Abstract

This paper aims at separating the respective influences of tropical and midlatitude variability on the development and life cycle of tropical temperate troughs (TTTs) over southern Africa in austral summer (November–February). Cluster analysis is applied to 1971–2000 40-yr ECMWF Re-Analysis (ERA-40) daily outgoing longwave radiation (OLR) anomalies to identify TTTs and monitor tropical convection. The same analysis applied to the zonal wind stretching deformation at 200 hPa (ZDEF) characterizes midlatitude transient perturbations. Results based on the comparison between these two classifications first confirm that midlatitude baroclinic waves are a necessary condition for TTT development, but they are not sufficient. Roughly 40% of those occurring in austral summer are associated with a TTT. They tend to be stronger than the baroclinic waves not associated with TTT development. In the tropics, additional conditions needed to form a TTT consist of an excess of latent energy over the Mozambique Channel, mostly because of moisture advections and convergence from the Atlantic and Indian Oceans. Taken together, these conditions are highly favorable for deep atmospheric convection over and near southern Africa and seem to explain a large fraction of TTT variability.

1. Introduction

In southern Africa and the nearby southwest Indian Ocean, the first satellite data revealed cloud bands oriented from the northwest to the southeast and linking the tropics to midlatitude circulation (Harangozo and Harrison 1983). These bands, commonly referred to as tropical temperate troughs (TTTs) and embedded in the south Indian Ocean convergence zone (Cook 2000), develop at the synoptic scale and are responsible for significant amounts of rainfall during austral summer (Harrison 1984, 1986; Crimp et al. 1998; Todd and Washington 1999; Washington and Todd 1999; Tyson and Preston-White 2000; Hart et al. 2012). TTTs are seen as an interaction between tropical convection and midlatitude transient perturbations (Harangozo and Harrison 1983; Harrison 1984; D’Abreton and Lindesay 1993; Todd et al. 2004): Lyons (1991) states that TTTs typically form when a tropical disturbance in the lower atmosphere is coupled with a midlatitude trough in the upper atmosphere. Fauchereau et al. (2009) and Pohl et al. (2009) showed that TTTs tend to propagate eastward, from southern Africa to the Mozambique Channel and southern Madagascar. Their preferential location has a strong influence on intraseasonal and even interannual rainfall variability (Harrison 1984, 1986; Todd and Washington 1999; Washington and Todd 1999).

Figure 1 summarizes the main mechanisms reviewed in the literature to be at the origin of TTT formation. Todd and Washington (1999) found that TTTs are related to moisture convergence supplied by a strong easterly (westerly) flux from the Indian (Atlantic) Ocean. Reason et al. (2006), Chikoore and Jury (2010), or Hart et al. (2010) indicate that the so-called Angola/Botswana low (actually over southern Angola/northern Namibia) or the heat low (Racz and Smith 1999) developing in summertime over the Kalahari both favor low-level penetration of moisture flux from the tropical southeastern Atlantic and could thus be another key mechanism for their initiation and development. Recent work analyzed interactions between TTTs and major modes of large-scale variability, such as the El Niño–Southern Oscillation (ENSO) and the Madden–Julian oscillation. Pohl et al. (2009) show that there is no clear relationship between TTT events and the Madden–Julian oscillation, whereas TTT occurrences are increased during La Niña conditions (Fauchereau et al. 2009; Pohl et al. 2009; Hart et al. 2010; Manhique et al. 2011; Ratna et al. 2012). Sea surface temperature off the southern African coasts was also found to play a significant role in their formation (Crimp et al. 1998). Moisture fluxes associated with warm temperature anomalies over the Agulhas Current (Indian Ocean) increase the occurrences and/or persistence of TTTs (Williams et al. 2007; Manhique et al. 2011; Vigaud et al. 2012). Fauchereau et al. (2009) also suggest that the TTT systems located over the Mozambique Channel are related to the positive phase of the subtropical Indian Ocean dipole (Behera and Yamagata 2001; Reason 2001). Todd and Washington (1999), Hart et al. (2010), and Manhique et al. (2011) finally suggest that TTT events are related to Southern Hemisphere planetary waves (wavenumber 3 or 4).

Fig. 1.

Southern Africa political boundaries and key summer season synoptic features: Angola low (AL), Kalahari heat low (HL), South Atlantic high (SAH), and south Indian Ocean high (SIH).

Fig. 1.

Southern Africa political boundaries and key summer season synoptic features: Angola low (AL), Kalahari heat low (HL), South Atlantic high (SAH), and south Indian Ocean high (SIH).

Despite many studies, the physical mechanism involving (or leading to an interaction between) both tropical convection and midlatitude transient perturbations to influence the development and variability of TTTs in time and space are still not well understood. The aim of this paper is to separate the respective influences of tropical convection and midlatitude transient baroclinic waves favoring the initiation and influencing the life cycle of TTT systems.

This study is organized as follows: Section 2 presents the datasets and the methodology used for this work. Section 3 is devoted to a brief presentation of TTT climatology, including their spatial and temporal variability and contribution to South African summer rainfall. Section 4 focuses on the mechanisms associated with TTT development (i) by analyzing the specific influence of midlatitude transients and tropical convection and (ii) by attempting to separate the necessary and sufficient conditions for TTT development. The results are finally summarized and discussed in section 5.

2. Data and methods

a. Observation and atmospheric data

As in the literature (Todd and Washington 1999; Washington and Todd 1999; Fauchereau et al. 2009; Vigaud et al. 2012), this study concentrates on the summer rainfall season (November–February), when TTT events are most frequent. Atmospheric fields (zonal U, meridional V, and vertical ω components of the wind; air temperature; and specific humidity) are derived from the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40; Uppala et al. 2005). They are used over the period 1971–2002 at a 1.5° × 1.5° spatial resolution.

Composite of moisture fluxes and of moist static energy (MSE) were computed. Holton (1992) indicates that MSE can be an alternative to the equivalent potential temperature, especially when convection is studied. It is defined as

 
formula

where gZ represents potential energy (where g is the gravitational constant and Z is the geopotential height above the surface), CpT is sensible energy (where Cp is the specific heat at constant pressure and T is the absolute air temperature in kelvins), and LQ is latent heat (where L is the latent heat of vaporization and Q is water vapor specific humidity).

