A method used to develop predictive models for estimating cumulative traffic volumes for Federal Highway Administration Long-Term Pavement Performance (LTPP) test sections is presented. The purpose of developing cumulative traffic prediction models for each LTPP site is so that distress and roughness development can be compared with cumulative traffic volumes for the various pavement types and conditions in the LTPP study. There are two types of LTPP traffic data: historical estimates and monitoring data. Predictive models from this study are based on the combined monitoring and historical LTPP traffic data available in January 2000 and designated Level E data. The models attempt to predict the cumulative kilo-equivalent single-axle loads for the LTPP test sections corresponding to any profiling date or distress survey date. The modeling method is described and consists of fitting an exponential growth curve to each site's data. As a result of the high variability of the LTPP traffic data and some apparent errors in the data, considerable judgment in the form of deleting and changing the reported LTPP data was required before the modeling process. Examples of reported versus backpredicted traffic curves are presented for various traffic conditions encountered. Some differences in historical and monitoring traffic data are shown. Some trends in the traffic data that affect the fitting of exponential growth-type predictive models to the data are described. Methods used to address these trends during model development are discussed.
Abstract