5:15 PM - 6:30 PM
[SCG45-P17] Effective removal of oceanic variation from ocean bottom pressure gauge data using the MSSA
Keywords: Multi-channel Singular Spectrum Analysis, ocean bottom pressure gauge, oceanic model
1. Introduction
Ocean bottom pressure gauges (OBPs) can observe vertical movement of the seafloor continuously, and are useful to detect slow movement due to slow slip events. The OBP data include components such as oceanic tide, oceanic variation, and instrumental drift. To detect crustal movement, these components must be removed. To remove the oceanic variation, several methods were proposed in the previous studies such as average subtraction using neighboring OBPs (e.g. Suzuki et al. 2016), fitting of a parametric model (Sato et al. 2017), and subtraction of a data-assimilated oceanic model (e.g. Inazu et al. 2012). This study conducts the Multi-channel Singular Spectrum Analysis (MSSA) for the observed data and an oceanic model, and divides into components. Then, we remove only good correlation components of the oceanic model from the observed data. This method can extract vertical movement better, because this method may remove incompleteness of the oceanic model.
2. Method
As an oceanic model, we use a four-dimensional variational ocean reanalysis
for the Western North Pacific over 30 years (FORA-WNP30) (Usui et al. 2017). The OBP data are obtained at off Boso Peninsula from 2013-2015, which include the 2014 Boso slow slip event.
From the observed data, we remove the oceanic tide using Baytap08 (Tamura et al. 1991), and the drift using a linear fitting. We conduct the MSSA for the removed data and the FORA-WNP30 data, and calculate correlation between them for each component. We reconstruct oceanic model data using only good correlation components, then remove them from the removed data.
3. Results
The residual between the removed data and the reconstructed oceanic model has standard deviation of about 1.0 hPa. On the other hand, the residual between the removed data and the original oceanic model has about 1.6 hPa. This suggests that the present method can remove oceanic variation better than the previous methods.
We fit a parametric model to the residual between the removed data and the reconstructed oceanic model, and obtain standard deviation of about 0.8 hPa.
Acknowledgments
We thank the captains and crew of R/V Hakuho-maru and R/V Natsushima of JAMSTEC for their support. This work was supported by the Earthquake and Volcano Hazards Observation and Research Program authorized by the Ministry of Education, Culture, Sports, Science and Technology. This work was supported by JSPS (25287109).
Ocean bottom pressure gauges (OBPs) can observe vertical movement of the seafloor continuously, and are useful to detect slow movement due to slow slip events. The OBP data include components such as oceanic tide, oceanic variation, and instrumental drift. To detect crustal movement, these components must be removed. To remove the oceanic variation, several methods were proposed in the previous studies such as average subtraction using neighboring OBPs (e.g. Suzuki et al. 2016), fitting of a parametric model (Sato et al. 2017), and subtraction of a data-assimilated oceanic model (e.g. Inazu et al. 2012). This study conducts the Multi-channel Singular Spectrum Analysis (MSSA) for the observed data and an oceanic model, and divides into components. Then, we remove only good correlation components of the oceanic model from the observed data. This method can extract vertical movement better, because this method may remove incompleteness of the oceanic model.
2. Method
As an oceanic model, we use a four-dimensional variational ocean reanalysis
for the Western North Pacific over 30 years (FORA-WNP30) (Usui et al. 2017). The OBP data are obtained at off Boso Peninsula from 2013-2015, which include the 2014 Boso slow slip event.
From the observed data, we remove the oceanic tide using Baytap08 (Tamura et al. 1991), and the drift using a linear fitting. We conduct the MSSA for the removed data and the FORA-WNP30 data, and calculate correlation between them for each component. We reconstruct oceanic model data using only good correlation components, then remove them from the removed data.
3. Results
The residual between the removed data and the reconstructed oceanic model has standard deviation of about 1.0 hPa. On the other hand, the residual between the removed data and the original oceanic model has about 1.6 hPa. This suggests that the present method can remove oceanic variation better than the previous methods.
We fit a parametric model to the residual between the removed data and the reconstructed oceanic model, and obtain standard deviation of about 0.8 hPa.
Acknowledgments
We thank the captains and crew of R/V Hakuho-maru and R/V Natsushima of JAMSTEC for their support. This work was supported by the Earthquake and Volcano Hazards Observation and Research Program authorized by the Ministry of Education, Culture, Sports, Science and Technology. This work was supported by JSPS (25287109).