Tropical gradient influences on Caribbean rainfall

Authors


Abstract

[1] Interbasin and intrabasin gradients play an important role as a part of a regional system of Caribbean climate drivers, which include the Atlantic warm pool (AWP) and the Caribbean low-level jet (CLLJ). When the Caribbean is conditioned to be wet between May and November, near-surface geopotentials in the Caribbean are lower than in the nearby eastern tropical Pacific and east tropical Atlantic. As a result, there is vertical ascent in the Caribbean through to the middle troposphere which connects to zonal circulations with both the eastern tropical Pacific and the eastern tropical Atlantic. The Caribbean Sea is also warm, and there is a moderate easterly flow regime, indicating a weakening of the trade winds. Deviations from this state caused by changes in one or both sides of the Pacific-Caribbean and Caribbean-Atlantic circulations (and diagnosed by changes in their geopotential gradients) reasonably track the transition of the Caribbean from wet to dry and vice versa on intraseasonal and interannual time scales. The study also uses changes to the gradients to offer insight into why the Caribbean region is projected to be drier during its traditional rainy season in the face of warmer surface temperatures under global warming. The Caribbean seemingly enters into a “July” mode, which persists for the duration of the boreal summer. The mode is characterized by higher (lower) geopotentials in the Caribbean (Pacific and Atlantic), a stronger CLLJ, and anomalous descent in the Caribbean in spite of the warmer surface temperatures.

1. Introduction

[2] The climate of the Caribbean region, best characterized by its bimodal precipitation regime (Figure 1), is understudied. Notwithstanding, a number of recent studies examining climatic features resident wholly or partially in the Caribbean Sea provide insight into the processes that influence rainfall over the Caribbean island chain. For example, a series of recent studies examining the Atlantic warm pool (AWP) link variability in its areal extent to interannual variations of summertime rainfall over the Intra-Americas as well as to tropical Atlantic cyclone activity [Wang et al., 2006, 2008a, 2008b; Wang and Lee, 2007; Wang and Enfield, 2003, 2001]. The AWP is characterized by warm sea surface temperatures (SSTs) which appear in early boreal summer in the Gulf of Mexico and the far western Caribbean Sea [Wang and Enfield, 2001]. The warm water expands eastward as the year progresses, and by September and through November the entire Caribbean basin exceeds 28°C. The warm SSTs associate with a warmer and moister troposphere, reduced sea level pressures, weaker easterlies, less vertical wind shear, and weakened subsidence [Knaff, 1997]. Consequently warmest waters coincide with the peak in Caribbean rainfall [Taylor et al., 2002].

Figure 1.

Precipitation climatology for the Caribbean. The base period is 1958–1998. Monthly means are averaged over the area 10°N–20°N, 65°W–83°W. Units are mm/month. The source is the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) data set [Xie and Arkin, 1997].

[3] Similarly, recent studies of the Caribbean low-level jet (CLLJ) associate changes in its strength with variations in the hydroclimate of the Intra-American region [Muñoz et al., 2008; Whyte et al., 2008; Wang, 2007]. The CLLJ is characterized by an intensification in the trades (below 600 hPa) between 70°W and 80°W and along an east-west axis at approximately 15°N, which peaks in February and July [Muñoz et al., 2008; Whyte et al., 2008; Wang, 2007]. The later peak influences orographic rainfall along the Caribbean coast of Central America [Magaña and Caetano, 2005; Magaña et al., 1999] and helps modulate the severity of the Caribbean midsummer rainfall minimum in July through an enhancement of surface moisture flux divergence [Whyte et al., 2008; Muñoz et al., 2008; Gamble and Curtis, 2008].

[4] Prior to these recent research thrusts, however, some earlier studies also noted the importance of gradients in SSTs and/or sea level pressure between the Caribbean and the eastern tropical Pacific and between the Caribbean and the eastern tropical Atlantic for Caribbean rainfall [Wang et al., 2010; Stephenson et al., 2005, 2008; Spence et al. 2004; Chen and Taylor, 2002; Giannini et al., 2000; Enfield and Alfaro, 1999]. The gradients, dependent on sign, modulate both the early (May–June) and late (September–November) Caribbean rainfall seasons on interannual and longer timescales via circulation changes they induce in and around the tropical Atlantic region [Wang et al., 2010; Stephenson et al., 2005; Spence et al., 2004; Chen and Taylor, 2002; Giannini et al., 2000; Enfield and Alfaro, 1999]. For example, a warm (cool) tropical Pacific and cool (warm) tropical Atlantic SST configuration is associated with a dry (wet) Caribbean late rainfall season through a reduction (increase) in moisture flux convergence over the region [Taylor et al., 2002; Enfield and Alfaro, 1999].

