SEARCH

SEARCH BY CITATION

Keywords:

  • tropical temperate troughs;
  • sea surface temperature;
  • rainfall variability;
  • Mozambique;
  • cluster analysis;
  • Southern Hemisphere planetary waves;
  • moisture fluxes

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

An analysis of the inter-annual frequency of the main summer rainfall producing synoptic system [tropical temperate troughs (TTTs)] over Mozambique and the Southwest Indian Ocean and their relationships with regional sea surface temperatures and circulation patterns is presented for the January–March season. Correlations and composite anomalies based on annual frequency of occurrence of TTTs over Mozambique suggest that TTT occurrence is related to La Niña and Southern Hemisphere planetary waves (wavenumber-3 or -4) that modulate the subtropical high pressure cells in the three oceanic basins. These modulations lead to enhanced easterly winds over the South Indian Ocean and onshore moisture fluxes towards southeastern Africa. A stronger than average Angola low, particularly linked to La Niña, also plays an important role in the frequency and intensity of the TTT, by enhancing the wind and moisture convergence over southeastern Africa. Moisture fluxes over the warm Agulhas current and tropical southeast Atlantic also affect the rainfall variability over southern Africa to some extent during summers with a high frequency of TTT. Positive sea surface temperature (SST) anomalies over the tropical and subtropical South Indian Ocean that are linked to El Niño-Southern Oscillation appear to play a significant role in causing dry conditions over Mozambique. These positive SST anomalies help promote convective activity offshore, and also weaken the northern half of the South Indian Ocean high pressure cell. As a result, the easterly winds are weakened leading to less onshore moisture fluxes than in the wet seasons. Copyright © 2009 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

Statistical and modelling studies (Jury et al., 1996; Reason and Mulenga, 1999; Reason, 2002; Rouault et al., 2003; Washington and Preston, 2006) have shown significant relationships between rainfall anomalies in southern Africa and different SST patterns in the neighbouring oceans. In particular, warm (cool) SST anomalies over the central tropical Indian Ocean and cool (warm) SST anomalies to the south of Madagascar which often occur during El Niño (La Niña) have been associated with dry (wet) conditions over the bulk of southern Africa (Jury et al., 1996; Rocha and Simmonds, 1997a, 1997b; Reason et al., 2000; Mulenga et al., 2003; Washington and Preston, 2006).

Another SST pattern of relevance to southern African summer rainfall is the subtropical South Indian Ocean dipole (SIOD—Behera and Yamagata, 2001; Reason, 2001; Suzuki et al., 2004) that consists of SST anomalies of opposite sign in the Southwest and southeast Indian Ocean. Positive rainfall anomalies tend to be observed when the warm (cool) pole is located in the west (east) of the subtropical South Indian Ocean (Reason, 2001, 2002). Fauchereau et al. (2003) and Hermes and Reason (2005) showed that this SIOD can often occur at the same time as a similar dipole in the South Atlantic and South Pacific Oceans and is related to surface heat flux anomalies generated by modulations to the atmospheric wavenumber-3 or -4 pattern in the mid-latitude Southern Hemisphere.

Tropical temperate troughs (TTTs—Figure 1 shows examples of the cloud bands associated with these systems) have been recognised as the main summer synoptic rain-producing systems over southern Africa (Harrison, 1984; Washington and Todd, 1999; Tyson and Preston-Whyte, 2000; Reason et al., 2006). Collectively over the austral summer, they contribute to the South Indian convergence zone (SICZ—Cook, 2000). However, the mechanisms that influence either the frequency of occurrence of these systems over southern Africa or their intensity are not well understood. The significant relationship between rainfall over southern Africa and SST patterns in the South Indian Ocean may imply that these SST patterns play an important role in influencing the position, frequency and/or intensity of TTT. However, these relationships between TTT and SST in the Indian Ocean have not been thoroughly explored.

thumbnail image

Figure 1. Satellite images showing cloud bands typical of TTTs over southern Africa on 2nd and 7th December 2005 (courtesy of the Mozambique Institute of Meteorology)

Download figure to PowerPoint

Some studies have shown relationships between the position of the SICZ and El Niño-Southern Oscillation (ENSO) (Tyson and Preston-Whyte, 2000; Cook, 2001; Fauchereau et al., 2009). These studies suggest that during La Niña years, the SICZ tends to be located over land and consequently wet conditions prevail (Fauchereau et al., 2009), while during El Niño, the SICZ shift northeastwards and tend to be located over the South Indian Ocean with consequent dry conditions over land (Cook, 2001). Exploring relationships between the SICZ and SST in the South Indian Ocean would help to better understand how ENSO affects the position and intensity of the SICZ and also the links between rainfall over southern Africa and SST in the South Indian Ocean. The aim in this study is to explore the relationship between the inter-annual frequency and intensity of TTT and SST in the Indian Ocean. The atmospheric mechanisms influencing the TTT will also be explored.