Moisture fluxes are defined as

 
formula

where qυ and υυ represent the specific humidity and horizontal velocity at a given tropospheric level.

For consistency with the atmospheric fields described above, tropical convection is estimated using ERA-40 daily outgoing longwave radiation (OLR). OLR values being strongly dependent on the model physics and little constrained by data assimilation, the quality of these data was tested through extensive comparisons with National Oceanic and Atmospheric Administration (NOAA) daily OLR (Liebmann and Smith 1996; Fauchereau et al. 2009). The results appear to be extremely robust, and none of our conclusions is qualitatively modified by this choice.

To analyze TTT effects on rainfall, we use daily rainfall observations over South Africa, Lesotho, and Swaziland, provided by the rain gauge records compiled in the Water Research Commission database by Lynch (2003). A total of 7858 stations, over the period 1970–2000, were extracted from a network of more than 11 000 stations and already used in Pohl et al. (2007).

b. Zonal stretching deformation

Zonal stretching deformation at 200 hPa (ZDEF; Widlansky 2010; Widlansky et al. 2011) can be viewed as a proxy of midlatitude transients that allows characterizing the behavior of the midlatitude baroclinic perturbations (Vigaud et al. 2012). It is defined as

 
formula

where U and x represent the wind zonal velocity and longitude at a given grid point, respectively.

Regions of negative ZDEF correspond in the upper troposphere to a decrease in the zonal component of the wind toward the east (Widlansky 2010; Widlansky et al. 2011), characteristic of midlatitude troughs.

c. Clustering approach

TTT systems are identified following the method used in Fauchereau et al. (2009) and Vigaud et al. (2012). Daily OLR and ZDEF anomalies (after removal of the mean annual cycle) are partitioned using a k-means clustering (Cheng and Wallace 1993; Michelangeli et al. 1995). To reduce dimensionality and ensure linear decorrelation between input variables, empirical orthogonal function (EOF) analysis was applied to the OLR and ZDEF data prior to clustering. The k-means clustering algorithm allows regrouping a series of observations (days) into k regimes minimizing the sum of intraregime variances (Cheng and Wallace 1993; Michelangeli et al. 1995). The Euclidean distance is used to measure similarities between each day and a given regime. In this study, k-means clustering is applied onto austral summer (November–February) ERA-40 daily OLR and ZDEF anomalies. Following the classifiability indexes (not shown) and in agreement with previous studies (Fauchereau et al. 2009; Vigaud et al. 2012), we retain seven (five) classes for OLR (ZDEF).

3. TTT climatology

a. Frequency and propagation

Figure 2 shows the composite OLR anomalies for each OLR cluster, together with corresponding ZDEF anomalies. Classes 5, 6, and 7 (Figs. 2e–g) show the negative OLR anomalies oriented from northwest to southeast and linking the tropics (15°S) to the midlatitudes (45°S). They correspond to the typical signature of TTT events. Negative OLR anomalies (enhanced convection) are surrounded to the west and east by positive ones (indicative of reduced convection). Class 5 locates large-scale increased convective activity over southeastern South Africa while in classes 6 and 7 convection is shifted eastward, over the Agulhas Current and nearby Indian Ocean, materializing the longitudinal variations of TTTs depicted in previous works (Todd and Washington 1999; Washington and Todd 1999; Todd et al. 2004; Fauchereau et al. 2009; Pohl et al. 2009). These classes compare remarkably well with Fauchereau et al. (2009), who used NOAA daily OLR (Liebmann and Smith 1996).

Fig. 2.

Mean daily OLR anomalies (shaded; W m−2) and 200-hPa ZDEF anomalies [dashed (solid) contours corresponding to positive (negative) anomalies, starting at ±200 × 10−7 s−1 with a 10 × 10−7 s−1 interval], for the seven OLR regimes, over the period November–February from 1971 to 2000. Only 95% significant anomalies according to a t test are displayed.

Fig. 2.

Mean daily OLR anomalies (shaded; W m−2) and 200-hPa ZDEF anomalies [dashed (solid) contours corresponding to positive (negative) anomalies, starting at ±200 × 10−7 s−1 with a 10 × 10−7 s−1 interval], for the seven OLR regimes, over the period November–February from 1971 to 2000. Only 95% significant anomalies according to a t test are displayed.

Other classes (Figs. 2a–d) do not seem be related to tropical temperate interactions: classes 1, 2, and 3 (class 4) are solely related to tropical (temperate) processes with no significant anomalies recorded in the midlatitudes (tropics). Note that negative OLR anomalies in classes 1, 2, and 4 present also some northwest–southeast oriented structures. Fauchereau et al. (2009) indicate indeed that class 4 tends to occur just before a TTT event and could thus be considered as a possible precursor. In contrast, 10% of their TTT events were followed by class 2, which can thus be interpreted as a possible decaying phase of a TTT. ZDEF anomalies projected onto the three TTT classes show an alternation of positive and negative ZDEF anomalies located in the extratropics on both sides of the cloud band, suggesting that the longitudinal variability in the location of TTTs is also associated with a similar shift of the midlatitude transient perturbations. Largest negative (positive) ZDEF anomalies associated with the three TTT classes present the same northwest to southeast direction as OLR but are located east (west) compared to maximum convection. In agreement with previous studies (Widlansky 2010; Widlansky et al. 2011; Vigaud et al. 2012), this suggests a consistent longitudinal lead between ZDEF negative anomalies and convection in subtropical regions. In other classes (classes 1–4), ZDEF anomalies are weaker than during TTT events. Moreover, there is no continuous band of negative and positive ZDEF anomalies collocated with negative OLR anomalies.

Figure 3 details the longitudinal profiles of OLR, ZDEF, and convergence anomalies averaged between 24° and 36°S, for the three TTT classes 5, 6, and 7. These latitudes correspond to the largest convective and dynamical anomalies associated with TTTs (Figs. 2e–g). Negative OLR anomalies are located between 20° and 45°E, between 30° and 60°E, and between 50° and 70°E, respectively, illustrating different locations of convection (Figs. 2e–g). They are out of phase with mass convergence anomalies in the upper atmosphere (200 hPa). Regions of enhanced (reduced) convection exhibit positive convergence (divergence) anomalies at 200 hPa and vice versa at 850 hPa (not shown), indicating a consistent baroclinic structure favoring deep convection. When comparing variations of OLR and ZDEF anomalies, the indices vary in phase and minimum (maximum) ZDEF values are found east (west) of the cloud band, in agreement with Figs. 2e–g. In addition, Fig. 3 indicates that the cloud band corresponds spatially to the change of sign (from positive to negative) of ZDEF. More importantly, it also clearly establishes that the structure of the cloud band remains basically unchanged from one class to another, in terms of magnitude and relative phase relationships between OLR, ZDEF, and mass convergence anomalies, except for their absolute longitudinal location.

