Deformation rates derived from GPS measurements made at two continuously operating stations at Leh (34.1°N, 77.6°E) and Hanle (32.7°N, 78.9°E), and eight campaign sites in the trans-Himalayan Ladakh spanning 11 years (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008), provide a clear picture of the kinematics of this region as well as the convergence rate across northwestern Himalaya. All the Ladakh sites move 32-34 mm/year NE in the ITRF2005 reference frame, and their relative velocities are 13-16 mm/year SW in the Indian reference frame and *19 mm/year W with reference to the Lhasa IGS station in southeastern Tibet. The results indicate that there is no statistically significant deformation in the 200-km stretch between the continuous sites Leh and Hanle as well as between Leh and Nubra valley sites along the Karakoram fault, whereas the sites in and around the splayed Karakoram fault region indicate surface deformation of 2.5 mm/year. Campaign sites along the Karakoram fault zone indicate a fault parallel surface motion of 1.4-2.5 mm/year in the Tangste and western Panamik segment of the Karakoram fault, which quantifies the best possible GPS-derived dextral slip rate of 3 mm/ year along this fault during this 11-year period. Baselines of Ladakh sites show convergence rates of 15-18 mm/year with respect to south India and 12-15 mm/year with respect to Delhi in north India and Almora in the Himalaya *400 km north-northeast of Delhi. These constitute an arc normal convergence of 12-15 mm/year across the western Himalaya, which is consistent with arc normal convergence all along the Himalayan arc from west to east. Baseline extension rates of 14-16 mm/year between Lhasa and Ladakh sites are consistent with the east-west extension rate of Tibetan Plateau.

SUMMARY Correlation of the coda of Empirical Green's functions (EGFs) from ambient noise can be used to reconstruct EGFs between two seismic stations deployed different periods of time. However, such method requires a number of source stations deployed in the area surrounding a pair of asynchronous stations, which limit its applicability in cases where there are not so many available source stations. Here, we propose an alternative method, called two-station C2 method, which uses one single station as a virtual source to retrieve surface wave phase velocities between a pair of asynchronous stations. Using ambient noise data from USArray as an example, we obtain the interstation C2 functions using our C2 method and the traditional cross-correlation functions (C1 functions). We compare the differences between the C1 and C2 functions in waveforms, dispersion measurements, and phase velocity maps. Our results show that our C2 method can obtain reliable interstation phase velocity measurements, which can be used in tomography to obtain reliable phase velocity maps. Our method can significantly improve ray path coverage from asynchronous seismic arrays and enhance the resolution in ambient noise tomography for areas between asynchronous seismic arrays.

<p><span>Ambient noise tomography (ANT) based on empirical Green’s functions (EGFs) retrieved from cross-correlation functions (CCFs) of ambient noise is widely used to construct shear-wave velocity structures. EGFs from ambient noise can be treated as virtual seismograms with one station working as a virtual source and the other station working as a receiver. We propose a method named two-station C</span><sup>2</sup><span> method(Rao et al., 2021), using one single station as a virtual source to obtain surface waves between a pair of asynchronous stations. This method can significantly improve ray path coverages and enhance the resolution in ANT for areas between asynchronous seismic arrays.</span></p><p><span>In our method, we select three stations, called a station triplet, which share the same great-circle path. We take one long-term station as a virtual source rather than using a number of stations as sources in the C</span><sup>3</sup><span> method(Stehly et al., 2008; Ma and Beroza, 2012; Spica et al., 2016; Zhang et al., 2019). We use data from the USArray to demonstrate the feasibility of our method in retrieving surface waves from asynchronous stations.</span></p><p><span>Due to the harsh environment and inaccessibility of most of parts of the plateau, it is nearly impossible to deploy a large-scale synchronous seismic array across Tibet. In the past few decades, several isolated arrays have been deployed in Tibet at different periods of time. ANT has been applied to Tibet to generate phase velocity maps using these seismic arrays (e.g., Yang et al.,2012;Xie et al.,2013; Shen et al., 2016). However, due to the fact that these seismic arrays were not deployed synchronously, inter-array paths between asynchronous arrays cannot be obtained from the traditional C</span><sup>1</sup><span> method, resulting in low resolution in the gaps of these seismic arrays.</span></p><p><span>We applied our method to the two seismic arrays (Z4 and X4) deployed in NE Tibet. The Z4 array was deployed from July 2007 to July 2008 and X4 from September 2008 to September 2009. For these two arrays, if we follow the C</span><sup>1</sup><span> method, we can get at most 153 paths for Z4 array and 300 paths for X4 array. But no crossing-array paths can be obtained. Fortunately, there is a permanent Chinese National Seismic Network (Zheng et al., 2010) deployed across China. We can take the permanent stations from the Chinese National Seismic Network as source stations and obtain C</span><sup>2 </sup><span>functions following our method. Here, to illustrate the application of our C</span><sup>2</sup><span> method for these two arrays, we select 153 permanent stations from the Chinese National Network as virtual sources. And, using these stations and our C</span><sup>2</sup><span> method for these two arrays, we can retrieve 413 C</span><sup>2</sup><span> functions with the source stations located within 5 degrees of the great-circle paths of receiver station pairs. The path coverage is improved by over 91%. Combining C</span><sup>1</sup><span> and C</span><sup>2</sup><span> paths, we can much better image the structures between these two arrays.</span></p>