Extreme value theory analysis of high emitter trends across four US cities from 1995 to 2021.

Extreme value theory provides a direct means to characterize the distribution of high emitters within a vehicle fleet and calculate statistical confidence intervals for comparisons. Defining a "high emitter" as the maximum emitter in a random sample of N vehicles implies in the limit of large N that high emitters follow an extreme value distribution, comprised of three distinct domains. The analysis of over twenty years of roadside remote sensing emissions measurements in Chicago, Denver, Los Angeles and Tulsa reveals clear differences between gasoline vehicle high emitter distributions across pollutants (hydrocarbons (HC), carbon monoxide (CO) and nitric oxide (NO)), but very similar behavior across the four cities. As Federal emissions regulations became more stringent, HC and CO high emitters evolved from distributions bounded by the stoichiometric limits of HC and CO formation in an engine to progressively longer tailed distributions. In contrast, NO high emitter distributions have changed little over this time period. These distinct behaviors likely reflect differences in the failure mechanisms leading to high HC, CO, versus NO emissions. Whereas average HC, CO and NO fleet emissions fell dramatically over these two plus decades by factors of about 4, 6 and 7, respectively, mean high emitter emissions declined by less than half of these. The paper examines in detail similarities and differences in high emitter distributions versus pollutant, city and vehicle type, how these changed over a period of roughly twenty years and the policy implications that can be drawn.