Fig. 3.

Longitudinal profiles of OLR (blue), 200-hPa ZDEF (blue dots), and wind convergence at 200 hPa (red) anomalies for the classes (a) 5, (b) 6, and (c) 7, averaged over the 36°–24°S latitudinal band. Blue (orange) shadings denote regions of enhanced (suppressed) convection.

Fig. 3.

Longitudinal profiles of OLR (blue), 200-hPa ZDEF (blue dots), and wind convergence at 200 hPa (red) anomalies for the classes (a) 5, (b) 6, and (c) 7, averaged over the 36°–24°S latitudinal band. Blue (orange) shadings denote regions of enhanced (suppressed) convection.

Figure 4a presents the 1971–99 summer daily calendar of ERA-40 TTT class occurrences. TTTs being synoptic systems, sequences lasting less than 2 days have been removed. Remaining sequences (Fig. 4a) typically last about 3–5 days, with an average and median duration of 4.9 and 3 days. These results indicate that the average is strongly influenced by outliers (6% of the sequences last more than 8 days): the statistical distribution of TTT persistence is thus very skewed. According to our k-means analysis, TTTs represent about 29% of the days (999 days in total over 1971–99) of the austral summer season. Class 5 is larger (391 days) than classes 6 (319 days) and 7 (289 days). A large majority of TTT events (69%; 140 out of 204 days) are initiated with class 5, move in class 6, and finish in class 7, in agreement with their average eastward propagation (Fauchereau et al. 2009; Pohl et al. 2009; Hart et al. 2012). About 31% (69%) of sequences are stationary (propagative). Only five events propagate westward (i.e., sequences start with class 6 and terminate in class 5). The number of seasonal occurrences (events) of each TTT class during the 1971–2000 period is shown in Fig. 4b (Fig. 4c). On average, one records 34 days of TTTs (Fig. 4b), representing about 7 events season−1 (Fig. 4c). This is much more than Ratna et al. (2012), who find 1.83 events season−1 but restrict their work to strong events located over the African continent only. The number of days and events varies greatly from one year to another, highlighting their marked interannual component. Fauchereau et al. (2009) and Ratna et al. (2012) showed indeed that there is a clear increase in the number of days of class 5 (class 6) during La Niña events (the positive phase of subtropical Indian Ocean dipole), whereas class 7 occurrences do not seem to be significantly related to well-identified modes of large-scale climate variability.

Fig. 4.

(a) TTT class occurrences in ERA-40, over the period November–February from 1971 to 2000. (b) Mean number of TTT class occurrences per austral summer (November–February) season. (c) Mean number of TTT sequences (i.e., events) per season.

Fig. 4.

(a) TTT class occurrences in ERA-40, over the period November–February from 1971 to 2000. (b) Mean number of TTT class occurrences per austral summer (November–February) season. (c) Mean number of TTT sequences (i.e., events) per season.

From analysis of Figs. 3 and 4, one can therefore state that (i) classes 5–7 recurrently tend to succeed each other in time (Fig. 4a) and (ii) they mostly differ by their longitude but their intrinsic properties and structures are basically unchanged (Fig. 3). These two results taken together allow concluding that they can be interpreted as an average eastward propagation of TTT structures, concerning both their convective and dynamical component and without any modification of their relative phasing.

b. Contribution to South African rainfall

Figures 5 and 6 display the contribution of TTTs to seasonal rainfall in South Africa. During occurrences of class 5, significant positive (negative) rainfall anomalies take place in eastern (western) South Africa (Fig. 5a). Classes 6 and 7 are associated with dry anomalies over the whole country (Figs. 5b,c) and enhanced convection being located farther east (Figs. 2f,g). Figure 6 examines how the class-5 TTTs contribute to seasonal rainfall over South Africa. Spatially (Fig. 6a), their contribution increases from west (20%) to east (40%), in agreement with Harrison (1986) and Hart et al. (2012). These results are fairly consistent with the literature, suggesting that the methodology used to track TTTs is robust and physically consistent. Temporally (Fig. 6b), the correlation between the seasonal frequency of class 5 and summertime rainfall is positive and significant (r = 0.42). On average, class 5 TTTs account for about 19% of seasonal rainfall over the whole country but their contribution varies greatly from one year to another (ranging from 1% in 1981 to 45.5% in 1999). The years corresponding to the five largest contributions of class 5 TTTs are 1971, 1977, 1987, 1994, and 1999. These years correspond to (i) maximum number of occurrences of class 5 during the period (21, 23, 31, 22, and 33, respectively) and (ii) above normal seasonal rainfall but not the wettest years in South Africa. While the number of occurrences of class 5 is related to La Niña events (Fauchereau et al. 2009), a correlation between the contribution of class 5 TTTs and the seasonal mean multivariate ENSO index (Wolter and Timlin 1993; not shown) shows no clear relationship, indicating that the contribution of continental TTTs to South African rainfall is not clearly ENSO dependent.

Fig. 5.

Station rainfall anomalies associated with OLR classes (a) 5, (b) 6, and (c) 7. Only 95% significant anomalies according to a t test are displayed.

Fig. 5.

Station rainfall anomalies associated with OLR classes (a) 5, (b) 6, and (c) 7. Only 95% significant anomalies according to a t test are displayed.

Fig. 6.

(a) The 1971–2000 mean contribution of class 5 to South Africa rainfall. (b) Time series of the contribution for each season over South Africa. The black line shows the long-term mean, and the correlation with seasonal rainfall is labeled.

Fig. 6.

(a) The 1971–2000 mean contribution of class 5 to South Africa rainfall. (b) Time series of the contribution for each season over South Africa. The black line shows the long-term mean, and the correlation with seasonal rainfall is labeled.

