5:15 PM - 6:45 PM
[SSS03-P03] Detection of slow slip event from seafloor pressure record using PCA: A case study in Hikurangi Subduction zone
Keywords:Seafloor geodesy, Ocean bottom pressure gauge, Hikurangi Subduction Zone
Ocean Bottom Pressure gauge (OBP) is an essential sensor for observing seafloor crustal deformation, although noise reduction is a critical issue for tectonic signal observations. Oceanographic fluctuations are a significant source of noise for slow slip event (SSE) observations due to their closed time scales. Previous studies have attempted to reduce these oceanic noises from OBP data for seafloor geodetic observations (e.g., Wallace et al., 2016). Wallace et al. (2016) used the relative time series from the reference site to reduce the common oceanographic fluctuations. However, conventional methods are required to sort for each marine region because the characteristics of oceanographic fluctuations vary between areas. Therefore, no systematic method for transient event detection from OBP data focused on the spatial distributions. Otsuka et al. (2023) focused on the spatial scale difference between oceanographic fluctuations and SSEs and constructed the anomaly detection method using Principal Component Analysis (PCA). The authors conducted synthetic SSE-like ramp signals to the observed time series and evaluated the SSE detectability from the spatial distribution variations of OBP data. In this study, we applied this PCA-based method for the OBP observatory in Hikurangi margin (HOBITSS), where the SSE occurrence was reported in GNSS and OBP, to try to detect the transient event.
HOBITSS monitored the seafloor pressure in 2014–2015. During the observation period, two SSEs were detected by in-land GNSS observations (Warren-Smith et al., 2019). The first event (Event-I) was also observed by OBP (Wallace et al., 2016; Muramoto et al., 2019). The transient pressure variations were also observed by OBP in the same period as Event-II (He et al., 2020). However, the major SSE slip distribution during Event-II was estimated in the southward of HOBITSS based on the inversion from GNSS data (Warren-Smith et al., 2019), and the slip amount in the HOBITSS region was relatively small. In this study, we extract the time series with a 50-day time window and compare the spatial distribution of each PC with the PC distributions calculated from the whole observation period (285 days). The similarity of PCs can be evaluated by their NIP (Normalized Inner Product) (Otsuka et al., 2023). We used PC2 to compare PCs due to their temporal stabilities.
When applying PCA to HOBITSS data, the common and sea-depth-dependent components appeared in PC1 and PC2, respectively. These characteristics are similar to the Nankai region, where the PCA-based method was constructed by Otsuka et al. (2023). Therefore, we determined this PCA-based method can be applied to the Hikurangi region. The temporal variations of PC1 and PC2 were stable in large value (~1.0) because their components had temporal stabilities. In contrast, PC3 and PC4 had large temporal perturbations, indicating these components are unstable. Focusing on the PC2 time series, the NIP value decreased temporally during the SSE periods reported by Warren-Smith et al. (2019). The more significant decrease in NIP value appeared in the latter SSE period (Event-II). Therefore, we focused on Event-II and evaluated the spatial pattern of the displacement.
We assumed the transient displacement was included in the PC2 because the time series of PC1 was stable at more than 0.95. According to the transient pressure variation of PC2, the uplift and subsidence appeared in up- and down-dip stations along the dip direction, respectively. This spatial displacement pattern is consistent with the reverse fault on the plate boundary. Although the estimated slip amount in the HOBITSS region (Warren-Smith et al., 2019) is small, the displacement pattern was consistent with the OBP stations. In the presentation, we will discuss the slip distribution due to the SSE, including the displacement observed by OBPs.
HOBITSS monitored the seafloor pressure in 2014–2015. During the observation period, two SSEs were detected by in-land GNSS observations (Warren-Smith et al., 2019). The first event (Event-I) was also observed by OBP (Wallace et al., 2016; Muramoto et al., 2019). The transient pressure variations were also observed by OBP in the same period as Event-II (He et al., 2020). However, the major SSE slip distribution during Event-II was estimated in the southward of HOBITSS based on the inversion from GNSS data (Warren-Smith et al., 2019), and the slip amount in the HOBITSS region was relatively small. In this study, we extract the time series with a 50-day time window and compare the spatial distribution of each PC with the PC distributions calculated from the whole observation period (285 days). The similarity of PCs can be evaluated by their NIP (Normalized Inner Product) (Otsuka et al., 2023). We used PC2 to compare PCs due to their temporal stabilities.
When applying PCA to HOBITSS data, the common and sea-depth-dependent components appeared in PC1 and PC2, respectively. These characteristics are similar to the Nankai region, where the PCA-based method was constructed by Otsuka et al. (2023). Therefore, we determined this PCA-based method can be applied to the Hikurangi region. The temporal variations of PC1 and PC2 were stable in large value (~1.0) because their components had temporal stabilities. In contrast, PC3 and PC4 had large temporal perturbations, indicating these components are unstable. Focusing on the PC2 time series, the NIP value decreased temporally during the SSE periods reported by Warren-Smith et al. (2019). The more significant decrease in NIP value appeared in the latter SSE period (Event-II). Therefore, we focused on Event-II and evaluated the spatial pattern of the displacement.
We assumed the transient displacement was included in the PC2 because the time series of PC1 was stable at more than 0.95. According to the transient pressure variation of PC2, the uplift and subsidence appeared in up- and down-dip stations along the dip direction, respectively. This spatial displacement pattern is consistent with the reverse fault on the plate boundary. Although the estimated slip amount in the HOBITSS region (Warren-Smith et al., 2019) is small, the displacement pattern was consistent with the OBP stations. In the presentation, we will discuss the slip distribution due to the SSE, including the displacement observed by OBPs.