Seismic loss assessment for İstanbul under a probable earthquake sequence in Marmara Sea

Abstract

Earthquakes are natural phenomena that typically occur as a series of events. Past experience shows that, especially after large earthquakes, many aftershocks occur following the mainshock event. Moreover, some of these aftershocks can be devastating and can significantly increase the overall seismic damage. However, existing seismic loss assessment studies are typically based on the mainshock event alone, and the number of studies that take aftershocks into account is quite limited. This limitation highlights the need to incorporate aftershock effects into seismic loss assessment frameworks. Accordingly, this thesis investigates the extent to which destructive aftershocks can increase the existing losses caused by a mainshock event, using a probable earthquake sequence scenario in the Marmara region and its estimated losses in Istanbul as a case study. During the 1999 Izmit and Duzce earthquakes, still debated as an aftershock or triggered second mainshock, as well as the 2011 Van earthquake sequence, in which a damaging M5.6 aftershock followed the M7.1 mainshock, subsequent events affected mainshock-damaged buildings and increased the total seismic losses. Recently, the 6 February 2023 Kahramanmaras earthquake sequence in Turkey is one of the most devastating examples. Therefore, these events have clearly demonstrated the importance of incorporating successive earthquakes such aftershocks and damage accumulation effects into seismic loss assessment frameworks. In this study, as the mainshock scenario an M7.5 rupture model which is identified as a probable scenario for the Sea of Marmara in the IMM–JICA microzonation study (JICA-IMM, 2002), was adopted. For the aftershock scenarios, three M6.5 rupture models are defined as assumed to rupture subsections (i.e. Cinarcik and Central Marmara) of North Anatolion Fault that have not ruptured during the mainshock event. An M6.5 aftershock scenario was considered sufficiently strong to cause additional damage to the building stock damaged by the mainshock. These scenarios are: Case-Mid, located in the central region of the mainshock rupture, and Case-West and Case-East, located at the western and eastern ends, respectively. This thesis uses the open-access exposure dataset for Istanbul, provided within the ESRM20 project for loss assessment of an earthquake sequence (Crowley et al., 2021). The Istanbul exposure dataset shared in ESRM20 consists of detailed building stock information and population data. This study models potential vulnerability under aftershocks by proposing a simple systematic vulnerability modification method. In this method, an economic loss ratio threshold is first defined to identify buildings that are considered to have collapsed after the mainshock event. These buildings are then excluded from the input building stock used in the aftershock loss analyses. For mainshock-damaged buildings below this threshold, the vulnerability model used in the aftershock analyses is modified as a function of the damage sustained by the structure during the mainshock. The increase in vulnerability is represented by reducing the ground motion intensity measure values corresponding to specific loss ratios; in other words, the vulnerability model is shifted towards the left along the horizontal axis. This assumption is based on the idea that a structure damaged during the mainshock may reach the same loss level under lower ground-motion intensity values in a subsequent aftershock, thereby reflecting the increased vulnerability caused by damage accumulation. Subsequently, sensitivity analyses were conducted using alternative functions for the damage accumulation model. According to earthquake sequence loss estimation results for Istanbul, the findings indicate that, in all earthquake sequence scenarios, considering the mainshock and aftershock together tends to increase the total losses compared to the mainshock-based case, including structural, nonstructural, contents, and occupant losses. When comparing the impact of aftershock locations on seismic losses, the Case-Mid scenario, in which the epicentres of the mainshock earthquake and the aftershock are closest together, shows the greatest increase in damage for all the scenarios. This finding is consistent with the existing literature. The Case-West scenario, which is closer to the European side of Istanbul and has a higher level of exposure, was the second scenario to increase losses the most. The Case-East scenario, which is closer to the Asian side and has a relatively lower level of exposure, increased losses the least of these three scenarios. However, while the Case-East scenario results in the lowest level of additional damage of the three aftershock locations, it should be noted that this is still considerable. According to result based on the benchmark 0.1 extreme-limit linear damage accumulation model, structural losses increase between 5.3% and 89% when all scenarios are considered. In the most devastating Case-Mid aftershock scenario, the loss increase ranges from 9.2% to 89%, with a median increase of 18.7%. When the damage accumulation sensitivity analysis results for structural losses are compared, it is observed that the highest aftershock-induced loss increase ratio (LIR) is obtained from the 0.1 extreme limit rapidly decaying model with a median LIR value of 46%. Based on the 0.1 extreme-limit linear damage accumulation model, nonstructural losses increase by between 4.8% and 42.3% when all scenarios are considered. In the most devastating Case-Mid aftershock scenario, the increase ranges from 10.5% to 42.3%, with a median increase of 18.9%. When the damage accumulation sensitivity analysis results for nonstructural losses are compared, it is observed that the highest aftershock-induced loss increase ratio (LIR) is obtained from the 0.1 extreme limit rapidly decaying model with a median LIR value of 18.9%. When the all scenarios are considered, the aftershock-induced LIR ranges from 1.0% to 77.5%. For contents, which is one of the loss components where no damage accumulation is assumed, the LIRs range from 1.8% to 15.1% when all scenarios are considered. For occupants, which is the another loss component where no damage accumulation is assumed, the LIRs range from 0.1% to 33.5% when all scenarios are considered. In most of the worst-case aftershock scenarios, the districts which densely exposed are identified as high-risk under the mainshock scenario remain among the most critical districts in terms of aftershock-induced losses. However, under the median-case aftershock scenarios, the spatial distribution of aftershock risk becomes more uniform, with risk being spread more broadly across the central districts of Istanbul. The comparison between the M6.5 event considered as a mainshock and as an aftershock scenario following the M7.5 mainshock shows that, for both structural and nonstructural components, the median loss ratios obtained from the M6.5 aftershock scenario are higher than those obtained from the M6.5 mainshock scenario. This finding clearly indicates the progression of damage in the post-mainshock exposure due to the aftershock following the M7.5 mainshock. Finally, this study will contribute to decision-making processes related to disaster management strategies, temporary shelter planning, and rapid disaster response, thereby improving planning in the post-mainshock phase.