4. Mechanisms for tropical temperate interactions

a. Association with midlatitude transient perturbations

Vigaud et al. (2012) used the Weather Research and Forecasting Model (WRF) spectral nudging capabilities above the planetary boundary layer over southern Africa, which resulted in prescribing midlatitude transients in phase within their ensemble simulation. They showed their influence on TTT development and propagation, and qualified midlatitude transient perturbations as a necessary but not sufficient condition for the formation of TTT systems. The aim of this section is to pursue their analyses by investigating in details the relationship between midlatitude transient perturbations (i.e., atmospheric Rossby waves) and convective activity over southern Africa and the nearby Indian Ocean. To analyze the specific influence of midlatitude perturbations on TTT development, we apply k-means clustering onto austral summer (November–February) ERA-40 daily ZDEF anomalies. The number of clusters retained here is k = 5 (see section 2c) and corresponding classes are referred to as regimes A, B, C, D, or E. Their anomaly patterns are shown in Fig. 7.

Fig. 7.

(a)–(e) Mean daily ZDEF anomalies (shaded; s−1) and negative OLR anomalies (contours; starting at −15 W m−2 with a 2 W m−2 interval) for the five 200-hPa ZDEF regimes A–E, over the period November–February from 1971 to 2000. Only 95% significant anomalies according to a t test are displayed.

Fig. 7.

(a)–(e) Mean daily ZDEF anomalies (shaded; s−1) and negative OLR anomalies (contours; starting at −15 W m−2 with a 2 W m−2 interval) for the five 200-hPa ZDEF regimes A–E, over the period November–February from 1971 to 2000. Only 95% significant anomalies according to a t test are displayed.

From west to east, all regimes are characterized by an alternation of positive and negative ZDEF anomalies, corresponding to an increase and a decrease of the zonal wind velocity toward the east, respectively. The largest amplitudes are mostly located south of 30°S and the patterns are oriented northwest to southeast, as for TTT convection. Basically, the regimes show the same spatial structure shifted in longitude, representing the northeastward propagation of transients with time (Vigaud et al. 2012). These recurrent regimes are accompanied by deep convection, located northwest of negative ZDEF anomalies. Figure 7 thus gives a picture symmetrical to Fig. 2. The joint analysis of both classifications will help separating tropical (OLR classes) and temperate (ZDEF regimes) influences on TTT formation.

Concomitance between ZDEF regimes and OLR classes is given in Table 1. While the chi-square test is useful to assess the overall significance between ZDEF regimes and OLR classes, Neu’s test (Neu et al. 1974) allows identifying more precisely which OLR classes and ZDEF regimes are significantly associated (i.e., more or less frequently than the average).

Table 1.

Contingency table between regimes of ZDEF and classes of OLR. Italic values are overrepresented, and the other values are underrepresented.

Contingency table between regimes of ZDEF and classes of OLR. Italic values are overrepresented, and the other values are underrepresented.
Contingency table between regimes of ZDEF and classes of OLR. Italic values are overrepresented, and the other values are underrepresented.

Midlatitude transient perturbations, except those associated with regime E, are often associated with the three classes corresponding to TTTs (classes 5, 6, and 7). TTT events developing over southern Africa (i.e., class 5; Fig. 2e) co-occur preferentially with regimes A and B (midlatitude transients located south of Madagascar; Figs. 7a,b). Similarly, class 6 (class 7) exhibits maximum simultaneous occurrences with regimes B and C (regimes C and D). These statistical relationships between regimes of ZDEF and classes of OLR confirm that TTT events are closely related to midlatitude transient perturbations and that their eastward propagation is in phase with that of midlatitude transients, hereby corroborating Vigaud et al. (2012) and Fig. 3.

Table 1 also shows that roughly 41% of midlatitude perturbations are associated with TTT events. This therefore implies that some transients are related to tropical temperate interactions while others are not. In the following section, we attempt to identify what differentiates these two categories and to point out which “phenomenon” is needed in addition to a midlatitude transient to cause a TTT. Possible candidates include intrinsic properties of the temperate perturbations (associated atmospheric instability, intensity, or location) or additional conditions in the tropics (favoring for instance the initiation and/or development of convective activity).

b. Differentiating temperate and tropical temperate perturbations

In this section and for brevity, we only consider all days of regime A (Fig. 7a) and constitute two distinct samples: those related (349 days) and those not related to continental TTTs (class 5: 170 days; Fig. 2e). Please note, however, that the results described here can be extended to the two other OLR classes interpreted as TTT events (classes 6 and 7) and to other ZDEF regimes B–D. Analyses are based on lagged composite maps [wind vertical velocity at 500 hPa (ω500), MSE, ZDEF, and moisture fluxes] 3 days before and 4 days after the initiation of the corresponding TTT events. MSE and moisture fluxes were integrated between the surface and 700 hPa.

Figure 8 shows lagged composites of ω500 for regime A transient perturbations that are not associated with a TTT, those associated with a TTT, and the differences between these two categories. Maps are shown on a larger domain in order to identify possible larger-scale configurations. In the midlatitudes, anomaly patterns are indeed reminiscent of a wave train, with a wavelength of roughly 50°–70° in longitude. This is coherent with the typical wavelength of an atmospheric Rossby wave. Eastward propagations show a phase speed of roughly 5° day−1, a result similar to Renwick and Revell (1999) over the South Pacific. All perturbations are associated with significant anomalies of uplift over southern Africa and nearby oceans from 2 to 1 day prior the occurrences of regime A. Even in the case of a purely temperate wave disturbance, associated uplift reaches the tropics (15°–20°S) over the African continent. This suggests that midlatitude transients recurrently favor atmospheric instability over tropical southern Africa.

Fig. 8.

(left) Lagged ω500 (Pa s−1) anomalies associated with the occurrences of ZDEF regime A that are not associated with a TTT event (as inferred by OLR classes 5–7). Only 95% significant anomalies according to a t test are displayed. (center) As in (left), but for ZDEF regime A occurrences associated with a TTT event. (right) The ω500 differences between ZDEF regime A occurrences associated with a TTT event and others. Only 95% significant differences according to a two-tailed t test are displayed.

Fig. 8.

