1. Introduction
Regional climate may be controlled by a number of factors, including remote forcing through teleconnection processes (Bjerknes 1969), regional forcing through sea surface temperatures of adjacent oceans (e.g., Mason 1995), and local forcing through feedback mechanisms with the land surface soil moisture (e.g., Douville et al. 2001) and vegetation (e.g., Zheng and Eltahir 1998). Each of these may have distinct signals that can either oppose or reinforce each other. As a result, climate in a given region may exhibit a complex spatial and temporal pattern in which the local response to large-scale forcing through teleconnections can become difficult to quantify.
Global teleconnections in climate have been studied for many years and these are well documented (e.g., Mo and White 1985; Trenberth et al. 1998). The most prominent examples include the El Niño–Southern Oscillation (ENSO; Yarnal 1985; Yarnal and Kiladis 1985; Ropelewski and Halpert 1987), and the North Atlantic Oscillation (Rogers 1984). The mechanics of these teleconnections are complex and often fluctuate over time, highlighting the importance of continued work in this field.
Evidence exists that the three main Southern Hemisphere convergence zones (viz., the South Atlantic convergence zone, the South Indian convergence zone, and the South Pacific convergence zone), are connected in some way (Todd and Washington 1999). This evidence makes a good case for the existence of teleconnections in the Southern Hemisphere. However, there is still not consensus about the physical mechanisms behind these teleconnections.
A number of ways in which ENSO signals are possibly communicated around the Southern Hemisphere are suggested in the literature. Cook (2000) describes two pathways for the ENSO signal to reach southern Africa. The first is an atmospheric mechanism where Rossby waves are generated by an adjustment of the equatorial Walker circulation and transmit the signal. The second is an atmosphere–ocean interaction where the atmospheric response to Pacific Ocean sea surface temperature (SST) anomalies causes Indian Ocean warming that in turn changes southern Africa rainfall through sensitivity to more local SSTs. Peterson and White (1998) suggested that changes in SST anomalies in the Pacific Ocean induce anomalies in the subtropical ocean through changes in the Hadley cell. These anomalies then propagate southward and around the Southern Hemisphere as the Antarctic Circumpolar Wave (ACW). This phenomenon has an 8-yr cycle but has only been apparent in the SST and sea level pressure fields during the 1982–94 period. Reason et al. (2000) propose that the ENSO signal is transported to the Indian Ocean at two distinct time scales, namely the quasi-biennial (2–2½ yr) and the low interannual frequency (2½–7 yr) bands.
In the southern Africa region, general circulation patterns and teleconnections associated with wet and dry years in the summer rainfall regions of South Africa are well documented (Mason and Tyson 1992; D’Abreton and Lindesay 1993; D’Abreton and Tyson 1995; Jury et al. 1996; Mason and Jury 1997; Reason and Mulenga 1999). However, these teleconnections only explain part of the climatic variability in the region. Local forcing is also evident from adjacent oceans, particularly the warm Agulhas Current, the strongest western boundary current in the Southern Hemisphere, and its inertial recirculation over the southwest Indian Ocean. This current has been the focus of several regional studies into its influence on summer rainfall along the southeast coast of South Africa (Jury et al. 1993; Mason 1995) and over the subcontinent as a whole (Reason 2001). Signals in the midlatitude eddies have also been shown to have a strong influence on the development of tropical-temperate troughs that are a major contributor to summer rainfall in southern Africa (Todd and Washington 1999) and in winter over Australia (Wright 1997). Such land-based convergence zones (Cook 2000) can in turn be affected by low-frequency signals as described by Reason et al. (2000).
Evidence of a coherent ENSO signal is hard to find in rainfall of the winter rainfall regions of South Africa [southwestern Cape (SWC)] and southwestern Western Australia (SWWA; Mason 1995; Smith et al. 2000; Reason et al. 2002). However, local SSTs do appear to be of significance; for example, the presence of the Leeuwin Current off the Western Australian coast and the resulting lack of cool upwelled SST there contributes toward this region being considerably wetter on average than the SWC (Gentilli 1991; Reason et al. 2002). GCM experiments (Reason 1998, 2001) suggest that winter SWC rainfall is sensitive to SST anomalies off the South African south coast. Notwithstanding this local SST forcing, there are also associations between regional rainfall in these regions and large-scale circulation including the circumpolar trough (Ansell et al. 2000) and the Antarctic Oscillation (Smith et al. 2000; Reason et al. 2002).
One way of improving our understanding of regional climate variability is to find associations between regional rainfall and the large-scale circulation. This may be achieved through a study of atmospheric energetics. Since the general circulation was explained in terms of the energy cycle by Lorenz (1955), various attempts to estimate the energy cycle have been made (e.g., Oort 1964; Holopainen 1970; Peixoto and Oort 1974). Wiin-Nielsen (1962) made a further contribution to the understanding of the energy cycle by noting that the conversion of potential energy into kinetic energy goes mainly into the baroclinic (shear) component. The shear component then helps maintain the barotropic (vertical mean) component.
Although a case may be made that atmospheric energetics are really only meaningful on a global scale, Huang and Vincent (1985) argued that they can still reveal a lot about atmospheric processes when applied to smaller regions. Eastin and Vincent (1998) also used this approach to investigate properties of the winter jet over Australia. This paper will also use the atmospheric energy cycle as a basis for understanding large-scale processes in the atmosphere. In addition, we use daily fields of the barotropic and baroclinic components of the kinetic energy to study the high-frequency atmospheric characteristics on a hemispheric scale and to explore possible links between the general circulation and regional rainfall. These kinetic energy fields have the added benefit of being related to physical atmospheric phenomena.
