goal or standard (NARSTO 2000). For that reason, observation-based models provide a useful adjunct to emissions-based models in an attainment demonstration but are not a substitution for them.
Scientists have had over three decades of experience in developing and applying a number of models for air quality management, and in that time, a number of insights into what is working well and what is not working well have been gained.
The dynamic partnership existing between the technical and regulatory communities promotes continuous improvement and advancement in air quality modeling and management. On the one hand, the evolving needs of the regulatory community promoted and catalyzed the scientific and engineering communities to develop increasingly sophisticated air quality models; on the other hand, insights gained from advanced models have prompted members of the regulatory community to rethink their approaches to air quality management.
As a general rule, models should be subjected to comprehensive performance evaluations using detailed data sets from the atmosphere before using them in regulatory applications (Roth 1999; Seigneur et al. 2000). Unfortunately, because of the lack of adequate atmospheric data (such as those on pollutant concentrations), that rule is all too often overlooked by the regulatory community.8 The consequences of this practice can be and, in some instances, have been the promulgation of less than optimal or even ineffective control strategies (see Box 3-4).
Model results are sometimes used inappropriately by regulators. One example is the practice in attainment demonstrations of using a model to predict pollutant concentrations years or even decades into the future without recognizing that input parameters, which are based on estimates of future socioeconomic factors, for the simulation are highly uncertain.
Substantial delays occurred in incorporating new scientific insights from models into policy design. One example is the reliance on VOC