Ocean Floor Positioning With Temporal Variation in Acoustic Velocity

We are developing an ocean floor positioning system to monitor the crustal deformation on the ocean floor using GPS and acoustic ranging. We determine the location of a vessel by Kinematic GPS positioning to measure the distance between the vessel and a benchmark unit on the floor using acoustic ranging. We have installed benchmark units to carry out a continuous monitoring at Suruga Bay in 2002. In this study, we report the results of the observation using the benchmark units at two sites situated in the northwestern and northeastern part of the Suruga Bay on October and November 2002, respectively. Each site is composed of three units within radius of about 200m. The first measurement was performed during their initial installation and the subsequent measurements were undertaken twice a year for each site. The acoustic ranging for each unit was repeatedly done from 200 to 400 times during each measurement in a one-day observation period. Throughout the observation period, we let the vessel moved to a certain distance adrift with the wind and current for a several hours while sending and receiving acoustic signals. Each drift made an observation line and was repeated which covered suitable area. In parallel with the acoustic ranging measurement, we also measured the acoustic velocity of ocean water using a CTD profiler. It showed that the velocity averaged over the ray path sometimes varies up to 3m/s (0.2%) during one-day measurement. According to a theoretical research, the positioning error might be up to 1 m if the position of the units were determined using homogeneous acoustic velocity. However, it should be noted that the actual velocity varies with time. Thus, analysis must be done considering the temporal variation in velocity structure. We first tried to determine the position of the units and velocity by grid-search method on assumption that velocity should be homogeneous in time and space. Residuals of the travel time were up to 1 ms. At that time, the actual velocity variation measured with CTD profiler. The measured velocity gave a good explanation for the residual of travel time. Furthermore, we tried to determine the benchmark position assuming temporal variations in the velocity structure, in which velocity varied with time but was constant for respective drift lines. Residual of the travel time reduced to about 0.1 ms. In this case the positions and the velocity were estimated for each unit independently. The temporal variation for estimated velocities were similar in pattern for all tree units in each site although they were different in absolute value of up to 2m/s (0.13%). The temporal variation patterns of velocity were similar even though the geometrical relationship between the wake of the vessel and the locations of the units were different. This may suggests plausibility of the model of the temporal variation in acoustic velocity. In contrast, the difference in absolute velocities between the units suggests that a trade-off between acoustic velocities and distance has not been solved yet. The measurements were done four times for northwestern site and three times for northeastern site from 2002 to 2003. Based on these observations, we estimated the location of the weight centers of the three units in the each site through the above method. As a result, repeatability of positioning was about 50 cm for both horizontal and vertical directions for both sites. Lastly, we will also attempt to address and discuss the result of analysis in constraining three units simultaneously using a varying velocity.