(left) Lagged ω500 (Pa s−1) anomalies associated with the occurrences of ZDEF regime A that are not associated with a TTT event (as inferred by OLR classes 5–7). Only 95% significant anomalies according to a t test are displayed. (center) As in (left), but for ZDEF regime A occurrences associated with a TTT event. (right) The ω500 differences between ZDEF regime A occurrences associated with a TTT event and others. Only 95% significant differences according to a two-tailed t test are displayed.

The main differences between the two categories of midlatitude waves mostly consist of larger ω500 anomalies (either positive or negative) for transients associated with a TTT event, thus forming a stronger wave pattern, constituted by an alternation of ascending and subsiding anomalies. This is coherent with the average structure of a typical TTT, with suppressed convection on the east and west of the cloud band (Fig. 2). This result also shows that the location of ω500 anomalies is basically unchanged between the two categories of midlatitude waves, suggesting that the location of the Rossby wave perturbation is not discriminating. It is worth noting that ω500 anomalies are not only stronger over southern Africa during the TTT event but also over the Atlantic sector, 1–3 days prior to the initiation of associated deep convection over Africa. This is more obvious for the subsiding anomalies occurring over the southern Atlantic Ocean between day −3 and day −1. This suggests that Rossby waves leading to TTT systems are already stronger a few days before their formation. Also important is the short-lived stationary enhancement of atmospheric convection over the subtropical southern Atlantic (northeast of the South Atlantic convergence zone) 3 days prior to the development of the TTT over southern Africa (Fig. 8, right).

Figure 9 extends these analyses to the regional differences of ZDEF (Fig. 9a) and MSE (Fig. 9b). From 3 to 2 days before the TTT (days −3 and −2), atmospheric convection gradually intensifies southwest of South Africa (Fig. 9a), while it slowly decreases over the subtropical Atlantic farther west (Fig. 8). This regional pattern is reminiscent of the typical class 1 convective anomaly pattern (Fig. 2a), suggesting that this class could be considered as a precursor for continental TTT events. Indeed, more than 15% of the days ascribed to class 1 are followed by class 5. The largest positive ZDEF differences are located over the southern Atlantic Ocean, corroborating the fact that midlatitude troughs prefiguring continental TTT systems tend to be of larger amplitude. At the same time, significant positive MSE differences, corresponding to an increase in the energy content in the lower layers of the atmosphere, develop above the Agulhas Current, off the Eastern Cape Province coastline (Fig. 9b).

Fig. 9.

(a) Differences of 200-hPa ZDEF (shaded; s−1) and OLR (contours; 5 W m−2; only negative differences are represented) between occurrences of ZDEF regime A associated with a TTT and others. Only 95% significant differences according to a two-tailed t test are displayed. The thick dashed box shows the ZDEF regional index. (b) As in (a), but for MSE integrated between 1000 and 700 hPa (shaded; 102 J kg−1). The thick dashed box shows the MSE regional index.

Fig. 9.

(a) Differences of 200-hPa ZDEF (shaded; s−1) and OLR (contours; 5 W m−2; only negative differences are represented) between occurrences of ZDEF regime A associated with a TTT and others. Only 95% significant differences according to a two-tailed t test are displayed. The thick dashed box shows the ZDEF regional index. (b) As in (a), but for MSE integrated between 1000 and 700 hPa (shaded; 102 J kg−1). The thick dashed box shows the MSE regional index.

At day −1, the midlatitude trough is strengthened and shifted northeast over the southern Atlantic Ocean (Figs. 8 and 9a). Convection and MSE increase and propagate southeastward over the continent and offshore over the Agulhas Current system, respectively. MSE is located southeast of the deep convective activity, as inferred from OLR differences, indicating a lead of MSE on convection. Taken together, these results suggest that MSE provides instability to the lower layers, while in the upper layers midlatitude synoptic activity also favors uplift motion in the free atmosphere. This forms a highly favorable environment for the development of convection.

Complementary analyses show that the excess of energy above Agulhas Current region mostly corresponds to increased latent energy (not shown). This moisture may be because of local evapotranspiration or from a remote source, through advections. During the formation of TTT events, latent heat fluxes at the surface are not significantly increased (not shown), suggesting that it could be mostly advected from remote regions.

Moisture flux differences between transients forming a TTT and others are shown in Fig. 10. Like for Fig. 8, maps are presented on an enlarged domain in order to identify their possible origins. Three days before a TTT event, westerly moisture fluxes form over the tropical southeast Atlantic Ocean at 10°S, continuously between South America and southern Africa. They penetrate over the subcontinent over Angola and Namibia. At day −2 a cyclonic vortex develops there, deviating moisture fluxes southward (D’Abreton and Tyson 1995; Cook et al. 2004; Hermes and Reason 2009; Vigaud et al. 2007) and reminiscent of semipermanent Angola low (Rouault et al. 2003; Reason and Jagadheesha 2005; Reason et al. 2006; Fig. 1). During the following day (day −1), continental northwesterly moisture fluxes are reinforced and converge over eastern South Africa and Mozambique with northeasterly fluxes originating from the Mozambique Channel.

Fig. 10.

As in Fig. 9, but for moisture fluxes integrated between 1000 and 700 hPa (vectors; g kg−1 m s−1). Cold (hot) colors show negative (positive) OLR differences (W m−2). Only 95% significant differences according to a two-tailed t test are displayed.

Fig. 10.

As in Fig. 9, but for moisture fluxes integrated between 1000 and 700 hPa (vectors; g kg−1 m s−1). Cold (hot) colors show negative (positive) OLR differences (W m−2). Only 95% significant differences according to a two-tailed t test are displayed.

At day 0 and by construction, convection is organized at large-scale, thereby forming the northwest–southeast oriented band linking the tropics over southern Africa (15°S) and the midlatitude over the Indian Ocean (45°S), typically representative of TTT systems. Figure 9a (Fig. 9b) shows significant positive MSE (ZDEF) differences located directly to the east (west) of the TTT, presenting the same direction as the cloud band and vertical velocity differences (Fig. 8). A strong cyclonic circulation takes place over southern Africa while the northerly moisture fluxes over Mozambique Channel develop northward, reinforcing and shifting the moisture convergence differences northeastward (Fig. 10). Important poleward transport of energy and momentum occur in the lee of the location of the cloud band, in fair agreement with the patterns discussed by Todd and Washington (1999), Todd et al. (2004), and Ratna et al. (2012).

