Multinomial Prediction Intervals for Micro-Scale Highway Emissions

Abstract

Listen

Translate article

Legislation passed in the early 1990’s requires that the air quality impacts of individual transportation projects be evaluated. Most of the current modeling practices to predict these air quality impacts use regionally-based average pollutant emissions rates without acknowledging that local traffic conditions and vehicle fleet compositions can vary widely. Yet it is known that some of the major health problems resulting from pollutants are immediate and localized, such as asthmatic reactions. There is a need for prediction methods that utilize small time-period variability in traffic volumes, combined with localized measures of emissions rates, to predict localized pollutant levels. Technology has been developed for measuring micro-scale levels of certain pollutants, but the simultaneous collection of micro-scale (e.g. 5-minute) traffic counts is expensive and impractical. In contrast, automated hourly count volumes are ubiquitous and available for most roadways. In this paper, we present a method for constructing prediction intervals for localized pollutant levels when only the total traffic volume count is known. The method utilizes micro-scale traffic volume counts and emissions factors previously collected on comparable roadways. To demonstrate how the prediction methods can be applied, we utilize micro-scale emissions data collected as part of a University of California, Davis experiment to predict carbon monoxide (CO) concentrations during worst-case meteorological conditions.