Climatic variations in the tropics are mainly manifest in rainfall. As in other parts of the tropics, interannual variability also affects Central America and the Caribbean. More remarkable here are long-term variations, as noted in much earlier literature (Portig, 1957, 1958; Hastenrath, 1963a, 1963b, 1967a, 1967b, 1976). The recent development and effective accessibility of high-quality long-term data bases invited the present exploration of pertinent circulation mechanisms. Section 2 presents some background, Section 3 describes data and methods, Section 4 deals with interannual variability, Section 5 explores the long-term variations, Section 6 addresses other key regions and circulation systems in the tropical Atlantic sector, and a synthesis is offered in the closing Section 7.
The annual cycle of circulation in the realm of the tropical Atlantic and eastern Pacific is extensively documented in our atlas (Hastenrath and Lamb, 1977). Some orientation is given in Figure 1. From the North Atlantic subtropical high, the Northeast tradewinds (Figure 1(c)) blow into a low pressure trough (Figure 1(b)) hydrostatically induced by a band (Figure 1(a)) of warmest sea surface temperature (SST), where they meet the cross-equatorial airstream from the southern hemisphere. This confluence entails convergence, the intertropical convergence zone (ITCZ). The confluence of airstreams from the two hemispheres and the ITCZ also dominate the eastern Pacific. The map of vertical motion (Figure 1(d)) shows strongest subsidence over the North Atlantic subtropical high (Figure 1(b)) and strongest upward motion in the realm of the ITCZ, with SST maximum, low pressure trough, and wind confluence (Figure 1(a), (b), and (c)).
The band of warmest waters, low pressure trough, and convergence are located farthest South in boreal winter and are displaced northward in boreal summer. May to October is the rainy season in Central America and the Caribbean. The ITCZ over the eastern Pacific reaches its northernmost position in May–June and September–October, with a retreat southward in July–August (JA), as already discovered by Alpert (1945) and further documented by Hastenrath and Lamb (1977) and Hastenrath (2002). The upward motion at 500 mb over the Caribbean is reduced in JA, as compared with May–June and September–October (Hastenrath, 1966, 1967a, 1968). Concurrently, the precipitation activity in Central America and the Caribbean peaks around May–June and September–October, with drier conditions in JA (Hastenrath, 1967a, 2002). This is distinctly different from the annual cycle in West Africa, where the ITCZ reaches its northernmost position around July–August–September, and the regions further south experience double rainfall peaks earlier and later in the summer half-year, associated with the double passage of the ITCZ (Hastenrath, 1985, pp. l60–166).
The roots of this study reach back half a century. On a walk around the Laguna de Chanmico (Figure 2(b)) in the countryside of El Salvador in 1961, I followed a path leading to below the water level and with trunks of trees flooded. Inquiries from people living in the area indicated that the water level dropped from the mid 1930s to 1947. Subsequent investigations, reported before (Hastenrath, 1963a, 1963b, 1967b), yielded complementary findings for other lakes: the Laguna del Llano dropped from the second half of the 1930s to about 1947–1950; the Lago de Coatepeque stood low around 1930, rose to the mid 1930s, and then dropped to around 1948; the Laguna de Apastepeque rose from the early 1930s to about 1934, then dropped to 1949, and rose again to 1958. The passage of a hurricane in June 1934 was widely remembered still in the early 1960s for the torrential floods it brought. While this single and for El Salvador rare weather event may not be responsible for the observed long-term variations in lake level, it is of interest in the context of the large-scale circulation setting. Stimulated by these findings from El Salvador, evidence was collected from adjacent countries of Central America and the Caribbean (Hastenrath, 1967b): in Guatemala, Lake Amatitlan stood low around 1961; in Nicaragua, Lago de Managua stood high around 1932 and Lago de Nicaragua around 1933. Later work (Hastenrath, 1976) collected and evaluated raingauge records in Central America and the Caribbean (Figures 1(a), 2(a), and 3). This largely confirmed the alternation of protracted dry and wet regimes indicated by the lake-level observations. Portig (1957, 1958) endeavoured an understanding of precipitation variations on both synoptic and long-term timescales in terms of the large-scale circulation.
