Assessment of Long-Term Pavement Performance Program Manual Distress Data Variability: Bias and Precision

Abstract

The use of manual survey methods within the Long-Term Pavement Performance (LTPP) program for the collection of distress data has drastically increased both in intensity and in coverage over the past couple of years. Because these surveys are conducted by individual raters whose biases can lead to variability between raters, it was hypothesized that distress data variability existed and that it could potentially be quite large. Thus, the purpose of the presented study was to quantify manual distress data variability, with special emphasis on the bias and precision of the data. Results from seven LTPP program distress rater accreditation workshops conducted during the period from 1992 to 1995 were used as the only source of data. On the basis of analyses of these data, both the apparent bias and the precision for the common distress type-severity level combinations were quantified. It was also concluded from this study that individual rater variability for any given distress type-severity level combination is typically large and increases as the distress quantity increases; however, when all distress type-severity level combinations are viewed in terms of a single composite number such as the pavement condition index value, there is excellent agreement between the individual raters, the group mean, and the ground truth value, and individual rater variability is also quite small. Because LTPP program distress data are to be used in the development of pavement performance prediction models, improvements in variability are highly desirable to ensure that they serve their intended purpose. Recognizing that the LTPP program distress raters are experienced individuals, such improvements are not envisioned to come through additional training. It is the authors’ contention that the only way of achieving the desired improvement is through the conduct of group consensus surveys.