Abstract

This study contrasts the pattern of low-frequency (LF) and high-frequency (HF) climate variability in the eastern Caribbean. A low-pass Butterworth filter is used to study oscillations in rainfall and regional SST on time scales of greater and less than 8 yr in the period 1901–2002. The results show that the southern and northern Antilles are dominated by HF variability, whereas rainfall fluctuations in the eastern Antilles oscillate at quasi-decadal periods over the 102-yr record. In the southern Antilles, the HF rainfall signal derives from a late-summer response to the ENSO phase: warm and dry versus cool and wet. In the northern Antilles, the HF signal relates to a combination of an ENSO and North Atlantic Oscillation (NAO) phase: a warm ENSO and a negative NAO bring wetter conditions, while a cool ENSO and a positive NAO bring drier conditions. The early rainfall LF signal in SST is characterized by a dipole between the North Atlantic and South Atlantic and is associated with cross-equatorial winds that promote convection in the Caribbean. The study analyzes the upper-ocean structure—in particular, a low (high) salinity signal in the tropical North Atlantic (North Pacific) that relates to LF (HF) climate variability.

1. Introduction

The Caribbean climate is characterized by an annual cycle of rainfall showing a bimodality of the rainy season over the whole Caribbean area, whereas the range of temperature and solar insolation is small throughout the year. The rainy season runs from May to November (boreal summer) when the ITCZ shifts north of the equator (Giannini et al. 2000; Enfield and Alfaro 1999). The distribution of the precipitation is rather heterogeneous due to island topography, the proximity of the subtropical ridge and trade wind subsidence, and variations in sea surface temperature (SST) induced by solar insolation, evaporation, and upwelling, as highlighted in Gamble and Curtis (2008). Several studies have analyzed the interannual variability of Caribbean rainfall (Enfield and Alfaro 1999; Giannini et al. 2000, 2001; Chen and Taylor 2002; Taylor et al. 2002). They show that the rainfall season exhibits maxima of precipitation in May and September and a relative minimum in July called the midsummer dry spell (Gamble and Curtis 2008). Previous work suggested that the early and late rainy seasons should be analyzed independently because different mechanisms modulate the rainfall variability (Chen and Taylor 2002). The early rainy season appears to be more under the influence of the tropical North Atlantic Ocean SST (Enfield and Alfaro 1999; Giannini et al. 2000, 2001; Chen and Taylor 2002; Taylor et al. 2002). A warmer tropical North Atlantic in spring induced by weaker trade winds (Giannini et al. 2000) enhances the amount of precipitation in April–June because the Atlantic ITCZ is shifted northward. Chen and Taylor (2002) also show that the early rainy season following the mature phase of a warm ENSO event tends to be wetter than normal, which is in agreement with Giannini et al. (2001), who suggest that the warming of the tropical North Atlantic is indirectly related to ENSO(+1). For the late rainy season, the rainfall variability is more related to the east equatorial Pacific SST and concurrent responses to ENSO(0). Late-season rains often end early with the gradual warming of east equatorial Pacific SST around August (Chen and Taylor 2002) and ENSO-driven westerly wind shear over the tropical North Atlantic. The intraseasonal variability of the rainfall also seems to be relatively important, especially during September–November. Intraseasonal variability has been related to the Madden–Julian oscillation (30–90 days), which favors rainfall when the associated prevailing easterly winds diminish (Martin and Schumacher 2011). Here, the focus is on interannual to multidecadal oscillations, and subseasonal fluctuations are not considered.

Previous studies have found a distinct role for the North Atlantic high pressure cell in decadal to multidecadal climate oscillations (Rajagopalan et al. 1998; Latif 2001; Enfield and Cid-Serrano 2010; Wang et al. 2009) that modulate Caribbean rainfall and SST in coastal upwelling zones off West Africa and Venezuela (Jury and Rodriguez 2011). Jury and Gouirand (2011) found a decadal mode in eastern Caribbean annual rainfall related to the presence of an SST tripole over the Atlantic Ocean (cool: 35°–20°N, warm: 20°–5°N, cool: 5°N–20°S with respect to increased rainfall). Wet years were associated with a weakening of Atlantic trade winds and higher pressure over the eastern Pacific and South Atlantic, which together favor the development of easterly waves. Thus, quasi-decadal and 2–7-yr oscillations play important roles in Caribbean rainfall variability.

The present analysis is aimed at distinguishing low-frequency (LF) and high-frequency (HF) forcing of Caribbean climate. The rainfall variability is divided into temporal cycles of greater and less than 8 years, using wavelet analysis with a focus on the summer wet season. The first objective is to determine the spatial distribution of low- and high-frequency rainfall variability in the Caribbean area. We then calculate the covariability of Caribbean rainfall and regional SST in those two frequency bands. Our final objective is to determine the role of the subsurface Atlantic and Pacific Oceans in LF and HF climate variability.

2. Data and methods

Monthly gridded rainfall across the eastern Antilles islands (10a°–20°N, 68–57°W) at a resolution of 0.5° from gauge-interpolated fields by the University of East Anglia Climate Research Unit (Mitchell and Jones 2005) in the period 1901–2002 are employed to study the rainfall variability. The choice of this particular area follows from the work of Jury and Gouirand (2011). National Oceanic and Atmospheric Administration (NOAA) monthly sea surface temperature (Smith et al. 2008). on a 2° × 2° grid in the region 20°S–35°N, 180°–20°W in the same time period is used to evaluate the influence of SST on the Antilles climate.

A number of climate indices are compared with the rain and SST fields: the Atlantic multidecadal oscillation (AMO) SST index (www.cdc.noaa.gov/Timeseries/AMO/); the North Atlantic Oscillation (NAO) air pressure index (www.jisao.washington.edu/data_sets/nao/) (Hurrell 1995); the Pacific Niño-3 SST index for the region 5°N–5°S, 150°–90°W (Min et al. 2005); and SST indices for the North and South Atlantic from the NOAA SST datasets: 5°–20°N and 35°W–5°E, 5°–20°S and 75°–25°W, respectively. All fields and indices are detrended as suggested in Behera and Yamagata (2010) and normalized prior to analysis.

To focus on the summer climate, May–November means and overlapping 3-monthly averages are calculated. We used a low-pass recursive Butterworth filter (cut off at 8 yr) for the rainfall and SST time series to obtain the low frequency, and the residuals from these time series are referred to as high frequency. The cutoff at 8 yr was selected primarily to isolate ENSO variability from other signals and considering that Caribbean LF rainfall variability is near 10 yr (Jury and Gouirand 2011). The LF and HF rainfall and SST fields were subjected to a singular value decomposition (SVD) to determine their relationship by maximized covariance. A further explanation of this method is given in Moron and Gouirand (2003) and Tootle et al. (2008). Each SVD is represented by heterogeneous correlation maps for each variable and a pair of associated time scores as in Enfield and Alfaro (1999). The temporal heterogeneous correlation is assessed with the “random phase” test (Ebisuzaki 1997). Results of LF and HF SVD time score wavelet structures obtained in pre- and post-1950 periods were not appreciably different.

Monthly National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) (Kalnay et al. 1996) latent heat flux and meridional and zonal wind fields have also been analyzed to support the notion of a wind–evaporation–SST (WES) feedback mechanism. They are available in the period 1949–2002 and accessible online (http://iridl.ldeo.columbia.edu/SOURCES/.NOAA/.NCEP-NCAR/CDAS-1/.MONTHLY/.Diagnostic/). Anomalies of latent heat flux and surface winds have been calculated based on the 10 (13)-yr extracted from the 20 warmest (coldest) tropical North Atlantic SSTs of the period 1901–2002 (see section 3c), selected from the May–July 3-month averaged HF SST mode 1 SVD scores. Latent heat flux and wind velocity fields have been filtered in the same way as rainfall and SST data before composite analysis. Anomalies are tested with a local Student’s t test applied to the warm composite to ascertain that the values are significantly different from zero at the two-sided 90% level.

We study upper-ocean structure and atmospheric coupling using composite differences for summer periods when the rain LF and HF mode 1 SVD scores are in high (more than 1σ) and low (less than −1σ) phases (eight seasons each since 1958), using Simple Ocean Data Assimilation (SODA) reanalysis fields (Carton and Giese 2008) of temperature, salinity, currents, wind stress, and sea surface height analyzed as north–south sections averaged over the Atlantic (60°W–20°E) and Pacific (150°–90°W) longitudes. A check of data density for these analyses indicates adequate coverage for most composite features of interest. For salinity sections, the observations increase from five observations per grid per year in the South Pacific to more than 100 observations per grid per year north of 20°N. In the Atlantic, observations are more evenly distributed and exceed 50 per grid per year north of 10°S. The standard error and standard deviation for salinity exceeds 0.1 ppt above 100-m depth from 10°S to 15°N. Thus, composite differences should exceed these values to achieve statistical significance.

3. Results

a. Distribution of Antilles rainfall variability

The wavelet power spectrum (using Gabor wavelet) is calculated for Antilles area-averaged unfiltered monthly precipitation to determine how the periodicities evolve from 1901 to 2002 (Fig. 1a). Areas of local significance in the time-varying spectrum are noted above and below 8 yr; however, there is no overall global significance at those frequencies. Before 1940 the rainfall variability is of generally lower amplitude and not significant. The Antilles rainfall exhibits a 3–6-yr frequency (significant at 90%) from 1945 to 1960 and from 1967 to 1973 and after 1990. A 9–10-yr periodicity appears between 1970 and 1990.

