15 OCTOBER 2002 INOUE ET AL. 2897

Bimodal Distribution of Tropical Cyclogenesis in the : Characteristics and Environmental Factors

MASAMICHI INOUE Coastal Studies Institute, and Department of Oceanography and Coastal Sciences, Louisiana State University, Baton Rouge, Louisiana

ITSUKI C. HANDOH AND GRANT R. BIGG School of Environmental Sciences, University of East Anglia, Norwich, Norfolk, United Kingdom

(Manuscript received 15 August 2001, in ®nal form 22 April 2002)

ABSTRACT Tropical cyclogenesis critically depends on the presence of warm water at the sea surface. For the North Atlantic basin as a whole, the tropical storm season starts in May, peaks in September, and then declines, generally following the seasonal warming and cooling of sea surface temperature. In the Caribbean, in contrast, there is a distinct bimodal distribution in the number of tropical storms formed, with peaks in June and October separated by a signi®cant minimum in July. The timing of the observed minimum in tropical cyclogenesis appears to be related to the strengthening of the easterly trade over the Caribbean associated with the onset of the so-called veranillo, or midsummer drought (MSD), previously recognized over south-central Mexico, , and parts of the Caribbean. It appears that the observed minimum in cyclogenesis is caused by a combination of environmental factors related to the strengthening of the easterly trade winds across the Caribbean Basin. The strengthening easterly trade winds and their associated changes in stress curl give rise to enhanced in the southwestern Caribbean. This appears to trigger an enhanced local atmosphere±ocean coupling, giving rise to very unfavorable conditions in several environmental variables including cooler sea surface temperature (SST), higher sea level pressure (SLP), increase in outgoing longwave radiation (OLR), and decrease in precipitable water content (PRW). Moreover, strengthening trade winds result in increases in tro- pospheric vertical wind shear (VSH). Except for OLR, these environmental variables become least favorable for southwestern Caribbean cyclogenesis in July. In contrast, the transition from weak to intense convective activity in the eastern Paci®c results in weaker trade winds in the Caribbean in October. The resulting westerly wind anomalies lead to weakening upwelling, warmer SST, enhanced , and moist air coupled with weaker VSH in the southwestern Caribbean. All variables, except OLR, then become most favorable for cy- clogenesis. In the rest of the Caribbean, some of the conditions, primarily SST related, are not fully met. Nevertheless, the southwestern Caribbean appears to dominate the rest of the Caribbean in terms of setting the stage for either favorable or unfavorable conditions for cyclogenesis in the whole Caribbean Basin. Therefore, ocean±atmosphere interaction over the southwestern Caribbean appears to play an integral role in both suppressing and enhancing tropical cyclogenesis in the Caribbean on an annual basis.

1. Introduction referred to collectively as tropical storms). In the North Atlantic as a whole, the tropical storm season starts in Cyclogenesis of tropical storms critically depends on May, peaks in September, and then declines. Peaking of the presence of warm water at the sea surface (Malkus and Riehl 1960; Carlson 1971; Wendland 1977; Shapiro cyclogenesis in September also coincides with maxi- 1982; Gray 1984; Shapiro and Goldenberg 1998). Con- mum tropical wave activity emanating from West Africa sequently, the tropical storm season generally follows (Thorncroft and Hodges 2001). The importance of west- the annual movement of the sun. The annual cycle of ward-propagating disturbances emanating from West the number of tropical storms (NTS) covering the period Africa (often referred to as tropical, African, or easterly January 1886±December 1999 for the whole Atlantic is waves), in the tropical cyclogenesis over the North At- shown in Fig. 1. It is noted that tropical storms in this lantic and eastern Paci®c Oceans, has been recognized study include tropical storms and hurricanes (hereafter since the 1940s (Dunn 1940). Satellite images have been very useful in tracking tropical waves across the At- lantic. During their westward propagation across the Corresponding author address: Dr. Masamichi Inoue, Coastal North , tropical waves encounter variable Studies Institute, Dept. of Oceanography and Coastal Sciences, Lou- isiana State University, Baton Rouge, LA 70803. environmental conditions. Under favorable conditions, E-mail: [email protected] those tropical waves can turn into tropical storms in-

᭧ 2002 American Meteorological Society

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 2898 JOURNAL OF CLIMATE VOLUME 15

