20th CENTURY DROUGHTS IN SOUTHERN AFRICA: SPATIAL ANDTEMPORAL VARIABILITY, TELECONNECTIONS WITH OCEANIC
AND ATMOSPHERIC CONDITIONSYVES RICHARDa,*, NICOLAS FAUCHEREAUa, ISABELLE POCCARDa, MATHIEU ROUAULTb and
SYLWIA TRZASKAa
a Centre de Recherches de Climatologie, Uni�ersite de Bourgogne, Franceb Oceanography, Uni�ersity of Cape Town, South Africa
Recei�ed 14 February 2000Re�ised 21 January 2001
Accepted 26 January 2001
ABSTRACT
Southern African rainfall does not show any trend to desiccation during the 20th century. However, the subcontinentexperienced particularly severe droughts in the 1980s and at the beginning of the 1990s and the magnitude of theinterannual summer rainfall variability shows significant changes. Modifications of the intensity and spatial extensionof droughts is associated with changes in ocean–atmosphere teleconnection patterns.
This paper focuses mostly on the well-documented 1950–1988 period and on late summer season(January–March). A principal component analysis on southern African rainfall highlights modifications of therainfall variability magnitude. The 1970–1988 period had more variable rainfall, and more widespread and intensedroughts than the 1950–1969 period.
To investigate the potential modifications of the associated ocean–atmosphere teleconnection patterns, a compositeanalysis is performed on sea-surface temperature (SST) and National Center for Environmental Protection (NCEP)atmospheric parameters, according to the 5 driest years of both sub-periods. Significant changes are shown inocean–atmosphere anomaly patterns coincident with droughts for both sub-periods. The 1950–1969 droughts wereassociated with regional ocean–atmosphere anomalies, mainly over the southwest Indian Ocean region. In contrast,during the 1970–1988 droughts near-global anomalies were observed in the tropical zone, corresponding to ElNino–Southern Oscillation (ENSO) phenomenon.
With the exception of southwest Africa and some southern African coastal areas, austral summer (i.e.October–March) is the main rainfall season over much of southern Africa. Moreover, the late summerseason (January–March) often represents more than 40% of the annual amount. The southern Africanrainfall time series does not seem to exhibit any trend to desiccation or abrupt shifts during the 20thcentury (Tyson, 1986; Hulme, 1992, 1996) except over some limited areas like the Lowveld of SouthAfrica (Mason, 1996). However, southern Africa summer rainfall experiences a high degree of interannualvariability, with recurrent wet and dry spells. In this way, southern Africa was struck by particularlysevere droughts in the 1980s and summers at the beginning of the 1990s (Harsch, 1992), which led todecrease in crop and stock production (Vogel, 1994).
* Correspondence to: Centre de Recherches de Climatologie, Universite de Bourgogne, UMR CNRS 5080, France.
Southern African summer rainfall variability is known to be linked to the El Nino–SouthernOscillation (ENSO) phenomenon (Mo and White, 1985; Lindesay, 1988; Ropelewski and Halpert, 1987,1989, 1996; Shinoda and Kawamura, 1996; Rocha and Simmonds, 1997). During warm ENSO events, dryconditions generally occurred over much of southern Africa. This influence is strongest in thesoutheastern part of the subcontinent and to the northeast of South Africa (Matarira, 1990; Richard,1996) and maximum during the late summer season of January–March (Lindesay, 1988; Lindesay andVogel, 1990; Poccard, 2000). However, warm ENSO events do not always lead to dry conditions oversouthern Africa (Mason and Mimmack, 1992). A different hypothesis has been suggested: thestratospheric Quasi-Biennial Oscillation (QBO) may play a significant role in the modulation of the ENSOsignal over South Africa (Mason and Tyson, 1992; Mason and Lindesay, 1993; Jury et al., 1994). In SouthAfrica, Kruger (1999) has also shown that the low-frequency rainfall variability (18–20 yearQuasi-Periodic Oscillation, see Tyson et al., 1975) could modulate the impact of ENSO events in such away that, during the ‘wet phases’ of this ‘oscillation’, a moderating effect is evident on the severity ofwarm ENSO events impacts, so that even above-average rains can be experienced during such events.
