Circulation Pattern Controls of Summer Temperature Anomalies in Southern Africa

Rising global temperatures resulting from both anthropogenic emissions and natural variability impact virtually all aspects of livelihood, ranging from the mortality rate (Allen and Lee, 2014; Lee and Sheridan, 2018; Sheridan and Lee, 2018; Sheridan and Dixon, 2017; Sheridan et al., 2019) and labor productivity (Knittel et al., 2020) to the rate and magnitude of climate extremes (Alimonti et al., 2022), among others. Climate predictions and projections are important for mitigating the consequences of rising temperatures by enhancing preparedness against extreme climate events as well as developing climate policies against future dangerous climate change. A major challenge in forecasting climate events, such as extreme temperatures, stems from the incomplete understanding of the physical mechanisms that drive the climate (Flato et al., 2014). Therefore, there is a need for studies that aim to improve the understanding of the atmospheric processes that govern the Earth’s climate. Enhanced knowledge of the atmospheric processes that govern the climate can serve as a benchmark for improving climate forecasts. In the regional context of southern Africa, which is vulnerable to anthropogenic climate change due to its economic conditions and geographical location within the descent region of the Hadley circulation (Baudoin et al., 2017), this study aims to enhance the understanding of the large scale circulations that modulate the summer temperatures in southern Africa.

Several studies have hinted at the role of climate drivers in modulating atmospheric circulations over southern Africa (Reason and Rouault, 2005; Lyon and Mason, 2007; Malherbe et al., 2014; Dieppois et al., 2015; Hoell et al., 2015; Manatsa et al., 2017) and other regions in the globe (Pirhalla et al., 2022). Among the climate drivers, the sea surface temperature (SST) patterns of the Subtropical Indian Ocean Dipole (SIOD) and the Southern Annular Mode (SAM) directly impact the climate of southern Africa because they exist in the mid-latitude oceans of the Southern hemisphere (i.e., the subtropical Indian Ocean, and the Southern Ocean). In addition, the Indian Ocean Dipole (IOD) and El Niño Southern Oscillation (ENSO) have a remote influence on the climate of southern Africa (Reason and Jagadheesha, 2005). Moreover, the imprint of the ENSO signal can be found in the SAM, IOD, and SIOD modes since all the climate drivers generally influence SST anomalies in the southwest Indian Ocean as well as the anticyclonic circulation of the Mascarene High. (Ding et al., 2012; Morioka et al., 2015; Ibebuchi, 2021a, b). Therefore, this study also investigates the impact of climate drivers on temperature variability in southern Africa.

The link between synoptic circulations and precipitation in southern Africa has been widely documented (Lennard and Hegerl, 2015; Engelbrecht and Landman, 2016; Wolski et al., 2018; Burls et al., 2019; Mahlalela et al., 2019). Also, studies have highlighted the possible impact of the SAM on temperature anomalies in southern Africa. For example, Gillett et al. (2006) used regression analysis to investigate the impact of positive SAM on temperatures in the Southern Hemisphere (with the inclusion of southern Africa), however, the negative SAM was not considered and the associating physical processes were not analyzed in detail. Also, for other regions in the Southern Hemisphere, such as Australia and South America, studies have addressed the relationship between the SAM and surface temperatures (Hendon et al., 2007; Silvestri and Vera, 2009). Further, the majority of available works of literature on the IOD and SIOD which address the climate of southern Africa, focus on the relationship between the climate drivers and rainfall variability in southern Africa (Behera and Yamagata, 2001; Reason, 2001; Hoell et al., 2015; Ibebuchi, 2023a, 2023b), with temperature responses receiving less attention. Given that these climate drivers impact the amplitude and frequency of circulation types (CTs), this study goes further to decompose in time, the distinct circulations in southern Africa and then analyze the mechanisms through which the individual CTs impact surface temperature and relative humidity (that contributes to heat stress) in southern Africa. Thus, this study adds major value to the literature in terms of improving the understanding of the physical processes through which large-scale circulations influence the spatial variations and magnitude of summer temperatures in southern Africa.

