Evaluating climate change-related risks on road surface deterioration under various SSP scenarios using XG-boost and LSTM : the case of cambodia

Abstract

Climate change has become a pressing issue for public agencies responsible for the development and maintenance of road infrastructure, especially in tropical regions, where increasing temperatures and fluctuating amounts of annual rainfall could exacerbate the deterioration process and, accordingly, shorten the life span and increase the road maintenance costs. However, in Cambodia which climate-related risk factors pose the greatest threat to the surface condition remain unknown. Therefore, this study develops an interpretable climate-informed pavement deterioration framework for data-scarce tropical road networks by integrating agency-based IRI, pavement age, and traffic records, CMIP6-based SSP climate projections, machine learning techniques (i.e. LSTM and XG-boost), traditional techniques (i.e. MMH) and SHAP-based interpretation. The framework is demonstrated on Cambodia’s National Road 6, in the period of 2025-2054, under the three SSPs (i.g. SSP1-2.6, SSP2-4.5, and SSP5-8.5) in the period of 2025-2054, to quantify how future temperature and precipitation changes may alter IRI progression and pavement life expectancy under maintenance and no-maintenance policies.