Uncertainty And Sensitivity Analysis of the iRAP Road Safety Risk Assessment Model for Vehicles

Roads are among the most vital infrastructures developed by humans, serving as the backbone of terrestrial transportation systems worldwide. Given their critical role, systematic evaluation of road hazards and associated risks has become a priority for governments and agencies that invest substantial resources in monitoring and improving road safety. To quantify these risks, the International Road Assessment Programme (iRAP) has proposed a standardized methodology and a comprehensive road safety risk score formula that incorporates numerous roadside and roadway attributes, such as geometry, roadside hazards, delineation quality, and protection measures. In this study, we apply a Monte Carlo simulation framework to the iRAP vehicle risk score formula to explore the uncertainty propagation and parametric influence of input variables. For the main analysis, we generated 32,768 simulated scenarios using a Sobol quasi-Monte Carlo design, covering the full spectrum of possible combinations of categorical and continuous attributes in a statistically representative manner. The sampling strategy was designed to ensure adequate coverage of both common and extreme road conditions, thereby capturing the realistic variability of field data. The resulting distribution of road risk scores was then analyzed to quantify expected ranges and probabilistic behavior under uncertain input conditions. Subsequently, sensitivity analysis was performed to identify the most influential parameters governing the final safety score. Both correlation-based (Spearman and Pearson) and variance-based (Sobol indices) techniques were employed to evaluate the relative contribution of each factor. The findings reveal the parameters that most critically affect the iRAP score, providing essential guidance for targeted interventions. Furthermore, the outcomes can be integrated into automated and AI-driven road assessment systems, enhancing their interpretability by quantifying how sensitive the computed risk score is to variations in specific attributes.

Keywords: Road safety risk assessment, Sensitivity analysis, Simulation-based analysis, Probabilistic modeling, Decision.