THE STRONG ASSOCIATION BETWEEN WESTERN SAHEL MONSOON RAINFALL AND INTENSE ATLANTIC HURRICANES

By Christopher W. Landsea and William M. Gray

Department of Atmospheric Science Colorado State University Fort Collins, CO 80523

(Manuscript received 24 June 1991, in final form 4 November 1991)
Reprinted from Journal Of Climate, Vol. 5, No. 5, May 1992


AMS Copyright Notice

© Copyright 1992 American Meteorological Society (AMS). Permission to use figures, tables, and brief excerpts from this work in scientific and educational works is hereby granted provided that the source is acknowledged. Any use of material in this work that is determined to be "fair use" under Section 107 or that satisfies the conditions specified in Section 108 of the U.S. Copyright Law (17 USC, as revised by P.L. 94-553) does not require the Society's permission. Republication, systematic reproduction, posting in electronic form on servers, or other uses of this material, except as exempted by the above statements, requires written permission or license from the AMS. Additional details are provided in the AMS Copyright Policies, available from the AMS at 617-227-2425 or amspubs@ametsoc.org. Permission to place a copy of this work on this server has been provided by the AMS. The AMS does not guarantee that the copy provided here is an accurate copy of the published work.


Abstract Tables
Text References

ABSTRACT

Seasonal variability of Atlantic basin tropical cyclones is examined with respect to the monsoonal rainfall over West Africa. Variations of intense hurricanes are of most interest as they are responsible for over three-quarters of United States tropical cyclone spawned destruction though they account for only one-fifth of all landfalling cyclones. Intense hurricanes have also shown a strong downward trend during the last few decades. It is these storms which show the largest concurrent association with Africa's Western Sahel June to September rainfall for the years 1949 to 1990.

Though the Sahel is currently experiencing a multidecadal drought, the relationship between Atlantic tropical cyclones and Western Sahel rainfall is not dependent upon the similar downward trends in both data sets. A detrended analysis confirms that a strong association still exists, though reduced somewhat in variance explained. Additionally, independent data from the years 1899 to 1948 substantiate the existence of the tropical cyclone-Western Sahelian rainfall association.

The fact that the Sahel periodically experiences multidecadal wet and dry regimes suggests that the current Sahel drought which began in the late 1960's could be a temporary condition that may end in the near future. When this occurs, the Atlantic hurricane basin -- especially the Caribbean islands and the United States East Coast -- will likely see a large increase in intense hurricane activity associated with abundant Sahel rainfall similar to the period of the late 1940's through the 1960's.

1. Introduction

Mechanisms of rainfall variability in the African Sahel received little attention until the severe drought years of 1972 and 1973. The magnitude of the drought in those years was such that it caused widespread crop failure and the deaths of thousands of people. Since that time, a variety of studies, both in observational and numerical modeling methods, [see reviews by Hastenrath (1988, 1990b)], have been employed for analysis and predictions of Sahel rainfall.

In contrast to this, seasonal variability of Atlantic basin tropical cyclones has intrigued researchers for almost a century. Garriott (1906) attempted to relate the exceptionally high hurricane activity in September 1906 to fluctuations in Atlantic basin surface pressure. Subsequently, many researchers have attempted to relate the frequencies and intensities of tropical cyclones to sea surface temperature anomalies (Ray, 1935), general circulation patterns (Namias, 1955; Ballenzweig, 1958), as well as surface pressure variations (Shapiro, 1982a,b).

In the early 1980s, Gray (1984a, 1984b) showed strong predictive relationships of seasonal tropical cyclone activity to the presence or absence of an El Nino ño event, the phase and trend of the stratospheric Quasi-Biennial Oscillation (QBO), and sea level pressure anomalies in the Caribbean basin. Gray has since modified the forecast scheme (Gray, 1990a, 1990b) to include 200 mb Caribbean zonal wind anomalies and discounted temporal variability of the QBO phase. Yearly forecasts and verifications exhibiting some skill in seasonal predictions (Gray, 1990d) have been issued since 1984.

However, the fact that there is a strong interrelationship between seasonal intense hurricane activity in the Atlantic basin and the amount of rainfall which fell in the Sahel had not been fully realized until very recently. This is somewhat understandable because named storm activity (which receives more attention in the literature) has not seen the types of decadal changes that the more infrequent intense hurricanes have. Yet it is quite plausible that such a relationship might be present since it is known that the same weather systems which contribute especially large amounts of rainfall to the Sahel also frequently serve as the genesis points for tropical cyclones (Riehl, 1945). These, of course, are the African-spawned easterly waves.

The first mention of a possible interannual connection between Sahel rainfall and tropical cyclones variability appears in class notes from Carlson and Lee (1978). They speculated that ``the severe drought conditions" over the African Sahel might be related to ``the diminution of hurricane activity'' in the Atlantic basin. They did not pursue this further, however, nor did they appear to recognize the primary association of intense hurricane activity and Sahel rainfall conditions.

Gray (1987) revived interest in the possible Sahel rainfall and hurricane association when he noted the similarity in his seasonal measure of hurricane intensity and duration -- Hurricane Destruction Potential -- and the strong downward trend in the Sahel rainfall, which continued through much of the 1980's. However, it took a failed seasonal forecast in 1989 (where Gray attributed his underestimation to conditions associated with a return of rainfall to the Sahel) to stress the importance of the intense hurricane--Sahel rainfall relationship (Gray, 1989, 1990c) and to stimulate research on this topic.

2. Definitions

Much confusion can arise when discussing the various categories of tropical cyclones. The United States (U.S.) Department of Commerce National Oceanic and Atmospheric Administration (NOAA) (1977) technically defines tropical cyclones as non-frontal low pressure synoptic--scale systems that develop over tropical or subtropical waters and have a definite organized circulation. Two additional requirements that should be added: the cyclone must have a relatively warmer tropospheric inner core structure than the environment (i.e., a ``warm--core") and the maximum intensity of tangential winds should be located in the lower troposphere.

The tropical cyclone designation is a broad term under which various strength systems in the Atlantic basin are divided into:

Often, the term ``tropical storm'' refers to any cyclone of tropical storm or hurricane strength. This type of reference will be avoided. In this paper, tropical storms and hurricanes will be collectively referred to as named storms [in deference to the fact that since 1950 all tropical cyclones that were of at least tropical storm force were given a name for identification (Neumann et al., 1987), though some cyclones were determined to be of tropical storm strength after the fact and thus lack a formal name].

The ``hurricane'' status can be further categorized by the severity of the cyclone. The Saffir--Simpson (Simpson, 1974) Hurricane Scale provides a measure of intensity from a Category 1 (Minimal) to Category 5 (Catastrophic) hurricane. Table 1 summarizes the delineation of strength and effect. For purposes of this study it became instructive to consider the Saffir--Simpson Category 3, 4, and 5 cyclones collectively as intense or major hurricanes .

One method for objectively determining the seasonal amount of tropical cyclone activity is through the summed duration of each storm. This partially removes the subjectivity involved in categorizing the intensity of tropical cyclones. Dunn and Staff (1965) first utilized the term hurricane days . A seasonal total of hurricane days is the amount of days in which a hurricane existed (two existing simultaneously count as two days). However, their calculations and a similar one done by Gray (1984a) count a full hurricane day even if the cyclone was of hurricane strength for just six hours of that 24 hour period. The computations for this report count days in one quarter increments (six hour periods). This will slightly reduce the numbers obtained by these earlier studies. Similar calculations have been performed for intense hurricane days and named storm days.

Another way to measure tropical cyclone variability on a seasonal basis is through Gray's (1987) Hurricane Destruction Potential or HDP . This measure gives a combined reflection of both the intensity and duration of the tropical cyclones. It is defined as the sum of the sustained wind speed (in knots) squared for every six hours that the cyclone is of hurricane strength (65 kt or 33 m/s). Hurricane Destruction Potential is defined to approximate the idea that hurricanes can cause damage relative to the square of their wind speed (i.e., a proxy of the summed kinetic energy of the cyclone's maximum winds). Both the intense hurricane days and HDP are used extensively in this report as measures of seasonal activity.