[5] In this study, we consider again the importance of interbasin and intrabasin atmospheric gradients for Caribbean rainfall, particularly in light of the new insights gained from the more recent AWP and CLLJ studies. That is, we consider the role that the gradients play as a part of a regional system of Caribbean climate drivers which include the AWP and CLLJ (see, for example, the conceptual model of Gamble and Curtis [2008]). In section 3 we examine the relative roles of Pacific-Caribbean and Caribbean–east Atlantic gradients in the intraseasonal changes associated with the bimodal pattern of the Caribbean rainfall season. In section 4 we examine the role of gradient strengthening and/or weakening in the determination of Caribbean wet season anomalies due to El Niño–Southern Oscillation (ENSO) events. Finally, in section 5 we use knowledge garnered about the gradient dynamics and the interplay with both the AWP and CLLJ to gain insight into the underlying dynamics associated with an end-of-century drying projected for the Caribbean under global warming. A summary and discussion of the main results of the study are given in section 6. Details about the data sets and methodology used are given in section 2.

2. Data and Methodology

[6] The data for the study are primarily from NCEP/NCAR Reanalysis monthly means [Kalnay et al., 1996]. This global atmospheric analysis is based on numerical model output and observation with variables expressed on a 2.5° × 2.5° grid. The period analyzed was 1958 to 1998, and the domain considered was 40°S–50°N and 180°W–20°E. The domain includes the Pacific and Atlantic Oceans, portions of North and South America and West Africa.

[7] The following variables were analyzed at the levels noted: (1) 1000 hPa surface temperatures, (2) 1000 and 700 hPa geopotential heights, (3) streamlines at 925 and 700 hPa, and (4) vertical velocity at 500 hPa and up to 200 hPa.

[8] The surface temperature over the oceans is used as a proxy for SSTs as our analysis suggests it as comparable to other SST data sources [e.g., Kaplan et al., 1998]. Surface temperature changes will provide insight into the role of the AWP. Differences in geopotential between the eastern Pacific and Caribbean Sea, and between the Caribbean Sea and the eastern Atlantic, are used to diagnose the state and impact of zonal gradients on circulation. The general direction of the streamlines in the lower atmosphere over the south and western Caribbean (south of Jamaica) offers insight into the activity of the CLLJ. Vertical velocity at the 500 hPa level is used to diagnose coupling of low-level convergence (divergence) and high-level divergence (convergence). We restrict our analysis to the Caribbean rainfall season which runs from May to November [Taylor et al., 2002].

[9] The variables are also analyzed utilizing end-of-century data from the PRECIS regional climate model run over a Caribbean and near-Caribbean domain. PRECIS is a limited area atmospheric and land surface model which was run at 50 km resolution. The resolution allowed for the smaller islands of the eastern Caribbean to be represented in the model which is an improvement over global climate models. A description of the PRECIS model's physics is given by Jones et al. [2003, 2004]. The model was run in the region as a part of a multicountry collaborative effort to generate downscaled future projections for the Caribbean. Details of the PRECIS-Caribbean project are given by Taylor et al. [2007], while details of the experiments undertaken, validation, and end-of-century projections for the Caribbean are given by Campbell et al. [2011]. In brief, the model was forced at its lateral boundaries by the simulations of a high-resolution global model (HadAM3P) with a horizontal resolution of 150 km × 150 km. PRECIS experiments were done for two time slices: 1961–1990 (baseline) and 2071–2100 (future) using the A2 and B2 IPCC Special Report on Emission Scenarios (SRES) [Intergovernmental Panel on Climate Change, 2000]. It is the results of the A2 and baseline simulations which are used in this study.

[10] Anomaly maps of the selected variables were generated for the domain. To examine intraseasonal variability the analyzed anomaly maps are derived by first generating climatological maps for the 1958–1998 base period. Anomalies are then calculated by subtracting from, for example, the climatological map for May the mean map derived from averaging all the climatological maps for the entire Caribbean rainy season months (May to November). References to intraseasonal anomalies are therefore references to monthly departures from the mean wet season state. The process is repeated for each month of the wet season.

[11] To examine interannual variability, the climatology for the base period is removed from the entire data set and the resulting anomaly maps composited by month for El Niño and La Niña years. El Niño (La Niña) years are defined as those in which the Niño-3 index exceeds (is less than) a threshold of 0.5°C (−0.5°C) for a minimum of five overlapping seasons. El Niño years since 1958 are 1963, 1965, 1969, 1972, 1976, 1982, 1987, 1991, 1993, and 1997. The La Niña years are 1964, 1970, 1971, 1973, 1975, 1985, and 1989. A Student's t test is used to determine regions where composited anomalies are significantly different for each month from the climatological intraseasonal departure for the same month as derived using the methodology outlined above.

[12] To examine the future state of the Caribbean under global warming, monthly anomaly maps for the period 2071–2100 are calculated with respect to the model's baseline climatology. Mean anomaly maps for each calendar month are produced by averaging like months over the entire future period. The climate change maps are examined for what they reveal about future changes in the variables being analyzed and how these may in turn reflect possible changes in the regional large-scale circulation dynamics. The biases of the PRECIS model with respect to Caribbean region climate are discussed by Campbell et al. [2011].