Apart from the global-scale ENSO influence, many authors have pointed out that interannual rainfallvariability over large parts of central and eastern South Africa is linked to sea-surface temperature (SST)anomalies over surrounding Indian and South Atlantic oceans (Walker, 1989, 1990; Richard, 1993; Rochaand Simmonds, 1997; Reason and Lutjeharms, 1998; Reason, 1999; Reason and Mulenga, 1999). In thisway, warm SST anomalies in the central equatorial Indian Ocean are frequently associated with dryconditions over southern Africa. At a more regional scale, the influences of the Agulhas current systemon the southern African rainfall have been investigated. Reason and Mulenga (1999) point out that theSST of Mozambique Channel and immediately east of South Africa show strong links with South Africansummer rainfall. In the southwest Indian Ocean, which corresponds to the Agulhas recirculation region,interannual events of warming and cooling, whose magnitude is generally at least comparable to thoseassociated with strong ENSO events, are linked to interannual rainfall variability (Reason, 1999).Atmospheric modelling experiments support this link between SST anomalies and southern Africanrainfall (Reason, 1998).
In the above-cited papers, the teleconnections with SST, both at global or regional scales, are generallyconsidered with the underlying hypothesis of their stability through time. However, noticeable changeshave been detected in the world-wide SST patterns at a global scale: a Pacific warming since themid-1970s (consistent with the intensification of the ENSO warm events), an anomaly in the meridionalgradient in tropical Atlantic during the post-1970 period and a long-term warming trend in the SouthernHemisphere extra-tropics (Fontaine et al., 1998). The Indian Ocean warming after 1970 (Trzaska et al.,1996) caused the strengthening of the relationship between the central Indian and central Pacific oceansSST during the more recent period (Lanzante, 1996). Wang and Ropelewski (1995) also found that theENSO-scale variability could be higher (both in frequency and amplitude) when the climate low frequency‘basic state’ is relatively warm. These large-scale changes in SST patterns could be responsible for thehigher Southern Oscillation Index (SOI)–southern African rainfall relationship that was detected inobserved datasets and from results of general circulation model (GCM) simulations (Moron and Ward,1998; Richard et al., 2000).
This paper focuses on the January–March summer season, considered the rainiest season as well as themost directly related to the tropical circulation (D’abreton and Lindesay, 1993). Rainfall variability andespecially drought variability during the 20th century is studied. The hypothesis is that, concurrently withglobal scale modifications of SST anomaly patterns, southern African rainfall features, and especiallydroughts, show significant changes.
This paper is organized as follow: data are presented in Section 2. Section 3 is divided into foursub-sections: the first investigates January–March southern African rainfall variability during the 20thcentury. Drought features variability is developed in the second section. The oceanic and atmosphericconditions associated with southern African droughts are explored in the third section. The statisticalrelationships between January–March rainfall variability and both SOI and southwest Indian Ocean SSTindexes are presented here too. Section 4 provides a discussion of the results, followed by Section 5, whichconcludes the paper.
The ‘Centre Recherches de Climatologie’ (CRC) dataset is an original monthly rainfall datasetwhich has been compiled from more than 570 rain gauges. The rainfall data has been tested andverified on the 1950–1988 period, and it covers the whole of the African continent south of theSahara (Bigot et al., 1994, 1995). This dataset, with a good spatial density of observation networkover southern Africa, has been interpolated onto a regular 1°×1° grid in order to obtain a relativehomogeneous density network (Figure 1). In order to extend the study to the beginning of thecentury, a synthetic rainfall index, named southern African Rainfall Index (SARI) hereafter, isconstructed by extracting 28 grid-points from the Climate Research Unit (CRU) dataset (Hulme, 1992,1996) over the 1900–1998 period. These points correspond to a rainfall covariance area definedpreviously (Richard et al., 2000). In this study, the CRC dataset is used in order to capture the spatialbehaviour of rainfall, thanks to its good spatial resolution, whereas SARI is used to show the timevariations of rainfall at a century scale.
2.2. The ENSO index
The Southern Oscillation Index (SOI) is used rather than the Nino3’s SST index because it has beenconsidered more reliable over almost a 100-year period (1900–1998). Atmospheric pressure at Tahitiand Darwin has not changed through the century, which is not the case for the SSTs (Ropelewski andJones, 1987). In this study, SOI values are averaged from July (year −1) to March (year 0), in orderto reduce intermonthly variability.
2.3. Sea-surface temperature
The UK Meteorological Office Sea-Surface Temperature (MOHSST) dataset is utilized in the study.The dataset (Bottomley et al., 1990) is available on a 5°×5° grid and quasi-global coverage (from60°N to 40°S) over the 1946–1998 period.
