Being able to explain, predict, and understand changes in tropical storm activity is of great societal importance in terms of economic and social impact, and has been studied intensely by scientists around the globe. There is a great variation in tropical storms on different time scales, varying from intra-seasonal to multi-decadal, and a great deal of argument about whether tropical storm frequency and intensity are sensitive to climate change. Villarini et al. (2010) examined the empirical understanding between tropical storm frequency and large-scale climate conditions by examining the climate indices that tropical storms are often associated with. The group of scientists modeled not only the North Atlantic climate basin, which is the typical target of studies, but also expanded their analysis to include all tropical storm activity that lasted longer than two days and was recorded as U.S. landfall events. Only tropical storms that last longer than two days were recorded since shorter storms are likely to produce negative results. The authors found that it would be best to use a family of models with Atlantic and tropical storms as covariates. —Brian Nadler
Villarini, G., Vecchi, G.A., and Smith, J.A., 2010. Modeling the dependence of tropical storm counts in the North Atlantic basin on climate indices. Monthly Weather Review 138, 2681–2705.
Villarini and colleagues used a statistical approach to examine the relationship between tropical storm count and climate indices, expanding on further studies by also including tropical storms recorded as U.S. landfall events, rather than only covering the North Atlantic basin. A Poisson distribution model was also utilized to examine the dependence of counts on climate indices, accounting for over-dispersion or under-dispersion of tropical storm counts. The model is able to predict inter-annual variability, however, another model will be necessary to examine decadal variability as well. For all the models, Atlantic and tropical sea surface temperatures (SST<!–[if supportFields]>XE “sea surface temperature (SST)”<![endif]–><!–[if supportFields]><![endif]–>) are retained as significant covariates, supporting an idea proposed by Vecchi that the increases or decreases in Atlantic SST are preferable to the values of tropical SST in predicting tropical storm count in the U.S. land areas and the North Atlantic basin. The Poisson model of distribution was determined to be the best method for evaluating the data. The scientists suggest running a further experiment modeling U.S. landfall count with the overall storm count for the North Atlantic in order to get a wider swath of data that would be much more accurate in predicting tropical storm changes and reduce anomalies in data.
Expanding such studies will allow for a better understanding of the ways that we can further predict tropical storm variability, along with patterns and variation.