INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 5: (5) Published online in Wiley InterScience ( DOI:./joc.93 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA AMIN T. MAPANDE a,b * and C. J. C. REASON a a Oceanography Department, University of Cape Town, Rondebosch 77, South Africa b Tanzania Meteorological Agency, PO Box 356, Dar es Salaam, Tanzania Received 6 January 4 Revised March 5 Accepted 6 March 5 ABSTRACT The evolution of the circulation anomalies associated with wet and dry austral summers (October April) over western Tanzania ( 4.75 S, 9 37 E) are investigated. It is found that wet (dry) years are characterized by weaker (stronger) equatorial westerlies over the western Indian Ocean that lead to less (more) export of equatorial moisture away from East Africa. These anomalies arise from an anticyclonic (cyclonic) anomaly over the tropical western Indian Ocean during wet (dry) austral summers. In addition, enhanced (reduced) westerly moisture flux from the southern Congo basin occurs during the wet (dry) seasons. Large-scale modulation of the Indian Ocean Walker cell is also evident in both cases, but particularly for the dry years. During wet seasons, the rains typically begin earlier and end later, tend to be more evenly distributed and end more gradually. For dry seasons, the reverse tends to be the case. An inverse relationship between anomalies in Niño 3.4 seasurface temperature and those in dry-spell frequency over the region was found. Copyright 5 Royal Meteorological Society. KEY WORDS: interannual rainfall; Tanzania; regional circulation; anomalies; dry spell frequency. INTRODUCTION Tanzania has a large rural population dependent on rain-fed agriculture and, therefore, is vulnerable to the impacts of climate variability. The northern part of the country tends to receive rain during the October December (OND; short rains) and March May (long rains) seasons, similar to Kenya, whereas the southern and western regions are more like most of southern Africa in experiencing a single rainy season during austral summer. In this study, we consider the western Tanzanian region (4.75 S, 9 37 E), which has not been widely studied hitherto, and which exhibits a single wet season from October to April. Previous work by Ogallo (988, 989), Hastenrath et al. (993), Kabanda and Jury (999), Indeje et al. () and Reason et al. (), amongst others, has found evidence that the El Niño southern oscillation (ENSO) phenomenon influences Tanzanian climate. However, the western Tanzanian region of interest here tends to lie within the transition region between the areas of strong ENSO impact in which above (below) average rainfall tends to occur over East Africa (southern Africa) during an El Niño event and the reverse during La Niña. In addition to ENSO, other large-scale modes manifest in the neighbouring oceans may be important. In the Indian Ocean, the Indian Ocean dipole (IOD; Saji et al., 999) or Indian Ocean zonal mode (Webster et al., 999) is known to influence East African OND rainfall strongly. For the late summer (January April), evidence has been presented that dipole-like sea-surface temperature (SST) patterns over the south Indian Ocean (Behera and Yamagata, ; Reason,, ) and Benguela Niños in the tropical southeast Atlantic (Hirst and Hastenrath, 983; Rouault et al., 3) influence rainfall over mainly the central and * Correspondence to: C. J. C. Reason, Oceanography Department, University of Cape Town, Rondebosch 77, South Africa; cjrdegs.uct.ac.za Copyright 5 Royal Meteorological Society

