The question of whether tropical Atlantic storms will increase or decrease in frequency and intensity during the coming decades has received considerable scientific attention, primarily because of the enormous physical, emotional and financial damage wreaked by such events. The financial damage of tropical storms has increased to around US$26 billion per year on average and this is projected to double by the end of the century, even without any climatic changes1.

Those storms that make landfall over highly populated areas (such as Hurricane Katrina in 2005) cause far greater harm. Writing in Nature Climate Change, Villarini and Vecchi2 report that future emissions of greenhouse gases are unlikely to significantly change the number of tropical Atlantic storms. However, there are intriguing suggestions that different types of anthropogenic emission could affect the number of storms in subtly diverse ways.

Historical records show considerable variability in the number of Atlantic tropical storms from year to year and decade to decade, with no significant trend. For instance, 2005 was a very active season with 15 hurricanes, whereas 2009 had just three; 2010 was once again very active, but no landfalls were observed for the United States.

This variability is probably driven by both local Atlantic and remote factors, and also by random chaotic fluctuations in the weather. However, recent research indicates that the sea surface temperature (SST) of the tropical Atlantic relative to the rest of the tropical oceans is a key metric for determining the likely number of storms.

An important issue for making projections of tropical storms is the pattern of future SST changes — will the tropical Atlantic warm more or less than the rest of the tropical oceans? Identifying such indicators is important because these storms are relatively small features and are not fully resolved, even in state-of-the-art climate simulations. Instead, Villarini and Vecchi rely on an empirical relationship between these climate indicators and the number of storms.

Using the latest set of global climate simulations (the Coupled Model Intercomparison Project phase 5) to make projections of the indicators then allows them to infer the future number of tropical storms.

Villarini and Vecchi find that the magnitude of SST changes depends strongly on future greenhouse-gas emissions, but that the pattern of such changes at the end of the twenty-first century does not. This might suggest that the frequency of tropical storms is less likely to significantly change in the long term, but there is a wide spread in projections from different climate simulators and the sign of any change in tropical storm numbers is ambiguous. More positively, significant progress in reducing uncertainty in the long- term outlook, which is relevant for mitigating future damages, can be made by improving our physical understanding and simulations.

However, successful adaptation to future changes requires predictions for the near term, or next few decades. On these timescales, the natural variability of climate will dominate and we are unlikely to see a signal of change in the frequency of tropical storms emerge from the background levels of variability. This is important because a sequence of years with relatively high or low numbers of storms cannot necessarily be used as an indication of longer-term trends.

Furthermore, different climate processes are important for the number of storms in this near-term period, in particular because of emissions of anthropogenic sulphate aerosols. Both the pattern of the emissions themselves and the effects on climate are highly uncertain, but Villarini and Vecchi find indications that suggest reductions in aerosol emissions (particularly around the Atlantic Basin) may increase the relative SST and therefore the frequency of tropical storms. Ironically, the improvement of air quality that results from the reduction in aerosol emissions may come at the price of increased atmospheric warming and more frequent tropical storms.

How can we make progress on these issues? Improved understanding of the role of aerosols and the likelihood of changes in aerosol emissions would help to narrow uncertainty in projections considerably. Also, further understanding of the magnitude and causes of natural variability in tropical storms is essential. Furthermore, it may be possible to make skilful forecasts of SST variability — and hence of tropical storm numbers — a few years ahead.

However, arguably the most important required development is in higher- resolution simulations of the atmosphere and ocean, removing the need to rely on an empirical relationship between the number of storms and climate indicators.

It is now possible to more explicitly resolve tropical storms in global models of the atmosphere but the required computing power is too great to allow the long centennial simulations necessary to make longer- term projections. The insurance industry, in particular, has realized the importance of these developments and has started to support academic research to complement the scarce observations available so far with the synthetic storms produced by numerical simulation. This will allow us to more robustly quantify risk.

There are three primary features of tropical storms that affect the damage that they cause: frequency, intensity and where they make landfall. The study by Villarini and Vecchi suggests that the frequency may not change significantly, but says nothing about the other two factors. There is plenty of room for further research into understanding how the number of intense landfalling hurricanes will change in the future, which is what really matters to people, economies and ecosystems.