4. Some claims of trends in extreme events are insufficiently supported. In some cases trends are inferred and trend lines are drawn on figures when the data do not appear to justify it. Key issues are whether a given time series is long enough to infer or deduce a trend, whether the underlying data are of sufficient homogeneity to draw conclusions, and whether the trend is statistically significant. In the case of the latter, there are many instances where small changes in the start time for the time series produce changes in the magnitude of the trend that are probably non-significant. One notable example of this is the apparent difference in the trend for Atlantic tropical cyclones when the start time is 1880 versus 1900. In this case and others, such a difference may reflect a problem with the trend assessment technique rather than an actual signal.


In general, the word “trend” is used too loosely and often interchanged with “variation” or “increase.” These words should be associated with precise statistical definitions. Furthermore, when statements are made, the authors should indicate whether the claim is based on rigorous statistical analysis of a particular dataset (or datasets), expert elicitation, or the informed judgments of the authors.


5. The levels of uncertainty inherent in the trends should be discussed in more detail. The levels of uncertainty associated with trends (both observed and projected) in various types of extreme events should be elaborated upon. For an observed trend in a particular type of extreme event or variable, please discuss the underlying scientific and technical reasons for that uncertainty, and discuss its implications for projected trends in the extreme event or variable in question. When feasible, compare and contrast the issue for a particular variable to the underlying issues for other variables. As an example, consider that it is reasonable to assert that more is known about observed trends in temperature than trends in heavy precipitation, and even less is known about trends in the frequency and/or intensity of tropical cyclones. What are the technical reasons for this and what are the implications for projected trends?


6. Some cited material is not yet published. Some key literature cited in the report was not available to the peer review committee. In many cases the literature in question was cited as “to be submitted.” The committee understands that all literature cited and used as scientific evidence in SAP 3.3 should have reached at least “in press” status by August 2007. The authors of the SAP should ensure that all cited works are publicly available before the public release of the SAP. The committee further recommends that the authors use caution in drawing too heavily on papers and information that are not yet scientifically mature. The authors should minimize reliance on “grey literature” and non-refereed works.


7. The discussion of drought should be strengthened. The discussion of drought is not consistent among the chapters and sometimes contradictory (in terms of observed trends). In some cases apparent trends for particular geographic regions are used to make statements on broader geographic trends that are not justified. This discussion could be strengthened by including a figure for precipitation analogous to Figure 2 in Chapter 1. Notwithstanding the discussion of observed trends and projections, the background information on droughts should better define the different



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement