changes shown for the United States in Figure O.2 relative to the re-plots of Milly et al. (2005) results in USCCSP (2008b). In general, those results showed more of an east-west divide in runoff changes across the United States, with modest increases in the east, and decreases across most of the west, which were most severe in the Colorado River and Great Basins. We extracted from our multi-model suite the same 12 models analyzed by Milly et al. (from the approximately 20 that were available to us), and analyzed the results in the same way as did Milly et al. (2041-2060 minus 1900-1970 means). When so analyzed, the results were quite similar to those shown in USCCSP (2008a). The differences in patterns in our results as compared with USCCSP (2008b) and Milly et al. (2005) apparently come about because of (a) differences in the 1900-1970 based period (used by Milly et al.  versus 1971-2000 base period that we used, and (b) differences in the number of models considered (nearly double in our analysis relative to those available to Milly et al. (2005). As in Milly et al. (2005) and USCCSP (2008a), we have limited our analysis to annual runoff volumes.
Our overall conclusion therefore is that streamflow in many temperate river basins outside Eurasia will decrease as global temperature increases, with the greatest decreases in areas that are currently arid or semi-arid. Streamflow across most of the United States will decrease, although there is considerable disagreement among models aside from the Southwest, where most models project decreases, and Alaska, where most models project increases. Runoff sensitivities are approximately constant (per degree of global warming) in the range 1-2ºC warming. There is strong agreement among models that runoff in the Arctic and other high-latitude areas, including Alaska, will increase.
The impact of climate change on the frequency of large fires and extent of area burned by wildfires depends mainly on the type of vegetation (fuel) and the future weather and climate. In many regions that have been examined, the projected climate change will cause large changes in wildfire frequency and extent. In a broad sense, wildfires will increase in regions that are dominated by forests that are already prone to fire in the current climate, while warming (without a sufficient increase in precipitation) will cause a decrease in wildfires in some shrub and grassland (fuel limited) regions that are prone to fire in the present climate.
The probability of wildfire depends on the availability of fuel, the moisture content of the fuels, and the likelihood of ignition. Ignition of most large