velopment regions has been attributed to the seasonal changes of position and intensity of the intertropical (ITCZ) and vertical wind shears (Gray 1968). This annual shifting of major development regions could give rise to the bimodal distribution in cyclogenesis in the Caribbean and in the Gulf of Mex- ico. However, this does not explain the uniqueness of the distinct bimodality for Caribbean cyclogenesis in comparison to the neighboring Gulf of Mexico. Previous studies have identi®ed several environmen- tal variables in addition to warm sea surface tempera- tures that are necessary for tropical cyclogenesis (e.g., Gray 1968; Landsea et al. 1998). Those variables in- clude low sea level pressure (SLP; Gray 1968; Shapiro 1982; Gray et al. 1993; Knaff 1997); low outgoing long- wave radiation (OLR), which is an indicator of deep atmospheric convection (Zhang 1993) necessary for FIG. 1. Climatology of the number of tropical storms formed in the whole North Atlantic (solid line) and in the Caribbean Basin tropical cyclogenesis (Gray 1968); low tropospheric (dotted line). Based on the period 1886±1999. vertical wind shear (VSH; Gray 1968; Goldenberg and Shapiro 1996; Landsea et al. 1998); and high precipi- table water content (PRW; Miller 1958; Malkus and cluding the strongest Atlantic hurricanes (e.g., Landsea Riehl 1960; Emanuel 1986). It should be noted that most 1993). In fact, it appears that tropical waves initiate most of these variables are not independent but closely in- Atlantic tropical (e.g., Landsea et al. 1998). terrelated. These variables have been identi®ed as cru- In various subbasins within the North Atlantic, how- cial in providing favorable conditions for hurricane for- ever, deviations from the general pattern of cyclogen- mation in the Caribbean Basin (Landsea et al. 1998), esis for the whole North Atlantic have been noted pre- which lies within the critical 10Њ±20ЊN latitude belt viously (e.g., Cry and Haggard 1962). One prominent known as the main development region of tropical example is the Caribbean Basin, where the observed storms in the North Atlantic (Goldenberg and Shapiro occurrences of tropical storms show a distinct bimodal 1996). The objective of this study is to examine the distribution with peaks in June and October separated climatological annual cycle in the relevant environmen- by a distinct minimum in July (Fig. 1). The temporal tal variables that might explain the observed bimodal dip in July is statistically signi®cant (paired t test of distribution in tropical cyclogenesis in the Caribbean. June and July, p Ͻ 0.05, n ϭ 114 yr), while the number reaches its seasonal maximum in October. Within the 2. Data and analysis whole North Atlantic, the Caribbean Basin is the only subbasin with such a distinct bimodal distribution of We examined climatological monthly mean data of NTS (Ballenzeig 1959; Cry and Haggard 1962); NTS, SST, SLP, OLR, VSH, and PRW. Additionally, though, a similar but much less distinct bimodal dis- surface wind vectors (SWV) are examined in order to tribution of NTS is found in the Gulf of Mexico (Cry explore a possible atmosphere±ocean link through wind and Haggard 1962; Fig. 2a). Curiously, its uniqueness stress acting on the sea surface. NTS are derived from has not attracted much attention in the published lit- all the tropical storms formed in the North Atlantic for erature. the period 1886±1999, reported by the National Hur- It is well known that there is a seasonal shift of de- ricane Center in Miami, . Both tropical storms velopment regions for tropical storms in the North At- and hurricanes are included primarily to increase the lantic (Cry and Haggard 1962; Gray 1968). During the number of samples; however, the general outcome of early part of the hurricane season (June), major devel- this study is still applicable even if only hurricanes were opment regions are located in the western Caribbean considered (not shown here). The climatological month- and in the Gulf of Mexico (Cry and Haggard 1962; Cry ly mean data used here include the 1982±99 SST from 1965). In July the favored genesis region shifts to the the National Centers for Environmental Prediction east, namely the Lesser Antilles and the southwestern (NCEP) optimal interpolation (Reynolds and Smith North Atlantic (Cry and Haggard 1962). In August and 1994); the long-term SLP, VSH, and PRW taken from the ®rst half of September, activity is noted in all regions the NCEP±NCAR (National Center for Atmospheric with principal contributions coming from the Lesser An- Research) reanalysis of 1958±97 (Kalnay et al. 1996); tilles and the southeastern North Atlantic (Cry and Hag- and the 1978±95 series of the National Oceanic and gard 1962). In October the major development region Atmospheric Administration (NOAA) interpolated OLR shifts back to the western Caribbean and the Gulf of (Chelliah and Arkin 1992). VSH is represented by the Mexico (Gray 1968). This shifting pattern of major de- magnitude of the wind vector difference between 200-

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 15 OCTOBER 2002 INOUE ET AL. 2899

