Traffic volume counts are used by many Departments of Transportation (DOTs) in planning,
traffic operations, and asset management programs. Traffic counts are usually collected using
sensor-based monitoring tools at limited locations in a network. The sensor-based method
excludes low-class roads due to the cost involved. Therefore, in general, there is no volume data
count available for local roads; even though they make up the majority of our highway network.
An extensive literature review from this paper revealed that using traditional factor approach,
regression-based models, and artificial neural network models failed to present network-wide
traffic/truck volume estimation because they rely on traffic counts for model development and
they all have inherent weaknesses. Moreover, their traffic estimates have high estimation errors.
Traditionally, four-step model (FSM) is based on traffic analysis zones (TAZs) structure which
conveniently uses existing census geography to take advantage of socioeconomic data available
from Statistics Canada. However, the coarse zone structure used in such models tends to
exaggerate the intrazonal trips resulting in biased and unbalanced trip distribution over roadway
network and high estimation errors. Also, their purpose is to guide infrastructure development
and therefore, they are not appropriate as a tool for estimating traffic at network wide, including
low-class roads.
This paper develops a high-fidelity travel demand model (HFTDM) capable of achieving
network-wide traffic volume estimation with improved accuracy. This will require using all
functional class roadways and spatially disaggregating census-based coarse TAZ structure into
fine zones. A case study using an areal interpolation technique, which is based on fine-scale
grids, road density and a detailed road network was developed for the Beresford/Bathurst area in
the province of New Brunswick. Finally, a few conclusions and recommendations regarding this
paper are given.