A further motivation for diagnostic studies into the link between large-scale variability and local climate is to assist with the development of extended-range prediction systems. In southern Africa, where variability is high and predictability is low, GCMs tend to reproduce the large-scale circulation better than local features such as rainfall (Tennant 2003). If these links are better understood then downscaling methods could be developed and applied more effectively to predict regional rainfall.
The objective of this paper then is to explore some of the prominent associations between the general circulation and December–January–February (DJF) rainfall in the summer rainfall regions over the central interior of South Africa and the June–July–August (JJA; hereafter all monthly periods will be designated by the first letter of each respective month) rainfall in the winter rainfall areas of southwestern South Africa and Western Australia. The southwestern Western Australian region is included to contrast large-scale processes and teleconnections that may influence winter rainfall on a near-hemispheric extent. The focus of these analyses will be on the atmospheric energy cycle and on daily circulation statistics.
2. Data and methodology
This paper is concerned with atmospheric energetics and later uses the self-organizing map (SOM) technique to analyze the daily variability. The formulation of the energetics set in this analysis is summarized below. We also give an overview of the SOM technique that is comparatively unknown in the meteorological community.
a. Energetics
b. Self-organizing maps
The SOM algorithm is a technique to reduce the degrees of freedom in a dataset by forming a predefined number of archetypes. These archetypes, or nodes, are identified by the SOM as dominant modes within the span of a dataset, such that the distribution of the nodes represent the observed distribution. One could describe the process as a nonlinear projection of the probability density function of high-dimensional input data onto a two-dimensional array of nodes. The SOM technique is different from other cluster techniques in that nodes with similar characteristics or patterns are placed near each other such that there is a smooth transition of archetypes across the SOM array.
The SOM algorithms were initially developed at the Helsinki University of Technology (software available online at http://www.cis.hut.fi/research/som-research) (Kohonen 1995) and are now used in a wide range of applications. A review of SOMs and the application to synoptic climatology is described in detail by Hewitson and Crane (2002). Further applications include extreme events (Cavazos 1999, 2000), defining seasons (Tennant and Hewitson 2002) and in GCM evaluation (Tennant 2003). Since the SOM methodology is relatively unknown, a brief description is given here.
The process of training a SOM begins with a random initialization of the reference spatial patterns of each node in the map. The number of nodes in the map is subjectively chosen according to the amount of generalization that is required. In this study a 12-node (4 × 3) SOM was chosen, allowing on average about 8 days season−1 yr−1 to populate each node. The next stage consists of a two-phase iterative training process where input data is presented and assigned to a best-matching node, that is, the node that has the least difference to the input data pattern. The spatial pattern on this node is adjusted toward the input pattern. This process is repeated for adjacent nodes such that they effectively span the variance structure of the data space. In the first phase, the best matching node and a neighborhood of nodes are adjusted toward the input pattern, and the spatial nodes converge to the dominant variance structure of the data, thus developing the broad mapping of the SOM. The second phase then develops the finer aspects of the SOM array, such that only nodes in the immediate vicinity of the best-matching node are updated.
Once the SOM is trained each input data sample, and any other forecast or testing data, can be associated with a best-matching node (archetype) as part of the algorithm output. The mapping coordinates for each data element may then be used to calculate frequencies of each archetype. These frequencies can be stratified according to various criteria (e.g., wet or dry periods) and contrasted with each other. In this study the interannual changes in these frequencies are of interest and we seek to contrast wet and dry years.
This whole process is now illustrated with a 4 × 3 SOM of DJF barotropic kinetic energy over the Southern Hemisphere. Twelve nodes were chosen to reduce the dimensionality of the data to a fair degree while still allowing a reasonable number of synoptic features to remain. The data were first converted to a Southern Hemisphere polar-stereographic projection to make the grid spacing more uniform across the domain. This was done to avoid the SOM training process placing too much emphasis on polar circulation features owing to a clustering of grid points that is characteristic of fixed latitude–longitude projections.
At a first glance SOM nodes may look similar as the time-averaged pattern often dominates these fields. For visualization purposes the SOM patterns can be plotted as departures from the overall time average. These departures can further be expressed as a percentage change of the time average to highlight significant changes in the patterns. Selected nodes from the 4 × 3 node SOM are shown in Fig. 1. The corner nodes represent the extremes of the variance structure in the data and show distinct modes of variability. The top-left and bottom-right corners represent a latitudinal shift in the barotropic kinetic energy in the Atlantic and Indian sectors of the Southern Ocean. The top-right and bottom-left corners represent a different mode of variability over the Australian region with a wavelike structure in the South Pacific Ocean. The actual number of days mapping to each node in a SOM varies from one application to another. However, in this study this mapping is fairly uniform across the SOM, not differing by more than a factor of 2. One characteristic is that the corner nodes represent the highest number of samples as these nodes attempt to incorporate the outer limits of the data variability.
Areas of strong node pattern anomalies are shown where these exceed one standard deviation of the interannual variability of the original data (Fig. 1). This is where the utility of SOMs is clear, as the algorithm can identify 12 different patterns in the input data. Since each input sample gets associated with a best-matching node, we have on average about 8% of the input data (roughly 172 days) being uniquely associated with each of the 12 nodes in this SOM. The variability of reasonably sized areas in each node (particularly toward the corners of the SOM) still exceeds the standard deviation of the interannual variability, demonstrating how the SOM can determine 12 different modes of variability common to this fairly large number of samples. Those nodes where the frequency appears related to wet and dry years in the different regions are used to determine the important modes of variability that may play a role in rainfall variability of these regions. This relation is determined by seeking those nodes that have maximum frequency increases (decreases) during dry (wet) years and vice versa.
c. Rainfall data
Regional rainfall characteristics derived using South African Weather Service daily station rainfall data have been shown to be important for determining links between regional rainfall and the general circulation (Tennant and Hewitson 2002). However, these differ from one region to another. Tennant and Hewitson (2002) found that wet and dry DJF seasons in the summer rainfall areas of South Africa are best defined as the highest and lowest quintile of sorted regional rainfall, respectively. The number of regional rain days exceeding 1 and 20 mm are also closely linked to regional season total rainfall (Tennant and Hewitson 2002). Therefore, in order to include all these rainfall characteristics in this analysis, a wet season is defined where the total rainfall for a given season lies in the highest quintile and the rain days exceeding 1 and 20 mm fall into either the fourth or fifth quintile. The dry seasons are defined analogously. The wet and dry seasons used in the analyses in this paper are listed in the quintile column in Table 1.