3. Data

a. Atlantic Basin Tropical Cyclones

The positions and intensities (sustained wind speed and minimum surface pressure) of all Atlantic basin tropical cyclones of at least tropical storm strength have been archived and are continually being updated by the National Hurricane Center (NHC) in Miami, Florida. (The `Atlantic basin' is defined as the tropical and subtropical regions north of the equator in the Atlantic Ocean, the Caribbean Sea, and the Gulf of Mexico.) This data set extends from 1886 to 1990 and is described in detail by Jarvinen et al. (1984). This ``Best Track'' data set (as it is known since it is composed of the ``best'' estimate of positions and intensities in a post-analysis of all data available) or HURDAT (short for HURricane DATa) has been used quite extensively in our Tropical Meteorology Project at Colorado State University.

We have followed the recommendations by Neumann et al. (1987) to use tropical cyclone statistics based upon data since the mid-1940's, when organized aircraft reconnaissance began, since this ``probably best represents Atlantic tropical cyclone frequencies''. The same logic follows for the day to day assessment of the intensity of individual storms; again because in the earlier period ``storms that were detected could have been mis-classified as to intensity''.

Another consideration in using tropical cyclone frequency and intensity data is the subjectivity inherent in the categorization. Satellite based tropical cyclone intensity estimations are pattern recognition and empirical methods that do not directly measure the storm's winds (Dvorak, 1977, 1984). Aircraft reconnaissance usually records maximum wind speeds at flight levels (usually 850 or 700 mb). These are instantaneous values and do not necessarily represent a very accurate estimate of maximum surface winds. Nevertheless, satellite and aircraft data usually give rather close independent intensity estimates (Sheets, 1990). [Landsea (1991b) has determined that the strength of the intense Atlantic hurricanes were slightly overestimated 2--3 m/s during the mid 1940's to the late 1960's. However, this bias is small and does not alter the main conclusions of this study.]

Tropical cyclones that have affected the United States mainland have a longer period of reliability concerning their frequency and intensity. This is because of the large coastal populations that the U.S. has along the Gulf of Mexico and Atlantic Ocean. It is unlikely that any tropical storms or hurricanes were unnoticed crossing the coastline since the late 1800's. Hebert and Taylor (1975) have made a Saffir--Simpson categorization of all U.S. landfalling hurricanes from 1900 to 1974. They based their estimates primarily upon minimum central pressures of the storm at time of landfall. Neumann et al. (1987) have also analyzed 1899 and 1975 through 1986. This paper has extended the analysis period through 1990 using NHC's end of the season write-ups (Case and Gerrish, 1988; Lawrence and Gross, 1989; Case and Mayfield, 1990; and Mayfield and Lawrence, 1991).

Besides the various categories of landfalling hurricanes, this report also makes use of tropical storms that affected the U.S. This information was also gleaned from the Neumann et al. (1987) reference book and the recent annual reports.

Recently, Avila and Clark (1989) have revived the annual summary of Atlantic basin tropical disturbances in the Monthly Weather Review. These articles are companion papers to the annual summary of Atlantic Basin tropical cyclones (see the original paper by Simpson et al. (1968)). Since 1967, when daily satellite analysis made it possible, NHC has attempted to report on the origins of tropical cyclogenesis and the numbers and varieties of tropical disturbances. L.A. Avila of NHC provided a detailed list of the origins of all Atlantic basin tropical cyclones of at least tropical storm strength from 1967 to 1990.

b. African Rainfall Data

The Sahel, lying between approximately 11 and 20° in Africa, is the region which separates the hyperarid Sahara Desert to the north and the rain forests along the Gulf of Guinea and the Congo River basin to the south. It is only during three to five months in the summer and early fall when the Sahel receives substantial precipitation. This rainfall is due to the annual cycle of the Intertropical Convergence Zone (ITCZ) which reaches its northernmost position at that time. Over West Africa, the ITCZ takes the form of low level southwesterlies (monsoonal flow) which converge with low level northeasterlies. An excellent qualitative overview of West African meteorology is given by Hayward and Oguntoyinbo (1987).

For this study, ``African rainfall'' refers to any form of precipitation. However, since temperatures in the Sahel range from 15 to 45C the only non-rain precipitation possible is hail, that occurring only very infrequently.

The main statistical method used for analyzing rainfall variations is the area-average normalization developed by Kraus (1977). To best look at the regional aspect of rainfall variations, Kraus attempted to combine stations without inducing a bias toward any station or any subgrouping of stations. Using mean percentages for a group of stations would favor the drier stations which experience huge percentage variations of rainfall. Using mean absolute deviations could slant the regional value toward stations with higher average rainfall. To avoid either of these problems, Kraus used a normalization of rainfall based upon the mean and standard deviation at each station. This method is currently utilized by several observational researchers in Sahel rainfall variability (Hastenrath, 1990a; Lamb et al ., 1990; Nicholson, 1986; and Shinoda, 1989) as well as by the U.S. Climate Analysis Center (CAC) (U.S. Department of Commerce, 1990) in their real-time analysis of U.S. and worldwide precipitation patterns.

Using the following definition for the mean, r i , and for the variance, Sigma (i)2 at station i,

EQUATION 1

EQUATION 2

where rij is the period (e.g. weekly, monthly, multi-month) rainfall at station i during year j and Ji is the number of years of data station i contains.The normalization for station i in year j is

EQUATION 3

which essentially tells how many standard deviations from the mean that rainfall is for that particular year. The resulting regionally averaged Index value of xij for the year j is defined as

EQUATION 4

where I j is the number of stations in the region available in year j.

We have spent considerable effort in the gathering and analyzing of the African rainfall data from a number of sources. Appendix A lists our African rainfall data sources and the many individuals who have assisted us.

The beginning point in the rainfall analysis was chosen to correspond to the first year Gray (1984a, 1984b) used in his studies---1949. His analysis was limited to that date because of the lack of tropical stratospheric data before then; this data being essential due to the strong relationship between the stratosphere QBO and Atlantic basin tropical cyclones (Gray, 1984a; Shapiro, 1989). This starting date also agrees well with Neumann et al.'s (1987) assessment (mentioned earlier) that quantitative climatological studies on Atlantic basin tropical cyclones are most valid since the mid 1940's.

4. Intense Hurricane Variability

Landsea (1991b) has documented the intra- and interseasonal climatology of intense Atlantic hurricanes (Saffir--Simpson Category 3, 4, and 5). Three main features were identified which differentiate these cyclones from weaker hurricanes and tropical storms. First, intense hurricanes have experienced the largest linear downward trend and show the most year to year variation. The reduction in major hurricanes during the last twenty years has also been observed in U.S. landfalling intense hurricanes, especially along the U.S. East Coast and the Florida peninsula (Sheets, 1990). Second, these storms account for three-quarters of all U.S. tropical cyclone spawned damage even though they strike on average only twice in three years. Third, since over 98% of the intense hurricane activity occurs after July, this later starting date of their annual cycle allows a prediction of occurrence to be made on 1 August -- two months after the beginning of the ``official'' hurricane season.

Figure 1 portrays the time series of intense hurricane activity as measured both by seasonal number of intense hurricanes and by seasonal number of days in which an intense hurricane was present. Note the extreme reduction in recent years as accentuated by the linear trend lines. The 42-year means for the intense hurricanes and the intense hurricane days are 2.5 and 5.7, respectively. While all tropical cyclone activity exhibit positive correlations with Sahel rainfall, it is the intense hurricanes that show the strongest association. This holds true even after these strong downward trends are removed from the tropical cyclone data.