3. Intraseasonal Variability

3.1. Mean Climatic State for the Wet Season

[13] Figures 2a2f depict the mean climatic state of the domain for the Caribbean wet season (i.e., averaged over May to November). By shading heights in excess of 150 geopotential meters (gpm) (Figure 2a), the four anticyclones of the Atlantic and eastern Pacific are easily identifiable. Lower troposphere streamlines over the Caribbean region (Figure 2b) reveal the importance of the North Atlantic Subtropical High (NASH) in maintaining an easterly regime over the Caribbean and the presence of a jet region in the southern Caribbean basin. The jet is associated with horizontal low-level convergence over a large part of the southwestern Caribbean basin (indicated by the shading in Figure 2b). Surface temperatures exceed 27°C (shaded) over the southwestern Caribbean (Figure 2d), thereby satisfying the oceanic condition for tropical cyclogenesis [Gray, 1968], while temperatures in the eastern tropical Pacific and eastern Atlantic are by comparison cooler.

Figure 2.

Climatological maps for the tropical storm season (May-November). (a) Geopotential (geopotential meters, gpm) at 1000 hPa. Geopotentials in excess of 150 gpm (less than 100 gpm) are shaded darker (lighter). (b) Streamlines at 925 hPa. Shading indicates horizontal velocity convergence. (c) Geopotential (gpm) at 700 hPa. Geopotentials in excess of 3160 gpm (less than 3140 gpm) are shaded darker (lighter). (d) Surface temperature (°C). Shading denotes temperatures exceeding 27°C. (e) Vertical velocity (Pa/s) at 500 hPa. Signs are reversed so positive (shading) indicates ascent. (f) Vertical velocity (Pa/s) along 20°N. Signs are reversed so positive (light shading) indicates ascent. The base period is 1958–1998.

[14] In Figure 2e positive vertical velocities at 500 hPa occur over the tropical eastern Pacific, central and southwestern Caribbean, the Amazon basin, and the equatorial Atlantic. A vertical cross section of vertical velocity along 20°N (Figure 2f) also shows subsidence over the depth of the tropical troposphere between 140°W and 110°W (tropical Pacific), and 40°W and 20°W (eastern Atlantic), and ascent between 110°W and 60°W (Caribbean), except at 85°W, where lower tropospheric subsidence interrupts. This pattern is consistent with a zonal circulation between the tropical Pacific and the Caribbean Sea, and between the Caribbean Sea and the eastern Atlantic. In both cases a rising limb occurs over the Caribbean Sea. The divergent wind at 850 hPa and 200 hPa (not shown) confirms the coupling of the lower and upper atmosphere for each arm of the zonal circulation.

[15] In the 10°N–30°N latitudinal band, the Caribbean Sea and specifically the southwestern Caribbean basin exhibit a minimum in geopotential which is evident through the lower troposphere (Figures 2a and 2c). There are comparatively higher geopotentials to the west (eastern Pacific) near the surface (Figure 2a) and to the east (eastern Atlantic) up through the lower troposphere. The resulting interbasin and intrabasin geopotential gradients are dynamically consistent with the regions of ascent and descent noted above, i.e., lower (higher) geopotentials corresponding to ascent (descent), and therefore represent a useful diagnostic for the Pacific-Caribbean and Caribbean-Atlantic circulations for the mean Caribbean wet season.

[16] The climatological characteristics of the Caribbean basin from May to November are summarized in Table 1. In Pacific-Caribbean gradient and Caribbean-Atlantic gradient columns, the region with higher geopotential is noted first.

Table 1. Climatological Characteristics of the Caribbean During the Mean Wet Season (May–November)
Vertical Velocity (500 hPa)aPacific-Caribbean Gradient (Surface to 700 hPa)bCaribbean-Atlantic Gradient (Surface to 700 hPa)bSurface TemperaturecLower and Middle Troposphere Streamlinesd
  • a

    Positive indicates ascent.

  • b

    W-E (E-W) implies higher geopotential anomalies to the west (east).

  • c

    The temperature 27°C is the critical value for tropical cyclogenesis.

  • d

    Easterlies indicate the direction of the prevailing low-level winds over the southwestern Caribbean.

PositiveW-EE-W>27°Ceasterlies

3.2. Bimodal Variability

[17] Figures 3 and 4 show intraseasonal geopotential (700 hPa) and streamflow (925 hPa) anomalies, respectively. Shading in Figure 3 shows regions of positive vertical velocity anomalies at 500 hPa (i.e., ascent), while shading in Figure 4 shows regions of positive surface temperature anomalies (warming). In all cases anomalies are also with respect to the mean wet season state depicted in Figure 2. We use Figures 3 and 4 to track the intraseasonal changes in the geopotential gradients and comment on the state of the AWP and CLLJ as the Caribbean is being conditioned for rain starting in May and through the peak of the season in October (see Figure 2). They are also instructive for what they suggest about dominant influences in July and August when a southwest extension of the NASH gives rise to the midsummer rainfall minimum [Curtis and Gamble, 2008; Gamble et al., 2008; Taylor et al., 2002].

Figure 3.

Intraseasonal geopotential (gpm) anomalies at 700 hPa for the Caribbean wet season (May–November). Shading depicts regions of positive vertical velocity anomalies at 500 hPa indicating ascending motion.

Figure 4.

