Christopher W. Landsea
Department of Atmospheric Science, Colorado State University

William M. Gray
Department of Atmospheric Science, Colorado State University

Paul W. Mielke, Jr.
Department of Statistics, Colorado State University

Kenneth J. Berry
Department of Statistics, Colorado State University

Weather 49, 273-284, (1994)

For the US eastern and Gulf coasts, the beginning of June marks the start of a watchfulness of the typically tranquil subtropical weather for the possibility of a visit from a destructive hurricane. The recent examples of Hurricanes Hugo, striking the Virgin Islands, Puerto Rico, and South Carolina in the U.S., and Andrew, striking the Bahamas and Florida and Louisiana in the U.S., have provided ample evidence to coastal residents that Atlantic hurricanes are not to be ignored.

Until recently, coastal inhabitants had only the climatology of tropical cyclones as a guide as to how active a particular hurricane season would be. There was no way to tell whether the upcoming season would be very busy as in 1969 with 12 hurricanes or extremely quiet as in 1982 with only 2 hurricanes. Since the early 1980s, the second author and his colleagues at Colorado State University have translated a knowledge of the physical mechanisms for hurricane variability into a systematic scheme for the prediction of seasonal hurricane activity several months in advance.

  Previous forecasting results

With the completion of the 1993 hurricane season, ten years of seasonal forecasting have been attempted. The original forecasting procedures were written up in Gray (1984a,b), but have been substantially reworked and improved since. Forecast techniques have been developed from the analysis of data going back to 1950. The results so far have been encouraging. Table 1 presents the seasonal forecasts for both named storms and hurricanes from various starting times, along with their verification. Named storms are those tropical cyclones having one-minute average winds of at least 18ms-1 at the surface.

Usually, these tropical cyclones are given names to prevent confusion when more than one storm is present and to facilitate watches and warnings for the public. Named storms which intensify to at least 33 ms-1 winds are called hurricanes. Note that in general the forecasts have provided a better measure of the upcoming season than climatology and that the early August forecasts have been more skillful than the early June forecasts. Additionally, in the last two years we have also been issuing an extended-range forecast by early December of the previous year. We have also recently been issuing seasonal forecasts of intense (or major) hurricanes, those cyclones with sustained winds of at least 50 ms-1.

The early August forecast may seem to be more of a "nowcast" than a prediction when one recalls that the ``official" Atlantic hurricane season extends from June through November. However, an inspection of the seasonal variation of named storms and hurricanes reveals that only 11% and 6% of the annual named storm and hurricane activity (as measured by days in which these cyclones are present) respectively, occurs before 1 August on average (Landsea 1993).

Figure 1 demonstrates the extreme seasonality exhibited by intense hurricanes of the Atlantic basin. Less than 2% of the intense hurricane activity occurs before 1 August and 95% occurs in the three months of August through October. Additionally, the small amount of activity that does occur in June or July has absolutely no predictive value for the entire season: a busy (e.g., two or three named storms) June and July can precede a very active year (such as 1990 when 14 named storms occurred) or a very quiet year (such as 1986 when only six named storms were observed). Alternatively, quiescent (e.g., with no named storms observed) June and July years can either precede very active years (such as 1988 with 12 named storms) or very quiet seasons (such as 1983 with only four named storms observed).

While the results obtained so far have been very successful, forecasts issued in the coming years have room for further improvements. It is hoped that both the forecasts for named storms and hurricanes can be improved to where up to nearly two-thirds of the variability can be anticipated, in other words, errors on the order of only one named storm or hurricane on the forecast. Additionally, in the last few years, more emphasis has been placed on predicting the intense hurricanes. These hurricanes are responsible for over 70% of U.S. tropical cyclone caused damage even though they only account for 20% of the tropical cyclones affecting the United States (Landsea 1993).

In the development of the forecast, we wish to use the longest reliable time series of predictors and hurricane data that is possible. Because of our use of tropical stratospheric data, we begin our collection of information in 1950. Currently, the methodology which we use to produce the forecast is a cross-validated (or jackknifed) procedure in which each year is hindcasted independently of the other years of data available. For example, we would use the data from the years 1950-1967 and 1969-1993 to develop the linear equation that would test the predictors for the year 1968. We repeat this for all years available to get an idea of the skill expected. For future years (a true forecast), we then utilize all available years of data. This procedure is detailed in Gray et al. (1992a, 1993a, 1994).