5. African Rainfall Variability

As described in the Introduction, Gray (1987) hypothesized that the general downtrend in Atlantic basin tropical cyclone activity had a connection to Sahel rainfall conditions. While the numbers of named storms and hurricanes showed little multidecadal (1949--1969 and 1970--1986) variation, the numbers and durations of stronger hurricanes (in intense hurricanes, hurricane days, and HDP) varied by a factor of 2 to 1 between these periods. Gray noted that the analysis of Sahel rainfall by Lamb (1985) also showed a large downward trend in the data. More quantitatively, the twenty station Lamb Index, for the years 1949 to 1990, is associated with intense hurricanes by a linear correlation coefficient of r = 0.61 and with intense hurricane days at r = 0.64. However, since Lamb's Index includes the western and central Sahel and covers the multi-month period of April to October, the signal that Sahel rainfall has with Atlantic basin tropical cyclones may not be maximized. The following section isolates which regions and time frames best relate the two phenomena.

a. Monthly Rainfall Variability

The most dominant feature in causing precipitation over the continent of Africa is the ITCZ. Its latitudinal extent as measured by the 100 mm isohyet covers over 15° (1650 km) meridionally during the Northern Hemisphere's summer. Direct extratropical influences are relatively minor, only producing precipitation in the most poleward > 30° latitude) extent of the continent. It is during the Northern Hemisphere's monsoonal season when substantial rainfall reaches north of 11°N that the Sahel gets almost all of its precipitation (June to September). Precipitation during these four months showed the strongest individual month associations with intense hurricane activity.

Table 2 presents an analysis of the individual months using linear correlation coefficients for individual station monthly rainfall versus seasonal intense hurricane days (Figure 2) . While each map shows areas of both positive and negative correlation, these were not accepted as showing meaningful significance unless they persisted for at least three months at a correlation of |r|> 0.15 (a subjective threshold of just 2 percent of variance explained) over a region greater than 3° square ~100,000 km2. Only one coupled region appears to meet the criteria: a strong positive correlation throughout the Sahel with an associated weaker negative correlation along the Gulf of Guinea. The Sahel consistently shows this positive relationship from June to September, while the negative dipole is seen for July through September. It is the western and central Sahel which shows the highest correlations versus intense hurricane days with several monthly correlation values of at least r = 0.50. August has the strongest association of any month with the seasonal hurricane activity. The presence of an inverse relationship of Sahel rainfall with the Gulf of Guinea region to its south is consistent with the findings of Nicholson (1986). She showed that the Sahel and the Gulf of Guinea often (but not always) show opposite anomalies during the same year: when the Sahel has abundant rainfall, the area along the Gulf of Guinea is usually dry; conversely, when the Gulf of Guinea receives above normal amounts of rainfall, the Sahel is often drier than normal.

Thus, the strongest monthly rainfall associations with intense hurricane activity occur in the western and central Sahel during the months of June to September with a maximum in August. The following section presents the combined association of the four month total rainfall.

b. June to September Rainfall Variability

Using combinations of the monthly correlations shown in the previous section, it was determined that June to September Sahel rainfall has the strongest concurrent relationship with intense hurricane days. It is this four month period that comprises most of the precipitation that occurs over the Sahel.

Confirming what was also observed in the Lamb Index, the Sahel shows increasingly positive correlations with tropical cyclones as the intensity category of the tropical cyclone increases. The three analyses in Figure 3show linear correlations for named storms, hurricanes, and intense hurricanes versus individual station June to September rainfall. The correlations were computed using data from 1949 to 1989, but as mentioned previously, many stations have very spotty temporal coverage. Because of this, stations with even 12 years of data are included in the analysis (though several in the Sahel have over 95 percent coverage).

Though all three analyses show a basic structure of positive correlations throughout the Sahel (strongest at the westernmost portion) and a weaker area of negative correlations along the Gulf of Guinea, it is the magnitudes of the associations that are of note. For the correlation with named storms, the highest values are only r = 0.35. For the strongest hurricanes, values of r = 0.65 and higher are seen in the Western Sahel. Figure 4 is an enlargement of North Africa with the correlation coefficients for intense hurricane days versus rainfall.

6. Western Sahel Rainfall Index

a. Development of Index

An accepted way of creating rainfall indices for a particular region is to combine stations through their mean standard deviations for that particular period (as discussed in section 2). Thirty-eight stations are utilized in such a method to create the June to September Western Sahel Rainfall Index. Figure 5shows the locations of the stations which comprise the index.

Countries included in the region are Senegal, Gambia, northern Guinea-Bissau, southern Mauritania, and western Mali. All available stations within the boundary are utilized if they provided at least 30 years of rainfall data in the period 1949 to 1990. Though this data threshold may appear somewhat low (only 74 percent of years needed), the quantity of data available over the Western Sahel is actually more reliable than over much of the rest of tropical Africa.

Rainfall in the Sahel has shown a strong tendency for year-to-year persistence during the last four decades. As seen in Figure 6, consistently wet anomalies were observed from 1950 to the mid-1960's and then almost continuous dry anomalies were experienced from 1968 onward, with above average rainfall only in the years of 1969, 1975, 1988, and 1989. The tendency for decreasing precipitation amounts with time is highlighted by the boldface trend line which gives the least squares best fit the data. Removal of the linear trend in the rainfall data is shown in a detrended analysis in Figure 7. The strong correlations between the Western Sahel rainfall and tropical cyclones are not dependent upon concurrent trends in the data sets, however.

Table 2 provides the mean standard deviations of rainfall (the same information as in Figure 6) as well as the number of stations that were included in each year's calculations. Data for the years of 1949 to 1950 and 1987 to 1990 are not as complete because the data sets from which several stations are drawn were not available for those time periods.

Information on the individual stations utilized in the Western Sahel Index are provided in Table 3 . This region covers a wide meridional area (11 to 20°N) and a large range of rainfall means. The individual stations show a high degree of correlation with the entire Western Sahel Index, with only 3 of 38 stations showing a correlation of less than r = 0.62. This type of internal consistency is vital where a single index is utilized in portraying precipitation for a widespread region.

Table 4 shows a month by month analysis of the combined 38 station means and standard deviations including the percent contribution to the main June to September rainfall period and to the entire rainy season of May to November. Though not shown, the remaining months, December to April, receive in total, only about 1% of the annual mean. Note that the main rainfall month is August, with both July and September receiving considerable amounts as well. It is also seen that with the higher means in July through September there is lower overall variability (i.e., a lower coefficient of variation). The higher variability during the fringe months of the monsoon (May to June and October to November) is likely due to typically having no rainfall or else receiving considerable amounts of rain (25--100 mm). Nevertheless, Table 4 documents that the June to September period chosen for inclusion in the Western Sahel Rainfall Index covers the large majority of the annual rainfall of this region.

b. Association with Tropical Cyclones

The concurrent associations which the Western Sahel Index has with various Atlantic basin tropical cyclone parameters are presented in Table 5 , including a detrended analysis. The goal of focusing on the relationship seen in the Lamb Index is achieved. In agreement with Fig. 3 and 4, the association is strongest for the most intense hurricane activity and weakest for the numbers of named storms. Figure 8depicts the scatter plot of the Western Sahel rainfall--intense hurricane days relationship, where 58 percent of the variance between the two are explained by a linear regression(Figure 9). Note that except for named storms, all parameters are significant beyond the 0.005 level using the one-tailed test (Spiegel, 1988).

However, the methodology of selecting a posteriori the region with the highest association induces artificial skill into the association (Davis, 1976; Shapiro, 1984). It is likely that the amount of bias is not large, as the Lamb Index (developed without regard to the intense hurricane activity) also showed a linear correlation of r = 0.64 with intense hurricane activity. Therefore, a caveat is recognized that significance testing for the relationships may slightly overestimate the degree of association. Additionally, independent data, though less reliable, verify the existence of the association as discussed later.

One uncertainty regarding this relationship is whether the strong correlations are simply due to coexisting trends in the data sets. One test that can be done is to break the record into two periods and then check the correlations. The obvious years that the break should be used would be after 1969, as the long term drought began in 1970 and has proceeded relatively uninterrupted since. The correlations with intense hurricane days then become:

which are just moderately lower than the overall relationship of r = 0.76 (a reduction of 14 percent of the variance explained). Since the trend is essentially removed by stratifying the data into two separate time periods, this confirms that the relationship is not trend dependent.