FIG. 2. Climatological monthly time series over the hurricane season: (a) total number of the tropical storms formed over the period 1886±1999 (NTS) of the Caribbean and Gulf of Mexico, (b) SST for the period 1982±99, (c) SLP from the NCEP±NCAR reanalysis of 1958±97, (d) OLR for the period 1978±95, (e) VSH from the NCEP±NCAR reanalysis of 1958±97, and (f) PRW from the NCEP±NCAR reanalysis of 1958±97. All of these except NTS are averaged over the southwestern Caribbean (10Њ±15ЊN, 80Њ±75ЊW). Units are shown on the vertical axes. and 850-mb surfaces on which intense westerly and To examine the spatial variability of these environ- easterly ¯ows, respectively, are found (Landsea et al. mental variables and how they might be related to the 1998). The horizontal resolution of the SST data is 1Њ observed decrease in NTS in July in the Caribbean, ϫ 1Њ, while other variables have a resolution of 2.5Њϫ differences (month-to-month anomalies) between July 2.5Њ. Data uniformity was achieved by interpolating all and June (July±June) are shown in Figs. 3a,c, and 3e, the 2.5Њϫ2.5Њ data on to the 1Њϫ1Њ grid. while the counterparts for October can be interpreted Climatological monthly mean values of those vari- from the October±September differences shown in Figs. ables estimated for the southwestern Caribbean (10Њ± 3b,d, and 3f (see the ®gure caption for details). Loca- 15ЊN, 80Њ±75ЊW) are presented in Figs. 2b±f (the reason tions of all the historical tropical storms formed in the for selecting this particular region will be discussed be- North Atlantic (for the period 1986±99) bounded by 0Њ± low). There appear to be good correlations among those 35ЊN and 100Њ±50ЊW are presented in Figs. 3e and 3f variables in terms of providing favorable/unfavorable for July and October, respectively. In July, SST gen- environmental conditions for tropical cyclogenesis; that erally increases in the North Atlantic due to seasonal is, more active cyclogenesis correlates with warmer SST, warming (Fig. 3a). The exception to this is in the south- lower SLP, lower OLR, weaker VSH, and more PRW. western Caribbean, where it decreases. The widespread Less active cyclogenesis in July appears to be accom- cooling trend evident in the eastern Paci®c in July, off panied by cooler SST, higher SLP, higher OLR, stronger the coast of Mexico and Central America, is due to VSH, and less PRW. Figures 2b±f effectively explain diminished solar radiation and stronger easterly winds the minimum in cyclogenesis in July and the maximum (MaganÄa et al. 1999). In the eastern Paci®c, during April in October in the southwestern Caribbean. All the pa- and May, the SSTs increase, reaching a maximum rameters except OLR display a temporal maximum or (Ͼ29ЊC) by mid- or late June, leading to the onset of minimum in July or October. In July this region expe- the rainy season. However, as convective activity in the riences the coldest SST, maximum SLP,maximum VSH, ITCZ in the eastern Paci®c increases and the trade winds and minimum PRW for the entire tropical sea- intensify along with the formation of an anticyclonic son. Conversely, in October this region experiences the circulation over the western coast of Mexico, the SSTs warmest SST, minimum SLP, minimum VSH, and max- begin to lower, signaling the onset of the midsummer imum PRW. drought (MSD; MaganÄa et al. 1999).

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 2900 JOURNAL OF CLIMATE VOLUME 15

FIG. 3. Month-to-month changes in environmental variables: (a) Jul±Jun and (b) Oct±Sep differences of SST (color shaded in units of ЊC) and SWV (reference vector is shown at bottom right in units of m s Ϫ1); (c) Jul±Jun and (d) Oct± Sep differences of SLP (color shaded in units of mb) and VSH (contoured with intervals of 2 m s Ϫ1 with dashed lines being negative anomalies); (e) Jul±Jun and (f) Oct±Sep differences of OLR (color shaded in units of 5 W m Ϫ2). Locations of tropical storm formation in Jul and Oct are plotted by ®lled circles in (e) and (f), respectively. The most unlikely and likely regions for the tropical cyclogenesis (where conditions for SST, SLP, VSH, OLR, and PRW are met) in Jul and Oct are circled by thick solid lines in (e) and (f), respectively; (g) Jul±Jun and (h) Oct±Sep differences (color shaded in units of ЊC) of the Levitus and Boyer marine temperature climatology in the top 500 m of the water column along 77ЊW in the Caribbean. Jul and Oct temperature pro®les are also contoured in (g) and (h), respectively, with the contour interval of 1ЊC.

The area showing positive SLP anomalies (Fig. 3c) land-based convection over the Nicaraguan±Costa Ri- covers much of the domain extending from the eastern can rainforest appears to be at least partly contributing Paci®c to the western Atlantic. This is due to the sea- to the strengthening easterly winds in the southwest Ca- sonal intensi®cation of the Bermuda high in July. ribbean. Strengthening of easterly trade winds associated with On the North Atlantic side, negative SST anomalies the onset of the MSD is evident along the steepest gra- of up to 0.5ЊC in July are limited to the southwestern dient regions in the SLP anomaly ®eld shown in Fig. Caribbean (Fig. 3a), and are collocated with positive 3c, and it appears to extend from the eastern Paci®c well VSH anomalies (stronger VSH; Fig. 3c). Stronger VSH into the southwestern Caribbean (Fig. 3a). The inten- is primarily due to the enhanced surface easterlies, and si®cation of the easterly winds in the southwestern Ca- it is associated with the onset of the MSD. The zone of ribbean must be part of the large-scale intensi®cation stronger VSH extends from the eastern Paci®c (15Њ± of the easterly winds associated with the onset of the 20ЊNat100ЊW) into the southwestern Caribbean (10Њ± MSD. It is interesting to note that there is some indi- 15ЊNat75ЊW). Development of stronger vertical shear cation of an enhanced land-based convection (negative and the suppression of tropical storm development in OLR anomalies) over the Nicaraguan±Costa Rican - the western Caribbean at this time of the year has been forest (Fig. 3e), where a local surface wind convergence previously noted (Gray 1968). The implications for the appears to be located (Fig. 3a). Therefore, the enhanced resulting air±sea interaction over the negative SST