The paper is organised as follows: Section 2 presents the data and methods used in the study while Section 3 discusses the methods used to select the clusters. The relationships between the inter-annual variability of the clusters, rainfall, SST and atmospheric circulation are presented in Section 4 while Section 5 contains the discussion and conclusions.

2. Data and methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

This study concentrates on the late summer rainfall season, i.e. January–March (JFM) and uses daily outgoing long-wave radiation (OLR) (Liebman and Smith, 1996) as a proxy for convective activity associated with the TTTs. The TTTs are identified using cluster analysis (CA). A non-hierarchical clustering method (Gong and Richman, 1995) is used in this study. Non-hierarchical methods start from either (1) an initial partition of items into groups or (2) an initial set of seed points, which form the nuclei of clusters. One way to start is to randomly select seed points from among the items or to randomly partition the items into initial groups (Johnson and Wichern, 1982). The widely applied non-hierarchical K-means method is used here as follows (Gong and Richman, 1995; Wilks, 2006):

  • Step 1.
    Specify k seed points as a set of centroids. The seed points are randomly identified.
  • Step 2.
    Compute the Euclidian distance between the current data vector (xi, i = 1, …, n) items and each of the centroids; then assign each item to the cluster having the nearest centroid to attain the initial partition.
  • Step 3.
    If xi is already a member of the group whose mean is closest, repeat step 2 for xi+1. Otherwise, xi is reassigned to the group whose mean is closest, and step 1 is repeated.

This method has been previously used for atmospheric pattern identification over southern Africa (Fauchereau et al., 2009). Prior to clustering, empirical orthogonal function (EOF) analysis was applied to the OLR data in order to reduce dimensionality. Based on the EOFs, 38 loading patterns henceforth called principal components (PCs) that explain 70% of the variance were retained and the CA was applied to these loading patterns. Nine different partitions from two to ten clusters were considered and, based on a classifiability index (Michelangeli et al., 1995), one partition was selected. The classifiability index compares the similarity of the clusters to the members of each partition. The higher the classifiability index, the more similar are the members of each partition and the more robust is the classification (Solman and Menéndez, 2003).

Composite, correlation and regression analyses are then used to study rainfall, atmospheric and SST patterns associated with the TTT. The atmospheric data used in the analyses are geopotential height (hgt), mean sea level pressure (mslp), net latent heat fluxes at surface, and wind, all from the National Centre for Environmental Prediction (NCEP) two reanalysis (Kanamitsu et al., 2002). Composites of moisture fluxes, calculated as the product of wind and specific humidity (u × q and v × q), were also computed. The SST data used are derived from the more recent HadSST2 monthly data at 1 × 1° resolution (Rayner et al., 2006). For rainfall, the monthly CRU TS 2.0 (Mitchell et al., 2004) data are used. Student's t test is applied to test the significance of the composites and correlations.

3. Cluster selection

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

The classifiability index (Figure 2) shows high values for a partition of two clusters (above 0.8); however, selecting only two clusters might discard potentially important results. Since the next highest values are for partitions of three to five clusters, with approximately the same value, it was decided to select a partition of five clusters. The clusters of interest in the present study are those related to TTT over southern Africa and the SWIO, and particularly those linked to rainfall over Mozambique. Therefore spatial patterns of OLR and rainfall over Mozambique are used to select the clusters for the present study.

thumbnail image

Figure 2. Classifiability index as a function of the number of clusters

Download figure to PowerPoint

Figure 3 shows the composite anomalies of OLR for the selected partition of five clusters. The anomalies represent the difference between days when a particular cluster is observed and the rest of the period.

thumbnail image

Figure 3. (a–e) Composite (days when a particular cluster is observed minus the rest of the period) of OLR anomalies (contour interval is 5 W m−2) for the selected partition of five clusters. Shaded areas are significant at 5% student's t test. Darker (lighter) shading shows negative (positive) significant anomalies. Dashed (solid) contours denote negative (positive) values

Download figure to PowerPoint

Cluster1 exhibits negative anomalies over the western side of southern Africa and also in a diagonal band extending from eastern Madagascar towards the southeast. This cluster also exhibits two diagonal bands of positive anomalies, one extending from southern Mozambique towards the SWIO and another from Tanzania towards the eastern Indian Ocean. Cluster2 exhibits a dipole pattern with positive anomalies extending from southern Mozambique/Zimbabwe towards the central South Indian Ocean and a negative band extending south-eastward from the northern tip of Madagascar. Cluster3 also exhibits a dipole pattern with negative anomalies extending from central southern Africa, straddling south and central Mozambique towards the SWIO. Positive anomalies exist to the north and extend from Tanzania towards the mid-latitude South Indian Ocean. The anomalies of cluster2 and cluster3 show opposing signs over southern Africa/SWIO. Cluster2 suggests enhanced (reduced) convection over SWIO (southern Africa). The reverse is true for cluster3. Clusters 2 and 3 are reminiscent of TTT activity located either over the land or over the SWIO (Reason et al., 2006).

