Characterizing Uncertainty in Transportation Air Quality Modeling
Research Idea Scope
Regulatory-mandated air quality analysis, such as SIP analysis, transportation conformity determinations, and project level evaluations typically require the use of multiple models (model
chain) to analyze a complex chain of land use trends, travel demand, vehicle activity characteristics, emission rates, dispersion characteristics, future projected background concentrations, and/or ambient concentrations of different pollutants. These types of analysis also require some professional judgments about uncertainty, since no formal uncertainty analysis currently exists.
While it is well-known that uncertainty exists in models, research has not assessed the amount of uncertainty that exists. A formal quantitative uncertainty analysis would help analysts understand how much confidence to place in the various steps in a required air quality analysis and support the preparation and review of these required evaluations by state DOT staff. This analysis should use a combination of comprehensive probability distributions and comprehensive sensitivity analysis to understand the propagating effect of land use and transportation model parameters on emissions and air quality modeling. This is especially critical for comparing when these estimates to the thresholds set by the EPA under the National Ambient Air Quality Standards (NAAQS) and other regulatory frameworks.
Define the specific models (travel demand models, traffic simulation models, emission factor models [for both regional and project applications], and dispersion models for regional or project
analysis). Suggest one or two selected travel demand and traffic simulation models, MOVES for emission factors, and CAL3QHCR and AERMOD for the dispersion component).
Identify the variables in each of the models along with the typical range of values used in the model.
Examine the uncertainty (accuracy and precision of the values where this may be known) of the outputs.
Document/illustrate the modeling and data process flow and present the results of uncertainty both within steps of the modeling chain and cumulatively along the modeling chain.
Urgency and Payoff
In addition to the known uncertainties of simulation models and scientific assessments, other factors compound the overall uncertainty in modeling results.
1.Many state DOTs lack sufficient funding, staff time, expertise, and data to analyze the target parameter(s), and criteria-specific uncertainties, in detail.
2.Engineering procedures and methodologies have historically been based on deterministic analyses.
3.The value of uncertainty analysis is not widely understood.
4.Inability to consider uncertainty in the existing regulatory framework. For example, the
transportation conformity regulations require a “bright line” analysis – the analysis result is modeled to be either above, at, or below acceptable emissions or air quality levels, and whether the project passes its test relies on the modeled point outcome. There is no place in the existing regulatory process to allow for a “passing” test result due to an uncertainty range, rather than a specific point estimate.
5. Mitigation at the project, program and regional level is based upon the same “bright line” analysis.
A quantitative assessment of uncertainty for various regulatory-mandated air quality analyses is proposed to put specific point estimates in the context of a comprehensive probability distribution and sensitivity analysis to better represent the uncertainty of the final numbers. Once the analysis is complete, the study results can be used to suggest refinements, as appropriate.
While uncertainty may be acceptable for comparative analysis such as between alternatives, this research could lead to better decisions being made in future legislation, rulemaking, regional transportation planning and project level NEPA actions. The research could potentially be used to reduce costs or streamline regional planning and/or project development.
The research will not resolve all questions or issues, or put an end to the uncertainty of the model projections, but it should enable users to define a “confidence interval” for the modeled results.
The propagation of uncertainty is a practical effort to define the uncertainty of each variable and the cumulative uncertainty associated with combinations of the variables and outputs from the modeling chain.
Jackie Ploch Texas Department of Transportation 512/416-2621