A second method for checking for trend induced associations would be to remove any linear trend from both data sets. Table 5 showed the detrended time series of the Western Sahel Rainfall Index. Though both the rainfall index and the intense hurricane activity show very substantial decreasing linear trends, removal of the trends has little effect. Detrended intense hurricane days correlate with the detrended Western Sahel Rainfall Index at r = 0.68 (Fig. 9), again moderately lower than the r = 0.76 value presented earlier for the original relationship. The remaining tropical cyclone parameters Table 5 show even less of a reduction in the detrended analysis.

For independent verification of the association, rainfall data from Landsea et al. (1991) can be utilized in conjunction with reliable data regarding landfalling intense hurricanes along the US East Coast (see Fig. 10). In the study, a five station Western Sahel rainfall index was created for the years 1899 to 1990. For the years 1949 to 1990, the five station index correlates at r = 0.91 (r = 0.85 in a detrended analysis) versus the larger 38-station index shown here. Thus, the five-station index likely represents the rainfall time series adequately. For intense hurricanes striking the US East Coast, the 38-station index correlates at r = 0.45 (r = 0.29 in a detrended analysis, significant beyond the 0.05 level) for the years 1949 to 1990. The five-station index correlates independently at r = 0.20 (r = 0.19 in a detrended analysis, significant beyond the 0.10 level) versus the US East Coast intense hurricanes for the years 1899 to 1948. Thus, after accounting for uncertainties due to less reliable rainfall and tropical cyclone data in the earlier period and after removing the substantial trend mainly in the later period, a confirmation of the Western Sahel--Atlantic tropical cyclone association is seen in earlier independent data.

c.`Wet' Versus `Dry' Years

To facilitate discussions of `wet' and `dry' Western Sahel rainfall years, the June to September Western Sahel Rainfall Index values are ranked from wettest to driest for 1949 to 1990 in Table 6 . This will allow composite analyses to be performed for groupings of wet versus dry seasons.

Another use of this ordering by rainfall is to use rank correlations versus tropical cyclone activity. The correlation coefficient between ranked rainfall data and ranked intense hurricane days is r = 0.81, with 65% of the variance explained. This is slightly higher than the linear correlation between the two (r = 0.76), suggesting that a non-linear relationship as measured by the rank correlation may be the best description of the association.

The reason for compositing parameters with respect to the wettest and driest Western Sahel rainfall seasons is that physical differences between the two regimes are accentuated. In a single season other factors (such as El Niño, stratospheric QBO, etc) which show as much control on the tropical cyclone variance as the Sahel rainfall may obscure the tropical cyclone---African rainfall signal. This problem is much reduced in a composite analysis. For the following analyses, the ten wettest years (mean Western Sahel Index value of sigma = 0.95) are contrasted with the ten driest years (mean value of sigma = --0.94).

Table 7 details how various tropical cyclone parameters differ in the ten wettest and ten driest Western Sahel years. Significant differences are seen in every category but are most pronounced for the strongest hurricane activity: a five to one ratio in all intense hurricane numbers and a ten to one ratio in the intense hurricane days. Again, the statistical significance may be slightly exaggerated because of the rainfall region selection methodology. While the statistical significance for landfalling systems are not as high as the entire Atlantic basin statistics, this is understandable due to the small landfalling data set being tested. Day to day characteristics of the mean steering flow are crucial to allowing landfall along the U.S. coastline; thus, variability on a seasonal basis for U.S. landfalling cyclones may contain more `noise'.

The U.S. landfalling intense hurricanes can be roughly separated into two regions: the U.S. East Coast (including the Florida peninsula) and the U.S. Gulf Coast (including the panhandle of Florida). Intense hurricanes which strike the two regions have differing characteristics such as the time of year which they come on shore and the origins of the storms (Landsea, 1991b). Figure 10 illustrates the approximate separation point between the East and Gulf Coasts for intense hurricane landfall. Note that while the East Coast experiences extreme differences in wet versus dry years in Table 7 , the Gulf Coast shows only a moderate modification. Similarly to the U.S. East Coast, the Caribbean Sea region also shows a strong modulation of hurricane and intense hurricane numbers with respect to Western Sahel rainfall. Since wet/dry differences in tropical cyclone frequency (Fig. 11) are accentuated with a component of duration (e.g. the named storm days, hurricane days, intense hurricane days, and HDP in Table 7 ), the cyclones are affected in their longevity in addition to their intensity. This assertion is supported by Table 8 , in which the durations of the various tropical cyclones are contrasted by intensity. The most striking results are seen in intense hurricane category where wet regime intense hurricane lasted over twice as long on average than dry regime intense hurricanes. Figure 13 demonstrates the combination of more numerous and longer lasting intense hurricanes in the wettest years versus the fewer, shorter lived intense hurricanes in the driest years by the paths these cyclones took. This figure also highlights the differing number of strikes along the U.S. coastlines and the Caribbean Sea.

One possible mechanism that can account for the variations of tropical cyclones with Western Sahel rainfall is the differences which occur in the genesis. Although just more than half of the tropical storms and hurricanes in the Atlantic basin have been clearly demonstrated to originate from easterly waves, over three-quarters of all intense hurricanes have their origin from these waves (Landsea, 1991b).

A method to study the Western Sahel's rainfall effect on tropical cyclone genesis is to stratify the origins of tropical cyclones by Western Sahel Rainfall Index. However, as noted earlier, reliable estimates of the genesis point of Atlantic basin cyclones has only been available since 1967. Utilizing the ten wettest versus the ten driest Sahel years would not be possible because the majority of the wet years occurred before that date. An alternative would be to name all years since 1967 from Table 2 with positive rainfall anomalies as ``wet'' years and those seasons with negative anomalies as ``dry'' years. This stratification splits the 24 years available into 5 wet years (with a mean value of Sigma = 0.48) and 18 dry years sigma= --0.68).

Table 9 shows the analysis of these wet and dry year composites. What is uncovered is a relative 10 to 20 percent higher incidence of easterly wave contribution to named storms, hurricanes, and intense hurricane formation when the Western Sahel is wet. Also, wet years show dramatic increases over dry years in the absolute numbers of easterly wave spawned cyclones: 54% more named storms, 93% more hurricanes, and 136% more intense hurricanes. Thus, part of the increase in the total numbers of cyclones spawned in wet years is due to more storms forming from easterly waves. The non-African-wave spawned cyclones, though not explicitly described in the table, show a slightly higher incidence in the dry years. It is hypothesized that in the wet years general circulation changes and local conditions over West Africa favor more easterly waves to become tropical cyclones, especially those developing into intense hurricanes. During the dry years, the general circulation is unfavorable for easterly wave spawned storm development, while the midlatitude spawned systems are slightly enhanced.

7. Physical Mechanisms

Two mechanisms appear to account for the consistent year-to-year covariation between Western Sahel rainfall and intense Atlantic basin hurricane activity (Gray, 1990c). The first, as illustrated in Figure 12, is that the general circulation is likely altered to be unfavorable to tropical cyclogenesis and intensification during Sahel drought years and more favorable during abundant rainfall years in the Sahel. During the drought years, stronger upper tropospheric westerly winds are developed which typically cause more vertical wind shear -- a feature that has long been known (e.g. Gray, 1968) to be detrimental to tropical cyclones. As Kidson (1977) pointed out, dry years in the Sahel are typically accompanied by a weaker 200 mb tropical easterly jet over West Africa. Preliminary evidence of the circulation differences in the Atlantic basin are seen in Figure 13which shows a comparison of Western Sahel rainfall and Caribbean basin 200 mb zonal wind anomalies. Note that the higher amounts of Western Sahel rainfall are associated with easterly zonal wind anomalies.

Another plausible influence of Western Sahel monsoonal strength variations on tropical cyclone activity is in the interannual modification of easterly waves. These waves were shown by Dunn (1940), who termed them westward traveling ``isallobaric waves'', to act as the `seedling' circulations for tropical storms and hurricanes in the Atlantic basin. It was shown in Table 9 that over 90% of all intense hurricanes originate from easterly waves in wet Western Sahel years.