Fig. 1.

(a) Wavelet analysis on unfiltered monthly rainfall averaged over the whole Antilles area; wavelet coefficient are shaded according to their absolute value from 0 to 122 (blue to red). Thick lines show 95% confidence; thin lines show 90% confidence for a white-noise process. Shaded area shows the cone of influence (Torrence and Compo 1998). Dotted red line on the right panel represents the variance of the time series, which is also the expected level of the white noise (so there are no significant time scales). (b) Temporal and spatial distribution of Antilles rainfall in dominant frequency bands, as a percentage of total variance.

Fig. 1.

(a) Wavelet analysis on unfiltered monthly rainfall averaged over the whole Antilles area; wavelet coefficient are shaded according to their absolute value from 0 to 122 (blue to red). Thick lines show 95% confidence; thin lines show 90% confidence for a white-noise process. Shaded area shows the cone of influence (Torrence and Compo 1998). Dotted red line on the right panel represents the variance of the time series, which is also the expected level of the white noise (so there are no significant time scales). (b) Temporal and spatial distribution of Antilles rainfall in dominant frequency bands, as a percentage of total variance.

To determine the temporal and spatial distribution of rainfall variability at each grid point in the Antilles (Fig. 1b), a second (Haar) wavelet analysis is performed following the technique of Curtis and Gamble (2008). The northern Antilles have more than 50% of rainfall variance in the 3–5-yr frequency band, 20%–30% in the 5–10-yr band, and ~15% in the 10–21-yr band. The eastern Antilles displays less than 50% of variance in the 3–5-yr frequency band; the rest is divided between 5 and 10 yr (15%–35%) and 10 and 21 yr (16%–33%). The southern Antilles (west of Trinidad) has around 67% of variance in the 3–5-yr frequency band and 15%–40% in the 5–10-yr frequency band with little variance above 10 yr. Thus, the southern and northern Antilles exhibit more HF modulation compared with the eastern islands, where LF is more significant.

b. SVD analysis of summer rainfall and SST

The SVD analysis of LF (periodicity above 8 yr) summer rainfall–SST covariance yields two LF modes representing 66% and 22% of the covariance (Figs. 2a,b,d,e). The LF mode 1 rainfall shows a significant positive correlation over the northeastern Antilles associated with a positive SST influence over the tropical North Atlantic Ocean. Correlations in the South Atlantic are insignificant. LF mode 2 has a mixed rainfall pattern with a significant correlation over the eastern Antilles and the coast of Venezuela associated with positive SST influence over the equatorial Atlantic and a negative SST influence in the central Pacific. The LF mode 1 summer rain time score is correlated with AMO and North Atlantic SST indices (+0.65, +0.73, respectively), indicating that the LF rainfall signal over the eastern Antilles islands derives from the tropical North Atlantic. The LF mode 2 summer rain pattern shows a dipole structure and its time score is weakly associated with NAO and AMO (−0.41, −0.45, respectively), and more influenced by an equatorial antiphase SST pattern between the Atlantic and central Pacific.

Fig. 2.

SVD correlation between May and November rainfall and associated SST for LF (a)–(c) mode 1 and (d),(e) mode 2 and (c),(f) associated rainfall (blue) and SST (red) time scores. SVD correlation between May and November rainfall and (g),(h) associated SST for HF mode 1 and (i) associated rainfall (blue) and SST (red) time scores. SST time scores have been divided by a factor 2.5. Significant areas are stippled.

Fig. 2.

SVD correlation between May and November rainfall and associated SST for LF (a)–(c) mode 1 and (d),(e) mode 2 and (c),(f) associated rainfall (blue) and SST (red) time scores. SVD correlation between May and November rainfall and (g),(h) associated SST for HF mode 1 and (i) associated rainfall (blue) and SST (red) time scores. SST time scores have been divided by a factor 2.5. Significant areas are stippled.

