1. Introduction
Low-level jets have been identified in a variety of locations around the globe including North America [i.e., the Great Plains low-level jet (GPLLJ); Bonner 1968; Ting and Wang 2006], South America (Virji 1981; Berbery and Collini 2000), and the easterly low-level jet of the Caribbean (CLLJ; Amador 1998; Amador et al. 2000). Interest in the CLLJ has increased in the past decade because of the importance of the CLLJ in transporting moisture from the tropical Atlantic into the Caribbean Sea, Gulf of Mexico, and the continental United States, hence, influencing rainfall both locally in the Caribbean and Central America and remotely in the United States (Mestas-Nuñez et al. 2007; Wang 2007; Amador 2008; Cook and Vizy 2010).
The CLLJ is a localized amplification of the large-scale circulation of the North Atlantic subtropical high (NASH) and is located in the Caribbean Sea, between northern South America and the islands of the Greater Antilles (13°–17°N, 70°–80°W, Fig. 1a). Unlike the GPLLJ, the CLLJ is present throughout the year and has two maxima (February and July) and two minima (May and October) in its semiannual cycle (Wang and Lee 2007; Wang 2007; Muñoz et al. 2008; solid line Fig. 2a). For most of the year the maximum wind speed (up to 16 m s−1) is located at 925 hPa, but as the jet weakens from its July peak, the jet maximum moves upward and deepens (Cook and Vizy 2010). The boreal summer maximum is related to the strengthening and westward extension of the NASH (Wang and Lee 2007; Wang 2007; Muñoz et al. 2008; Cook and Vizy 2010). During boreal summer, the CLLJ splits into two branches. The southerly branch connects with the GPLLJ (Cook and Vizy 2010) and an easterly branch traverses Central America (Fig. 1a). The boreal winter CLLJ maximum may be in part due to increased heating over northern South America associated with the South American monsoon (Cook and Vizy 2010).
Seasonal mean JJA 925-hPa wind speed (shaded contours, 2, 4, 6, 8, and 10 m s−1) and direction (vectors) from (a) NCEP–DOE reanalysis 2 and four different IPCC AR4 simulations: (b),(c) coupled CMIP simulations and (d),(e) uncoupled AMIP simulations. Only two models, MIROC3.2(hires) in (b),(d) and MRI in (c),(e) are shown for brevity. In (a), the white box indicates the region for calculating the CLLJ index and the thick black box indicates the region for calculating the GPLLJ index. Thin black box indicates the averaging region for Caribbean area-averaged quantities. NASH indicates the approximate climatological center of the North Atlantic subtropical high. Wind vector sizes are not consistent between panels.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Seasonal mean JJA 925-hPa wind speed (shaded contours, 2, 4, 6, 8, and 10 m s−1) and direction (vectors) from (a) NCEP–DOE reanalysis 2 and four different IPCC AR4 simulations: (b),(c) coupled CMIP simulations and (d),(e) uncoupled AMIP simulations. Only two models, MIROC3.2(hires) in (b),(d) and MRI in (c),(e) are shown for brevity. In (a), the white box indicates the region for calculating the CLLJ index and the thick black box indicates the region for calculating the GPLLJ index. Thin black box indicates the averaging region for Caribbean area-averaged quantities. NASH indicates the approximate climatological center of the North Atlantic subtropical high. Wind vector sizes are not consistent between panels.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Seasonal mean JJA 925-hPa wind speed (shaded contours, 2, 4, 6, 8, and 10 m s−1) and direction (vectors) from (a) NCEP–DOE reanalysis 2 and four different IPCC AR4 simulations: (b),(c) coupled CMIP simulations and (d),(e) uncoupled AMIP simulations. Only two models, MIROC3.2(hires) in (b),(d) and MRI in (c),(e) are shown for brevity. In (a), the white box indicates the region for calculating the CLLJ index and the thick black box indicates the region for calculating the GPLLJ index. Thin black box indicates the averaging region for Caribbean area-averaged quantities. NASH indicates the approximate climatological center of the North Atlantic subtropical high. Wind vector sizes are not consistent between panels.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Observed (solid), CMIP mean (dashed), and AMIP mean (dotted) annual cycle of various quantities. (a) Zonal wind; (b) precipitation with averaging area 10°–25°N, 90°–55°W (thin black box in Fig. 1a); (c) SLP, and (d) SLP gradient at 12.5°–17.5°N, 70°–80°W (white box in Fig. 1a); and (e) SST and (f) SST gradient at 12°–16°N, 70°–80°W.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Observed (solid), CMIP mean (dashed), and AMIP mean (dotted) annual cycle of various quantities. (a) Zonal wind; (b) precipitation with averaging area 10°–25°N, 90°–55°W (thin black box in Fig. 1a); (c) SLP, and (d) SLP gradient at 12.5°–17.5°N, 70°–80°W (white box in Fig. 1a); and (e) SST and (f) SST gradient at 12°–16°N, 70°–80°W.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Observed (solid), CMIP mean (dashed), and AMIP mean (dotted) annual cycle of various quantities. (a) Zonal wind; (b) precipitation with averaging area 10°–25°N, 90°–55°W (thin black box in Fig. 1a); (c) SLP, and (d) SLP gradient at 12.5°–17.5°N, 70°–80°W (white box in Fig. 1a); and (e) SST and (f) SST gradient at 12°–16°N, 70°–80°W.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Although this study will focus predominantly on the local structure and characteristics of the CLLJ in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) models, there are many connections between the CLLJ and SSTs in both the Atlantic and Pacific. The relationship between ENSO and the CLLJ varies seasonally. During the summer, when the CLLJ is at a maximum, a strong CLLJ occurs in conjunction with warm Pacific SST anomalies, through teleconnections with SLP in the Atlantic (Amador et al. 1999, 2000, 2006; Wang 2007; Amador 2008). Atlantic SSTs may be related to the CLLJ through the North Atlantic Oscillation, again through the SLP (Wang 2007). In addition to these remote influences, the CLLJ has also been shown to be associated with the Madden–Julian oscillation (Martin and Schumacher 2011a).