2 356 A. T. MAPANDE AND C. J. C. REASON eastern parts and western parts of southern Africa respectively. However, western Tanzania seems to lie outside the main areas of influence of these modes. Therefore, it is of interest to examine the interannual rainfall variability of this region. In addition to these large-scale modes, the evolution over the western Indian Ocean of the northeast monsoon and its transitions to and from the southwest monsoon near the beginning and end of the wet season are important. Considerable intraseasonal variability and breaks in convective activity exist during the monsoon, so that the wet season is typically characterized by a number of wet and dry spells. Despite the importance of such wet and dry spells for water resources, agriculture and the prevalence of diseases such as malaria and Rift Valley fever, relatively little work has been done on the frequency of these spells during the rainy season or its onset and withdrawal. Thus, there is a need to understand how these fundamental characteristics may vary on interannual time scales. To help improve understanding of western Tanzanian rainfall variability, this study examines the evolution of the regional circulation associated with anomalously wet and dry seasons. In addition, the interannual variability of the onset and withdrawal of the rainy season and the frequency of dry spells within it are investigated.. DATA AND METHODOLOGY The data used in this study to investigate the evolution of wet and dry spells over western Tanzania include pentad and monthly rainfall, National Centers for Environmental Production (NCEP) National Center for Atmospheric Research reanalyses (Kalnay et al., 996), National Oceanic and Atmospheric Administration (NOAA) outgoing longwave radiation (OLR), and monthly NOAA extended reconstructed SST (Smith and Reynolds, 4). Moisture fluxes are calculated from the winds and specific humidity. The rainfall data consist of mean monthly records at six stations within the period from the Tanzania Meteorological Agency and pentad (5-day mean) Climate Prediction Center merged analysis of precipitation (CMAP) gridded rainfall (Xie and Arkin, 997). The latter merges of satellite, model and rain-gauge data with a spatial resolution of.5.5. The region of western Tanzania (4.75 S, 9 37 E) studied herein has a unimodal rainfall regime with a wet season extending from October to April. Six rainfall stations (Kigoma, Tabora, Sumbawanga, Iringa, Mbeya and Songea; Figure ) were used to analyse interannual rainfall variability over the region. S Kenya 4S 6S 8S Congo Kig Tab Iri Sum Mbe Indian Ocean S Zambia Son 4S 6E 8E 3E 3E 34E 36E 38E 4E 4E 44E 46E 48E 5E Figure. The six rainfall stations in western Tanzania used for identifying wet and dry years during (the station names for Kigoma, Tabora, Sumbawanga, Iringa, Mbeya and Songea are abbreviated)

3 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA 357 Anomalies Anomalies (a) Years (b) Years.5 Sumbawanga (Oct-Apr) Seasonal Rainfall Anomalies Songea (Oct-Apr) Seasonal Rainfall Anomalies ( ).5 Anomalies Anomalies (c) Years (d) Years.5 Tabora (Oct-Apr) Seasonal Rainfall Anomalies ( ).5.5 Kigoma (Oct-Apr) Seasonal Rainfall Anomalies ( ).5 Anomalies Anomalies (e) Years (f) Years Figure. Standardized anomalies of station rainfall for the austral summer (October April): (a) Iringa; (b) Mbeya; (c) Sumbawanga; (d) Songea; (e) Tabora; (f) Kigoma. The year on the horizontal axis corresponds to the OND of the given season Figure plots standardized departures in October April rainfall for the 968/69 998/99 summer seasons for each station where the year on each axis refers to the OND months of a given season. Although there is some variation in the magnitude of the anomalies at the different stations, inspection of the time series shows that the six wettest seasons are , 97 7, , 979 8, and and the six driest seasons are 969 7, , , 99 93, and The years selected after 979 are the same as those derived by spatially averaging CMAP data (Xie and Arkin, 997) over western Tanzania. Since the station data are only available as monthly totals, pentad (5 day) CMAP rainfall totals, averaged over 4.75 S, 9 37 E, are used to analyse the variability in onset and withdrawal and in dry-spell frequency during the wet and dry seasons starting in Note that CMAP data are not available prior to 979. Onset of the rainy season was determined as the first pentad with rainfall of at least mm followed

