The annual frequency of Atlantic hurricanes—particularly of major (category 3–5) hurricanes—exhibits coherent interdecadal variability (Fig. 1a, b). Hurricane seasons during the period from the late 1960s to the mid-1990s were comparatively quiescent; however, during periods before and after this, seasonal hurricane activity was comparatively active. Hurricanes are known to be modulated by their ambient environmental conditions, particularly by SST, through its relationship with thermodynamic potential intensity7, and by VWS, which inhibits hurricanes from reaching or maintaining their potential intensity8,9,10. During the more active (or more quiescent) hurricane periods, SST in the hurricane main-development region (MDR) is anomalously warm (or cool) and VWS is anomalously weak (or strong) (Fig. 1c, d).

Figure 1: Interdecadal variability of basin-wide Atlantic hurricane frequency and MDR environmental conditions. a–d, Time series (grey bars/lines) showing the detrended annual basin-wide frequency of hurricanes (a) and major hurricanes (b), and the main development region VWS (c) and SST (d). Thick black lines show the time series smoothed with a filter centred on the 11-year mean. The MDR is defined here as the region 10° N to 20° N, and 275° E to 340° E. Full size image Download PowerPoint slide

The interdecadal SST and VWS variability in the MDR, or simply in the tropical North Atlantic, correlates well with basin-wide hurricane activity. But SST and VWS variability in the MDR does not provide an adequate picture when considering intrabasin hurricane activity. In particular, the patterns of variability exhibit pronounced intrabasin differences between the MDR and regions closer to the United States (US) coast (Fig. 2). During periods of anomalously warm SSTs and weak VWS in the MDR, which coincide with enhanced basin-wide activity, the VWS along the US coast tends to be anomalously high, while the SST anomalies in the coastal region are substantially smaller than those found in the MDR. Of particular relevance to hazard exposure and mortality risk is the relationship between basin-wide activity and US landfall activity13,14,15,16,17,18,19. Here, I address the question of whether the patterns of variability seen in Fig. 2 project onto this relationship.

Figure 2: Patterns of Atlantic VWS and SST variability. a–c, Leading principal component (PC) analysis loading patterns of VWS (a) and SST (b), and their associated principal component time series (c). The US east coast is shown in grey in b, and is delineated with thick black lines in a. The regions outlined with dotted white lines in a are the hurricane MDR (tropical North Atlantic) and the region through which hurricanes approaching the US coast track. The loading patterns have units of local standard deviation from local mean. Full size image Download PowerPoint slide

As discussed above, VWS prevents hurricanes from reaching or maintaining their thermodynamic potential intensity. That is, VWS acts as an intensity-braking mechanism, and hurricanes that move into regions of higher VWS are expected to weaken (or to intensify more slowly). Thus, enhanced VWS along the US coast during more active periods of hurricane activity would be expected to weaken hurricanes that approach or move along the US coast. This signal becomes clear when looking at intensification rates of hurricanes near the US coast (denoted by the northernmost white-outlined region in Fig. 2a) during active and quiescent periods (Fig. 3 and Table 1). The mean intensification rates between the active and quiescent periods are well separated for hurricanes, and particularly for major hurricanes, for which the mean intensification rates are negative in both active periods and positive during the quiescent period (Table 1). The mean rates tend to be near zero, as steady-state intensity is most likely, except for major hurricanes during the quiescent period, which are most likely to intensify by 5 knots (kt) in any given 6-hour period (Fig. 3a–d). Although the differences in the mean are statistically significant (at 90% confidence or greater), they are fairly small. The more pronounced differences are found in the variances: the distributions in intensification rate for the quiescent period are less leptokurtic than for the active periods, with distinctly broader tails. For major hurricanes near the US coast, the variance in intensification rate is two times (or three times) greater during the quiescent period than during the subsequent (or prior) active periods. This elevated variance, or volatility, in intensification rates during quiescent period—which is not just a manifestation of the smaller sample size (see Methods)—would be very likely to introduce substantial additional challenges to forecasting and warning as these hurricanes and major hurricanes approach or move along the US coast.