  1 December forecasting

The 1 December forecast is based upon several predictors (Gray et al. 1992a). These predictors include those based upon the extrapolated state of the quasi-biennial oscillation (QBO) - the zonal winds at 50 mb, 30 mb and the vertical shear of the zonal winds between the two levels, August and September precipitation within the western Sahel, and August to November precipitation along the Gulf of Guinea; and a long-term forecast of El Niño-Southern Oscillation (ENSO) (Gray et al. 1993b) for the following year hurricane season. The recent addition of ENSO to the scheme adds four additional predictors:

from 5 N to 5 S and 90 W to 150 W averaged for the past 27 months.
Table 2 lists these predictive groupings and Fig. 2 shows the location of these various predictors.

The relationship that the QBO has to Atlantic hurricanes is hypothesized to be due to variations in the vertical wind shear between the upper troposphere and the lower stratosphere over the tropical North Atlantic during the height of the hurricane season, August through October. Because of the very regular changes of the QBO phenomena, successful long range extrapolations of the mean stratospheric zonal winds can be made almost a year in advance. For this 1 December forecast time, an extrapolation of mean following-year September QBO conditions is made based upon November information. An east QBO phase (typically lasting 12 to 15 months) forces strong easterly winds in the lower stratosphere between 10 and 15 N and with it large amounts of vertical wind shear from the upper troposphere to the lower stratosphere. The west QBO phase (typically lasting 13 to 16 months) allows for weak easterly winds in the tropical North Atlantic stratosphere and for low amounts of vertical wind shear. Large amounts of this vertical wind shear is hypothesized to inhibit genesis and intensification of tropical cyclones. Thus years of west QBO phases typically have 50% more named storms, 60% more hurricanes, and 200% more intense hurricanes than occur in east QBO phases within the Atlantic basin.

The two West African rainfall indices are needed for Atlantic tropical cyclone forecasting because of the intimate tie between concurrent seasonal amounts of intense hurricane activity and seasonal rainfall in the Sahel of West Africa (Landsea and Gray 1992). We have found that the previous year rainfall along the Gulf of Guinea and in the Sahel itself provides a very dependable indication of future Sahel rainfall (and thus Atlantic hurricane activity). The Sahel rainfall correlation to its previous year rainfall is reflected in the strong tendency for anomalies of precipitation to continue from year to year. This persistence is likely due to a combination of global sea surface temperature forcing (Lamb 1978; Folland et al. 1986) and changes in the land surfaces including desertification which may reinforce drought conditions (Nicholson 1988; Xue and Shukla 1993). The positive feedback between the Gulf of Guinea rainfall in August through November to Sahel rainfall/Atlantic hurricanes the following year appears to result from changes in available moisture for the North African monsoon through long-term storage in the soil and biosphere (Gray et al. 1992a). While the previous year Sahel rainfall accounts for only about 5% of the intense hurricane variability, the Gulf of Guinea rainfall anomalies provide a much stronger predictor of up to 40% of the variability forecasted in the intense hurricane activity.

Finally, the latest addition to the 1 December forecast methodology is an inclusion of the forecasted state of ENSO 10 months later. The previously observed eastern equatorial Pacific SSTA, sea level pressures measured at Darwin, Australia, 100 mb temperature anomalies at Singapore, and the earlier described QBO winds and Western Sahel rainfall can be used to give a quite skillful extended range forecast of ENSO conditions. The physical link between ENSO and Atlantic hurricanes is in the alteration of the upper tropospheric flow patterns. In general, when a warm episode (i.e. El Niño) is occurring, the atmosphere over the tropical North Atlantic and Caribbean are observed to have much higher tropospheric vertical shear due to increased westerlies at 200 mb. Conversely, when a cold episode (i.e. La Niña) is observed during the hurricane season, there is reduced vertical wind shear due to easterly 200 mb wind anomalies. As stated earlier, large vertical wind shear is detrimental to tropical cyclone formation and intensification. Moderate to strong El Niño events have been observed to reduced the number of hurricanes by 44% from non-El Niño years (Gray 1984a).

Overall, the 1 December forecasting should be able to provide insight into about 40% to 50% of the variability (or 45% to 60% in the newest version including ENSO effects) of the tropical cyclone activity. Fig. 3 demonstrates the observed differences in intense hurricanes for the ten hindcasts for the most active TC seasons and the ten hindcasts for the quietest TC seasons. Note the very large differences in observed intense hurricane tracks indicating a very substantial amount of skill present in these hindcasts. We feel that this is a startling result considering that this forecast is issued six months before the start of the ``official" hurricane season and eight months before the active portion of the hurricane season.

