Empirical fragility and ROC curves for masonry buildings subjected to settlements

In the Netherlands, the potential damage to the building stock due to subsidence phenomena has recently received increased awareness. However, evaluating and predicting damage to buildings in subsiding areas is a complex task that requires associating the vulnerability of exposed structures with the intensity of the subsidence hazard. Considering the widespread presence of subsidence-related damage to the built heritage, the focus of this study is to provide empiricalbased insights to assess and forecast subsidence damage to masonry buildings. A rich dataset with manual levelling measurements was collected comprising 386 surveyed masonry buildings, mainly low-rise (terraced) houses built before 1950. Of the total set of buildings, 122 cases rest on shallow foundations and 264 on piled foundations. For each building, the recorded damage is related to the settlement, calculated from the bed-joint levelling measurements, using four different intensity parameters, namely differential settlement, rotation, relative rotation and deflection ratio. These four parameters are appraised in their capacity to effectively predict the intensity of the damage. The Receiver Operating Characteristic (ROC) method is used to evaluate the relative efficacy of the selected hazard parameters. The rotation, the relative rotation (angular distortion) and the deflection ratio are observed as the most accurate when predicting the intensity of damage, while the differential settlement appears less accurate. Additionally, the dataset was used to generate empirical fragility curves where the probability of damage is described as a function of the aforementioned parameters. Thresholds were set to distinguish between the light damage and the functional and structural damage state. At a relative rotation of 1/500 masonry buildings on shallow foundations were observed to reach or exceed light damage with a probability of 13%, and functional and structural damage with 5%. The availability of the bed joint levelling measurements made it possible to classify eight recurrent settlement profiles, including both symmetric and asymmetric profiles, associated with both the overall deformation and the rigid rotations of the surveyed buildings.