Chapter 4 Global and Regional Modelling
Effective implementation of policy to decrease ambient pollutant levels requires knowledge of the contribution of pollutant sources (e.g., from natural sources versus those due to human activities, or from local versus regional versus intercontinental sources) to the observed pollutant distributions. While observational methods provide clear evidence for intercontinental transport, and can help fingerprint specific source contributions (see Chapter 2), these measurements are generally not in operational networks, and are often not regionally representative. Thus at present, observation-based approaches alone cannot provide the information on source attribution (S/A) and
source/receptor (S/R) relationships needed to make better informed policy decisions about
hemispheric pollution. Numerical models incorporate our current understanding of the physical and chemical processes which control atmospheric composition, and are designed to predict the
relationships between emissions and the resulting pollution distributions in the atmosphere. These models allow us to explore the importance of different processes and to attribute observed
enhancements in pollutant concentrations to particular sources. In this chapter we discuss methods for quantifying intercontinental transport using numerical models, and summarize recent estimates of the magnitude and uncertainty in these effects, both from the published literature and from the model intercomparison organized under the Task Force (hereafter referred to as the HTAP intercomparison).
The capabilities and limitations of current models are analyzed, along with the sensitivity of S/R relationships to future changes in emissions and climate. Further activities needed to improve the modelling capabilities and the estimates of hemispheric transport of pollutants are identified.
4.1.1. Modelling approaches
Chemical transport models (CTMs) provide a representation of the chemical, transport and removal mechanisms that control the distributions of constituents in the atmosphere, as introduced in Chapter A1, based around a given set of emissions. They reflect our current understanding of these processes, which is informed by theoretical, laboratory and field experiments that focus on particular components of the system. A number of different types of models are useful for exploring
intercontinental transport, and we outline here the different approaches that may be taken,
summarizing their strengths, requirements and limitations. These range from point-based dispersion models, representing the transport and evolution of a single plume of pollutants from a source region, to regional or global-scale grid-based models that represent large-scale atmospheric composition.
Atmospheric dispersion models are commonly used to quantify ground-level pollutant concentrations and deposition <50 km downwind of point, line and area sources. These models have been evaluated at larger scales (~1000 km) with observations from field studies, e.g., CAPTEX (Cross-Appalachian Tracer Experiment) and ETEX (European Tracer Experiment), to build
confidence in their use in scientific, regulatory and policy applications [Galmarini et al., 2004; Klug et al., 1992]. Lagrangian particle dispersion models are now the most commonly used tools for meteorological based source attribution studies. These models initiate backward trajectories from a receptor location and calculate the transport pathways of air masses (often referred to as particles because they are assumed to be infinitesimally small) following the winds resolved in atmospheric analyses or forecasts. The models also include a stochastic step to represent the effects of unresolved turbulence and convection [Stohl et al., 2002]. The parameterised “random walk” shuffles particles, each weighted with the same tracer mass, so that their sum within a volume represents the effects of advection and diffusive mixing on passive tracer concentration [Legras et al., 2003]. The models
(described in more detail in Chapter 2) have not yet been developed to include chemical or aerosol processing or detailed descriptions of deposition processes where non-linearities in chemical transformation may be important.
Eulerian (gridded) CTMs are the most commonly used tools for studying intercontinental transport. CTMs divide the atmosphere into discrete grid-boxes in which transport and chemical processes are calculated. A schematic of the eulerian modelling approach is shown in Figure 4.1.
Typical resolutions are tens of vertical levels between the Earth‟s surface and the tropopause and a few degrees horizontally in global models to tens of kilometres or less in regional models. Common modes of operation are “off-line”, where a CTM ingests large volumes of meteorological data from a driving numerical weather prediction model or global atmospheric circulation (i.e., climate) model and “free-running”, where tracers are included directly in a global atmospheric circulation model which is often driven by observed sea surface temperatures (SSTs) for present-day simulations. In most studies to date that have estimated intercontinental S/R relationships, there is no coupling from the atmospheric chemistry back to the meteorology which determines the transport and dispersion of air pollution. CTMs require pollutant emission inventories and schemes to represent processes relevant to intercontinental transport such as advection, turbulent mixing, convection, chemical production and loss, wet scavenging and dry deposition. Many of these processes operate on scales smaller than resolved explicitly by the models, and so are heavily parameterized, increasing
uncertainty in the capabilities of current CTMs to represent quantitatively intercontinental transport.