Intraseasonal streamline anomalies at 925 hPa for the Caribbean wet season (May–November). Shading depicts regions of positive surface temperature anomalies indicating warming.

[18] Table 2 also summarizes the main changes in the respective variables by month for the Caribbean region as gleaned from analysis of lower tropospheric geopotential and streamline anomalies. The signs of the intraseasonal 500 hPa vertical velocity and surface temperature anomalies are shown. The streamlines column indicates the direction of the anomalous low-level streamlines in the southwest Caribbean, and the geopotential anomaly gradients between the eastern tropical Pacific and the Caribbean Sea and between the Caribbean Sea and the eastern tropical Atlantic in the lower troposphere are also given. In determining the gradient we consider points located at 140°W (eastern Pacific), 80°W (Caribbean Sea) and 30°W (eastern Atlantic) along 20°N and summarize the patterns evident in the lower troposphere, i.e., at 1000 hPa (not shown) through the 700 hPa (Figure 4). As in Table 1, higher geopotentials are written first, so for the Pacific-Caribbean gradient column, W-E indicates higher geopotentials exist over the Pacific compared to the Caribbean Sea and implies that the rising limb of the zonal circulation over the Caribbean Sea is enhanced. In the Caribbean-Atlantic gradient column, E-W implies higher geopotentials exist over the eastern Atlantic than over the Caribbean and the rising limb over the Caribbean Sea is enhanced.

Table 2. Intraseasonal Changes Over the Caribbean Basina
MonthVertical Velocity (500 hPa)bPacific-Caribbean GradientcCaribbean-Atlantic GradientcSurface Temperature AnomaliesdStreamlinese
  • a

    Anomalies are with respect to the mean wet season state.

  • b

    Negative (positive) vertical velocity anomalies indicate enhanced subsidence (ascent).

  • c

    The Pacific-Caribbean gradient is the difference in lower tropospheric (1000 through 700 hPa) geopotential anomalies between the eastern Pacific (140°W) and the Caribbean Sea (80°W) along 20°N. The Caribbean-Atlantic gradient is the same but for differences between the Caribbean Sea and the eastern Atlantic (30°W). W-E (E-W) implies higher geopotential anomalies to the west (east); B implies anomalies of near-equal magnitude.

  • d

    Negative (positive) surface temperature anomalies indicate enhanced warming (cooling).

  • e

    The general direction of streamline anomalies at 925 and 700 hPa. SSE, south-southeasterly; SW, southwesterly; NW, northwesterly.

May−veW-EB−vesoutherlies, 925 hPa; westerlies, 700 hPa
June+veBE-W+vedeep SSE
July−veE-WB+vedeep easterly
August−veE-WW-E+vedeep easterly
September+veBE-W+vedeep SW
October+veW-EE-W+ve east of 80°Wdeep westerly
November+ve east of 80°WW-EW-E−venortherly, 925 hPa; NW, 700 hPa

[19] In tandem, Figures 3 and 4 and Table 2 suggest that in the early rainfall season (MJJ), it is during the month of June that the Caribbean is most conducive to rain. Positive vertical velocity anomalies at 500 hPa occur over the north, west, and southeastern Caribbean (Figure 3b) suggesting an enhanced rising limb of the zonal circulations. The enhancement is largely driven by changes in the zonal circulation on the Atlantic side where the geopotential anomalies suggest an E-W gradient (Figure 3b). The atmospheric gradient reflects the zonal Atlantic temperature gradient as western Caribbean surface temperature anomalies are positive (Figure 4b) because of the appearance of the AWP. There are also south-southeasterly streamlines (Figure 4b) which extend up to 700 hPa (not shown) indicative of enhanced low-level convergence in the southwestern Caribbean. Interestingly, at the start of the rainy season in May, save for the Pacific-Caribbean gradient (W-E), the Caribbean region is less conducive to tropical convection as compared to its mean wet season state (see Table 2). At the surface (not shown) there are northwesterly streamline anomalies north of the Caribbean which extend through to the mid to lower troposphere and southerly low-level flow anomalies over the Caribbean basin (Figure 4a). These suggest the importance of tropical-extratropical influences in May for producing rain.

[20] In July, the NASH is reestablishing its presence resulting in positive geopotential anomalies over much of the Caribbean region, with smaller anomalies in the eastern and equatorial Pacific (Figure 3c). The resulting unfavorable Pacific-Caribbean gradient in the lower troposphere (Table 2) weakens the rising limb over the Caribbean. As a result negative vertical velocity anomalies exist over most of the central Caribbean basin for both July and August (Figures 3c and 3d). In August, ascent in the Caribbean is further diminished by the addition of an unfavorable gradient in the Atlantic, as captured in Figure 3d by the lower geopotentials and positive vertical velocity anomalies between 10°N and 20°N just west of Africa. There is also a strengthening of the CLLJ as indicated by easterly streamline anomalies in the south Caribbean (Figures 4c and 4d) which extend deep into the lower troposphere (not shown). The conditions are consistent with the onset of the midsummer minimum in rainfall which occurs in spite of Caribbean surface temperatures of the correct sign to support convection (Figures 4c and 4d). That is, in July and August, unfavorable lower tropospheric gradients are induced by the incursion of the NASH, which result in less rain because of diminished ascent and stronger surface easterlies, in spite of warm waters in the Caribbean.

