The incidence of these events is unevenly distributed in time, but their important societal impacts have led to efforts to forecast these systems since time immemorial. In the Caribbean, formal records of tropical cyclone occurrence have been kept since Padre Benito Viñes attempted to forecast the arrival of storms. The first to combine traditional observing practices with modern quantitative methods was Padre Benito Viñes (Fernández Partagás and Diaz, 1996; Viñes 1877 and 1895), who developed a systematic method of operational hurricane forecasting in Cuba during the 1860s. Padre Viñes’ forecasting approach (Viñes, 1898) informed the first hurricane warning procedures for the US National Weather Service (NOAA, 2010b). In only six of the past 110 years was the Caribbean spared from the passage of at least one tropical cyclone (Fig. 3). Thus, in this section we will explore the spatial and temporal variations of hurricanes affecting the Caribbean and relate these variations to the large-scale weather and ocean patterns of the region.
Tropical cyclones impacting the Caribbean basin generally form in the far eastern Atlantic Ocean (Fig. 3). Partitioning the tracks of hurricanes (category 1 or greater) that affected the western and eastern Caribbean (Fig. 3, left and right columns, respectively) reveals that many of the weaker storms impacting one region do not impact the other (compare Figs. 3b and 3c), whereas almost all of the category 2 or stronger hurricanes impact nations in both the western and eastern Caribbean (Figs. 3d–3g; Fig. 1). Contrasting the mean storm tracks depicted in Fig. 3 with the climatological “most likely” tracks by month (Fig. 4) emphasizes the increased risk of hurricane passage in the peak hurricane season months of August through October (Fig. 4). In August and September, hurricanes most often form in the “Main Development Region (MDR)” off west Africa and increase in intensity (stronger winds) as they traverse across the tropical Atlantic Ocean, often arriving in the Caribbean region packing “plenty of punch” (Fig. 5).
The situation is somewhat different in October: the majority of hurricanes form in the Gulf of Mexico and have much less time to intensify before they reach the western Caribbean; in contrast, these storms rarely affect the eastern Caribbean (compare Fig. 3 right panels and Fig. 4).
Environment favorable for tropical cyclone activity in the Caribbean
To understand the factors governing the development of tropical cyclones in the North Atlantic and their passage through the Caribbean, we examine the average North Atlantic weather patterns in the peak of the hurricane season (Fig. 6).
Tropical cyclones impacting the Caribbean predominantly form off the west coast of Africa in the months of August through October (Figs. 3, 4 and 6). Throughout these months, the Bermuda High is strong, but is further north than in the remainder of the year. This means that lower pressures generally prevail in the Atlantic south of about 30 N. These lower pressures also lead to weaker vertical wind shear near the Equator than at other times of the year. Finally, the ocean temperatures in the Caribbean exceed 27 C, a critical threshold for the reliable development of active thunderstorms. These factors satisfy the necessary conditions needed for tropical cyclone formation preferentially off West Africa, where the ocean temperatures also exceed the critical 27 C threshold and the low pressures (dark contour roughly outlining the African coastline) signify the presence of the West African monsoon. These same factors also combine to create an environment favorable for tropical cyclone intensification as they cross the near-equatorial region of the North Atlantic on their passage towards the Caribbean basin.
If tropical cyclones follow such clear-cut patterns, then why are they so difficult to forecast? Confounding factors, such as the dry air of the Saharan Air Layer, SAL (Fig. 7), small regions of cool ocean water or large vertical wind shear (the latter known as “jets”), interactions with land or other weather systems all complicate the evolution of individual tropical cyclones. Substantial progress is being made on pinning down each of these complicating factors. For example, the SAL occurs when dust storms from the Saharan desert create deep layers of dry and dusty air that are carried over the North Atlantic Ocean (Braun, 2010; Dunion and Velden, 2004). Interactions between weather systems in the tropics and at higher latitudes transport this dust in complicated ways (Fig. 7). Because of its temperature and moisture content, the SAL is a region of strong vertical wind shear, so it carries with it two negative influences on a tropical cyclone: the dry air and the large change of wind speed with height. Satellite diagnostics (Fig. 7) are used to observe the SAL (Dunion and Velden, 2004) and resulting information is fed to numerical models to forecast evolution of tropical storms. This approach aids in understanding and forecasting the influence of the SAL on tropical cyclone lifecycles.
In a similar fashion, satellites can be used to track the spatial and temporal variability of ocean temperatures (Fig. 6) and also the three-dimensional atmospheric wind patterns.
Through the combination of observational and computer modeling technologies, theoretical model development and physical insight, our understanding of the complex factors governing the evolution of individual tropical cyclones has advanced over recent decades. This understanding gives us a framework for relating the systematic changes in the atmosphere and oceans due to ENSO and other climatic signals with longer-term variations in tropical cyclone activity. We will use this to our advantage now in examining the influence of ENSO on Caribbean storminess.