Cluster4 exhibits positive anomalies over southern Africa and over the nearby South Indian and South Atlantic Oceans. The patterns of this cluster suggest significantly reduced convective activity over southern Africa. Cluster5 exhibits negative anomalies extending eastward from Tanzania across the equatorial South Indian Ocean. The spatial distribution of this pattern appears to be associated with the southernmost position of the Inter-tropical Convergence Zone (ITCZ). This cluster also exhibits positive anomalies over the SWIO. Over southern Africa, the anomalies are weak.

Based on the annual frequencies (Figure 4) of each cluster, composites of CRU2 rainfall data are presented as the difference between years with above average (Table I) and years with below average occurrence for each cluster. The composites of CRU2 rainfall data (Figure 5) show significant rainfall anomalies over Mozambique for clusters 2 and 3. Cluster2 (Figure 5(b)) exhibits negative rainfall anomalies over the bulk of southern Africa, with significant anomalies extending in a diagonal band from Zimbabwe/Zambia southeast to central and southern Mozambique. Over Tanzania, the anomalies are positive. Cluster3 (Figure 5(c)) exhibits an anomaly pattern that is symmetric with cluster2, with positive anomalies over the bulk of southeastern Africa. Positive anomalies are observed extending from central southern Africa to southern and central Mozambique. Over Tanzania and northern Mozambique, the anomalies are negative but weak. In general, the other three clusters show weak patterns, suggesting that the inter-annual variability of rainfall over southern and central Mozambique is not clearly linked to these clusters (1, 4, 5). Given the apparent importance of clusters 2–3 for Mozambique rainfall, the SST and atmospheric circulation associated with these two clusters are explored in Section 4.

thumbnail image

Figure 4. Annual frequency of the clusters for the selected partitions. Vertical axes show the number of days each cluster is observed per year, the dot dashed lines show the mean number of observations

Download figure to PowerPoint

thumbnail image

Figure 5. (a–e) Composites of rainfall anomalies (contour interval is 20 mm). Shaded areas are significant at 5% student's t test. Darker (lighter) shading shows positive (negative) significant anomalies. Dashed (solid) contours denote negative (positive) values. Box in (c) indicates the averaged area for the rainfall and OLR indices used for the correlations with SST

Download figure to PowerPoint

Table I. Years used in the composites of the annual frequencies of the clusters
  Cluster2 (JFM)Cluster3 (JFM)
YearsAbove normal1979, 1980, 1987, 1992, 1995, 1998, 2002, 20031981, 1990, 1991, 1996, 1997, 1999, 2000, 2001
 Below normal1981, 1982, 1983, 1984, 1985, 1986, 1988, 1989, 1990, 1991, 1993, 1994, 1996, 1997, 1999, 2000, 20011979, 1980, 1982, 1983, 1984, 1985, 1986, 1987, 1988, 1989, 1992, 1993, 1994, 1995, 1998, 2002, 2003

4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

In this section, inter-annual relationships between ENSO, regional SST and clusters 2 and 3 are explored using composite, correlation and regression analyses. Composite (years of high frequency minus years of low frequency of the cluster) analyses are first applied to the raw data and, second, after removing the linear influence of ENSO from the SST and geopotential height by linear regression. The aim is to explore other patterns (independent of ENSO) of SST anomalies over the Indian and Atlantic Oceans that may influence the atmospheric circulation patterns that are associated with the anomalies of rainfall over Mozambique. However, it should be noted that linear regression will not isolate non-linear ENSO interactions with other modes.

Correlations between the annual frequencies of the clusters and the Niño 3.4 index (Table II) suggest that a strong relationship exists between the annual frequency of cluster2 and ENSO. Correlations are positive and significant from 7-months lag (JJA) to 0-month lag (JFM). These strong correlations between cluster2 and the Niño 3.4 index, with significant lead time may indicate the importance of this index for seasonal prediction of JFM rainfall in eastern southern Africa during El Niño years. Cluster3 shows consistent negative correlations with Niño 3.4 index, but the correlations are not as strong as for cluster2, and are only significant (at p ≤ 0.05) for 1- and 0-month lag. These correlations may suggest that for cluster3 other atmospheric or oceanic forcing mechanisms besides ENSO may play an important role.

Table II. Lagged correlations between annual frequencies of clusters 2 and 3 and Nino3.4 index from 12- to 0-month lag
 JFMFMAMAMAMJMJJJJAJASASOSONONDNDJDJFJFM
  1. Numbers in bold show anomalies with p ≤ 0.05.

