Résumé
The 1905 M8 Tsetserleg-Bulnay earthquake sequence, which occurred 14 days apart in Mongolia, constitutes one of the major continental strike-slip earthquake sequence ever documented. Although the surface ruptures associated with these two events are well-preserved due to dry climate, they had only been mapped in details along short sections. Here, using sub-metric optical satellite images 'Pleiades' with ground resolution of 0.5 m, complemented by field observation, we mapped in details the entire surface ruptures and we extensively measure co-seismic and cumulative offsets to understand earthquake propagation mechanisms and rupture processes for the two events. Preliminary results show that 1) the NE-SW trending Tsetserleg rupture is characterized by an average 2.5 m left-lateral slip on its central segment, and by a number of second-order antithetic and dip-slip ruptures around its main trace, 2) the E-W trending Bulnay rupture shows 4-6 m of left-lateral slip with high complexity in rupture geometry on its western part, where partitioning can be often observed. In contrast, co-seismic slip increases up to 8-10 m along the eastern part with the rupture trace being highly localized. Interestingly, this change in rupture behavior seems to correlate with a major change in the local geology, 3) the NNW-SSE trending Teregtiyn rupture, one of the subsidiary segments of the Bulnay rupture, is characterized by about 4 m of right-lateral slip, 4) another subsidiary branch of the Bulnay rupture, the N-S trending Dungen rupture is composed of a series of en-echelon tension cracks with no visible through going rupture. The geometry of the cracks suggests right-lateral shear. We show that the apparent complexity of the rupture geometry for the Tsetserleg-Bulnay sequence is actually compatible with the regional stress field (NNE-SSW trending compression). Our detail mapping shows that the Tsetserlegrupture, despite its large size, did not connect with the main Bulnay fault. Similarly, no evidence for connection between these two faults during previous events could be found. Eventually, our preliminary results imply that the variation in rupture geometry and their related slip distribution seem to be controlled by pre-existing geologic structures, local surficial conditions, as well as dynamic rupture processes.