[21] During October which is the peak of both the late rainy season and tropical cyclogenesis in the tropical Atlantic [Inoue et al., 2002], the rising limb over the Caribbean Sea is again reinforced by favorable geopotential gradients on both the Pacific and Atlantic sides (Figures 3f and 4f and Table 2). There are also positive surface temperature anomalies in the Caribbean and deep westerly anomalies (Figure 4f) with the latter indicating weaker trades and weaker vertical shear. September shares most of the characteristics of October (Table 2). It is worth noting, that in September and October, the Caribbean Sea is warmer than the eastern Pacific, with the highest anomaly zonal gradient of temperature occurring in October. It is likely the strong surface temperature gradient which is driving the strengthened Pacific (descent)–Caribbean (ascent) circulation pattern at this time. Wang et al. [2010] show that during boreal summer and fall, coincident with the AWP expansion, there is a strong atmospheric circulation characterized by ascending motion over the AWP and descending motion over the southeastern Pacific.

[22] In November the Atlantic side is also inhibitive for the Caribbean rising limb of the zonal circulation (Figure 3g and Table 2). The SST requirement may also not be sufficient for cyclogenesis as indicated by cooler anomalies over the Caribbean (Figure 4g). The streamline anomaly map (Figure 4g) again suggests tropical-extratropical influences in November as seen in May.

4. Interannual Variability

[23] A number of studies document that during the El Niño year there is a tendency for the Caribbean to be dry and tropical cyclone activity to be depressed during the late rainfall season, with an opposite signal in the La Niña year [Jury and Enfield, 2010; Taylor et al., 2002; Giannini et al., 2000; Gray, 1984]. In the El Niño+1 year (the year of El Niño decline) the tendency is for a wet early rainfall season [Chen and Taylor, 2002; Taylor et al., 2002]. We examine the role of changes to the interbasin and intrabasin atmospheric gradients in determining these conditions in tandem with changes in the CLLJ and AWP.

[24] Figures 5a, 5c, 5e, and 5g show composite anomalous streamflow for June, July, August, and October of El Niño years (with respect to the climatology of the base period) with regions of positive surface temperature anomalies shaded. Figures 6a, 6c, 6e, and 6g show the same for La Niña. The darkest shading indicates the warmest waters. Figures 5b, 5d, 5g, and 5h are similar but for 1000 hPa geopotential anomalies with regions of positive vertical velocity anomalies (ascent) at 500 hPa shaded. June and October are shown as they represent the peak of the early and late rainfall seasons (see section 4) when the geopotential gradients yield a circulation pattern characterized by ascent in the Caribbean. July and August are shown as they represent the midsummer rainfall minimum months when the circulation pattern is weakened or reversed because of the NASH. Figures 5 and 6 are complemented by Table 3, which summarizes by month the El Niño composite fields which show a statistically significant change (and the sign of that change) from the intraseasonal anomaly maps of Figures 3 and 4 as determined by a Student's t test.

Figure 5.

El Niño composite anomalies for June, July, August, and October. (a, c, e, g) Streamline anomalies at 925 hPa. Shading depicts regions of positive surface temperature anomalies. Darker shading indicates the largest anomalies. (b, d, f, h) Geopotential (gpm) anomalies at 700 hPa. Shading depicts regions of positive vertical velocity anomalies at 500 hPa, indicating ascending motion.

Figure 6.

As in Figure 5 but for the composite of La Niña years.

Table 3. Summary of Significant Differences in Selected Large-Scale Parameters Over the Caribbean During the Composite El Niño Year and the La Niña Yeara
MonthVertical VelocityPacific-Caribbean GradientCaribbean-Atlantic GradientSurface Temperature AnomaliesStreamlines
  • a

    Columns and abbreviations are as defined for Table 2. For the gradients, significant difference was determined for the 1000 hPa level.

El Niño
May E-W   
June E-W  southerly, 925 hPa
July-veE-W +vedeep easterly
August   +veeasterly, 925 hPa
September    SW, 925 hPa
October E-W   
November E-W   
 
La Niña
May−ve   westerly, 700 hPa
June−veE-WW-E−venortherly
July W-EW-E−vewesterly
August+veW-EW-E westerly, 925 hPa
September    westerly, 925 hPa
October W-EE-W−ve 
November W-E −venortherly, 925 hPa

[25] Table 3 suggests that it is the Pacific-Caribbean arm of the circulation which is significantly altered during the El Niño and in such a manner that ascent in the Caribbean is inhibited. Figures 5d, 5f, and 5h show that the Pacific-Caribbean geopotential gradient is biased toward higher (lower) Caribbean (east Pacific) geopotentials at 1000 hPa from July onward. There is evidence of the bias up to the 700 hPa level in July and August (not shown). An E-W Pacific-Caribbean gradient represents a weakening of the mean circulation in June and October which features Caribbean ascent, but a strengthening of that observed during July and August which favors descent (see the Pacific-Caribbean gradient column in Table 2).