The pattern of JJA rainfall characteristics in the winter rainfall region does not suggest that quintiles provide as clear definition of wet and dry years as for the summer region. Notwithstanding, quintiles of JJA rainfall do largely agree with the definition of dry and wet categories using one standard deviation around the median in the standardized precipitation index (SPI) (Hayes et al. 1999; Table 1). In a previous study of this region (Reason et al. 2002), a longer winter rainfall season (MJJAS) was considered, but in order to keep the methodologies between the summer and winter rainfall areas consistent, this paper will consider the winter rainfall period to be JJA and use the quintile categorization.
Reason et al. (2002) found associations between SWC winter rainfall and hemispheric modes such as the Antarctic Oscillation. It seems logical to expect that these patterns may also affect other winter rainfall regions in the Southern Hemisphere and therefore we also consider the SWWA region, which is located at similar latitudes to the SWC and is also a predominantly winter rainfall region. To do so, gridded rainfall data from New et al. (2000) over southwestern Australia were used to define wet and dry seasons over that region. These were obtained analogously to the quintile method above. The wet and dry JJA seasons in SWWA are given in Table 2.
3. Results
The mechanisms and characteristics of rainfall in the summer and winter rainfall areas of South Africa are quite distinct and are discussed separately in this section. Rainfall in the former region results mostly from convection and the latter mostly from frontal action. We begin with the large-scale seasonal-averaged energy processes and then consider the characteristics of daily circulation statistics.
a. Summer rainfall region
The vertical structure of the large-scale meridional Hadley and Ferrel cells is shown using a two-dimensional streamfunction [see Eq. (7)]. Positive values denote circulation in a clockwise sense. On the hemispheric scale, the northern Hadley cell is stronger during dry years in central South Africa (Fig. 2a) than in wet years (Fig. 2b). The boundary between the southern Hadley and Ferrel cells is also closer to the equator during dry years. This pattern is also evident in the jet streams that are shifted equatorward during dry years as contrasted with wet years (Fig. 3a). These composite patterns exceed one standard deviation of the respective interannual variability and are consistent with a stronger Hadley cell transporting increased amounts of angular momentum to the subtropics and thereby increasing the kinetic energy of the jet and/or causing an equatorward latitudinal shift in its position.
Although strictly speaking the meridional cells are not defined over a longitudinal sector, in a qualitative sense some important features emerge for African longitudes (15°–35°E). During dry years, both the southern Hadley cell and the southern Ferrel cell strengthen considerably (Fig. 2c). Changes to the subtropical jet in the African sector (Fig. 3b) confirm this, with an increase in speed throughout the column at about 45°S while shrinking in latitudinal extent, shown by decreases at 60° and 30°S.
During wet years, the vertical boundary between the southern Hadley and Ferrel cells tilts and the Ferrel cell assumes a dual-core nature with the northern core positioned around 35°S (Fig. 2d). The relative weakening of the jet in the African sector at 200 hPa around 30°S during dry years compared to wet years appears to be related to the northern core of the Ferrel cell.
Interannual longitudinal shifts in the equatorial Walker cells can be depicted using the seasonal-averaged divergent component of moisture flux for a latitude band from 5°N to 15°S (Fig. 4). Eastward (westward) flux is indicated by positive (negative) values. Using this convention, areas of convergence around 140°E (where the flux from the east and to the west meet) and divergence around 100°W (where the flux to the west and east originate) can be deduced. Shifts in the convergence zone around 140°E, related to shifts in the Walker cell over the western Pacific Ocean, correlate well with South African summer rainfall. Eastward shifts of the convergence zone, typically during ENSO events, generally correspond to dry years. However, two interesting features emerge. First, not all dry summers in southern Africa relate to ENSO, and second, longitudinal shifts in the Walker cells between 50°W (western Atlantic) and 50°E (western Indian Ocean) are relatively insignificant. Walker cells are useful in understanding climate variability over the Pacific Ocean but the latter point above of small interannual longitudinal shifts in Walker cells raises questions about using these cells in the African region to explain the atmospheric response over the continent to ENSO events. This supports the notion that other mechanisms, such as those mentioned in the introduction, must exist whereby the ENSO signal reaches southern Africa.
Returning to the energy cycle it is again clear that summer rainfall in central South Africa is closely tied to latitudinal shifts of midlatitude wave activity. Dry years are clearly associated with a northward displacement of mean kinetic energy (Fig. 5a). In fact there is a distinction between all wet and all dry years between 42° and 32°S such that the mean kinetic energy values for all the dry years lie north of the wet years in this latitudinal band.
All of these large-scale features discussed above are essentially seasonal composites of the daily circulation that may mask out underlying processes. Therefore, in order to probe more deeply into the higher-order variability, the remainder of this section examines daily fields of kinetic energy of the barotropic (vertical mean) and baroclinic (shear) components of the wind for the Southern Hemisphere through a SOM analysis. This will attempt to explain the observed anomalies in large-scale circulation during wet and dry years.
