Deteksi Otomatis dan Penentuan Hiposenter Gempa Mikro Menggunakan EQTransformer dan NonLinLoc:Studi Kasus Bandung, Jawa Barat

Ramadhan, Fathur Rizky (2026) Deteksi Otomatis dan Penentuan Hiposenter Gempa Mikro Menggunakan EQTransformer dan NonLinLoc:Studi Kasus Bandung, Jawa Barat. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017221051-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Sesar Lembang merupakan sesar aktif sepanjang 29 km yang membentang di bagian utara Kota Bandung, Jawa Barat, dengan laju geser sinistral 1,95–3,45 mm/tahun dan potensi membangkitkan gempa bumi Mw 6,5–7,0 (Daryono et al., 2019). Penelitian ini bertujuan mengidentifikasi kinerja pre-trained model dalam mendeteksi kejadian gempa mikro di wilayah Bandung menggunakan pipeline EQTransformer–PyOcto–NonLinLoc, serta mengkaji keterkaitan spasial hiposenter terelokasi dengan Sesar Lembang dan kompleks vulkanik Bandung Utara. Data seismogram tiga komponen dari 30 stasiun jaringan 3M pada periode Mei–Agustus 2014 diolah secara otomatis melalui tahapan preprocessing (ObsPy), deteksi dan picking fase P dan S berbasis deep learning menggunakan EQTransformer, asosiasi fase menggunakan PyOcto, penentuan hiposenter awal menggunakan NonLinLoc, dan relokasi double-difference menggunakan HypoDD dengan penggabungan data waveform cross correlation (idat=3). EQTransformer menghasilkan 505.249 deteksi, dimana 93,6% memiliki pasangan fase P dan S dengan probabilitas deteksi rata-rata 0,83, namun 85,0% picks fase P memiliki SNR di bawah 5 sehingga diperlukan quality control bertingkat. PyOcto mengasosiasikan 17.075 picks menjadi 2.606 event, yang kemudian disaring menjadi 1.382 event melalui tiga kriteria QC konsistensi. Penentuan hiposenter oleh NonLinLoc dan penerapan lima kriteria QC menghasilkan 120 event berkualitas baik dengan RMS rata-rata 0,040 s. Relokasi HypoDD dengan nilai damping optimal 30 berdasarkan analisis L-curve menghasilkan katalog akhir 70 hiposenter terelokasi dengan kedalaman 0–42 km (dominan <20 km) dan magnitudo lokal M_L 0,8–2,7 (median M_L 1,86 dan rata-rata M_L 1,76), yang seluruhnya tergolong gempa mikro. Analisis penampang vertikal memperlihatkan bahwa penampang utara–selatan (A-A' dan C-C') yang memotong jalur Sesar Lembang secara tegak lurus menunjukkan konsentrasi hiposenter di sekitar trace Sesar Lembang pada kedalaman 015 km, sedangkan penampang barat–timur D-D' yang melintas melalui kompleks vulkanik Bandung Utara (Gunung Burangrang, Gunung Tangkuban Parahu, dan Gunung Bukit Tunggul) menunjukkan distribusi hiposenter yang tersebar secara lateral tanpa membentuk kolom vertikal di bawah satu gunung api.
===============================================================================================================================
The Lembang Fault is a 29 km-long active fault extending along the northern edge of Bandung City, West Java, with a left-lateral slip rate of 1.95–3.45 mm/yr and the potential to generate earthquakes of Mw 6.5–7.0 (Daryono et al., 2019). This study aims to evaluate the performance of a pre-trained deep learning model in detecting microearthquake events in the Bandung region using an EQTransformer–PyOcto–NonLinLoc pipeline, and to examine the spatial relationship between relocated hypocenters and the Lembang Fault and the North Bandung volcanic complex. Three-component seismograms from 30 stations of the 3M network during the period May–August 2014 were processed through an automated pipeline comprising preprocessing (ObsPy), deep learning-based seismic phase detection and picking using EQTransformer, phase ssociation using PyOcto, initial hypocenter determination using NonLinLoc, and double-difference relocation using HypoDD with waveform cross-correlation data integration (IDAT=3). EQTransformer produced 505,249 detections, of which 93.6% contain paired P and S phases with a mean detection probability of 0.83; however, 85.0% of P-phase picks have SNR below 5, necessitating multi-stage quality control. PyOcto associated 17,075 picks into 2,606 events, which were subsequently filtered to 1,382 events through three consistency-based QC criteria. Initial hypocenter determination by NonLinLoc followed by the application of five QC criteria yielded 120 high-quality events with a mean RMS of 0.040 s. HypoDD relocation with an optimal damping value of 30 based on L-curve analysis produced a final catalog of 70 relocated hypocenters at depths of 0–42 km (predominantly <20 km) and local magnitudes of ML 0.8–2.7 (median ML 1.86; mean ML 1.76), all of which are classified as microearthquakes. Vertical cross-section analysis reveals that the north–south profiles (AA' and C-C') cutting perpendicularly across the Lembang Fault trace show a concentration of hypocenters around the fault trace at depths of 0–15 km, whereas the west–east profile (D-D') traversing the North Bandung volcanic complex (Gunung Burangrang, Gunung Tangkuban Parahu, and Gunung Bukit Tunggul) shows laterally distributed hypocenters without forming a distinct vertical column beneath any single volcano.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	EQTransformer, gempa mikro, HypoDD, magnitudo lokal, Sesar Lembang EQTransformer, HypoDD, Lembang Fault, local magnitude, microearthquake
Subjects:	Q Science > QE Geology > QE538.8 Earthquakes. Seismology
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geophysics Engineering > 33201-(S1) Undergraduate Thesis
Depositing User:	Fathur Rizky Ramadhan
Date Deposited:	24 Jul 2026 01:54
Last Modified:	24 Jul 2026 07:15
URI:	http://repository.its.ac.id/id/eprint/137117

Actions (login required)

View Item