A New Statistical Framework for Estimating Carbon Monoxide Impacts at Intersections

The computer program CAL3QHCR has been recommended by the U.S. Environmental Protection Agency (EPA) for modeling carbon monoxide (CO) concentrations at intersections. EPA's guidelines for modeling CO concentration ([CO]) levels at roadway intersections outline a procedure to identify intersections that should undergo a more detailed CO analysis by running CAL3QHCR, and this procedure uses intersection level-of-service (LOS) as one of its major defining factors. However, it is possible that intersections can exhibit the same intersection LOS but different levels of [CO], depending on factors such as intersection orientation, intersection geometry, total traffic volume, local meteorological condition (e.g. wind speed and wind direction), and emission factors. A new statistical framework for determining whether an intersection should be modeled for CO emission impact using CAL3QHCR and for estimating [CO] levels is presented for use at the intersection design level. The proposed statistical framework is based on not only the intersection LOS (as EPA's current criterion) but also on other major modeling factors, such as intersection orientation, intersection geometry, traffic volume, wind speed, wind direction, and vehicle emission factors, to predict [CO] levels. The proposed statistical model is much simpler than CAL3QHCR so that it can be used by traffic engineers at the intersection design level to approximate the [CO] level. Ideally then any potential exceedance could be mitigated at the design level. In addition, the new statistical model better represents the potential of CO exceedance than EPA's current LOS D criterion. The dependent variable, modeled [CO] level, used in this study is the output of the computer program CAL3QHCR rather than actual measured field [CO]. Thus, we are assuming that CAL3QHCR is a "perfect" model for estimating [CO] at intersections. In addition, a hypothetical typical urban traffic pattern rather than real traffic data was used in developing the statistical models. Therefore, the proposed models might not be applicable to areas that have a different traffic pattern from the one used in this study.