[S-CG44] Science of slow-to-fast earthquakes

[SCG44-P21] Comparison of spatiotemporal distributions of offshore tremor activities in the northern part of the Hikurangi subduction margin, New Zealand

*青山 都和子¹、望月 公廣¹、山下 裕亮²、山田 知朗¹ (1.東京大学地震研究所、2.京都大学防災研究所)

キーワード：ヒクランギ沈み込み帯、微動、海域地震観測

The Hikurangi margin is located off the east side of the North Island of New Zealand, where the Pacific Plate subducts beneath the Australian Plate at 50-60 mm/year in the northern section and 20-30 mm/year in the southern section of the margin (Wallace et al., 2004). Offshore the North Island, the Hikurangi Plateau, a large igneous province with numerous seamounts, subducts. At the Hikurangi margin, slow slip events (SSEs) occur at recurrence intervals of 18-24 months offshore at depths less than 15km, and last one to three weeks (Wallace and Beavan, 2014; Wallace et al., 2012). In 2014-2015, Hikurangi Ocean Bottom Investigation of Tremor and Slow Slip (HOBITSS) experiment deployed ocean bottom seismometers (OBSs) and absolute pressure gauges (APGs). During the observation, a large SSE occurred in September 2014 beneath the HOBISS network, and its slip distribution were determined in detail by using both onshore GNSS and offshore APG data (Wallace et al, 2016). Offshore tremors accompanying the SSE were first identified by Todd et al. (2018) concentrated in the down-dip side of the subducted seamount that is located near the largest slip of the SSE. They apply a band-pass filter in two frequency ranges, 4-10 Hz and 12-20 Hz to isolate tremor with energy at lower frequencies and remove local earthquake detection that retain energy above 12 Hz, respectively. The tremor activity lasted for about three weeks after the SSE. Another offshore seismic observation was conducted around the subducted seamount from October 2018 to October 2019 using 5 OBSs. A large SSE occurred in March 2019 under the network, and similar tremor activity accompanying the SSE was observed (Yamashita et al., 2021). Three-dimensional P-wave velocity structure has been revealed around the subducted seamount from a 3-D seismic survey in 2018, and the subducted seamount has been better imaged (Arai et al., 2020). The 2018 tremor activity was found to be located in the up-dip side of the subducted seamount.
In order to compare in detail the tremor distributions between the 2014 and 2018 activities, we apply the same detection and location method employing the envelope correlation technique to the 2-4 Hz band-passed OBS data from the two observations. This method detects much more tremors than the method used by Todd et al. (2018). However, the detected events include not only tremors but also regular earthquakes. We try to remove earthquakes from the detected events by investigating the frequency content of the event waveforms and by applying an AI-based earthquake signal detector, EQTransfomer (Mousavi et al., 2020). We found that earthquake discrimination by investigating frequency content may not work properly for events at times with large low-frequency ambient noise on the OBS data. By incorporating the AI-based earthquake detection method together, we may be able to better remove regular earthquakes. After removal of earthquakes from the tremor catalog, we will discuss differences in spatiotemporal distributions of tremor activities in 2014 and 2018.