Figure 1. Rain gauge network and late summer (January–March) mean rainfall in mm (1950–1988). Thick contour: grid-pointsselected to determine the SARI (1990–1998)
The National Center for Environmental Prediction (NCEP) and the National Center for AtmosphericResearch (NCAR) have completed a reanalysis project to produce a 50-year record of global atmosphericanalyses with fixed data assimilation (Kalnay et al., 1996). In this paper, monthly wind, pressure, airtemperature and upward longwave radiation flux have been selected at 850 and 200 hPa over the1950–1988 period. It is important to note that a recent study (Poccard et al., 2000) has shown thatdifferent abrupt shifts occur in some NCEP parameters (e.g. in 1967/1968), mainly due to assimilation ofobservations in the NCEP model and problems concerning land surface processes parameterization. Thisproblem must be kept in mind when interpreting the results.
3. RESULTS
3.1. Rainfall �ariability during the 20th century
Twenty year running means and running standard deviations (with a 1-year lag) are computed from thestandardized SARI time series over the 1900–1998 period. The January–March rainfall during the 20thcentury does not show any long-term trend of desiccation nor abrupt shift over the southern Africanregion (Figure 2). However, since the beginning of the 1980s, the annual rainfall amount has slightlydecreased.
Moreover, the January–March interannual rainfall variability seems to have experienced significantchanges, leading to three major sub-periods if the interannual variability magnitude is considered. Thestrongest rainfall anomalies (more than 2 S.D.) are found at the beginning of the century (e.g. severedrought of 1922) and during recent decades. Anomalies are weaker (not more than 1 S.D.) between 1935and the middle of the 1960s.
A principal component analysis (PCA) is performed over the 1950–1988 period on the high spatialresolution CRC dataset. The first eigenvector (PC1, not shown, see Richard et al., 2000) extracts 31.9%of the total variance and the others describe less than 10%. This first PC is highly correlated with SARI(r=0.97) over the common period. Both CRC PC1 and SARI highlight modifications of the rainfallvariability magnitude, with higher anomalies during the recent period. This could be due to changes in themagnitude of rainfall anomalies and/or changes in the spatial extension of these anomalies. To addressthese hypotheses more satisfactorily, especially the second one, the CRC dataset is chosen because of itshigher spatial resolution (1°×1°). The goal is to check if the intensity and the spatial extension ofdroughts are the same for both sub-periods.
Figure 2. January–March SARI (1900–1998). Bars: January–March standardized rainfall. Central curve: 20-year running mean.Curves: 20-year running S.D. Grey: S.D. envelope (area between −1 S.D. and +1 S.D.)
3.2. Drought features associated with low and high January–March rainfall �ariability periods(1950–1969 and 1970–1988)
The CRC dataset only allows the consideration of the 1950–1988 period. According to the above results,there are two different sub-periods, if the magnitude of the interannual variability is considered. The1950–1969 period has relatively low variability, and 1970–1988 has higher variability. It must be notedthat the year 1970 is only indicative and is used to separate two samples for the composite analysis. Bothsub-periods are chosen to form two January–March composites selected from the driest quartile of the PC1scores. During 1950–1969, the composite sample includes 1951, 1960, 1964, 1965 and 1968. During1970–1988, it includes 1970, 1973, 1982, 1983 and 1987. The Student’s t-test allows the comparison of the15 or 14 other years of both dry composites to the mean of the two sub-periods (1950–1969 and 1970–1988).
During the low variability period (1950–1969), the spatial extent of droughts affecting southern Africais not homogeneous (Figure 3(A)). Only 25 of the 104 grid-points have standard deviation anomalies lessthan −1. They are found mainly over Zimbabwe. The rainfall deficits are less significant in Zambia andthe Republic of South Africa (RSA) (except Free State). Tanzania, which is integrated in a different rainfallcovariability region (Richard et al., 1998), does not experience these droughts.
Between 1970 and 1988, droughts are more extensive with 48 of the 104 grid-points showing a standarddeviation less than −1 (Figure 3(B)). The droughts spread northwards compared to the former sub-period.They affect northern Zambia and Mozambique more significantly. Droughts are also more intense inNamibia and South Africa.
3.3. Specific oceanic and atmospheric conditions associated with January–March droughts (1950–1969and 1970–1988)
The SST and atmospheric parameter anomalies for both dry sub-periods are presented and analysed inthis section. The statistical significance is tested with a Student’s t-test, except for the wind for which theHottelling test is used.
Figure 4(A) shows a SST anomaly composite for the low variability sub-period (1950–1969). Significantanomalies are not widespread. Nevertheless, negative anomalies are observed in the subtropical southwestIndian Ocean (Mozambique Channel, Agulhas Current and Agulhas retroflection) and in the southwestAtlantic Ocean.