Cluster2− 0.13− 0.11− 0.010.190.360.490.530.560.590.620.650.660.67
Cluster3− 0.15− 0.15− 0.18− 0.20− 0.21− 0.23− 0.26− 0.29− 0.32− 0.34− 0.370.410.45

Composite SST anomalies associated with cluster2 show warm and significant anomalies over the tropical and subtropical South Indian Ocean and cool anomalies over the SWIO and mid-latitudes in the South Atlantic (Figure 6(a)). Lagged composites (with SST leading) display persistent warm anomalies over the subtropical Indian Ocean for periods longer than 3-months lag. Figure 6(b) shows the composite anomalies at 3-months lag (OND anomalies). This pattern is consistent with the negative SIOD (Behera and Yamagata, 2001; Reason, 2001) in which positive (negative) anomalies are observed over the eastern (western) side of the South Indian Ocean. However, in our results, the warm pole is stronger than the cool pole which does not show significant anomalies. The pattern is also similar to El Niño associated anomalies in which warm SST anomalies are observed over the bulk of the tropical and subtropical Indian Ocean (Mo, 2000; Reason et al., 2000). As already indicated, both of these SST patterns have been associated with dry conditions over subtropical southern Africa (Jury et al., 1996; Reason, 2001, 2002). At 0-month lag, when the linear influence of ENSO is removed, the anomalies become insignificant over both the Indian and Atlantic Oceans (figures not shown).

thumbnail image

Figure 6. Composite anomalies of SST (contour interval is 0.2 degrees Celsius) associated with cluster2 at 0-lag (a) at 3-months lag with SST leading the cluster observation (b). Dashed (solid) contours denote negative (positive) values. Shaded areas are significant at 5% student's t test. Darker (lighter) shading shows regions with negative (positive) significant anomalies. The box in (a) shows the averaged area for the SST index used in the correlations (see text)

Download figure to PowerPoint

Composite SST anomalies linked to cluster3 (Figure 7) exhibit a dipole pattern anomaly with a positive pole over the SWIO and the mid-latitude South Atlantic and negative anomalies over the central-eastern subtropical South Indian and Atlantic Oceans. The significance over the positive anomalies appears larger. These patterns are reminiscent of the co-evolving subtropical SST dipoles in the South Indian and Atlantic Oceans analysed by Fauchereau et al. (2003) and Hermes and Reason (2005). When lagged composites are considered, the patterns of anomalies described above weaken and vanish completely at 3-months lag (figures not shown). When the influence of ENSO is removed, the positive pole over the SWIO south of Madagascar and that in the southwest Atlantic remains robust, but the negative pole in the central-eastern regions is less significant (Figure 7(b)).

thumbnail image

Figure 7. Same as Figure 6 except for cluster3 at 0-months lag (a) and at 0-lag but with the linear influence of ENSO removed (b)

Download figure to PowerPoint

The links between rainfall and the patterns of SST in the South Indian Ocean are explored further using correlations between time series averaged in the boxes indicated in Figure 5(c) (rainfall) and Figures 6(a) and 7(a) (SST). Tables III and IV show the corresponding correlation coefficients.

Table III. Correlations of time series of SST (warm pole) with the annual frequencies of OLR and rainfall (averaged in the box shown in Figure 5(c)) associated with cluster2
Season (SST)OLR/warmRainfall/warm
 rprp
  1. Numbers in bold indicate values significant at p < 0.05.

SON0.280.18− 0.270.18
OND0.30.14− 0.310.13
NDJ0.370.07− 0.380.06
DJF0.430.030.440.03
JFM0.420.040.470.02
Table IV. Correlations of time series of SST (warm and cold poles) with the annual frequencies of OLR and rainfall associated with cluster3 (averaged in the box shown in Figure 5(c))
Season (SST)OLR/warmOLR/coolRainfall/ warmRainfall/ cool
 rprprprp
  1. Numbers in bold indicate values significant at p < 0.05.

SON0.00.980.340.10.00.99− 0.380.06
OND− 0.280.180.250.230.250.22− 0.310.14
NDJ0.520.010.260.220.480.02− 0.290.15
DJF0.530.0060.370.070.470.02− 0.370.07
JFM0.510.010.410.040.430.030.400.05

For cluster2, only warm anomalies are considered in the correlations because cool anomalies are in general weak (Figure 6). Correlations between inter-annual anomalies of SST and rainfall show negative values, while correlations between inter-annual anomalies of SST and OLR show positive values. A consistent increase of the coefficients, as the lag decreases, is observed in the correlations of SST with both rainfall and OLR (Table III). The strongest correlations are observed at 1- and 0-month lag (significant at p < 0.05). At 2-months lag, the correlations weaken, but significant (at p < 0.1) values are observed for both OLR and rainfall. Weak correlations are still observed at 3- and 4-month lags.