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 15 OCTOBER 2002 INOUE ET AL. 2901

FIG.3.(Continued) anomalies in the southwestern Caribbean are consistent can be compared to ®ve storms having formed in June with suppressed convection over this region indicated in the same region (not shown). by positive OLR anomalies (Fig. 3e) and positive SLP It is notable that the rest of the Caribbean (outside anomalies (Fig. 3c), and negative PRW anomalies (not of the southwestern Caribbean) also shows a signi®cant shown). It should be noted that negative SST anomalies drop in NTS from 19 in June (not shown) to 4 in July. being discussed here do not indicate actual cold SST In July, the rest of the Caribbean experiences small pos- but rather refer to relative month-to-month changes. De- itive SST anomalies (Fig. 3a) and negative VSH (Fig. spite relatively small SST anomalies, they could have 3c), both of which might suggest slightly more favorable signi®cant impact on the variation in deep atmospheric conditions for tropical cyclogenesis in July compared convection, because of their high background temper- to June. However, this region experiences positive SLP atures (above 27Њϳ28ЊC; Zhang 1993). The regions (Fig. 3c) and positive OLR anomalies (Fig. 3e) that circled by the thick solid lines in Fig. 3e are where appear to overcome the opposing effect of positive SST negative SST and PRW, and positive SLP, OLR, and and negative VSH. It is worth noting that the Gulf of VSH anomalies, occur simultaneously, suggesting that Mexico, with a similar but less distinct bimodal distri- in those regions tropical storm formation is less likely bution of NTS (Fig. 2a), does not exhibit similar changes in July compared to June. On the North Atlantic side, in environmental variables in July, with only positive the only encircled region is located in the southwestern SLP anomalies contributing toward the suppression of Caribbean, a region slightly smaller than the box (10Њ± cyclogenesis. 15ЊN, 80Њ±75ЊW) used for Fig. 2. In fact, only two trop- In contrast to July, a completely reversed situation ical storms have ever formed in July in this region over arises in October (Figs. 3b,d, and 3f) when tropical cy- the period 1886±1999 (Fig. 3e), and both of them were clogenesis in the Caribbean reaches its peak (Fig. 1). In formed along the periphery of the encircled region. This October compared to September, the southwestern Ca-