  1 June forecasting

The 1 June seasonal tropical cyclone forecast incorporates elements from the 1 December forecast as well as adding in more timely information from the most recent few months (Gray et al. 1994). There are 13 predictors in five groups as listed in Table 2 used in this forecast. Fig. 2 shows the locations of these various predictors. Three again are for extrapolating the state of the QBO expected during September - zonal winds at 50 mb, 30 mb, and the vertical shear between the two layers. Four predictors involve North African surface parameters. Two of these, the Gulf of Guinea and western Sahel rainfall, were described in the previous section. The other two North African predictors are the anomalous surface temperature and sea level pressure gradients from February through May of the current year. The remaining six predictors involve conditions over the Caribbean Sea (April to May sea level pressure anomalies and 200 mb zonal wind anomalies) and more up to date information regarding the current strength and trend of ENSO.

For the QBO predictors, the only change between the predictors used here at the 1 June time frame and the previous year 1 December initial time is that the QBO extrapolation is only a four month lead. Otherwise, the three predictors - 50 mb and 30 mb zonal winds and the vertical shear between the two levels - are identical to the ones used earlier.

For the four north African surface predictors, two the August to November Gulf of Guinea rainfall and the August and September western Sahel rainfall - are identical to the predictors used in the 1 December forecast. The other two predictors relate to the pre-rainy season conditions over sub-Saharan North Africa. We find that, when zonal surface temperature and sea level pressure gradients during February through May are relaxed as the monsoon onsets into the Sahel that the Sahel rainfall and Atlantic hurricane activity are stronger than normal. Conversely, when the surface temperature and sea level pressures have tightened gradients from the west coast to the interior, that the Sahel rainfall is reduced and the Atlantic hurricane activity is quieter than normal. These surface conditions act to alter the strength of the southwesterly monsoon flow into the Sahel. The amount of intense hurricane activity observed after relaxed temperature/pressure gradients over North Africa is 200% more than observed after tightened gradients.

Over the Caribbean, we utilize April and May sea level pressure and 200 mb zonal wind anomalies as predictors for the hurricane season. The sea level pressure anomalies give an indication of how strong and/or how far north the intertropical convergence zone (ITCZ) is setting up for the hurricane season (Gray 1984b). A stronger and/or more poleward position of the ITCZ will increase the likelihood of tropical cyclone formation. As described earlier, 200 mb westerly anomalies tend to increase the vertical wind shear and thus block tropical cyclogenesis and intensification. The 200 mb winds over the Caribbean Sea are also a good measure of the direct influence of the strength of ENSO, as suggested earlier. Both the sea level pressure anomalies and the 200 mb zonal winds over the Caribbean have a tendency to persist into the heart of the hurricane season and thus are useful as predictors of the hurricane activity.

The final four predictors give indications of the current strength of ENSO and its trend in the previous few months. Thus the April and May SSTA and the standardized Tahiti - Darwin sea level pressure anomalies or the Southern Oscillation Index (SOI) values as well as the change of these parameters from January/February to April/May are used as predictors. These values provide good indications about how ENSO will likely behave during the crucial hurricane months of August through October.

With the use of these 13 predictors, we estimate through the hindcast testing that we should have between 50% and 70% of the variability explained by 1 June. This is a substantial improvement over the skill levels that are suggested for our 1 December forecasts. If the atmosphere behaves in future years as it has over the last four decades, then substantial future year forecast skill is available.

  1 August forecasting

For the final forecast initial time at 1 August, we are able to use information that extends right up to the start of the active portion of the hurricane season. Nine predictors in four predictor groupings as shown in Table 2 are used in the current version of the forecast scheme; all but one of which are simply updates of predictors described earlier. Fig. 2 again shows the locations of these various predictors. As before the QBO measures of 50 mb and 30 mb zonal winds and the vertical shear between the two levels through July are extrapolated two months forward to September. One predictor is the identical previous year August through November Gulf of Guinea rainfall. Also as before, the Caribbean Sea sea level pressure anomalies and 200 mb zonal wind anomalies are utilized, but now updated for the months of June and July. In addition, we use the June and July values of SSTA and SOI for a current indication of ENSO's phase and strength. The Caribbean Sea and ENSO predictors are useful as a consequence of their strong tendency to persist through the remainder of the hurricane season.