For cluster3, the warm pole displays strong and significant correlations (at p < 0.05) from 2-months lag (NDJ) with both OLR and rainfall (Table IV). When the lag is increased, the correlations weaken significantly and at 4-months lag no correlations are observed. The warm pole over the subtropical SWIO has been shown to force positive anomalies of rainfall over subtropical southern Africa (Reason and Mulenga, 1999; Reason, 2002). Correlations with the cool pole for cluster3 are observed at all lags considered here, however, they are not stable. For example, at 4-months lag (SON) correlations are significant (at p ≤ 0.1), and they weaken between 3- and 2-months lag, while at 1-month lag, the correlations again become significant (at p ≤ 0.1), and they display strongest values at 0-month lag (significant at p ≤ 0.05). The reason for these changes in the significance of correlation is not clear. However, similar behaviour is observed in the composites of ENSO-related rainfall over southeastern Africa in Nicholson and Kim (1997). Specifically over central and southern Mozambique, the rainfall response to ENSO forcing is strong between October and November but weakens between December and January before strengthening again between February and April (see Figure 4 in Nicholson and Kim, 1997). This result may indicate that the correlations observed here between rainfall/OLR and the SST anomalies in the Indian Ocean are linked to ENSO forcing.

To explore the relationships between the atmospheric circulation and also the role of the Indian and Atlantic Ocean SST on the frequencies of the clusters, composites of geopotential height, wind, moisture fluxes and net latent heat fluxes are computed.

The 500 hPa geopotential height pattern associated with cluster2 displays positive and significant anomalies over the entire tropical belt and almost all of southern Africa (Figure 8). This pattern of geopotential height anomalies is also observed at lower and upper levels (figures not shown) and is typical of El Niño (Reason et al., 2000; Tyson and Preston-Whyte, 2000). In the mid-latitudes of the Southern Hemisphere, the patterns of geopotential height anomalies tend to be negative, suggesting a northward shift of the subtropical jet and more dry westerly air being advected over subtropical southern Africa during seasons with high frequencies of cluster2.

thumbnail image

Figure 8. Composites of geopotential height (contour interval is 4 m) anomalies at 500 hPa for the annual frequencies of cluster2. Dashed (solid) contours denote negative (positive) values. Shaded areas show values significant at 5% student's t test. Darker (lighter) shading shows regions with positive (negative) significant anomalies

Download figure to PowerPoint

Composites of moisture fluxes at 850 hPa associated with cluster2 (Figure 9) show relative convergence over the warm SST anomalies in the central South Indian Ocean. This low-level convergence of moisture fluxes may indicate the position of the SICZ to the northeast of Madagascar and also may represent weakening of the South Indian Ocean trade winds, and consequently weak onshore moisture fluxes. The weakening of the onshore moisture fluxes may explain the negative anomalies of rainfall (Figure 5(b)) over the bulk of southern Africa.

thumbnail image

Figure 9. Composite of moisture fluxes anomalies (Kg m s−1 with arrows scaled at 1 unit per degree latitude) at 850 hPa associated with annual frequency of cluster2

Download figure to PowerPoint

Composite anomalies of geopotential height associated with the annual frequency of cluster3 display a wavenumber-3 or -4 pattern (Figure 10) with positive anomalies over the three ocean basins. Over southwestern Africa, a low-pressure anomaly is observed at 500 hPa (Figure 10(a)), however, it displays the strongest values at 700 hPa level (figure not shown). This pattern suggests a deep Angola low, the source region of the TTT (Reason et al., 2006), and it has been associated with La Niña and positive anomalies of rainfall over southern Africa (Tyson and Preston-Whyte, 2000; Reason and Jagadheesha, 2005). However the low pressure cell is located to the west of the largest rainfall anomalies [compare Figure 5(c) and 10(a)), suggesting that it is not only the regional air mass uplift over southern Africa that contributes to the observed anomalies of rainfall during La Niña. Low-level convergence of winds and of moisture flux over southeastern Africa associated with enhanced pressure gradient between the northwest sector of the South Indian Ocean anticyclone and southern and central Mozambique, together with the deeper Angola low also plays an important role (Figures 10(a) and 11). Note that the pattern of moisture fluxes over southeastern Africa associated with cluster3 (Figure 11) is similar to that observed in Todd and Washington (1999) for major TTT events located over southern Africa/SWIO in the late summer (January).

thumbnail image

Figure 10. (a and b) Same as Figure 8 except for cluster3 with raw data (a) and the linear influence of ENSO removed (b)

Download figure to PowerPoint

thumbnail image

Figure 11. Same as Figure 9 except for cluster3

Download figure to PowerPoint

When the linear influence of La Niña is removed, the wavenumber-3 or -4 structure with the centres over the three ocean basins remains robust and the anomalies are strengthened on the northern and western sides of the anticyclones in the lower- and mid-troposphere (Figure 10(b)). However, the geopotential gradient over the northern and western sides of the anticyclones weakens significantly. Furthermore, the low-pressure cell over southern Africa also weakens significantly. This result is consistent with studies showing that changes in this low-pressure cell are linked to La Niña (Tyson and Preston-Whyte, 2000; Reason and Jagadheesha, 2005).