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 2902 JOURNAL OF CLIMATE VOLUME 15 ribbean experiences positive SST (Fig. 3b), negative to modulating the easterly winds and providing feedback SLP, negative VSH (Fig. 3d), and negative OLR (Fig. from the ocean to the atmosphere. 3f) anomalies, all contributing toward more favorable conditions for cyclogenesis. SWV anomalies in October 3. Annual modulation of upwelling in the (Fig. 3b) indicate the development of an atmospheric Caribbean convergence anomaly in the southern Caribbean where warmer SST and negative OLR anomalies are found. Comparing Figs. 3a and 3b, we notice a reversal of Development of westerly wind anomalies (actually the SWV anomaly pattern over the southwestern Carib- weakening easterly winds) in the Caribbean (Fig. 3b) bean between July±June and October±September. To ex- results in weaker VSH (negative VSH anomalies) in amine the oceanic response to this changing surface most of the Caribbean Basin (Fig. 3d). Despite the peak- wind forcing, a north±south vertical section of temper- ing of tropical cyclogenesis in October in the whole ature in the top 500 m of the water column along 77ЊW Caribbean Basin, the southwestern Caribbean still re- was constructed based on the climatological monthly mains as the least favorable region for cyclogenesis (Fig. atlas (Levitus and Boyer 1994). Figure 3g shows the 3f). This is due to the fact that the coldest SSTs in the July temperature section and July±June temperature dif- Caribbean are always found in the southern half of the ferences, while Fig. 3h shows the October temperature basin due to the prevailing upwelling maintained by the section and October±September temperature differenc- predominant easterly trade winds (Gordon 1967). es. Due to the prevailing easterly trade winds in the It appears that there could be distinct coupling be- Caribbean Basin, persistent coastal upwelling is evident tween the ocean and atmosphere in the southwestern along the coast of (Figs. 3g,h). Fur- Caribbean, which might play an important role in mod- thermore, the strongest easterly trade wind is located ulating the environmental conditions for cyclogenesis. near the midbasin, giving rise to negative wind stress Interestingly, on the intraseasonal timescale, it has been curl over the southern half of the basin, while the north- reported that the Madden±Julian oscillation (MJO) mod- ern half is exposed to positive wind stress curl (Gordon ulates hurricane activity in the Gulf of Mexico and the 1967). The resulting Ekman pumping forces open-water western Caribbean (Maloney and Hartmann 2000). upwelling (away from the coast) in the southern half of Tropical cyclogenesis and activity are enhanced over the basin and downwelling in the northern half of the the basins during the MJO westerly phase. It is worth basin (Gordon 1967). Consequently, SSTs in the south- noting that the structure of wind anomalies over the ern half are relatively low, while SSTs in the northern western Caribbean during the MJO easterly and westerly half are higher throughout the annual cycle. The July phases are similar to those plotted in Figs. 3a,b, re- and October temperature sections clearly indicate large- spectively. We will, therefore, examine in more detail scale upwelling in the southern half and downwelling the reversal of the trade wind anomalies and its impact in the northern half of the basin. The observed pattern on the air±sea interaction over the southwestern Carib- of upwelling and downwelling appears to be consistent bean. with the distribution of phytoplankton in the Caribbean The dry air period of July, indicated by a decrease Basin previously observed (MuÈller-Karger et al. 1989; in PRW over this time (Fig. 2e), is caused by the in- MuÈller-Karger and Castro 1994). It is notable that both vasion of stable and rather dry maritime trade wind air of these dynamical signals penetrate down to at least formed by the enhancement of the Atlantic subtropical 500 m, as evident in Figs. 3g,h. There is an indication high and the consequent southward displacement of the of an eastward-¯owing coastal countercurrent between ITCZ (Garbell 1947). Enhancement of the surface east- 10Њ and 11ЊN, while the westward-¯owing Caribbean erly trades and rather constant upper (at 200- Current (Gordon 1967) occupies between 11Њ and 17ЊN. mb isobaric surface) from June to July results in an The latter manifests itself in the northward-sloping iso- increase of VSH. Toward the west, part of the ITCZ therms, that is, warmer water to the north and colder still remains over the Nicaraguan±Costa Rican region, water to the south of the basin. This is superimposed represented by negative OLR