Previous studies (Walker, 1989; Reason and Mulenga, 1999) have depicted a positive correlation betweensummer rainfall and regional SST. Particularly, Walker describes a mechanism associated with local oceanic‘warm events’, which are preceded by easterly wind anomalies across the southwest Indian Ocean oversource regions of the Agulhas Current. These ‘warm events’ correspond to higher rainfall. Is this anobservation of the inverse mechanisms during the ‘cold events’ associated with the 1950–1969 droughts?
The January–March 1950–1969, 850-hPa NCEP temperature composite (not shown) is consistent withthe SST composite. Over this sub-period, the 200-hPa geopotential composite (Figure 4(B)) shows markednegative anomalies above the abnormally cold water and air masses of the western subtropical Indian andAtlantic oceans. These low features indicate a northward displacement of the jet stream. The westerlywinds are significantly abnormally strong near 20°–25°S over Argentina and South Africa (Figure 4(C)).On the other hand, the westerly winds are abnormally weak near 35°–40°S. The result for southern Africacould be a reduction of deep convection that would be displaced to the northeast above the Indian Ocean(Figure 4(D)). This outgoing longwave radiation (OLR) dipole pattern between southern Africansubcontinent and Indian Ocean is consistent with the results of Jury (1992, 1996), Jury et al. (1992) andJury and Pathack (1993). They highlighted that summer convection variability over southern Africa canbe monitored by spatial changes in OLR, and that decreased convection over the subcontinent during dryyears is often compensated for by increased convection to the east of Madagascar.
Unlike the former two decades, the 1970s and 1980s droughts are associated with warm and morewidespread SST anomalies (Figure 5(A)). Moreover, the concerned sectors are not located in thesubtropical latitudes, but are more centred on the equator. These warm anomalies correspond to thoseobserved during ENSO events (Nicholson, 1997). Four droughts of the 1970–1988 period correspond tothe end of ENSO events (1970, 1973, 1983 and 1987). During the first sub-period, none of the driest yearscorresponds to ENSO. Only the January–March 1982 drought cannot be considered to be associated withENSO that only developed in spring of that year.
The 850-hPa NCEP air temperature composites also show positive anomalies above low latitudes of theEast Pacific, west of South America and east equatorial Atlantic Ocean (not shown). Strong abnormalhigh temperatures are also observed over most of the southern African subcontinent. Abnormal highfeatures at 200 hPa are found in the Tropics (Figure 5(B)) with substantial wind anomalies (Figure 5(C)).Tyson (1984), Matarira and Jury (1992), Shinoda and Kawamura (1996) have shown that highergeopotential heights over the subcontinent indicate weaker subtropical troughs during dry years. Toddand Washington (1999) also found enhanced 200-hPa geopotential heights in the tropical zone during dryJanuarys. In addition, the upper-level easterly wind anomalies above the Central and West Pacific Oceanand westerly anomalies above the equatorial Atlantic typical of ENSO, westerly anomalies aboveSouthern Africa and the nearby Atlantic and Indian oceans are observed (Figure 5(C)). They aresignificant south of South Africa and over the Mozambique Channel and are accompanied by markednegative OLR anomalies (Figure 5(D)). Positive anomalies of OLR mean a deficit in deep convectionabove the subcontinent and a decrease of rainfall above southern Africa.
3.4. Relationship between SARI, southwest Indian Ocean SSTs and SOI o�er the whole century
Two distinct oceanic regions have been highlighted by previous composite analysis. Droughts of the1950–1969 period—marked by a relatively low interannual variability—are associated with significantnegative SST anomalies, mainly in the southwest Indian Ocean. In contrast, droughts of the more recentperiod (1970–1988)—marked by a high January–March rainfall variability—are associated withsignificantly positive SST anomalies over the Pacific region, in a pattern that corresponds to the ENSOsignal. The aim of this section is to investigate the variations within the whole century of the statisticalrelationship between southern African January–March rainfall (through the SARI index), the SOI, andthe SSTs of the southwest Indian Ocean.
Figure 6 shows the variation of SOI values from 1900 to 1998. The 20-year running mean and standarddeviation envelopes are superimposed. SOI (Figure 6) and SARI (Figure 2) interannual variability seemto follow the same behaviour: strong variability from the beginning of the century to 1930, then weak upto the end of the 1960s and strong again from the 1970s. Allan et al. (1995) have already discussed thisSOI decadal variability. Based on these observations, we have divided the 20th century into three parts(1900–1933, 1934–1969 and 1970–1998) to study the relationship variability between SARI and SOI.Correlations between SARI and SOI are calculated over the 1900–1998 period and over the three formersub-periods. For the whole period, the correlation value is 0.44 (Table I), a correlation level similar to
Figure 6. Standardized pressure difference between Tahiti and Darwin (SOI) (1900–1999). Central curve: 20-year running mean.Curves: 20-year running S.D. Grey: S.D. envelope (area between −1 S.D. and +1 S.D.)
