Ocean Bottom Crustal Deformation -Error estimation and discussion about analysis model-

We have been developing a system of ocean-bottom crustal deformation observation by using GPS/Acoustic ranging techniques. In this system, we determine the position of a vessel by a Kinematic GPS technique and measure distances between the vessel and seafloor benchmarks by means of acoustic ranging. Then we can determine the positions of benchmarks. We have installed benchmark units in two regions, Suruga Bay and Kumano basin along the Nankai trough. Each benchmark unit consists of three transponders with a certain spacing. In each region, the acoustic ranging to the three units is repeated several thousands times for each observation within two days. We keep the vessel sail slowly for several hours along planned trajectory, transmitting and receiving acoustic signals. We also measured the acoustic velocity of the sea water using a CTD profiler several times during each observation. The CTD measurement data are used for determining the reference sound velocity structure in the sea. We assume that the sound velocity is constant within a certain interval but changes with time. We estimate a velocity correction coefficient for each time interval and the positions of the three transponders simultaneously. In determining the length of interval, longer interval should be suitable for modest velocity change and shorter interval should be suitable for a steeper change in sound velocity. We can determine the length of the interval on the basis of AIC. We determined the positions of benchmark units repetitively in both regions of Suruga Bay and Kumano basin. In Suruga Bay, the positions of the benchmark units, which were measured twice in two days in June 2005, were concentrated in a circle whose radius was 5 cm in horizontal. In Kumano basin, we detected 20 cm south-southeastward co-seismic movement between two observations before and after an Mw 7.4 earthquake. We carried out additional two observations at the Kumano site after the earthquake. To evaluate the accuracy of the positioning and the effect of temporal variations in sound velocity, we carried out a numerical experiment. Travel times with random noises are synthesized and resolved to locate the benchmark units. The results prove that locations of the benchmarks are slightly affected by temporal change in the sound velocity. The positioning is mainly affected by the given random errors in acoustic travel-time. Thus, this experiment indicates that our observation could achieve a better accuracy (a few cm) than those estimated, and that spatial variations in sea water velocity affect positions of benchmarks. To overcome these errors, we need measurements of acoustic ranges three or four times more than the present observations. Such measurements could give significantly improved spatial variation to get a better accuracy as high as 2-3 cm.