Under the background of climate change and human activities, the non-stationary characteristics caused by the trend of hydrometeorological extreme value series in the the Yarlung Zangbo River basin has imposed impat on the time-varying evolution characteristics of risk of extremes. This paper constructs a dynamic generalized extreme value distribution (GEV) model by taking the physical driving factors related to human activities and climate change as the covariates of model parameters. Taking the extreme daily high temperature and low temperature series of the the Yarlung Zangbo River basin as the research object, this paper quantitativelyevaluates the impact of human activities and climate change factors represented by urbanization on the risk of extreme high temperature and low temperature events in the basin. Based on a phase-wise model optimization strategy, the optimization process of the optimal non-stationary GEV model is greatly simplified. This article proposes a slope comparison method based on linear regression to achieve attribution and stripping of the response relationship between human activities, climate change, and extreme temperature event risk. The analysis results show that the extreme high temperature and extreme low temperature sequences in Yarlung Zangbo River show a significant upward trend of 5%. The large-scale climate oscillation factor is a significant driving factor for the nonstationary characteristics of the high temperature series at Yarlung Zangbo River, rather than the non-stationary characteristics of the extreme low temperature series at both stations. Human activities represented by urbanization have had a certain weakening effect on the risk of extreme high temperature events at Yarlung Zangbo River. Climate change related factors, especially global warming and local warming effects, are the main driving factors for the continuous increase in extreme high temperature risks.