Unforced Decadal-Scale Global Mean Warming and Cooling in a Hierarchy Of Models

The current global warming hiatus has been attributed in some studies to cooler eastern tropical Pacific sea surface temperatures (Kosaka and Xie 2013) and increased deep ocean heat uptake (Palmer et al. 2011, Balmaseda et al. 2013, Trenberth et al. 2013, Watanabe et al. 2013, England et al. 2014). While some models can reproduce hiatus periods with similar characteristics as the observed, these characteristics have not been shown to be a robust feature of multiple climate models, and the physical cause of this change in Pacific SST remains unclear.

In this study, we examine the statistics and spatial patterns of warming and cooling periods in a hierarchy of models, in order to explain which physical mechanisms can produce decadal-timescale trends in global mean temperature in a climate model. Examining how cooling trends occur and their contrast to warming trends in a variety of model configurations may provide insight on the mechanism of the current hiatus. The hierarchy consists of an aquaplanet version of CAM4 coupled to a slab ocean mixed layer, CAM4 with continents coupled to a slab ocean, the fully-coupled CCSM4 pre-industrial control (piControl) run, as well as a 5-member CCSM4 with rcp4.5 forcing. We analyze the occurrence and probability distribution to quantitatively compare the characteristics of hiatus periods among model runs. We also consider the spatial patterns of 10- and 15-yr hiatus periods and accelerated warming periods (defined as having trends lower than -0.16 C/period and higher than 0.16 C/period, respectively).

Hiatus periods are found on both 10- and 15-yr timescales in all model configurations, but with a range of likelihoods and statistical distributions. In the fully coupled CCSM4 piControl run, both hiatus and accelerated warming decades occur twice as often than the run with same forcing and a slab ocean. Coupling to a dynamical ocean also increases the spread in simulated trends, and 21st century forcing further increases this spread. It appears that as the hiatus period lengthens from 10 years to 15 years, the features of the hiatus depend less on the hierarchy, and the spatial patterns and frequencies become more similar.

There are similarities in spatial patterns among model setups, but revealing differences as well. For example, an eastern tropical pacific cold/warm tongue was present in all model configurations but was stronger on 10-yr timescales in fully coupled models, suggesting a contribution from the low-frequency tail of ENSO. A handful of individual hiatus periods in all model runs share a signal in either the north central Pacific extratropics or the southeastern Pacific extratropics and could be attributed to warming or cooling in the tropics. These results highlight the complexity of the relationship between forcing, atmosphere, and ocean dynamics that leads to hiatuses in surface temperature trends.