3. Data and methods
An index CRB (Carib) of annual rainfall in the Central American–Caribbean region has been compiled from 48 stations as all-station average normalized departures for the series 1921–1974 (Hastenrath, 1976) and later updated to 1986. An index CGP (Central Great Plains) of annual precipitation over the period 1921–1986 on the US Central Great Plains is the average of totals at ten stations obtained from High Plains Regional Climate Center, University of Nebraska, Lincoln (http:www.phrcc.unl.edu). An index SHL (Sahel) of annual rainfall in the Sahel zone of sub-Saharan Africa is available from Nicholson (1985) and Lele and Lamb (2010).
Fields of SST are available from the Extended Reconstructed Sea Surface Temperature (ERSST.V3) data set (Smith et al., 2008), with a spatial resolution 2° latitude-longitude squares, for the years 1901–2008. This source was used for map analysis and also the compilation of an index PAC (Pacific) of JA SST in the equatorial central Pacific, 10°N–10°S, 180–190°W.
Fields of 1000-mb height Z, U, and V wind components and omega 500-mb vertical motion, with a 2° latitude-longitude resolution for the years 1921–1986, were obtained from the Twentieth Century Reanalysis V2 data set provided by the NOAA/OAR/ESRL/PSD, Boulder, Colorado, USA, from their website at http://www.esrl.noaa.gov/psd/ (Compo et al., 2006, 2011).
The fields of SST, Z, wind, and vertical motion were analysed for JA. Interannual variability was explored with correlation, presented in Section 4, Figures 4 and 5, and Table I. Concerning long-term variations, extreme regimes were identified from the CRB precipitation series as follows: 1931–1938 (Wa, WET), 1939–1947 except 1942–1943 (Da, DRY), 1950–1956 (Wb, WET), and 1971–1978 (Db, DRY). Differences in the fields between contrasting regimes were mapped and significance calculated by t-test, as presented in Section 5 and Figures 6, 7, and 8. A complementary evaluation in Section 6 is based on pertinent indices. For northern (N) and southern (S) domains of the tropical Atlantic, indices were compiled for JA and April–May (AM) of T (SST) and Z, denoted as TN, TS, dT = TN − TS, ZN, ZS and dZ = ZN − ZS. An index NEB (Northeast Brazil) of March–April–May–June rainfall in Brazil's Nordeste is available from earlier work (Hastenrath et al., 2009). From the archive of the North Atlantic Oscillation (NAO) accessible at http://www.cru.uea.ak.uk/cru/data/nao.htm, series were compiled for JA and AM. The archive of Southern Oscillation (SOI) available from http://reg.bom.au/climate/current/soihtml.shtml is the source for JA and AM SOI values.
Table I. Matrix of correlation coefficients for 1921–1986, in hundredths
4. Interannual variations
The variations of circulation and rainfall over the time span 1921–1986 are explored in Table I and Figures 3, 4, and 5. Consistent with Figure 3, Table I shows for index CRB (Figure 3(a)) a strong inverse correlation with SST in the equatorial Pacific PAC (Figure 3(c)), modest negative correlation with precipitation in the US Central Great Plains CGP (Figure 3(b)), and modest positive correlation with Sahel rainfall SHL. Further, CGP has a modest inverse correlation with SHL and weak positive with PAC. These are associations, while causalities are explored in the correlation maps shown in Figures 4 and 5.
