1. Introduction

The importance of westward-propagating disturbances, referred to as tropical waves (also called African or easterly waves), has been recognized for more than half a century (Dunn 1940). These disturbances, which are usually traced back to Africa, can lead to the formation of tropical cyclones over the Atlantic and eastern Pacific basins. Typically, the strongest Atlantic hurricanes of the year originate from tropical waves (e.g., Landsea 1993).

Initially, the waves were detected as series of centers of falling and rising sea level pressure moving from east to west across the islands of the Caribbean Sea. When upper-air observations became available, it was noted that those pressure centers were accompanied by a cyclonic wind shift (typically from northeast to southeast) that was largest near the 700-mb level. Currently, satellite images are an extremely valuable supplement to surface and upper-air data for tracking tropical waves.

Tropical waves have been the focus of research for a long time. Riehl (1945, 1954) first described the dynamics of easterly waves, and a more detailed analysis and compilation began in 1967 by Simpson et al. (1968). Carlson (1969a,b) carried out synoptic analyses describing the characteristics of tropical waves over west Africa and the tropical Atlantic. Burpee (1972) studied the origin and structure of tropical waves in the lower troposphere of North Africa. In addition, several papers about the structure and properties of tropical waves were published after the Global Atmospheric Research Project (GARP) Atlantic Tropical Experiment (GATE) (e.g., Thompson et al. 1979; Burpee and Reed 1982). Furthermore, there have been some attempts to analyze and forecast tropical waves using global models (e.g., Reed et al. 1988).

This article is primarily concerned with operational analysis of tropical waves. The main purpose here is to tabulate and summarize certain weaker synoptic-scale systems of 1994 and 1995, namely, tropical waves and tropical depressions. Nature has provided the unique opportunity to compare some prevailing meteorological conditions between a nonactive 1994 season and a very active or near-record 1995 season.

2. Data and analysis

Tropical waves were identified and tabulated using satellite imagery, surface observations, and upper-air data across the Tropics from Africa to Central America. In general, vertical time sections were constructed with the available 0000 and 1200 UTC upper-air data from 1 May to 30 November. The time sections were prepared for several stations and, in conjunction with daily satellite imagery, are used to locate the waves. The presence of a lower-tropospheric cyclonic wind shift, a notable 24-h surface pressure change, and/or the propagation of an organized and distinct cloud mass are the primary parameters used in locating the wave axis. Figure 1 is a vertical time section of the wind for Trinidad for 1–15 September 1994 and 1995.

Since some tropical waves have very poor cloud definition, it is difficult to track such waves across the data-sparse oceanic areas, and there is a certain amount of noise in their identification and positioning. In general, when deep convection was minimal or absent and the cloud pattern was not well defined, the position of such waves was simply extrapolated.

Over the past few years, improvements in the analysis and initialization in global models have made it possible to depict more accurately the westward propagation of lower- to middle-tropospheric vorticity centers along the tropical belt. Longitude versus time (Hovmöller) diagrams of the meridional component of the wind were sometimes used to depict the (U.S. Environmental Modeling Center) Medium Range Forecast Model’s representation of tropical waves. This was also used as complementary information to track the waves.

3. Census and comparison of the 1994 and 1995 Atlantic systems

Tropical cyclone activity in the Atlantic was below normal in 1994. The season featured a total of seven named tropical cyclones, three of which became hurricanes. None of the hurricanes hit the United States but the combination of Alberto, Beryl, and Gordon at tropical-storm strength caused $1 billion in damage. Torrential rains associated with the remnants of Alberto caused 30 fatalities in Alabama and Georgia. Rainfall and mud slides associated with Gordon during its tropical storm stage killed 1122 people in Haiti.

In contrast to 1994, tropical cyclone activity was near a record in 1995. There were 19 named storms, 11 of which became hurricanes. Three tropical storms and two hurricanes made landfall in the continental United States. Tropical cyclones in the Atlantic basin were responsible for an estimated 123 deaths, and U.S. damages were over one billion dollars each in Hurricane Marilyn and Hurricane Opal. In addition, it is estimated that non-U.S. damages of $2.5 billion occurred in the Caribbean from Hurricane Luis and $1.5 billion occurred in Mexico from the combination of Roxanne and Opal. Detailed summaries of the 1994 and 1995 hurricane seasons are included in Avila and Rappaport (1996) and Lawrence et al. (1998), respectively.