4 358 A. T. MAPANDE AND C. J. C. REASON by three consecutive pentads having average rainfall of at least mm/pentad. Withdrawal of the rainy season was determined as that pentad followed by three consecutive pentads of not more than mm per pentad of rain. The period analysed each year started with pentad 48 (late August) and ended with pentad 8 (June) so as to encompass fully any early start or late withdrawal of a particular season. 3. INTERANNUAL VARIABILITY Figure shows that each station is characterized by substantial interannual variability in summer rainfall. In addition, some of the stations show interannual persistence in rainfall anomalies, which hints at variability near the lower frequency part of the interannual range. To investigate this possibility, a wavelet analysis of the summer rainfall data was performed. This analysis revealed the largest power at time scales of 5 6 years throughout the record, but particularly in the 97s and 98s, and also at 3 years during the 99s. These time scales are consistent with the peaks found in earlier work using spectral analysis (Nicholson and Entekhabi, 986). Of the six identified extreme wet and dry seasons, it is seen that three in each case occur during an ENSO event, whereas only 98 (wet) and 99 (dry) correspond to an IOD event. However, during the period plotted in Figure, positive (negative) IOD events occurred during 97, 977, 98, 994 and 997 (97, 989, 99 and 996). Figure shows that all the positive IOD events correspond to seasons with above-average rainfall; but, except for 98, the magnitudes are relatively weak. Of the negative IOD years, only 99 is dry at all six stations, and 97 and 989 are rather wet. This result suggests that any IOD western Tanzanian rainfall relationship is nonlinear with respect to IOD phase and relatively weak in terms of the summer rainfall season as a whole, although it is recognized that larger anomalies occur if only the early summer (OND) is considered. Since three out of the six seasons occur during ENSO for both the wet and dry years, it is also of interest to analyse the three non-enso seasons separately to see whether different mechanisms are responsible. NCEP reanalyses are used to examine the composite anomaly evolution for the wet and dry summers and comparison is made between the three non-enso (968 69, , 979 8) and three ENSO seasons (97 7, and ). Monthly plots corresponding to the approximate onset, mid and withdrawal periods of the rainy season over western Tanzania in October, December and April are shown. During October, the moisture flux plot for the wet composite indicates a strong lower tropospheric anticyclonic anomaly over the western Indian Ocean (Figure 3(a)). This anomaly weakens the equatorial mean westerly export of moisture away from East Africa. At the same time, the northeasterly anomaly near the southern Tanzanian coast helps to increase moisture penetration into the western half of the country via the northeast monsoon. North and east of Madagascar, westerly anomalies imply a deceleration of the mean low-level southeasterly trade wind flow over the western south Indian Ocean. As a result, increased low-level moisture convergence is expected over western Tanzania, as indeed is the case (not shown). A secondary moisture source is the tropical southeast Atlantic, where anomalous westerlies extending into southern Congo and Zambia lead to convergence of low-level moisture over western Tanzania with that originating from the north Indian Ocean. Negative OLR anomalies, and hence increased rainfall, are evident over Zambia and western Tanzania (Figure 3(b)). Lower (85 hpa) and upper ( hpa) level velocity potential plots (Figure 4) show evidence of increased low-level (upper level) convergence (divergence), and hence uplift, over the western Indian Ocean Tanzania, consistent with increased rainfall. In December, near the peak of the season, the centre of the anticyclonic moisture flux anomaly over the Indian Ocean moves east, allowing westerly anomalies to extend from the southern Congo basin right across Tanzania (Figure 5(a)). This pattern acts to increase (decrease) the climatological flux towards (away from) East Africa that exists roughly north (south) of about 3 S in December and hence increase the climatological flux reaching Tanzania. In addition, the westerly anomalies over the land transport more lowlevel moisture from the Congo basin, as well as act to decelerate the mean flow from the south Indian Ocean, leading to substantial low-level moisture convergence over the country (not shown) and rainfall. Low-level convergence and large negative OLR anomalies are apparent over Tanzania northern Mozambique northern

5 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA S W (a) E E 3E 4E 5E 6E 7E 8E 3 4S W E E 3E 4E 5E 6E 7E 8E (b) Figure 3. Composite of wet year anomalies for October (onset of rainy season): (a) the 85 hpa moisture flux (a scale vector of g kg m s is shown); (b) OLR (contour interval W m ). The wet years, determined from the monthly station rainfall data are , 97 7, , 979 8, and Zambia (Figure 5(b)). The velocity potential plots (Figure 6) show enhanced overturning in the eastern Africa westernindian Ocean sector and uplift over Tanzania. Taken together, this, together with the southward shift of the intertropical convergence zone over southern Tanzania during this month, suggests a substantial rainfall increase in December. As the rainy season draws to a close in April, relatively weak anticyclonic moisture flux anomalies exist over the tropical western Indian Ocean (Figure 7(a)). Relative to the mean flux, these anomalies imply less (more) low-level moisture transport into Kenya (southern Tanzania) and reduced moisture export by the equatorial westerlies away from East Africa. Such conditions are favourable for ongoing wet conditions over western Tanzania during this month. Thus, the OLR anomalies show negative values over this region (Figure 7(b)). Furthermore, low-level moisture convergence (not shown) is evident over western Tanzania, where westerly anomalies emanating from the southern Congo converge over western Tanzania with easterly flow from the Indian Ocean.