During the following days, cloud band, ZDEF, ω500, and MSE differences continue to propagate eastward and weaken. The cyclonic circulations associated with TTT events over southern Africa remain located over the subcontinent and gradually weaken between days 0 and +2 (Fig. 10). MSE (ZDEF) differences vanish completely over the Mozambique Channel at day +4 (day +3) (Fig. 9b).

These results suggest that, on average, TTT events are associated with (i) intense midlatitude troughs; (ii) moisture convergence over southern Africa, originating from the Atlantic basin on the west and the southwest Indian Ocean and Mozambique Chanel on the east; and—in agreement with (ii)—(iii) an excess of moist static energy (and more particularly lower-layer humidity) over the Agulhas Current region. Two main questions arise now. (i) To what extent can the amplitude of the temperate Rossby wave and the lower-layer moisture convergence and air humidity above southern Africa discriminate the transient perturbations forming TTT events and others? (ii) Can the results raised above, based on composite analyses including a large number of events, be verified for each TTT recorded during the period?

c. Necessary versus sufficient conditions

To assess the usefulness of the results presented above to explain the formation of TTTs, ZDEF and MSE indices are computed for each occurrence of regime A associated with (i) a TTT event and (ii) others (i.e., not related to a TTT). To that end we averaged ZDEF and MSE anomalies south of the tip of the South Africa (37°–44°S, 24°–33°E; Fig. 9a) and south of the Mozambique Channel (27°–33°S, 34°–45°E; Fig. 9b), respectively. These areas correspond to the maximum positive ZDEF and MSE differences between midlatitude transient perturbations associated and those not associated with TTTs (Fig. 9). Figure 11 presents the scatterplot intersecting the corresponding ZDEF and MSE regional indices.

Fig. 11.

Scatterplot intersecting the ZDEF and MSE regional indices for each day of ZDEF regime A. Red (blue) crosses correspond to occurrences associated (not associated) with TTT events. Red (blue) dot represents the centroid of the corresponding sample.

Fig. 11.

Scatterplot intersecting the ZDEF and MSE regional indices for each day of ZDEF regime A. Red (blue) crosses correspond to occurrences associated (not associated) with TTT events. Red (blue) dot represents the centroid of the corresponding sample.

On average, the Rossby waves associated with TTTs have larger ZDEF (indicative of a more intense perturbation: Fig. 8) and lower-layer MSE than others. A total of 100% of the midlatitude transients associated with tropical convection display positive ZDEF anomalies. Four (out of 41 events) are characterized by negative MSE anomalies. This seems to indicate that our previous results are valid for a large majority of TTT events.

However, some of midlatitude transients not related to TTTs also have strong positive ZDEF and MSE anomalies. This suggests a priori that we identified once again some highly favorable conditions/mechanisms but that do not lead systematically to TTT events. To understand why these cases of strong ZDEF and MSE do not lead to TTT developments, their OLR and moisture flux anomalies are plotted individually (not shown). It appears that, in these cases, maximum convection takes place farther west. Of these disturbances, 67% are indeed ascribed to OLR classes 1 (Fig. 2a; convection over western southern Africa) and 4 (Fig. 2d; convection in southwest southern Africa), followed a few days later by a TTT. It was indeed noted previously that these classes could be considered as possible precursors for TTT events (section 3a; Fauchereau et al. 2009). The robustness of the mechanism identified here, involving both a strong Rossby wave perturbation in the midlatitudes and MSE anomalies over the Agulhas Current region mostly caused by moisture advections from nearby oceans, is thus probably greater than what Fig. 11 may suggest at first view.

In the end, (i) midlatitude Rossby waves are confirmed to be a necessary condition for TTTs, in agreement with Vigaud et al. (2012); however, TTT systems are associated on average with waves that are abnormally strong and favor large-scale atmospheric instability in the free atmosphere over subtropical and tropical southern Africa. (ii) Positive latent energy anomalies over the Agulhas Current system supply lower-layer moisture, which, associated with the large-scale instability associated with the Rossby wave, form favorable conditions for the development of atmospheric convection. Such latent energy seems to have a remote origin and is partly advected over the region by convergent lower-layer moisture fluxes originating from the southern Atlantic basin and the Mozambique Channel. Such regional convergence seems to be favored by a transient reinforcement of the so-called Angola low (Reason and Jagadheesha 2005) that leads the TTT event by at least 3 days. Over the Atlantic sector TTT development seems to be embedded in large-scale circulation anomaly patterns, involving continuous moisture fluxes between South America and Africa, as well as a transient enhancement of atmospheric convection over the subtropical southern Atlantic, occurring three to two days prior to the TTT development over southern Africa. Similar precursors, potentially useful for operational TTT forecasting, could not be found over the Indian Ocean sector: moisture fluxes there develop concomitantly with the TTT itself.

5. Conclusions and discussion

This study attempts to identify favorable conditions leading to the genesis and development of tropical temperate troughs (TTT) over southern Africa and the southwest Indian Ocean during the austral summer season (November–February). In agreement with previous work, TTTs appear to be systematically associated with a midlatitude transient perturbation, interpretable as an atmospheric Rossby wave. The reverse is not true, only 41% of the Rossby waves occurring in summer being associated with deep convective activity linking the tropics and the midlatitudes. Thus, temperate perturbations can be seen as a necessary condition for TTT development, but are not sufficient.

Based on these two categories, we attempted to identify the “phenomena” that differentiate the midlatitude transients that are and those that are not associated with TTTs. A first result shows that the Rossby waves associated with TTTs are on average stronger than others. This result is verified during the TTT occurrence but also 3 days prior to the initiation of its convection. Both categories of Rossby waves seem however to have similar longitudinal and latitudinal locations and are all likely to favor atmospheric instability over southern Africa as far as 15°S in the tropics.

