Comparison of Modeling and Measurement Accuracy of Road Condition Data

Abstract

The condition of a road network and its deterioration rate can be estimated by using measurements and statistical models. The purpose of this paper is to provide tools for assessing the accuracy of the condition information based on measured and modeled values. In this study, International Roughness Index (IRI) measurements over 3 years (2000–2002) were used. A logarithmic transformation was applied to the measured IRI values. One half of the data set was used for developing regression models that predict road roughness. These models were validated using the other half of the data set. The comparison of the residual distribution in the logarithmic regression model and measurement accuracy in logarithmic terms facilitates direct consideration of the relative accuracies. The Taguchi loss function was applied to estimating the losses incurred when measured and modeled values were used. The decision of taking new measurements depends on the relative accuracies of the measurement and the modeling, the cost of measurement, and the losses incurred to the road users and the maintaining agency due to untimely maintenance.