Figure 4 presents maps of correlation with the rainfall in Central America and the Caribbean, CRB. For abundant rainfall, Figure 4(a) shows positive correlations, or anomalously warm waters, extending in a band from the coast of West Africa into the Caribbean and Gulf of Mexico, plausibly favourable for rainfall, and negative correlations or cold waters in the eastern Pacific. Figure 4(b) features negative correlations or low pressure over much of the tropical North Atlantic except the northeastern extremity of the map area and positive correlations or high pressure over the eastern Pacific, hydrostatically plausible with the negative SST correlations in Figure 4(a). Regarding the wind field, Figure 4(c) indicates Atlantic tradewinds weaker in the south and stronger in the north, broadly consistent with the pressure patterns conveyed by Figure 4(b); over the eastern Pacific, consistent with the higher pressure apparent in Figure 4(b), flow from the south is enhanced, favouring a more northward position of the ITCZ. The map of vertical motion shown in Figure 4(d) features over the eastern equatorial Pacific enhanced subsidence, consistent with the cold waters and increased surface pressure shown in Figure 4(a) and (b); enhanced upward motion is found over the tropical North Atlantic and much of the Caribbean. In context, the maps in Figure 4 depict circulation characteristics plausibly favourable for rainfall in Central America and the Caribbean. Consistent with Figure 4 are patterns found in earlier studies (Hastenrath, 1984; Hastenrath et al., 1987).
Figure 5 presents maps of correlation with the index of SST in the equatorial Pacific PAC. For warm PAC, Figure 5(a) shows positive correlations not only in the eastern Pacific but also in much of the tropical North Atlantic. Figure 5(b) features positive correlations, or low pressure, over much of the tropical North Atlantic, contrasting with negative values at the northern extremity of the map area and negative values over the eastern Pacific. Figure 5(c) displays for warm PAC reduced/enhanced easterlies over the northern/southern part of the tropical North Atlantic and reduced southwesterlies over the eastern Pacific. For the eastern Pacific causalities look straightforward: with warm waters comes lower pressure and with that reduced southwesterly wind. For the Atlantic sector such interrelationships appear to be less obvious, but may be understandable from earlier empirical and modelling work on teleconnections (Kucharski et al., 2008). This shows that with anomalously warm waters in the equatorial central Pacific, an upper tropospheric wave train develops from the Pacific to the North Atlantic. There it entails reduced subsidence in the north and reduced upward motion in the south; this serves to decrease the surface pressure in the north and increase it in the south, resulting in weaker meridional pressure gradient and thus weaker tradewinds; the weaker windstress forcing has as consequence warmer surface waters. The map of vertical motion shown in Figure 5(d) features over the eastern equatorial Pacific enhanced upward motion, consistent with the warm waters and reduced surface pressure as in Figure 5(a) and (b); in the tropical Atlantic sector, enhanced subsidence is found in the south and prevalence of enhanced upward motion in the north. Overall, the ensemble of correlation maps shown in Figure 5 appears plausible. Furthermore, with reference to Figure 5, one should now try to understand the causalities for the strong inverse correlation between CRB and PAC, as shown in Table I: in the eastern Pacific, warm surface waters hydrostatically induce low pressure, which leads to weaker southerly wind component and more southerly position of the ITCZ; moreover, in the Atlantic sector, the enhanced surface pressure and easterlies are unconducive to precipitation activity. Thus, one may appreciate the causalities through which anomalously warm Pacific waters may affect the rainfall in Central America and the Caribbean. Consistent with Figure 5 are patterns found in earlier studies (Hastenrath, 1984; Hastenrath et al., 1987).
5. Long-term variations
Figure 3(a) reveals for CRB alternations between protracted and contrastingly extreme regimes of rainfall, namely WET Wa in 1931–1938 and Wb in 1950–1956, and DRY Da in 1939–1947 and Db in 1971–1978. Applying the procedure developed in a recent article in this journal (Hastenrath and Polzin, 2010), we analyse the alternation of circulation characteristics in the succession between the contrasting protracted rainfall regimes, Wa—Da—Wb—Db.
Figure 6 presents the circulation differences between the DRY regime Da 1939–1947 minus the WET Wa 1931–1928. Figure 6(a) shows warming in a band from the coast of West Africa into the Gulf of Mexico and some cooling in the Caribbean. Figure 6(b) features pressure drop over Central America and the Caribbean, some rise in the central part of the tropical North Atlantic, and a drop in the northeastern part of the map area. Consistent with this is Figure 6(c) of the wind field, featuring enhanced tradewinds over the Caribbean and northwesterly wind differences over the central part of the tropical North Atlantic. Figure 6(d) shows greatly enhanced subsidence over the central part of the tropical North Atlantic, whereas Figure 6(c) indicates enhanced pressure. In a comparison of Figure 6versus 4, the sign of correlation should plausibly be inverted. Thus, the SST pattern in Figure 6(a) is remarkably consistent with Figure 4(a), as are the patterns of pressure in Figures 6(b) and 4(b), of wind in Figures 6(c) and 4(c), and of vertical motion in Figures 6(d) and 4(d).
Figure 7 presents the differences between the WET regime Wb 1950–1956 minus the DRY Da 1939–1947. Figure 7(a) shows favourable warming in the Caribbean and the northern edge of the map area and some cooling in the central part of the tropical North Atlantic and in the eastern Pacific. Figure 7(b) features pressure rise in the northeastern part of the map area, a drop in the central part of the tropical North Atlantic, and a rise in the eastern equatorial Pacific. Consistent with this is Figure 7(c) of the wind field, featuring decreased tradewinds over the Caribbean and adjacent Atlantic, with enhanced easterlies further North. Likewise, comparison of Figure 7(d) and (b) shows enhanced downward/upward motion in the domains of increased/decreased pressure. In particular, Figure 7(b), (c), and (d) are broadly consistent with the correlation maps shown in Figure 4(b), (c), and (d); less pronounced are the similarities between the SST maps shown in Figures 7(a) and 4(a). Comparison of Figures 6(b) and 7(b) shows for Da minus Wa negative/positive pressure differences in the northeastern/central part of the tropical North Atlantic and broadly inverse pattern for Wb minus Da. Likewise, broadly inverse patterns are found in Figures 6(a) and 7(a) for SST, in Figures 6(c) and 7(c) for the wind field, and in Figures 6(d) and 7(d) for vertical motion. This is noteworthy in relation to Portig's (1957, 1958) perception that zonal shifts in the position of the North Atlantic subtropical high affect the rainfall in the Caribbean and Central America.
Figure 8 presents the circulation differences between the DRY regime Db 1971–1978 minus the WET Wb 1950–1956. Figure 8(a) features marked cooling most pronounced in a band extending from West Africa into the Caribbean and the eastern Pacific except its equatorial region. Figure 8(b) shows pattern characteristics indicative of some hydrostatic forcing from the SST pattern shown in Figure 8(a), in particular a band of pressure increase over the eastern tropical Atlantic and Central America and the Caribbean. Plausibly consistent with Figure 8(b) is the map in Figure 8(c) of the wind pattern, featuring enhanced tradewinds over the southern part of the tropical North Atlantic extending into the Caribbean and reduced southwesterlies over the eastern Pacific. Figure 8(d) shows some consistency with Figure 8(c), with enhanced subsidence over the tropical North Atlantic, especially in a band extending from West Africa to the Caribbean. In a comparison of Figure 8versus 4, the sign of correlation should plausibly be inverted. Thus, the maps show similarities of patterns. Again with reference to the early work of Portig (1957, 1958) comparison of Figures 7(b) and 8(b) shows for Wb minus Da positive/negative pressure differences in the northeastern/central part of the tropical North Atlantic, and broadly inverse patterns are found in Figures 7(a) and 8(a) for SST, in Figures 7(c) and 8(c) for the wind field, and Figures 7(d) and 8(d) for vertical motion. Compared with Figures 6 and 7, the maps in Figure 8 exhibit more extensive high statistical significance. It may be conjectured that this may in part reflect the higher quality of the observational data sets in the second half of the twentieth century.
Comparing the rainfall in Central America and the Caribbean with that in the US Central Great Plains (Figures 2 and 3 and Table I), some tendency for inverse variations is apparent. This raises interest in the literature on droughts in the US Central Great Plains. Following the early studies of Namias (1955, 1983), the subject has recently received renewed attention (Cole et al., 2002; Hoerling and Kumar, 2003; Hoerling et al., 2009; Schubert et al., 2004). In modelling experiments, Schubert et al. (2004) found for anomalously cold waters in the equatorial Pacific 200-mb topography standing high in the midlatitudes and low in the tropics and drought conditions in the US Great Plains. Hoerling et al. (2009) found for 1946–1956 cold Pacific SST associated with drought in the Southern Plains and adjacent Southwest. In contrast, they concluded that ‘drought severity over the Northern Plains during 1932–1939 was likely triggered instead by random atmospheric variability’. Thus, the US Central Great Plains versus Central America and the Caribbean may share some common mechanisms of inverse variability of rainfall, no unique causalities.
6. Complementary exploration
The present appraisal of long-term variations in Central America and the Caribbean invites complementary consideration of other key regions in the tropical Atlantic sector, namely the West African Sahel and Brazil's Nordeste (Figure 9). In this section, the three regions shall be referred to as Carib, Sahel, and Nordeste, respectively. The annual cycle and interannual variability of circulation in the tropical Atlantic sector and rainfall variations in key regions have been explored in earlier work (Hastenrath and Lamb, 1977; Hastenrath, 1985; Hastenrath et al., 1987, 2009; Kucharski et al., 2008; Lele and Lamb, 2010). The rainy season in Carib and Sahel is centred around JA and in Nordeste around AM. The rains in Carib and Sahel are favoured by extreme northerly and in Nordeste by extreme southerly position of the ITCZ in the course of the annual cycle. The latitude position of the wind confluence is controlled by the low pressure trough and warmest ocean surface, as also illustrated in Figure 1. Accordingly, the SST and sea-level pressure in the northern (N) and southern (S) parts of the tropical Atlantic are of immediate interest, with delineation of domains in Figure 9 as in our previous paper in this journal (Hastenrath and Polzin, 2010). Information is represented by the SST indices TN, dT, and pressure indices ZN, dZ, considered in Table II for JA and Table III for AM.
Table II. Matrix of correlation coefficients in hundredths, 1921–1986: July–August.a
The relevance of SST variations in the tropical Atlantic for rainfall in key regions has been recognized in earlier literature. Regarding Brazil's Nordeste, a chain of teleconnections has been identified, involving also the SOI (Kucharski et al., 2008). Results on the Sahel have been presented in a recent paper in this journal (Hastenrath and Polzin, 2010). Extensive literature has been devoted to meridional overturning circulation and thermohaline circulation (Bryden et al., 2005; Knight et al., 2005; Zhang and Delworth, 2005; Schott et al., 2006; Cunningham et al., 2007; Delworth and Zeng, 2008; Richardson, 2008; Zhu and Jungclaus, 2008; Legler, 2009). However, this extensive literature has not produced evidence that could account for the observed long-term variations in the SST field of the tropical Atlantic.
Tables II and III exhibit associations between rainfall in three key regions and circulation characteristics in the tropical Atlantic sector. Warm/cold TN and low/high ZN and strong/weak dT and dZ favour a far northerly/southerly position of the ITCZ, and thus wet/dry conditions in Carib and Sahel versus dry/wet in Nordeste. The correlations in Tables II and III should be appreciated accordingly. Thus, in the JA season (Table II), SHL has correlations negative with ZN and dZ and positive with dT. For CRB correlations are of the same sign yet smaller, as other long-term processes explored in Section 5 prevail. In the AM season (Table III), NEB has correlations broadly inverse to those of CRB and SHL, consistent with the seasonal migration of equatorial trough and ITCZ.
Correlations with NAO, related to North Atlantic subtropical high, are presented in the second lowest line of Tables II and III. Correlations are near zero, except for a small positive value in JA with CRB, conceivably related to pressure near the Azores.
