Abstract:By use of the monthly maximum wind sequences of 70 years and 20 years from 86 national meteorological observation stations in Guangdong Province, the threeparameter Generalized Pareto Distribution (GPD) function is used to calculate the returnperiod wind speeds of 86 national meteorological observation stations, respectively. The four parameter estimation methods such as Moment Estimation (MOM), Maximum Likelihood Estimation (MLE), Likelihood Moment Estimation (LM) and Probability and Weight Moment Estimation (PWM) are used based on the PeaksOverThreshold (POT) sampling. The characterization parameter estimation indicators such as rootmeansquare error (RMSE), fitting relative deviation (〖WTBX〗R〖WTBZ〗f) and Komogorov Test with a reliability of 0.05 (〖WTBX〗K〖WTBZ〗f) are used to test the fitting effect. The results show that Probability Weight Moment Estimation (PWM) based on POT has the best fitting effect.