A New High-Resolution Seismic Catalog for Southwestern Australia (2020–2025) and Analysis of Long-Term Clustering Behavior

Abstract

We present a new machine-learning-based catalog of southwestern Australia, a stable intraplate zone primarily comprised of the Archean aged Yilgarn craton and the continent’s most seismically active region. About 29,000 events were located between 2000 and May 2025 with 43% of these presumed to be related to anthropogenic mining based on location and temporal filtering. Most (75%) events were located following the new SWAN (2P, 2020) and WA Array networks (WG, 2022), which collectively added ∼340 stations from 2020 to 2025 and were the first to target this region in detail. We observe a very high degree of spatially correlated clustering, which contains power-law, Omori-type mainshock–aftershock behavior as well as low-volume and low-magnitude atemporal clustering we label as “drip-type” behavior. Drip-type clustering is presumed to reflect the long-tail baseline activity following the cessation of temporally correlated behavior following large earthquakes, but may also be unrelated to past activity. As such, the identification of drip-type clusters could be used to infer the location of prehistoric seismicity and future seismic risk. Three recent significant earthquake sequences were also analyzed in detail: Arthur River (2022), Gnowangerup (2023), and Wyalkatchem (2024), which is still producing significant seismicity as of publication. In each, the distribution of hypocenters is shallow ( < 5 km) but mostly disorganized, no clear fault plane could be resolved, and the largest event in the sequence was preceded by a significant but smaller magnitude earthquake by several weeks to months. All three sequences also show centroid moment tensor solutions consistent with the expected west–east compression regime in southwest Australia. “Drip-type” activity preceded both Gnowangerup and Wyalkatchem, but the earthquakes at Arthur River sequence were the first at that location in our catalog.