5:15 PM - 7:15 PM
[SCG55-P28] Quantitative comparison of underwater sound speed structures derived from GNSS-Acoustic observations and ocean models
Keywords:GNSS-A positioning, ocean model
GNSS-Acoustic (GNSS-A) positioning technique is one of the seafloor geodetic observations. This technique measures seafloor displacement as the translation of a seafloor transponder array (array displacement), which consists of 3–6 seafloor transponders. A critical issue of this method is the treatment of an underwater sound speed structure (SSS). Recent studies have represented a SSS as a combination of a horizontally-stratified SSS and a multiple-layered gradient structure, estimating the corresponding sound speed parameters (e.g., Tomita & Kido, 2024). Although sufficient variation in the incident angles of an acoustic raging dataset is essential for accurate estimation of the sound speed parameters expressing the gradients (the gradient parameters), adequate data collection is challenging when using a Wave Glider because of its low navigation speed. Recently developed ocean models provide precise SSSs derived from temperature and salinity fields, and may potentially offer constraints on the gradient parameters. This study investigates the applicability of ocean models to GNSS-A.
We employed the ocean models of JCOPE2M (Miyazawa+, 2017; 2019) and MRI.COM (Tsujino+, 2015; Sakamoto+, 2018). In the GNSS-A positioning method of Tomita & Kido (2024), multiple-layered gradients are represented by a time-invariant slowness gradient profile, and the depth integral of the slowness gradient profile is estimated as the gradient parameter. This approach allows for a quantitatively comparison between GNSS-A and the ocean models because we can easily calculate a slowness gradient profile above a GNSS-A site from the ocean models.
First, we calculated slowness gradient profiles at four GNSS-A sites (CHOS and G19 off Ibaraki, and KAMN and G04 off Kamaishi) for 500 randomly selected days during 2013–2019 profiles to investigate the local characteristics of the slowness gradient. We found no significant gradient at the deep portion (>1500 m) for any sites and a temporally stable gradient at the shallow portion (0–800 m) of CHOS and G19, likely due to the Kuroshio extension.
Next, we calculated the gradient parameters from the ocean models and compared them with those estimated by GNSS-A. This analysis focused on CHOS and KAMN where data quality was sufficient for the gradient parameter estimation (Watanabe+, 2021). At CHOS, direction of the gradients obtained from the ocean models is consistent with that estimated by GNSS-A. This suggests that the characteristic wavelength of the sound speed heterogeneity off Ibaraki is longer than the horizontal resolution of the ocean models (JCOPE2M: ~9 km, MRI.COM: ~3 km). However, the gradient intensity did not show a consistent agreement. At KAMN, neither the gradient direction nor intensity was consistent between the ocean models and GNSS-A. These discrepancies may be attributed to small scale heterogeneity with wavelengths shorter than the horizontal resolutions of the ocean models (<~3 km).
Finally, we calculated array displacements using the gradient parameters derived from the ocean models (ocean model solutions). Assuming that the true values of the array displacement were the solutions when the gradient parameters are solved as the variables in GNSS-A, we compared the ocean model solutions with the solutions without considering the gradients (zero-gradient solutions). The ocean model solutions were slightly more accurate than the zero-gradient solutions for both ocean models at CHOS. At KAMN, however, the ocean model solutions showed cases of both improvement and deterioration.
This study demonstrates the potential of employing ocean models as a constraint on gradient parameters when a long-wavelength spatial heterogeneity dominates as shown at CHOS. However, further investigations are necessary to assess the influence of small scale heterogeneities using a higher resolution ocean model and datasets from other sites in future studies.
We employed the ocean models of JCOPE2M (Miyazawa+, 2017; 2019) and MRI.COM (Tsujino+, 2015; Sakamoto+, 2018). In the GNSS-A positioning method of Tomita & Kido (2024), multiple-layered gradients are represented by a time-invariant slowness gradient profile, and the depth integral of the slowness gradient profile is estimated as the gradient parameter. This approach allows for a quantitatively comparison between GNSS-A and the ocean models because we can easily calculate a slowness gradient profile above a GNSS-A site from the ocean models.
First, we calculated slowness gradient profiles at four GNSS-A sites (CHOS and G19 off Ibaraki, and KAMN and G04 off Kamaishi) for 500 randomly selected days during 2013–2019 profiles to investigate the local characteristics of the slowness gradient. We found no significant gradient at the deep portion (>1500 m) for any sites and a temporally stable gradient at the shallow portion (0–800 m) of CHOS and G19, likely due to the Kuroshio extension.
Next, we calculated the gradient parameters from the ocean models and compared them with those estimated by GNSS-A. This analysis focused on CHOS and KAMN where data quality was sufficient for the gradient parameter estimation (Watanabe+, 2021). At CHOS, direction of the gradients obtained from the ocean models is consistent with that estimated by GNSS-A. This suggests that the characteristic wavelength of the sound speed heterogeneity off Ibaraki is longer than the horizontal resolution of the ocean models (JCOPE2M: ~9 km, MRI.COM: ~3 km). However, the gradient intensity did not show a consistent agreement. At KAMN, neither the gradient direction nor intensity was consistent between the ocean models and GNSS-A. These discrepancies may be attributed to small scale heterogeneity with wavelengths shorter than the horizontal resolutions of the ocean models (<~3 km).
Finally, we calculated array displacements using the gradient parameters derived from the ocean models (ocean model solutions). Assuming that the true values of the array displacement were the solutions when the gradient parameters are solved as the variables in GNSS-A, we compared the ocean model solutions with the solutions without considering the gradients (zero-gradient solutions). The ocean model solutions were slightly more accurate than the zero-gradient solutions for both ocean models at CHOS. At KAMN, however, the ocean model solutions showed cases of both improvement and deterioration.
This study demonstrates the potential of employing ocean models as a constraint on gradient parameters when a long-wavelength spatial heterogeneity dominates as shown at CHOS. However, further investigations are necessary to assess the influence of small scale heterogeneities using a higher resolution ocean model and datasets from other sites in future studies.