When we published a blog post summarizing Colorado State’s (CSU) 2011 Atlantic hurricane season outlook Wednesday, readers expressed considerable skepticism about the credibility of such outlooks. So I thought it would be interesting to take a look at how accurate CSU’s April outlooks have been over the years. I found that most years they’ve been “in the ballpark” even if they haven’t demonstrated skill from a statistical standpoint.
Let’s begin by reviewing some of the comments in response to my Tuesday blog post on CSU outlook’s (which calls for a very active season with 16 named storms, nine hurricanes and five major hurricanes in 2011):
Every year these self promoters bring out the doom and gloom predictions of impending disaster. How about WAPO going over their predictions for the last ten years and seeing how accurate they’ve been.
Given how badly their 2010 predictions failed, I place no stock in the 2011 prediction whatsoever.
I’m now going to address these comments, point by point...
TravisBickle1 wants to see CSU’s predictions over the last 10 years and how accurate they’ve been. We can do even better, showing you the last 16 years...
|Year||Predicted storms||Observed storms|
The table indicates the following:
* In 12 out of 16 years, CSU has been within four storms of the actual number
* In 9 out of 16 years, CSU has been within three storms of the actual number
* CSU has only had “big misses” of more than 5 storms in three years (1995, 2005, 2006)
To me, the table above illustrates CSU has generally been in the ballpark in terms of the number of storms.
On the other hand, CSU has performed statistical analysis that demonstrates its April forecast “skill” offers minimal improvement over climatology. Nonetheless, the analysis finds that its forecast update made in early June (just as hurricane season is beginning) has shown considerable skill
In its own self-assessment last fall, CSU wrote:
...our forecasts are successful at forecasting whether the season will be more or less active than the average season by as early as December of the previous year. We tend to have improving skill as we get closer in time to the start of the hurricane season.
Did the 2010 forecast “fail” (per jsmith021961)?
Not really. Last April, CSU predicted an active season with 15 named storms, and it was very active, with 19 named storms. Four off isn’t bad. The public perception may have been it wasn’t an active season because very few of storms that developed impacted U.S. soil.
Does CSU predict “higher than usual” activity “every year”?
Well, as it turns out, it is true that CSU has yet to issue an April forecast for a below average number of named storms. The long-term average number of named storms is 10, and since 1995, they’ve never predicted fewer than that. But, for the most part, they’ve been right not too! Only twice since 1995 have there been fewer than 10 storms - in 1997 when there were 8 and in 2009 when there were 9. And both of those years, their forecast was within three storms of the actual number.
Also, there’s a legitimate scientific explanation for why they’ve predicted average to above average activity every since 1995. Beginning in 1995 (coincidentally), we entered a cyclical multi-decadal period of elevated hurricane activity, that is expected to continue for another 10-15 years. So we shouldn’t expect many outlooks for below average activity for quite some time...