2019 CSCE Annual Conference - Laval (Greater Montreal) Conference
Dr. Tamer El-Diraby, University of Toronto
Understanding the condition of roads is important to transportation departments. Therefore, modeling and predicting the deterioration of roads is crucial to road asset management. This paper provides a practical decision-support tool for predicting the condition of asphalt roads in the short term under a changing climate. Users have the option of running a predictive model under different values of climate stressors. The prediction of deterioration is done via machine learning. More than a thousand examples of road sections from the Long-Term Pavement Performance (LTPP) database were used in the process of model training. The results are implemented in a web-based platform, which includes a map with an interactive dashboard. Users can query any road, input its data, and get relevant predictions about its deterioration in two, three, five and six years. Since the attributes used for model training include important climatic features, such as temperature and precipitation, the users can run the model multiple times under different climate scenarios and get relevant analytics. This type of analysis is highly important knowing that climate change will be a significant factor in future infrastructure decision making. The results of this study could be useful to municipalities, policy-makers and transportation agencies who deal with large networks of assets under a varying climate.