As noted by Wang (2007) and Cook and Vizy (2010), the CLLJ is geostrophic to first order, thus it is controlled by gradients in pressure (geopotential). As the NASH expands and strengthens the meridional pressure gradient across the Caribbean Sea is increased and the CLLJ strengthens (Wang 2007; Muñoz et al. 2008; Whyte et al. 2008; Cook and Vizy 2010). The opposite occurs when the CLLJ contracts. Wang (2007) and Muñoz et al. (2008) show that the semiannual cycle of the CLLJ is in phase with the semiannual cycle of the meridional SLP gradient across the Caribbean Sea. Meridional gradients in sea surface temperature (SST) across the region also show a semiannual cycle that is in phase with that of the CLLJ. This suggests a feedback between the atmosphere and the ocean that acts to reinforce the CLLJ through the effect of opposite values of wind stress curl and hence upwelling on either side of the jets zonal axis (Wang 2007). Using an idealized modeling study, Wang et al. (2008) also showed that the magnitude of the SST anomaly in the Caribbean influences the strength of the CLLJ through interactions with the NASH. This result is also echoed by the results of Rauscher et al. (2011). When Caribbean SST is anomalously warm (cold), the CLLJ is anomalously weak (strong).
Annual rainfall in the Caribbean exhibits a bimodal structure (solid line in Fig. 2b), with an initial maximum in May, a minimum around July–August, and a second maximum in September–October (Jury et al. 2007; Gamble et al. 2008). The minimum that separates the two peaks in rainfall has been termed the midsummer drought (MSD) based on a similar minima of rainfall that occurs on the Pacific coast of Central America (Magaña et al. 1999). The use of MSD in this study will be specific to the Caribbean. The minimum in rainfall associated with the MSD occurs simultaneously with the summer maximum of the CLLJ (solid line in Fig. 2a). It is postulated that the CLLJ is a major contributor to the MSD through the increase in moisture flux divergence in the Caribbean, which acts to suppress convection and decrease rainfall (Magaña et al. 1999; Wang 2007; Muñoz et al. 2008; Whyte et al. 2008). An anomalously strong CLLJ is also associated with reduced precipitation over the Caribbean throughout the year (Cook and Vizy 2010). While drying is seen across much of the Caribbean when the CLLJ is strong, the Caribbean coast of Central America has enhanced rainfall (through orographic enhancement and large-scale low-level convergence at the jet exit), which deprives the Pacific coast of Central America of moisture and decreases rainfall (Amador 1998; Cook and Vizy 2010).
Observational studies of the CLLJ have increased in recent years, including a field experiment in 2001 called Experimento Climático en las Albercas de Agua Cálida (ECAC; Amador 1998), but there have been limited investigations into its representation in regional or global climate models (GCMs), such as the IPCC AR4 models. Wang et al. (2008) used the National Center for Atmospheric Research (NCAR) Community Atmosphere Model (CAM) to investigate the influence of the size of the Atlantic warm pool (AWP) on summer climate, including the CLLJ, as discussed previously. The representation of the climate of the Caribbean by the IPCC AR4 coupled models is somewhat lacking, with SSTs too cold and precipitation underestimated by up to 30% (Neelin et al. 2006; Christensen et al. 2007). However, the role of the CLLJ in these simulations has not been examined in detail. The relationship between the CLLJ, moisture transport, and precipitation both locally and remotely (including extremes and the MSD; Magaña et al. 1999; Wang 2007; Cook and Vizy 2010; Durán-Quesada et al. 2010; Martin and Schumacher 2011a), in addition to its impact on easterly waves and tropical storms (Serra et al. 2010), highlight the importance of having a realistic CLLJ in GCMs.
This paper will begin with a brief discussion of the observational and IPCC AR4 data (section 2), followed by an analysis of CLLJ properties (section 3) and the connection between the CLLJ and Caribbean precipitation (section 4). Results related to the simulation of the CLLJ impact on the United States is shown in section 5, followed by conclusions in section 6.
2. Data and methodology
a. Observations
A variety of observational and reanalysis datasets are used to obtain the most accurate and spatially complete climatology of the CLLJ, with which the model output can be compared. All data is monthly resolution for the period 1979–2008. Wind (1000–600 hPa) and sea level pressure are obtained from the National Centers for Environmental Prediction (NCEP) Department of Energy (DOE) reanalysis II (Kanamitsu et al. 2002) at 2.5° resolution. Reanalysis data will be referred to as observational throughout this study. Precipitation data is from the Global Precipitation Climatology Project (GPCP), which is a combination of satellite products and gauge observations (Huffman et al. 2001). Monthly GPCP data is available at 2.5° resolution. The third dataset used was the Hadley Centre Sea Ice and Sea Surface Temperature dataset, version 1 (HadISST1) to investigate the links between the CLLJ and SST. The HadISST dataset is at 1° resolution and uses a combination of in situ and satellite observations to provide global coverage (Rayner et al. 2003).
Calculation of area mean values are used in this study to investigate the annual cycle and month-to-month variability of quantities associated with the CLLJ. Three different averaging regions are used: the entire Caribbean region in this study is defined as lying between 10°–25°N and 55°–90°W (thin black box in Fig. 1a), the CLLJ region is defined as 11°–17°N and 70°–80°W (white box in Fig. 1a), and the GPLLJ region is defined as 25°–35°N and 95°–100°W (thick black box in Fig. 1a). Meridional gradients in the CLLJ region are calculated at each longitude and then averaged over the CLLJ box. The same averaging regions are used for the model output. A CLLJ and GPLLJ index are calculated by removing the average annual cycle of zonal (CLLJ) or meridional (GPLLJ) winds from the time series. The CLLJ index is then multiplied by −1 to create an index where a stronger CLLJ is positive.
b. Model output
Output from 19 coupled ocean–atmosphere GCMs [i.e., the Coupled Model Intercomparison Project (CMIP) type] and 12 atmospheric only GCMs [i.e., the Atmosphere Model Intercomparison Project (AMIP) type] are examined. These models are a subset of those used as part of the IPCC AR4 and are available through the Program for Climate Model Diagnostics and Intercomparison (PCMDI). All output (wind, SLP, precipitation, and SST) are monthly mean values for the last 30 yr of the climate of the twentieth-century simulations (20C3M) in order to make a fair comparison with the observational data. All models are configured differently with varying resolutions and parameterizations. A summary of the models used in this study, as well as pertinent information concerning their configuration, is shown in Table 1. While some models had multiple ensembles for the 20C3M simulations, we use only one realization from each model to ensure model mean results are not skewed toward models with a large number of ensembles. All model output is kept at its original resolution, rather than interpolating onto a consistent grid, in order to make conclusions about the impact of resolution on the structure of the CLLJ in the models. Throughout this paper, for purposes of brevity, not all model output will be presented in every figure. Presented model output is chosen as representative of the entire model ensemble.
List of IPCC AR4 models used in this study. Flux correction indicates those models using heat (H), water (W), or no (N) flux correction. AMIP indicates which models also have atmosphere only (AMIP type) simulations. Further model details, including references, can be found at the PCMDI Web site (http://www-pcmdi.llnl.gov).