The SVD analysis of HF data yields a mode 1 with 82% of the covariance, while the other modes have a weak covariance. The correlation between the HF mode 1 time scores of rain and SST is 0.50. The HF mode 1 pattern for rain is focused on the southern Caribbean in the zone west of Trinidad, and the SST pattern exhibits significant correlations over the eastern Pacific, consistent with ENSO (Figs. 2e,f). There is also a weak alternating SST pattern in the North Atlantic in the latitudes 0°–30°N. Hence, HF rainfall variability over the southern Caribbean is largely modulated by the eastern Pacific ENSO phase.

c. Seasonal modulation of LF and HF signals

An SVD was performed on overlapping 3-month averaged rainfall and SST from May to November, that is, May–July (MJJ), June–August (JJA), July–September (JAS), August–October (ASO), and September–November (SON) to follow the evolution of the signal from early to late summer in both LF and HF. The objective is to detect how ocean–atmosphere coupling evolves as suggested by Giannini et al. (2000) and by Hu et al. (2011). To evaluate the intermonthly evolution of the SST–rainfall relationship, an extended SVD has been performed using HF seasonal anomalies. The results (not shown) indicate a persistent rainfall signal over the northern Antilles during the months of May, June, and July (r ~ +0.45) but a decline in August–September (r < +0.2). During the months of September, October, and November, the correlation is from 0.35 to 0.58 with a rainfall signal center over the southern Antilles.

The LF mode 1 for the MJJ season captures 82% of the covariance and the rain correlation map reveals a widespread significant (>+0.5) correlation associated with a SST dipole between the tropical North and South Atlantic (Figs. 3a,b). The LF mode 2 for MJJ presents only 20% of the covariance and the correlations are weak between the SST and the rain field. The LF mode 1 for JJA is similar to MJJ except the covariance is 65%. The rainfall pattern associated with SST is weak in the LF mode 1 for JAS and ASO. The LF mode 1 for SON reaches maximum covariance over the southern Caribbean, similar to the HF pattern (Fig. 1b), indicating a weak significance (not shown).

Fig. 3.

SVD correlation between (left) rainfall and (right) SST for (a),(b) MJJ season LF mode 1, and (c),(d) MJJ and (e),(f) SON season HF mode 1. Significant areas are stippled.

Fig. 3.

SVD correlation between (left) rainfall and (right) SST for (a),(b) MJJ season LF mode 1, and (c),(d) MJJ and (e),(f) SON season HF mode 1. Significant areas are stippled.

The correlations between LF mode 1 rain time scores and climate indices confirm a strong relation with the tropical Atlantic (Table 1) that tends to weaken in late summer. Early summer LF rainfall variability falls under control of Atlantic ocean–atmosphere coupling. Increased rain coincides with a warmer (cooler) tropical North (South) Atlantic and NAO in the negative phase (Table 1), meaning a weaker Azores anticyclone and reduced trade wind evaporation. The increase of rainfall could be related to the Atlantic SST dipole that modulates the ITCZ at a decadal time scale (Servain et al. 1999) via a WES mechanism (Xie and Philander 1994). This mechanism could be sustained by decadal ocean–atmosphere processes linked to the AMO or NAO.

Table 1.

Zero-lag cross correlation of SVD mode 1 Antilles rainfall scores and various climate indices for LF and HF in 3-month seasons. Bold values are significant above the 95% confidence level after a Monte Carlo test. ATLN = North Atlantic. ATLS = South Atlantic.

Zero-lag cross correlation of SVD mode 1 Antilles rainfall scores and various climate indices for LF and HF in 3-month seasons. Bold values are significant above the 95% confidence level after a Monte Carlo test. ATLN = North Atlantic. ATLS = South Atlantic.
Zero-lag cross correlation of SVD mode 1 Antilles rainfall scores and various climate indices for LF and HF in 3-month seasons. Bold values are significant above the 95% confidence level after a Monte Carlo test. ATLN = North Atlantic. ATLS = South Atlantic.

For the HF band, the MJJ and SON maps are shown in Figs. 3c–f: MJJ has a distinct heterogeneous correlation pattern compared with the following season, while JJA is the transition period. The SVD correlation pattern and Table 1 indicate that MJJ rainfall is associated with warmer SST in the tropical North Atlantic. Xie and Philander (1994), Chang et al. (1997), and Hu et al. (2011) suggest a WES feedback from a combination of a previous winter ENSO event and a NAO phase to explain the warming of the tropical North Atlantic and consequent displacement of the Atlantic ITCZ and Caribbean summer rainfall. To determine the covariability of ENSO and the NAO phase, we considered the 20 warmest and 20 coldest tropical Atlantic SST fields from the HF SST score 1. The results show that 80% (85%) of the warmest (coldest) years over the tropical North Atlantic follow a warm (cold) Pacific ENSO phase; 70% (65%) of the warmest (coldest) years follow a negative (positive) NAO phase. The WES feedback associated with ENSO and the NAO phase appears to explain the delayed warming of the tropical North Atlantic. This mechanism implies that the pressure gradient anomaly related to a warm ENSO phase and a negative NAO phase decreases the northeast trade winds north of 10°N and hence the evaporation or latent heat flux producing a warming of the North Atlantic (Hu et al. 2011). The northward shift of the ITCZ associated with the warming explains the rainfall increase over the Caribbean.

