Preliminary analysis of nonlinear site response at the S-net seafloor sites during three Mw 7 class earthquakes

Abstract

In this paper, we investigated the characteristics of nonlinear site response (NLSR) at 23 S-net seafloor sites using strong-motion records obtained during three Mw 7 class earthquakes that occurred directly beneath the network. During the earthquakes, horizontal peak accelerations as large values as 1,400 and 1700 cm/s2 were recorded at the land (KiK-net) and S-net stations, respectively. The S-net is a large-scale inline-type seafloor observation network for earthquake and tsunami in the Japan Trench area. Characterization of NLSR is important because, in most common cases, it can cause a reduction of higher frequency components and a shift of predominant site frequency to lower one. Obtaining high-quality strong-motion records at seafloor sites is extremely difficult and expensive. Some of the records from the three earthquakes used in this study were contaminated by the rotations of the sensor houses, resulting in the ramps and offsets after the arrival of strong S-wave phases. We used a time window of 10 s starting from the S-wave onset, that avoided the ramps and offsets mostly. Using the horizontal-to-vertical spectral ratio (HVSR) technique, we found that the selected S-net sites might have experienced substantial degrees of NLSR during the three earthquakes with peak accelerations greater than about 60 cm/s2. To investigate that the obtained features of NLSR were realistic or not at the S-net sites, we examined the NLSR at nine KiK-net sites on land where high-quality strong-motion records were obtained. We found that the KiK-net sites experienced various degrees of NLSR during the three earthquakes, and the obtained characteristics of NLSR at the KiK-net and S-net sites were comparable. We found that the NLSR affected the ground motions at frequencies mainly higher than 1 Hz at both Kik-net and S-net sites. Despite these similarities, by analyzing the spectral ratios between two horizontal component records, we suspected that the induced rotations contributed to some extent in exaggerating the degree of NLSR at the S-net sites, primarily when the components perpendicular to the cable axes were used. We concluded that consideration of induced rotational effects is necessary to understand the NLSR at the S-net sites better.