3. CLLJ properties
a. Location
All CMIP and AMIP models captured an enhancement of the low-level wind field in the Caribbean Sea between South America and the Great Antilles that was evident throughout the annual cycle. As all models were able to produce an enhancement of the low-level wind in the Caribbean, horizontal resolution was not a factor in the ability of models to accelerate the flow in the Caribbean. Climatological June–August (JJA) 925-hPa wind speed and direction from example model simulations are shown in Fig. 1. The location of the jet remained constant in other seasons and JJA was chosen as both the CLLJ and GPLLJ are strongest.
Numerous similarities are observed between all simulations in Figs. 1b–e and the observations in Fig. 1a. The strong easterly trade winds across the Atlantic strengthen as they enter the Caribbean Sea, forming a CLLJ. These winds then weaken close to Central America, turn northward across the Yucatan Peninsula, and form a southerly jet across the south-central United States (the GPLLJ). Details of the IPCC AR4 models representation of the springtime GPLLJ are presented in Cook et al. (2008), who show the ability of the models to produce a realistic GPLLJ.
Despite the clear similarities of the general flow structure in the region, maximum wind speeds in the CLLJ simulations sometimes varied significantly from observations. This variation in wind speeds will be addressed further in section 3b. Figures 1b–e also show that maximum wind speeds are reduced from CMIP to AMIP simulations, while the CLLJ location remained in the Caribbean.
A feature that is clearly seen in Figs. 1b,d and not in the reanalysis (Fig. 1a), is the overly strong easterly flow over Central America and in the east Pacific. The splitting of the CLLJ into an easterly and southerly component over Central America is observed during JJA; however, four of the CMIP models have winds at least 4 m s−1 larger than observations in this region (e.g., Figs. 1b,d). A lack of observations in this region may be influencing the reanalysis, as Amador (2008) shows the CLLJ to be strong (in excess of 20 m s−1 at 1.5 km in February) at one location in western Central America using data from the local sounding network.
b. Annual cycle
The CLLJ has a distinct semiannual cycle, with maxima in February and July as discussed in section 1. The seasonal cycle of zonal wind in the CLLJ region (white box in Fig. 1a) for observations and the CMIP (dashed) and AMIP (dotted) model means is shown in Fig. 2a. The CMIP and AMIP model means do not capture the semiannual cycle of the CLLJ; the simulated CLLJ remains almost uniform throughout January–July before reaching a minimum in September/October. The relative peak in CLLJ strength in July is not captured by either the coupled or uncoupled model mean. The weakened semiannual cycle is also seen in two GCMs in Amador (2008) and four in Amador et al. (2010). The magnitude of the AMIP model mean is also consistently less than that of the CMIP mean and is closer to the observed values, especially in the latter portion of the year.
Figure 2b shows the annual cycle of Caribbean area-averaged precipitation. The CMIP and AMIP models appear unable to accurately produce a MSD in July and August, while simultaneously not producing a peak in the CLLJ. The relationship between the CLLJ, rainfall, and the MSD will be addressed further in section 4. Also of importance is the underestimation of precipitation by the CMIP model mean and overestimation by AMIP model means, but that is outside the scope of this investigation (Martin and Schumacher 2011b).
The annual cycle of several quantities that have been shown to be important in the development and maintenance of the CLLJ, such as SLP and SST, as well as their meridional gradients across the region, are shown in Figs. 2c–f. During the observed July peak in CLLJ, SLP increases due to the expansion of the NASH (Fig. 2c) and SLP and SST gradients also have a distinct peak (Figs. 2d,f), as expected (Wang 2007; Muñoz et al. 2008). It is well know that the CMIP models underestimate SSTs in the Caribbean region (Misra et al. 2009) and this is clearly demonstrated in Fig. 2e; however, SST meridional gradients across the Caribbean Sea are overestimated, contributing to the overly strong CLLJ.
Both AMIP and CMIP means show a small peak in SLP during June (a month earlier than observed) and a large dip during September–October. However, the SLP gradient does not show a midsummer peak in either CMIP or AMIP model means. This uniformity of the SLP gradients from January to July is playing an essential role in the uniformity of the simulated CLLJ during these months. The magnitudes of the SLP gradients are similar to the observations, with the CMIP mean being consistently higher than the AMIP mean leading to the stronger CLLJ in the CMIP models.
The overestimation of the magnitude of the CLLJ in the CMIP models is consistent with results presented by Wang et al. (2008) based upon the theory of Gill (1980). The atmospheric response to an off-equatorial heating anomaly (such as anomalously warm SSTs in the Caribbean) is atmospheric Rossby waves resulting in low SLP to the northwest of the heating (Gill 1980). Hence, when considering the CMIP simulations, which have anomalously weak SSTs in comparison to observations (Fig. 2e), anomalously cold Caribbean SSTs during summer lead to a stronger NASH, a weaker continental low over Mexico and the southwest United States, stronger meridional SLP gradients across the Caribbean, and, hence, a stronger CLLJ. However, this theory does not adequately explain the January–June overestimate in CLLJ magnitude in the AMIP simulations, where SST is prescribed (i.e., not anomalously low as in the CMIP simulations).
While the simple theory of Gill (1980) can explain the overestimate of the CLLJ in CMIP simulations, the lack of a summer peak in magnitude is still unexplained. As seen in Fig. 2c, SLP in the Caribbean increases from a minimum in May to a maximum in July in conjunction with the increase in CLLJ magnitude. The spatial extent of this SLP increase is shown in Fig. 3, maps of the difference between May and July SLP for observations and a selection of models. SLP is observed to increase across much of the Northern Atlantic between May and July as the NASH strengthens (Fig. 3a). Expanding and increasing SLP westward into the Caribbean, Mexico, and Central America is essential for increasing the SLP gradient in the southern Caribbean and causing the observed summer peak in CLLJ magnitude. The models have a wide representation of this May–July SLP difference structure and results are presented from one more successful simulation {the Model for Interdisciplinary Research on Climate 3.2, high-resolution version [MIROC3.2(hires); Figs. 3b,d]} and one less successful simulation [the NCAR Community Climate System Model (CCSM); Figs. 3c,e].