Figure 3: Probability distributions of observed intensification rates near the US coast. a, b, Probability density distributions of 6-hour intensity changes (ΔV) for hurricanes (a) and major hurricanes (b) during comparatively active periods (1947–1969 and 1993–2015) and an intervening quiescent period (1970–1992). Units are knots [1 knot (kt) = 0.51 m s–1] per 6 hours. Data are provided in bins of 5-kt resolution. Error bars show ± σ (s.d.) from the mean probability density for that bin (based on bootstrap sampling). c, d, Actual counts of 6-hour intensification rates for hurricanes (c) and major hurricanes (d). e, f, Empirical cumulative distribution functions of hurricane (e) and major hurricane (f) intensification rates and their 95% confidence bounds (dashed lines). All data are taken from the region denoted by the northernmost white-outlined region in Fig. 2a, and only intensification rates over water are included; that is, weakening due to landfall does not contribute. Full size image Download PowerPoint slide

Table 1: Means and variances of 6-hour intensification rates near the US coast Full size table

As tacitly expected, there are fewer basin-wide hurricanes during the period of quiescence, and substantially fewer basin-wide major hurricanes (Table 1). But when a hurricane or major hurricane is near the US coast, the probability of intensification is substantially greater during the period of basin-wide quiescence (Fig. 3 and Table 2). The probability that a hurricane near the US coast during the quiescent period would intensify by 10 kt or more in the following 6 hours was roughly twice that of a hurricane near the coast during the active periods (Fig. 3e and Table 2). Moreover, the probability of intensification by 15 kt or more in 6 hours was two times (or three times) more likely during the quiescent period than in the subsequent (or prior) active period. For major hurricanes, the differences are even larger. A major hurricane near the US coast during the quiescent period was about two times (or four times) more likely to intensify by 10 kt or more, and three times (or six times) more likely to intensify by 15 kt or more in the following 6 hours than in the subsequent (or prior) active period (Fig. 3f and Table 2). Rapid intensification near the coast poses a major risk, because it is difficult to forecast and shortens public warning time20,21.

Table 2: Probabilities of exceedance of 6-hour intensification rates near the US coast Full size table

In the hurricane MDR, anomalously warm SSTs occur concurrently with anomalously weak VWS, and vice versa, so that the two factors operate in concert to either enhance or inhibit basin-wide hurricane activity11,12. This is not the case, however, when the region along and near the US coast is considered. In this region, during periods of anomalously warm MDR SSTs, VWS is anomalously strong and the local SST anomalies are substantially weaker. That is, the VWS and SST in this region tend to operate in concert to inhibit intensification during periods when the MDR is conducive to it. This region is where hurricanes approaching the US coast must track, and thus the environmental conditions act as a coastal barrier during periods when basin-wide activity is elevated. The shift in intensification rates of major hurricanes near the US coast between periods of active and quiescent hurricane seasons is particularly pronounced, and the probability that a major hurricane near the US coast would undergo rapid intensification was comparatively much higher during the last quiescent period.

The patterns of variability and co-variability of SST and VWS described here are congruent with the patterns forced by the Atlantic meridional mode (AMM)11,12. The AMM is the leading mode of coupled ocean–atmosphere variability in the Atlantic, and operates on interannual to interdecadal timescales. The AMM can be described as an intrinsic dynamic mode that is established through a wind–evaporation–SST feedback process12, which provides a plausible physical/dynamical foundation for understanding the relationship between MDR hurricane activity and the suppression of intensification along the US coast. Further empirical and numerical modelling explorations of this potential linkage, particularly on interdecadal timescales, are warranted.

The results described here raise questions regarding what might be expected if environmental conditions were to shift back towards the pattern of the previous quiescent period22,23, as well as what might be expected as the tropics continue to warm. The VWS pattern shown in Fig. 2 exhibits marked interdecadal variability but no trend since 1948. However, there is a marked SST trend that projects onto the interdecadal variability, which can be seen from the trend in the leading principal component time series (blue line in Fig. 2c). The SST trend pattern (Fig. 4) shows substantial and significant increasing trends throughout the MDR, but essentially no trend along the US coast. Considering this in tandem with the pattern of VWS variability (Fig. 2a), it seems that the effects of interdecadal VWS variability on hurricanes near the US coast have not been strongly compounded or offset by SST variability or trends. That is, the amplitude of both the interdecadal variability and the SST trend have minima along the US coast (Figs 2b, 4), and seem to have only a minor role in modulating conditions there. It is not clear that this behaviour will remain stationary under projected continued warming of the tropics, but there is the possibility that future warming will not strongly affect the control of VWS on these hurricanes.

Figure 4: Pattern of Atlantic SST trends. SST trend pattern over the period 1948–2015 in units of standard deviation per decade. Hatching shows regions where the trends are not significant. Full size image Download PowerPoint slide

As a closing note, the inverse relationship between variability in the MDR and variability in the region along the US coast may help to explain the weak relationship between the frequency of hurricanes that make US landfall and the frequency of basin-wide hurricanes13,14,15,16,17,18,19,24, by tempering US landfall frequency increases during active periods. Similarly, the present ‘drought’ of US major hurricane landfalls25,26 could plausibly be explained, in part, by this relationship. Given the potential impacts on US coastal hazards and risk, this relationship merits further observational and modelling study.