The Chelungpu fault ruptured during the September 1999 M w 7.6 Chi-Chi earthquake, in Central Taiwan. This event was characterized by coseismic displacements increasing along strike and updip, from south to north. Previous studies suggested that such lateral variations also existed in the long-term fault slip rate, but this has not yet been clearly documented. To address this, we investigate deformed fluvial terraces along the Choushui and Tatu-Wu rivers, in the southern and central segments of the thrust fault. Optical ages of~13 ka to~38 ka obtained for these terraces enable estimation of fault slip rates of 5.8 ± 2.0 and 10.3 ± 1.6/À3.0 mm/yr for the southern and central segments, respectively. We combine these findings with the fault slip rate determined by other authors for the northern segment. Statistical analysis of the data suggests lateral variations in the long-term fault slip rate, with values increasing toward the north. This pattern in the slip rate, averaged over approximately tens of thousand years, is similar to that observed for coseismic displacements during the Chi-Chi earthquake. The similarities in the deformation pattern observed for one earthquake or cumulated over several events suggest that the Chi-Chi earthquake could be characteristic of the earthquakes breaking the Chelungpu fault, with respect to slip distribution. Our results also allow for discussing the plausible evolution of major rivers draining the foothills of central Taiwan over the last~40-60 kyr.
IntroductionVariable coseismic slip along fault ruptures is a common observation [e.g., Vigny et al., 2011] and seems to follow generic characteristics that could be related to the physical properties of the fault zones [e.g., Manighetti et al., 2005]. For earthquakes along subduction zones, variable coseismic slip is often correlated to lateral variations in the interseismic coupling of the plate interface [e.g., Chlieh et al., 2008;Métois et al., 2012]. Such variability is probably reproduced over several seismic cycles because variable coseismic slip correlates to longer-term features, such as gravity anomalies [Song and Simons, 2003] or topography [Cubas et al., 2011] of the forearc. For continental faults, lateral slip variability has been observed but the consistency of this variability over different timescales has not been clearly and systematically documented on the same faults [e.g., Manighetti et al., 2001Manighetti et al., , 2005. Lateral variations in long-term slip rates [Tapponnier et al., 2001] and/or cumulative slip [Manighetti et al., 2001] have been reported and are usually attributed to the lateral long-term growth of faults. Continental faults in active mountain ranges, such as in the Himalayas [Ader et al., 2012] or in Taiwan [Ching et al., 2011;Hsu et al., 2009], seem to be fully locked during the interseismic period preceding an earthquake, even though laterally variable coseismic slip has been documented [Dominguez et al., 2003;Xu et al., 2009;Yu et al., 2001]. The question of why such variations ...