In addition, there were five tropical depressions that did not intensify into tropical storms in 1994 and two in 1995. Only those depressions are described here. In 1994, three tropical depressions formed from disturbances of nontropical origin (e.g., cold fronts, upper lows) and nine developed from tropical waves. In 1995, 19 of the 21 total number of tropical depressions developed from tropical waves.

In terms of tropical depression frequency, 1994 was below average with a total of only 12 tropical depressions. On the other hand, in 1995 the total number of tropical depressions jumped to 21. The average for the period of 1967–94 is 19. In both years, there were tropical depressions in every month of the season. In 1994, September was the most active month with five tropical depressions, but only two intensified to tropical storms and none reached hurricane status. In 1995, August was the most active month with eight tropical depressions, and only one failed to become a tropical storm. Four of the tropical storms reached hurricane status. Figures 2a and 2b show the Atlantic tropical depression tracks for 1994 and 1995, respectively.

4. Tropical waves and their environment during 1994 and 1995

Figures 10a and 10b show that during 1994 and 1995 70 and 63 tropical waves crossed the northwest coast of Africa and moved westward over the tropical Atlantic, the Caribbean, Central America, and appeared to move into the eastern Pacific. The average period of the waves in 1994 was 3.0 days, which is very similar to the period that was observed during the previous three years. The first tracked wave exited Africa on 2 May and the last one crossed Dakar during late November. In 1995, the first wave was counted on early May and the last one on the middle of November. The average period of the 1995 waves was 3.4 days.

Tropical waves in 1994, as in the previous three years, appeared to be poorly defined in the wind and cloud fields relative to other years wherein large amplitude waves were observed to move off the coast of Africa (e.g., Simpson et al. 1968). Vertical time sections from several rawinsonde stations consistently showed that the wind shifts associated with the waves were not significantly marked. In general the convective clusters propagating westward on satellite images were not really distinct. For example, one can notice in Fig. 11a, a sequence of daily satellite images from late August through September of 1994, that the tropical wave associated with Debby was the only easily discernible system on satellite images during that climatologically most active period. In Fig. 1a, one can also detect that with the exception of the tropical wave associated with Debby on 10 September, only small cyclonic wind shifts were observed at Trinidad associated with the other waves during that period. The low amplitude of the waves observed since 1991 may partially explain the lack of tropical cyclone activity in the deep Tropics.

In contrast with 1994, during 1995, tropical waves were characterized by their large amplitude and impressive cloud representation on satellite images. In fact, by the time some tropical waves exited the coast of Africa, there was already a low- to middle-level circulation and/or a surface low associated with the waves. If time sections for Trinidad during 1994 and 1995 are compared, one can see that during 1995 the wind shifts associated with tropical waves were distinct. The maximum wind change occurred at the low and middle levels and one could imply from the data that these waves already had an embedded vortex. In addition, one can observe that the upper-level winds (250 mb and above) were predominantly from the east (Landsea et al. 1997), whereas in 1994 these winds were in general from the west and northwest. This suggests that the 1995 waves were moving through an environment of prevailing low vertical wind shear. Indeed, the cloud pattern for the tropical Atlantic during September of 1995 (Fig. 11b) shows that many of the waves had already spawned tropical cyclones.

