In the presentation, I summarized the development of my linear joint inversion approaches using hr-GPS, teleseismic body and surface waves, regional seismic, campaign GPS, InSAR and tsunami datasets, to investigate the kinematic rupture patterns of large earthquakes. In several different studies of rupture models of the 2011 Tohoku earthquake, 2012 Costa Rica earthquake, 2013 Craig earthquake, 2010 Mentawai earthquake, 2010 Chile earthquake and 2014 Iquique earthquake, I adopted each available dataset progressively in my joint inversion algorithm. As noted in this presentation, the teleseismic datasets provide good temporal resolution of the rupture process, while geodetic datasets have good spatial resolution. Tsunami datasets have good spatial resolution of slip near the trench. The joint inversion combines the advantage of each dataset, yielding stable and high resolution rupture models with detailed spatial and temporal information. Resolving a robust and detailed rupture model helps us to understand co-seismic rupture properties, such as depth dependent energy release patterns, super-shear rupture, and tsunami excitation. Comparing the inter-seismic locking pattern and post-seismic stress release pattern with the co-seismic rupture model helps to investigate the locking and releasing behavior of the fault plane through the earthquake cycle, the stress release level of large earthquakes and the relationship between the main shock ruptures, aftershocks and non-seismogenic deformation.