sistencies likely result from the rather large biases in the climatology of the atmosphere and ocean in the tropical Pacific in current generation of global climate models (e.g., the double Intertropical Convergence Zone (ITCZ) common to many of the models). Whether or not the spatial, temporal and amplitude characteristics of ENSO change in a warmer world, however, the associated far-field impacts of ENSO will be different for several reasons. For example, the pattern and amplitude of the mid-latitude wintertime climate anomalies associated with ENSO will change as the mid-latitude jets shift poleward; places that experience drought during the warm (Australia) or cold (central United States) phase of ENSO and that are projected to dry with increasing global average temperature will experience enhanced drought conditions associated with identical ENSO cycles.
The literature on temperature (as well as that on precipitation) extremes clearly suggests that any method trying to link changes in extremes of maximum and/or minimum temperature, and their consequent effects on the number of very hot or very cold days, and on the duration of hot and cold spells, to global or even local average temperature changes would fall short. As a recent indication of this, a paper by Ballester et al. (2010) shows how an accurate scaling of temperature extremes would have to involve not only average temperature change under future scenarios, but also change in temperature variability and in the skewness of its distribution. No paper has addressed explicitly the quantitative differences across multiple scenarios (or stabilization targets) of temperature extremes that we could straightforwardly utilize to describe expected changes under different atmospheric concentrations.
We chose here to adopt the perspective offered by Battisti and Naylor (2009)—BN09 from now on—on the expected changes in the likelihood of experiencing extremely hot or unprecedented average summer temperatures. This way of characterizing warming from the perspective of the tail of the distribution of average temperature has the strength of utilizing model output (seasonal averages of surface temperature, or TAS) whose reliability has been more extensively corroborated than other parameters representing more traditional definitions of extreme or rare events (e.g., daily temperatures exceeding high thresholds). We adapt BN09’s analysis to our report’s focus on different stabilization targets and make it consistent with its reliance on pattern scaling in order to infer geographically detailed projections of temperature changes as a linear function of changes in global average temperature (see Methods, Section 4.5).