Reappraisal of the arc-arc collision in Taiwan

We propose a new model of the arc-arc collision in Taiwan. In this model, the Luzon arc has been colliding with the S Ryukyu forearc migrating southwestward with respect to Eurasia. The collision, thus, occurs in the NNW direction and the western Luzon arc is obducted and the eastern one is subducted. This explains the apparent lack of the collision front of the S. Ryukyu forearc with Taiwan, the NNW strike of the subducted slab edge beneath NE Taiwan, and the SW propagation of the S. Ryukyu Trench. The subducted slab migrates westward with respect to the mantle beneath the Eurasian continent. The negative E-W Bouguer gravity anomaly in the Nanao Basin may be caused by buckling of the shallow portion of the slab due to resistance of the mantle against this migration.


INTRODUCTION
There still remain a lot of enigmas in the collision of the Luzon arc in Taiwan.If the motion of the Philippine Sea plate (PH) with respect to the Eurasian plate (EU) did not differ much from that of Seno et al. (1993) since the Pliocene, a hypothetical transform fault connecting the northern tip of the Manila Trench and the southern tip of the Ryukyu Trench with a strike of the PH-EU motion direction (~N307°E) is necessary, as shown by many previous papers.If the northern end of the accretionary prism on land near Taipei is the relict northern tip of the Manila Trench, the initial intersection point of the transform with the Ryukyu Trench should be ~200 km further east of the SW terminus of the present Ryukyu Trench.Therefore, Suppe (1984) assumed propagation of the S Ryukyu Trench toward SW, transferring the collided Luzon arc to Eurasia.However, there is no indication of such transfer in the forearc near the Yonaguni Island.On the other hand, Lallemand et al. ( 2001) assumed propagation of a fracture to the west along the passive margin of Eurasia.In this case too, there is no indication that such a fracture existed offshore Taiwan distinctly young in a geological history.

LATERAL MIGARATION OF S. RYUKYU FOREARC
The interplate earthquakes in this southernmost Ryukyu Trench have slip vectors (~N345°E) deviated clockwise than the PH-EU motion (e.g., Kao et al., 1998).This indicates that the S Ryukyu forearc has been migrating SW with respect to EU due to the oblique subduction, advocated geologically by Kuramoto and Konishi (1989).Some workers (e.g., Kimura, 1996) have thought that the S Ryukyu forearc has been colliding with Taiwan due to this migration, similarly to the collision of the S Kuril forearc with W Hokkaido.However, the Ryukyu Trench abuts the NE coast of Taiwan (See the free-air gravity map of Hsu et al., 1998) and there appears no collisional feature in the north of this terminus of the trench.Reconstructing a velocity diagram between PH, EU and the S Ryukyu forearc (RA), the collision front of the Luzon arc and that of RA are separating from each other, and if the latter exists, it could be discriminated from the former.

THE WESTERN EDGE OF THE PHILIPPINE SEA SLAB
Beneath the Ilan plain, NE Taiwan, there is intermediate-depth seismicity representing the subducting PH slab down to ~200 km (e.g., Kao and Rau, 1999).The western boundary of this seismicity is elongated in the NNW direction.If PH northeast of the transform fault is subducting along the Ryukyu Trench in the PH-EU relative motion direction, we expect that the western boundary of the subducted slab is elongated in the NW direction, which is different from that observed.

A PROPOSED MODEL
We propose a new model of the collision in Taiwan to resolve the above enigmas.This model was stimulated by and extended from the idea of Hsu and Sibuet (1995) that the Ryukyu arc is extending further SW beneath Taiwan to the Tainan Basin.In our model, the Ryukyu forearc has been migrating SW since the start time of the collision or earlier.Thus, the Luzon arc has been colliding with RA, not EU.Hsu and Sibuet (1995) thought that the Ryukyu arc became inactive before the collision.In their model, therefore, the collision occurs in the PH-EU relative motion direction.In contrast, in our model, the collision occurs between the Luzon arc and RA in the NNW relative motion direction (Fig. 1).The Luzon arc, migrating NW with respect to EU, is bisected along this NNW direction, to the western part colliding and obducted over RA, and the eastern part subducted beneath RA (Fig. 1).The bisecting line migrates SW with respect to EU at a velocity of 5 cm/yr.Therefore, the S Ryukyu Trench propagates with this speed toward SW, but this does not imply that a trench is newly created, because the Ryukyu forearc is also migrating with the same speed.

IMPLICATIONS
In this model, because RA extends beneath the collisional wedge, there is no collision front between RA and EU.However, the subduction of the buoyant eastern half of the Luzon arc would have caused a northward bend of the S Ryukyu Trench, as seen in trenches where buoyant features are subducting.The reason why the subduction of the eastern Luzon arc was possible, not having caused collision, would be lubrication due to dehydration of the subducted serpentinized forearc.On the basis of the dehydration embrittlement hypothesis, the double seismic zone beneath NE Taiwan (Kao and Rau, 1999) implies dehydration from the westernmost part of the slab mantle (Yamasaki and Seno, 2003).Seno (2005) demonstrated that forearcs can be subducted when slab dehydration accompanies with it.The subducted slab has a westward velocity to EU.This is because the NNW bisecting line is oblique to the PH-EU motion (Fig. 1).The western edge of the subducted slab would, then, be obstructed by the mantle beneath the Eurasian continental margin.The Bouguer gravity anomaly shows a series of large negative anomalies of a wavelength of ~60 km in the Nanao Basin region, diminishing to the east (Hsu et al., 1998).This anomaly is a peculiar feature that is not seen in other part of the Ryukyu forearc.We regard this caused by buckling of the westernmost PH slab due to the obstruction mentioned above.This buckling is partly recognized in the seismic crustal structure (Wang et al., 2004).
The proposed model of the collision of the Luzon arc with the S Ryukyu forearc, actively moving southwestward, thus seems to resolve most of the enigmas cited above.The idea should be elaborated and be tested more by observed data in the future.

Figure 1 A
Figure 1 A schematic illustration of the collision of the Luzon arc with the migrating S Ryukyu forearc The NNW green line bisects the Luzon arc into the colliding western part and the subducted eastern part.Collision proceeds from the left to the right.