It's not rocket science although as usual the biases introduced by mans assumptions are often harder to model than
simple observational trends, they just create noise around definite natural trends.
While the statistics of 30 - year (or longer) NAO
trends and associated surface climate impacts can not be reliably determined from the short
observational record, we have made use of a
simple relationship between the statistics of
trends of any length and the statistics of the interannual variability, provided the time series is Gaussian (Thompson et al. 2015).