scope:

CONTEXT OF THIS STUDY

This document describes studies conducted to evaluate the performance of selected air quality models that are used to simulate the atmospheric behavior of pollutants emitted from industrial area and volume sources. This assessment of model performance follows completion of the first phase of a comprehensive area/volume source model evaluation study being funded by the American Petroleum Institute (MI) and the Chemical Manufacturers Association (CMA). The study is being conducted by Ogden Environmental and Energy Services Company (Ogden). Work conducted during Phase I included:

• identification of available area/volume source models;

• evaluation of the reasonableness of the models' mathematical representations of physical processes; and

• testing of the models' responses to variations in key input parameters.

This model performance evaluation report constitutes Phase II of the overall study. Development of recommendations for the design of future field measurements to provide data for a more complete model evaluation study will be the subject of the third and final phase of this effort.

BACKGROUND

Phase I showed that, although several types of area/volume source models are in widespread and increasing use, there have been only very limited studies that provide quantitative information on their accuracy and reliability. It was also shown that the available models utilize widely differing dispersion algorithms, and that little progress has been made in determining which approaches produce the most realistic concentration estimates. A series of tests was conducted in Phase I with six of the currently employed models to: (1) evaluate the response of the alternate models to variations in key input parameters; (2) compare the relative magnitudes of the predicted concentrations for different models with the same input conditions; and (3) provide the basis for at least an intuitive evaluation of the models' reasonableness in terms of conformance with established physical principles. These tests have led to the selection of certain models to be tested further in Phase II, using the available data from appropriate field and wind tunnel tracer experiments.

Upon completion of the Phase I analyses, a detailed model performance evaluation protocol was developed for Phase II. This protocol specified the objectives, models, data sets and analysis procedures to be incorporated in this study phase. Several iterations of the protocol were prepared in close coordination with the API Air Modeling Task Force, whose members contributed significantly to its refinement. The final version of the protocol is reflected in Sections 2 through 5 of this report. In addition, several of the field data sets used in this study were made available by API Task Force members, who had direct personal knowledge of the objectives, procedures and data sets generated by these experiments. The involvement of these individuals in the development of the Phase II study design provided important insight into the appropriate uses of the resulting data for this model performance evaluation.

The models that have been retained for Phase II and the rationale for their selection are discussed in Section 2. Section 3 identifies field and laboratory studies that have produced data that are at least potentially useful for evaluating the performance of area/volume source models, and establishes criteria for selection of data sets for this purpose. Next, the data sets generated by these studies are examined in detail to determine: (1) their general appropriateness for area/volume source model testing; and (2) whether the experimental data are reasonably complete in terms of providing values for all important model input parameters, as well as measured concentrations for comparison with model predictions. The success of this effort was limited, in that none of the identified studies resulted in data sets that meet all the criteria for a rigorous model performance evaluation. In fact, most of the studies that have produced potentially usable information were actually designed for purposes other than dispersion model development or testing. Based on the results of this data search, it was decided that a limited model performance evaluation would be conducted, using the relatively small data sets developed from six historical tracer experiments. In addition to providing useful information on the models tested, this effort will help determine the data needs for a more comprehensive model performance evaluation.

The selected data sets and the manner in which they are used in the model performance evaluation are described in detail in Section 4. The statistical measures that were employed in evaluating each model's performance individually and in comparing the different models are the methods developed and utilized by Sigma Research Corporation under contract to API. The specific procedures that have been used to analyze the results of model simulations versus measured concentrations for the Phase II model performance evaluation studies are discussed in Section 5. The results obtained from these analyses are presented in Section 6. Conclusions and recommendations are given in Section 7.