July minus May SLP difference (hPa). Contours at −4, −3, −2, −1, 1, 2, 3, and 4 hPa with positive indicating an increase in SLP between May and July. (a) Observations are shown in conjunction with example output from two models: (b),(d) MIROC3.2(hires) and (c),(e) NCAR CCSM. Output from both (b),(c) CMIP and (d),(e) AMIP simulations are shown for each model.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
July minus May SLP difference (hPa). Contours at −4, −3, −2, −1, 1, 2, 3, and 4 hPa with positive indicating an increase in SLP between May and July. (a) Observations are shown in conjunction with example output from two models: (b),(d) MIROC3.2(hires) and (c),(e) NCAR CCSM. Output from both (b),(c) CMIP and (d),(e) AMIP simulations are shown for each model.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
July minus May SLP difference (hPa). Contours at −4, −3, −2, −1, 1, 2, 3, and 4 hPa with positive indicating an increase in SLP between May and July. (a) Observations are shown in conjunction with example output from two models: (b),(d) MIROC3.2(hires) and (c),(e) NCAR CCSM. Output from both (b),(c) CMIP and (d),(e) AMIP simulations are shown for each model.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
In general, all models captured the strengthening of the NASH in the northern Atlantic, although this strengthening was often too strong and displaced to the north of observations, as seen in Figs. 3c,e. Although all models showed strengthening of the NASH, the increase in SLP in the western Atlantic and Caribbean was problematic. In the majority of simulations, such as those for NCAR CCSM (Figs. 3c,e), little to no increase in SLP was observed in the western Atlantic and Caribbean. With little to no change in SLP and, hence, SLP gradients, the CLLJ strength remained relatively constant in these simulations. For the few models that did see a westward expansion of the NASH from May to July [e.g., MIROC3.2(hires); Figs. 3b,d], magnitudes were often slightly too large. Little improvement was observed between CMIP and AMIP simulations, with the same problems occurring in both simulation types, as seen in Fig. 3. Possible suggestions as to why this westward expansion is not accurately captured by the models include poor simulation of Amazon or Central American precipitation, which causes weaker overturning. The reduced subsidence over the western Atlantic thus reduces the SLP increase.
c. Vertical structure
The structure of the CLLJ in the vertical is another metric for assessing the models ability to accurately represent the CLLJ. All material in section 3 investigated the CLLJ structure only at 925 hPa. Figure 4a shows the annual cycle of the vertical structure of the zonal wind (averaged over the CLLJ area as shown by the white box in Fig. 1a). The maximum at 925 hPa throughout the year is evident, with a stronger and deeper CLLJ in July as compared to January/February. Minima are observed throughout the lower atmosphere in April and October, as expected from previous studies (Cook and Vizy 2010). Area-averaged winds reach a maximum of approximately 11 m s−1 between 950 and 850 hPa in July. Again, four simulation examples are shown in Figs. 4b–e, which represent the ensemble of model output.
Annual cycle (repeated twice) of the vertical (1000–600 hPa) profile of zonal wind averaged over the CLLJ index area (white box in Fig. 1a). The contour interval is 2 m s−1 up to −8 m s−1 and 1 m s−1 at higher wind speeds. Shading begins at −8 m s−1 and dotted contours indicate easterly winds. As in Fig. 3, (a) observations are shown in conjunction with example output from two models: (b),(d) GFDL_2_1 and (c),(e) NCAR CCSM. Output from both (b),(c) CMIP and (d),(e) AMIP simulations are shown for each model.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Annual cycle (repeated twice) of the vertical (1000–600 hPa) profile of zonal wind averaged over the CLLJ index area (white box in Fig. 1a). The contour interval is 2 m s−1 up to −8 m s−1 and 1 m s−1 at higher wind speeds. Shading begins at −8 m s−1 and dotted contours indicate easterly winds. As in Fig. 3, (a) observations are shown in conjunction with example output from two models: (b),(d) GFDL_2_1 and (c),(e) NCAR CCSM. Output from both (b),(c) CMIP and (d),(e) AMIP simulations are shown for each model.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Annual cycle (repeated twice) of the vertical (1000–600 hPa) profile of zonal wind averaged over the CLLJ index area (white box in Fig. 1a). The contour interval is 2 m s−1 up to −8 m s−1 and 1 m s−1 at higher wind speeds. Shading begins at −8 m s−1 and dotted contours indicate easterly winds. As in Fig. 3, (a) observations are shown in conjunction with example output from two models: (b),(d) GFDL_2_1 and (c),(e) NCAR CCSM. Output from both (b),(c) CMIP and (d),(e) AMIP simulations are shown for each model.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
All models were able to capture the CLLJ at 925 hPa. However, because of the overly strong CLLJ in almost all simulations, the depth of the CLLJ (approximated by the 8 m s−1 contour, which is shaded in Fig. 4) is too great, extending above 800 hPa throughout the first half of the year. The uniformity of the CLLJ annual cycle, as discussed in section 3b, is evident throughout the lower atmosphere in the majority of simulations and in all four example model outputs shown in Fig. 4.
A positive aspect of the simulations is the deepening of the CLLJ during July. This deepening is observed in all models despite the uniformity and too large magnitude of the CLLJ. In some cases (Figs. 4b,d), the reduction in depth of the CLLJ during April is also captured by the models, even though the magnitude at 925 hPa remains constant. Some models (Fig. 4c) have a secondary peak in summer around 600 hPa that is not seen in observations. Improvement in the strength of the CLLJ at all levels is seen in the AMIP simulations, with some models showing slight minima in April (Fig. 4d). However, the general structure of the annual cycle is similar between the two simulation types.
d. Variability
In addition to the climatological structure and annual cycle of the CLLJ, the interannual variability of the CLLJ was studied (Fig. 5). The interannual variability was calculated monthly using the standard deviation of the time series for each month (e.g., the standard deviation of all the Januarys, etc.). The interannual variability of the CLLJ index is presented in Fig. 5a. Similar to Wang (2007) and Muñoz et al. (2008), the largest variance occurs in September and October when the CLLJ is climatologically weak. This strong boreal fall variability may influence the development and track of easterly waves and tropical storms in the region from year to year (Wang 2007). The two peaks of variance in February and September seen in Wang (2007), who use NCEP–NCAR reanalysis, is seen in Fig. 5a but a third peak in May is also evident in the NCEP–DOE II reanalysis, which may be due to differences in averaging area and time periods between the two studies. The CMIP and AMIP model means, which are also shown in Fig. 5a, show the ability of the models to represent both the magnitude and annual cycle of the interannual variability of the CLLJ, although the fall peak is 1 month earlier in the AMIP mean and 1 month later in the CMIP mean.
As in Fig. 2, but for standard deviations.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
As in Fig. 2, but for standard deviations.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
As in Fig. 2, but for standard deviations.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
As seen in Figs. 5c–f, the dominant quantity affecting the interannual variability of the CLLJ is the SLP gradient (Fig. 5d), which shows a strong October peak and a similar annual cycle to the CLLJ index variability in Fig. 5a. Both the CMIP and AMIP model means underestimate the strength of this peak, but have early (late) peaks corresponding to the early (AMIP) and late (CMIP) peaks in the CLLJ index standard deviation. The timing of the peaks in standard deviation of precipitation in the models is also incorrect (Fig. 5b). The standard deviations of SLP, SST, and the SST gradient are relatively constant throughout the seasonal cycle (aside from a February peak in SLP), which is well represented by the model means. The exception is the magnitude of the SST gradient, which has a standard deviation in the CMIP model mean of approximately twice the observed value, indicating an additional problem with the SSTs in this region.
4. Relationship of the CLLJ with local rainfall
a. Annual correlations
The connection between the CLLJ and Caribbean rainfall has been previously identified, with an anomalously strong CLLJ associated with reduced precipitation over the Caribbean and enhanced precipitation along the Caribbean coast of central America (Amador 1998; Cook and Vizy 2010). This is illustrated by the significant (at 95% level) regression coefficient [−2.57 mm day−1 (m s−1)−1] between the annual CLLJ index and Caribbean area-averaged precipitation anomaly time series (Table 2). This significantly negative regression coefficient reinforces the theory that when the CLLJ is strong, moisture divergence in the Caribbean and transport out of the region is large, leading to a reduction in precipitation amounts.
Regression coefficients [mm day−1 (m s−1)−1] between the annual CLLJ index and the Caribbean area-averaged precipitation anomaly. Values in bold are significant at the 95% significance level and values in bold and italic are significant at the 99% significance level.
Despite the ability of both the CMIP and AMIP models to develop a CLLJ and precipitation with a similar structures to observations (e.g., Fig. 2), not all models produce a significant negative regression coefficient between the CLLJ index and Caribbean area-averaged precipitation anomalies. Table 2 shows that only six CMIP models have a significant negative regression coefficient and three CMIP models actually have small positive regressions (although they are not significant). Magnitudes of the regression values were often considerably larger than observed (e.g., −4.32 for mpi), suggesting that precipitation anomalies reduce too rapidly with an increasingly strong jet.
An improvement is seen in the AMIP models, with 10 out of 12 AMIP models producing significant negative regression values, none having positive coefficients, and magnitudes general closer to that of observations. This suggests that the errors in the relationship between the CLLJ and precipitation seen in the majority of CMIP models were not simply caused by errors in the atmospheric component of the model. The large CMIP and small AMIP errors suggest that the incorrect interaction between the ocean and atmosphere or the anomalously cold SSTs in the CMIP models are affecting the CLLJ–rainfall relationship. It is important to note that local, small-scale connections between the CLLJ and precipitation in the Caribbean are not captured by this annual, area-averaged approach. Higher resolution than is available from both observations, reanalysis, and GCM output would be necessary to further investigate these relationships.
b. The midsummer drought
The CLLJ has an important relationship with precipitation during the MSD, as discussed by Magaña et al. (1999), Wang (2007), Muñoz et al. (2008), and Whyte et al. (2008). In addition to the annual, area-averaged regression coefficients between the CLLJ index and rainfall anomalies, the spatial distribution of correlation values is also important for determining the simulation abilities of the models. The observed correlation map between the CLLJ index and precipitation anomalies for August is shown in Fig. 6a. August was chosen as it coincides with the MSD and a strong CLLJ variation. Results for July were similar but slightly weaker in this dataset. Observations show extensive negative correlations across most of the Caribbean, western tropical Atlantic, and Central America, except for the Caribbean coast of Central America (orographic and jet exit convergence influence is strong in this region), agreeing well with Cook and Vizy (2010).
Maps of correlation coefficients between precipitation anomalies and CLLJ index for August. (a) Observations; (b),(c) CMIP output; and (d),(e) AMIP output. Model output presented from (b),(d) MIROC3.2(medres) and (c),(e) NCAR PCM. Contour interval is 0.1, with negative correlations dashed (indicating increased precipitation anomalies with reduced CLLJ strength) and positive correlations solid. Shading indicates correlations significantly different from zero at the 95% confidence level.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Maps of correlation coefficients between precipitation anomalies and CLLJ index for August. (a) Observations; (b),(c) CMIP output; and (d),(e) AMIP output. Model output presented from (b),(d) MIROC3.2(medres) and (c),(e) NCAR PCM. Contour interval is 0.1, with negative correlations dashed (indicating increased precipitation anomalies with reduced CLLJ strength) and positive correlations solid. Shading indicates correlations significantly different from zero at the 95% confidence level.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Maps of correlation coefficients between precipitation anomalies and CLLJ index for August. (a) Observations; (b),(c) CMIP output; and (d),(e) AMIP output. Model output presented from (b),(d) MIROC3.2(medres) and (c),(e) NCAR PCM. Contour interval is 0.1, with negative correlations dashed (indicating increased precipitation anomalies with reduced CLLJ strength) and positive correlations solid. Shading indicates correlations significantly different from zero at the 95% confidence level.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Correlation maps from two different models are shown in Figs. 6b–e. These models were chosen to be representative of the model ensemble, with the Model for Interdisciplinary Research on Climate 3.2, medium-resolution version [MIROC3.2(medres); Figs. 6b,d] having a small negative annual regression coefficient in the CMIP run and a significant negative regression in its AMIP run. The NCAR Parallel Climate Model (PCM) model (Figs. 6c,e) had significant negative regression values in both simulations. It is seen from the model correlation maps that the models produce similar patterns to the observations. Significant negative correlation values across the Caribbean are simulated, although the region of negative correlation in the simulations is not as spatially extensive as the observations. Noticeable improvements are seen in the AMIP simulations in comparison to the CMIP simulations, with a larger spatial extent of the negative correlations across the entire domain. Similar results were observed for other months (not shown).
Primary deficiencies in the model simulations occur in Mexico and the east Pacific. Correlations are close to zero across much of Mexico in many simulations (e.g., Figs. 6b,c,d) and any negative correlations that are simulated are not as widespread as observations (e.g., Fig. 6e). An interesting region of positive correlations in the models (increased precipitation with a stronger CLLJ) is seen in the east Pacific off the coast of South America, which is not in the observations. One possible explanation is that the representation of the Central American terrain in the models is lacking, leading to excess moisture transport into the east Pacific. Model errors in this region may also be due to problems in simulation the low-level westerly Choco jet (Poveda and Mesa 2000; Durán-Quesada et al. 2010), which is an important moisture transport mechanism in the far eastern Pacific.
