Probabilistic Pavement Performance Prediction Models: A Systematic Meta-Review Towards a Next-Generation Conceptual Framework

Pavement infrastructure deteriorates progressively under the combined influence of multiple uncertainty-related factors making accurate deterioration prediction essential for effective Pavement Management Systems (PMSs). Probabilistic modelling approaches have received increasing attention for capturing uncertainty and variability of pavement deterioration and have been examined across several review studies. However, a systematic review that comprehensively investigates and evaluates the existing probabilistic pavement performance prediction models (PPPMs) has yet to be established. Accordingly, this study presents a systematic meta-review that synthesizes current knowledge, identifies key research gaps and proposes a conceptual framework for PPPM aligned with current research needs. Primarily, the analysis was based on a systematic literature review (SLR) of ten review papers identified in the Web of Science (WoS) between 2010 and 2026. The in-depth content analysis revealed five PPPM categories: Markov-chain, Bayesian approaches, fuzzy models, grey theory and logistic regression dominating the probabilistic PPPM field. The findings highlight a shift towards hybrid, multi-parameter, and duration-based models for effectively representing pavement deterioration over time. Thus, an adaptive conceptual framework of a Dynamic Bayesian Network (DBN) is proposed that integrates the dynamic interdependencies of multiple pavement performance indicators (PPIs), the dynamic updating of model parameters and the impact of extreme events on PPPM.