Latent heat fluxes from NCEP re-analysis were composited in order to explore possible physical mechanisms linking SST over the South Indian Ocean and atmospheric circulation associated with the rainfall anomalies. The patterns for the composite of net latent heat flux for cluster2 are not significant (figure not shown). Different from cluster2, the composite of net latent heat flux for cluster3 shows strong positive anomalies over the southeast Indian Ocean, off Australia and also over the tropical SWIO, east of Madagascar (Figure 12). These anomalies are consistent with strong easterly anomalies of wind over the same regions (Figure 11) and suggest that increases in latent heat flux play a role in the cool pole of SST associated with cluster3 (Figure 7) consistent with Hermes and Reason (2005). South and west of South Africa, significant positive anomalies in the net latent heat fluxes are also observed. SST anomalies south of South Africa in the Agulhas region are weak, but off the west coast of South Africa, negative anomalies prevail (Figure 7). Strengthening of the south-easterlies over the west coast of South Africa (Figure 11) may be responsible for the positive latent heat flux anomalies observed here. Strong positive anomalies of latent heat fluxes are also observed in central southern Africa, apparently linked to a surplus of moisture over the region. Lagged composites (OND, NDJ and DJF: 3-, 2- and 1-months lag, respectively) show that the anomalies of latent heat are strongest in JFM and weaken significantly in OND (figures not shown), indicating that the anomalies are not associated with pre-season soil moisture. Over the warm pole in the SWIO, the anomalies of net latent heat flux are in general weak and with mixed sign. Positive (negative) anomalies are observed over the western (eastern) side of the SST pole. Negative anomalies of latent heat flux indicate that the ocean is gaining heat and this suggests atmospheric forcing the SST pole (Zhang and McPhaden, 1995; Hermes and Reason, 2005).

thumbnail image

Figure 12. Composite anomalies of net latent heat fluxes (contour interval is 10 W m−2) for cluster3. Dashed (solid) contours denote negative (positive) values. Positive values indicate upward fluxes. Shaded areas show values significant at 5% student's t test

Download figure to PowerPoint

5. Discussion and conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

The main objective in this study was to explore the mechanisms by which the SST over Indian Ocean may influence the inter-annual variability of the TTTs and the associated rainfall anomalies over southern Africa. The TTT events were identified using cluster analysis. Two clusters were considered in the study: cluster3 representing the seasonal location of TTT over land and cluster2 the seasonal location of TTT offshore. These are the typical seasonal mean position of the TTT during wet and dry years over southern Africa particularly during ENSO years (see Figure 13.21 in Tyson and Preston-Whyte, 2000).

Composites of rainfall for clusters 2 and 3 displayed significant anomalies over subtropical southern Africa, indicative of the importance of these two clusters for the JFM seasonal rainfall in the sub-region.

Cluster3 was found to be associated with Southern Hemisphere planetary waves (wavenumber-3 or -4), suggesting that planetary wave modulation of the South Indian anticyclone, such that the anticyclone is strengthened and shifted north-westward, is an important mechanism for generating the TTT. The mechanisms for generating the TTT are associated with the strengthening of the easterly winds which result in stronger than average moisture fluxes towards southeastern Africa. The modulation of the South Indian anticyclone by Southern Hemisphere planetary waves was previously suggested as being important in the evolution of the subtropical South Indian Ocean SST dipole mode (Fauchereau et al., 2003; Hermes and Reason, 2005). However, in this study these planetary waves were shown to be important in the inter-annual variability of the TTT.

La Niña was also found to influence the intensity and frequency of TTT. Composites of geopotential and moisture fluxes suggested that the influence occurs through enhancement of the pressure gradient between the tropical and subtropical regions over the SWIO and also through strengthening of the Angola low over central southern Africa.

Low-level easterly anomalies of wind over the tropical SWIO, resulting from increased pressure gradients and the strengthening of the South Indian anticyclone, were associated with increased net latent heat fluxes. As a result, cool SST anomalies evolve in the tropical and subtropical South Indian Ocean. These links between cool anomalies of SST, low-level winds, latent heat and moisture fluxes may in part explain the previously identified links between cool SST anomalies in the tropical South Indian Ocean and rainfall over southern Africa (Reason, 2002; Washington and Preston, 2006).