6 36 A. T. MAPANDE AND C. J. C. REASON N N 3 E E 3E 4E 5E 6E 7E 8E 9E E E E Oct 968, 978, 979, 97, 98, 986 (a) e+6 5e e+6 e+5 N N 3 E E 3E 4E 5E 6E 7E 8E 9E E E E Oct 968, 978, 979, 97, 98, 986 (b) 3.5e+6 e+6 5.5e+6 Figure 4. As Figure 3, except (a) lower and (b) upper level velocity potential 3.. Comparison of non-enso and ENSO wet summers An examination of the individual seasons indicated that the circulation anomalies for the three non-enso wet seasons are rather similar to those for the wet seasons, except that there is a strengthening (weakening) of the common features in the tropics of the western Indian (Pacific) Ocean. To aid comparison, Figure 8 shows the austral summer difference between the non-enso and ENSO wet seasons, highlighting the stronger uplift over East Africa and neighbouring ocean in the former. In addition, the non-enso wet summers show warmer SST anomalies near the Tanzanian Kenyan coast, aiding local evaporation and moisture convergence. Warmer SST anomalies in the non-enso composite also exist near Angola. This feature is important, as it helps strengthen the Angola low (Cook et al., 4) and the westerly moisture flux anomalies over the southern Congo and western Tanzania (Figures 3(a) and 5(a)), which were shown above to be associated with increased rainfall over the region. To shed further light on the conditions associated with the non-enso wet summers, transects of zonal wind just upstream of the region were examined. Again, the major features are similar, but the low-level westerlies from the southern Congo basin are stronger at 8 E, just upstream of western Tanzania, in the non-enso composite than in the ENSO composite at the start of the rainy season (October), and to a lesser extent midway through it in December (Figure 9). Taken together, these differences in velocity potential and

7 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA S W E E 3E 4E 5E 6E 7E 8E (a) 3 4S W E E 3E 4E 5E 6E 7E 8E (b) Figure 5. As Figure 3, except for December (middle of rainy season) zonal wind suggest that the regional circulation anomalies that lead to wetter summers are the same for the non-enso and ENSO cases, but that the former are stronger. 3.. Dry summers Examination of the dry composite 85 hpa moisture flux, OLR and velocity potential fields for October, December and April shows that the important modulations to the equatorial westerlies and south Indian Ocean trade winds discussed above for the wet case are essentially reversed. Thus, there are stronger equatorial westerlies taking more moisture away from East Africa, particularly in October and December of the dry composite, as well as a cyclonic anomaly northeast of Madagascar. Less low-level moisture is advected over western Tanzania from the southern Congo basin. Since the patterns are more-or-less opposite to the wet case, they are not shown for brevity; however, their reversal lends credence to the suggestion that the modulations of the atmospheric circulation over the Indian Ocean are important for influencing western Tanzanian rainfall. The OLR patterns suggest that most of tropical eastern Africa will be dry during these years, not just western Tanzania, whereas they imply enhanced convection over the eastern Indian Ocean. This suggests that