New developments in modelling intercontinental transport at finer scales include nested-grid, global-to-regional modelling, which allows for inclusion of a regional window over a source or receptor continent with a higher spatial resolution embedded in a global domain. Higher resolution nested-grid models can resolve better the localized regions of intense upward advection and
convection responsible for lifting chemicals from the near-surface environment, a mechanism poorly represented in coarser-resolution global simulations [Chen et al., 2009b; Wang et al., 2004].
Figure 4.1. Schematic of the Eulerian forward modelling approach. The model represents various processes that impact pollutant transport, chemistry and removal of pollutants. These models can be applied in S/R and source attribution studies, where emissions from specific source regions are followed through the atmosphere.
4.1.2. Model methods for quantifying source contributions to intercontinental transport From scientific and policy perspectives it is useful to estimate the contribution of specific emission sources to air pollution levels. The distinction between source attribution (S/A), the contribution of a particular source to the absolute concentrations observed at a given location, and
source-receptor (S/R) relationship, the relative extent to which concentrations at a specific location change when a particular source is perturbed in an arbitrary manner, is made in Chapter 1, Section 1.3. Both approaches rely heavily on models for quantification, as independent verification from observations is difficult (see Chapter 2). The use of models for these source contribution applications is illustrated in Figure 4.1. For inert tracers, the approaches give similar results as the contribution of all sources at a particular receptor is additive [Seibert and Frank, 2004], and this is a good
approximation for most aerosol species [Liu et al., 2009]. However, for many reactive species such as ozone (O3), which have a non-linear dependence on precursor concentrations, the contribution of a given source change at a downwind receptor location is dependent on the magnitude of all other sources, not the specified source alone. Over intercontinental distances additional O3 precursor emissions typically lead to successively smaller increases in O3 over receptor regions, reflecting the reduced efficiency of O3 formation [Liu et al., 1987].
Two different approaches have been used to quantify the contribution of a specific source regionto the pollution amounts at a receptor area (see section 1.3.2). The emission sensitivity approach requires two simulations: 1) a standard „control‟ run that includes all emission sources and is designed so that predicted pollution levels can be directly compared with observations and 2) an additional simulation where emissions from a specific source region or category are perturbed. The contribution from that particular source is then determined by the difference between the predicted distributions from the two runs [Chin et al., 2007; Fiore et al., 2009; Jacob et al., 1999; Park et al., 2004; Wild and Akimoto, 2001; Wild et al., 2004a]. When relatively small emission changes are applied, this approach provides the sensitivity of receptor region responses around current conditions.
The second approach is the tagged tracer method, in which the distribution of idealized pollutanttracers is calculated, each of which is emitted or produced only inside a particular designated source region. The concentration of each pollutanttracer tagged by its source region then represents the contribution of that source region. Many independent source regions or categories can be tagged within a single model simulation, but this may require special model design. This method has been used in studies addressing a wide variety of issues using both global and regional chemistry transport models [Auvray and Bey, 2005; Bian et al., 2007; Fiore et al., 2002; Li et al., 2002; Li et al., 2008;
Sudo and Akimoto, 2007; Wang et al., 1998]. The method provides a good estimate of source contributions for directly-emitted primary pollutants, and has been shown to work well for most aerosol species [Liu et al., 2009]. However, it is problematic for catalytically-produced secondary pollutants such as O3 where rapid chemical interconversion precludes a simple representation of the source term. Application of this approach for O3 has either involved complex tagging of key precursors along with simplified assumptions about production [e.g., Horowitz et al., 1998], or a simpler tagging of O3 produced over particular regions [e.g., Sudo and Akimoto, 2007]. Although this latter approach is far simpler, it is domain-based rather than source-based, and therefore does not reflect the contribution of precursor emissions from a given source region. Due to the nature of the advection algorithms used in models, neither of these tagging approaches is entirely conservative, so typically the sum of the fractional contribution from all source regions does not equal unity.
