TNA/Enso Link Followup

[MWatkins]

Just a follow-up to this excellent discussion from Don Sutherland:

http://www.storm2k.org/phpbb2/viewtopic.php?t=57244

after the super bowl last night (I know, I know) I ran some multi-variate analysis against 4 predictors, all of which are available from the CPC.

The 4 Predictors used were:

1. TNA (Tropical North Atlantic Index)
2. ENSO Multivariate Index (A multivariate ENSO index)
3. PDO
4. NAO

Here's the methodology:

1. August through December values for each of the indices were averaged to represent the value used in the calculation. These indices were used to determine a link to the subsequent year's activity. For example...ENSO, TNA and other indices from 2002 were used to predict Named Storms, Hurricane landfalls and Major Hurricane landfalls for 2003 etc.

2. Regression analysis was run to determine if any of the variables in combination with other variables, or alone, showed a statistically significant correlation between Aug to Dec values (of those indices) and subsequent year hurricane activity.

There were a couple of conclusions...

1. Due to the relative infrequency of landfall data for Major hurricanes, I was unable to find a statistically significant (95% confidence) predictor for US landfalling hurricanes from any of the 4 indices.

2. Due to the relatively small number landfalling US hurricanes, I was unable to find a statistically significant predictor for US landfalling hurricanes from any of the 4 indices.

3. NAO and PDO, independently or in combination with any other variable, were not statistically significant predictors for any storm metric.

4. However, when combined together, TNA and the ENSO Multivariate index were very strong, statistically significant predictors of the next years number of Named storms. The key there is next years...not the current years which is already a known indicator.

According to the analysis...strong TNA values alone through the month of August through December suggest increased named storm activity the following season. When this is coupled with neutral to slightly negative (cool) ENSO conditions...the two make a very strong set of predictors.

I am working to model these results into a data set that would help create a statistical model for the upcoming year...if I'm repeating somebody else's research just let me know...again I wanted to look at this deeper based on Don's excellent discussion from a few days ago.

MW

[donsutherland1]

Great work, Mike!!!