To determine how the WES mechanism relates to changes of SST in the tropical North Atlantic, latent heat flux and surface wind anomalies were calculated for the February–April (FMA) and April–June (AMJ) seasons preceding the warm and cold phases of the tropical North Atlantic (Fig. 4) using a subset of years from the SST HF mode 1 time scores to correspond with the NCEP data starting in 1949. Seven out of 10 yr correspond to a warm ENSO and negative NAO phase in late winter. Negative anomalies of latent heat flux over the tropical North Atlantic are associated with southwesterly wind anomalies over the same area in FMA, while in AMJ the signals weaken as the tropical North Atlantic warms. In 8 out of 13 yr, a cold ENSO event corresponds with a positive NAO phase. The latent heat flux and wind anomalies exhibit stronger trade winds over the tropical North Atlantic in FMA. As stated earlier, the AMJ anomalies weaken while the tropical North Atlantic cools. These results support the WES mechanism: changes in trade winds and evaporation in late winter produce a gradual response in the tropical North Atlantic in spring.

Fig. 4.

(left) Latent heat flux anomalies (from −10 to 10 W m−2, from blue to red) and (right) wind vector anomalies during the warm and cold phases of the tropical North Atlantic, defined as the 10 warmest and 13 coldest years of the period 1949–2002 from the HF mode 1 of the SST. Black contour (left) and gray shading (right) defines the 0.1 significance level according a Student’s t test.

Fig. 4.

(left) Latent heat flux anomalies (from −10 to 10 W m−2, from blue to red) and (right) wind vector anomalies during the warm and cold phases of the tropical North Atlantic, defined as the 10 warmest and 13 coldest years of the period 1949–2002 from the HF mode 1 of the SST. Black contour (left) and gray shading (right) defines the 0.1 significance level according a Student’s t test.

In late summer, the influence of a concurrent Pacific ENSO event on Caribbean rainfall increases as the delayed WES influence on Atlantic SST decreases (Figs. 3e,f; Table 1). In SON the strongest rainfall signal is located along the coast of Venezuela and the southern Caribbean Antilles, corresponding with an increase of rainfall during the cold ENSO phase. In an ENSO warm phase, more divergent trade winds limit precipitation in the southern Antilles (Ropelewski and Halpert 1987).

d. Composite ocean structure in LF and HF phases

To determine the role of the subsurface Atlantic and Pacific Oceans in Antilles climate variability, the structure of the oceanic thermocline layer has been analyzed. Composite differences were calculated for LF and HF, by averaging eight summers with scores more than +1σ and subtracting the average of eight summers with scores less than −1σ based on the SVD mode 1 Antilles rainfall time scores. Depth sections and graphs averaged over Atlantic longitudes (60°W–20°E) for LF and Pacific longitudes (150°–90°W) for HF are considered. In the Atlantic LF sections and graphs (Figs. 5a–f), warm waters are seen to extend from 0° to 20°N to a depth of 150 m. The equatorial warm layer remains below the surface. In contrast there are cool conditions from 5° to 25°S to a depth of 150 m. Deeper layers show negative temperature differences. The warm surface layer north of the equator is associated with a higher sea level, more westerly wind stress, and a fresh layer that extends to a 50-m depth at 7° and 20°N, indicative of persistent rainfall on the flanks of passing easterly waves. The composite difference of −0.1 ppt is comparable to the standard deviation of 0.2.

Fig. 5.

LF mode 1 composite difference (high − low phase) (left) depth sections and (right) line graphs for (a) temperature, (b) salinity, (c) U (shaded) and schematic meridional and vertical current components (VW) (arrows) based on composite data, (d) sea surface height, (e) zonal wind stress, and (f) meridional wind stress averaged over Atlantic longitudes 60°W–20°E.

Fig. 5.

LF mode 1 composite difference (high − low phase) (left) depth sections and (right) line graphs for (a) temperature, (b) salinity, (c) U (shaded) and schematic meridional and vertical current components (VW) (arrows) based on composite data, (d) sea surface height, (e) zonal wind stress, and (f) meridional wind stress averaged over Atlantic longitudes 60°W–20°E.

The zonal currents show eastward differences near the surface and westward differences on the equator below 200 m. Meridional current differences are southward across the equator in the upper layer. In contrast the meridional wind stress differences are northward across the equator and weak elsewhere. The composite Atlantic Ocean LF dipole structure drives cross-equatorial flows in both the ocean and atmosphere.

