Comments on: End-of-2007 Hurricane-Global Warming Update http://cstpr.colorado.edu/prometheus/?p=4299 Wed, 29 Jul 2009 22:36:51 -0600 http://wordpress.org/?v=2.9.1 hourly 1 By: Harry Haymuss http://cstpr.colorado.edu/prometheus/?p=4299&cpage=1#comment-9301 Harry Haymuss Sat, 29 Dec 2007 12:00:14 +0000 http://sciencepolicy.colorado.edu/prometheusreborn/?p=4299#comment-9301 Trenberth calling someone else's action "shameful" is ludicrous. This is the same Trenberth, after all, who suddenly became a good soldier (as opposed to a good scientist) when Chris Landsea refused to toe the dogmatic line necessary to link hurricanes to global warming. See also this: http://www.ucar.edu/communications/quarterly/summer05/catarina.html "Although Catarina was later tagged by some as a possible sign of climate change, the waters over which it formed were actually slightly cooler than average. However, "the air was much colder than normal," says Dias. This produced the same type of intense upward heat flux that fuels hurricanes, normally seen in warmer waters." Trenberth calling someone else’s action “shameful” is ludicrous. This is the same Trenberth, after all, who suddenly became a good soldier (as opposed to a good scientist) when Chris Landsea refused to toe the dogmatic line necessary to link hurricanes to global warming. See also this:
http://www.ucar.edu/communications/quarterly/summer05/catarina.html

“Although Catarina was later tagged by some as a possible sign of climate change, the waters over which it formed were actually slightly cooler than average. However, “the air was much colder than normal,” says Dias. This produced the same type of intense upward heat flux that fuels hurricanes, normally seen in warmer waters.”

]]> By: mattstat http://cstpr.colorado.edu/prometheus/?p=4299&cpage=1#comment-9300 mattstat Sat, 29 Dec 2007 04:50:21 +0000 http://sciencepolicy.colorado.edu/prometheusreborn/?p=4299#comment-9300 Roger, The thing about Emanuel's 2005, and other similar, papers, is that they do some odd things with the hurricane statistics, which are not optimal, and which make it too easy to find trends that do not exist. For example, Emanuel adds up a function of wind speed over all storms within a year (storm 1 + storm 2 + ...) as a proxy for intensity. He then plots this sum year by year, over-plotting a 5-year running average through the points because he feels the actual data are too noisy. It is far better to try to model the distribution of storm intensity within a year. You then track this distribution from year to year, using a physically realistic probability model. This is a lot more complicated, and is what I attempted in my papers (I also model intensity as a three-dimensional entity: storm days, track length, and wind speed). The other common mistake is to run X-year running averages through measures like storm number. This leads to physically unrealistic estimates: for example, it can lead you to say something like, "there were 5.6 storms last year," which is, of course, impossible. You can have 5 storms or 6, but you can never have 5.6: the probability models you choose should reflect this impossibility, which they do not, in many papers. But the biggest mistake is to fail to provide measures of uncertainty. Even if you feel you can say that "Next year there will be 6.3 storms", this is of little use unless you can also say something like, "And there's a 90% chance that there will be between 1 and 10 storms." To say there will be exactly "6.3 storms" is to be overconfident to the highest degree. But many predictions, especially in "global warming" scenarios, are unfortunately like this. Also sorry to hear about Trenberth's unfortunate comment. Matt Briggs Roger,

The thing about Emanuel’s 2005, and other similar, papers, is that they do some odd things with the hurricane statistics, which are not optimal, and which make it too easy to find trends that do not exist.

For example, Emanuel adds up a function of wind speed over all storms within a year (storm 1 + storm 2 + …) as a proxy for intensity. He then plots this sum year by year, over-plotting a 5-year running average through the points because he feels the actual data are too noisy.

It is far better to try to model the distribution of storm intensity within a year. You then track this distribution from year to year, using a physically realistic probability model. This is a lot more complicated, and is what I attempted in my papers (I also model intensity as a three-dimensional entity: storm days, track length, and wind speed).

The other common mistake is to run X-year running averages through measures like storm number. This leads to physically unrealistic estimates: for example, it can lead you to say something like, “there were 5.6 storms last year,” which is, of course, impossible. You can have 5 storms or 6, but you can never have 5.6: the probability models you choose should reflect this impossibility, which they do not, in many papers.

But the biggest mistake is to fail to provide measures of uncertainty. Even if you feel you can say that “Next year there will be 6.3 storms”, this is of little use unless you can also say something like, “And there’s a 90% chance that there will be between 1 and 10 storms.”

To say there will be exactly “6.3 storms” is to be overconfident to the highest degree. But many predictions, especially in “global warming” scenarios, are unfortunately like this.

Also sorry to hear about Trenberth’s unfortunate comment.

Matt Briggs

]]>