Since it has now been established that dry years in the Sahel correspond to less intense hurricane seasons, it is possible that: 1) the number of easterly waves per year is decreased (as suggested by Druyan, 1989), 2) that the mean latitude where the waves travel is altered, or 3) the amplitudes of the waves are diminished (also as suggested by Druyan). The first hypothesis appears to not be valid. Avila and Clark (1989) have shown that the number of waves originating over Africa is very stable. Yearly averages of 58 waves with a standard deviation of only 5 waves are observed. The second hypothesis is possible but not likely to affect tropical cyclone numbers and strengths unless the mean latitude is altered substantially. Personal communication from satellite experts R. Zehr and L. Avila do not suggest that this has happened.

Thus, the third hypothesis is the one needing most consideration. Figure 14 gives an idealized view that in wet Western Sahel years (June to September) a larger number of waves emanating from Africa have strong amplitudes (reflected in the streamlines at 700 mb and surface pressure) and have more concentrated, persistent deep convection. Some of these strong waves eventually develop into intense Atlantic hurricanes. In dry years, substantially fewer waves would be so organized and would show weaker or negligible convection as well as having a weaker signature in the 700 mb flow field and surface pressure.

The next obvious question then is do weaker easterly waves cause the Sahel drought (as the waves with their embedded squall lines are the major contributors to the monsoonal rainfall) or are weaker easterly waves a result of the drought conditions? It is also very possible that there is a feedback between the two and that there is no clear cause and effect. This question remains open at this time, but research is currently being conducted to help answer it.

8. Summary

Seasonal Atlantic basin tropical cyclone activity, especially that of intense hurricanes, is strongly related to concurrent Western Sahel rainfall. The association is shown to be strong even after substantial linear trends have been removed from the data sets. Additionally, an independent analysis of earlier data confirms the existence of the tropical cyclone--Western Sahel rainfall association. Plausible physical mechanisms have also been introduced to explain the association. An understanding of intense hurricane variations is extremely important because of the huge potential destruction these storms can cause and the recent emphasis on possible greenhouse gas warming impacts upon their occurrence.

This new emphasis has taken the form of an American Meteorological Society and the University Corporation for Atmospheric Research joint statement (AMS Council and UCAR Board of Trustees, 1988) suggesting that a potential greenhouse gas impact would be ``a higher frequency and greater intensity of hurricanes''. Recent devastating cyclones with Hurricane Gilbert in 1988, the strongest storm ever measured in the Western Hemisphere (Willoughby et al., 1989), and Hurricane Hugo in 1989, ``the most costly hurricane in the U.S. history'' (Case and Mayfield, 1990), have only raised concerns higher. However, with the results shown here, the modulation of intense hurricane activity in the Atlantic basin is codependent upon rainfall conditions in the Sahel. It is sadly ironic that when the Western Sahel obtains a reprieve from its multidecadal drought with a return of significant rainfall, it is very probable that the Atlantic basin, especially along the U.S. East Coast and the Caribbean islands will experience many more destructive intense hurricanes. One should not identify a temporary return of intense hurricanes such as occurred in 1988 and 1989 as being even partially influenced by greenhouse gas warming. The association recognized here is probably not anthropogenic, but is likely due to the natural variations of the atmospheric-oceanic general circulation.

Acknowledgements The bulk of this paper was gleaned from research toward the first author's Masters Thesis at Colorado State University with the second author's tutelage (Landsea, 1991a). The first author had the privilege of attending the WMO's Symposium on ``Meteorological Aspects of Tropical Droughts with Emphasis on Long-Range Forecasting'' at Niamey, Niger in the spring of 1990. While there I had many excellent discussions of the Western Sahel rainfall and tropical cyclone association with Yinka Adebayo, Birama Diarra, Leonard Druyan, Graham Farmer, Stefan Hastenrath, Peter Hutchinson, Peter Lamb, Kevin Lane, Robert Livezey, Tim Palmer, Randy Peppler, Chet Ropelewski, M.V. Sivakumar, and Neal Ward. My time was well spent as I gained a greater appreciation for the severity of Sahelian drought conditions as well as for the people attempting to solve the mysteries of long term droughts. It was truly a wonderful experience.

The authors are grateful for the African rainfall and Atlantic basin tropical cyclone data as detailed in section 3 and Appendix A. Much valuable computer programming expertise was contributed by Richard Taft, William Thorson, and Todd Massey. Barbara Brumit, Laneigh Walters, and Judy Sorbie-Dunn have provided important manuscript, data analysis, and figure drafting assistance. Additional valuable discussions were had with Professors Wayne Schubert, Roger Pielke, Duane Stevens, David Randall, Paul Mielke, Kenneth Berry and Edward Prill; with Research Associates John Sheaffer and Paul Ciesielski; with fellow Gray project students Steve Hodanish, Ray Zehr, Dan Mundell, Steve Hallin, John Knaff, and Mike Fitzpatrick here at CSU; and with Lloyd Shapiro and Stan Goldenberg at the Hurricane Research Division in Miami.

Appendix A

CSU AFRICAN RAINFALL DATA BASE

The primary data base for African rainfall is the ``World Monthly Surface Station Climatology'' (WMSSC) managed by W.M.L. Spangler and R.L Jenne at the National Center for Atmospheric Research (NCAR). Though WMSSC provides surface data (precipitation, temperatures, surface pressures, sea level pressures, and others) globally, our interest was primarily precipitation (rainfall) data over Africa. The data collection, extending from the mid-1800's to 1988, is available mainly from the efforts of Prof. S.E. Nicholson of Florida State University for additional historical data. Including stations located on nearby islands (Azores, Canary, Madeira, Seychelles, and Madagascar), the WMSSC data set has 584 African stations with rainfall information. The Monthly Climatic Data for the World (MCDW) by the National Climatic Data Center (U.S. Dept. of Commerce, 1989) has provided an official monthly updating for the same stations through December 1989. The data presented in MCDW is essentially what will be used in the next updating of WMSSC.

A secondary data set that supplements the WMSSC information was supplied to us by G. Farmer of the A.I.D. FEWS Project in Arlington, VA. He provided us with additional information on 36 WMSSC stations and data on 75 agricultural raingauges that have been available since 1951. These stations are all in the Sahel region from Senegal in the west to Chad in the east. In general, the quality on the 36 stations that overlapped with the WMSSC stations was higher in the AID stations (e.g. less rounding to the nearest 10 mm, less missing data, fewer unreasonable outliers). Therefore, in combining the data sets together, we chose the AID data as more reliable in the case where there was a 3 mm or greater difference between WMSSC and AID for monthly rainfall amounts.

Additionally, E. O. Oladipo of Ahmadu Bello University provided us with monthly rainfall data for stations in Nigeria. Recently, Nigeria has been a difficult country to obtain data for. Professor Oladipo has graciously provided us with data for 13 stations from 1970 to 1988. Since the data was coming directly from the stations involved, this rainfall data was considered to be of high quality.

P.J. Lamb also provided us with the 20 stations used in his index of West Sahel rainfall. He has used this data in connection with his research in the underlying physical mechanisms for Sahel drought (Lamb, 1982) as well as in a real--time monitoring of the Sahel in conjunction with CAC (Lamb et al., 1990). The data was very similar in quality to WMSSC yet provided information for some stations which had missing data. Accordingly, we utilized Lamb's data to fill data gaps but considered it lower priority than WMSSC, AID, and Nigeria data in duplications.

The final addition that was used in this study was more recent data provided us by D. Miskus and R.J. Tinker of CAC. They provided us with CAC's best estimates of monthly rainfall as they are able to infer from daily reports on the Global Telecommunications System (GTS) going back to 1960. This information provided us excellent preliminary data that will temporarily fill in gaps until we receive updated reports from NCDC. Miskus and Tinker have also been very generous in providing us data in real-time so as to enable us to monitor the Sahel (as it now relates to Prof. W.M. Gray's (1990a, 1990b, 1990d) seasonal hurricane forecasts). Since these are estimates (and can be somewhat erroneous due to missing and mistaken reports) and not the official report, this data is only used in absence of any other information (i.e., WMSSC, AID, Nigeria and Lamb).