In the lower latitudes, conditions favoring lower-layer atmospheric instability over tropical southern Africa appear as another intuitive candidate that could add its own influence to midlatitude synoptic-scale variability. Results identify a few features in the tropical atmosphere that appear to be of primary importance for TTT genesis: some of them were already mentioned in the literature but their precise influence remained to be quantified, which we attempted to achieve in this work. Statistically, the development of a TTT is accompanied by strong positive moist static energy (MSE) anomalies over the Agulhas Current system, off the South African east coasts. Among the components of MSE (viz., potential, sensible, and latent energy), the latter appears as the main contributor. This excess in lower-layer air humidity leads TTT-associated convection in time and space; it develops and strengthens about 2 days prior to the maximum convection and rainfall. Instead of a solely local origin because of enhanced evaporation over the Agulhas Current system, results suggest that air humidity is partly supplied by advections from remote regions, with this transport being of course likely to combine its effects with local processes. In agreement with the literature, a TTT coincides with moisture convergence over southern Africa, due on the one hand to a reinforcement of the so-called Angola low that favors flux penetration from the Atlantic basin toward southern Africa on the west and on the other hand to easterly moisture fluxes from the nearby Indian Ocean and Mozambique Channel on the east. Fluxes are then forced to deviate southward, hereby causing the poleward export of momentum reported in previous works (Todd and Washington 1999; Todd et al. 2004; Fauchereau et al. 2009). Another promising result is that, unlike their Indian Ocean counterparts, Atlantic moisture fluxes seem to be embedded in a large-scale circulation pattern linking South America to southern Africa from about 3 to 2 days prior to the development of the TTT. This could be important for their real-time prediction.

Considering these elements together helps understanding how a temperate baroclinic instability is capable to generate (or is associated with) tropical convection as far as 15°S, with tropical moisture fluxes and convergence being equally important to destabilize the air mass and sustain convective activity over southern Africa and nearby oceans. Although this work helped understanding how tropical and temperate variability combine their effects to favor the development of convective activity from the tropics to the midlatitudes, it partly failed at identifying unequivocally some sufficient conditions leading to TTT development. This attempt could however be doomed by the nonlinear, stochastic behavior of the atmosphere and more particularly of deep moist atmospheric convection. Other components of the regional climate system could also be involved (including, for instance, more directly the continent–atmosphere and ocean–atmosphere couplings).

In the future, we plan to assess the potential predictability of each of the components identified here to be relevant for TTT formation. Atmospheric instability and moisture fluxes over the southern Atlantic sector seem to constitute the most robust potential precursors. Although it was previously noted (Crétat and Pohl 2012) that the location of TTT-associated maximum rainfall is subject to huge uncertainties, once again, because of the chaotic component of the climate system, more studies are needed to eventually improve their operational prediction and anticipate their effects on the societies and environment.

Acknowledgments

This work is a contribution to the LEFE/IDAO VOASSI program funded by CNRS. The authors thank two anonymous reviewers for their constructive comments, as well as Jérôme Vialard, Claude Kergomard, and Mathieu Rouault for helpful discussions. ERA-40 data used in this study have been freely obtained from the ECMWF data server. Calculations were performed using high-performance computing (HPC) resources from DSI-CCUB, Université de Bourgogne.