The overall pattern of SST associated with cluster3 is similar to the pattern discussed in Washington and Preston (2006) and associated with two extreme rainfall years (1974 and 1976) in southern Africa. To explore the influence of this SST pattern on southern Africa rainfall, these authors used several general circulation model experiments. Their results suggest that a cool SST pattern east of Madagascar forces a low-level anticyclone that is accompanied by significant low-level anticyclonic wind and moisture flux anomalies. Similar circulation patterns are observed in the present study; however, the analyses here also suggest a link with large-scale Southern Hemisphere planetary waves. These results may indicate that both cool SST to the east of Madagascar and large-scale planetary waves play an important role in forcing low-level pressure anomalies in the Mascarene region that in turn contribute to increased frequencies of TTT over southern Africa.

Warm anomalies over the SWIO, south of Madagascar, were observed to be associated with a high frequency of TTT. These warm anomalies of SST have been also previously associated with positive anomalies of rainfall over southeastern Africa (Reason and Mulenga, 1999; Reason, 2002). The associations between rainfall and the warm anomalies of SST may be in part explained by the forcing exerted by the Southern Hemisphere planetary waves on both rainfall (discussed here) and on the positive anomalies of SST over SWIO, south of Madagascar (discussed in Hermes and Reason, 2005).

The composites of SST associated with the negative anomalies of rainfall over southern and central Mozambique (cluster2) were characterised mainly by warming over the tropical and subtropical Indian Ocean, typical of ENSO (Reason et al., 2000). Low-level westerly wind anomalies were observed to converge over the position of the largest warm SST anomalies in the central South Indian Ocean. This result suggests that the SST anomalies are forcing this convergence and is in agreement with other studies on the role of the SST on forcing dry conditions over subtropical southern Africa (Rocha and Simmonds, 1997a, 1997b; Reason, 2001, 2002).

While the influence of ENSO in summer rainfall over southern Africa is fairly well understood, large-scale Southern Hemisphere planetary waves appear to play a significant role as well, an aspect that has not been sufficiently explored in previous work. It appears from the analyses in this study that these planetary waves may also explain much of the statistical relationship between the SIOD and rainfall over southeast parts of southern Africa, as they have an impact on both.