The remainder of this study is structured as follows. Section 2 presents the data and methods employed in this study. The results are presented comprehensively in section 3. Section 4 provides a discussion of the results and section 5 presents the conclusions.

High-resolution reanalysis data is obtained from ERA5 datasets (Hersbach et al., 2020). The relevant ERA5 reanalysis datasets to this study include sea level pressure (SLP), 850-hPa wind vectors, relative humidity, specific humidity, and geopotential height, and 1000-hPa geopotential height. Gridded 2-m average temperature observations are obtained from the Climate Prediction Center (CPC) Global Unified Temperature dataset (https://psl.noaa.gov/data/gridded/data.cpc.globaltemp.html). All data sets are obtained at daily temporal resolution from 1979 to 2021. The horizontal resolution of the ERA5 reanalysis data is 0.25° longitude and latitude, while the gridded CPC 2-m temperature data has a resolution of 0.5° longitude and latitude. Also, correspondence between ERA5 and gridded CPC temperature data has been reported (Roffe and van der Walt, 2023), which strengthens the choice of using both data sets for our analysis.

This current study represents incremental work from previous research that applied circulation typing to investigate circulations in southern Africa at multiple spatial scales (Ibebuchi, 2023a), hence the same classified CTs are used. A major advancement in this study is that these same CTs are now being used to investigate the mechanisms through which circulation patterns modulate the magnitude and spatial variations of austral summer (i.e., December to February, DJF) temperature in southern Africa. We consider austral summer to be coincident with the time when absolute temperatures are highest in the study region, and thus, most impactful. The spatial extent for the CT classification is 0° to 50.25°S and 5.25°E to 55.25°E (Fig. 1). The adjacent maritime regions that act as moisture sources to the landmass were included. Further, when the CTs are linked to temperature over the landmass, southern Africa might be defined to include the landmass from about 10°S and poleward. This is typically done to make a distinction with the tropical landmass. Nonetheless, we have included the tropical latitudes (i.e., up to 0° latitude) to capture variability in the cross-equatorial northeast trade winds. This is because the cross-equatorial northeast trade winds play a role in defining convergence within the Angola Low and the Mozambique Channel that both impact the climate of southern Africa. However, our analysis focuses on the temperature anomalies under the distinct CTs on land from poleward of 10°S.

Figure 1.  The standardized SLP composite of z-score for the classified CTs.

The obliquely rotated T-mode (variable is time series and observation is grid points) principal component analysis (Compagnucci and Richman, 2008) applied in a fuzzy approach is used for the classification of the CTs. The details of the classification approach are given in Appendix A. The classification of the CTs with rotated T-mode PCA is a time decomposition technique. Singular value decomposition is applied to the correlation matrix containing the correlation between the SLP fields at the distinct daily time series in the analysis period. The resulting PC loadings which designate the amplitude of the set of circulation patterns for a given day are used to assign CT(s) within the signal range for a given day, which implies that the CTs that occurred on the day in question.

We apply a composite map of different climate variables to examine the relationship between the classified CTs and DJF temperatures in southern Africa. We have used SLP composites to investigate the horizontal variations of pressure under each of the CTs, which can also be suitable for characterizing continental heating (i.e., in areas with continental lows) as well as boundary layer stability and convective activity over the adjacent oceans. We defined the 1000–850-hPa thickness of the atmospheric layer as the vertical distance in meters between two pressure levels. This is justified because iterative pre-evaluations show that the 1000–850-hPa thickness can accurately capture the spatiotemporal variations of 2-m mean temperature values in the study region. It follows that positive thickness values are related to positive temperature anomalies and negative thickness values are related to negative temperature anomalies. In addition to analyzing the SLP and atmospheric layer thickness, composites of 850-hPa wind velocities are analyzed to establish the prevailing winds under each CT. Thus, the analysis of the SLP, wind direction, and thickness enables the documentation of low-level cold and warm air advection under each CT.

