As a coherent mode variability, the Atlantic multi-decadal oscillation/variability (AMO/AMV) plays an important role in modulating the regional and global climate. As the governing mechanisms of AMO is still under intense debate current, the definition of AMO actually is not universal either. The challenges to define AMO from observations rises from the combined effects of nature variability and changes of the external forcing (mainly the anthropogenic forcings). To isolate the effects from these two sources on North Atlantic sea surface temperature (SST), removal of the trend from external forcing becomes a necessary step. However, we found that the resulting AMO index and its regional and global influence heavily depends on the way how the external forcing induced trend is removed. A simple removal of the global or linear trend will inevitably alias some forced response into the resulting AMO index since the changes of the external forcing is not linear in time. Removal of the global mean temperature time series will partially remove the influence of AMO on global mean temperature too. Such as, the AMO can induce a peak-to-peak global mean temperature change at the order of 0.2°C. Here we propose that the best practice to define AMO is to remove the quadratic fit of the global mean temperature time series or via remove the trend from ensemble empirical mode decomposition (EEMD) analysis. Both latter methods provide a trend closely related to time evolution of the radiative forcing induced by changes in anthropogenic forcing. The resulting AMO index is not too different from the AMO index defined in more traditional ways (Such as Trenbeth and Shea, 2006), but the regional impact show significant changes in certain region.