The development of an annual average daily traffic estimation model for county roads

Dadang Mohamad, Purdue University

Abstract

Most counties and local transportation agencies do not have the facilities and resources necessary to have any traffic monitoring programs, particularly for low volume roads. However, estimates of traffic volumes and their composition are important for planning, design, and operation of these roads, as well as for the allocation of highway funds. This study has developed a traffic estimation procedure for county roads that can be implemented with limited resources. The procedure estimates average daily traffic volumes with alternative traffic prediction models that incorporate relevant socioeconomic and demographic variables. Field traffic data were collected from 40 out of 92 counties in Indiana. A statistical design of experiment was developed for grouping and selecting these counties. The selection of a county was based on population, state highway mileage, per capita income, and the presence of interstate highways. To select road sections within those counties, additional criteria--surface type (paved and gravel) and location (urban and rural)--were used. The study combined randomness and subjective judgments to select road sections for the installation of automatic traffic counters. Four automatic traffic counters were used in each selected county. Two traffic counters were installed on county paved roads and two counters on county gravel roads. Two types of counters, a traditional rubber-tube traffic counter which counts only traffic volume and a vehicle magnetic imaging traffic counter which counts not only traffic volume but also truck percentage, were used in the present study. Most counters installed on the selected road sections were based on the standard 48-hour traffic counts. Then, the obtained average daily traffic (ADT) was converted to annual average daily traffic (AADT) by means of adjustment (monthly variation) factors. Traffic data obtained from the counters were analyzed before the development of the AADT prediction models. Multiple regression analyses were conducted to develop the AADT prediction models, which are reliable and easy to use. Two different AADT prediction models were developed for county roads, one for paved road and the other for gravel road. There were quantitative and qualitative predictor variables used in the model development. Because too many predictor variables were involved in the model development, a principal component analysis was also conducted. To validate the developed models, additional field traffic data were collected from eight randomly selected counties. The accuracy measures of the validation showed the high accuracy of the prediction models. After various statistical analyses and tests, the study identified that models without reasonable outliers gave a better estimate. The statistical analysis also found that the independent variables employed in the models were statistically significant. The number of independent variables included in the models was kept to a minimum.

Degree

Ph.D.

Advisors

Sinha, Purdue University.

Subject Area

Transportation|Civil engineering|Statistics|Public administration

COinS