Date of Award

December 2019

Degree Name

Master of Science (MS)


Civil and Environmental Engineering


Baris Salman


Data Resampling, Markov Chains, Pavement Condition Ratings, Pavement Deterioration, Prediction Modeling, Project Management

Subject Categories



Pavement deterioration prediction models play an important role in pavement management systems and allow decision makers to plan maintenance and rehabilitation works in advance. There are several factors affecting the deterioration of pavement sections such as traffic loading, freeze and thaw, snow plowing, construction quality as well as pavement thickness and age. Developing pavement deterioration prediction models can provide decision makers with many benefits. First of all, pavements can be treated before they reach undesired levels of service and this would extend the lifetime. Secondly, required budget for expenses can be provisioned, which helps public authorities to generate more precise cost estimates during budgeting. Finally, management philosophies can be better extended into long term planning, offering the potential to minimize life cycle costs.

This thesis describes the development of stochastic network-level pavement deterioration prediction models, which can assist local agencies (particularly the City of Syracuse) in making strategic investment decisions regarding maintenance, repair and rehabilitation activities. Based on the literature, various methods can be used in developing prediction models such as regression analysis, Markov chains and Artificial Neural Networks. In this research, Markov chains have been selected to develop the most convenient prediction models for Syracuse, NY for two reasons. First, the original dataset included pavement condition ratings; however, data on other important factors, such as pavement equivalent single axle load, pavement thickness and age, were missing. Markov chains-based models can be simply generated by making use of pavement condition ratings and the process of pavement deterioration. Second, Markov chains offer robust results for network-level prediction models.

Historical pavement condition ratings were incorporated into a Markovian model to develop a probabilistic pavement deterioration model. Separate models were developed for avenues, streets, roads and different pavement types. Pavement types were divided into three categories in this study depending on their sub-base structures and pavement thicknesses. Results were also validated with the bootstrap method, which is a common resampling technique to estimate statistics in a population such as bias, variance and confidence intervals.

Overall, this study demonstrated that Markov chains can be used in generating network level pavement deterioration prediction models in the absence of some of the key input variables. The methodology and findings of this research can help decision makers in generating network level maintenance, repair and rehabilitation plans for the City of Syracuse.


Open Access

Available for download on Saturday, January 09, 2021

Included in

Engineering Commons