Deep-Learning-Based Construction of a High-Resolution Earthquake Catalog for the 2023 Jishishan Ms 6.2 Earthquake Sequence

On 18 December 2023, an \((M_{\mathrm{s}})\) 6.2 earthquake struck Jishishan, Gansu Province, northwestern China. The resulting aftershock sequence provides an important opportunity to investigate the seismogenic structure and postseismic activity in the source region. In this study, we analyzed continuous waveform data recorded by 31 seismic stations within 250 km of the epicenter from 19 to 31 December 2023 to construct a high-resolution earthquake catalog. Seismic phases were automatically detected using the deep-learning phase picker PhaseNet, followed by phase association and preliminary location with the REAL algorithm. Absolute locations were then refined using VELEST, and relative relocations were performed using HypoDD. The resulting catalog contains more than 3200 earthquakes, more than three times the number reported in the routine catalog of the China Earthquake Networks Center (CENC), substantially improving the completeness of the aftershock sequence. The relocated events are mainly distributed at depths of 5--16 km, with an estimated rupture length of approximately 18 km and a width of about 13 km, and they define two major aftershock clusters. Fault-structure analysis based on the E2F program indicates that the aftershocks exhibit a clear NW--SE alignment and are primarily concentrated between the LJS-NF and LJS-SF faults. These observations suggest that the Jishishan earthquake sequence likely ruptured a secondary or previously unmapped fault structure located between the two major mapped faults. The high-resolution earthquake catalog presented here provides new constraints on the detailed structure of the aftershock sequence and contributes to a better understanding of the seismogenic setting of the region.