Each day in the input data was assigned to one of the 12 nodes after the training of the SOM by finding a best-matching node with the least difference from the input data pattern, and the frequency of days related to each node has been calculated. When choosing those years with a particular rainfall characteristic (e.g., wet or dry), these frequencies will change if a relationship exists between the field and rainfall in the region being considered. In the case of summer rainfall over the central interior region of South Africa, this frequency change is marked. During dry (wet) years the frequency of days associated with nodes at the bottom right of the SOM are increased (decreased) and the frequency of days associated with nodes at the top left are decreased (increased, Fig. 6a). Basically the two associated SOM nodes (right column in Fig. 6a) are in opposition to each other with an equatorward shift in the barotropic kinetic energy during dry years. In both nodes the areas of significant variability are mainly in the Atlantic/African/Indian sector of the Southern Ocean. Furthermore, the frequency change of most nodes in the SOM, stratified by dry and wet years, is significant at the 99.5% level using the t test and exceeds one standard deviation of the interannual variability. This means that the change in circulation statistics between wet and dry years is significantly greater than the natural interannual variability of these statistics.
Similar contrasts are found for the baroclinic (vertical shear) kinetic energy, with dry (wet) years being associated with an increase (decrease) in the frequency of days associated with nodes on the left side of the SOM array and a decrease (increase) in days at the top right of the SOM (left column in Fig. 6b). The main mode of variability in this case lies in the South Pacific Ocean and the western South Atlantic Ocean (the latter is climatologically an important area of cyclogenesis; Jones and Simmonds 1993). These findings demonstrate that the variability of summer rainfall in South Africa is related to the large-scale patterns of variability in the Southern Hemisphere.
When relating these archetype frequency shifts to the circulation it is clear that dry (wet) years in South Africa are associated with a maximum northward (southward) displacement and increase (decrease) in the barotropic kinetic energy over the South Atlantic. The baroclinic component is a little more complex in that there are two leading areas of variability. The first is the tropical Pacific Ocean region where a decrease in tropical shear kinetic energy and a corresponding increase in subtropical energy (Fig. 6b), typical of an ENSO signal, is associated with dry years. The reverse pattern is found for wet years, although in these years baroclinic energy is reduced throughout most of the Southern Hemisphere subtropics. The second is over the South Atlantic Ocean. Here an increase (decrease) in baroclinic kinetic energy in the South Atlantic, particularly in the western sector, is associated with dry (wet) years.
The pattern that emerges here for dry years is that baroclinic energy generated in the western South Atlantic Ocean increases the barotropic jet to the east of this area (i.e., near southern Africa) through barotropic decay (Wiin-Nielsen 1962). As a result, a more zonal and northward-shifted westerly jet would weaken the subtropical South Indian Ocean anticyclone. The South Indian Ocean anticyclone is an important circulation system that transports moisture from the southwest Indian Ocean into the interior of southern Africa and a weakening of this system would ultimately cause a decrease in the summer rainfall. The weaker anticyclone was also modeled by Cook (2000), for idealized ENSO events that were designed to simulate dry years in southern Africa. The modulated zonal jet would likely also decrease the number of westerly waves that can interact with the tropical disturbances to form the tropical extratropical cloudbands that are so important for summer rainfall. This process of energy exchange suggests how weather patterns over South America and southern Africa may be linked. Zooming out to a hemispheric scale, the latitudinal shift in the barotropic kinetic energy is consistent with the changes to the meridional cells and jet stream discussed earlier, demonstrating the link between these features and the energy cycle.
b. Winter rainfall region
Whereas summer rainfall in southern Africa is mostly a result of convective activity that is typically controlled by tropical and extratropical disturbances and their interaction, the winter rainfall over the SWC and SWWA is predominantly driven by the passage of midlatitude waves. This difference becomes apparent when contrasting wet and dry years for the winter rainfall areas.
The distinction in the large-scale Hadley cells between wet and dry years over the winter rainfall region is not as clear as that for the summer rainfall areas of South Africa (Figs. 2e,f). The southern Hadley cell is slightly stronger (weaker) and the austral Ferrel cell is weaker (stronger) during wet (dry) seasons. Over African longitudes, the southern Ferrel cell is now much stronger during wet years (Figs. 2g,h). Changes to the jet between wet and dry years are consistent both zonally and in the African sector (Figs. 3c,d). The austral jet shifts toward the equator during wet years and toward the pole during dry years when it also appears to become more concentrated in a narrow band.
There is no obvious distinction in seasonal mean kinetic energy (Figs. 5b,c) between wet and dry years in the SWC or SWWA. This is in contrast to the summer months, but may be related to the bifurcation of the jet in the Australian region during winter (Trenberth 1991), that could obscure zonal-averaged fields. Nevertheless, these patterns do not reveal a great deal in terms of explaining the large-scale influences in interannual variability of winter rainfall areas in South Africa and Australia. Part of the difficulty for the SWC may be that the large-scale circulation anomalies appear to resemble a wavenumber-1 anomaly for wet winters, whereas wavenumber-3 anomalies are more apparent for dry winters, at least when derived for the full May–September season (Reason et al. 2002).
Turning to a SOM analysis, daily circulation characteristics of the barotropic and baroclinic kinetic energy for the Southern Hemisphere do demonstrate robust associations with winter rainfall in the SWC. These associations are seen in anomaly frequency maps of a SOM array trained on daily JJA data. The relative change in frequency of some of the archetypes of barotropic kinetic energy during dry and wet years compared to all years is high (Fig. 7a). During dry (wet) years the number of days in those seasons associated with the node toward the bottom-left corner (central-right portion) of the SOM double in number. A compensating halving of the frequency in the same areas occurs during wet (dry) years. Similar patterns are evident in SOM node frequencies for the baroclinic kinetic energy (Fig. 7b). These frequency changes, coupled with the fact that they are spatially coherent toward one side of the SOM (similar archetypes are placed near each other in the SOM), shows how the daily circulation statistics are quite different between dry and wet years. The nodes of maximum contrast between wet and dry years are also placed far from each other on the SOM indicating that either opposing signs of a dominant mode or different modes characterize wet or dry years.