[26] From Table 3, other significant El Niño changes include lower tropospheric easterly anomalies in the southern Caribbean in July and August (Figures 5c and 5e) indicative of a strengthened CLLJ. This is consistent with the results of Whyte et al. [2008]. For the same months significant positive temperature anomalies exist over the Caribbean (Table 3), but although the Caribbean is warm the east Pacific is warmer (Figures 5c and 5e). During an El Niño, then, the interbasin temperature gradient enhances an already unfavorable geopotential gradient in place because of the southward incursion of the NASH and intensifies midsummer drought conditions.

[27] Table 3 and Figures 5g and 5h also suggest that subsequent El Niño drying in the late rainfall season [Giannini et al., 2000; Taylor et al., 2002] is likely contributed to by a perpetuation of the unfavorable geopotential gradient driven by the interbasin temperature differences. (Recall that the NASH would have by then migrated northward). There are however other features conditioning the Caribbean which may offset the gradient influence including that the AWP is at its maximum eastward extent (Figure 5g), there are westerly flow anomalies at the surface (Figure 5h), and there is evidence of the Pacific North Atlantic (PNA) pattern (Figure 5h) which would yield a wetter north Caribbean (above 18°N).

[28] Figure 6 shows composite anomaly maps for La Niña years, while Table 3 summarizes the significant changes for the wet season months. It is important to note that three of the La Niña years composites also correspond with El Niño+1 years during which warm temperature anomalies appear in the eastern tropical Atlantic during the first half of the year [Enfield and Mayer, 1997; Taylor et al., 2002]. There is evidence of this in Figure 6a which shows that in June the Caribbean is cool (see also Table 3) but warm anomalies and ascending motion (Figure 6b) exist north of 10°N in the east tropical Atlantic. The result is that in May and June the ascending branch in the Caribbean is inhibited by an unfavorable Caribbean-Atlantic gradient. Table 3 suggests that the Pacific-Caribbean gradient is also unfavorable, and the region is likely experiencing a La Niña “drought.”

[29] In July and August the Caribbean-Atlantic geopotential gradient remains unfavorable largely because of the intrabasin temperature gradient (Figures 6c and 6d). The Pacific-Caribbean gradient is however becoming favorable and there is ascending motion in the Caribbean (Table 3 and Figures 6d and 6f). This is likely due to the fact that although the Caribbean still has cool anomalies, the Pacific is cooler due to the deepening La Niña. A favorable Pacific-Caribbean gradient would offset the opposing impact of the NASH on the gradient. There is evidence in both Figures 6c and 6e of westerly anomalies in the southern Caribbean suggesting a weaker CLLJ and ascent over the Caribbean east of 80°W. The intensity of the midsummer rainfall minimum during La Niña years may be a function of the strength of the La Niña and its ability to impact the Pacific-Caribbean geopotential gradient.

[30] By October, however, cold anomalies have established themselves in the east Pacific and the differential cooling, i.e., a cooler Pacific and relatively warmer Caribbean configuration should favor Caribbean ascent. Further enhancement comes from the Atlantic side which has slightly higher geopotentials (Figure 6h and Table 3) making the region conducive to rain. The low-level streamlines (Figure 6g) and region of Caribbean ascent (Figure 6h) suggest that storms typical of this time of year may track through the eastern Caribbean and north of eastern Cuba and Hispaniola as the far southwestern Caribbean is still unfavorable for cyclogenesis.

5. Future Changes

[31] There are strong indications that the Caribbean region will be warmer and drier in the mean by the end of the century under global warming. Various GCM-based rainfall projections show rainfall decreases (−6.8% ± 15.8%) in both annual and June–August rainfall totals [Nurse and Sem, 2001]. Campbell et al. [2011] show projected end-of-century changes in monthly, seasonal, and annual rainfall for a Caribbean rainfall index under A2 warming scenario as simulated by the PRECIS regional model. They show that by 2071–2100 annual rainfall totals decrease by 20% primarily due to a robust June–October drying signal in which rainfall decreases approach 35% in some months. The projected decreases also exceed the magnitude of historically observed variations obtained from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) data set [Xie and Arkin, 1997; see also Campbell et al., 2011, Figure 5].

[32] Figure 7 shows the anomalous state of the PRECIS Caribbean domain at the end of the century under the A2 SRES storyline. The anomaly maps are generated by subtracting from the future climatology that of the present as simulated by the PRECIS model, and then averaging the difference maps for the mean Caribbean rainfall season. The present-day climatology as simulated by PRECIS is a fair representation of the observed large-scale conditions, as shown by Campbell et al. [2011], and can be similarly described by the state depicted in Table 1. Figure 7, then, represents deviations from this state, with statistically significant differences shaded.

Figure 7.

Climatological maps for the mean Caribbean wet season (May–November) for 2071–2100 relative to 1961–1990. (a) Geopotential (gpm) at 1000 hPa. (b) Streamlines at 925 hPa. (c) Geopotential (gpm) at 700 hPa. (d) Surface temperature (°C). (e) Vertical velocity (Pa/s) at 500 hPa. Signs are reversed so positive indicates ascent. Shading in Figures 7a, 7e, and 7f indicates differences statistically significant at the 90% level. In Figures 7c and 7d the entire domain is significantly different.