8 36 A. T. MAPANDE AND C. J. C. REASON N N 3 E E 3E 4E 5E 6E 7E 8E 9E E E E Dec 968, 978, 979, 97, 98, 986 (a).4e+6 6 e+6 N N 3 E E 3E 4E 5E 6E 7E 8E 9E E E E Dec 968, 978, 979, 97, 98, 986 (b) 3e+6 e+6 e+6 3e+6 Figure 6. As Figure 4, except for December the ascending branch of the local Walker circulation is strengthened over the eastern Indian Ocean during dry years with relative subsidence over Tanzania, as indicated by upper and lower level velocity potential plots. Such a coherent modulation of the Walker circulation across the Indian Ocean during these dry years is more pronounced than the reverse modulation during the wet composite. 4. INTERANNUAL VARIABILITY IN DRY SPELLS The analysis discussed above indicates how the regional circulation evolves at various stages of the rainy season. In general terms, there is typically a significant increase in rainfall as the rainy season progresses from October to December, and often a relatively sharp decrease in April or May. However, similar to other tropical regions, the rainy season over western Tanzania is characterized by a number of wet and dry spells on synoptic and intraseasonal scales. The number of dry spells during the rainy season and how this number varies from one year to the next are of particular interest to farmers and other user groups. In addition, it is of interest to see how these dry spells are distributed temporally through each of the extreme wet and dry seasons examined in Section 3.

9 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA S W E E 3E 4E 5E 6E 7E 8E (a) 3 4S W E E 3E 4E 5E 6E 7E 8E (b) Figure 7. As Figure 3, except for April (near the end of the rainy season) Dry-spell frequency and its interannual variability have been examined for southern Africa as a whole by Usman and Reason (4), who found that a robust relationship existed between anomalies in this frequency and that in Niño 3.4 SST. To see whether this relationship also holds true for the western Tanzanian region, Figure plots the standardized anomaly in Niño 3.4 SST and in dry-spell frequency calculated over western Tanzania. The same criterion of a dry spell has been used as in Usman and Reason (4), namely as a pentad with less than 5 mm of rainfall. This criterion is very similar to that used by the Famine and Early Warning System of the USAID. The data used to construct Figure is pentad CMAP rainfall, since only monthly station data are available thus, the series starts in Evident in Figure is that the identified extreme wet and dry seasons discussed in Sections and 3 also tend to show above- and below-average numbers of dry spells. In terms of the ENSO relationship, Figure suggests that for most of the record there is an inverse relationship between anomalies in dry-spell behaviour and Niño 3.4 SST. However, the El Niño event is a prominent exception. This inverse behaviour is opposite to that found for southern Africa as a whole (Usman and Reason, 4), where the two series tend to be in phase and may reflect the proximity of western Tanzania to equatorial East Africa, which shows

10 364 A. T. MAPANDE AND C. J. C. REASON N N 3 E E 3E 4E 5E 6E 7E 8E 9E E E E Oct to Apr. : 969, 979, 98 minus 97, 983, 987 (a) N N 3 E E 3E 4E 5E 6E 7E 8E 9E E E E Oct to Apr. : 969, 979, 98 minus 97, 983, 987 (b) 3N N N 3 4S W E E 3E 4E 5E 6E 7E 8E 9E E E E Oct to Apr. : 969, 979, 98 minus 97, 983, 987 (c) Figure 8. Difference in (a) lower and (b) upper level velocity potential and (c) SST between non-enso wet summers and ENSO wet summers an opposite-signed ENSO rainfall impact to that of southern Africa. Figure also suggests that the 99s tended to show more seasons with large increases in dry-spell frequency, with the mid to late 98s indicating the reverse. This aspect may reflect the near-decadal-scale variability in ENSO that is known to occur over eastern and southern Africa (Allan, ; Allan et al., 3).