Inspection of the circulation archetypes (SOM nodes) related to these large frequency changes reveal some interesting results. Dry years in the SWC have increased barotropic kinetic energy in a band from 30° to 45°S extending across the South Atlantic and Southwest Indian Ocean and across the Pacific Ocean, with a band of decreased energy in the subtropics over the South Atlantic and southern Africa (Fig. 7a). Wet years have a very similar pattern, but in reverse, demonstrating the role of the large-scale circulation in rainfall in this region.
The baroclinic kinetic energy differences are also clear. Here, dry (wet) years are associated with enhanced energy between 30° and 40°S and decreased energy north and south of that over the western South Atlantic Ocean (Fig. 7b). Over the Indian and western South Pacific Oceans a pattern of decreased (increased) baroclinic kinetic energy in the subtropics and increased (decreased) kinetic energy in the midlatitudes south of that, is associated with dry (wet) years in the SWC.
As is the case for summer rainfall, increased baroclinic kinetic energy in the southwest South Atlantic (Fig. 7a) coincides with increased barotropic energy downstream (viz., the SWC where cyclogenesis is reduced and cyclolosis becomes dominant; Jones and Simmonds 1993) resulting in drier conditions. This pattern is a meridional component of variability that has been shown to be modified by local SSTs (Reason at al. 2002). On a hemispheric scale, the significant changes in the barotropic energy in the Atlantic and South Pacific Oceans resemble a wavenumber-1 pattern. This is especially apparent for the wet years and is consistent with the wavenumber-1 pattern suggested by Reason et al. (2002).
In the case of SWWA there are also clear distinctions in daily circulation statistics between wet and dry years. However, wet (dry) years in this region have increased (decreased) barotropic kinetic energy over the region (Fig. 8a), in contrast to the situation in the SWC. This result is consistent with findings that anomalously higher sea level pressures over Western Australia and a southward shift of the circumpolar trough are associated with dry years in SWWA (Allan and Haylock 1993; Ansell et al. 2000). It also shows that the rainfall-producing mechanisms operate differently in SWWA in comparison to the SWC, as the jet splits over Australia during winter. It is during wet years that the barotropic portion of the jet in the midlatitudes is farther north than usual over the Indian Ocean. This coincides with increased (decreased) baroclinic kinetic energy in the midlatitudes (subtropics).
This situation is consistent with Allan and Haylock (1993) who suggested that modulations in the winter anticyclone over Australia and the longwave trough to the south influence rainfall in SWWA. For dry years in SWWA a wavenumber-1 feature becomes apparent that resembles the Antarctic Oscillation (Fig. 8a), which was found to correlate significantly with SWWA rainfall (Ansell et al. 2000; Smith et al. 2000). As already mentioned this Antarctic Oscillation feature can also be seen for the SWC (Fig. 7a). The wave has an equatorward displacement of barotropic energy centered on the South Atlantic and a compensating poleward displacement in the South Pacific. Comparison of Figs. 7a and 8a show that the circulation represented by the barotropic kinetic energy during wet (dry) years in the SWC tend to correspond with dry (wet) years in SWWA. This would support a large-scale hemispheric mode of variability.
4. Discussion
Knowledge of large-scale systems, such as the meridional Hadley cells and zonal Walker cells, assist in understanding observed teleconnections. The findings in this paper basically support the notion of an ENSO signal enhancing the Hadley cells that in turn increase the strength of the subtropical jet. However, changes to the Walker cells forced by ENSO do not appear to play such an important role in DJF rainfall in southern Africa. Cook (2001) suggests that the South Indian convergence zone shifts northeastward forced by upper-tropospheric convergence over the Indian Ocean and Indonesia in response to an eastward shift of the Walker cell during ENSO. However, the local Hadley circulation was shown to be an important factor in the subseasonal behavior of the Indian monsoon during ENSO events (Slingo and Annamalai 2000) and tended to override the effects of the Walker cell. In addition, there was considerable intraseasonal variability despite persistent ENSO forcing through the Walker cell. A similar situation is likely in southern Africa and this could explain why not all El Niño events correspond to dry years and not all dry years correspond to El Niño events.
The results highlight the need to study the shorter-period fluctuations and how they react to the larger-scale persistent forcing from SSTs and associated circulations. This was recognized by Godfred-Spenning and Reason (2002) in their investigation into the Australian summer monsoon. Cook (2000) also showed how interactions between transient and stationary eddies in the Tropics and midlatitudes explain how the South Indian convergence zone moves and strengthens or weakens.
The movement of this convergence zone has been identified as the major cause of interannual variability (Cook 2001). It is primarily composed of tropical-temperate troughs that are generally seen as the major rainfall producing system in southern Africa (Mason 1995; Mason and Jury 1997; Todd and Washington 1999). Numerical simulations of tropical-temperate troughs over South Africa have confirmed that these systems serve as a means of transferring moisture and energy from the Tropics to the higher latitudes (van den Heever et al. 1997). It is then plausible to link variability in the convergence zones to larger-scale energy exchange. For the summer season, it is apparent that we need to understand the causes of variability in the three Southern Hemisphere convergence zones as it has been shown that these are all linked through teleconnections (Mo and White 1985; Todd and Washington 1999).
