Comparative Declustering Approaches for Seismic Data: Insights from Gardner‑Knopoff, Gruenthal, Reasenberg, and Uhrhammer in the Kathmandu Valley

Abstract

Declustering is a crucial process that filters out dependent events to focus on independent mainshocks, thereby improving the accuracy of seismic hazard analysis. This study analyzes the spatiotemporal features of background seismicity in the Kathmandu Valley, Nepal using four widely applied declustering methods: Gardner and Knopoff, Gruenthal, Uhrhammer, and Reasenberg. A total of 3190 events from 2000 to 2023 that was analyzed using ZMAP software, revealing significant differences in the performance of each method. Statistical tools including the Allan factor, coefficient of variation Morisita index, and the inhomogeneous L-function were employed to assess spatial clustering, and temporal clustering behaviour. The results indicate that clustering persists in all declustered catalogs, with Reasenberg exhibiting the highest residual temporal clustering and Gruenthal the lowest. Spatial clustering analysis using the Morisita index revealed that Uhrhammer retained the strongest clustering, whereas Gruenthal produced the most homogeneous spatial distribution. Furthermore, L-function analysis with Monte Carlo simulations demonstrated that none of the declustered catalogs fully adhered to a homogeneous Poisson process, as all methods yielded p-values of zero. These findings underscore the limitations of current declustering techniques in completely removing clustering effects, emphasizing the need for refined methodologies in seismic hazard analysis. Future research should explore alternative declustering strategies and statistical models to better represent earthquake sequences in tectonically active regions like the Kathmandu Valley.