| The network-level Pavement Management System (NPMS) is responsible for the planning and allocation of agency resources. It is at the network level where needs estimation and the development of multi-year programs greatly depend on accurate performance models. These models are dominantly empirical and often utilized without regard for uncertainty. Some local models have been shown to provide an R2 as low as 14%. Without demonstrating reasonable accuracy, these models cannot be considered as a rational means of supporting decisions in the NPMS.;This research fulfills a timely need to transition the state of practice in network-level pavement management to mechanistic modeling while quantifying prediction uncertainties. Researchers and practitioners have long recognized the advantages of mechanistic modeling based on fundamental engineering properties, yet empirical methods remain dominant at the network level.;The Nebraska PMS serves as a model case for implementation. As with any PMS, visual distress surveys are typically combined into an index to provide an overall measure of performance. Decision makers utilize these familiar indices in a number of facets. In project-level design, all U.S. highway agencies are in the process of implementing the Mechanistic-Empirical Pavement Design Guide (MEPDG). Therefore, as states move forward with the MEPDG implementation, maintaining the role these local indices play is critical for upholding the system's continuity. However, utilizing the MEPDG output directly in calculating performance indices becomes problematic, as local distresses and MEPDG distresses are not always congruent.;Procedures were developed to calculate local performance indices of flexible pavements from locally calibrated MEPDG distress output. When implemented in a 5-year needs assessment of 86 sections, M-E models yielded a 70% improvement in accuracy over the needs estimated with current linear models. In addition, a reliability concept was implemented to address the as-built construction variability as well as the primary sources of model error. This approach affords engineers the ability to support recommended multi-year programs with measured risk and state-of-the-practice theory. |