anomalies (Fig. 3e). The on top of the existing wind-driven upwelling and down- observed dry period in July corresponds to the MSD welling. In July, subsurface (Ͻ500 m) to surface cooling previously noted as a dry period in the middle of the by as much as 0.5Њ±1.0ЊC over 11Њ±15ЊN in Fig. 3g is rainy season of Mexico and Central America (MaganÄa a clear indication of enhanced upwelling. This area of et al. 1999). surface cooling is collocated with the unfavorable region Annual suppression of tropical cyclogenesis in the of tropical storm formation (encircled by a thick solid Caribbean appears to be related to the intensi®cation of line in Fig. 3e). The stronger upwelling is driven by easterly trade winds associated with the onset of the enhanced easterlies in July (Fig. 3a). The meridional MSD. There is a strong indication that the local air±sea length scale of the cooling zone is much broader than interaction plays an integral role in modulating envi- the local ®rst baroclinic Rossby radius of deformation ronmental conditions necessary for annual modulation (less than 100 km along the South American coast; Chel- of tropical cyclogenesis in the Caribbean. In the next ton et al. 1998). Hence, not only coastal upwelling but section, we will examine the role of oceanic response also open ocean divergence (DIV) due to Ekman pump-

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 15 OCTOBER 2002 INOUE ET AL. 2903

FIG. 4. Climatological daily time series through a calendar year: (a) the observed LPD between Kingston (Jamaica) and Panama City (Panama). SLP of the NCEP±NCAR grids covering the stations are taken; (b) the divergence caused by wind stress curl (DIV) averaged over the south- western Caribbean (10Њ±15ЊN, 80Њ±75ЊW). Units are shown on the vertical axes. ing associated with negative wind stress curl anomalies tropical high pressure system) and the low pressure cen- should be responsible for this. In October, subsurface ter located in Central America. Thus, we examine the to surface warming by as much as Ͼ1ЊC appears, which large-scale pressure difference (LPD) between Panama is primarily limited to the southern half of the basin City (Panama) and Kingston (Jamaica) that character- (Fig. 3h), but the overall vertical temperature structure izes the pressure gradient across the easterly trade winds of the upwelling is still maintained. It is noted that the in the Caribbean Basin. Figures 4a and 4b show a cli- surface warming revealed by the Levitus data (Fig. 3h) matological mean daily time series of LPD and DIV, is not limited to the southern half of the basin in com- respectively, the latter being averaged over the region parison to the warming indicated by the NCEP data (Fig. used in Figs. 2b±f. LPD attains a temporal maximum 3b). The overall warming trend in the upper 400 m of from June to July. This appears to be the precursor of the water column in October in the southern half of the the July event. In fact, DIV rapidly increases through basin appears to be the manifestation of weaker up- June by 3 m dayϪ1. This causes cold water from a depth welling due to the weakened easterlies. The large me- of 90 m to be upwelled all the way to the surface over ridional length scale associated with the warming trend a period of a month in addition to the permanent cooling in October suggests that it is the consequence of open- by the large-scale background upwelling. This event water downwelling via Ekman pumping. It should be causes the cooling of the subsurface water, but does not noted that some of the features revealed by the vertical affect the near-surface water (top 30 ϳ 40 m) in June sections, such as the eastward-¯owing countercurrent (not shown) probably due to a rapid seasonal warming near the South American coast, might be due to oceanic by the incoming solar radiation. However, the strong eddies previously noted in this region by numerical upwelling signal continues until mid-July and then model studies (Murphy et al. 1999), and by satellite weakens. Hence, SST attains a temporal minimum dur- observations (Andrade and Barton 2000). ing July rather than June (Fig. 2b), which appears to be It is interesting to consider the climatological annual synchronized with suppressed convection through a lo- regulation that enhances the easterly trade winds, and, cal air±sea coupling over the southwestern Caribbean. in turn, upwelling. It is noted that the easterly trade In the eastern Paci®c, there is no signi®cant DIV signal winds in the Caribbean are maintained by a geostrophic (not shown), although there are a few limited regions balance between the Bermuda high (North Atlantic sub- where some of the environmental conditions are less

