The development of a procedure to forecast traffic volumes on urban segments of the state and interstate highway systems. (Volumes I and II)

Sunil Kumar Saha, Purdue University

Abstract

This study has developed a simplified traffic forecasting procedure for the highly traveled interstate network (both rural and urban), as well as for urban segments in the remainder of Indiana's state highway system. The procedure forecasts traffic volumes with alternative traffic models that incorporate the important socioeconomic and demographic variables, and the growth experienced by traffic alone. The effects of fundamental socioeconomic variables have been reflected in the volume forecasts from elasticity models. The forecasting models, especially lag-AADT models, reflect the effects of local conditions on traffic volume and its growth. The study combines statistical analysis with subjective judgment to develop models that are reliable and easy to use. After various statistical analyses and tests, the study identified two efficient models--elasticity and lag-AADT--to forecast traffic volumes on urban segments in the state and interstate highway system in Indiana. These models are developed using traffic data from coverage count stations, and data for various local and state level demographic and economic predictor variables. The key element in the elasticity model is the accuracy of input variable forecasts. Forecasts of input variables, entered in the elasticity models for which no reliable forecasts are available from outside sources, are generated in this study. Box-Jenkins and autoregressive time series procedures are used to forecast vehicle registration. The accuracy measures of these forecasts and the combination of forecasts have showed the high accuracy of these forecasts. The statistical analysis found that the predictor variables employed in the models are statistically significant; no other variables will provide significant additional predictive power to the models. The number of predictor variables employed in the models was kept to a minimum. An extensive evaluation of performance measures revealed the high quality of forecasts from both elasticity and lag-AADT models. Since there were no overwhelming reasons for preferring one method over the other, the traffic estimates recommended for design and planning purposes were simple averages of the lag-AADT and the elasticity forecasts. The research results, in general, showed that the accuracy of combined forecasts has outperformed the accuracy of forecasts from its constituent models.

Degree

Ph.D.

Advisors

Fricker, Purdue University.

Subject Area

Civil engineering|Transportation|Urban planning|Area planning & development

COinS