[33] Figure 7a not only shows that higher surface geopotentials exist over the Caribbean at the end of the century, but also that geopotentials of opposite sign are over the nearby eastern tropical Pacific and northern South America. This suggests that the Pacific-Caribbean gradient which is biased to Caribbean ascent in the present-day mean wet season is being weakened. This is supported by both Figures 7e and 7f. In Figure 7e, negative 500 hPa vertical velocity changes exist over the Caribbean while positively signed vertical velocity changes occur over the nearby tropical Pacific. Similarly, Figure 7f shows negative vertical velocity changes indicative of increased subsidence over the Atlantic (55°W–95°W) and positive changes indicative of enhanced ascent over the Pacific (95°W–115°W) relative to the model 1960–1991 baseline. Circulation between the Pacific and Caribbean is seemingly altered.

[34] We note also that by the end of the century, there are easterly flow anomalies in the lower troposphere over the south Caribbean, northeasterly near the surface (Figure 7b) and strongly easterly through 700 hPa (not shown). The southwestern Caribbean is in fact dominated by anomalous anticyclonic flow in the lower levels. The flow pattern of the domain may reflect the influence of a stronger than normal or earlier or even prolonged influence of the NASH [Rauscher et al., 2008]. This would contribute to the weaker Pacific-Caribbean wet season gradient and is reminiscent of what occurs in climatological July (see section 3.2). We also note that in the far north Caribbean there is anomalous westerly flow which is consistent with some projections for the region which suggest that the future tendency in that region may be for slightly enhanced rainfall [see also Campbell et al., 2011].

[35] There are higher surface temperatures over the entire domain (Figure 7d) as would be expected under a global warming scenario. Changes over the Caribbean Sea are however one degree less than over the nearby eastern Pacific (see Figure 7d). A warmer Pacific–cooler Atlantic configuration would contribute as well (i.e., in addition to a stronger or prolonged NASH) to the weakening of the Pacific-Caribbean wet season gradient yielding significant drying in the Caribbean as projected by the modeling studies. This is not unlike the pattern seen during July and August of the El Niño year.

[36] Future monthly changes in the analyzed fields with respect to the baseline monthly climatology were also examined. Because the pattern for each month is largely captured by the mean pattern for the season (Figure 7), these changes are not shown. We note however that near-surface geopotential anomalies over the Caribbean bias the Pacific-Caribbean gradient unfavorably (i.e., inhibits ascent) in all months but November. Subsidence, as indicated by negative vertical velocity anomalies in the Caribbean, is also bolstered by the Caribbean-Atlantic gradient which depicts higher geopotentials over the Caribbean relative to the nearby Atlantic in the PRECIS domain for all months except September. There are also stronger trades for all months indicative of a strengthening of the CLLJ beyond July. Together, all the fields suggest a drier Caribbean in the mean for all the traditional rainy months in spite of the background of higher surface temperatures due to global warming.

[37] Table 4 is the summary of the changes in the state of the analyzed fields for the end-of-century Caribbean wet season as gleaned from Figure 7. In the mean, the future Caribbean is characterized by negative vertical velocity through the middle troposphere and higher geopotentials especially in comparison to the east Pacific. The implication is that of a weakened ascending limb over the Caribbean in the Pacific-Caribbean zonal circulation and a stronger CLLJ.

Table 4. Future Climatological Changes for the Mean Caribbean Wet Season (May–November) Relative to Baseline Monthly Climatology for the Southwestern Caribbean as Simulated by the PRECIS Model
Vertical Velocity (500 hPa)Pacific-Caribbean Gradient (Surface to 700 hPa)Caribbean-Atlantic Gradient (Surface to 700 hPa)Surface TemperatureLower and Middle Troposphere Streamlines
−veE-WW-E+venortheasterlies

6. Summary and Discussion

[38] Climatic phenomena of the Intra-American seas including the AWP and CLLJ are important determinants of the climatic state of the Caribbean island territories. We also suggest that interbasin and intrabasin atmospheric gradients are equally important considerations. Whereas the AWP conditions the Caribbean waters for convection and the CLLJ varies the strength of convergence and divergence in the south Caribbean and alters vertical shear, the zonal gradients modulate the circulations between the Caribbean and the eastern tropical Pacific and the Caribbean and the eastern tropical Atlantic. The study shows that when the Caribbean is conditioned to be wet in the mean, it is characterized by warm surface temperatures, upward vertical velocity through to the middle troposphere, a moderate easterly flow regime, and the ascending arm of a zonal circulation that spans both the eastern equatorial Pacific and eastern tropical Atlantic (see Table 1).

[39] Deviations from this state serve as a diagnostic for the processes which transition the region between wet and dry on intraseasonal, interannual or longer timescales. The study suggest that changes to the Pacific-Caribbean and/or Caribbean-Atlantic geopotential gradients can be influenced by either strong colocated surface temperatures anomalies which yield zonal temperature gradients, e.g., as in the phases of AWP expansion or El Niño/La Niña influences, and/or can be due to higher surface pressures induced by the southward movement of the NASH in July and August. The strength of the gradients and the Caribbean region's conduciveness to rain would then be determined by the relative state of each phenomena at any given time.