To further investigate the interaction between the CLLJ and MSD in the models, composite annual cycles of models with and without a MSD were calculated. Each IPCC AR4 model (both AMIP and CMIP) was categorized by whether it captured the Caribbean MSD. A simple definition for the MSD was established. A model was categorized as MSD if it simulated at least a 0.1 mm day−1 reduction in area-averaged precipitation between June and July, as the largest reduction in the area-averaged precipitation is seen between June and July (Fig. 2). The AMIP models better captured the MSD, with 7 of 12 categorized as MSD. Only 6 of 19 CMIP models fell into this category (as expected from Fig. 2b). The same variables in Fig. 2 were then calculated for these MSD and no-MSD model composites and are shown in Fig. 7.
Observed (solid), CMIP mean (dashed), and AMIP mean (dotted) annual cycle of quantities shown in Fig. 2. Red lines show means of models that captured the MSD and blue lines show means of models without a MSD.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Observed (solid), CMIP mean (dashed), and AMIP mean (dotted) annual cycle of quantities shown in Fig. 2. Red lines show means of models that captured the MSD and blue lines show means of models without a MSD.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Observed (solid), CMIP mean (dashed), and AMIP mean (dotted) annual cycle of quantities shown in Fig. 2. Red lines show means of models that captured the MSD and blue lines show means of models without a MSD.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
It is clear from Fig. 7b that the models that capture the MSD (red lines) not only do better at simulating summer rainfall, but the entire annual cycle is improved and both CMIP and AMIP MSD model means are closer to observations. The CLLJ annual cycle is also improved, with a more distinct maximum in July and magnitudes closer to the observations (Fig. 7a). As seen in observations, the CLLJ July peak coincides with the increase in SLP gradients and SST gradients across the southern Caribbean. The MSD composited models show a much better structure than the no-MSD composites (Figs. 7d,f), although SLP gradient changes are still weaker than observations.
By compositing the models by those that capture the MSD and those that do not, the representation of the SST in the CMIP models is greatly improved in the MSD composites (Figs. 7e,f). The models with a MSD have larger SSTs (although they are still approximately 1°C below observations) and greatly improved SST gradients in the southern Caribbean. Whether this improvement in SST gradient is the cause or effect of a better CLLJ simulation is not possible to determine from these simulations alone, although the August peak in SST gradient perhaps suggest the SST gradient is a response to the CLLJ rather than a generating factor. These results reiterate the importance of improving CMIP simulations of SST in this region.
5. Connection with U.S. climatology
a. The Great Plains low-level jet
Numerous studies have shown the importance of the CLLJ as a moisture and momentum source to the United States and the GPLLJ (Mo et al. 2005; Mestas-Nuñez et al. 2007; Wang et al. 2008; Cook and Vizy 2010). The observed CLLJ and GPLLJ indices (as described in section 2a and Fig. 1) were regressed against each other for each month of the year and the results are shown in Table 3. The significant positive regression coefficients in January–April agree with the results of Cook and Vizy (2010), in that the GPLLJ forms temporarily during these cold months when the CLLJ is strong and hence, a positive regression coefficient is observed. Positive values are also seen throughout the rest of the year (except November) and are significant in June, July, and September when both the CLLJ and GPLLJ are strong and the AWP is large (particularly in September).
Regression coefficients between the CLLJ index and the GPLLJ index for observations by month. Values in bold are significant at the 95% significance level and values in bold and italic are significant at the 99% significance level.
The connection between the CLLJ and the GPLLJ is important for the accurate simulation of the U.S. climate. Figure 8 shows scatterplots and accompanying regression lines between the February CLLJ and GPLLJ indices for observations, as well as each CMIP (Fig. 8a) and AMIP (Fig. 8b) ensemble members. Similar patterns and results were seen for other months (not shown). For the CMIP model ensemble (Fig. 8a), the majority of models (11 of 19) produce a significant positive regression coefficient between the CLLJ and GPLLJ. The AMIP ensemble (Fig. 8a) however, shows quite a different result, with only the minority (4 of 12) models producing a significant positive correlation despite the improved simulation of the February CLLJ in the AMIP models.
Scatterplots between February CLLJ and GPLLJ indices for (a) CMIP models and (b) AMIP models. Observations are shown in top left of (a) and (b). Regression line (red) is shown only if significant at the 95% level (based on Student’s t test). Averaging areas for indices as shown in Fig. 1a.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Scatterplots between February CLLJ and GPLLJ indices for (a) CMIP models and (b) AMIP models. Observations are shown in top left of (a) and (b). Regression line (red) is shown only if significant at the 95% level (based on Student’s t test). Averaging areas for indices as shown in Fig. 1a.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
Scatterplots between February CLLJ and GPLLJ indices for (a) CMIP models and (b) AMIP models. Observations are shown in top left of (a) and (b). Regression line (red) is shown only if significant at the 95% level (based on Student’s t test). Averaging areas for indices as shown in Fig. 1a.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
b. Relation with U.S. rainfall
The ability of the models to simulate the connection between the CLLJ and U.S. precipitation was also investigated for February, as the models produce a strong connection with the GPLLJ and Cook and Vizy (2010) have shown a link with U.S. precipitation in this month. The observed correlation between the CLLJ index and precipitation anomalies for February is shown in Fig. 9, and as for the August maps (Fig. 6) it matches well with the results of Cook and Vizy (2010). A region of significant positive correlations, showing a stronger CLLJ leads to increased precipitation, is evident across the south-central United States and the Midwest. Little correlation is observed in the Caribbean itself, but a region of significant negative correlation is seen in the western Atlantic. This negative correlation region is likely due to increased subsidence in this region of the NASH when the CLLJ is strong.
As in Fig. 6, but for February and NCAR PCM is replaced by NCAR CCSM.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
As in Fig. 6, but for February and NCAR PCM is replaced by NCAR CCSM.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
As in Fig. 6, but for February and NCAR PCM is replaced by NCAR CCSM.
Citation: Journal of Climate 24, 22; 10.1175/JCLI-D-11-00134.1
The models produce varying results, as seen in Fig. 9. In both chosen models, the CLLJ and the GPLLJ indices have a significant positive regression coefficient (both CMIP and AMIP), but differing precipitation correlations are evident. The region of negative correlations in the western Atlantic is farther west in all the model simulations, consistent with the westward displacement of the NASH (Fig. 2c) and the overly strong CLLJ in February.
Both CMIP models (Figs. 9b,c) show positive correlations across the central United States, although it is shifted southeast in both simulations. The AMIP models (Figs. 9d,e) show little to no positive correlation with precipitation over the central United States, despite being significantly correlated with the GPLLJ. This suggests that although the CLLJ itself may be continuing northward into the GPLLJ in the AMIP models, it is not transporting sufficient moisture to influence the central U.S. precipitation. This lack of moisture may be due to the AMIP models producing too much Caribbean rainfall, which leaves little moisture available for northward transport. It may also be due to inaccurate air–sea moisture fluxes over the Gulf of Mexico, which do not input more moisture into the jet as it transitions from the Caribbean to the United States.
6. Conclusions
Using 19 coupled and 12 uncoupled model runs from the IPCC AR4, the ability of the models to produce a CLLJ has been investigated. Previous studies of the CLLJ had either been purely observational (Wang 2007; Muñoz et al. 2008; Whyte et al. 2008; Cook and Vizy 2010) or with few GCM studies (Amador 1998; Wang et al. 2008). The CLLJ is an important feature for IPCC AR4 models to reproduce, as it has a large impact on both local and U.S. climate, including easterly waves and tropical storms (Serra et al. 2010). Although the IPCC AR4 output varied in horizontal resolution from 1.125° to 5° and contained a multitude of different parameterization configurations, all were able to develop and maintain a CLLJ with similar features to the observed CLLJ, with no clear impact of resolution on the results.
The seasonal cycle of the CLLJ was more challenging for the models to simulate. The observed semiannual cycle was not seen in either CMIP or AMIP models, with uniform magnitudes between January and July followed by a minimum in September and October. The uniformity of the CLLJ throughout the first half of the year was a result of the NASH being too uniform in its strength and location. The lack of a westward and southward extension of the NASH in July meant that meridional SLP gradients across the Caribbean were not enhanced and hence the CLLJ remained uniform in magnitude. The correct simulation of the structure, strength, and evolution of the NASH is essential for the correct development of an accurate CLLJ in the models. The poor simulation of Atlantic and South American precipitation by the IPCC models (Biasutti et al. 2006; Richter and Xie 2008) may be impacting the NASH and thus the CLLJ.
The magnitude of the CLLJ was also a problem with the majority of models. At 925 hPa and throughout the lower atmosphere, the models regularly overestimated the strength of the CLLJ with CMIP models having greater magnitudes than AMIP models. This was most evident in the first half of the year. The overestimate of CLLJ strength along with anomalously cold SSTs in the CMIP models agrees with the results presented by Wang et al. (2008) based on the theory of Gill (1980). Despite the overestimate of magnitude by the models, the vertical deepening of the CLLJ during July was well captured by all the models, despite not having a CLLJ peak at this time. This indicates that separate processes must be controlling the deepening and strengthening of the CLLJ, with the deepening better represented in the models.
During the summer months, and particularly August, the CLLJ is highly negatively correlated with precipitation anomalies in the Caribbean. As the CLLJ strength increases, moisture is transported away from the Caribbean leading to drier conditions. Despite some problems with the simulation of precipitation by the models (and the subject of another study), the models showed considerable promise in the simulation of the CLLJ–rainfall relationship in the Caribbean, particularly during August. The AMIP models performed better than the CMIP models in the strength and structure of the precipitation correlations, suggesting that the atmospheric component of the model is performing well and the influence of the cold SSTs and/or incorrect moisture fluxes in the CMIP models may be leading to poorer performance in the CLLJ–rainfall coupling.
A link between the ability of models to produce a summer CLLJ peak and the MSD was established, although the cause and effect could not be explicitly determined. The models that did produce a distinct MSD produced better area-averaged precipitation throughout the entire annual cycle, had an improved CLLJ annual cycle (although magnitudes were still large in January–June), and had improved magnitudes and annual cycle of SLP and SLP gradients in the Caribbean. These improvements were seen in both the CMIP and AMIP models. An important result from the MSD compositing was the improvement in the SST and especially the SST gradient in the CMIP models that produced a MSD. Although SSTs were still lower than observations in the MSD composite, this highlights the importance of correctly simulating both mean SSTs and SST gradients for reproducing accurate Caribbean climate.
Connections between the CLLJ and U.S. climate in the models were found to be quite variable, with CMIP models outperforming AMIP models. Not only did CMIP models reproduce the positive correlation between the CLLJ and GPLLJ during the cold season, they also showed a better positive correlation between the CLLJ and central U.S. cold season precipitation. Even AMIP models that captured the CLLJ–GPLLJ correlation were unable to accurately reproduce the CLLJ–U.S. rainfall correlations. This leads to the hypothesis that the northward branch of the CLLJ is not transporting enough moisture to the GPLLJ. This may be due to too much moisture being used for precipitation in the Caribbean region (AMIP models overestimate Caribbean precipitation), not enough moisture being fluxed into the lower atmosphere over the Gulf of Mexico, or a combination of both. Further investigation into the moisture fluxes and transport in the models is necessary. However, it is also important to consider possible errors in the reanalysis, especially in data-sparse regions such as the Caribbean and surrounding areas. Amador (1998) show that NCEP–NCAR reanalysis underestimates moisture flux in the northern Gulf of Mexico between 1973 and 2004 in the entry region to the GPLLJ.
The ability of the IPCC AR4 models to simulate a realistic and accurate CLLJ has been examined in detail and related to local and U.S. climate. While the models show pleasing results across the range of horizontal resolutions and model configurations of the ensemble, further model improvements and understanding of the observations are needed to fully reproduce the observed CLLJ structure and impacts in GCMs. Additional investigations of the relationship between the CLLJ and large-scale climate features such as the North Atlantic Oscillation (NAO) and ENSO in the GCMs would be beneficial in further examining the model performance.
Acknowledgments
This research has been supported by NASA Grant NNX10AG89G. The authors thank Chunzai Wang and two anonymous reviewers for comments that improved the quality of the manuscript. The authors acknowledge the modeling groups, the PCMDI and the World Climate Research Programme’s Working Group on Coupled Modelling (WGCM), for their roles in making the WCRP CMIP3 multimodel dataset available. Support of this dataset is provided by the Office of Science, U.S. Department of Energy. The monthly GPCP combined precipitation data were developed and computed by the NASA Goddard Space Flight Center’s Laboratory for Atmospheres as a contribution to the GEWEX Global Precipitation Climatology Project. NCEP_Reanalysis 2 data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, from their Web site (http://www.esrl.noaa.gov/psd/).