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 2904 JOURNAL OF CLIMATE VOLUME 15 favorable for tropical cyclogenesis in July compared to panied by higher VSH. Intensi®cation of upwelling June (Fig. 3e). According to MaganÄa et al. (1999), in leads to lower SST, and weaker atmospheric convection. the eastern Paci®c, suppression of cyclone activity dur- This might, in turn, lead to subsequent weakening of ing the MSD is due to cooling of SST caused by di- easterly trade winds in the southwestern Caribbean, sig- minished solar radiation and stronger easterly winds naling the cessation of stronger upwelling, eventually (presumably by enhanced evaporation), and to subse- leading to the reversal of environmental conditions. In quent decrease in deep atmospheric convective activity. October, a peak in tropical cyclogenesis coincides with Thus, annual upwelling-induced suppression of tropical the simultaneous occurrences of favorable environmen- cyclogenesis is unique to the southwestern Caribbean. tal conditions, triggered by the weakening easterly trade This mechanism for modulating cyclogenesis cannot be winds. The question as to what happens in the rest of applied to explain the intraseasonal analogs by MJO the Caribbean deserves some consideration. It should (Maloney and Hartmann 2000), since the lifetime of be noted that SST is the major variable whose behavior enhanced upwelling is longer than MJO's easterly phase. is reversed in the rest of the basin compared to the Thus, while VSH seems to be a critical environmental southwestern region; that is, there is slight warming factor for modulating cyclogenesis on an intraseasonal rather than cooling in July, and cooling rather than MJO timescale and does not accompany oceanic re- warming in October (Figs. 3a and 3b). It is also inter- sponses, our analysis emphasizes the importance of local esting to note that the October ocean temperature section ocean±atmosphere feedback to the probability of cyclo- (Fig. 3h) indicates more extensive surface warming than genesis on a longer timescale. This is con®rmed by the what is suggested in Fig. 3b based on the NCEP data. near-permanent absence of cyclogenesis between His- Hence, the spatial extent of the unfavorable and favor- paniola and the South American coast (Figs. 3e,f). The able region might be larger than the southwestern Ca- anomaly maps of Fig. 3 do not suggest the existence of ribbean as indicated by the NCEP SST datasets. Nev- this feature, but there is a continuous presence of east- ertheless, the southwestern Caribbean appears to dom- erlies there, hence, upwelling and cooler SST and higher inate the rest of the Caribbean in terms of setting the OLR. stage for either favorable or unfavorable conditions for It is also important to note that the recent numerical cyclogenesis in the whole Caribbean Basin. It is inter- model studies (Murphy et al. 1999) as well as obser- esting to note that one of the current Atlantic basin vational studies using satellite data have identi®ed nu- seasonal hurricane forecast models (Gray et al. 2000) merous oceanic eddies in the Caribbean Basin (Andrade includes June±July Caribbean Basin sea level pressure and Barton 2000). It is conceivable that wind-forced and zonal wind anomalies as predictive parameters. The annual modulation of upwelling in the southern basin dry air of the maritime trades is most likely to be main- (downwelling in the northern basin) could generate cold tained by the air±sea interaction triggered by enhanced (warm) eddies. Noting that Hurricane Opal in October upwelling, providing a link between the onset of the 1995, intensi®ed over a warm eddy propagating in the MSD and tropical cyclogenesis in the Caribbean. Thus, Gulf of Mexico (Bosart et al. 2000), variability of cold a next step in investigating the reason for the Caribbean and warm eddies in the Caribbean Basin might hold a tropical cyclogenesis bimodality will be the formulation key to understanding the detailed processes involved in of a simple coupled ocean±atmosphere model to ex- individual tropical storm formation there. amine the detailed processes involved. Previously, clear examples of ocean±atmosphere interaction impacting the environmental conditions often come from large- 4. Discussion and conclusions scale interannual variability, such as El NinÄo±Southern In this paper, it is shown that annual suppression (en- Oscillation (e.g., Philander 1990) and the Equatorial At- hancement) of tropical cyclogenesis in the Caribbean in lantic Oscillation (Handoh and Bigg 2000). It is inter- July (October) appears to be related to the concurrent esting, therefore, to note that local atmosphere±ocean occurrences of various environmental variables provid- interaction in the Caribbean appears to be an important ing favorable (unfavorable) conditions. The most inter- element in modulating environmental conditions for esting observation is that the variability of large-scale tropical cyclogenesis on an annual basis. upwelling in the southwestern Caribbean due to the an- Finally, it is noteworthy to mention that similar bi- nual intensi®cation and weakening of the easterly trade modal distributions of NTS are found in other parts of winds, appears to play an integral role in modulating the world, such as in the Bay of Bengal, southeast Indian the environmental conditions for tropical cyclogenesis Ocean, and the South Paci®c (Atkinson 1971). To ex- in the Caribbean. It appears that the annual intensi®- amine their links to ocean±atmosphere coupling, we per- cation of the easterly trade winds in July, which is as- formed a singular value decomposition analysis (e.g., sociated with the onset of the MSD, results in the in- Wallace et al. 1992) of long-term mean daily SST and tensi®cation of upwelling in the southwestern Carib- OLR over the global tropical band (30ЊS±30ЊN). When bean, triggering the sequence of events leading to the the ®rst two harmonics of the annual cycle are excluded, formation of unfavorable conditions for tropical cyclo- the gravest coupled mode, with its periodicity of 6 genesis. Strengthening easterly trade winds are accom- months, clearly shows decrease in SST and increase in