Since tropical waves appeared to be so active and spawned so many more tropical cyclones over the Atlantic in 1995 as compared to 1994, one could ask the question: “Were these waves stronger over west Africa in 1995 in comparison to the previous year?” A clue to this may be found by considering the dynamics of the waves. It has been established that tropical waves have cold cores in the lower troposphere and warm cores aloft (e.g., Reed et al. 1977). As long as the waves produce deep convection not far from the trough axis, as they typically do over west Africa, the loss of energy from the lifting of cooler air at low levels is compensated by heating and lifting aloft. However, once these waves enter the tropical Atlantic, if deep convection is suppressed, there will be a sink of energy and the systems will weaken. Based on satellite images, fields of outgoing longwave radiation during 1994 and 1995 (Climate Analysis Center 1994, 1995), and rainfall data (Landsea 1996, personal communication), the waves were not more active in terms of the amount of deep convection or associated rainfall over west Africa in 1995 as compared to 1994. However, over the eastern tropical Atlantic, the amount of deep convection appeared to be greater in late summer of 1995 as compared to the previous year, which implies a less stable atmosphere during 1995. This decreased stability is also consistent with positive sea surface temperature anomalies over the eastern Atlantic during the summer of 1995 (Landsea et al. 1998) and negative anomalies over the comparable region and period of 1994 (Avila and Rappaport 1996).

Some characteristics of the lower-tropospheric environment of the waves during 1994 and 1995 can be inferred from Figs. 12a,b, which depict the average low-level (approximately 900 mb) wind anomalies (from the 1974–94 mean) for August, September, and October (ASO) of 1994 and 1995, respectively. In 1994 (Fig. 12a), as occurred in 1993, an anomalous monsoon-type flow extended northward across the tropical Atlantic to near 17°N, where a large anomalous cyclonic flow prevailed. More significantly, there was an intrusion of anomalous northwesterly (and, presumably, dry) flow over the western Atlantic into the Tropics, and also an equatorward branch of anomalous flow feeding air from a stable region of the subtropics into the monsoonal region of west Africa. These also could have been inhibiting factors. The region of anomalous cyclonic flow centered along approximately 18°N from 50°W to Africa in Fig 12a is probably related to the primary path of the vorticity maximum associated with the waves during the period. Since the region of anomalous cyclonic flow was located over relatively cool sea surface temperatures (Avila and Rappaport 1996), the development of deep convection associated with the waves moving through that area was very limited (this reiterates what was discussed earlier from a static stability point of view). Only the few systems that moved westward through lower latitudes were able to produce deep convection (e.g., Debby). In general those systems that propagated westward over low latitudes (between 5° and 10°N), moved over South America and did not have the opportunity to develop.

In 1995, the pattern of wind anomalies for ASO (Fig. 12b) changed from 1994. In 1995, there was a significant intrusion of moist tropical air from the eastern Pacific into the Caribbean Sea and into the tropical Atlantic. Contrasting with 1994, the monsoonal flow prevailing in 1995 was over a region where the ocean was warmer than normal, primarily over the eastern Atlantic. In 1994, the eastern Atlantic was cooler than normal, as noted above. In addition, in ASO 1995, there were two large low-level cyclonic wind gyres, one centered just to the northeast of Puerto Rico and the other over the eastern Gulf of Mexico. These gyres were superimposed by a large area of anticyclonic flow and diffluence aloft. This persistent pattern of anomalous low-level confluence, upper-level diffluence, and warm ocean probably favored the waves to develop. Moreover, the southwesterly anomalies in the low-level flow pattern over the tropical Atlantic basin during ASO 1995 would also correspond to weaker northeast trade winds. This would contribute to a weaker vertical shear, since the easterlies typically decrease with height and become westerlies in the upper troposphere over that area. The weaker shear over the Atlantic basin during 1995 has been confirmed by Landsea et al. (1998); this decreased shear favored the development of more storms from waves in 1995.

It may be argued that the existence of the large low-level anomalous gyres in ASO of 1995 was due to the presence of a number of Atlantic hurricanes. Whereas this may be partially true, one can see, in Fig. 12b, that there is a flow of air, on a very broad scale, feeding from low latitudes into the gyres. This is suggestive of a large-scale circulation anomaly that favored cyclogenesis from tropical waves in the Atlantic basin during the peak months of the 1995 Atlantic season.

