12/14/2011

Hurricane forecasts apparently no better than a random guess

This is almost too hard for me to believe. You would have to think we know enough to narrow down the range of possibilities a little bit. But if we really can't forecast the number of hurricanes a few months in advance, what about longer term forecasts? Philip J. Klotzbach and William M. Gray write this:

We have suspended issuing quantitative forecasts at this extended-range lead time, since they have not proved skillful over the last 20 years. We attribute the primary reasons for the lack of skill of our early December forecast due to the breakdown of several long-term relationships that worked well in many years of hindcast data, but not in real-time forecasting.

We would never have issued a seasonal hurricane forecast that did not show significant skill on many years of hindcast data. In addition, no statistical or dynamical models have shown skill at predicting El Niño – Southern Oscillation (ENSO) at this extended forecast lead time of 9-12 months. . . .

Labels:

3 Comments:

Blogger Martin G. Schalz said...

Chaos theory at it's finest...

The sheer amount of data that would have to be collected in order to accurately predict huricanes is way beyond huge. Not to mention simply trying to process the data in an attempt to predict future outcomes based upon previous observations. 'Tis more than enough to give the fastest Cray machine the equivalent of a digital migrane.

Cray computers are, and have been used for weather prediction, but they are limited in long term accuracy.

12/14/2011 5:42 PM  
OpenID DougHuffman said...

N. N. Taleb calls reality fractally complex in his development of the Ludic fallacy at the root of the failure of induction. We whitewash with normalcy and gamble on constancy to our doom. The Black Swan: The Impact of the Highly Improbable.

12/15/2011 6:05 AM  
OpenID DougHuffman said...

Coincidental with the burgeoning realization of complexity, comes this announcement of Maximal Information-based Nonparametric Exploration data mining software.

Reshef, DN et al., Detecting novel associations in large data sets, Science, DOI: 10.1126/Science1205438

12/16/2011 6:36 AM  

Post a Comment

Links to this post:

Create a Link

<< Home