Correlations with SOI are shown in the lowest line of Tables II and III. For the JA season (Table II) and SHL, the sign of correlations is consistent with that for PAC given in Table I, and with patterns found in earlier research (Hastenrath et al., 1987), although the chain of causalities has not yet been identified. For the JA season and CRB, the correlation is understood from the role of SST in the equatorial Pacific manifest from the exploration in the preceding Section 5. For the AM season (Table III) and NEB, earlier diagnostic analysis (Kucharski et al., 2008) has identified the chain of causalities leading from the January SST departures in the equatorial Pacific to the SST and pressure departures in the tropical North Atlantic that affect the latitude position of equatorial trough and ITCZ and thus Nordeste rainfall. On the basis of the empirical circulation diagnostics, our decade-long seasonal forecasting in real time (Hastenrath et al., 2009) had among its four predictors as least an index of January SST in the equatorial Pacific, PAC.Jan, a representative of the SOI (correlation = − 0.67). The positive correlations between SOI and NEB in the preceding months of January–February–March are + 0.20, + 0.25, and + 0.23, much stronger than the value for AM shown in Table II.
The subtropical high pressure belt, tradewinds, and ITCZ are displaced northward in boreal summer, which is the rainy season in Central America and the Caribbean, with peaks around May–June and October–November and reduced precipitation in JA. The region experiences interannual variability and alternations between protracted multiannual dry and wet regimes. Favourable for abundant rainfall in the region are warm surface waters, low pressure, and weak tradewinds on the Atlantic side and over the eastern Pacific enhanced southerlies and a northward displaced ITCZ. A more northerly position of the circulation systems in the Atlantic sector is also conducive to rainfall in the Sahel, so that a modest positive correlation between rainfall in the two regions is plausible. Warm waters in the equatorial Pacific hydrostatically entail low pressure and thus reduced southwesterlies and a more southerly ITCZ position, unfavourable for rainfall in the region. With such background of interannual variability, most remarkable is the alternation of protracted successive dry and wet regimes Da, Wa, Db, and Wb.
From Wa to Da, the waters cooled in the Caribbean and became warmer in a band from the coast of West Africa into the Gulf of Mexico. The pressure rose in the central part of the tropical North Atlantic, and consistent with that the tradewinds over the Caribbean accelerated, unconducive to precipitation activity. The rainfall in the US Central Great Plains changed from the 1930s to the 1940s broadly inversely to that in Central America and the Caribbean. While the literature recognized that the causalities remain uncertain, it may be noted that the observed westward shift of the Atlantic subtropical high may have been a contributing factor.
From Da to Wb, the waters warmed in the Caribbean and cooled in the eastern equatorial Pacific. The pressure rose in the northeastern, dropped in the central part of the tropical North Atlantic, and rose over the eastern equatorial Pacific. Consistent with this, the tradewinds over the Caribbean and adjacent Atlantic became weaker. The Db minus Wb differences in the SST, pressure, and wind fields are in the sense characteristic of interannual variability of rainfall in the Caribbean and Central America.
From Wb to Db, the waters cooled in a band from the coast of West Africa to the Caribbean and warmed in the eastern equatorial Pacific. The pressure rose most strongly in a band from West Africa to the Americas, and consistent with that the tradewinds over the Atlantic and the Caribbean accelerated, unconducive to precipitation activity. As for Wa to Da and Da to Wb, the Wb to Db differences in the pressure and wind fields are in the sense characteristic of interannual variability of rainfall in Central America and the Caribbean.
Variations of meridional SST and pressure gradients in the tropical Atlantic are associated with rainfall anomalies in Central America and the Caribbean, the West African Sahel, and Brazil's Nordeste. The NAO shows no association with the rainfall variations in Central America and the Caribbean, in the West African Sahel, nor Brazil's Nordeste. The SOI has correlations with rainfall in all three regions, although for the Sahel, the chain of causalities remains to be ascertained.
In context, the analysis identifies circulation patterns associated with and forcing the interannual and longer term variability of rainfall in Central America and the Caribbean. Essential for this exploration was the creation and effective accessibility of valuable novel data sets.
This study was supported by the Variability of Tropical Climate Fund of the University of Wisconsin Foundation. Essential was the effective access to two recent valuable data sets: the Twentieth Century Reanalysis V2 data provided by the NOAA/OAR/ESRL PSD for the information on pressure, wind, and vertical motion and the ERSST.V3 for data of sea surface temperature. We fondly recall exchanges of thought with colleagues in the region in the past century.