The one new and very important predictor not in the 1 December or 1 June forecasts is the rainfall anomaly in the western Sahel during June and July. Since the rainy season usually commences during these two months, this rainfall index gives a good idea of the early summer strength of the monsoon in its effect on the Sahel. Typically, the use of June and July rainfall provides an excellent indication of how rainy the remaining two months of August and September of the rainy season will be (Bunting et al. 1975). Again, because of the strong concurrent correlation between Atlantic tropical cyclone activity and seasonal Sahel rainfall, this June and July western Sahel rainfall provides an excellent precursor signal for about to commence hurricane activity, particularly the expected intense hurricane activity to follow in the August through October period. Typically, 230% more intense hurricanes occur in seasons immediately following wet western Sahel June and July years than in dry western Sahel June and July years. Note that these more recent rainfall measurements replace the previous year August and September western Sahel rainfall anomalies.

The skill levels based upon hindcast testing is about 45% up to about 60% of the variability explained by 1 August. While this is a substantial improvement over the hindcast skill available by 1 December, it is somewhat lower than what we believe is possible by 1 June, two months earlier. This is because a reworking of the forecast scheme for the 1 June lead time has just recently been completed - the improvements to which (Gray et al. 1994) actually allowed it to perform better than the older version (Gray et al. 1993a) of the 1 August scheme. We are currently working on altering the 1 August forecast methodology to include the North African surface temperature and sea level pressure gradients as well as the time change of the ENSO parameters. With the addition of these predictors, we should be able to improve the performance of the 1 August forecast beyond that seen for the 1 June forecast. Note that this forecast methodology is predicated on the assumption that the tropical cyclone activity in the future will behave (i.e. have the same characteristics) as what has been observed in the last four decades. Because this may not exactly hold true, some degradation of the skill suggested in the hindcast simulations is to be expected.

  1994 seasonal forecast

The forecast that the second author issued in late November 1993 for the 1994 hurricane season was based upon a somewhat mixed set of indicators (Gray 1993). The QBO which was in transition from a westerly to an easterly phase in November, was extrapolated to be in a strongly easterly phase by September 1994. The extrapolated winds for 50 mb and 30 mb are -20 ms-1 and -24 ms-1, respectively. Thus the QBO wouldfavor a quieter tropical cyclone season.

For the North African rainfall indices, while the western Sahel during August and September 1993 was moderately dry (at -0.48 standardized deviations), the Gulf of Guinea during August through November experienced near normal rainfall amounts (-0.05 standardized deviations). Both of these prior year rainfall indices are the highest amounts observed since 1989. (Note that both of these composite rainfall measurements are slightly revised from those reported in the Gray (1993) report due to updated data since received.) Thus these African rainfall indices suggested only slightly below normal rainfall conditions for the coming 1994 rainy season in the Sahel as well as near normal hurricane activity. However, since the last two decades in the Sahel have been very dry compared to the longer period mean climatology, a return to near normal rainfall conditions would, relatively, indicate an increase in rainfall as well as Atlantic intense hurricane conditions over recent year measurements.

The remaining four predictors used to give an indication of ENSO suggest that there would be a return for the first time in four years to moderately cool SSTA conditions during the 1994 hurricane season. Of all the predictors, ENSO is the most influential. A change over to cool ENSO conditions for 1994 represents a primary change in global weather patterns. Thus the ENSO factors would hint toward a more active hurricane season than has occurred in the last three hurricane seasons.

Overall, the seasonal forecast is expressed in terms of a linear equation:

Forecast tropical cyclone activity = Median conditions + QBO adjustment (suppressing factor in 1994) + African rainfall adjustment (neutral factor in 1994) + ENSO adjustment (enhancing factor in 1994)

The combination of nine predictors in these three groupings gave 1994 seasonal forecast numbers that are average or slightly above average, based on mean conditions of the last 50 years. The late November 1993 forecast for 1994 is for ten named storms (108% of the long-term mean), six hurricanes (105%), two intense hurricanes (92%), 60 named storm days (130%), 25 hurricane days (107%), and seven intense hurricane days (150%). Conditions in the tropics are being carefully monitored during the winter and springs months of 1994 to see the extent to which the late November forecast will need modifying. For example, for the early June forecast, to be issued on 7 June 1994, we will have a better evaluation of ENSO conditions for this coming hurricane season. ENSO has always been notoriously unpredictable, especially during the last few years. (Most ENSO forecasters did not anticipate the continuation of the 1991-1992 warm event through the 1993 season.) Our new extended range ENSO forecast scheme should add future improvement to our 1 December extended-range forecast.