Thus, the data sets used in this paper and the relative order of priority can be summarized as:

highest priority:  Nigeria Data 
                   AID Data  
                   WMSSC and MCDW Data 
                   Lamb Data 
lowest priority:   CAC Estimated Data 
               

Landsea (1991) has additional detailed information regarding individual station rainfall data.

REFERENCES

AMS Council and UCAR Board of Trustees, 1988: The changing atmosphere - challenges and opportunities. Bull. Amer. Meteor. Soc ., 69, 1434--1440.

Avila, L. A. and G. B. Clark, 1989: Atlantic tropical systems of 1988. Mon. Wea. Rev ., 117, 2260--2265.

Ballenzweig, E. M., 1959: Relation of long-period circulation anomalies to tropical storm formation and motion. J. Meteor ., 16, 121--139.

Carlson, T. N. and J. D. Lee, 1978: Tropical Meteorology. Pennsylvania State University, Independent Study by Correspondence, University Park, Pennsylvania, 387 pp.

Case, R. A. and H. P. Gerrish, 1988: Atlantic hurricane season of 1987. Mon. Wea. Rev ., 116, 939--949.

Case, R. A. and M. Mayfield, 1990: Atlantic hurricane season of 1989. Mon. Wea. Rev ., 118, 1165--1177.

Davis, R. E., 1976: Predictability of sea-surface temperature and sea-level pressure anomalies over the North Pacific Ocean. J. Phys. Oceanogr. , 6, 249--266.

Druyan, L. M., 1989: Advances in the study of Subsaharan drought. Int. J. Climate, 9, 77--90.

Dunn, G. E., 1940: Cyclogenesis in the tropical Atlantic. Bull. Amer. Meteor. Soc ., 21, 215--229.

Dunn, G. E. and Staff, 1965: The hurricane season of 1964. Mon. Wea. Rev., 93, 175--187.

Dvorak, V. F., 1977: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev .,103 , 420--430.

Dvorak, V. F., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Technical Report NESDIS 11, U.S. Dept. of Comm., Washington, DC, 47 pp.

Garriott, E. B., 1906: The West Indian hurricanes of September, 1906. Mon. Wea. Rev .,34, 416--423.

Gray, W. M., 1968: Global view of the origins of tropical disturbances and storms. Mon. Wea. Rev .,96 , 669--700.

Gray, W. M., 1984a: Atlantic seasonal hurricane frequency. Part I: El Nino and 30mb Quasi-Biennial Oscillation influences. Mon. Wea. Rev ., 112 , 1649--1668.

Gray, W. M., 1984b: Atlantic seasonal hurricane frequency. Part II: Forecasting its variability. Mon. Wea. Rev ., 112 , 1669--1683.

Gray, W. M., 1987: Cyclic trends in hurricane destruction potential and discussion of factors which will be used in the 1987 seasonal forecast. 9th National Hurricane Conference, Orlando, FL, 3 April 1987, 29 pp.

Gray, W. M., 1989: Summary of 1989 Atlantic tropical cyclone activity and seasonal forecast verification. Colo. State Univ., Dept. of Atmos. Sci. Paper issued on 20 November 1989, Fort Collins, CO 80523, 37 pp.

Gray, W. M., 1990a: Forecast of Atlantic seasonal hurricane activity for 1990. Colo. State Univ., Dept. of Atmos. Sci. Paper issued on 5 June 1990, Fort Collins, CO 80523, 15 pp.

Gray, W. M., 1990b: Updated forecast of Atlantic seasonal hurricane activity for 1990. Colo. State Univ., Dept. of Atmos. Sci. Paper issued on 3 August 1990, Fort Collins, CO 80523, 15 pp.

Gray, W. M., 1990c: Strong association between West African rainfall and U.S. landfall of intense hurricanes. Science,249 , 1251-1256.

Gray, W. M., 1990d: Summary of 1990 Atlantic tropical cyclone activity and seasonal forecast verification. Colo. State Univ., Dept. of Atmos. Sci. Paper issued on 28 November 1990, Fort Collins, CO 80523, 29 pp.

Hastenrath, S., 1988: Climate and Circulation of the Tropics. Reidel, Dordrecht, Boston, Lancaster, Tokyo, Second Printing, 455 pp.

Hastenrath, S., 1990a: The relationship of highly reflective clouds to tropical climate anomalies. J. Climate , 3 , 353--365.

Hastenrath, S., 1990b: Tropical climate prediction: A progress report, 1985--90. Bull. Amer. Meteor. Soc ., 71 , 819--825.

Hayward, D. F. and J. S. Oguntoyinbo, 1987: The climatology of West Africa . Barnes and Noble Books, Totowa, New Jersey, 217 pp.

Hebert, P. J. and G. Taylor, 1975: Hurricane experience levels of coastal county populations--Texas to Maine. Special Report, National Weather Service Community Preparedness Staff and Southern Region, 153 pp.

Jarvinen, B. R., C. J. Neumann and M. A. S. Davis, 1984: A tropical cyclone data tape for the North Atlantic basin, 1886--1983: Contents, limitations, and uses. NOAA Technical Memorandum NWS NHC 22, Miami, FL, 21 pp.

Kidson, J. W., 1977: African rainfall and its relation to the upper air circulation. Quart. J. Roy. Meteor. Soc ., 103 , 441--456.

Kraus, E. B., 1977: Subtropical droughts and cross-equatorial transports. Mon. Wea. Rev ., 105, 1009--1018.

Lamb, P. J., 1982: Persistence of Subsaharan drought. Nature , 299 , 46--48.

Lamb, P. J., 1985: Rainfall in Subsaharan West Africa during 1941--83. Z. Gletsh. Glaz ., 21 , 131--139.

Lamb, P. J., C. F. Ropelewski and D. H. Portis, 1990: Real-time monitoring of Subsaharan rainy season. Extended Abstracts of Papers Presented at the Third WMO Symposium on Meteorological Aspects of Tropical Droughts with Emphasis on Long-Range Forecasting, WMO/TD - No. 353, 163--166.

Landsea, C. W., 1991a: West African monsoonal rainfall and intense hurricane associations. MS Thesis, Atmospheric Science Paper No. 484, Colo. State Univ., Ft. Collins, CO 80523, 272 pp.

Landsea, C. W., 1991b: Intra and interseasonal climatology of intense Atlantic hurricanes. To be submitted to J. Climate .

Landsea, C. W., W. M. Gray, P. W. Mielke, Jr., and K. J. Berry, 1991: Multidecadal variations of Sahel monsoon rainfall and intense US landfalling hurricanes. Submitted to J. Climate , 26 pp.

Lawrence, M. B. and J. M. Gross, 1989: Atlantic hurricane season of 1988. Mon. Wea. Rev ., 117 , 2248--2259.

Mayfield, M. and M. B. Lawrence, 1991: Atlantic hurricane season of 1990. Mon. Wea. Rev. , 119 , 2014--2026.

Namias, J., 1955: Secular fluctuations in vulnerability to tropical cyclones in and off New England. Mon. Wea. Rev ., 83 , 155--162.

Neumann, C. J., B. R. Jarvinen, A. C. Pike and J. D. Elms, 1987: Tropical cyclones of the North Atlantic Ocean, 1871-1986. National Climatic Data Center in cooperation with the National Hurricane Center, Coral Gables, FL 186 pp.

Nicholson, S. E., 1986: The spatial coherence of African rainfall anomalies: Interhemispheric teleconnections. J. Climate Appl. Meteor ., 25 , 1365--1381.

Ray, C. L., 1935: Relation of tropical cyclone frequency to summer pressures and ocean surface-water temperatures. Mon. Wea. Rev ., 63 , 10--12.