The Pacific HF sections and graphs (Figs. 6a–f) show an equatorial La Niña signal, with cool differences extending to a depth of almost 200 m, surrounded by weak warm differences below 150 m. The sea surface height graph has a distinctive dip on the equator that is consistent with the cold tongue. In the salinity difference section, the key feature is a salty zone extending from 5° to 20°N, indicative of weakened convection in the east Pacific ITCZ. The composite difference of +0.06 ppt is comparable to the standard deviation of 0.1. Although not significant, this feature is consistent with a distinctive meridional wind stress graph, reflecting divergence north of the equator. The zonal current section has equatorial westward differences above 150 m and eastward differences below. The meridional current differences diverge strongly over the equatorial cold tongue as expected. The zonal wind stress graph illustrates strengthened trade winds in the latitudes 20°–30°S. The composite Pacific Ocean HF signal is consistent with the onset stage of ENSO. In general, the ocean sections indicate zonal and meridional overturning circulations that complement the air–sea interactions and extend their influence to the thermocline—contributing the additional “memory” needed to sustain oscillations in specific frequency bands.

Fig. 6.

HF mode 1 composite difference (high − low phase) depth sections (left) and line graphs for (a) temperature, (b) salinity, (c) U (shaded) and schematic VW currents (arrows) based on composite data, (d) sea surface height, (e) zonal wind stress, and (f) meridional wind stress, averaged over Pacific longitudes 150°–90°W.

Fig. 6.

HF mode 1 composite difference (high − low phase) depth sections (left) and line graphs for (a) temperature, (b) salinity, (c) U (shaded) and schematic VW currents (arrows) based on composite data, (d) sea surface height, (e) zonal wind stress, and (f) meridional wind stress, averaged over Pacific longitudes 150°–90°W.

4. Discussion

Our study has examined the distribution of low- and high-frequency rainfall signals in the Caribbean Antilles. The southern and northern Caribbean are dominated by HF variability, whereas rainfall fluctuations in the eastern Antilles are more LF in nature. The differences may be related to the proximity of the northern and southern islands to the American continent, whereas the eastern Antilles have a “marine” character where slow adjustments by the ocean predominate. The climate of the southern Caribbean is more directly affected by ENSO and HF fluctuations of the ITCZ. The northern Antilles falls more under the influence of the subtropical jet and associated continental climate regimes.

The HF signal is most significant after August and derives from the onset of an ENSO event. This finding is in agreement with Enfield and Alfaro (1999), wherein a decrease of the rainfall over Venezuela was associated with the ENSO warm phase. In the northern Caribbean, rainfall is modulated by North Atlantic SSTs in the early summer, in agreement with Chen and Taylor (2002) and Spence et al. (2004). Changes of tropical North Atlantic SSTs in spring tend to follow the ENSO phase (Giannini et al. 2000) and NAO phase (Hu et al. 2011): warming after a warm ENSO event and a negative phase of NAO, and cooling after a cold ENSO event and a positive NAO. Moron and Gouirand (2003) previously related warm ENSO events to negative NAO phases in winter. Our results explain how WES feedback operates in the context of HF and LF climate variability. Changes in trade wind evaporation modulate Atlantic SST and ITCZ disposition (Xie and Philander 1994; Chang et al. 1997; Xie and Carton 2004), producing corresponding changes in Caribbean rainfall. The HF early summer rainfall in the northern Antilles refers to the north tropical Atlantic variability associated with decaying ENSO(+1) and NAO forcing, whereas late-summer rainfall variability over the southern Antilles is linked to concurrent Pacific ENSO forcing.

LF rainfall variability over the eastern Antilles represents 50% of the total variance and derives from the SST dipole between the North and South Atlantic. This pattern is consistent with the Atlantic dipole mode noted by Servain et al. (1999) and Xie et al. (2005). Servain et al. (1999) have shown that the latitudinal location of the ITCZ is related to the Atlantic SST dipole at a decadal time scale. Xie et al. (2005) suggested that this dipole affects hurricane activity through a stronger West African monsoon, spawning more easterly waves that then pass over warmer SST favorable to cyclogenesis. They also found that lower air pressure west of Africa fits into the negative NAO phase that favors westward hurricane tracks across the Caribbean. Our results support this concept, reflecting more frequent and intense easterly waves bringing a low-salinity signal in the tropical North Atlantic (cf. Fig. 5b). We also suggest that the WES feedback associated with a warming of the tropical North Atlantic could influence LF rainfall over the Antilles via ocean–atmosphere coupling between the NAO and the AMO.

