Maps similar to those in Figure 3 of BN09 will result from each shift in mean temperature, except they will correspond to different atmospheric concentrations of CO2 (and their implied expected global average warming) rather than to different future periods. The main difference in our approach compared to the analysis in BN09 is the choice of shifting the distribution uniformly to the right rather than trying to build a new distribution of average temperature anomalies on the basis of model output. This is a choice dictated by the use of pattern scaling, which lacks the availability of an ensemble of fully coupled model runs for each CO2 target.
It could be argued that our analysis represents a conservative estimate of the expected changes in extreme seasonal temperatures, because we are assuming no change in the climatological distributions of temperatures besides a shift of their central locations. In particular, no change in the variability of seasonal average temperature is taken into consideration. Some studies (see for example Scherrer et al., 2005; Fischer and Schar, 2009 for the European region; Giorgi and Bi, 2005; Kitoh and Mukano, 2009 for global patterns) have shown that future variability of summer temperature is projected to increase, in association with drying.
In Figures 4.8 and 4.9 we show the resulting likelihoods of exceeding the 95th percentile, or the warmest anomaly of current average JJA and DJF temperatures (1971-2000 of 20C3M simulations) for the three levels of global warming. The patterns become redder (higher likelihood) as we look down each column (larger global average warming implies greater chances of exceeding the thresholds) and bluer (smaller likelihood) as we look across rows (higher thresholds make exceedances rarer).
A general increase in atmospheric water vapor is predicted by essentially all climate models as temperatures increase, and an upward trend in column integrated water vapor has been observed in many regions (Trenberth et al., 2005). The consequences of this increase for the distribution of mean precipitation has been discussed in Section 4.2. As outlined below, models and simple theories also suggest that this increase in water vapor will increase the intensity of heavy rainfall events. Increasing trends in extreme precipitation have been documented in many regions, including much of North America (USCCSP, 2008c), but evidence for associated increases in floods is not compelling to date (Lins and Slack, 1999, 2005; WWAP, 2008).
As articulated by Trenberth (1999) and many others, precipitation in storms is related mostly to atmospheric moisture convergence rather than