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC 15 OCTOBER 2002 INOUE ET AL. 2905

OLR during July±August over the western Caribbean the association of El NinÄo and West African rainfall with Atlantic Sea and northern equatorial Atlantic Ocean, and during major hurricane activity. J. Climate, 9, 1169±1187. Gordon, A. L., 1967: Circulation of the . J. Geophys. February±March over the western Paci®c and southeast Res., 72, 6207±6233. Indian Oceans. The mode is another harmonic of the Gray, W. M., 1968: Global view of the origin of tropical disturbances annual cycle, as there is an out-of-phase relationship in and storms. Mon. Wea. Rev., 96, 669±700. spatial structure between the Northern and Southern ÐÐ, 1984: Atlantic seasonal hurricane frequency. Part II: Fore- casting its variability. Mon. Wea. Rev., 112, 1669±1683. Hemisphere tropical basins. However, the coupled mode ÐÐ, C. W. Landsea, P. W. Mielke Jr., and K. J. Berry, 1993: Pre- could cause a temporal suppression of cyclogenesis, dicting Atlantic basin seasonal activity by 1 while the annual movement of the sun provides a fun- August. Wea. Forecasting, 8, 73±86. damental monomodal NTS. This points to an ocean± ÐÐ, ÐÐ, ÐÐ, and ÐÐ, cited 2000: Synopsis of 2000 Atlantic atmosphere coupling mechanism suppressing tropical basin seasonal hurricane forecast. [Available online at http:// typoon.atmos.colostate.edu/forecast/.] cyclogenesis, and leading to bimodal distributions of Handoh, I. C., and G. R. Bigg, 2000: A self-sustaining climate mode NTS in those tropical regions, as well as the south- in the tropical Atlantic, 1995±97: Observations and modelling. western Caribbean. Quart. J. Roy. Meteor. Soc., 126, 807±821. Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re- analysis Project. Bull. Amer. Meteor. Soc., 77, 437±471. Acknowledgments. This work was partially funded by Knaff, J. A., 1997: Implications of summertime sea level pressure NATO CRG and the Coastal Studies Institute, Louisiana anomalies in the tropical Atlantic region. J. Climate, 10, 789± State University. We thank Susana Towery for compi- 804. lation of the tropical storm data. ICH was ®nancially Landsea, C. W., 1993: A climatology of intense (or major) Atlantic hurricanes. Mon. Wea. Rev., 121, 1703±1713. supported by the School of Environmental Sciences of ÐÐ, G. D. Bell, W. M. Gray, and S. B. Goldenberg, 1998: The the University of East Anglia and the COAPEC thematic extremely active 1995 Atlantic hurricane season: Environmental program of the Natural Environment Research Council. conditions and veri®cation of seasonal forecasts. Mon. Wea. Rev., 126, 1174±1193. Levitus, S., and T. P. Boyer, 1994: Temperature. Vol. 4, World Ocean REFERENCES Atlas 1994, NOAA Atlas NESDIS 4, 117 pp. MaganÄa, V., J. A. Amador, and S. Medina, 1999: The midsummer drought over Mexico and Central America. J. Climate, 12, 1577± Andrade, C. A., and E. D. Barton, 2000: Eddy development and 1588. motion in the Caribbean Sea. J. Geophys. Res., 105, 26 191± Malkus, J. S., and H. Riehl, 1960: On the dynamics and energy 26 201. transformations in steady-state hurricanes. Tellus, 12, 1±20. Atkinson, G. D., 1971: Forecasters' guide to tropical . Maloney, E. D., and D. L. Hartmann, 2000: Modulation of hurricane U.S. Air Force Tech. Rep. 240, Air Weather Service (Mac), 360 activity in the Gulf of Mexico by the Madden±Julian oscillation. pp. Science, 287, 2002±2004. Ballenzeig, E. M., 1959: Relation of long-period circulation anom- Miller, B. I., 1958: On the maximum intensity of hurricanes. J. Me- alies to tropical storm formation and motion. J. Meteor., 16, teor., 15, 184±195. 121±139. MuÈller-Karger, F. E., and R. A. Castro, 1994: Mesoscale processes Bosart, L. F., C. S. Velden, W. E. Bracken, J. Molinari, and P. G. affecting phytoplankton abundance in the southern Caribbean Black, 2000: Environmental in¯uences on the rapid intensi®- Sea. Cont. Shelf. Res., 14, 199±221. cation of Hurricane Opal (1995) over the Gulf of Mexico. Mon. ÐÐ, C. R. McClain, T. R. Fisher, W. E. Esaias, and R. Varela, 1989: Wea. Rev., 128, 322±352. Pigment distribution in the Caribbean Sea: Observations from Carlson, T. N., 1971: An apparent relationship between the sea-sur- space. Progress in Oceanography, Vol. 23, Pergamon, 23±64. face temperature of the tropical Atlantic and the development Murphy, S. J., H. E. Hurlburt, and J. J. O'Brien, 1999: The connec- of African disturbances into tropical storms. Mon. Wea. Rev., tivity of eddy variability in the Caribbean Sea, the Gulf of Mex- 99, 309±310. ico, and the Atlantic Ocean. J. Geophys. Res., 104, 1431±1453. Chelliah, M., and P. A. Arkin, 1992: Large-scale interannual vari- Philander, S. G. H., 1990: El NinÄo,LaNinÄa, and the Southern Os- ability of outgoing longwave radiation anomalies over the global cillation. Academic Press, 293 pp. Tropics. J. Climate, 5, 371±389. Reynolds, R. J., and T. M. Smith, 1994: Improved global sea surface Chelton, D. B., R. A. deSzoeke, M. G. Schlax, K. El Naggar, and N. temperature analyses using optimal interpolation. J. Climate, 7, Siwertz, 1998: Geographical variability of the ®rst baroclinic 929±948. Rossby radius of deformation. J. Phys. Oceanogr., 28, 433±460. Shapiro, L. J., 1982: Hurricane climatic ¯uctuations. Part II: Relation Cry, G. W., 1965: Tropical cyclones of the North Atlantic Ocean: to large-scale circulation. Mon. Wea. Rev., 110, 1014±1023. Tracks and frequencies of hurricanes and tropical storms, 1871± ÐÐ, and S. B. Goldenberg, 1998: Atlantic sea surface temperature 1963. Tech. Paper 55, U.S. Weather Bureau, Washington, DC, and tropical cyclone formation. J. Climate, 11, 578±590. 148 pp. Thorncroft, C., and K. Hodges, 2001: African easterly wave vari- ÐÐ, and W. H. Haggard, 1962: North Atlantic tropical cyclone ability and its relationship to Atlantic tropical cyclone activity. activity, 1901±1960. Mon. Wea. Rev., 90, 341±349. J. Climate, 14, 1166±1179. Dunn, G. E., 1940: Cyclogenesis in the tropical Atlantic. Bull. Amer. Wallace, J. M., C. Smith, and C. S. Bretherton, 1992: Singular value Meteor. Soc., 21, 215±229. decomposition of sea surface temperature and 500-mb height Emanuel, K. A., 1986: An air±sea interaction theory for tropical anomalies. J. Climate, 5, 561±576. cyclones. Part I: Steady-state maintenance. J. Atmos. Sci., 43, Wendland, W., 1977: Tropical storm frequencies related to sea surface 585±604. temperature. J. Appl. Meteor., 16, 477±481. Garbell, M. A., 1947: Tropical and Equatorial Meteorology. Pitman, Zhang, C., 1993: Large-scale variability of atmospheric deep con- 237 pp. vection in relation to sea surface temperature in the Tropics. J. Goldenberg, S. B., and L. J. Shapiro, 1996: Physical mechanisms for Climate, 6, 1898±1913.

Unauthenticated | Downloaded 09/30/21 04:00 AM UTC