REFERENCES
Amador, J., 1998: A climatic feature of the tropical Americas: The trade wind easterly jet. Top. Meteor. Oceanogr., 5, 91–102.
Amador, J., 2008: The Intra-Americas sea low level jet: Overview and future research. Ann. N. Y. Acad. Sci., 1146, 153–188.
Amador, J., V. Magaña, and J. Pérez, 1999: Dynamics of the low-level jet over the Caribbean Sea. Preprints, 23rd Conf. on Hurricanes and Tropical Meteorology, Dallas, TX, Amer. Meteor. Soc., 868–869.
Amador, J., V. Magaña, and J. Pérez, 2000: The low level jet and convective activity in the Caribbean. Preprints, 24th Conf. on Hurricanes and Tropical Meteorology, Fort Lauderdale, FL, Amer. Meteor. Soc., 114–115.
Amador, J., E. Alfaro, O. Lizano, and V. Magaña, 2006: Atmospheric forcing of the eastern tropical Pacific: A review. Prog. Oceanogr., 69, 101–142.
Amador, J., E. Alfaro, E. Rivera, and B. Calderón, 2010: Climatic features and their relationship with tropical cyclones over the Intra-Americas seas. Hurricanes and Climate Change, J. Elsner et al., Eds., Springer, 149–174.
Berbery, E., and E. Collini, 2000: Springtime precipitation and water vapor flux over southeastern South America. Mon. Wea. Rev., 128, 1328–1346.
Biasutti, M., A. Sobel, and Y. Kushir, 2006: AGCM precipitation biases in the tropical Atlantic. J. Climate, 19, 935–958.
Bonner, W., 1968: Climatology of the low level jet. Mon. Wea. Rev., 96, 833–850.
Christensen, J., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 909–917.
Cook, K., and E. Vizy, 2010: Hydrodynamics of the Caribbean low-level jet and its relationship to precipitation. J. Climate, 23, 1477–1494.
Cook, K., E. Vizy, Z. S. Launer, and C. Patricola, 2008: Springtime intensification of the Great Plains low-level jet and Midwest precipitation in GCM simulations of the twenty-first century. J. Climate, 21, 6321–6340.
Durán-Quesada, A., L. Gimeno, J. Amador, and R. Nieto, 2010: Moisture sources for Central America: Indentification of moisture sources using a Lagrangian analysis technique. J. Geophys. Res., 115, D05103, doi: 10.1029/2009JD012455.
Gamble, D., D. Parnell, and S. Curtis, 2008: Spatial variability of the Caribbean mid-summer drought and relation to North Atlantic high circulation. Int. J. Climatol., 28, 343–350.
Gill, A., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106, 447–462.
Huffman, G., R. Adler, M. Morrissey, D. Bolvin, S. Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observation. J. Hydrometeor., 2, 36–50.
Jury, M., B. Malmgren, and A. Winter, 2007: Subregional precipitation climate of the Caribbean and relationships with ENSO and NAO. J. Geophys. Res., 112, D16107, doi:10.1029/2006JD007541.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP-DOE AMIP-II reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1643.
Magaña, V., J. Amador, and S. Medina, 1999: The midsummer drought over Mexico and Central America. J. Climate, 12, 1577–1588.
Martin, E., and C. Schumacher, 2011a: Modulation of Caribbean precipitation by the Madden–Julian oscillation. J. Climate, 24, 813–824.
Martin, E., and C. Schumacher, 2011b: The relationship between tropical warm pool precipitation, sea surface temperature, and large-scale vertical motion in IPCC AR4 models. J. Atmos. Sci., in press.
Mestas-Nuñez, A., D. Enfield, and C. Zhang, 2007: Water vapor fluxes over the Intras-Americas Sea: Seasonal and interannual variability and associations with rainfall. J. Climate, 20, 1910–1922.
Misra, V., S. Chan, R. Wu, and E. Chassignet, 2009: Air-sea interaction over the Atlantic warm pool in the NCEP CFS. Geophys. Res. Lett., 36, L15702, doi:10.1029/2009GL038737.
Mo, K., M. Chelliah, M. Carrera, R. Higgins, and W. Ebisuzaki, 2005: Atmospheric moisture transport over the United States and Mexico as evaluated in the NCEP regional reanalysis. J. Hydrometeor., 6, 710–728.
Muñoz, E., A. Busalacchi, S. Nigam, and A. Ruiz-Barradas, 2008: Winter and summer structure of the Caribbean low-level jet. J. Climate, 21, 1260–1276.
Neelin, J., M. Munnich, H. Su, J. Meyerson, and C. Holloway, 2006: Tropical drying trends in global warming models and observations. Proc. Natl. Acad. Sci. USA, 103, 6110–6115.
Poveda, G., and O. Mesa, 2000: On the existence of Lloró (the rainiest locality on Earth): Enhanced ocean-land-atmosphere interaction by a low-level jet. Geophys. Res. Lett., 27, 1675–1678.
Rauscher, S., F. Kucharski, and D. Enfield, 2011: The role of regional SST warming variations in the drying of Meso-America in future climate projections. J. Climate, 24, 2003–2016.
Rayner, N., D. Parker, E. Horton, C. Folland, L. Alexander, D. Powell, E. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.
Richter, I., and S.-P. Xie, 2008: On the origin of equatorial Atlantic biases in coupled general circulation models. Climate Dyn., 31, 587–598.
Serra, Y., G. Kiladis, and K. Hodges, 2010: Tracking and mean structure of easterly waves over the Intra-Americas Sea. J. Climate, 23, 4823–4840.
Ting, M., and H. Wang, 2006: The role of the North American topography on the maintenance of the Great Plains summer low-level jet. J. Atmos. Sci., 63, 1056–1068.
Virji, H., 1981: A preliminary study of summertime tropospheric circulation patterns over South America from cloud winds. Mon. Wea. Rev., 109, 599–610.
Wang, C., 2007: Variability of the Caribbean low-level jet and its relations to climate. Climate Dyn., 29, 411–422.
Wang, C., and S.-K. Lee, 2007: Atlantic warm pool, Caribbean low-level jet, and their potential impact on Atlantic hurricanes. Geophys. Res. Lett., 34, L02703, doi:10.1029/2006GL028579.
Wang, C., S.-K. Lee, and D. Enfield, 2008: Climate response to anomalously large and small Atlantic warm pools during the summer. J. Climate, 21, 2437–2450.
Whyte, F., M. Taylor, T. Stephenson, and J. Campbell, 2008: Features of the Caribbean low level jet. Int. J. Climatol., 28, 119–128.