Nine of the 12 tropical cyclones in 1994 and 19 of 21 in 1995 developed from tropical waves over the Atlantic basin. Those tropical cyclones are represented in Figs. 10a,b by lines attached to the main “stream” of waves. Note that only Chris attained hurricane strength south of 21°N in 1994. In 1995, several of the systems reached hurricane strength in the deep Tropics. Moreover, four of them reached category 3 or higher intensity on the Saffir–Simpson hurricane scale (Simpson 1974) along that belt.

The development of eastern Pacific tropical cyclones from tropical waves has been suggested previously by Frank and Clark (1980). During the summer months it is typical for weak vertical shear and monsoonal-type flow to prevail over the warm Pacific waters to the south of Mexico. Therefore, tropical cyclogenesis in this region is frequently favored. In 1994 as well as 1995, it is likely that tropical waves again played a significant role in the formation of many of the eastern Pacific tropical storms (Pasch and Mayfield 1996). These waves usually take more than a week to traverse the Atlantic and Central America after emerging from west Africa. In Figs. 11a,b, one can observe the westward propagation of the convection associated with some of the tropical waves crossing Central America from the Caribbean into the eastern Pacific, leading to the development of some of the tropical cyclones in the latter basin. For example, in 1995 (Fig. 11b) there was a westward progression of an area of cloudiness across Central America on 11–12 September. This same system, which can be seen crossing 100°W on 14 September, later developed into east Pacific Hurricane Juliette.

Figure 10 shows two longitude bands centered along 105°W and 120°W, where tropical waves contributed to the formation of several of the eastern Pacific tropical cyclones in 1994. The formation along 120°W during 1994 represented a westward shift from the climatological longitude of development. In some cases, particularly for those eastern Pacific tropical cyclones that formed closer to Central America, it is possible that the orographic effects of the adjacent mountainous landmass can aid in their formation, as theorized by Zehnder (1991). In 1995 the tropical cyclone activity in the eastern Pacific was rather limited, and development was clustered in a small region centered along 110°W. Rappaport et al. (1998) gave a possible explanation for the below average activity in this basin during 1995. Even though there was considerably less tropical cyclone activity in the eastern Pacific during 1995 compared to 1994, an inspection of Fig. 10 reveals that the percentage of cyclones that developed from tropical waves was about 75% in both years.

As noted above, there are some cases (in 1994, 1995, or other years) where eastern Pacific tropical cyclogenesis could not be traced back to a tropical wave, and the source of the preexisting disturbance was not clear. Typically these were situations in which the development occurred from a disturbance that was first detected within the ITCZ over the eastern north Pacific.

5. Comparison with other years

Figure 13a and Table 1 show the number of tropical waves from 1967 through 1995. In 1994, the number of waves (70) was somewhat higher than the average (61) but very similar to the previous four years. In 1995, the number of waves (63) was close to the average. It is emphasized that year to year variations in the total number of waves are probably not significant. Since the process of identifying tropical waves has not been uniformly applied over the years, one must be careful in interpreting the total number of such events especially when we are dealing with weak systems. However, this was not the case in 1995 since the waves were well represented in the wind and cloud fields and obviously the counting was more accurate.

Table 1 indicates that peaks in the number of tropical waves do not coincide with peaks in the number of tropical storms. This reaffirms Frank’s (1975) statement that the number of waves is not related to the total number of tropical storms. The large-scale atmospheric and oceanic environment exercises the largest control on the yearly variation in the number of tropical storms forming from tropical waves.

Table 1 and Figs. 13b and 13c show the seasonal totals of tropical depressions and storms since 1967. A trend toward fewer tropical depressions per year began around 1980, as indicated by Avila and Pasch (1992). In 1994, that trend continued with a total of only 12 depressions. This number is well below the more recent 1980–94 average of 15. In 1995, there were 21 depressions. This number is even higher than the long-term average and definitely much higher than the average observed since 1980 when the trend of having fewer depressions began. During the 1967–80 period, some nontropical cyclones may have been classified as tropical depressions. Examples of such cases have been documented by Pasch and Avila (1994). However, the number of tropical depressions in 1995 was realistic since these depressions were well documented with surface data, satellites, and reconnaissance aircraft.

