Paléosismologie Spatiale : segmentation et scénarios de ruptures sismiques. Applications à deux failles actives décrochantes asiatiques : la faille de Fuyun et la faille du Kunlun, Chine

To understand what are the parameters that control the slip distribution during large strike-slip earthquakes, it is necessary to document each earthquake in details. One key question is to understand how the geometry of the fault interacts with the rupture propagation during the earthquake, to initiate the rupture and possibly to stop it. To address such question, detailed information is needed about rupture geometry and slip distribution for as many as possible earthquakes, in order to try to decipher general rules, if they ever exist. Up to nowadays, the dataset of well-documented continental strike-slip earthquakes includes about 20 historical strike-slip earthquakes, for which accurate surface rupture maps are available. Displacement patterns for successive earthquakes on the same fault are even more rare. Hence, it remains difficult to assess which, if any, proposed earthquake recurrence models might work better in general, or on any given fault in particular. Using geo-referenced high-resolution optical satellite images (Quickbird satellite, pixel 70 cm) along with field observations, we first analysed the well-preserved August 11th, 1931, Fuyun strike-slip rupture (Ms 7.9), China. This event is one of the largest continental strike-slip earthquakes of the last hundred years. Thanks to the arid climate conditions that prevail in the Fuyun area, the analysis of these images allowed us to constrain the length and the geometry of the rupture zone. Detailed mapping revealed a linear right-lateral shear zone striking NNW that can be mapped for about 160 km, from the Kayirti River, in the north, to the Ulungur River, in the south. The rupture went bilateral through a series of jogs, both extensional and compressional, with the width varying from few hundred meters to about 2km. In the South, the rupture ended at the Almant Mountain, a 10-km-wide compressional jog. According to major changes in rupture geometry (jogs, fault branches and bends), the 1931 surface rupture can be divided into seven main geometric segments. Each of these first order segments could be further split into higher order segments, linked by geometric complexities at smaller scales. Our two targets in this work have been the Fuyun fault, affected by an earthquake in 1931 (Ms7.9), and the Kunlun fault affected by an earthquake in 2001 (Mw7.8). In both cases, we have been able to measure landform offsets that we can identify on the satellite images. These offsets are associated either to the last event or to older events. In that case they represent cumulative displacement. Along the Fuyun rupture, the 1931 earthquake slip distribution, defined by 290 offsets, shows an average slip of 6.3 m. The slip distribution is unusually flat over its entire length. Co-located larger offsets, interpreted as cumulative offsets, allow us to define 4 more events in addition to the most recent one. Although pre-1931 offsets remain to be dated, the slip distribution of each event seems to follow a similar flat pattern, supporting a characteristic earthquake model interpretation for the 5 last events. Exploration of several sections of the 2001 (Mw 7.8) Kunlun strike-slip surface rupture with similar methodology allowed us to identify cumulative offsets, in addition to the coseismic slip distribution, which is already well constrained by several other methods (Field observations, InSAR and SPOT images correlation). We observed that co-located larger offsets are very similar to the 2001 co-seismic offsets, supporting a characteristic-slip model. Slip variability along the 2001 rupture is significant and it seems to be correlated with the geometry of the surface rupture. Hence, unlike for the Fuyun event, in the case of the Kunlun earthquake it seems that the different fault segments involved in the 2001 rupture might not always break together. Instead, the fault segments could also break individually during smaller events, to accommodate over the long term the difference of slip between the different fault segments that form the Kunlun fault. Although it is difficult to differentiate between the slip- or the earthquake characteristic models, we propose that the behaviour of these two faults is organized and that they follow a characteristic pattern, when rupturing. The addition of some time new constrains in a near future (samples currently being processed) will help us to make progress in the understanding of the long term behaviour of these two faults.