The effect of spatial variability on the reliability of pavements

Spatial variability—the tendency of closely spaced data points to be better correlated than then those further apart—has largely been ignored in pavement engineering, although it is observed in nearly all contexts, from layer thicknesses to ride quality. This dissertation looks at three areas that have not received much research attention: the use of spatial statistics in characterizing the properties of pavement sections, the impact of spatial correlation on how variability must be handled in pavement design methods, and the effect of spatial variability on pavement response. It was found that spatial statistics could be used to characterize pavement variability, although many different processes, with different spatial scales, contribute to the overall variability, which complicates modeling. Further work should be performed to better understand the nature and causes of spatial variability. A general framework for handling variability in pavement performance is outlined, detailing various sources and types of variability in pavements, so that one can categorize the variability within a set of observed data, and determine what factors should be included in the prediction of variability, depending on the time and application of the prediction models. Major findings of this framework are that the language and definitions of structural reliability which have been adopted by pavement engineers, are not generally appropriate, and that ‘design method reliability,’ as implemented in current design methods may not be desirable. An area of concern is whether the effects of spatial variability run deeper than correlation between performance at various points. If the performance at one point is dependent on that in the neighborhood of point, then pavement design methods must, deal with the performance of all points simultaneously. However, simulations with realistic spatial variability show that there is an insignificant difference between the responses of these pavements and an idealized model where the surrounding pavement has the same structural properties as the point under consideration. Until the pavement shows signs of distress, various transverse cross-sections of the pavement can be analyzed independently, meaning that using Monte Carlo simulation to model variability in Mechanistic-Empirical design using layered elastic theory is feasible