This paper explains how parametric methods were used to model pavement deterioration on a network level. Pavement condition is rated on a visual scale from’1’ to ‘10’ (‘10’ being new). The data file consists of yearly pavement condition ratings from 1981 to the year 2000 for approximately 23,000 individual pavement sections. Right data censoring was applied where pavement improvements occurred and for the last data entry for each pavement section (Year 2000).

Duration distributins for each pavement condition level are right-skewed. To determine an appropriate distribution for the data, probability plots were prepared. Anderson-Darling goodness-of-fit statistics and Pearson Correlation Coefficients were then compared to determine the best-fitting distribution. Based upon these observations, the Weibill distribution was selected for the reliability analysis.

Next, parametric distribution analyses were performed to determine if significant differences exist between factor levels for factors identified to have a potential effect on pavement performance (e.g. traffic, climate). Chi-square testing was used to assess equality between Weibull shape and scale parameters to determine if significant differences existed.

Lastly, pavement deterioration curves were developed using Wiebull estimated mean durations at each condition level. Regression analyses were performed and quadratic formulas were developed representing each deterioration curve with R-squared values greater than 97 percent.