Flood Frequency Analysis (FFA) is a popular approach providing important information about the frequency of future extreme values. Employing this information can be very effective for better planning, design, and operation of flood protection defenses, which leads to more reliable watershed management. The present study, considering Haji-Ghoshan, Ghouchmaz, and Ghareh-Shour hydrometric stations of the Gorgan Roud River basin as the study area, examined the behavior of six different GPD parameter estimators using Monte-Carlo simulations, compares three automated threshold selection techniques based on POT and finally evaluates the performance of each threshold selection methods using NRMSE and relative index of agreement. The results indicate that in all stations the SE method provided smaller thresholds than other threshold selection techniques. Short tail distribution is provided for the normality of differences method in in Haji-Ghoshan and Ghouchmaz stations. The SE, MAX PV, and Cramer-von Mises/Kolmogorov-Smirnov methods provided the optimum thresholds based on\({r_d}\)and NRMSE in Haji-Ghoshan, Ghouchmaz and Ghareh-Shour stations, respectively.