Rank Correlation Method for Evaluating Manual Pavement Distress Data Variability

Abstract

Evaluations of surface distresses in pavements, such as cracking, bleeding, and raveling, are often used as one component of overall pavement condition indexes. Both manual and automated survey methods are available for pavement distress evaluation; however, all distress evaluations experience a certain level of variability in their results. How this level of variability is determined depends on the type of data collected during the pavement distress evaluations. When distress data are collected as ordinal values, the variability may be determined by comparison of pairs of ranked values. This paper presents one rank correlation method, Kendall’s correlation coefficient, and illustrates how it can be used to assess variability in ordinal distress data collected through manual surveys. Using Kendall’s correlation coefficient, variability between different raters and variability between multiple evaluations by one rater were determined for each individual distress type. As a result, the ratings for certain distresses such as bleeding were found to have a high level of variability. This information can be useful when used to develop training programs to reduce data variability.