Massey, C.I.; McColl, S.T.; Lukovic, B.; Sadashiva, V.K.; Kaiser, A.E. 2023 Development of earthquake shaking vulnerability functions for residential buildings located on slopes in New Zealand. Lower Hutt, NZ: GNS Science. GNS Science report 2023/40. 72 p.; doi: 10.21420/AJ0K-RX88
Abstract
The main objective of the research presented in this report is to explore how hillslopes in Christchurch influenced the damage to low-rise (mainly 1- and 2-storey) residential buildings located on hillslopes, from earthquake shaking only. The main outputs from this research are new vulnerability functions, and their associated uncertainties, for low-rise residential single-dwelling buildings on hillslopes. To generate the vulnerability functions, we have used a ‘deep dive’ of building damage claim information and repair and replacement cost estimates collated by Toka Tū Ake EQC from the 2010/11 Canterbury Earthquake Sequence (CES) events, mainly, the 4 September 2010 and 22 February 2011 earthquakes. The main outcome we hope to achieve from this research is that end users (e.g. re-insurers) have greater confidence in the vulnerability functions that they use to estimate loss, i.e. the models better represent the impacts. Only residential low-rise single-dwelling buildings within the Port Hills of Christchurch were selected for use in this research. The hazard intensity data used comes from published earthquake-induced ground shaking models developed for the four main earthquakes within the 2010/11 CES. These ground motions were then modified to capture topographic and material impedance (soil overlying rock) effects that might amplify the ground shaking, using published regional-scale amplification models, which were compiled and evaluated as part of this research. The variables used to predict the damage ratio ‘predictor variables’ in this assessment were chosen based on the variables previously found to influence earthquake shaking on slopes and the damage to buildings exposed to earthquake shaking. These variables relate to (1) earthquake shaking and amplification, (2) slope geology and morphology and (3) the building characteristics. Overall, our results show that the Peak Ground Accelerations (PGAs) experienced by buildings located on hillslopes during the 4 September 2010 and 22 February 2011 earthquakes were likely to have been higher than those simulated using models that do not account for topographic amplification. A combination of topography (steep and locally high slopes) coupled with material impedance contrasts (between soil overlying rock) influenced amplification of shaking at the surface and thus the recorded building damage. The vulnerability functions derived from this research indicate a positive trend between the damage ratios (cost to repair divided by the cost to replace) and earthquake PGA but a negative trend between the damage ratios and the building floor area and building plan area. In other words, in the 2010/11 CES, the damage ratios increased, with increasing PGA and decreasing building floor area. The latter suggests that smaller buildings cost relatively more to repair as a proportion of their replacement cost than larger ones. By adopting the amplified PGAs used in this research, the damage ratios (DR) associated with a given PGA reduce overall. For example, if we take a building on a slope with DR = 0.5, the non-amplified PGAat the building might be 0.3 g, but the amplified PGA might be 0.45 g, indicating that the damage ratio is more likely to be associated with a higher PGA. This suggests that,if vulnerability functions trained on non-amplified (free field) PGAs were used to forecast the damage ratios for buildings in the Port Hills, the levels of damage would be over-estimated, resulting in higher event losses once included in a loss model. The vulnerability functions proposed by this research could be greatly improved if earthquake-induced ground shaking models can be improved – more strong-motion instruments in urban areas on slopes – and more detailed building data (e.g. deep-dive data) incorporated. The variation between the earthquake-induced ground motions forecast by the different models, and the uncertainties inherent within each model, are the largest uncertainties in these analyses. Overall, these results suggest that the PGA models used to forecast losses would provide more accurate estimates if topographic amplification effects, albeit relatively simple ones like those listed in Eurocode 8, are considered in earthquake-induced ground-motion simulations (auths)
