric Environment
(2004) - et al.
Web-based system for decision support in case of emergency: ensemble modelling of long-range atmospheric dispersion of radionuclides
Environmental Modelling and Software
(2004) - et al.
PM and light extinction model performance metrics, goal, and criteria for three-dimensional air quality models
Atmospheric Environment
(2006) The Danish eulerian hemispheric Model – a three-dimensional air pollution model used for the Arctic
Atmospheric Environment
(1997)- et al.
Effect of aerosol-radiation feedback on regional air quality – A case study with WRF/Chem
Atmospheric Environment
(2012) - et al.
ENSEMBLE and AMET: two systems and approaches to a harmonised, simplified and efficient assistance to air quality model developments and evaluation
Atmospheric Environment
(2012) - et al.
Cloud processing of gases and aerosols in a regional air quality model (AURAMS)
Atmospheric Research
(2006) - et al.
Fully coupled online chemistry within the WRF model
Atmospheric Environment
(2005) - et al.
Natural volatile organic compound emission rate estimates for US woodland landscapes
Atmospheric Environment
(1994)