Confidence Level and Efficient Sampling Size of Roughness Measurements for Pavement Management in Maryland

Abstract

The international roughness index (IRI) is the primary index of the Maryland pavement management system and is the data source for performance trend analysis, budget allocation, and project selection. The Maryland State Highway Administration (MDSHA) makes a significant investment to collect, process, analyze, and store IRI data every year over the entire network of the roads. Such reliance on IRI data and investment instigated the investigation of the confidence level embedded in the data and of devising possible sampling scenarios. To evaluate the confidence level, repeatability errors of the measurements were assessed. IRI data were repeatedly collected over a designated test loop under normal operating conditions to mimic network-level data collection. Sampling scenarios were devised using the Monte Carlo method for network IRI data collected in 2008. The network was stratified on the basis of functional classifications. Sampling errors in estimating the percentage of the roads in each of five IRI categories established by MDSHA were evaluated. The results indicated that the repeatability error of the measurements was less than 7%. The sampling results indicated that the state may survey a third of the network each year and expect estimation errors of less than 0.5% for all IRI categories. As such, MDSHA would not incur a considerable sacrifice in the confidence of network condition evaluation.