Some 2017 Hurricane Odds

This season has been eventful, to say the least. Even though we’re still reeling from Harvey and Irma, and we’re nowhere close to assessing the damage, it’s interesting to step back and estimate the probabilities for some of 2017’s striking features. I’ve analyzed 100,000s of years of simulations from my stochastic tropical cyclone model (Hall and Yonekura, J Climate, 2013) to calculate probabilities for the following 2017 events:

  1. Three or more major hurricanes formed, with at least two reaching category 4 or higher (cat4+) (Harvey, Irma, Jose).
  2. Two or more cat4+ hurricanes spinning in the Atlantic simultaneously (Irma, Jose).
  3. Two or more cat4+ hurricane landfalls on the US within 15 days (Harvey, Irma).

I’m interested in the impact of climate variability on hurricane activity, and the Atlantic has been warm this season, so I’ve done the calculations for five values of seasonal-mean North Atlantic subtropical sea-surface temperature (SST): -2, -1, 0, +1, and +2, (units of standard deviation below and above the 1944-2015). Here are the annual probabilities, along with 1944-2015 historical counts from archives (HURDAT):

SST formation simultaneous landfall
+2 0.625 0.104 0.012
+1 0.462 0.061 0.009
0 0.288 0.034 0.006
-1 0.183 0.019 0.003
-2 0.108 0.010 0.002
HURDAT 21/72 1/72 0/72

It’s not too hard to get three majors with two Cat4+ in a season. Historically, it’s happened in 21 seasons in the 72-season period 1944-2015, in close agreement with the model’s SST-neutral annual probability of 0.288. The model’s warm-SST probability is much higher, 0.625. It’s much harder to get two Cat4+ storms spinning simultaneously, and it’s only happened once in the 1944-2015 period (2010). The model predicts an annual probability of only 0.034 for the SST-neutral state, but a higher 0.104 for the warm state. Getting two Cat4+ landfalls within 15 days is exceedingly rare. It hasn’t happened historically, and the model’s neutral-SST predicts a 0.006 annual probability. However, while still rare in the model’s warm prediction, at 0.012 it’s twice as likely.

So far in the 2017 season the subtropical North Atlantic is about 1.6 standard deviations above the long-term mean, and the model at that SST value has about 0.011 odds (interpolated between +1 and +2 SST) for two Cat4+ landfalls within 15 days. Apparently, the devastating pair of landfalls in 2017 is a case of elevated odds due to a warm Atlantic, combined with 1-in-90 bad luck.

Dr Hall is a climate scientist at the NASA Goddard Institute for Space Studies in New York City. He specialises in hurricanes and their relationship to climate.  He develops statistical hurricane and weather hazard models, and has worked as a consultant to RMS and reinsurance companies. Dr Hall can be reached at

Posted: Monday, September 11th, 2017