Development of an AI-Based Predictive Model for Flexible Pavement and Its Response to Climate Change Arid and Semi-Arid Regions

Abstract

Climate change poses significant risks to pavement performance, especially in arid and semi-arid regions where extreme temperatures, variable rainfall, and heavy traffic loads accelerate deterioration. Traditional models, such as the Long-Term Pavement Performance (LTPP) framework, often rely on U.S.-centric datasets and mechanistic-empirical assumptions that may not accurately represent local conditions in regions like Libya. To address this gap, we developed the Pavement Life Prediction Tool (PLPT), an AI-based decision-support system designed to classify flexible pavement failure risk levels (low, medium, high) using climatic and structural variables, including road age, rainfall, traffic volume, soil type, asphalt mix, and maximum temperature. The model applies supervised machine learning techniques specifically, a Decision Tree classifier and Logistic Regression to predict the decline in Pavement Condition Index (PCI) over a 5–30-year horizon without maintenance. Results show consistent classification performance across both algorithms, with soil properties, traffic volume, and maximum temperature emerging as dominant predictors of pavement life. These findings align with recent studies that highlight the reliability of AI approaches for pavement condition modeling and the importance of climate-sensitive adaptation strategies in infrastructure planning. By localizing predictive modeling to Libya’s climatic context, PLPT offers policymakers and transport authorities a practical tool to anticipate deterioration risks, optimize maintenance schedules, and enhance the resilience of road networks under climate variability.