Riehl, H., 1945: Waves in the easterlies and the polar front in the tropics. Misc. Rept., No. 17, Dept. Meteorology, University of Chicago, 79 pp.

Shapiro, L. J., 1982a: Hurricane climatic fluctuations. Part I: Patterns and cycles. Mon. Wea. Rev. , 110 , 1007--1013.

Shapiro, L. J., 1982b: Hurricane climatic fluctuations. Part II: Relation to large-scale circulation. Mon. Wea. Rev., 110 , 1014--1023.

Shapiro, L. J., 1984: Sampling errors in statistical models of tropical cyclone motion: A comparison of predictor steering and EOF techniques. Mon. Wea. Rev., 112 , 1378--1388.

Shapiro, L. J., 1989: The relationship of the quasi-biennial oscillation to Atlantic tropical storm activity. Mon. Wea. Rev ., 117 , 1545--1552.

Sheets, R. C., 1990: The National Hurricane Center -- Past, present, and future. Wea. Fore ., 5 , 185--232.

Shinoda, M., 1989: Annual rainfall variability and its interhemispheric coherence in the semi-arid region of tropical Africa: Data updated to 1987. J. Meteor. Soc. Japan , 67 , 555--564.

Simpson, R. H., 1974: The hurricane disaster potential scale. Weatherwise , 27 , 169--186.

Simpson, R. H., N. Frank, D. Shideler and H. M. Johnson, 1968: Atlantic tropical disturbances, 1967. Mon. Wea. Rev ., 96 , 251--259.

Spiegel, M. R., 1988: Statistics--Schaums Outline Series . Second Edition, McGraw--Hill, Inc., New York, NY, 504 pp.

U.S. Department of Commerce, NOAA, NCDC, National Environmental Satellite, Data, and Information Service, 1989: Monthly Climatic Data for the World, Asheville, NC, 24--36, 42, No. 1--12.

U.S. Department of Commerce, NOAA, Federal Coordinator for Meteorological Services and Supporting Research, 1977: National Hurricane Operations Plan, FCM 77-2, Washington, D.C., 104 pp.

U.S. Department of Commerce, NOAA, NWS, NMC, CAC, 1990: Weekly Climate Bulletin, 90, Washington, D.C.

Willoughby, H. E., J. M. Masters and C. W. Landsea, 1989: A record minimum sea level pressure observed in hurricane Gilbert. Mon. Wea. Rev ., 117 , 2824--2828.

TABLES

TABLE 1
Maximum sustained wind speed, minimum surface pressure, storm surge, and general damaging effects for the five Saffir--Simpson (Simpson, 1974) Hurricane Scale values.
Saffir-Simpson
Category
Maximum Sustained
Wind Speed (m/s)
Minimum Surface
pressure (mb)
Storm Surge
(m)
Potential
Damaging Effects
1 33 to 42 >980 1.0 to 1.7 Minimal
2 43 to 49 979 to 965 1.8 to 2.6 Moderate
3 50 to 58 964 to 945 2.7 to 3.8 Extensive
4 59 to 69 944 to 920 3.9 to 5.6 Extreme
5 >69 < 920 >5.6 Catastrophic


TABLE 2
Mean standard deviations and data availability (out of a possible 38 stations per year) for the June to September Western Sahel rainfall.
Year Standard
Deviation
Number of
Stations
Year Standard
Deviation
Number of
stations
1949 -0.10 24 1970 -0.45 37
1950 1.49 24 1971 -0.30 37
1951 0.32 37 1972 -1.16 38
1952 1.00 37 1973 -0.80 38
1953 0.55 38 1974 -0.23 38
1954 0.77 38 1975 0.34 38
1955 1.46 38 1976 -0.56 38
1956 0.40 38 1977 -0.91 37
1957 0.52 38 1978 -0.18 38
1958 1.28 37 1979 -0.54 38
1959 0.06 36 1980 -0.68 37
1960 0.51 35 1981 -0.34 37
1961 0.77 38 1982 -0.88 37
1962 0.19 37 1983 -1.31 37
1963 -0.09 38 1984 -1.16 37
1964 0.88 38 1985 -0.50 37
1965 0.54 37 1986 -0.37 37
1966 0.07 38 1987 -0.76 26
1967 0.75 37 1988 0.18 26
1968 -0.75 37 1989 0.58 25
1969 0.55 37 1990 -0.95 25


TABLE 3
The 38 stations used in the June to September Western Sahel Rainfall Index: station name and country, June to September mean and standard deviation (in mm), years of data in analysis, correlation coefficients versus the Western Sahel Index itself, and correlation coefficient versus intense hurricane days.
Rainfall data
Correlation coefficent
Station Mean SD Years Versus index Versus intense
hurricane days*
Nioro Du Sahel, Mali 483.9 159.75 42 0.809 0.511
Kayes, Mali 611.3 121.28 42 0.641 0.475
Kita, Mali 865.6 185.37 42 0.705 0.46
Segou, Mali 591.7 126.96 42 0.774 0.548
San, Mali 635.0 108.13 42 0.602 0.571
Kenieba, Mali 1045.2 241.41 40 0.768 0.50
Bamako/Senou, Mali 814.7 169.17 42 0.647 0.461
Koutiala, Mali 789.0 171.43 42 0.676 0.630
Bougouni, Mali 926.1 174.60 42 0.677 0.579
Sikasso, Mali 925.4 143.01 41 0.554 0.454
Atar, Mauritania 66.0 43.27 42 0.496 0.344
Akjoujt, Mauritania 58.7 45.29 40 0.628 0.555
Nouakchott, Mauritania 86.8 54.98 42 0.814 0.578
Boutilimit, Mauritania 133.3 76.44 42 0.734 0.615
Rosso, Mauritania 221.8 104.86 40 0.700 0.563
Kiffa, Mauritania 280.4 118.51 42 0.780 0.46
Saint Louis, Senegal 261.3 128.88 42 0.572 0.442
Podor, Senegal 232.3 120.71 42 0.706 0.620
Linguere, Senegal 399.7 115.33 42 0.747 0.632
Matam, Senegal 405.2 177.28 42 0.674 0.513
Dakar/Yoff, Senegal 421.2 177.83 42 0.853 0.680
Thies, Senegal 527.1 202.63 37 0.864 0.643
Diourbel, Senegal 544.5 177.29 42 0.864 0.632
Kaolack, Senegal 615.0 183.10 40 0.753 0.541
Tambacounda, Senegal 725.2 168.38 42 0.694 0.466
Bathurst/Yundum, Gambia 998.4 286.76 42 0.882 0.552
Bissau Airport, Guinea-Bissau 1542.1 337.73 40 0.773 0.574
Bansang, Gambia 823.7 290.44 34 0.781 0.562
Georgetown, Gambia 811.7 189.83 31 0.783 0.518
Basse Met, Gambia 842.1 221.00 34 0.774 0.698
Boghe, Mauritania 250.4 89.88 36 0.749 0.610
Selibaby, Mauritania 493.5 145.51 36 0.782 0.495
Louga, Senegal 336.3 157.19 31 0.825 0.570
Mbour, Senegal 576.8 216.54 36 0.834 0.538
Nioro Du Rip, Senegal 685.6 184.95 33 0.807 0.495
Velingara Casamance, Senegal 863.6 208.37 34 0.691 0.689
Sedhiou, Senegal 1094.6 303.71 36 0.764 0.553
Bambey Met, Senegal 538.5 151.72 36 0.871 0.586

*Significance level is 0.005 for all stations except 0.025 for Atar,Mauritania.


