Integration of Heuristics and Statistics to Improve the Quality of Network-Level Pavement Condition Data

Transportation agencies use pavement management systems (PMSs) to make efficient decisions about allocating available resources to the maintenance, rehabilitation, and renewal of their roadway networks. One of the most costly parts of the PMS process is collecting pavement condition data. The efficiency and reliability of decisions made based on PMSs depend upon the quality of this data. Thus, transportation agencies need to ensure that dollars invested in this data are well spent, and pavement condition data has the level of quality necessary to meet PMS requirements. Therefore, assessing and improving the quality of pavement management data is a major challenge for both researchers and practitioners. This study advances the quality assessment of network-level pavement condition data by answering the following questions: (a) How can we identify potential errors in pavement condition data used in PMSs? (b) How do multiple dimensions of error detection affect our ability to detect errors? (c) How does the accuracy of pavement condition data impact predictions of future road network performance? And (d) How do we measure multiple quality dimensions of pavement condition datasets? First, this research devises and implements a computational method to identify potential errors in pavement condition data, integrating conventional statistical methods and heuristics. Second, the effect of considering multiple dimensions of error detection in pavement condition data was investigated. These dimensions are based on data properties, including time series trends in pavement condition data, variability within uniform performance families, and the consistency between several performance indicators. Third, this research presents a quantitative assessment of the impact of data accuracy on the estimated remaining service life (RSL) of a roadway network as an overall measure of network health. Finally, it provides metrics for measuring data quality dimensions for pavement condition datasets. The developed technique was validated using pavement condition field data for a road network in Texas. The technique has the advantage of differentiating between extreme yet valid data points and potential errors. In addition, accounting for several properties of pavement condition data to identify potential errors improves the results of this technique. It is hoped that this research will enable pavement engineers to identify potential errors in pavement condition data, and more effectively assure data quality