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References

Atanásio J. Manhique would like to thank SIDA-SAREC (Swedish agency) for funding his PhD studies at the Department of Oceanography, University of Cape Town. This article is part of his PhD dissertation.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data and methods
  5. 3. Cluster selection
  6. 4. Inter-annual variability of the clusters and relationships with ENSO, SST and atmospheric circulation
  7. 5. Discussion and conclusions
  8. Acknowledgements
  9. References
  • Behera SK, Yamagata T. 2001. Subtropical SST dipole events in the southern Indian Ocean. Geophysical Research Letters 28(2): 327330.
  • Cook KH. 2000. The Southern Indian convergence zone and interannual rainfall variability over southern Africa. Journal of Climate 13: 37893804.
  • Cook KH. 2001. A Southern Hemisphere wave response to ENSO with implications for Southern Africa precipitation. Journal of the Atmospheric Sciences 58: 21462162.
  • Fauchereau N, Pohl B, Reason CJC, Rouault M, Richard Y. 2009. Recurrent daily OLR patterns in the Southern Africa/Southwest Indian Ocean region, implications for South African rainfall and teleconnections. Climate Dynamics 32: 575591, DOI 10.1007/S00382-008-0426-2.
  • Fauchereau N, Trzaska S, Richard Y, Roucou P, Camberlin P. 2003. Sea-surface temperature co-variability in the southern Atlantic and Indian Oceans in connections with the atmospheric circulation in Southern Hemisphere. International Journal of Climatology 23: 663677.
  • Gong X, Richman MB. 1995. On the application of cluster analysis to growing season precipitation data in North America east of the Rockies. Journal of Climate 8: 897931.
  • Harrison MSJ. 1984. A generalised classification of South African summer rain-bearing systems. Journal of Climatology 4: 547560.
  • Hermes JC, Reason CJC. 2005. Ocean model diagnosis of interannual co-evolving SST variability in the South Indian and South Atlantic Oceans. Journal of Climate 18: 28642882.
  • Johnson RA, Wichern DW. 1982. Applied Multivariate Statistical Analysis. Prentice-Hall, Englewood Cliff, NJ, USA; 594 pp.
  • Jury MR, Pathack B, Rautenbach CJdeW, Van Heerden J. 1996. Drought over South Africa and Indian Ocean SST: statistical and GCM results. The Global Atmosphere and Ocean System 4: 4763.
  • Kanamitsu M, Ebisuzaki W, Woollen J, Yang S-K, Hnilo JJ, Fiorino M, Potter GL. 2002. NCEP–DOE AMIP-II reanalysis (R-2). Bulletin of the American Meteorological Society 83: 16311643.
  • Liebmann B, Smith CA. 1996. Description of a complete (interpolated) outgoing longwave radiation dataset. Bulletin of the American Meteorological Society 77(6): 12751277.
  • Michelangeli PA, Vautard R, Legras B. 1995. Weather regimes: recurrence and quasi-stationarity. Journal of the Atmospheric Sciences 52: 12371256.
  • Mitchell TD, Carter TR, Jones PD, Hulme M, New M. 2004. A comprehensive set of high-resolution grids of monthly climate for Europe and the globe: the observed record (1901–2000) and scenarios (2001–2100). Tyndall Centre working paper n° 55.
  • Mo KC. 2000. Relationship between low-frequency variability in the Southern Hemisphere and sea surface temperature anomalies. Journal of Climate 13: 35993610.
  • Mulenga HM, Rouault M, Reason CJC. 2003. Dry summers over northeastern South Africa and associated circulation anomalies. Climate Research 25: 2941.
  • Nicholson SE, Kim J. 1997. The relationship of the el niño—southern oscillation to African rainfall. International Journal of Climatology 17: 117135.
  • Rayner NA, Brohan P, Parker DE, Folland CK, Kennedy JJ, Vanicek M, Ansell TJ, Tett SFB. 2006. Improved analyses of changes and uncertainties in sea surface temperatures measured in situ since the mid-nineteenth century: the HadSST2 dataset. Journal of Climate 19: 446469.
  • Reason CJC. 2001. Subtropical Indian Ocean SST dipole events and southern African rainfall. Geophysical Research Letters 28: 22252227.
  • Reason CJC. 2002. Sensitivity of the southern African circulation to dipole sea-surface temperature patterns in the South Indian Ocean. International Journal of Climatology 22: 377393.
  • Reason CJC, Allan RJ, Lindesay JA, Ansell TJ. 2000. ENSO and climatic signals across the Indian Ocean basin in the global context: part I, interannual composite patterns. International Journal of Climatology 20: 12851327.
  • Reason CJC, Jagadheesha D. 2005. A model investigation of recent ENSO impacts over southern Africa. Meteorology and Atmospheric Physics 89: 181205.
  • Reason CJC, Landman WA, Tennant W. 2006. Seasonal to decadal prediction of the southern African climate and its links with the variability of the Atlantic Ocean. Bulletin of the American Meteorological Society 87: 941955.
  • Reason CJC, Mulenga H. 1999. Relationships between South African rainfall and SST anomalies in the Southwest Indian Ocean. International Journal of Climatology 19: 16511673.
  • Rocha AMC, Simmonds I. 1997a. Interannual variability of south-eastern African summer rainfall. Part I: relationships with air-sea interactions processes. International Journal of Climatology 17: 235265.
  • Rocha AMC, Simmonds I. 1997b. Interannual variability of south-eastern African summer rainfall. Part II: modeling the impact of sea-surface temperatures on rainfall and circulation. International Journal of Climatology 17: 267290.
  • Rouault M, Reason CJC, Lutjeharms JRE, Beljaars ACCM. 2003. Underestimation of latent and sensible heat fluxes above the Agulhas Current in NCEP and ECMWF analyses. Journal of Climate 16: 776782.
  • Solman AS, Menéndez CG. 2003. Weather regimes in the South American sector and neighbouring oceans during winter. Climate Dynamics 21: 91104.
  • Suzuki R, Behera SK, Iizuka S, Yamagata T. 2004. Indian Ocean subtropical dipole simulated using a coupled general circulation model. Journal of Geophysical Research 109: C09001, DOI: 10.1029/2003JC001974.
  • Todd MC, Washington R. 1999. Circulation anomalies associated with tropical-temperate troughs in southern Africa and the southwest Indian Ocean. Climate Dynamics 15: 937951.
  • Tyson PD, Preston-Whyte RA. 2000. The Weather and Climate of Southern Africa. Oxford University Press: Cape Town; 396.
  • Washington R, Preston A. 2006. Extreme wet years over southern Africa: role of the Indian Ocean sea surface temperatures. Journal of Geophysical Research 112: D15104.
  • Washington R, Todd M. 1999. Tropical-temperature links in southern Africa and southwest Indian Ocean satellite-derived daily rainfall. International Journal of Climatology 19: 16011616.
  • Wilks DS. 2006. Statistical Methods in Atmospheric Sciences. Academic Press: California; 627.
  • Zhang JZ, McPhaden MC. 1995. The relationship between sea surface temperature and latent heat flux in the equatorial Pacific. Journal of Climate 8: 589605.