Coseismic surface rupture is the permanent deformation on the surface in the middle and upper parts of the earth’s crust during a strong earthquake occurred. It indicates to a certain extent the deep structural characteristics of the active faults and the pre-existing tectonic environment , so complete extraction of characteristic of coseismic surface rupture has an important influence on the understanding of geometry and structure of the active faults.

On November 14, 2001, a West Pass of Kunlunshan earthquake with magnitude of M_S8.1 occurred at Hoh Xil of no man’s land in Golmud City, Qinghai Province, northern Qinghai-Xizang Plateau. The instrumental epicenter was located near the Kunlun mountains at the junction of Xinjiang Uygur Autonomous Region and Qinghai Province within higher local altitude. The strong quake destroyed the Qinghai-Xizang Railway under construction and Qinghai-Xizang Highway connecting Xining to Lhasa, causing major economic losses to the country's lifeline projects. Due to the hypoxia, harsh natural environment, and inconvenient transportation of the epicenter area which difficult to conduct extensive field surveys. It is necessary to use high-resolution satellite archived images to study the time-varying characteristics of coseismic surface rupture induced by recent strong earthquakes, which has important reference significance for conducting research on historical strong earthquake events. This paper uses multi-temporal high-resolution satellite images （with resolution of 1−6 m） covering the coseismic surface rupture zone of the 2001 M_S8.1 mainshock in the west of Kunlun Pass as the main research data to discuss the changes of coseismic surface rupture zone at different segments under natural surface process and check the reasons for its changes.

We select USGS KeyHole images with a spatial resolution of 6 m as pre-earthquake images, and the multi-period image data within 20 years post- earthquake from Google Earth （1 m）, ALOS （10 m）, Chinese GF-2 （1 m）, and GF-7 （0.6 m）. We use a comprehensive method of machine-assisted visual interpretation to extract the distribution of the coseismic rupture zone. The research results show that: ① A nearly 400 km linear coseismic surface rupture zone was formed during the 2001 earthquake, no pre-existing rupture remains were found in the relatively shorter western section before the earthquake, while the fresh coseismic rupture was reoccurrence along pre-existing earthquake ruptures at the longer eastern section. ② In 2003, two years after the earthquake, the coseismic surface ruptures locating at lake surface, crossing river valleys, gullies and alluvial fans which been checked during the earthquake field survey were rapidly affected by seasonal water flow and temperature （lake ice melt） changes. In 2011, 10 years after the earthquake, the surface rupture segments gradually showed a significant fragmented distribution on the images. Usually, continuous coseismic surface rupture zones can only be identified on high-level terraces or locations far away from active alluvial fans; in 2021, 20 years after the earthquake, coseismic surface ruptures on low-level terraces are almost impossible to identify, while the rupture zones on high-level terraces are relatively intact. The total length of identifiable rupture zones sharply reduced to 64%. More than 500 coseismic rupture sections had disappeared from the images during the past 20 years. ③ The measurement of coseismic horizontal dislocation based on GF image 20 years after the earthquake shows that the distribution of horizontal dislocation still has multiple peaks with maximum measured horizontal coseismic slip of 8.6 m, which is consistent with previous field survey but slightly smaller, and this proved that high quality satellite images can be used to combine the discontinuous coseismic rupture zone to together the length.

The identification of whole coseismic surface rupture zones on images is mainly affected by the surface processes. It is also affected by the spatial resolution of the images and their acquisition time. The identification of coseismic surface rupture of strong earthquakes can rely on emergency scientific surveys and satellite and aerial image data shortly after the earthquakes. However, the identification of surface rupture zones in uninhabited areas and historical strong earthquakes mainly relies on long-elapsed time and satellite imagery. Through residual rupture zone information on older landforms, it can recover more complete surface rupture zones than can be directly observed at the surface ground. Using satellite remote sensing images, scholars can restore the rupture scale of paleo-earthquakes and historical earthquakes, the surface rupture zone can be traced back through the residual rupture traces on older landforms, which is helpful in estimating the total length of the surface rupture zone, its magnitude, etc. Our research shows that the transformation process of coseismic surface ruptures induced by strong earthquakes can be understood in detail using multi-temporal high-spatial resolution images, which provides a methodological reference for exploring a large number of active faults within the Qinghai-Xizang Plateau to assess their earthquake risk in the future.