摘要: |
In 1997 the Canadian Long-Term Pavement Performance (C-LTPP) project capitalized on an opportunity to have surface distress surveys completed twice at more than half of its test sections to provide some insight as to the variability between different raters. The major findings of that investigation are presented. Because only two distress surveys for each test section were available for comparison, traditional variance analysis techniques were not applicable. Instead, a technique involving an "agreement index," also known as "Cohen's weighted kappa statistic," was used to directly compare the levels of agreement between the surveys. The kappa statistic considers the likelihood of chance agreement and also allows the introduction of penalty values for individual cases of disagreement. The analysis results indicated that the level of agreement based on crack type was very high; less agreement was observed for severity level. Agreement increased significantly with increased data aggregation, indicating that less variability will be present for network-level analyses than for project-level analyses. The effect of rater experience was also investigated, although no firm conclusions could be made with the available data. Recommended methods for reduction of the variability of future distress surveys included reduction of the number of severity levels from five to three, reduction of the number of individual agency raters, more frequent training of raters, and the use of rating teams. In general, the results of the C-LTPP project distress variability analysis were in agreement with previous studies, including the recent study completed by the U.S. LTPP program. |