Table I. Correlation between SARI and SOI
Period 1900–1998 1900–1933 1934–1969 1970–1998
R 0.44* 0.47* 0.15 0.56*
* Correlations that are significant at the 95% confidence level.
Lindesay’s (1988) result. The correlation is 0.47 for the first sub-period, 0.15 for the second and 0.56 forthe last. This confirms the temporal instability of the statistical relationship between SOI and southernAfrican rainfall variability, and is in accordance with the results of Moron and Ward (1998).
Globally, there is a correspondence between periods of strong variability and strong teleconnections.This seems to confirm that southern African rainfall anomalies are only sensitive to ENSO, when itexhibits a strong interannual variability. Nevertheless, caution must be addressed because, even in astrong variability period, both 1925–1926 and 1997–1998 strong ENSOs were not associated withdroughts.
The SST index is constructed for the southwest Indian Ocean (referenced as SWIOSST). This areacorresponds to the negative anomaly area for 1950–1969 (Figure 4(A)). January–March SST is averagedbetween 20° and 30°S and between 30° and 60°E. This index can only be considered to be reliable for thesecond part of the century. There is a decrease in observations during World War II and 1945 has beendiscarded from the sample. Figure 7 shows the evolution of the standardized index SWIOSST for1946–1998. SWIOSST’s 20-year running mean and standard deviation envelopes are superimposed. Thereis a marked increase of the mean values, consistent with the temperature increase in the subtropical andextra-tropical latitudes of the Southern Hemisphere (Fontaine et al., 1998). Thus, there is a lack of coldevent in that area from 1965.
As for SOI, correlation coefficients between southern African rainfall (SARI) and SWIOSST arecalculated over the 1946–1998 period and for the 1946–1969 and 1970–1998 sub-periods (Table II). Forthe global period, the correlation is significant at the 95% confidence level but the value is weak becauseSARI is not characterized by a trend, contrary to SWIOSST. However, the correlation for bothsub-periods and the evolution of those correlations underline several points (Table II). For the 1946–1969sub-period, the association between southern African droughts and cold southwest Indian Ocean SST isalmost systematic, especially for intense droughts (not showed). This confirms the results from the1950–1969 composites (see Section 4). After 1970, the association between droughts and cold SST doesnot occur (not shown), and the correlation coefficient drops (Table II). A possible reason is a decrease inthe amplitude of the cold events due to an increase of mean Indian Ocean SST (Figure 7). Anotherpossibility is that stronger southern African droughts associated with ENSO events mask the relativelydry periods of January–March 1984 and 1994 which occurred in association with cold SWIOSST.
Figure 7. Standardized January–March SWIOSST (1946–1998). Central curve: 20-year running mean. Curves: 20-year runningstandard deviation. Grey: S.D. envelope (area between −1 S.D. and +1 S.D.)
Table II. Correlation between SOI and SWIOSST
Period 1946–1998 1946–1969 1970–1998
R 0.30* 0.49* 0.32
* Correlations that are significant at the 95% confidence level.
4. DISCUSSION
Three points in the above results need to be highlighted and discussed:
(i) There is a similarity between the spatial scale of southern African droughts and theocean-atmosphere mechanisms during both sub-periods. During 1970–1988, the droughts are moreintense and more widespread, and the associated oceanic and atmospheric abnormal conditions aremore global and coherent. In contrast, the 1950–1969 droughts are less extensive and less coherentand they are only connected to regional oceanic and atmospheric anomalies (southwest IndianOcean).
(ii) In spite of the large differences between the anomaly patterns during the droughts of the bothsub-periods, two major similarities are observed:– The enhanced meridional temperature gradient for the subcontinent and the nearby Indian Ocean.
In both sub-periods, the thermal gradient increases between the temperate/subtropical latitudesand the low latitudes compared to normal years. Before 1970, this increase is linked to abnormallycold temperatures south of 20°S; after 1970 to abnormally warm temperatures north of 20°S.
– The atmospheric circulation, mainly noticeable at 200 hPa. In both sub-periods, upper levelanomalies suggest a stronger and further north jet stream than usual. The scheme can thus bedrawn as follow: a strengthening and a northward displacement of the jet stream could lead to adisplacement of cyclonic systems tracks further north, a reduction of the tropical influences froma decrease of easterly wind component, less tropical temperate troughs and then less rains over thesummer rainfall area.
