Reply to Comments on “Multiyear Predictions of North Atlantic Hurricane Frequency: Promise and Limitations”

Abstract

In our original paper (Vecchi et al. 2013, hereafter V13),westated‘‘theskillintheinitializedforecastscomesinlargepart from the persistence of the mid-1990s shift by the ini-tializedforecasts,ratherthanfrompredictingitsevolution.’’Smith et al. (2013, hereafter S13) challenge that assertion,contending that the Met Office Decadal Prediction System(DePreSys) was able to make a successful retrospectiveforecastofthatshift.Westandbyouroriginalassertionandpresent additional analyses using output from DePreSysretrospective forecasts to support our assessment.S13 argue that the physical basis for multiyear hurri-cane prediction comes from (i) multiyear predictabilityofNorthAtlanticsubpolargyretemperature(SGT)and(ii) the impact of SGT changes on hurricane activity,thereby creating skillful multiyear predictions for hurri-canes. We agree that skillful multiyear predictions forNorthAtlanticSGTshavebeenmadewithDePreSysandother systems (e.g., Robson et al. 2012; Yeager et al.2012),includingtheGFDLpredictionsystem(Yangetal.2013; R. Msadek et al. 2013, unpublished manuscript).We also agree that there are physically plausible hypoth-eses by which SGT changes could influence hurricane ac-tivity (e.g., SuttonandHodson2005; Zhang andDelworth2006; Kang et al.2008; Dunstoneet al. 2011), although themagnitude of this influence is not clear.However, we disagree that this multiyear predictiveskill for SGTs has in fact produced skillful multiyearretrospective forecasts of hurricane activity, apart fromtheeffectofthepersistenceoftheshift.Hereweprovideadditional support for our assessment through the anal-ysis of retrospective forecasts of hurricane activity fromDePreSys (Smith et al. 2010), in addition to the workoriginally presented in V13. Output from the DePreSysretrospective forecasts was kindly provided by DougSmith.ThedatausedinthepresentreplyandinS13differfrom those in the original V13, with V13 using an SST-based index as a proxy for hurricane activity and S13 andthe present study counting low pressure systems in a dy-namical model. We have conducted analyses using theDePreSys hurricane retrospective forecasts for lead years1–5, 2–6, and 3–7; all leads yield similar results. In Fig. 1awe plot the number of forecast hurricanes for lead years1–5 (vertical axis) versus the number of observed hurri-canes (horizontal axis; 5-yr running mean centered onretrospective forecast year 3), with one symbol for eachinitializationyear.Asetofperfectretrospectiveforecastswould form a straight line with a 1:1 slope (i.e., for eachyear the observed and retrospective forecast hurricanecounts would be identical). Retrospective forecasts ini-tialized in the ‘‘preshift period’’ (blue circles) show anessentially random scatter, with no significant correlationbetween observed and forecast counts (the correlation is0.07). This indicates no skill.For the postshiftperiod (redcircles), the correlation is 20.47,againindicating noskill.Thus,fortheperiodsbeforeandaftertheshift,thereisnodemonstrated skill in multiyear retrospective forecasts ofhurricane counts, as was argued in V13 using a differenthurricane prediction methodology.