R ESOLUTION T EST

In order to check the reliability of the 3-D tomographic solution, the Checkerboard Resolution test (CRT; Grand, 1987) and the Restoring Resolution test (RRT; Zhao et al., 1992) were carried out. The checkerboard velocity model was constructed by assigning the velocity perturbations of ＋3％ and －3％ to the grid points and it lead to 6％

velocity discontinuity across the boundary of checkerboard boxes. The block size in CRT is the same as that in coarse 3-D inversion.

From the P-wave velocity results of CRT, it shows only the portion with depth shallow than 9 km is well return to the checkerboard model in Taipei Basin and Tatun region. The small green squares in Figure 31 and 32 represent the station distribution of the seismic networks in metropolitan Taipei and the small seismic monitoring at TVG.

The resolution is good within the region covered with the seismic stations. Moreover,

the Vp/Vs resolution is poor than P-wave velocity resolution and it is only good in layers at depths shallow than 6 km in target region. It is because that most of earthquakes occurred within the target region are shallow than 10 km and seldom earthquake occurred between 10 and 60 km (Figure 33). Those earthquakes with depth larger than 60 km are associated with the subducting slab. So, the poor resolution in the target region makes sense. Thus, the results on the shallow portion of the target region were discussed.

On the other hand, the reliability of the coarse solution was checked by RRT. The arrival time data was made by adding random errors with standard deviations of 0.02 s for P-wave and 0.05 s for S-wave to theoretical data set, respectively. It is because that S-wave is generally contaminated by the P-wave coda and then has larger picking errors.

Moreover, the same picking errors were used in Nakamichi et al., (2007) and its inversion region is approximately the same as the target region in this study. In order to understand the image recovery, a quantity called the “recovery rate” was introduced (Nakamichi et al., 2007). It is drawn in black when the tomographic result of coarse normal inversion is completely recovered or a little bit over-recovered and is approximately white when the resulting velocity perturbation is zero or has sign opposite the tomographic result of coarse normal inversion. If the tomographic result of RRT is widely over-recovered, the region is drawn in light purple. In Figures 35 to 39, most of area is over-recovered, especially beneath TVG. But the main pattern of the tomographic results does not have significant difference. So, it means the main geological trend and structures is approximately imaged upon the data set. Besides, the recovery results seem to correlate with the hypocenter and station distribution. It means it is deeply influenced by the ray-paths of seismic waves. Thus, the lack of events in the target region is a dominant influence in obtaining the detail tomographic images.

Figure 32. The checkerboard test result of Vp/Vs ratio.

Figure 33. The relocated earthquakes distribution.

Figure 34. Locations of profiles A to E.

Figure 35. Reliability test along profile AA’. (A) Results of the coarse normal inversion along AA’. (B) Results of the RRT along profile AA’. (C) Tomographic recovery rate along profile AA’.

Figure 36. Reliability test along profile BB’. (A) Results of the coarse normal inversion along BB’. (B) Results of the RRT along profile BB’. (C) Tomographic recovery rate along profile BB’.

Figure 37. Reliability test along profile CC’. (A) Results of the coarse normal inversion along profile CC’. (B) Results of the RRT along CC’. (C) Tomographic recovery rate along CC’.

Figure 38. Reliability test along profile DD’. (A) Results of the coarse normal inversion along profile DD’. (B) Results of the RRT along profile DD’. (C) Tomographic recovery rate along profile DD’.

Figure 39. Reliability test along profile EE’. (A) Results of the coarse normal inversion along profile EE’. (B) Results of the RRT along profile EE’. (C) Tomographic recovery rate along EE’.

5.4 3-D Velocity Model Result

Detail tomographic images and the relocated hypocenters distribution were shown in Figure 40 to 44. The earthquakes within 2 km to the profile are plotted. Figure 46 shows the recordings number of each station was used in this study. In Figure 40, Figure 41 and Figure 44, the Taipei Basin is well imaged as the high Vp/Vs ratio in the near-surface portion.

A significant lower Vp/Vs ratio zone which seems associated with the Chinshan and Shanchiao fault is observed through entire region. Chang (2004) studied the 3-D Vp, Vp/Vs and Qp structures beneath TVG and also found the significant high Vp, high Vs and low Vp/Vs ratio inclining toward southeast beneath TVG. Because the Chinshan fault is formed earlier than the eruption of Tatun volcanoes, they inferred the feature to the solidification igneous rock along Chinshan fault. But, the significant zone is a continuous feature through entire study region in this study (Figure 44). Because of the poor resolution of the depth larger than 9 km in Taipei Basin and 6 km in TVG, the feature might reflect the hard rock site relative to the upper soft sedimentary. Besides, most of the earthquakes occurred beneath Taipei Basin are located within this zone. The feature also well depicts the half-graben shape of Taipei Basin in Figure 44.

A high Vp and low Vp/Vs ratio region beneath TVG at subsurface zone should represents the solidified igneous rock (Figure 42, Figure 43). In Figure 45, the Vp and Vs perturbation maps also show the shallow portion with higher Vp and higher Vs relative to the surrounding region.

Result of depth larger than 10km in Figure 42 and 43 due to lack of earthquakes, it is not clear to identify that a region with relatively high Vp/Vs and low Vp exist or not. It may be caused by inadequate ray-paths passing through and the unsuitable ray tracing

algorithms used in this study. Zhao et al. (1992) pointed out the pseudobending technique fails to work when velocity discontinuities exist and it only can be applicable to continuous velocity models. So, if a magma chamber or fluid saturated zone really exists, the pseudobending technique may be inappropriate.

Based on the earthquakes locations, there seems that a seismogenic zone is perpendicular to the Chinshan fault beneath TVG. This zone with seismicity may correlate with the Chinshan fault. This part will be discussed in next section.

Figure 40. The tomographic profile along line AA’.

Figure 41. The tomographic profile along line BB’.

Figure 42. The tomographic profile along line CC’.

Figure 43. The tomographic profile along line DD’.

Figure 44. The tomographic profile along line EE’.

Figure 45. Vp and Vs perturbation maps at 2 and 4 km depth. The green square represents the seismic network in Taipei Basin and the gray square is the station in TVG.

Figure 46. The recording number for SIMUL2000. The cartoon shows seismic recording number in each station during executing SIMUL2000 (see Table 5).

Table 6. The recording number for SIMUL2000.

Station Number The recording number for SIMUL2000 (P phases + P-S time)

TWE 6049 TWU 1907 YM03 581 YM12 202 E038 126

TWC 6030 EGS 1757 YM06 567 TWCP 186 TB25 121

TWB1 5645 TAP1 1657 YM04 520 E027 161 E059 118

TWA 5459 NCU 1500 YM07 518 TB09 159 TB19 112

ENT 5136 TAP 1243 YM05 511 TB22 157 E007 109

TWS1 4856 TWZ 1149 YM10 332 TB28 151 E037 107

NSK 4639 TWX 893 E032 314 E027 146 TB32 107

NWF 4071 ANP 771 YM09 303 E055 144 E028 106

TWY 3180 YM01 704 YM11 247 TB02 144 E046 106

ILA 2324 YM08 582 E031 227 YM02 139 ．

． ．