One of the greatest uncertainties in the forcing of climate change is the effect of anthropogenic aerosols on radiation and clouds. By serving as seeds for cloud droplets, aerosols not only enhance the brightness of clouds, but may also alter their longevity and, locally, the precipitation they produce. These effects on clouds are sometimes called 'aerosol indirect effects', in contrast to the direct effects on solar radiation. It has even been proposed that the effects of aerosols on clouds could be deliberately exploited to reduce the strength of hurricanes, by purposeful seeding of clouds1. Writing in Nature Geoscience, Dunstone et al.2 show, using model simulations, that aerosols also affect the frequency of tropical hurricanes over the North Atlantic by altering surface temperatures, and suggest that aerosol–cloud interactions are a main driver of this effect.

With the advent of ever more comprehensive representations of aerosol–cloud interactions in global climate models, the potential effects of anthropogenic aerosols on atmospheric circulation are increasingly being recognised. On the basis of climate model simulations, it has been suggested that the growth in aerosol levels over the North Atlantic in the second half of the twentieth century was responsible for the drying of the Sahel3, the weakening of the Indian Monsoon4 and changes in North Atlantic weather5 over the same period. In all cases, aerosol effects on clouds are thought to have been important. Statistical analysis of satellite data also reveals a link between dust outbreaks in the Sahara and a reduction in hurricane activity over the Atlantic6. Whether anthropogenic aerosols influence tropical storm activity has remained uncertain.

Dunstone et al.2 assess the influence of anthropogenic aerosols on tropical storm activity in the Atlantic over the twentieth century, using climate model simulations from the Fifth Coupled Model Intercomparison Project, an international effort to simulate historical and future climate change with the latest generation of Earth system models. Their analysis focuses on an Earth system model (HadGEM2-ES) that incorporates a comprehensive representation of aerosol–cloud interactions, and simulates a particularly strong indirect aerosol effect. Like all current-generation climate models, it does not resolve individual hurricanes. Instead, periods of frequent and infrequent storm activity are identified using indices for weather favouring hurricane development. According to Dunstone and colleagues' interpretation of the simulations2, tropical storms were less frequent during periods characterized by strong aerosol increases over the North Atlantic. The simulated changes in activity match those expected based on observations. Performing additional simulations in which anthropogenic aerosols were removed, they show that variations in anthropogenic aerosol levels were largely responsible for the simulated variations in tropical storm activity over the twentieth century, such that strong increases in anthropogenic aerosols before 1990 were associated with low levels of storm activity, and decreased aerosol levels after 1990 were associated with more intense activity (Fig. 1). Many of the other climate model simulations assessed qualitatively confirm this result.

Figure 1: Tropical storm over the Atlantic.
figure 1


Dunstone et al.2 analyse climate model simulations to show that a reduction in anthropogenic aerosol levels may have triggered a rise in tropical storm activity over the North Atlantic Ocean since the 1990s.

The explanation offered for the effect of aerosols on hurricanes is a change in atmospheric circulation over the tropical Atlantic Ocean. Specifically, Dunstone et al.2 show that high levels of anthropogenic aerosols lowered surface air temperatures — in large part because of aerosol-induced changes in clouds — over the North Atlantic relative to the Tropics, resulting in a southward shift of the Hadley circulation. In the late twentieth century, a relative northward shift of the Hadley circulation is seen, coincident with more hurricanes.

It would be premature to consider the proposed influence of anthropogenic aerosols on tropical storm activity in the North Atlantic textbook knowledge at this stage, however. Although very good in many aspects — including in its capacity to simulate variability in cyclone indices — the HadGEM2-ES climate model fails to realistically reproduce other key aspects of North Atlantic climate, such as upper-ocean heat content7. As such, Dunstone et al. are right to call for more modelling studies.

Dunstone et al.2 provide convincing evidence for a significant influence of anthropogenic aerosols on tropical storm activity over the North Atlantic in the twentieth century. Verification of their findings through future modelling efforts remains more a matter of academic interest, however: looking to the future, they show that warming resulting from greenhouse gas emissions may outweigh any influence of anthropogenic aerosol reductions on storm activity.