A novel approach to produce road-level inventories of on-road greenhouse gas and air pollutant emissions

Abstract:

Emissions inventories are an important tool often built by governments to
manage and assess greenhouse gases and other air pollutants. High resolution
inventories, both in space and time, are necessary to capture local
characteristics of on-road transportation emissions in particular. Emissions
vary widely due to the local nature of the fleet, fuel, and roads and this
heterogeneity must inform effective emissions modeling on the urban level. In
addition, widespread availability of low-cost computing now makes high
resolution climate and air quality modeling feasible, but efforts to improve
inventories have not kept pace. There currently is a lack of inventories at
comparable resolutions. This motivated similar work such as the VULCAN project
which used county-level data to estimate on-road emissions. We are motivated
to improve upon this by using site-level traffic count data where available.

Here we show a new high resolution model of CO$_2$ emissions for the Portland,
OR metropolitan region. The backbone is an archive of traffic counter
recordings taken by the Portland Bureau of Transportation intermittently at
9,352 sites over 21 years and continuing today (1986-2006 data are summarized
here) and by The Portland Regional Transportation Archive Listing at 309
freeway sites. We constructed a regression model to fill in traffic network
gaps using GIS data such as road class and population density. After stepwise
testing of each of eighteen road classes (from minor streets to freeway), we
were able to select ten variables that are significant (P < 0.001) predictors
of traffic; particularly freeway, unimproved road, and minor streets. The
model was tested by holding back one-third of the data. The R$^2$ for the linear
model (based on road class and land use) is 0.84. The EPA MOVES model was then
used to estimate transportation CO$_2$ emissions using local fleet, traffic, and
meteorology data.