New approach for GNSS-A analysis with simultaneous estimation of seafloor position and sound speed variation

Japan Coast Guard has been developing the GNSS-Acoustic seafloor positioning technique (GNSS-A) since early 2000's and detected crustal deformation along the Japan Trench (e.g., Sato et al., 2011, 2013; Watanabe et al., 2014) and the Nankai Trough (e.g., Yokota et al., 2016).

Our GNSS-A system and analysis procedure are described by Yokota et al. (2018), fundamentally referring to Fujita et al. (2006). We determine the position of geometrical center of the seafloor station array, which is typically composed of four acoustic transponders. Well-aligned array of transponders enables us to estimate sound speed variations. In our method, both positions and sound speed variations are estimated from several thousands of acoustic travel time data. Because the vertical positions strongly correlate with sound speed, Fujita et al. (2006) divided the analysis procedure into two steps, i.e., positional estimation step and sound speed estimation step. In the positional estimation step, parameters for sound speed are fixed, and vice versa. This is a suitable way to obtain stable solutions under the limited spatial variation of ship track due to drifting observation which was performed at that time. Although we adopted spatially symmetric track by sailing observation using the hull-mounted system in 2009, this method is remained because it works well for horizontal displacements in centimeter and vertical displacements in decimeter level. However, when considering the stability of vertical solutions in centimeter order, the results by the conventional method have dependency on initial value of sound speed.

We then developed a new analysis program to simultaneously estimate position and sound speed variation, which provide little dependency on initial value. In addition, we updated basis functions for bulk sound speed variation (transponder-independent component) from temporal to spatio-temporal based, to directly derive the spatial variation of sound speed in the shallower part. Spatial gradient of sound speed in the deeper region (transponder-dependent component) can also be estimated. We found that the averaged value of spatial sound speed gradient has strong correlation with the horizontal position. This program enables us to investigate the differences between multiple sound speed variation models.