The frequency and intensity of Atlantic easterly waves and tropical cyclones ultimately influence Caribbean rainfall variability. These are modulated at LF through the Atlantic SST dipole and at HF by ENSO-modulated wind shear (Shapiro 1987; Shapiro and Goldenberg 1998; Goldenberg et al. 2001). The composite ocean depth sections highlight the meridional structure and overturning circulations that sustain climate oscillations in the Atlantic (LF) and Pacific (HF). Cross-equatorial flows in the Atlantic are thought to couple with the NAO phase to promote westward tropical cyclone tracks in the Caribbean. The composites show a low (high)-salinity signal in the tropical North Atlantic (North Pacific) in respect to LF (HF) climate variability. Although these features are noteworthy, they do not achieve statistical significance. Future work could investigate the rela--tionships between the NAO and AMO, and the ocean–atmosphere coupling that governs the quasi-decadal frequency and intensity of tropical cyclones that pass through the Caribbean.

Acknowledgments

We appreciate the comments, suggestions, and help from Prof. Vincent Moron and the three anonymous reviewers, which largely improved the manuscript. This paper was partly supported by a Campus Research Award from the School for Graduate Studies and Research, Cave Hill.

REFERENCES

REFERENCES
Behera
,
S.
, and
T.
Yamagata
,
2010
:
Imprint of the El Niño Modoki on decadal sea level changes
.
Geophys. Res. Lett.
,
37
,
L23702
,
doi:10.1029/2010GL045936
.
Carton
,
J. A.
, and
B. S.
Giese
,
2008
:
A reanalysis of ocean climate using Simple Ocean Data Assimilation (SODA)
.
Mon. Wea. Rev.
,
136
,
2999
3017
.
Chang
,
P.
,
L.
Ji
, and
H.
Li
,
1997
:
A decadal climate variation in the tropical Atlantic from thermodynamics air-sea interactions
.
Nature
,
385
,
516
518
.
Chen
,
A. A.
, and
M. A.
Taylor
,
2002
:
Investigating the link between early season Caribbean rainfall and the El Niño + 1 year
.
Int. J. Climatol.
,
22
,
87
106
,
doi:10.1002/joc.711
.
Curtis
,
S.
, and
D. W.
Gamble
,
2008
:
Regional variations of the Caribbean mid-summer drought
.
Theor. Appl. Climatol.
,
94
,
25
34
,
doi:10.1007/s00704-007-0342-0
.
Ebisuzaki
,
W.
,
1997
:
A method to estimate the statistical significance of a correlation when the data are serially correlated
.
J. Climate
,
10
,
2147
2153
.
Enfield
,
D. B.
, and
E. J.
Alfaro
,
1999
:
The dependence of Caribbean rainfall on the interaction of the tropical Atlantic and Pacific Oceans
.
J. Climate
,
12
,
2093
2103
.
Enfield
,
D. B.
, and
L.
Cid-Serrano
,
2010
:
Secular and multidecadal warnings in the North Atlantic and their relationships with major hurricane activity
.
Int. J. Climatol.
,
30
,
174
184
doi:10.1002/joc.1881
.
Gamble
,
D. W.
, and
S.
Curtis
,
2008
:
Caribbean precipitation: Review, model and prospect
.
Prog. Phys. Geogr.
,
32
,
265
276
.
Giannini
,
A.
,
Y.
Kushnir
, and
M. A.
Cane
,
2000
:
Interannual variability of Caribbean rainfall, ENSO, and the Atlantic Ocean
.
J. Climate
,
13
,
297
311
.
Giannini
,
A.
,
J. C. H.
Chiang
,
M. A.
Cane
,
Y.
Kushnir
, and
R.
Seager
,
2001
:
The ENSO teleconnection to the tropical Atlantic Ocean: Contribution of the remote and local SSTs to rainfall variability in the tropical Americas
.
J. Climate
,
14
,
4530
4544
.
Goldenberg
,
S. B.
,
C. W.
Landsea
,
A. M.
Maestas-Nuñez
, and
W. M.
Gray
,
2001
:
The recent increase in Atlantic hurricane activity: Causes and implications
.
Science
,
293
,
474
479
.
Hu
,
Z.-Z.
,
A.
Kumar
,
B.
Huang
,
Y.
Xue
,
W.
Wang
, and
B.
Jha
,
2011
:
Persistent atmospheric and oceanic anomalies in the North Atlantic from summer 2009 to summer 2010
.
J. Climate
,
24
,
5812
5830
.
Hurrell
,
J. W.
,
1995
:
Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation
.
Science
,
269
,
676
679
.
Jury
,
M. R.
, and
I.
Gouirand
,
2011
:
