Long-term prediction of the performance and durability of pavement represents a critical and vital issue in the pavement surface type selection process by the Kansas Department of Transportation (KDOT) using the life-cycle-cost analysis. Accurate prediction of roughness progression on Portland Cement Concrete (PCC) pavements is very important since the current model used by KDOT is based on the pavement serviceability guidelines (1993 AASHTO Design Guide). In this study, dynamic Artificial Neural Network (ANN) and statistical analysis approaches were used to develop reliable and accurate time-dependent roughness (International Roughness Index, IRI) prediction models for the newly constructed Kansas Jointed Plain Concrete Pavements (JPCP). To achieve this objective, data used in the model development process include construction and materials data as well as other inventory items, such as, traffic and climatic related data, which reflect the section-specific local conditions in Kansas.