5:15 PM - 6:30 PM
[SCG45-P07] Classification and estimation of petit-spot volcanos by multibeam backscatter data
Keywords:petit-spot, GLCM, acoustic backscatter image, classification
Petit-spots are small monogenetic volcano generated on the outer rise system prior to subduction of oceanic plate. It shows several important features of the Earth (e.g. the composition of asthenosphere: Hirano et al., 2006). Although the geochemical composition of petit-spot basalts recently provides new insights both into lithospheric and asthenospheric mantle prior to subduction (e.g. Yamamoto et al., 2014; Machida et al., 2015; Pilet et al., 2016), the surface structure of volcanic distributions and lava-flowing patterns are not enough discussed in their tectonic regime. For example, the detail distribution of petit-spot volcano generally follows the tectonic lines of surface environment. Petit-spot volcanic province after the subduction plausibly controls the occurrence of interplate earthquake, contrasting the pelagic sediment of the surface of subducting plate. However, we have found only a few locations of petit-spot volcanic field. Previously, the acoustic backscatter images have been macroscopically checked to discover a petit-spot on the KR04-08 cruise in 2004 (Hirano et al., 2006). In this research, we aim to establish a method to find new petit-spots by acoustic measurement data and to classify petit-spots by topographic features such as their distributions and lava-flowing patterns.
As the petit-spot volcanic edifices are small and monogenetic of less than 1 km3 and few hundred meters in height (Hirano et al., 2008), it is difficult to find them using onboard bathymetry for deep ocean floor. The acoustic backscatter image, on the other hand, is a powerful tool to visualize submarine lava flow. In this research, we picked up data from a petit-spot volcanic field Site C, off the Fukushima and Miyagi Prefecture, as a target area, where tens of strong backscatter zones has been recognized as the presence of petit-spot sites (up to 80 sites; Hirano et al., 2008). We conducted geophysical mapping of multibeam echo sounder in this area during three cruises (KR04-08, YK14-05 and KS18-09). The CARIS HIPS and SIPS software (Teledyne CARIS Inc., Ltd., Fredericton, Canada) was used for raw data processing. In this cleaning process, both CUBE algorism and manual editing were applied. The sound velocity was corrected by real-time data of the surface water velocity meter, and observations of XCTD (expendable conductivity, temperature, and depth) probe profiles.
Backscatter image shows many of high intensity locations that are similar to some locations of petit-spot that were already sampled. There is a variation of intensity. It is considered that sediments signature or measurement conditions (e.g., track direction, horizontal distance from the ship) can affect observed beam-pattern and scattering. Additionally, highly intensity zones have a variation of the area. Comparing with topography, small areas (approximately 1 km in diameter) are located flat field. In contrast, large areas (approximately 5 km in diameter) are located topographical height (approximately 500 m in height). Therefore, it is suggested that intensity variations are also caused by scattering on the slope or existing relatively large petit-spot. Furthermore, some of large areas show some peaks on topography. It suggests an eruption by multiple petit-spot activities.
In following analytical steps, we are applying clustering of gray level co-occurrence matrix (GLCM), in which the probability of neighboring some target pixel value with any neighbor pixel value can be estimated. It is calculated by each texel of any size subarea cut from image, then calculating some values to explain textual feature (e.g. regularly pattern area, strong signal area). The features of backscatter image provide an effective means to compile topography before the clustering features by k-means clustering (e.g. Honsho et al., 2015; Fakiris et al., 2019). Based on results of the textural feature classes that describe rough surface, and intensity of backscatter image, we aim to discuss the correlation between our classified groups and the geochemical and physical variation of petit-spot lavas.
As the petit-spot volcanic edifices are small and monogenetic of less than 1 km3 and few hundred meters in height (Hirano et al., 2008), it is difficult to find them using onboard bathymetry for deep ocean floor. The acoustic backscatter image, on the other hand, is a powerful tool to visualize submarine lava flow. In this research, we picked up data from a petit-spot volcanic field Site C, off the Fukushima and Miyagi Prefecture, as a target area, where tens of strong backscatter zones has been recognized as the presence of petit-spot sites (up to 80 sites; Hirano et al., 2008). We conducted geophysical mapping of multibeam echo sounder in this area during three cruises (KR04-08, YK14-05 and KS18-09). The CARIS HIPS and SIPS software (Teledyne CARIS Inc., Ltd., Fredericton, Canada) was used for raw data processing. In this cleaning process, both CUBE algorism and manual editing were applied. The sound velocity was corrected by real-time data of the surface water velocity meter, and observations of XCTD (expendable conductivity, temperature, and depth) probe profiles.
Backscatter image shows many of high intensity locations that are similar to some locations of petit-spot that were already sampled. There is a variation of intensity. It is considered that sediments signature or measurement conditions (e.g., track direction, horizontal distance from the ship) can affect observed beam-pattern and scattering. Additionally, highly intensity zones have a variation of the area. Comparing with topography, small areas (approximately 1 km in diameter) are located flat field. In contrast, large areas (approximately 5 km in diameter) are located topographical height (approximately 500 m in height). Therefore, it is suggested that intensity variations are also caused by scattering on the slope or existing relatively large petit-spot. Furthermore, some of large areas show some peaks on topography. It suggests an eruption by multiple petit-spot activities.
In following analytical steps, we are applying clustering of gray level co-occurrence matrix (GLCM), in which the probability of neighboring some target pixel value with any neighbor pixel value can be estimated. It is calculated by each texel of any size subarea cut from image, then calculating some values to explain textual feature (e.g. regularly pattern area, strong signal area). The features of backscatter image provide an effective means to compile topography before the clustering features by k-means clustering (e.g. Honsho et al., 2015; Fakiris et al., 2019). Based on results of the textural feature classes that describe rough surface, and intensity of backscatter image, we aim to discuss the correlation between our classified groups and the geochemical and physical variation of petit-spot lavas.