Accounting for biases in the analysis of building damage data for the 2023 M7.8 Türkiye/Syria earthquake sequence
In the aftermath of a large earthquake, several practical constraints affect the systematic collection of data for assessing building damage. The available data are therefore typically incomplete, frequently missing key information such as the type of construction, age and geographic location. Moreover, the observations often favour certain locations and damage grades. In the absence of robust methods, using such data directly for characterising the building stock or conducting fragility assessments can introduce systematic bias. Using the 2023 M7.8 Türkiye/Syria earthquake sequence as a case study, we propose a statistical model that enables joint inference over the inventory and fragility of the building stock while explicitly accounting for sources of data error and missing information. The results show higher-than-expected vulnerability amongst several building types, particularly mid- and high-rise reinforced concrete buildings, which performed extremely poorly during the 2023 earthquake sequence. The proposed Bayesian framework provides rigorous uncertainty quantification that can be propagated to future applications.