Hierarchical K-means clustering of earthquake data for in-depth spatial and magnitude analysis in Gujarat, India

This study aims to understand seismic patterns by analysing the spatial, magnitude, and depth distribution of seismic events in Gujarat, India, using hierarchical K-means clustering to analyse earthquake data in Gujarat, a region known for its high seismic activities. An unsupervised machine learning-based clustering algorithm is used to examine a dataset of earthquake occurrences from 1819 to 2023, focusing on earthquakes with a magnitude of two or above. Adopting a hierarchical approach in K-means clustering, we first identified six clusters at Level-1 based on latitude and longitude coordinates. To elucidate seismic patterns inside the major clusters, we sub-clustered each Level-1 cluster by magnitude and depth. The silhouette score, Calinski-Harabasz index, Davies-Bouldin index, and elbow technique ensured reliability and accuracy in determining the number of clusters. The compactness and separation of the six clusters at Level-1 were assessed using the Dunn-2 parameter, resulting in a score of 0.891035, indicating well-defined clusters. Distinct seismic clusters, together with their respective magnitudes and depths, were found, offering crucial information on seismic activity patterns and their geological correlations in the region. Expanding this research to encompass broader geographic regions could enhance hazard forecasting and disaster preparedness by detecting precursors in each cluster zone. Moreover, it provides essential insights for earthquake risk assessment, urban development and improves disaster management strategies.