Figure 14a displays the ratio of the number of tropical depressions of African origin to the total number of tropical depressions per year (ratio 1). Low values of ratio 1 indicate that a high number of depressions originated from upper-tropospheric cold lows or along frontal zones—that is, from nontropical “seedling” disturbances (Frank 1975). In 1994, ratio 1 was 0.8, a number that is similar to 1993 but substantially higher than that observed in 1991 or 1992, when many depressions originated from nontropical sources. In 1995, ratio 1 was 0.9.

The ratio between the number of tropical storms of African origin to the total number of tropical storms (ratio 2) has been found to be a useful parameter to describe the overall character of the hurricane season (Avila and Clark 1989). Figure 14b and Table 1 display the values of ratio 2 since 1967. The 1967–94 average contribution from tropical waves to the total number of storms is 0.61. In 1994, ratio 2 was 0.71, a value just above average, while in 1995, ratio 2 jumped to 0.89, a value much above normal.

Typically, tropical waves are the main source of hurricanes for the Atlantic basin. There have been exceptions, namely the 1971, 1972, 1977, 1983, 1984, 1991, and 1992 seasons, when tropical waves induced only a few hurricanes. In 1994, for the first time in three years, a tropical wave spawned a hurricane (Chris) at low latitudes. However, it was a hurricane only for a few hours. Table 1 shows that on average, 65% of the hurricanes in the Atlantic basin develop from tropical waves. In 1994, two of the three hurricanes developed from waves. However, although Gordon originated from a tropical wave, it underwent a quasi-subtropical transformation before reaching hurricane status at high latitudes. In 1995, all of the hurricanes were spawned by tropical waves.

It is of interest to note that 1991, 1992, 1993, and 1994 have been designated as years of generally moderate El Niño conditions, based on equatorial Pacific sea surface temperatures and atmospheric indices (Climate Analysis Center 1994). During those years, tropical waves, although prevalent, induced only a few storms and hurricanes. Gray (1984) speculated that a hostile environment, induced by El Niño episodes, prevented many of the waves from developing over the deep Tropics in the Atlantic basin.

Avila and Clark (1989) arbitrarily used ratio 2 to quantify the relative contribution to tropical storm development by tropical waves. Years in which ratio 2 is greater than or equal to 0.7 are termed African years. Non-African years are those for which ratio 2 is less than or equal to 0.5. Intermediate values correspond to average years. Ratio 2 for 1994 was 0.71, indicating that it was a borderline African year. However, hurricanes were not intense. In fact, of the African years, only 1987 produced fewer intense hurricanes. Ratio 2 was 0.89 in 1995 and there were several intense hurricanes. This is a more typical characteristic of an African year. Obviously, there are exceptions to the rule that African years feature a relatively large number of intense hurricanes. It appears that these exceptions often coincide with El Niño conditions, as alluded to above.

While ratio 2 provides information about tropical cyclone origin, the hurricane destruction potential (HDP) measures a hurricane’s potential for wind and storm surge destruction. The HDP is defined as the sum of the square of each hurricane’s maximum wind speed for each 6-h period of its existence (Gray et al. 1992). Table 2 summarizes the African, non-African, and average years, including the HDP and the total number of major hurricanes (category 3 or higher on the Saffir–Simpson hurricane scale). The average HDP of African years for the 1967–94 period was substantially larger than (more than double) the average HDP of the non-African years. The HDP for 1994 was only 15, which is a very low number for an African year, but for 1995, the HDP increased dramatically to 158.

This study again supports the concept that tropical waves play a dominant role as the precursors to major hurricanes, and that many of eastern Pacific tropical cyclones can likely be traced back to these waves. The significance of tropical waves is further highlighted by the fact that these systems are largely responsible for the enhancement of rainfall in the Caribbean area (Riehl 1954). Thus the two well-established climatological seasons (namely, wet and dry) in the tropical Atlantic basin can be attributed to the presence of (typically) stronger, rain-producing tropical waves during the summer and fall months. Therefore, it is important to continue to monitor and to study these systems.