Symmetric Nearest-Neighbor Analysis for Improved Earthquake Sequence Identification
Siyuan Zhang, & Heidi HoustonSubmitted August 30, 2026, SCEC Contribution #15547, 2026 SCEC Annual Meeting Poster #TBD
Spatiotemporal earthquake interaction is one of the most widely studied characteristics of seismicity, reflecting fundamental aspects of earthquake physics, such as nucleation mechanisms, stress transfer, and deformation modes. These interactions are commonly inferred from clustered or background earthquakes, often using the conventional nearest-neighbor analysis proposed by Zaliapin & Ben-Zion (2013). Natural seismicity typically exhibits a clustered mode, in which the proximity between an earthquake and its preceding nearest neighbor is smaller than that in a background mode that approximately follows a Poisson process, resulting in a bimodal distribution. However, the structure of individual earthquake clusters identified by the conventional analysis is not always physically reasonable. In particular, some identified foreshocks and aftershocks can be far from the mainshock. We show that the nearest-neighbor proximity to subsequent events exhibits a similar bimodality, with clearly distinguishable clustered and background modes. The similar bimodality demonstrates that our proposed alternative has equal power to separate clustered and background modes. We therefore combine both proximities and, as an improvement to the conventional analysis, propose a symmetric nearest-neighbor analysis.
Our new method treats any event as a potential center and identifies its preceding and subsequent related events. In this framework, each event in a sequence is connected to the central event either directly or sequentially, such that individual sequences are more closely attached to the mainshock. For example, aftershocks of the 2019 Ridgecrest M6.4 are removed from the sequence of the M7.1. Furthermore, the improved declustering algorithm can incorporate magnitude weighting without creating big seismicity holes after major earthquakes, which often occur in the conventional nearest-neighbor declustering. Although the new method does not partition the catalog into isolated clusters because of its bidirectional connections, it can center on moderate events within large aftershock sequences and help reveal previously obscured features of earthquake interactions. The topological features of foreshocks and aftershocks, quantified by their average leaf depths, evolve in swarm-like and burst-like manners, respectively, as a sequence develops. This likely reflects different underlying physical mechanisms and further favors our new method for the identification of foreshocks.
Citation
Zhang, S., & Houston, H. (2026, 08). Symmetric Nearest-Neighbor Analysis for Improved Earthquake Sequence Identification. Poster Presentation at 2026 SCEC Annual Meeting.
Related Projects & Working Groups
Seismology
