Calibration Helps Reduce Disagreement Between Different Pavement Condition Indices

To make expert estimates of pavement condition more accurate, the American Society for Testing and Materials (ASTM) split one of the original pavement distress categories, for which experts previously provided a single numerical estimate, into two subcategories to be estimated separately. While this split has indeed made expert estimates more accurate, there is a problem: to get a good understanding of the road quality, we would like to see how this quality changed over time, and it is not easy to compare past estimates (based on the old methodology) with the new estimates, which are based on the new after-split methodology. In this paper, we show that a linear calibration reduced disagreement between these two types of estimates – and thus, leads to a more adequate picture of how the road quality changes with time. 1 Formulation of the Problem How pavement condition is evaluated: general idea. To evaluate the pavement condition of a given road segments, experts evaluate several different characteristics of a pavement. These estimates e1, e2, . . . , are then combined into a linear combination a0 + a1 · e1 + a2 · e2 + . . . with appropriate weights ai. This linear combination is known as a Pavement Condition Index (PCI). The weights are selected in such a way that the PCI can take any value from 0 to 100. The more distresses, the lower the PCI.