  Outlook for future years

As discussed thoroughly in Gray (1990) and Landsea et al. (1992), the Atlantic tropical cyclones - especially the intense hurricanes - have experienced a strong decrease in numbers since the late 1960s that is linked to the severe two decade long drought that has afflicted Africa's Sahel (Fig. 4). We believe that the season to season links between the Sahel and the Atlantic hurricanes are primarily due to the associated changes in the upper tropospheric circulation which accompany changes in the monsoon structure and secondarily to changes in the strength of the easterly waves produced over North Africa. Sahel drought years are typically associated with strong upper level westerly "shearing" winds over the tropical North Atlantic as well as to a reduction in the percentage of strong easterly waves which propagate out of North Africa during the August to September period. Unfavorable alterations in both the upper tropospheric flow and easterly wave structure may make it difficult for tropical cyclogenesis and intensification.

There is a question as to whether the recent 25-year downturn in Sahel rainfall and in intense hurricane activity is a relatively permanent feature. Some have suggested that human-induced desertification is, to a large extent, responsible for a change of climate in West Africa (Charney 1975; Xue and Shukla 1993). However, an examination of the historical records and pre-historical proxies indicate that multidecadal drought and wet periods are a common occurrence in the Sahel (Nicholson 1989). As mentioned earlier, there is also good evidence that drought and wet periods are forced by global sea surface temperature changes. On a year to year basis, ENSO with its alterations of tropical Pacific sea surface temperature leads to drier Sahel rainfall conditions during warm events and more rain in cold events (Palmer et al. 1992; Ward 1992). Of note is the more frequent occurrence of moderate to strong El Niños (e.g. warm events) during the last two decades, contributing both to decreased Sahel rainfall and lowered Atlantic basin hurricane activity. On a multidecadal time scale, a warmer southern and cooler northern hemispheric sea surface temperatures leads to the long term Sahel drought episodes (Folland et al. 1986, 1991).

It is suggested that both Southern Minus Northern Hemispheric sea surface temperature changes and more frequent El Niños in the last two decades are being forced by alterations of the global oceanic conveyor belt. Recent research has indicated that the oceanic conveyor belt (or thermohaline circulation) is a global transport of water that is associated with the sinking of the surface cold, salty water in the North Atlantic Ocean as shown in Fig. 5. The deep water originating there then flows southward and then eastward through the Indian Ocean depths before upwelling in the Western Pacific Ocean. The water then completes the circuit as a returning surface flow westward in the Indian Ocean and then northward back up to the North Atlantic. The period of ocean particle transport through this large circulation is estimated to be 500 to 2000 years with multidecadal variations in strength (see review by Broecker (1991)). As hypothesized by Street-Perrott and Perrott (1990) and Gray et al. (1992b), the conveyor belt has slowed down during the last two to three decades. This has brought about the south/north dipole of sea surface temperatures by transporting less heat across the equator in the Atlantic as well as to allow more heat to build up in the western Pacific Ocean - thus spawning more frequent El Niños (Fig. 6).

Since this conveyor belt slow down is likely to be a temporary multidecadal fluctuation and not judged to be a permanent feature, it is to be expected that this global oceanic circulation will again speed up in coming years with concomitant increase in Sahel rainfall and Atlantic intense hurricane activity. When this may occur is a matter of speculation. If the past can be used as a guide then these drought conditions should be expected to abate in the next few years. The Sahel has now experienced over a quarter century (1968 to 1993) of nearly continuous drought conditions. With that in mind it is to be expected that rainier years lie ahead in the future. While this may have the positive impact of increasing agricultural production in the countries of the Sahel, ironically it will have a potentially disastrous influence around the Caribbean Sea and along the U.S. East Coast. Intense hurricanes similar to Hurricane Andrew are likely to become more frequent. This is especially ominous when consideration is given to the large increase in United States and Caribbean coastal populations and property buildup since the last active hurricane period of the 1940s to 1960s.


A large portion of this work has been supported by research grants from the U.S. National Science Foundation as well as a U.S. NASA Global Change Fellowship held by the lead author.