Decadal climate variability of the eastern Caribbean
.
J. Geophys. Res.
,
116
,
D00Q03
,
doi:10.1029/2010JD015107
.
Jury
,
M. R.
, and
E.
Rodriguez
,
2011
:
Caribbean hurricanes: Interannual variability and prediction
.
Theor. Appl. Climatol.
,
106
,
105
115
,
doi:10.1007/s00704-011-0422-z
.
Kalnay
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Latif
,
M.
,
2001
:
Tropical Pacific/Atlantic Ocean interactions at multi-decadal time scales
.
Geophys. Res. Lett.
,
28
,
539
542
.
Martin
,
E. R.
, and
C.
Schumacher
,
2011
:
Modulation of Caribbean precipitation by the Madden–Julian oscillation
.
J. Climate
,
24
,
813
824
.
Min
,
S.-K.
,
S.
Legutke
,
A.
Hense
, and
W.-T.
Kwon
,
2005
:
Internal variability in a 1000-yr control simulation with the coupled climate model ECHO-G. II: El Niño Southern Oscillation and North Atlantic Oscillation
.
Tellus
,
57A
,
622
640
.
Mitchell
,
T. D.
, and
P. D.
Jones
,
2005
:
An improved method of constructing a database of monthly climate observations and associated high-resolution grids
.
Int. J. Climatol.
,
25
,
693
712
,
doi:10.1002/joc.1181
.
Moron
,
V.
, and
I.
Gouirand
,
2003
:
Seasonal modulation of the El Niño–Southern Oscillation relationship with sea level pressure anomalies over the North Atlantic in October–March 1873–1996
.
Int. J. Climatol.
,
23
,
143
155
.
Rajagopalan
,
B.
,
Y.
Kushnir
, and
Y. M.
Tourre
,
1998
:
Observed decadal midlatitude and tropical Atlantic climate variability
.
Geophys. Res. Lett.
,
25
,
3967
3970
.
Ropelewski
,
C. F.
, and
M. S.
Halpert
,
1987
:
Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation
.
Mon. Wea. Rev.
,
115
,
1606
1626
.
Servain
,
J.
,
I.
Wainer
,
J. P.
McCreary
, and
A.
Dessier
,
1999
:
Relationship between the equatorial and meridional modes of climatic variability in the tropical Atlantic
.
Geophys. Res. Lett.
,
26
,
485
488
.
Shapiro
,
L. J.
,
1987
:
Month-to-month variability of the Atlantic tropical circulation and its relationship to tropical storm formation
.
Mon. Wea. Rev.
,
115
,
2598
2614
.
Shapiro
,
L. J.
, and
S. B.
Goldenberg
,
1998
:
Atlantic sea surface temperature and tropical cyclone formation
.
J. Climate
,
11
,
578
590
.
Smith
,
T. M.
,
R. W.
Reynolds
,
T. C.
Peterson
, and
J.
Lawrimore
,
2008
:
Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006)
.
J. Climate
,
21
,
2283
2296
.
Spence
,
J. M.
,
M.
Taylor
, and
A.
Chen
,
2004
:
The effect of concurrent sea surface temperature anomalies in the tropical Pacific and Atlantic on Caribbean rainfall
.
Int. J. Climatol.
,
24
,
1531
1541
.
Taylor
,
M. A.
,
D. B.
Enfield
, and
A. A.
Chen
,
2002
:
Influence of the tropical Atlantic versus the tropical Pacific on Caribbean rainfall
.
J. Geophys. Res.
,
107
,
3127
,
doi:10.1029/2001JC001097
.
Tootle
,
G. A.
,
T. C.
Piechota
, and
F.
Gutierrez
,
2008
:
The relationships between Pacific and Atlantic Ocean sea surface temperatures and Colombian streamflow variability
.
J. Hydrol.
,
349
,
268
276
.
Torrence
,
C.
, and
G.
Compo
,
1998
:
A practical guide to wavelet analysis
.
Bull. Amer. Meteor. Soc.
,
79
,
61
78
.
Wang
,
X.
,
D.
Wang
, and
W.
Zhou
,
2009
:
Decadal variability of the twentieth-century El Niño and La Niña occurrence from observations and IPCC AR4 coupled models
.
Geophys. Res. Lett.
,
36
,
L11701
,
doi:10.1029/2009GL037929
.
Xie
,
L.
,
T.
Yan
, and
L.
Pietrafesa
,
2005
:
The effect of Atlantic sea surface temperature dipole mode on hurricanes: Implications for the 2004 Atlantic hurricane season
.
Geophys. Res. Lett.
,
32
,
L03701
,
doi:10.1029/2004gl021702
.
Xie
,
S.-P.
, and
S. G. H.
Philander
,
1994
:
A coupled ocean–atmosphere model of relevance to the ITCZ in the eastern Pacific
.
Tellus
,
46A
,
340
350
.
Xie
,
S.-P.
, and
J. A.
C
arton
,
2004
:
Tropical Atlantic variability: Patterns, mechanisms, and impacts. Earth’s Climate: The Ocean-Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 121–142
.