One of the strongest findings in this paper supporting Southern Hemisphere teleconnections is how daily circulation archetypes of the entire Southern Hemisphere still relate so clearly to interannual variability in rainfall over South Africa. The frequency of particular archetypes changes dramatically between wet and dry years. Furthermore, the changes are such that archetypes of generally opposite characteristics are prevalent in either wet or dry years. This would not happen unless the energy cycle is linked on a daily scale around the hemisphere, supporting the idea of higher-frequency systems playing an important role in interannual variability. The persistent forcing by the Southern Hemisphere land–sea distribution and tropical SST anomalies could force atmospheric responses that communicate around the hemisphere through energy exchange as described herein.
In summer, cyclogenesis coincides with the strongest 500-hPa wind (Taljaard 1967). Thus, an active South Atlantic convergence zone will feed the barotropic energy south of South Africa through downstream barotropic decay (Wiin-Nielsen 1962; Trenberth 1991) and disrupt summer rainfall. Mason (1995) suggested that variations in South American convection would affect South African rainfall through teleconnections in the westerly waves. The findings in this study can now explain how this happens.
In winter, there are two main axes of cyclone tracks, from central South America southeastward to the central South Atlantic Ocean and from the Tasman Sea to the Drake Passage (Taljaard 1967; Jones and Simmonds 1993). The proximity of the former to the SWC and the fact that cyclones occur in an irregular pattern in the Southern Ocean (Taljaard 1967) suggests that variability of this axis of cyclone density would account for the interannual variability of SWC rainfall. Such variability is probably brought about by changes in energy conversion over the cyclogenetic area over South America.
There is no corresponding axis of cyclone density in the Indian Ocean, probably because southern Africa is dry during winter and the South Indian convergence zone weakens. However, when the South Atlantic cyclone track is active, barotropic kinetic energy is fed into the higher latitudes of the Indian Ocean and this affects the splitting of the jet over Australia, causing dry winters in SWWA. The more complex jet patterns during winter can explain why there are no clear zonal-averaged signals in rainfall variability, but the fact that there are teleconnections during winter, as with summer, is clear.
5. Summary and conclusions
During DJF seasons when the atmospheric energy cycle is intensified, perhaps through ENSO events, areas of preferred energy conversion, such as the South Pacific convergence zone, become more active through increased transient wave activity. The mechanism is that a stronger Hadley cell increases baroclinicity in the subtropics through an increased poleward flux of angular momentum. This process then enhances the jet and associated downstream barotropic energy. A stronger southern Hadley cell is clearly associated with dry summer conditions in South Africa. In addition to the zonal effect, there is a meridional dependency in the convergence zones that manifests through a northeast shift of the South Indian convergence zone during dry years. This meridional effect has been ascribed to increased baroclinic activity in the southwest South Atlantic Ocean increasing the barotropic energy south of South Africa. An increase in barotropic energy in this region then disrupts rainfall over the continent by weakening the Indian Ocean anticyclone and reducing midlatitude systems that interact with the tropical systems to produce the cloudbands.
During the austral winter, changes in the general circulation force different responses in regional rainfall. The South Indian convergence zone weakens considerably and the jet splits over Australia. Nevertheless, the energy cycle proceeds in much the same way as summer but with different consequences. Increased energy conversion through baroclinic action in the western South Atlantic Ocean forces increased barotropic energy farther downstream, which results in a more zonal jet and drier conditions in the SWC. Decreased energy conversion in the southwest Indian Ocean weakens the barotropic jet west of Australia and the subtropical branch of the jet over Australia, resulting in dry conditions in SWWA.
Intraseasonal variability is an important scale at which the climate affects society. Farmers and water resource managers in the region are interested to know parameters such as the start and end dates of the rainy season, and the frequency and duration of wet and dry spells within it. Thus, forecasts of event characteristics are really what endusers of seasonal outlooks require. The large spatial scale at which the teleconnections investigated in this paper take place, means that seasonal forecasts can begin to explore intraseasonal variability as part of the forecast product range. Traditionally such forecasts owe their skill to large-scale, slowly varying processes of the atmosphere and oceans. Now with a better understanding of teleconnections and energy exchange in the Southern Hemisphere we may be able to improve the accuracy of longer-range forecasts.
The challenge for South African variability is that the ENSO influence is subtle and there are a number of potential factors of importance. Perhaps as a result of its geographical location and relatively narrow zonal extent, regional oceanic influences (South Atlantic, South Indian, and Southern Oceans) may all contribute to southern Africa climate variability. These ocean influences with the very tight SST, topographic, and vegetation gradients in the region, complicate the task. This situation is in contrast to the relatively flat and broad Australian continent, close to the main centers of ENSO action, where seasonal predictability seems to be higher. An aspect not explored in this paper but that still needs work is to determine to what extent antecedent conditions play a role in climate variability. This has been identified in the case of the Indian monsoon (Slingo and Annamalai 2000), but not explicitly for southern Africa. An important question concerns the potential effect of winter and early spring rains (or lack thereof) or strong frontal systems or cutoff lows traveling anomalously far north during DJF since the experience of regional forecasters is that early onset of the summer rains is often associated with such midlatitude weather systems.
Acknowledgments
The authors sincerely thank the editor and anonymous referees for their valuable comments on this manuscript. This research was funded, in part, by the South African Government Department of Science and Technology Innovation Fund.
REFERENCES
Allan, R J., and M. Haylock, 1993: Circulation features associated with the winter rainfall decrease in southwestern Australia. J. Climate, 6 , 1356–1367.
Ansell, T J., C. J. C. Reason, I N. Smith, and K. Keay, 2000: Evidence for decadal variability in southern Australian rainfall and relationships with regional pressure and sea surface temperature. Int. J. Climatol., 20 , 1113–1129.
Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97 , 163–172.