[40] On the intraseasonal timescale, the region is favorable for rain in June largely because of warm waters and enhanced ascent driven by a favorable geopotential gradient on the Atlantic side. The idea of the tropical Atlantic being the driver of early season rainfall is therefore reinforced [Taylor et al., 2002] and makes the AWP a phenomenon of particular significance at this time of the year. When the region transitions to dry in July and August it is in spite of warm waters and so the AWP loses it primary modulating role. The southern Caribbean is now characterized by strong easterlies through the lower troposphere indicative of a strengthened CLLJ. Both can be attributed to the southward jog of the Azores High [Knaff, 1997; Gamble and Curtis, 2008] which results in higher Caribbean geopotentials and a weakening of Caribbean basin ascent associated with the Pacific-Caribbean gradient. This marks the onset of the region's midsummer rainfall minimum. The Caribbean's second rainfall peak in October occurs when the zonal circulation between the Caribbean and Pacific and the Caribbean and east Atlantic is reestablished such that ascent occurs over the Caribbean. This occurs in tandem with much warmer waters in the Caribbean because of a well-established AWP.

[41] It is an alteration of the Pacific-Caribbean circulation which is a primary modifier of the Caribbean rainfall regime during ENSO events. During the El Niño year the geopotential gradient is such that there are higher geopotentials on the Caribbean side because of a warm Pacific–cooler Caribbean temperature gradient, resulting in the ascending branch over the Caribbean being weak for much of the rainfall season. This contributes to the noted tendency for the region to be dry [Giannini et al., 2000; Taylor et al., 2002]. In July, in particular, the unfavorable gradient enhances the background conditioning of the region by the NASH (which includes an unfavorable atmospheric gradient and a stronger CLLJ), thereby intensifying the midsummer drying in the region [Whyte et al., 2008]. In the La Niña year, beginning in July and through the late rainfall season, the geopotential gradient is reversed because of a cool Pacific, relatively warmer Caribbean temperature gradient, thereby enhancing ascent in the Caribbean and contributing to favorable conditions for rain.

[42] Finally, the analysis helps explain why by the end of the century the mean state of the Caribbean region will be biased toward dry conditions. During the future Caribbean “wet season” the Caribbean is characterized by higher relative geopotentials and stronger low-level easterlies and atmospheric gradients that weaken regional ascent. In fact, the state of the Caribbean is reminiscent of its condition during the present-day midsummer rainfall minimum (see Table 3), prompting the suggestion that in the future the Caribbean under global warming may be in a perpetual “July” mode. This is consistent with the work by Rauscher et al. [2008], who similarly project an end-of-century intensification of the Meso-American midsummer drought because of stronger easterlies in June and July and an earlier displaced and stronger NASH. We extend the idea however to suggest that a stronger CLLJ will endure through the end of the traditional Caribbean rainy season, i.e., even beyond July (M. A. Taylor et al., Why dry? Investigating the future evolution of the Caribbean low-level jet to explain projected Caribbean drying, submitted to International Journal of Climatology, 2011).

[43] The stronger south Caribbean easterlies will be in part due to differential warming between the eastern tropical Pacific and western tropical Atlantic (see Figure 7d) which ensures that the Caribbean is still cooler. This is similar to what occurs during the late season of El Niño years. So, in tandem with the NASH's impact in July and August, the interbasin temperature gradient will maintain unfavorable atmospheric gradients throughout the entire rainy season. The resultant drying in the Caribbean from June through October is therefore consistent with the idea of a perpetual El Niño in the future [DiNezio et al., 2009; Vecchi and Soden, 2007; Vecchi et al., 2006]. There are clear implications for tropical cyclone formation and intensification within the Caribbean basin which would be diminished under this scheme. The idea, however, needs further investigation and represents future work to be undertaken.

[44] In general the results of this study point to the usefulness of variables which capture the state of the AWP, CLLJ, and Pacific-Caribbean and Caribbean-Atlantic atmospheric gradients for diagnosing and analyzing the climatic state of the Caribbean region at any given time, i.e., its predisposition to be wet or dry. In particular we emphasize the importance of including the interbasin and intrabasin atmospheric gradients in monitoring programs or new predictive model schemes for the Caribbean region because of the additional insight they provide into circulation characteristics and dominating processes. Additionally, with growing interest in the use of dynamical regional models and/or the downscaling of global model results for the region, the results of this study should prove useful in the development of diagnostic criteria for choosing models for the region, in the interpretation of the model results, and in the selection of parameters for statistical downscaling.

Acknowledgments

[45] The modeling portion of this work was in part funded by the Caribbean Community Climate Change Centre (Belize). Special thanks go to the Hadley Centre (UK) for the PRECIS model support. Completion of this work was also enabled by a University of the West Indies Research Fellowship and funding from the CARIBSAVE partnership. We thank the anonymous reviewers, whose comments significantly enhanced the paper.

Ancillary