Environmental variables for predicting hurricanes?
I'm trying to predict hurricane count in the Caribbean using a poisson regression, and I am following a book by Elsner and Jagger which gives details on how to do so using R.
Their book is concerned with hurricane landfalls in the US. However, since my area of interest is further south of this, I'm wondering if the environmental variables described in the book still apply.
The variables are:
1) Atlantic sea surface temperature (SST) derived from Atlantic Multidecadal Oscillation as an indicator of oceanic heat content:
https://www.esrl.noaa.gov/psd/data/correlation/amon.us.long.mean.data
2) Southern Oscillation Index (SOI) related to El Nino-Southern Oscillation (ENSO) as indicator of wind shear:
https://climatedataguide.ucar.edu/sites/default/files/SOI.signal.annstd.txt
3) Northern Atlantic Oscillation (NAO) as an indicator of steering flow:
https://crudata.uea.ac.uk/cru/data/nao/nao.dat
4) Sunspot number (SSN) as an indicator of upper air temperature:
https://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/Data/sunspot.long.data
Now, I ran my poisson regression model and found that NAO and SSN were statistically insignificant (p-values were greater than 0.4).
SST was significant with p-value less than 0.01 and SOI was significant with p-value less than 0.05.
Are these environmental variables suitable? They seem to deal with the North Atlantic and I'm therefore not sure if they would affect the southern Caribbean as much as the US.
The specific area of interest is the Windward Islands (Dominica, Martinique, Guadeloupe, St. Vincent and the Grenadines, Grenada and Trinidad and Tobago).