[Two figures] depict the framework of a comprehensive modeling system. An air quality model typically includes:
The detailed modular formation varies from model to model. It is possible, however, to formulate a general modular framework that is common to most three-dimensional modeling systems. First, the spatial and temporal resolutions of the modeling system must be defined. The spatial distribution of meteorological and chemical variables is approximated by three-dimensional gridded systems. The meteorological and the air quality models may have different grid structures over the same domain. For example, the meteorological model may use a system based on altitude (with respect to mean sea level), whereas the air quality model may use a terrain-following coordinate system. The output of the meteorological model will need to be processed to provide meteorological fields that match the gridded system of the air quality model. The emissions model uses a gridded spatial resolution that is consistent with that of the air quality model. The spatial resolution does not need to be uniform throughout the domain. In the vertical direction, meteorological and air quality models typically use a finer resolution near ground level than aloft. In addition, nesting of domains with different horizontal resolutons may be performed to accommodate the need for fine spatial resolution (e.g., on the order of 1-5 km) in critical source or receptor areas without penalizing the computational cost over the entire domain (where a larger horizontal grid size of the order of 20 km would be used).
The temporal resolution used in meteorological, emissions, and air quality modls is generally 1 h. (We define temporal resolution as time averaging for model inputs and outputs, not the integration step time used in a model, which is generally on the order of seconds or minutes.) All air quality ambient standards are based on 1-h or longer averaging times. Many meteorological data are routinely available with 1-h resolution.
The three-dimensional field of meteorological variables can be constructed by a diagnostic model that uses interpolation techniques to develop a three-dimensional field based on a discrete set of data or by a dynamic (or prognostic) model that solves the fundamental equations of mass, momentum, and energy to calculate the three-dimensional field of meteorological variables. Diagnostic models are useful if:
These constraints are not always met in the [study area] domain, even during the intensive measurement periods. Unless the observed meteorological data input into the air quality model can reproduce the complex four-dimensional dispersion patterns likely to be found, air parcels could be mislocated by more than 100 km in the horizontal and upward of 500-1,000 in the vertical after only 12 h of simulation time.
Thus, it appears that a dynamic modeling approach is needed for [this] study. However, a dynamic meteorological model must be able to simulate not only the mesoscale alpha (scales of 100-1,000 km) meteorological phenomenon, but also the mesoscale beta and gamma (scales of 1-100 km) effects that may be critical for determining the transport and dispersion of pollutants. A dynamic meteorological model is usually based on the so-called primitive equations defined [later]. The values of the wind velociy vector, temperature, pressure, and relative humidity are calculated in each grid cell of the meteorological modeling domain and transfer of mass, momentum, and energy between the grid cells is governed by the primitive equations. Some meteorological models also calculate turbulent diffusion, cloud water, and precipitation.
The air quality model is based primarily on the atmospheric diffusion equation that governs the emission, advection, diffusion, reaction, and removal of chemical species. The concentrations of gaseous, aerosol, and aqueous chemical species are calculated as a function of time in each grid cell of the modeling domain. The effect of chemistry on the concentrations of various species is calculated for each grid cell. Transfer of chemical species between the grid cells is governed by the transport and diffusion model. In addition, transfer of chemical species to the surface is calculated by the dry deposition module for gaseous species, the dry deposition and sedimentation module for aerosols, and the wet deposition module for dissolved species. Cloud and fog processes are generally treated in a fractional volume of grid cells that contain condensed water. The cloud/fog physics module calculates dynamic parameters such as updraft velocity, liquid water content, and precipitation rate. The updraft velocity is used in the transport equations to represent convective transport of chemical species due to clouds. The liquid-water content is a key input to the droplet chemistry module. Cloud/fog chemistry includes gas-phase and droplet chemistry; however, some air quality models neglect gas-phase chemical reactions in clouds and fog. The precipitation rate and droplet concentrations are the major input to the wet deposition model. Some wet deposition modules also consider the scavagening of gases and aerosols below the cloud base.
Inputs to these models or model components include initial and boundary conditions for meteorological variables and chemical (gas, aerosol, and droplet) concentrations. Other inputs such as location, time of year, terrain characteristics, and emission rates are also required. The sequence of the components shown in [the graphic] does not correspond to the actual organization of these components within the air quality model. Air quality models are often organized differently. For example, within a model time step, a component may be called several times by the main program subroutine. (A typical example is the equilibrium between the aerosol and gas phase, and droplets and gas phase, which can be recalculated at several occasions within a time step.) [The graphic] shows the flow of information from model components to other model components. However, the flow chart does not provide a detailed description of the input and output data of the various components of the modeling system, but only describes the interaction between these components.
There are several characteristics that are desirabe in an air quality modeling system. The modeling system should be modular. Modularlity means that components of the modeling system can be replaced by equivalent components without major redesigning of the overall modeling system or of the interfaces between the components. A modular modeling system will allow the substiution of a component such as a numerical integration scheme, gas-phase chemical mechanism, or aerosol thermodynamic model by a different but equivalent component. Thus, the sensitivity of the model predictions can be easily assessed.
Although the modeling system will be developed for the [San Joaquin Valley] domain, it should be applicable (i.e., transportable) to areas of similar regional scale. The [study area] domain includes many geographical features such as coastal areas, a large bay, mountain ranges, and urban and rural areas. Such features will need to be treated in the modeling system and should favor the design of a system that is portable to many other regions that have these features.
The modeling system should reflect advances in environmental science. New, more accurate model components should be integrated into the modeling system as they become available. The gas-phase chemical mechanism, aerosol thermodynamics, and aqueous-phase chemical mechanism should be updated as new kinetic, mechanistic, and thermodynamic data become available. The modeling system should also be transferable from the organization where it is developed to another organization equipped with adequate computational resources and staffed with qualified personnel. This transferability is important in allowing for quality control of the modeling system and promoting a wider range of applications. The model should be designed to take advantage of current (vectorization) and future (massively parallel processing) computing hardware.
A full version of the modeling system may be computationally demanding and may not be suitable for investigating a large number of scenarios. A version that has the major features of the system but includes some simplified model components should be considered for development. A simplified version would be computationally less expensive, but also less accurate. A simplified modeling system would be very useful as a screening technique when investigating various emission control scenarios. However, it would not replace the full version of the modeling system, since the limitation in model accuracy would likely prevent it from determining the effectiveness of specific emission control strategies.