Abstract: Generalized Pareto distribution (GPD) is commonly applied to frequency analysis of the extreme events in peak-over-threshold series (POTS). Previous studies suggested that the higher order probability weighted moment (HPWM) method is applicable to estimate the parameters of generalized extreme distribution and Pearson type-III distribution, but studies of its extension to GPD case are lacking. This paper derives a theoretical formula of HPWM tailored to GPD and presents an empirical method for estimation of its parameters. Results of our statistical experiments demonstrate that the parameters estimated using HPWM method of zeroth order and probability weighted moment method are nearly unbiased and both are more accurate than the conventional method of moment. This method was applied to the POTS (1946-2004) of the Yichang station on the Yangtze. The results show that this method generally gives a better fitting to the empirical distribution, especially for the upper tails.