Sensitivity of urban seismic damage predictions to input data detail: an application to Sanremo, Italy

The accuracy of seismic damage scenarios is of paramount relevance for various objectives inherent to disaster risk management and particularly critical to support effective emergency response and to address seismic mitigation policies. When transitioning from large-scale studies – such as national ones – to urban scale applications, the assessment does not often reflect a downscaling in the detail level of exposure, vulnerability, and hazard data, leading to inaccurate evaluations of damage and loss, as well as of the related uncertainty. While the value of local, high-resolution data is widely acknowledged, its quantitative impact on final damage predictions remains poorly constrained. This study addresses this gap by quantifying the sensitivity of urban damage predictions to varying levels of detail in hazard, vulnerability, and exposure data. Specifically, the study compares estimates derived from large-scale datasets against those based on refined, local information acquired on-site. To this aim, a multi-level comparative framework is applied to the Sanremo Municipality (Northwestern Italy), simulating ground-motion scenarios consistent with the 1887 M6.3 Ligurian Sea earthquake. Within this framework, ground shaking is estimated using ground-motion prediction equations, which are amended to account for site-specific amplification effects. This critical step compares results derived from national soil classification maps against detailed seismic microzonation studies. Building damage and consequences are then assessed using fragility curves. Outcomes from vulnerability models based on standard aggregated census-level data are compared to those derived from refined inventories and field inspections. The results show substantial discrepancies between the predicted scenarios. The use of local data, particularly site-specific amplification effects and building characteristics, leads to significant differences in damage intensity and, especially, its spatial distribution. This study underscores the critical importance of improving knowledge through acquisition of local data and provides a robust general framework to improve decision-making for disaster risk management.