The techniques discussed above are forward approaches to source contribution calculation. In these methods the impacts of emissions from a specific region are estimated for every grid cell within the modelling domain (e.g., globally). An alternative and complimentary approach is to focus on pollution levels at a particular receptor, and calculate only the contributions from those sources that contribute to pollution levels at this receptor. This receptor (or backward) approach can be thought of as running the models in reverse, and can be efficiently done using adjoint models [Carmichael et al., 2008], or for inert tracers using reverse particle models such as the FLEXPART model [Stohl et al., 1998]. The adjoint approach has been used in regional receptor-based source-contribution studies [Hakami et al., 2006], and in global O3 and aerosol applications [Henze et al., 2009; Zhang et al., 2009].
4.1.3. Role of coordinated model studies
The majority of studies summarized in this chapter have applied global or regional Eulerian models to explore the emission, transport and removal of O3, aerosols and their precursors. Many studies have focused on intercontinental transport from one or more source regions [e.g., Berntsen et
al., 1996; Bey et al., 2001], but there is little consistency between studies in terms of the techniques used, the spatial extent of source and receptor regions, or in the metrics applied to quantify transport downwind. Given the large variation in approaches and objectives of these studies it is hard to draw consistent conclusions from them or to identify and quantify key weaknesses in our current
understanding. To address this, five sets of coordinated model studies were performed as part of the HTAP intercomparison, adopting a coherent, standardized framework in the modelling approach, definition of regions, choice of meteorological year, and diagnostics to provide consistent estimates of the effects of intercontinental transport. The diversity of model results produced with this approach allows a more direct, quantitative assessment of some of the current uncertainty in our understanding of emissions, transport, and chemical or microphysical processes. These additional experiments are summarized in Table 4.1. The major intercomparison study (SR) focused on quantifying current source-receptor relationships for O3 and aerosols and characterizing model uncertainty, while
subsequent studies addressed specific aspects including the contribution of transport processes to this uncertainty (TP), ability to reproduce observed transport events (ES) and the likely effects of future changes in emissions (FE) and climate (FC).
Table 4.1. Summary of model studies performed under the HTAP intercomparison
Model Study Brief Description No. of
Participants SR: Source-receptor studies
Sensitivity study with 20% changes to anthropogenic emissions over four major source regions in 2001 to quantify source-receptor relationships
TP: Tracer process studies Repeat of SR with specified emissions and standardized tracers to
explore differences in model transport and mixing processes 25 ES: Event Simulations Explore model ability to reproduce specific intercontinental
transport events observed during the ICARTT campaign in 2004 7 FE: Future emissions studies Repeat of SR under different scenarios for future emissions in
2030 (RCP8.5) and 2050 (RCP2.6) conditions 4
FC: Future climate studies Repeat of SR under different climate conditions corresponding to
present-day and 2100 SRES-A2 climates 3
One advantage of this multi-model approach is that differences between models highlight the uncertainties associated with representation of key atmospheric processes. When combined with observational data, this allows identification of model weaknesses and a clearer assessment of model reliability. Additionally, ensemble mean results are often found to compare better with observations than those of any one model [e.g., Vautard et al., 2006]. In this chapter we draw on valuable
conclusions from previous model intercomparisons, in particular ACCENT/PHOTOCOMP [Dentener et al., 2006], AEROCOM [Textor et al., 2006], TRANSCOM [Law et al., 2008], and RETRO [Schultz et al., 2007] which provide additional insight into the strengths and weaknesses of current models.