Cavazos, T., 1999: Large-scale circulation anomalies conducive to extreme precipitation events and derivation of daily rainfall in northeastern Mexico and southeastern Texas. J. Climate, 12 , 1506–1523.
Cavazos, T., 2000: Using self-organizing maps to investigate extreme climate events: An application to wintertime precipitation in the Balkans. J. Climate, 13 , 1718–1732.
Cook, K H., 2000: The South Indian convergence zone and interannual rainfall variability over southern Africa. J. Climate, 13 , 3789–3804.
Cook, K H., 2001: A Southern Hemisphere wave response to ENSO with implications for southern Africa precipitation. J. Atmos. Sci., 58 , 2146–2162.
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.
Dima, I M., and J M. Wallace, 2003: On the seasonality of the Hadley cell. J. Atmos. Sci., 60 , 1522–1527.
Douville, H., F. Chauvin, and H. Broqua, 2001: Influence of soil moisture on the Asian and African monsoons. Part I: Mean monsoon and daily precipitation. J. Climate, 14 , 2381–2403.
Eastin, M D., and D G. Vincent, 1998: A 6-yr climatology of vertical mean and shear components of kinetic energy for the Australian–South Pacific jet stream. J. Climate, 11 , 283–291.
Gentilli, J., 1991: Homologous peri-oceanic west coast climates in the Southern Hemisphere. J. Roy. Soc. West. Aust., 74 , 15–33.
Godfred-Spenning, C R., and C. J. C. Reason, 2002: Interannual variability of lower-tropospheric moisture transport during the Australian monsoon. Int. J. Climatol., 22 , 509–532.
Hayes, M J., M D. Svoboda, D A. Wilhite, and O V. Vanyarkho, 1999: Monitoring the 1006 drought using the Standardized Precipitation Index. Bull. Amer. Meteor. Soc., 80 , 429–438.
Hewitson, B C., and R G. Crane, 2002: Self-organizing maps: Applications to synoptic climatology. Climate Res., 22 , 13–26.
Holopainen, E O., 1970: An observational study of the energy balance of the stationary disturbances in the atmosphere. Quart. J. Roy. Meteor. Soc., 96 , 626–644.
Huang, H-J., and D G. Vincent, 1985: Significance of the South Pacific convergence zone in energy conversions of the Southern Hemisphere during FGGE, 10–27 January 1979. Mon. Wea. Rev., 113 , 1359–1371.
Jones, D A., and I. Simmonds, 1993: A climatology of Southern Hemisphere extratropical cyclones. Climate Dyn., 9 , 131–145.
Jury, M R., H R. Valentine, and J. R. E. Lutjeharms, 1993: Influence of the Agulhas Current on summer rainfall along the southeast coast of South Africa. J. Appl. Meteor., 32 , 1282–1287.
Jury, M R., B. Pathack, C J. de W Rautenbach, and J. van Heerden, 1996: Drought over South Africa and Indian Ocean SST: Statistical and GCM results. Global Atmos.–Ocean Syst., 4 , 47–63.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Kohonen, T., 1995: Self-Organizing Maps. Springer, 362 pp.
Lorenz, E N., 1955: Available potential energy and the maintenance of the general circulation. Tellus, 7 , 157–167.
Mason, S J., 1995: Sea-surface temperature—South African rainfall associations, 1910–1989. Int. J. Climatol., 15 , 119–135.
Mason, S J., and P D. Tyson, 1992: The modulation of sea surface temperature and rainfall associations over southern Africa with solar activity and the quasi-biennial oscillation. J. Geophys. Res., 97 , 5847–5856.
Mason, S J., and M R. Jury, 1997: Climatic variability and change over southern Africa: A reflection on underlying processes. Prog. Phys. Geogr., 21 , 23–50.
Mo, K C., and G H. White, 1985: Teleconnections in the Southern Hemisphere. Mon. Wea. Rev., 113 , 22–37.
New, M., M. Hulme, and P. Jones, 2000: Representing twentieth-century space–time climate variability. Part II: Development of 1900–1998 monthly grids of terrestrial surface climate. J. Climate, 13 , 2217–2238.
Newell, R E., J W. Kidson, D G. Vincent, and G J. Boer, 1972: The General Circulation of the Tropical Atmosphere and Interactions with Extratropical Latitudes. Vol. 1. MIT Press, 258 pp.
Oort, A H., 1964: On estimates of the atmospheric energy cycle. Mon. Wea. Rev., 92 , 483–493.
Peixoto, J P., and A. Oort, 1974: The annual distribution of atmospheric energy on a planetary scale. J. Geophys. Res., 79 , 2149–2159.
Peterson, R G., and W B. White, 1998: Slow oceanic teleconnections linking the Antarctic Circumpolar Wave with the tropical El Niño–Southern Oscillation. J. Geophys. Res., 103 , 24573–24583.
Reason, C. J. C., 1998: Warm and cold events in the southeast Atlantic/southwest Indian Ocean region and potential impacts on circulation and rainfall over southern Africa. Meteor. Atmos. Phys., 69 , 49–65.
Reason, C. J. C., 2001: Evidence for the influence of the Agulhas Current on regional atmospheric circulation patterns. J. Climate, 14 , 2769–2778.
Reason, C. J. C., and H M. Mulenga, 1999: Relationships between South African rainfall and SST anomalies in the south west Indian Ocean. Int. J. Climatol., 19 , 1651–1673.
Reason, C. J. C., R J. Allan, J A. Lindesay, and T J. Ansell, 2000: ENSO and climatic signals across the Indian Ocean basin in the global context: Part I, Interannual composite patterns. Int. J. Climatol., 20 , 1285–1327.
Reason, C. J. C., M. Rouault, J L. Mélice, and D. Jagadeesha, 2002: Interannual winter rainfall variability in SW South Africa and large scale ocean–atmosphere interactions. Meteor. Atmos. Phys., 80 , 19–29.