(iii) The correspondence between periods of:– high SOI interannual variability;– high southern African rainfall variability;– intense and extended southern African droughts; and– strong and significant southern African rainfall/SOI correlations.
The higher frequency and/or intensity of ENSO events could be a consequence of the troposphericwarming in the Tropics observed since 1960 (Flohn and Kapala, 1989). Nevertheless, during the period1910–1930, high fluctuations of SOI occurred, although slightly lower to the fluctuations experienced inthe 1970s and 1980s (Allan et al., 1995). The similarity between those two periods lies more in the increasein temperature than in the level of temperature observed (WMO, 1999). These two periods haveexperienced an increase in SST and they correspond to high interannual fluctuations of SOI. During thetransition period (1934–1969), both SST and SARI variations are weaker. This could lead to thehypothesis that the strong ENSO events occurring during periods of SOI’s enhanced interannualvariability induce the most severe southern African droughts. Nevertheless, even during enhancedvariability and correlation periods, strong ENSO events were not systematically accompanied by severedroughts (e.g. the strong 1925–1926 ENSO was associated with slightly higher than normal rainfall). Alsoin the recent period, when the teleconnection is the highest, the strong 1997–1998 ENSO was notassociated with drought. Several studies have shown that the Indian Ocean plays an important role in theassociation between ENSO and southern African rainfall. The enhanced association of the 1970–1988period has thus been related to the warming Indian Ocean (Richard et al., 2000). But internal dynamicsof the Indian Ocean-atmosphere system during the 1997–1998 event (Webster et al., 1999) is also likelyto be responsible for the lack of the association observed in 1997–1998.
5. CONCLUSION
The aim of this paper was to study the rainfall variability and changes in interannual variability amplitudeof 20th century droughts in southern Africa. An analysis of the oceanic and atmospheric conditions linkedto these changes was conducted.
There are no significant changes in the January–March rainfall totals during the last century. However,summer rainfall shows a change in the intensity of interannual variability. Furthermore, the spatialextension of droughts shows three distinct successive periods. The 1900–1933 and 1970–1998 periodshave strong amplitude in their interannual variability, while the 1934–1969 period has a weak amplitude.The two well documented 1950–1969 and 1970–1988 sub-periods were the focus of our attention. Thesecond (1970–1988), characterized by strong interannual rainfall variability, experienced more intense andmore widespread droughts than the former (1950–1969). The SST and atmospheric parameter compositesshow that droughts before and after 1970 are not associated with the same anomaly patterns. The1950–1969 droughts are linked to regional oceanic and atmospheric anomalies. The 1970–1988 droughtsare associated with global tropical oceanic and atmospheric conditions mainly linked to ENSO. Eventhough the anomaly patterns associated with southern African droughts are different between bothsub-periods, there are some similar features which are noted: the enhanced surface thermal meridionalgradient to the south and southeast of the subcontinent and the strengthening and northwarddisplacement of the jet stream over southern African. Further investigations are needed to understandbetter the physical mechanisms responsible for the droughts, as well as a better knowledge of why, in therecent period, some ENSO events (e.g. in 1997–1998) are not accompanied by droughts over SouthernAfrica.
ACKNOWLEDGEMENTS
The authors wish to acknowledge Dr Mike Hulme (University of East Anglia) for providing the CRUrainfall dataset and Dr Henry Mulenga for his interesting comments. The authors also thank theNCEP/NCAR for making the reanalyses datasets available. The SOI is available on the NOAA web site.This work is integrated into the French–South Africa Program ‘Variability of the Summer South AfricanRainfall’, in which the University of Cape Town, the Centre de Recherches de Climatologie and theLaboratoire de Meteorologie Dynamique are involved. The authors thank the South African WaterResearch Commission for their support.
Allan RJ, Lindsay JA, Reason CJC. 1995. Multidecadal variability in the climate system over the Indian Ocean region during theAustral Summer. Journal of Climate 5: 1853–1873.
Bigot S, Roucou P, Moron V. 1994. Amelioration des donnees de precipitation en zone tropicale en vue de la validation des modelesatmospheriques de circulation generale: exemple de l’Amerique du sud et de l’Afrique. Publications de l’Association Internationalede Climatologie 7: 421–431.
Bigot S, Camberlin P, Moron V, Richard Y, Roucou P. 1995. Modes of rainfall variability in tropical Africa and their stabilitytrough times. In 21st Conference of the American Meteorological Society ‘Hurricanes and Tropical Meteorology ’, AmericanMeteorological Society: Boston, MA; 448–449.
