The earthquakes and the bathymetry in Taiwan regions are such that T-wave can be excited by relatively small (Mw > 3.3) earthquakes and be recorded by an OBS. The detecting threshold for the island station is a bit higher may due to the additional acoustic wave converted to elastic wave near the island.
We can predict the duration of the T-waves relatively well using the available 1000 m contour line in this region assuming Type 1 T-wave travel paths. Different paths correlate with different conversion points, and may represent different parts of the T-wave. But paths from multiple conversion points can arrive at the OBS at the same time.
It might be difficult to simulate the T-wave in Taiwan because there are so many paths from different conversion points on the 1000 m contour lines. In addition, elastic wave convert into acoustic waves at both sloping seafloor near the SOFAR-channel and abyssal seafloor also had been documented Park et al. [2001]. These imply that T-waves may have even more complicated travel paths.
Based on the synthetic waveform modeling using Type 1 path, we found that T-wave amplitude is correlated with the acceleration of the ground motions at the conversion points, which is the function of the earthquake radiation patterns and the travel path of the leg 1. The waveform amplitudes might decrease if there are gaps in the conversion points on the 1000 m contour lines. The synthetic waveforms were used to calculate the amplitudes of the T-wave at the different conversion points, most of the time, the S to T conversion generates larger T-wave amplitude than P to T conversion (Fig. 4). In terms of T-wave duration, we found that the Type 1 T-wave energy dominate the whole T-wave.
However, sometimes we can not fit the amplitudes of the T-wave very well, especially
28
the later part of the T-wave, which might be related to the 3D crustal structures generating coda that can not be simulated using the 1D velocity model. There is also T-wave energy arriving earlier than our predicted Type 1 time window. We interpret that there might be contribution from Type 2 T-wave in the observed data.
For the abyssal earthquakes, we have also observed Type 2 T-waves. We have interpreted the energy to be T-wave, instead of the S wave coda from Type 3, because it is usually rare to have such long duration of S wave coda (~ 50 sec). In addition, the amplitude of the observed T-wave almost stays as big as the Type 3 S to T conversion. If it were Type 3 T-wave, the amplitude should decay as time increases like that in the ground motion. We have also compared the ratio of the seismic ground acceleration with the pressure changes, and found that at time “a”, there is a lower ratio, showing the energy is from the water column to the ground near the OBS. We interpret that this phase is the direct path from the abyssal earthquake to the SOFAR-channel, then going down to seafloor near the OBS site. In short, from traveltime and from two way transfer function analyses, we have evidence showing the T-wave can also be excited from the deep sea [e.g. Yang and Forsyth, 2003], and propagate into SOFAR-channel before being recorded at deep sea by the OBS.
Our traveltime analyses of the 60 events shed some light on using the different phases of the T-wave for relocation purpose. We think in some cases, it is better to pick the arrival time of the maximum amplitude T-wave, and use a great circle path to relocate the event using T-wave. However, such simple case can easily be complicated by energy from multiple conversion points that arrive at the same time, or by strong radiation pattern effects, or by the availability of the conversion points in the region. In addition, the arrival time of the maximum amplitudes of the filtered ground motions sometimes does not coincide with that of the DPG. Sometime such discrepancy can be up to tens of seconds.
29
The cause of such discrepancy is still unknown. In short, although it is better to use the arrival time of the maximum amplitude for relocation purpose, there are still many factors that need to be considered.
It is difficult to pick the first arrival of the T-wave because of its emergent feature.
These early arrivals in our traveltime analyses have paths that follow the Snell’s law and Fermat’s law due to the large velocity contrast between elastic waves and the acoustic wave. Theoretically it is possible to use such early arrivals to help relocate the event. In reality, there is also some energy arriving before those “early arrival”. We interpret that these energy to be Type 2 T-wave, which has shorten acoustic leg, and thus the earlier arrival time. In other words, there are multiple mechanisms that can generate the early arrivals of the T-wave, making it difficult for us to pick it for relocation purpose.
For Type 1 path, sometime it is possible to use the gap in the T-wave to help locate the events. In Taiwan region, the 1000 m contour line is not continuous. There are gaps of conversion points available to excite T-wave in some particular locations, and thus some particular arrival times. We have seen several examples of T-wave with reduced amplitudes in some short time windows that are correlated with these gaps in conversion points (e.g. Fig. 3, 5 & 8). Such observation led us to interpret the reduced strength of the T-wave to be due to the missing conversion points at 1000 m. It also helps us to be able to use the gap to help event relocation.
Multiple conversions in one path will also reduce the T-wave amplitude. The island station has lower detection threshold, maybe due to the additional conversion from water column to the crust near the station.
Sometimes the geometry of 1000 m contour line can also reduce the T-wave amplitude by blocking the direct path in SOFAR-channel from the conversion point to
30
the station. From our study using island stations, we found that some of the T-wave energy can be blocked by the 1000 m contour line. The conversion points between 22.7°
N to 24° N do not have direct line of sight to the island station (Fig. 8). In such a shadow zone, conversion points produce T-wave that will first hit a cape to the north of the station before being recorded by the island station. Such paths will not generate strong T-wave.
On the other hand, for the earthquakes further to the north, they can generate strong T-wave from the conversion points north of the shadow zone. We interpret this as the reason that we see more T-wave from earthquakes to the north than to the west for this island station.
Using the FK code we were able to generate synthetic ground motions at the 1000 m contour line, then simulate the amplitude envelope of the T-wave recorded by the OBS.
The relatively good fits of the envelopes suggest that the T-wave amplitudes are affected by the earthquake radiation patterns. In other words, the T-wave waveforms contain information of the earthquake focal mechanism. It might be possible to extract earthquake focal mechanism using T-wave waveforms from one station by utilizing different ground motions at different conversion points along the 1000 m contour.
The ratio between ground accelerations and water pressure changes can be used as a tool to tell us whether the wave comes to the OBS from water column or from the crust.
The two-way transfer function of incident energy from crust to the water should have a same ratio. But in Figures 10 and 11, it seems S wave incident has higher ratios than that of the P incident. However, we document that we can use this two-way transfer function to discriminate whether the T-wave energy comes from the water column or from crust.
31