On North American decadal climate for 2011-20

The predictability of North American climate is diagnosed by taking into account both forced climate change and natural decadal-scale climate variability over the next decade. In particular, the "signal" in North American surface air temperature and precipitation over 2011-20 associated with the expected change in boundary conditions related to future anthropogenic greenhouse gas (GHG) forcing, as well as the "noise" around that signal due to internally generated ocean-atmosphere variability, is estimated. The structural uncertainty in the estimate of decadal predictability is diagnosed by examining the sensitivity to plausible scenarios for the GHG-induced change in boundary forcing, the model dependency of the forced signals, and the dependency on methods for estimating internal decadal noise. The signal-to-noise analysis by the authors is thus different from other published decadal prediction studies, in that this study does not follow a trajectory from a particular initial state but rather considers the statistics of internal variability in comparison with the GHG signal. The 2011-20 decadal signal is characterized by surface warming over the entire North American continent, precipitation decreases over the contiguous United States, and precipitation increases over Canada relative to 1971-2000 climatological conditions. The signs of these forced responses are robust across different sea surface temperature (SST) scenarios and the different models employed, though the amplitude of the response differs. The North American decadal noise is considerably smaller than the signal associated with boundary forcing, implying a potential for high forecast skill for 2011-20 North American climate even for prediction methods that do not attempt to initialize climate models. However, the results do suggest that initialized decadal predictions, which seek to forecast externally forced signals and also constrain the internal variability, could potentially improve upon uninitialized methods in regions where the external signal is small relative to internal variability.