REFERENCES

REFERENCES
Behera
,
S. K.
, and
T.
Yamagata
,
2001
:
Subtropical SST dipole events in the southern Indian Ocean
.
Geophys. Res. Lett.
,
28
,
327
330
.
Cheng
,
X.
, and
J. M.
Wallace
,
1993
:
Regime analysis of the Northern Hemisphere wintertime 500-hPa height field: Spatial patterns
.
J. Atmos. Sci.
,
50
,
2674
2696
.
Chikoore
,
H.
, and
M. R.
Jury
,
2010
:
Intraseasonal variability of satellite derived rainfall and vegetation over southern Africa
.
Earth Interact.
,
14
,
doi:10.1175/2010EI267.1
.
Cook
,
C.
,
C. J. C.
Reason
, and
B. C.
Hewitson
,
2004
:
Wet and dry spells with a particularly wet and dry summers in the South African summer rainfall region
.
Climate Res.
,
26
,
17
31
.
Cook
,
K. H.
,
2000
:
The south Indian convergence zone and interannual rainfall variability over southern Africa
.
J. Climate
,
13
,
3789
3804
.
Crétat
,
J.
, and
B.
Pohl
,
2012
:
How physical parametrizations can modulate internal variability in a regional climate model
.
J. Atmos. Sci.
,
69
,
714
724
.
Crimp
,
S. J.
,
J. R. E.
Lutjeharms
, and
S. J.
Mason
,
1998
:
Sensitivity of a tropical-temperate trough to sea-surface temperature anomalies in the Agulhas retroflection region
.
Water SA
,
24
,
93
101
.
D’Abreton
,
P. C.
, and
J. A.
Lindesay
,
1993
:
Water vapour transport over southern Africa during wet and dry early and late summer months
.
Int. J. Climatol.
,
13
,
151
170
.
D’Abreton
,
P. C.
, and
P. D.
Tyson
,
1995
:
Divergent and non-divergent water vapour transport over southern Africa during wet and dry conditions
.
Meteor. Atmos. Phys.
,
55
,
47
59
.
Fauchereau
,
N.
,
B.
Pohl
,
C.
Reason
,
M.
Rouault
, and
Y.
Richard
,
2009
:
Recurrent daily OLR patterns in the southern Africa/southwest Indian Ocean region, implications for South African rainfall and teleconnections
.
Climate Dyn.
,
32
,
575
591
.
Harangozo
,
S.
, and
M. S. J.
Harrison
,
1983
:
On the use of synoptic data indicating the presence of cloud bands over southern Africa
.
S. Afr. J. Sci.
,
79
,
413
414
.
Harrison
,
M. S. J.
,
1984
:
A generalized classification of South African summer rain-bearing synoptic systems
.
Int. J. Climatol.
,
4
,
547
560
.
Harrison
,
M. S. J.
,
1986
: A synoptic climatology of South African rainfall variations. Ph.D. dissertation, University of Witwatersrand, 341 pp.
Hart
,
N. C. G.
,
C. J. C.
Reason
, and
N.
Fauchereau
,
2010
:
Tropical–extratropical interactions over southern Africa: Three cases of heavy summer season rainfall
.
Mon. Wea. Rev.
,
138
,
2608
2623
.
Hart
,
N. C. G.
,
C. J. C.
Reason
, and
N.
Fauchereau
,
2012
:
Cloud bands over southern Africa: Seasonality, contribution to rainfall variability and modulation by the MJO
.
Climate Dyn.
,
41
,
1199
1212
,
doi:10.1007/s00382-012-1589-4
.
Hermes
,
J. C.
, and
C. J. C.
Reason
,
2009
:
Variability in sea-surface temperature and winds in the tropical south-east Atlantic Ocean and regional rainfall relationships
.
Int. J. Climatol.
,
29
,
11
21
.
Holton
,
J. R.
,
1992
:
An Introduction to Dynamic Meteorology.
Academic Press, 511 pp.
Liebmann
,
B.
, and
C. A.
Smith
,
1996
:
Description of a complete (interpolated) outgoing longwave radiation dataset
.
Bull. Amer. Meteor. Soc.
,
77
,
1275
1277
.
Lynch
,
S.
,
2003
: Development of a RASTER database of annual, monthly and daily rainfall for southern Africa. WRC Rep. 1156/1/03, 78 pp.
Lyons
,
S. W.
,
1991
:
Origins of convective variability over equatorial southern Africa during austral summer
.
J. Climate
,
4
,
23
39
.
Manhique
,
A. J.
,
C. J. C.
Reason
,
L.
Rydberg
, and
N.
Fauchereau
,
2011
:
ENSO and Indian sea surface temperatures with tropical temperate troughs over Mozambique and the southwest Indian Ocean
.
Int. J. Climatol.
,
31
,
1
13
.
Michelangeli
,
P.
,
R.
Vautard
, and
B.
Legras
,
1995
:
Weather regime occurrence and quasi-stationarity
.
J. Atmos. Sci.
,
52
,
1237
1256
.
Neu
,
C. W.
,
C. R.
Byers
, and
J. M.
Peek
,
1974
:
A technique for analysis utilization-availability data
.
J. Wildl. Manage.
,
38
,
541
545
.
Pohl
,
B.
,
Y.
Richard
, and
N.
Fauchereau
,
2007
:
Influence of the Madden–Julian oscillation on southern African summer rainfall
.
J. Climate
,
20
,
4227
4242
.
Pohl
,
B.
,
N.
Fauchereau
,
Y.
Richard
,
M.
Rouault
, and
C. J. C.
Reason
,
2009
:
Interactions between synoptic, intraseasonal, and interannual convective variability over southern Africa
.
J. Climate
,
33
,
1033
1050
.
Racz
,
Z.
, and
R. K.
Smith
,
1999
:
The dynamics of heat lows
.
Quart. J. Roy. Meteor. Soc.
,
125
,
225
252
.
Ratna
,
S. B.
,
S.
Behera
,
J. V.
Ratnam
,
K.
Takahasgi
, and
T.
Yamagata
,
2012
:
An index for tropical temperate troughs over southern Africa
.
Climate Dyn.
, 41, 421–441,
doi:10.1007/s00382-012-1540-8
.
Reason
,
C. J. C.
,
2001
:
Subtropical Indian Ocean SST dipole events and southern African rainfall
.
Geophys. Res. Lett.
,
28
,
2225
2227
.
Reason
,
C. J. C.
, and
D.
Jagadheesha
,
2005
:
A model investigation of recent ENSO impacts over southern Africa
.
Meteor. Atmos. Phys.
,
89
,
181
205
.
Reason
,
C. J. C.
,
W. A.
Landman
, and
W.
Tennant
,
2006
:
Seasonal to decadal prediction of the southern African climate and its links with the variability of the Atlantic Ocean
.
Bull. Amer. Meteor. Soc.
,
87
,
941
955
.
Renwick
,
J. A.
, and
M. J.
Revell
,
1999
:
Blocking over the South Pacific and Rossby wave propagation
.
Mon. Wea. Rev.
,
127
,
2233
2247
.
Rouault
,
M.
,
P.
Florenchie
,
N.
Fauchereau
, and
C. J. C.
Reason
,
2003
:
South east Atlantic warm events and southern African rainfall
.
Geophys. Res. Lett.
,
30
,
8009
,
doi:10.1029/2002GL014840
.
Todd
,
M.
, and
R.
Washington
,
1999
:
Circulation anomalies associated with tropical-temperate troughs in southern Africa and the south west Indian Ocean
.
Climate Dyn.
,
15
,
937
951
.
Todd
,
M.
,
R.
Washington
, and
P. I.
Palmer
,
2004
:
Water vapour transport associated with tropical-temperate trough systems over southern Africa and the southwest Indian Ocean
.
Int. J. Climatol.
,
24
,
555
568
.
Tyson
,
P. D.
, and
R. A.
Preston-White
,
2000
: The Weather and Climate of Southern Africa. Oxford University Press, 396 pp.
Uppala
,
S.
, and
Coauthors
,
2005
: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc.,131, 2961–3012.
Vigaud
,
N.
,
Y.
Richard
,
M.
Rouault
, and
N.
Fauchereau
,
2007
:
Water vapour transport from the tropical Atlantic and summer rainfall in tropical southern Africa
.
Climate Dyn.
,
28
,
113
123
.
Vigaud
,
N.
,
B.
Pohl
, and
J.
Crétat
,
2012
:
Tropical-temperate interactions over southern Africa simulated by a regional climate model
.
Climate Dyn.
, 39, 2895–2916,
doi:10.1007/s00382-012-1314-3
.
Washington
,
R.
, and
M.
Todd
,
1999
:
Tropical-temperate links in southern Africa and southwest Indian Ocean satellite-derived daily rainfall
.
Int. J. Climatol.
,
19
,
1601
1616
.
Widlansky
,
M.
,
2010
: Climate dynamics of the South Pacific convergence zone and similarities with other subtropical convergence zones in the Southern Hemisphere. Ph.D. dissertation, Georgia Institute of Technology, 154 pp.
Widlansky
,
M.
,
P. J.
Webster
, and
C. D.
Hoyos
,
2011
:
On the location and orientation of the South Pacific convergence zone
.
Climate Dyn.
,
36
,
561
578
,
doi:10.1007/s00382-010-0871-6
.
Williams
,
C. J. R.
,
D. R.
Kniveton
, and
R.
Layberry
,
2007
:
Climatic and oceanic associations with daily rainfall extremes over southern Africa
.
Int. J. Climatol.
,
27
,
93
108
.
Wolter
,
K.
, and
M. S.
Timlin
,
1993
: Monitoring ENSO in COADS with a seasonally adjusted principal component index. Proc. 17th Climate Diagnostics Workshop, Norman, OK, NOAA/NMC/CAC,
52
57
.