11 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA S S 8S S 9S 6S 3S 3N Oct : 986, 978, 979 minus 97, 98, 988, 8E to 8E S S 8S S 9S 6S 3S 3N Dec : 986, 978, 979 minus 97, 98, 988, 8E to 8E Figure 9. As Figure 8, except zonal wind transect through 8 E, just upstream of western Tanzania To see whether there are any obvious differences in the temporal distribution of the dry spells, Figure plots pentad rainfall from pentad 48 (late August) through the summer to pentad 8 (late June). Note that only the 979 8, and wet seasons and the , 99 93, and dry seasons are shown, since CMAP data are only available back to 979. Using the criteria stated in Section, it is evident from these plots and Table I that the onset of the anomalously wet seasons for occurs during the pentad range of (mid October to early November), whereas for the unusually dry seasons the onset typically occurs around pentads (early to late November). Anomalously wet seasons tend to end considerably later (pentads 97, late April to mid May) than the dry seasons (pentads 94 97, mid to late April). Thus, the three anomalously wet seasons lasted pentads, whereas the duration of the four drier seasons ranges from 9 to 34 pentads, or up to weeks shorter in length. Inspection of the individual time series plotted in Figure suggests that all years experience pentads of relatively intense rainfall (9 mm or more per pentad), but that the wetter seasons tend to have more of these spells than the drier seasons. In addition, relatively dry spells (less than 3 mm per pentad) tend to be more common during the drier seasons. Within the rainy season itself, the wet seasons tend to have more evenly

12 366 A. T. MAPANDE AND C. J. C. REASON Figure. Standardized anomalies in Niño 3.4 SST (light curve) and dry-spell frequency (dark curve) computed for western Tanzania Table I. Mean pentad and dates of the onset and withdrawal of the rains for wet and dry years over western Tanzania during Wet years Dry years Year Pentad and dates of the rain onset Pentad and dates of withdrawal of the rains Year Pentad and dates of the rains onset Pentad and dates of withdrawal of the rains November 6 3 April 7 November 6 April October 6 May 6 November 6 3 April October 5 May 6 November 6 April November 5 April distributed rains, although the dry season also shows this characteristic. In addition, three of the four dry seasons tend to show a more abrupt end to the season, whereas the wet seasons tend to show a more gradual decline of the rainfall intensity. In summary, both the duration of the season and the distribution of the spells of high and low rainfall intensity within the season are important for determining the rainfall totals experienced during the anomalous wet and dry years. Such intraseasonal characteristics have important implications for local agriculture and water resources. 5. SUMMARY AND CONCLUSIONS The evolution of the circulation anomalies associated with particularly wet and dry rainy seasons over western Tanzania has been investigated. It is evident that wetter than average seasons are characterized by an earlier onset of the rains and a later withdrawal, such that the overall season may be up to weeks longer than the drier than average seasons. Anomalously wet seasons are also typified by more evenly distributed rains with

13 INTERANNUAL RAINFALL VARIABILITY OVER WESTERN TANZANIA 367 Precipitation (mm/day) CMAP Pentad precipitation (9 37 E, 4.75 S) (979) (98) Pentad Number (Year) 8 Precipitation (mm/day) (987) CMAP Pentad precipitation (9 37 E, 4.75 S) (988) Pentad Number (Year) Precipitation (mm/day) CMAP Pentad precipitation (9 37 E, 4.75 S) (98) (983) Pentad Number (Year) Precipitation (mm/day) (99) CMAP Pentad precipitation (9 37 E, 4.75 S) (993) Pentad Number (Year) Precipitation (mm/day) (986) CMAP Pentad precipitation (9 37 E, 4.75 S) (987) Pentad Number (Year) Precipitation (mm/day) (993) CMAP Pentad precipitation (9 37 E, 4.75 S) (994) Pentad Number (Year) Precipitation (mm/day) (998) CMAP Pentad precipitation (9 37 E, 4.75 S) (999) Pentad Number (Year) Figure. Area-averaged time series of pentad (5-day) CMAP rainfall (mm/day) over western Tanzania (979 8, 98 83, wet years on the left and , 99 93, , dry years on the right)