TABLE 4
Monthly means, standard deviations, ratios, and coefficients of variability for the 38 stations Western Sahel Rainfall Index.
May Jun Jul Aug Sep Oct Nov Jun-Sep May-Nov
Mean 18 66 149 217 162 51 4 595 668
Standard Dev. 15 38 65 94 74 48 9 164 177
Coefficient Var.(%) 83 57 44 43 46 94 225 28 26
Percent of Jun-Sep Mean -- 11 25 36 28 -- -- -- --
Percent of May-Nov Mean 3 10 22 32 24 8 1 89 --


TABLE 5
Regression equations and correlation coefficients of June to September Western Sahel Rainfall Index (x) versus Atlantic basin tropical cyclone parameters (y) using data from 1949--1990 for both the original data sets and the detrended analysis. (Asterisks refer to significance level: 0.100 for `*', 0.025 for `**' and 0.005 for `***'.)
Tropical Cyclone
Parameter
Correlation
Coefficient
Regression
equation
Standard
error of y
Original data
Named Storms 0.35** y = 9.42 + 1.406 x ± 2.8
Named Storm Days 0.56*** y = 47.15 + 15.052x 16.7
Hurricanes 0.47*** y = 5.85 + 1.404x 2.0
Hurricane Days 0.67*** y = 23.73 + 12.192x 10.1
Int. Hur. 0.71*** y = 2.47 + 1.805x 1.3
Int. Hur. Days 0.76*** y = 5.68 + 5.713x 3.6
HDP 0.73*** y = 73.21 + 47.936x 33.8
Detrended Data
Named Storms 0.34** y = 0.05 + 1.862 x ± 2.8
Named Storm Days 0.54*** y = 0.12 + 19.053x 16.7
Hurricanes 0.41** y = --0.02 + 1.575x 2.0
Hurricane Days 0.60*** y = 0.13 + 13.681x 10.1
Int. Hur. 0.59*** y = 0.03 + 1.766x 1.3
Int. Hur. Days 0.68*** y = 0.04 + 6.115x 3.6
HDP 0.65*** y = 0.52 + 51.916x 33.8


TABLE 6
June to September Western Sahel Rainfall Index ranked by rainfall amounts in mean standard deviations from 1949 to 1990.
Rank Year Value
Index
Rank Year Index
Value
1. 1950 1.49 22. 1963 -0.09
2. 1955 1.46 23. 1949 -0.10
3. 1958 1.28 24. 1978 -0.18
4. 1952 1.00 25. 1974 -0.23
5. 1964 0.88 26. 1971 -0.30
6. 1954 0.77 27. 1981 -0.34
7. 1961 0.77 28. 1986 -0.37
8. 1967 0.75 29. 1970 -0.45
9. 1989 0.58 30. 1985 -0.50
10. 1969 0.55 31. 1979 -0.54
11. 1953 0.55 32. 1976 -0.56
12. 1965 0.54 33. 1980 -0.68
13. 1957 0.52 34. 1968 -0.75
14. 1960 0.51 35. 1987 -0.76
15. 1956 0.40 36. 1973 -0.80
16. 1975 0.34 37. 1982 -0.88
17. 1951 0.32 38. 1977 -0.91
18. 1962 0.19 39. 1990 -0.95
19. 1988 0.18 40. 1972 -1.16
20. 1966 0.07 41. 1984 -1.16
21. 1959 0.06 42. 1983 -1.31


TABLE 7
Summary statistics on the variability of tropical cyclone parameters from the ten wettest Western Sahel years and the ten driest years from 1949 to 1990. (Asterisks refer to significance level: 0.100 for `*', 0.025 for `**' and 0.005 for `***'.)
Tropical cyclone
Parameter
Climatology
(42 years)
Wettest Years'
mean
Percent
of normal
Driest years'
mean
Percent
normal
Ratio
wet/dry
Named Storms 9.41 11.2 120 7.7 83 1.45***
Named Storm Days 47.2 67.0 143 34.0 73 1.97***
Hurricanes 5.85 8.0 138 4.6 79 1.74***
Hurricane Days 23.8 39.1 165 13.2 55 2.96***
Intense Hurricanes 2.48 4.5 180 0.9 36 5.00***
Intense Hurricane Days 5.69 12.4 214 1.2 21 10.33***
HDP 73.4 131.2 175 33.9 45 3.87***
U.S. Landfalling
Named Storms 3.30 3.4 103 1.9 58 1.79*
Hurricanes 1.82 2.1 130 0.7 43 3.00**
Intense Hurricanes 0.72 1.2 167 0.2 28 6.00***
I. H. Gulf Coast 0.37 0.5 147 0.2 59 2.50*
I. H. East Coast 0.35 0.7 241 0.0 0 infty**
Caribbean Sea
Hurricanes 1.18 2.1 178 0.4 34 5.25***
Intense Hurricanes 0.71 1.5 211 0.2 28 7.50***


TABLE 8
Variation in durations of named storms, hurricanes, and intense hurricanes in the ten wettest and ten driest Western Sahel years from 1949 to 1990. (Asterisks refer to significance level: 0.100 for `*', 0.025 for `**' and 0.005 for `***'.
Tropical Cyclone
Parameter
Climatology
(42 Years)
Wettest Years'
mean
Percent
of normal
Driest years'
mean
Percent
of normal
Ratio
wet/dry
Named Storms 4.9 6.0 122 4.4 90 1.36***
Hurricanes 3.9 4.9 126 2.9 74 1.69***
Intense Hurricanes 2.3 2.8 122 1.2 52 2.33***


TABLE 9
Mean annual easterly wave contributions of tropical cyclone genesis for the Western Sahel Rainfall Index wet (5) and dry years (19) as well as the entire 24 year dataset.
Wet years Dry years Twenty-four years
Tropical Cyclone
intensity
All
disturbances (#)
Waves (#) Waves (%) All
disturbances (#)
Waves (#) Waves (%) All
disturbances (#)
Waves (#) Waves (%)
Named Storms 11.4 8.0 70 8.5 5.2 61 9.1 5.8 64
Hurricanes 7.2 5.8 81 5.0 3.0 60 5.4 3.6 67
Intense Hurricanes 2.8 2.6 93 1.4 1.1 80 1.7 1.5 87

LIST OF FIGURE CAPTIONS

1. Intense Atlantic hurricane activity from 1949 to 1990 by intense hurricanes (upper panel) and intense hurricane days (lower panel). The solid line is the least squares linear trend.

2. Smoothed values of linear correlation coefficient for individual station rainfall versus seasonal intense hurricane days for June, July, August, and September from 1949 to 1989. Contours are at r = ±0.15, 0.30 and 0.45.

3. Correlation coefficients of individual station June to September rainfall versus named storms, hurricanes, and intense hurricanes from 1949 to 1990. Contours indicate values of r = ±0.15, 0.30, 0.45, and 0.60. Positive correlations are within solid contours, while negative contours are indicated by dashed lines.

4. Same as Fig. 3 with June to September rainfall versus intense hurricane days.

5. Location of rainfall stations which comprise the Western Sahel Rainfall Index.

6. Mean standard deviations of rainfall for the 38 station June to September Western Sahel Index. The boldface line indicates the least squares best fit line to the data. Data presented are from 1949 to 1990.

7. Same as Fig. 6 with the downward linear trend removed from the data.

8. Scatter plot of 1949 to 1990 values of June to September Western Sahel Rainfall Index versus intense hurricane days. The dashed line is the least squares best fit.

9. Same as Fig. 8 with linear trends removed from both data sets.

10. U.S. coastal regions that experience differing responses of intense hurricane activity to Western Sahel rainfall. The approximate separation point is the Apalachee Bay of Florida.

11. Intense hurricane tracks by 6 seven year groupings based upon rankings of the June to September Western Sahel rainfall. Rain years 1--7 indicate the seven wettest years, while rain years 36--42 show the seven driest years.

12. Idealized portrayal of upper tropospheric wind patterns in wet (upper panel) versus dry (lower panel) Western Sahel years.

13. August to September Caribbean basin 200 mb Zonal Wind Anomalies (ZWA) versus the June to September Western Sahel rainfall for the years 1949 to 1990. (Negative ZWA indicate easterly anomalies and positive indicate westerly anomalies.)

14. Idealized portrayal of the easterly wave variations that are suggested to occur during wet (upper panel) versus dry (lower panel) Western Sahel years.


Return to Landsea biography
Return to Main page