Non-climatic discontinuities in the climate record should be detected in the climate change studies. In this paper, non-parameter and parameter methods are used to identify the inhomogeneity in extreme value series. It is found that the two phase generalized linear regression method that explicitly considers probability distribution of extreme values has higher power of detection than the Wilcoxon rank-sum test and two-phase linear regression method, and the more the extreme values are used, the higher the power of detection on the identification of single changepoint case. Its applicability is also examined in the multiple changepoints case.