14 368 A. T. MAPANDE AND C. J. C. REASON a more gradual end to the season, whereas for drier than average years the reverse tends to be true. Thus, both changes in season length and in rainfall intensity tend to contribute towards determining a wet or dry year. A tendency for the frequency in dry spells to track inversely that in Niño 3.4 SST was noted. There was also some evidence that dry-spell numbers during the 99s tended to be higher than those during the mid to late 98s. The interannual variability of summer (October April) rainfall over western Tanzania appears to be associated with modulations of the equatorial westerlies and southeast trade winds over the south Indian Ocean. In wet (dry) years, weaker (stronger) equatorial westerlies and an anticyclonic (cyclonic) anomaly over the southern tropics act to reduce (enhance) the export of equatorial moisture away from East Africa. In addition, increased (decreased) low-level moisture fluxes from the southern Congo basin tend to occur during the wet (dry) seasons, leading to enhanced (reduced) low-level moisture convergence over western Tanzania. Evidence for a large-scale modulation of the Indian Ocean Walker cell was also found, particularly for the dry years. ACKNOWLEDGEMENTS This study represents part of the first author s MSc thesis. Funding from the Commonwealth Secretariat for technical training is gratefully acknowledged. We thank Dr D. Jagadheesha for helpful discussions. REFERENCES Allan RJ.. ENSO and climatic variability in the last 5 years. In El Niño and the Southern Oscillation: Multiscale Variability, Global and Regional Impacts, Diaz HF, Markgraf V (eds). Cambridge University Press: Cambridge; Allan RJ, Reason CJC, Lindesay JA, Ansell TJ. 3. Protracted ENSO episodes over the Indian Ocean region. Deep-Sea Research II 5: Behera SK, Yamagata T.. Subtropical SST dipole events in the southern Indian Ocean. Geophysical Research Letters 8: Cook C, Reason CJC, Hewitson BC. 4. Wet and dry spells within particular wet and dry summers in the South African summer rainfall region. Climate Research 6: 7 3. Hastenrath S, Nicklis A, Greischar L Atmospheric hydrospheric mechanisms of climate anomalies in the western equatorial Indian Ocean. Journal of Geophysical Research 98: 9. Hirst AC, Hastenrath S Atmosphere ocean mechanisms of climate anomalies in the Angola tropical Atlantic sector. Journal of Physical Oceanography 3: Indeje M, Semazzi FHM, Ogallo LJ.. ENSO signals in East African rainfall seasons. International Journal of Climatology : Kabanda TA, Jury MR Interannual variability of short rains over northern Tanzania. Climate Research 3: 3 4. Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Leetmaa A, Reynolds R, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelewski C, Wang CJ, Jenne R, Joseph D The NCEP/NCAR 4-years reanalysis project. Bulletin of the American Meteorological Society 77: Nicholson SE, Entekhabi D The quasi-periodic behaviour of rainfall variability in Africa and its relation to the southern oscillation. Archiv für Meteorologie, Geophysik und Bioklimatologie, Serie A: Meteorologie und Geophysik 34: Ogallo LJ Relationships between seasonal rainfall in East Africa and southern oscillation. Journal of Climatology 8: Ogallo LJ The spatial and temporal patterns of the East African seasonal rainfall derived from principal component analysis. Journal of Climatology 9: Reason CJC.. Subtropical Indian Ocean SST dipole events and southern African rainfall. Geophysical Research Letters 8: 5 7. Reason CJC.. Sensitivity of the southern African circulation to dipole SST patterns in the south Indian Ocean. International Journal of Climatology : Reason CJC, Allan RJ, Lindesay JA, Ansell TJ.. ENSO and climatic signals across Indian Ocean basin the global context: part, inter-annual composite patterns. International Journal of Climatology : Rouault M, Florenchie P, Fauchereau N, Reason CJC. 3. South East Atlantic warm events and southern African rainfall. Geophysical Research Letters 3: Saji NH, Goswami BN, Vinayachandran PN, Yamagata T A dipole mode in the tropical Indian Ocean. Nature 4: Smith TM, Reynolds RW. 4. Improved extended reconstruction of SST ( ). Journal of Climate 7: Usman MT, Reason CJC. 4. Dry spell frequencies and their variability over southern Africa. Climate Research 6: 99. Webster PJ, Loschnigg JP, Moore AM, Leben RR The great Indian Ocean warming of ; evidence of coupled oceanic atmospheric instabilities. Nature 4(675): Xie P, Arkin PA Global precipitation: a 7-year monthly analysis based on gauge observations, satellite estimates and numerical model outputs. Bulletin of to American Meteorological Society 78: