Is the Skewness of the Distribution of Temperature Increasing?

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 4 February 2014: 9:00 AM
Room C205 (The Georgia World Congress Center )
Richard W. Katz, NCAR, Boulder, CO

Several researchers have recently claimed that the skewness of the probability distribution of temperature is increasing. If real, such an increase in asymmetry of the distribution would have significant implications for future temperature extremes. Yet analyses based on extreme value methods have not typically found trends in extreme temperature consistent with an increase in asymmetry.

These skewness calculations have been performed by dividing the historical record into subperiods consisting of one or more decades and, in effect, assuming stationarity within the subperiod. The fact that the mean (and possibly the standard deviation) of the probability distribution is generally increasing, even within a subperiod, makes this calculation problematic.

For a hypothetical statistical model in which only the mean (and in some cases the standard deviation) is increasing (i.e., no actual increase in skewness), it is shown that the present method of calculation can produce an erroneous apparent increase in skewness. Hence, whether the skewness of the distribution of temperature is really increasing remains an open question.