Implications of Uncertainty in Distress Measurement for Calibration of Mechanistic–Empirical Performance Models

Abstract

Discrepancies between mechanistic–empirical (M-E) model predictions and observed field performance are conventionally attributed solely to errors in the predicted pavement distresses. However, significant uncertainty may be inherent in the measured pavement distresses due to spatial variability, sampling errors, and measurement error. A statistical approach is described for correcting field calibration for the influence of measurement uncertainty. This statistical formulation quantifies the hidden but systematic downward bias in the distresses predicted by the “calibrated” model because of uncertainty in the distress measurements. Perhaps more important, the formulation enables quantitative evaluation of the true predictive accuracy of distress prediction models by correcting for the confounding influence of uncertainty in distress measurement. Application of the methodology is illustrated through flexible pavement rutting and rigid pavement faulting data from the calibration of the M-E pavement distress models originally developed in NCHRP Project 1-37A.