Retrospective assessment of rupture propagation probability in the 2023 Kahramanmaraş, Türkiye, earthquake considering multiple fault paths

Coseismic rupture length is a key parameter for predicting earthquake magnitude and strong ground motion. In contrast to the conventional assumption of fault segmentation, multi-segment earthquakes and partially ruptured earthquakes occur. Therefore, a probabilistic approach to evaluating rupture propagation plays a crucial role in hazard assessment. Some empirical studies have proposed rupture propagation probability at the fault complex and an associated method for calculating the final rupture length probability. However, in this context, case studies on actual surface rupture and objective validation are relatively limited. Here, we developed a semi-automatic method for measuring fault geometry and extracting possible rupture paths and applied it to the 2023 Kahramanmaraş earthquake sequence. The results showed that the area with relatively high rupture probabilities calculated for the Mw 7.8 Pazarcık earthquake was more extensive than the corresponding area obtained assuming rupture initiation from other segments. The actual fault rupture tends to follow the path with the least geometrical penalty. However, this cannot always be confirmed to be the case, potentially because of the uncertainty around the subsurface structure or the inherent limitation of the trace dataset used in the geometrical approach. During the Mw 7.5 Elbistan earthquake, the rupture propagated to the Doğanşehir fault, bypassing the geometrically advantageous Sürgü fault. This could be explained by the insufficient stress accumulated at the Sürgü fault, which may have forced the rupture to shift direction. These findings highlight the importance of incorporating geological parameters related to stress accumulation into a geometrical method. To explore the effect of these geological factors, we incorporated the normalized elapsed time since the last earthquake, which serves as a proxy for stress conditions and fault activity history. As a result, we find a significant probability decrease at the Pütürge segment for eastward propagation when recent stress release is considered. For westward rupture propagation, results for different assumed event times suggest that stress accumulation since before 1822 supports the long-distance rupture observed in 2023. This approach has a potential to enhance our understanding of rupture propagation paths in seismic hazard assessment.