Bottomley M, Folland CK, Hsiung J, Nevell RE, Parkers DE. 1990. Global Ocean Surface Temperature: Atlas GOSTA.Meteorological Office: Bracknell, UK.
D’abreton PC, Lindesay JA. 1993. Water vapour transport over Southern Africa during wet and dry early and late summer months.International Journal of Climatology 13: 151–170.
Fontaine B, Janicot S, Moron V, Roucou P, Trzsaka S. 1998. Anomalies de temperature de surface de la mer et precipitationstropicales: synthese de quelques travaux recents portant sur les precipitations au Sahel et dans le Nordeste. La Meteorologie 23:14–35.
Flohn H, Kapala A. 1989. Changes of tropical sea-air interaction processes over a 30-year period. Nature 338: 244–246.Harsch E. 1992. Drought devastates Southern Africa. Drought Network News 2: 17–19.Hulme M. 1992. Rainfall changes in Africa: 1931–1960 to 1961–1990. International Journal of Climatology 12: 685–699.Hulme M. 1996. Recent climatic changes in the World’s drylands. Geophysical Research Letters 23: 61–64.Jury MR. 1992. A climatic dipole governing the interannual variability of convection over the SW Indian Ocean and the SE Africa
region. Trends in Geophysical Research 1: 165–172.Jury MR. 1996. Regional teleconnection patterns associated with summer rainfall over South Africa, Namibia and Zimbabwe.
International Journal of Climatology 16: 135–153.Jury MR, Mac Queen C, Levey K. 1994. SOI and QBO signals in the African Region. Theoretical and Applied Climatology 50:
103–115.Jury MR, Pathack B. 1993. Composite climatic patterns associated with extreme modes of summer rainfall over southern Africa:
1975–1984. Theoretical and Applied Climatology 47: 137–145.Jury MR, Pathack B, Sohn BJ. 1992. Spatial structure and interannual variability of summer convection over southern Africa and
the SW Indian Ocean. South African Journal of Science 88: 275–280.Kruger AC. 1999. The influence of the decadal-scale variability of summer rainfall on the impact of El Nino and La Nina events
in South Africa. International Journal of Climatology 19: 59–68.Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M,
Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D. 1996. TheNCEP/NCAR 40-year Reanalysis Project. Bulletin of the American Meteorological Society 77: 437–471.
Lanzante JR. 1996. Lag relationships involving tropical sea surface temperatures. Journal of Climate 9: 2568–2578.Lindesay JA. 1988. South African rainfall, the Southern Oscillation and a Southern Hemisphere semi-annual cycle. Journal of
Climatology 8: 17–30.Lindesay JA, Vogel CH. 1990. Historical evidence for Southern Oscillation–Southern African rainfall relationships. International
Journal of Climatology 10: 679–689.Mason SJ. 1996. Climatic change over the Lowveld of south Africa. Climate change 32: 35–54.Mason SJ, Lindesay JA. 1993. A note of the modulation of Southern Oscillation–Southern African rainfall oscillations and the
Quasi-Biennial Oscillation. Journal of Geophysical Research 98: 8847–8850.Mason SJ, Mimmack GM. 1992. The use of bootstrap correlation coefficients for the correlation coefficient in climatology.
Theoretical and Applied Climatology 45: 229–233.Mason SJ, Tyson PD. 1992. The modulation of sea-surface temperature and rainfall associations over Southern Africa with solar
activity and the Quasi-Biennial Oscillation. Journal of Geophysical Research 97: 5487–5856.Matarira CH. 1990. Drought over Zimbabwe in a regional and global context. International Journal of Climatology 10: 609–625.Matarira CH, Jury M. 1992. Contrasting meteorological structures of intra-seasonal wet and dry spells in Zimbabwe. International
Journal of Climatology 12: 165–176.Mo KC, White GH. 1985. Teleconnections in the Southern Hemisphere. Monthly Weather Re�iew 113: 22–37.Moron V, Ward MN. 1998. ENSO teleconnections with climate variability in the European and African sectors. Weather 9:
287–295.Nicholson SE. 1997. An analysis of the ENSO signal in the tropical Atlantic and western Indian oceans. International Journal of
Climatology 17: 345–375.Poccard I. 2000. Djon,Universite� de Bourrgogne Etude diagnostique de nouvelles donnees climatiques: les reanalyses. Exemples
d’application aux precipitations. PhD thesis. Dijon, Universite de Bourgogne.Poccard I, Janicot S, Camberlin P. 2000. Comparison of Rainfall Structures between NCEP/NCAR Reanalyses and observed data
over tropical Africa. Climate Dynamics 16: 897–915.Reason C. 1998. Warm and cold events in the Southeast Atlantic/Southwest Indian Ocean region potential impacts on circulation
and rainfall over southern Africa. Meteorology and Atmospheric Physics. 69: 49–65.Reason C. 1999. Interannual warm and cool events in the subtropical/mid-latitude South Indian region. Geophysical Research
Letters 26: 215–218.Reason C, Lutjeharms J. 1998. Variability of the South Indian Ocean and implications for southern African rainfall. South African
Journal of Science 94: 115–123.Reason CJC, Mulenga H. 1999. Relationships between South African rainfall and SST anomalies in the South West Indian Ocean.
