FIGURE 16.18 Concentration-time profiles for 03, NOx, NMHC, H202, HCHO, and higher aldehydes (RCHO) predicted using four different chemical submodels: two carbon bond four models (CB4.1 and CB4-TNO), a RADM model (RADM2), and the EMEP model (adapted from Kuhn et al, 1998).
Monitoring and .Evaluation .Program model; Simpson, 1995), predictions are 27% higher than the mean whereas the lowest, the CB4-TNO version of the carbon bond 4 mechanism, predicts ozone concentrations 35% below the mean. Other studies in which the carbon bond 4 mechanism was tested against environmental chamber data have also found that it underpredicts 03 formation (e.g., Simonaitis et al., 1991). The sensitivity of predicted 03 by CB4 to the chemistry, particularly radical-radical reactions, has been discussed by Kasibhatla et al. (1991).
It is common in model intercomparisons that relatively good agreement is obtained for the major species NOx and 03 but the discrepancies can be larger for trace species such as HCHO and H202. This can be important since such species are radical sources that contribute substantially to continuing the chain oxidation of the organics. For example, for the model intercomparisons shown in Fig. 16.18, the range of predicted values expressed as the root mean square (rms) error for 03 is f0%. However, the rms errors for H202, HCHO, and higher aldehydes are 22, 23, and 48%, respectively.
Similar results have been reported in an intercom-parison study of models used to predict tropospheric ozone on a global scale (Olson et al., 1997). Agreement for 03 and NO, was reasonably good for relatively clean atmospheres, with a larger spread for predicted H202. However, introduction of VOC chemistry increased the range of model predictions substantially, with the rms error for 03 doubling and that for NOx more than doubling, from 15 to 40%.
Given the numerous potential uncertainties due solely to uncertainties in the chemistry, particularly the VOC chemistry and how it is incorporated into models, it is clearly important to understand which are likely to have the most important net effects on predicted concentrations. A number of studies have been carried out to address this issue (e.g., see Hough and Reeves, 1988; Hough, 1988; Dodge, 1989, 1990; Chock et al., 1995; and Olson et al., 1997). Sensitivity analyses have been performed for a number of models. For typical approaches, see Derwent and Hov (1988), Milford et al. (1992), and Gao et al. (1995, 1996).
Once a chemical submodel has been developed, it must be tested extensively prior to its application in comprehensive computer models of an air basin or region. This is done by testing the chemical submodel predictions against the results of environmental chamber experiments. While agreement with the chamber experiments is necessary to have some confidence in the model, such agreement is not sufficient to confirm that the chemistry is indeed correct and applicable to real-world air masses. Some of the uncertainties include those introduced by condensing the organic reactions, uncertainties in kinetics and mechanisms of key reactions (e.g., of aromatics), and how to take into account chamber-specific effects such as the unknown radical source.
These chemical submodels are then incorporated into more comprehensive models of the type shown schematically in Fig. 16.17. These can vary from relatively simple box models to large-scale regional models, briefly described in the following text.
c. Simple Mathematical Models
(1) Box models (including EKMA) One type of simple model that has been applied to predict pollutant concentrations is known as the box model (Fig. 16.19) (e.g., Schere and Demerjian, 1978). The air mass over a region is treated as a box into which pollutants are emitted and undergo chemical reactions. Transport into and out of the box by meteorological processes and dilution is taken into account.
The box model is closely related to the more complex airshed models described below in that it is based on the conservation of mass equation and includes chemical submodels that represent the chemistry more accurately than many plume models, for example. However, it is less complex and hence requires less computation time. It has the additional advantage that it does not require the detailed emissions, meteorological, and air quality data needed for input and validation of the airshed models. However, the resulting predictions are
Variable mixing height
Was this article helpful?