Potential evapotranspiration (PET) is a key factor involved in regional and global Earth system processes. This is particularly important for Arctic region, which faces amplified warming and is vulnerable to climate change. In this study, we estimate the PET during 1975-2014 at 190 meteorological stations in Siberian river basins using three frequently used models including a combination model (Penman–Monteith, PM), a radiation-based model (Priestley–Taylor, PT), and a temperature-based model (Hamon, H). Based on the estimated PET, the spatiotemporal variations, and differences among the three models are analyzed, and the PET changes are attributed to the changes in meteorological drivers. The results show that the annual values estimated by PM and PT models are close and are much larger than that estimated by H model. The trends of PET from PM model are generally insignificant while these from H model exhibit widespread increase in much of the Yenisey and Lena river basins. The contributions of changes in air temperature and net radiation to PET changes are less than 20% at most of stations, while the increasing VPD and decreasing wind speed contribute more than 80% to the changes of PET. This finding highlights the importance of VPD and wind speed in PET estimates in the Siberian river basins. Using different PET models can substantially affect estimates of aridity index. It further indicates that cautions must be taken when selecting PET model for hydrological and climate change studies in the arctic region.