International Journal of Climatology 19: 1651–1673.
Richard Y. 1993. Relations entre la variabilite pluviometrique en Afrique australe tropicale et la circulation oceano-atmospherique.PhD thesis, Aix-en-Provence.
Richard Y. 1996. La question de la linearite des relations en climatologie diagnostique: l’exemple de la sensibilite des pluiesstationnelles d’Afrique australe a l’oscillation australe. Mediterranee 4: 87–90.
Richard Y, Camberlin P, Beltrando G. 1998. Recherche de structures spatio-temporelles en climatologie: l’exemple de la variabilitepluviometrique en Afrique orientale. L’Espace Geographique 1: 31–40.
Richard Y, Trzaska S, Roucou P, Rouault M. 2000. Modification of the southern African rainfall variability/ENSO relationshipsince the late 1960’s. Climate Dynamics 16: 883–895.
Rocha A, Simmonds I. 1997. Interannual variability of south-eastern African summer rainfall. Part 1: Relationships with air-seainteraction processes. International Journal of Climatology 17: 235–265.
Ropelewski CF, Halpert MS. 1987. Global and regional scale precipitation and temperature patterns associated with ElNino/Southern Oscillation. Monthly Weather Re�iew 115: 1606–1626.
Ropelewski CF, Halpert MS. 1989. Precipitation patterns associated with the high index phase of the Southern Oscillation. Journalof Climate 2: 268–284.
Ropelewski CF, Halpert MS. 1996. Quantifying Southern Oscillation–Precipitation relationships. Journal of Climate 9: 1043–1059.Ropelewski CF, Jones PD. 1987. An extension of the Tahiti–Darwin Southern Oscillation Index. Monthly Weather Re�iew 115:
1606–1626.Shinoda M, Kawamura R. 1996. Relationships between rainfall over semi-arid Southern Africa, geopotential heights and sea-surface
temperatures. Journal of the Meteorological Society of Japan 74: 21–36.Todd M, Washington R. 1999. Circulation anomalies associated with tropical-temperate troughs in southern Africa and the south
west Indian Ocean. Climate Dynamics 15: 937–951.Trzaska S, Moron V, Fontaine B. 1996. Global atmospheric response to specific linear combinations of the main SST modes. Part
I: Numerical experiments and preliminary results. Annales Geophysicaes 14: 1066–1077.Tyson PD. 1984. The atmospheric modulation of extended wet and dry spells over South Africa, 1958–1978. International Journal
of Climatology 4: 621–635.Tyson PD. 1986. Climatic Change and Variability in Southern Africa, Cape Town. Oxford University Press: Oxford.Tyson PD, Dyer TGJ, Mametse MN. 1975. Secular changes in South African rainfall: 1880 to 1972. Quarterly Journal of the Royal
Meteorological Society 76: 340–341.Vogel CH. 1994. (Mis)management of droughts in South Africa: Past, present, future. South African Journal of Science 90: 4–5.Walker ND. 1989. Sea-surface temperature–rainfall relationships and associated ocean-atmosphere coupling mechanisms in the
Southern Africa region. C.S.I.R. Research Report 683, Stellenbosch, South Africa.Walker ND. 1990. Links between South African summer rainfall and temperature variability of the Agulhas and Benguela system.
Journal of Geophysical Research 95: 3297–3319.Wang XL, Ropelewski CF. 1995. An assessment of ENSO-scale secular variability. Journal of Climate 8: 1594–1599.Webster PJ, Moore AM, Loschnigg JP, Leben RR. 1999. Coupled ocean-atmosphere dynamics in the Indian Ocean during 1997–98.
Nature 401: 356–360.WMO. 1999. WMO statement on the status of the global climate in 1998. WMO 596: 12.