Rogers, J C., 1984: Association between the North Atlantic Oscillation and the Southern Oscillation in the Northern Hemisphere. Mon. Wea. Rev., 112 , 1999–2015.
Ropelewski, C F., and M S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Mon. Wea. Rev., 115 , 1606–1626.
Slingo, J M., and H. Annamalai, 2000: 1997: The El Niño of the century and the response of the Indian summer monsoon. Mon. Wea. Rev., 128 , 1778–1797.
Smith, I N., P. McIntosh, T J. Ansell, C. J. C. Reason, and K L. McInnes, 2000: South-west western Australian winter rainfall and its association with Indian Ocean climate variability. Int. J. Climatol., 20 , 1913–1930.
Taljaard, J J., 1967: Development, distribution and movement of cyclones and anticyclones in the Southern Hemisphere during the IGY. J. Appl. Meteor., 6 , 973–987.
Tennant, W J., 2003: An assessment of intra-seasonal variability from 13-yr GCM simulations. Mon. Wea. Rev., 131 , 1975–1991.
Tennant, W J., 2004: Considerations when using pre-1979 NCEP/NCAR reanalyses in the Southern Hemisphere. Geophys. Res. Lett., 31 .L11112, doi:10.1029/2004GL019751.
Tennant, W J., and B C. Hewitson, 2002: Intra-seasonal rainfall characteristics and their importance to the seasonal prediction problem. Int. J. Climatol., 22 , 1033–1048.
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.
Trenberth, K E., 1991: Storm tracks in the Southern Hemisphere. J. Atmos. Sci., 48 , 2159–2178.
Trenberth, K E., G W. Branstator, D. Karoly, A. Kumar, N-C. Lau, and C. Ropelewski, 1998: Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures. J. Geophys. Res., 103 , 14291–14324.
Van den Heever, S C., P C. D’Abreton, and P D. Tyson, 1997: Numerical simulation of tropical–temperate troughs over southern Africa using the CSU RAMS model. S. Afr. J. Sci., 93 , 359–365.
Wiin-Nielsen, A., 1962: On transformation of kinetic energy between the vertical shear flow and the vertical mean flow in the atmosphere. Mon. Wea. Rev., 90 , 311–322.
Wright, W J., 1997: Tropical–extratropical cloudbands and Australian rainfall. Int. J. Climatol., 17 , 817–829.
Yarnal, B., 1985: Extra-tropical teleconnections with El Niño–Southern Oscillation (ENSO) events. Prog. Phys. Geogr., 9 , 315–352.
Yarnal, B., and G. Kiladis, 1985: Tropical teleconnections associated with El Niño/Southern Oscillation (ENSO) events. Prog. Phys. Geogr., 9 , 524–558.
Zheng, X., and E. A. B. Eltahir, 1998: The role of vegetation in the dynamics of West African monsoons. J. Climate, 11 , 2078–2096.
Anomalies (calculated by subtracting the average of all SOM nodes) of the four corner nodes of the 12-node SOM of barotropic kinetic energy for daily DJF fields from Jan 1986 to Dec 2003 expressed as a percentage change of the actual node value. Areas where these anomalies exceed one standard deviation of the interannual variability of the original dataset are contoured.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
Mass streamfunction composites (109 kg s−1) depicting the zonally averaged meridional Hadley cells during (a) dry and (b) wet DJF years in central South Africa. (c), (d) Meridional circulations for the sector from 15° to 35°E. (e), (f), (g), (h) Analogous maps for JJA rainfall in the SWC are shown. Shading denotes areas where these differences exceed one standard deviation of the interannual variability.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
Vertical slices of the differences in (a) zonal and (b) 15°–35°E sector kinetic energy (J m−2) between dry and wet DJF rainfall in the central interior and (c) and (d) JJA rainfall in the southwestern Cape. Shading shows areas where the difference exceeds one standard deviation of the interannual variability.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
Zonally orientated divergent moisture flux averaged from 15°S to 5°N for dry summers (solid) and wet summers (dotted) in the central interior of South Africa. Values above (below) the zero line denote eastward (westward) moisture transport, indicating the lower branch of the Walker cell.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
Seasonal mean zonal kinetic energy in the Southern Hemisphere for dry (solid lines) and wet (dashed) years (defined in Table 1) during summer in the (a) central interior, and during winter in the (b) SWC and (c) SWWA.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
SOM-node frequency changes (%) for (upper row) dry and (lower row) wet DJF years in central South Africa, relative to the frequency for the (left) full period, with the node pattern anomaly (as per Fig. 1) of the node identified on the SOM by the (right) circle, for (a) barotropic kinetic energy and (b) baroclinic kinetic energy. (right) Contouring indicates areas where the mean of the node is significantly different to the grand mean at the 99.5% level using the t-test statistic. The original frequency of the highlighted node is given above the node pattern. (left) Shading on the SOM-node frequency change boxes indicates changes that are significant at the 99.5% level using the t test to test the difference of two proportions. Positive (negative) changes are shaded dark (light).
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
As for Fig. 6, but for a SOM trained on daily kinetic energy for JJA months and stratified according to wet and dry years in the SWC rainfall region.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
Same as in Fig. 7, but for the SWWA rainfall region.
Citation: Journal of Climate 18, 15; 10.1175/JCLI3464.1
Wet and dry seasons for the summer (northern interior) and winter (SWC) rainfall regions in South Africa using various criteria. DJF years given are related to the Jan period. For DJF quintile values, El Niño years are given in bold and La Niña years are italicized.
Wet and dry years over SWWA during the JJA months taken from the spatial average of New et al. (2000) for 30°–35°S, 115°–120°E.