Temporal variations in tropical cyclone activity impacting the Caribbean
As noted above, inter-annual tropical cyclone frequency (Fig. 8) and track variations are evidence of climate-scale4 modulation of weather in the global tropics. Global-scale changes in the atmosphere and oceans in response to the phases of the ENSO modulate the inter-annual frequency of North Atlantic tropical cyclones (e.g. Goldenberg and Shapiro, 1996; IPCC, 2007a and b; Sabbatelli and Mann, 2007). Other influences, such as the location and strength of the African monsoon, have also been linked to changes in North Atlantic tropical cyclone activity (Landsea and Gray, 1992) and even to subsequent US landfall locations (Gray and Landsea, 1992). The North Atlantic Oscillation (NAO) modifies these other signals over timescales of decades.
The impacts of these global climate signals are even evident over a region as small as the Caribbean. This is evident from a comparison of the surface wind, temperature and pressure patterns averaged only over ENSO years (Figs. 9b and 9c, Table 1) with the same patterns for all years (Fig. 6b). The large region of high pressure (around 1020 mb) in the northeast of the North Atlantic is the Bermuda high. In ENSO years (Fig. 9c), the Bermuda high gets stronger than normal, and extends a further westward into the Caribbean basin. Since the Bermuda high is the dominant weather feature for steering tropical cyclones in this region, these changes result in stronger winds from the east and so we see changes in the tropical cyclone tracks (Fig. 9a). These changes in pressure coincide with a slight cooling (about 0.5 C) of Caribbean ocean and a similar magnitude of warming in the waters of the northern Gulf of Mexico and east of Florida.
Examination of tropical cyclone and ENSO frequency by decade (Fig. 10) does not present a simple picture to explain long-term Caribbean tropical cyclone variability, although the persistent differences between the western and eastern Caribbean throughout the decades is noteworthy. Tropical cyclone activity in the western Caribbean was at a peak in the 1930s, dropped until the 1980s and built back to 1930s levels in the decade since 2000 (Fig. 10a). Activity in the eastern Caribbean differed from the west, apparently breaking down into three distinct phases: an active period in the 1930s through 1950s and again since 1990, with three decades of relatively low activity in the 1960s through 1980s. Interestingly, the phases of the eastern Caribbean generally correspond to active and inactive ENSO decades (Fig. 10b). Giannini et al. (2001) propose that differences in the impacts of ENSO on Caribbean rainfall in the last two decades compared to earlier times are due to changes in the phasing of the NAO and ENSO during these decades. Only the years 1980-2009 were used in the analyses presented in Fig. 9, so these spatial patterns cannot be used to describe variability in the earlier decades depicted in Fig. 10. These decadal-scale variations of tropical cyclone incidence in the western and eastern Caribbean regions are consistent with long-term changes in North Atlantic hurricane activity (e.g. Landsea et al., 1992; Goldenberg et al., 2001).
Long-term variations in the numbers of North Atlantic tropical cyclones (HURDAT; NHC, 2011) have generally been attributed to various climate cycles. However, the long-term changes already observed may also have been contributed to, at least in part, by the atmospheric and oceanic response to global warming (Mann and Emanuel, 2006). We will return to consideration of global warming later.
Landfall statistics in populated regions are robust since people were affected by landfalls and left diaries and other reports of their misadventures (Landsea, 2007). In contrast, the lack of observations out to sea leads to uncertainties in the annual numbers of North Atlantic tropical cyclones in the pre-satellite era (Landsea et al., 2006). Efforts have been made to account for these uncertainties in storm numbers out to sea by mimicking the regions covered by ship tracks in earlier times to sample the storm tracks over the last 30 years to develop probabilities of detection in the earlier part of the record (Mann et al., 2007; Vecchi and Knutson, 2008; Landsea et al., 2010).
The short period of record spanned by the historical hurricane database (NHC, 2011), and the even shorter 30-year period of the satellite era, lead us to explore additional lines of evidence through history for annual tropical cyclone frequency, as well as trends and variations. One approach to extending the historical records is through construction of multi-century proxy records (Donnelly and Woodruff, 2007; Nyberg et al., 2007). Sand deposits in coastal lakes due to storm passage have been used to develop proxy records of storm activity, and possibly even intensities of those storms. One major caveat is that a limited number of locations have been sampled to date, so it is not yet possible to deconvolve changes in annual – or even decadal – tropical cyclone frequency from changes in their average tracks. Since the region currently surveyed is so small geographically, any trends observed here cannot be generalized to the entire North Atlantic basin. Even systematic changes in Caribbean landfalls may just reflect a change in storm tracks over the period of study rather than a change in tropical cyclone numbers across the basin. Still, these studies are a different and fresh approach, providing an additional contribution to the interpretation of the available record and many (e.g. Donnelly and Woodruff, 2007 in Puerto Rico) are being undertaken in the Caribbean region.