Collecting Driving Data to Support Mobile Source Emissions Estimation

This paper describes a new sampling approach to collect driving data. Traditional data collection assesses the percentage of vehicle miles traveled by road facility and time of day, and collects data in proportion to how real-world driving occurs. While the traditional approach is reasonable, it would be better to establish statistical targets. Statistical targets minimize uncertainties about how well data represent real-world driving and allow confidence levels to be computed. We illustrate statistical sampling with a California study that collected over 180h of driving data. Road load power (RLP) was estimated based on speeds and accelerations and used as a surrogate for emissions-producing behavior. The variance of RLP (RPV) was used to establish statistical targets. The new method facilitated fine-tuned data collection by facility and time of day. The research team estimated, at a 90% confidence interval, that mean RVP was within ±11 to ±23% of the true mean among facilities studied. The method ensured that key facilities were adequately represented, thus providing a data resource to build facility-specific emissions tools.