Broecker, W. S. (1991) The great ocean conveyor. Oceanography, 4, pp.79-89.

Bunting, A. H., Dennett, M. D., Elston, J. and Milford, J.R. (1975) Seasonal rainfall forecasting in West Africa. Nature, 253, pp.622-623

Charney, J. G. (1975) Dynamics of desert and drought in the Sahel. Quart. J. Roy. Meteor. Soc., 101, pp.183-202

Folland, C. K., Palmer, T. N. and Parker, D. E. (1986) Sahel rainfall and worldwide sea temperatures, 1901-1985. Nature, 320, pp.602-607

Folland, C. K., Owen, J. A., Ward, M. N. and Colman, W. (1991) Prediction of seasonal rainfall in the Sahel region using empirical and dynamical methods. J. Forecasting, 10, pp.21-56 Gray, W. M. (1984a) Atlantic seasonal hurricane frequency: Part I: El Niño and 30 mb quasi-biennial oscillation influences. Mon. Wea. Rev., 112, pp.1649-1668

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

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

Gray, W. M. (1993) Extended range forecast of Atlantic seasonal hurricane activity for 1994. Dept. of Atmospheric Science paper issued on 19 November 1993, Colorado State University, Fort Collins, CO, 80523, U.S., 10 pp.

Gray, W. M., Landsea, C. W., Mielke, Jr., P. W. and Berry, K. J. (1992a) Predicting Atlantic seasonal hurricane activity 6-11 months in advance. Wea. Forecasting, 7, pp.440-455

Gray, W. M., Landsea, C. W., and Sheaffer, J. D. (1992b) Long period variations in African rainfall and hurricane related destruction along the US East Coast. Dept. of Atmospheric Science paper issued on 3 April 1992, Colorado State University, Fort Collins, CO, 80523, U.S., 37 pp.

Gray, W. M., Landsea, C. W., Mielke, Jr., P. W. and Berry, K. J. (1993a) Predicting Atlantic seasonal tropical cyclone activity by 1 August. Wea. Forecasting, 8, pp.73-86

Gray, W. M., Landsea, C. W., Mielke, Jr., P. W. and Berry, K. J. (1994) Predicting Atlantic basin tropical cyclone activity by 1 June. Wea. Forecasting, 9, pp.103-115

Gray, W. M., Sheaffer, J. D., Mielke, Jr., P.W., Berry, K. J. and Knaff, J.A. (1993b) Predicting ENSO 9-14 months in advance. In:Postprints of the 18th Annual Climate Diagnostics Workshop, National Oceanographic and Atmospheric Administration, Boulder, U.S., 4 pp.

Lamb, P. J. (1978) Large-scale tropical Atlantic surface circulation patterns associated with sub-Saharan weather anomalies. Tellus, 30, pp.240-251

Landsea, C. W. (1993) A climatology of intense (or major) Atlantic hurricanes. Mon. Wea. Rev., 121, pp.1703-1713

Landsea, C. W. and Gray, W. M. (1992) The strong association between western Sahel monsoon rainfall and intense Atlantic hurricanes. J. Climate, 5, pp.435-453

Landsea, C. W., Gray, W. M., Mielke, Jr., P. W. and Berry, K. J. (1992) Long-term variations of western Sahel monsoon rainfall and intense Atlantic hurricanes. J. Climate, 5, pp.1528-1534

Nicholson, S. E. (1988) Land surface-atmosphere interaction: Physical processes and surface changes and their impact. Progr. Phys. Geogr.,12, pp.36-65

Palmer, T. N., Brankovic, C., Viterbo, P. and Miller, M. J. (1992) Modeling interannual variations of summer monsoons. J. Climate, 5, pp.399-417.

Nicholson, S. E. (1989) Long-term changes in African rainfall. Weather,44, pp.46-56

Street-Perrott, F. A. and Perrott, R. A. (1990) Abrupt climate fluctuations in the tropics: The influence of the Atlantic circulation. Nature,343, pp.607-612.

Ward, M. N. (1992) Provisionally corrected surface wind data, worldwide ocean-atmosphere surface fields, and Sahelian rainfall variability. J. Climate,5, pp.454-475.

Xue, Y. and Shukla, J. (1993) The influence of land surface properties on Sahel climate. Part I: Desertification. J. Climate, 6, pp.2232-2245