predict dramatic dying in parts of sub-Saharan Africa, but others do not. Patterns of precipitation changes in middle and higher latitudes are more robust on large scales but remain uncertain on the regional scales of importance for impact and adaptation strategies. For example, the changes in precipitation in the western United States in wintertime are sensitive to the El Niño phenomenon in the equatorial Pacific, the response of which is itself uncertain. And especially over semi-arid land surfaces, arguments based on the assumption that relative humidities do not change can break down. Given these complexities, we consider the CMIP3 ensemble as providing our best estimates of the pattern of precipitation change accompanying global warming, with consistency or inconsistency across this ensemble provided some indication of uncertainties.
Atlantic hurricane activity has increased markedly in the past 20 years, concurrent with increases in ocean temperatures in the tropical North Atlantic region in which hurricanes are formed. But it is well established that hurricane frequency is dependent on other aspects of the large-scale climate as well as the local ocean temperatures. Many of the issues regarding detection, attribution, and projection of tropical cyclone trends in the Atlantic and throughout the tropics have recently been assessed by a WMO expert team (Knutson et al., 2010). We highlight a few of these issues here, basing the discussion in large part on this recent assessment, and refer the reader to Knutson et al. (2010) for a more detailed discussion.
Several detailed studies of the tropical storm record in the Atlantic have converged on a consensus view that the recent trend is more likely due to internal multi-decadal variability than a part of a century-long trend associated with greenhouse gas increases. While the raw data for Atlantic storms suggest a significant century-long trend, three separate lines of analysis cast doubts on the reality of this trend: (1) the frequency of landfalling storms, for which the completeness of the data is less of an issue than for basin-wide statistics, shows no significant long-term trend (Landsea, 2007); (2) estimates based on historical ship tracks indicate enough storms were likely missed in the early part of the century to account for most of the long-term trend in storm frequency (Chang et al., 2007; Vecchi and Knutson, 2008); and (3) the long-term trend in the raw data is primarily from short-lived storms (<3 days), which also suggests data artifacts are dominating the long-term