Given the high seasonality of temperature, we rely on composite anomaly values during DJF to enable a robust diagnosis of the temperature values and other variables to diagnose temperature patterns under each CT. The composite anomaly is calculated as the difference between the days assigned to a CT and the 1981–2010 DJF climatology of the variable in question. Regression analysis was used at each grid point in the region of assessment to investigate the association between climate drivers (SAM, ENSO, IOD, and SIOD) and temperature variability in southern Africa. The statistical significance of the association, based on the regression model, is tested at a 95% confidence level. To account for the false discovery rate in the multiple hypothesis testing, we used the Benjamini-Hochberg procedure (Benjamini and Yekutieli, 2001), which assumes that the hypotheses being tested are independent.

The standardized SLP composites for the days when the classified CTs occurred during the analysis period are shown in Fig. 1. The CTs in Fig. 1 were externally validated using SLP data from a different reanalysis product (Ibebuchi, 2022c) and a suite of CMIP6 general circulation models (Ibebuchi, 2022b). While the CTs are not confined to occur in any specific season, the spatiotemporal structure and dynamics of the CTs can also be influenced by seasonal variations in large-scale processes such as diabatic heating, thus the CTs tend to be dominant in specific seasons when the prevailing (seasonal) atmospheric condition favors their mechanisms (Fig. A1 in the Appendix). Pressure variations over the adjacent oceans are used to infer convective activity in these locations —anticyclonic (cyclonic) circulations over the oceans imply suppressed (enhanced) convection. Similarly, pressure variations over the landmass can be used to infer continental heating or atmospheric stability at the boundary layer — anticyclonic (cyclonic) circulations over the land imply a relatively stable (unstable) atmosphere.

Next, we analyze the composite anomaly patterns for DJF the classified CTs in Fig. 1 in terms of the 1000–850 hPa thickness (Fig. 2a), temperature (Fig. 2b), and 850-hPa relative humidity (Fig. 3), as well as the composite patterns of the 850-hPa wind vectors (Fig. 4). These figures will be interpreted coherently, based on the asymmetric patterns of the CTs (i.e., the positive phase and the negative phase). We focus our analysis of the composite anomaly maps in Figs. 2 and 3 mostly on the regions (i.e., group of grid points) with robust changes.

Figure 2.  Standardized composite anomaly maps of the (a) 1000– 850 hPa thickness) and (b) 2-m temperature anomaly during DJF for the classified CTs in Fig. 1.

Figure 3.  Standardized composite anomaly maps of 850-hPa relative humidity during DJF for the classified CTs in Fig. 1.

Figure 4.  Composite maps of SLP (black contours) and 850-hPa wind (green vectors) for the classified CTs in Fig. 1. The contour interval is 3 hPa.

The composite patterns generally reveal that under each CT, the DJF temperature anomalies are characterized by spatial heterogeneity (Fig. 2b). A possible explanation for the spatial heterogeneity of temperature anomalies under each CT can be due to the distinct regional climates in southern Africa. Under the asymmetry of each CT (i.e., the negative and positive phases) temperature variations are least in the deep tropics (10°S–0°) mostly because diabatic heating (and SLP) have fewer variations in the deep tropics. However, as noted earlier, the major focus of the temperature anomalies is in southern Africa. Figure 2a indicates that for the asymmetry of a given CT, depending on the circulation features, there appear to be coherent temperature patterns in the central parts of southern Africa (mostly at landmasses close to the southwest Indian Ocean), the southwestern parts of southern Africa (characterized by a Mediterranean type of climate), and at the southern tip of southern Africa. Thus, the regional climate types of the aforementioned regions contribute to defining the overall spatial heterogeneity of the temperature anomaly patterns under each CT.

Analysis of the CT1+/CT1– pairing indicates variations in the subtropical ridge south of South Africa. Under CT1+ the mid-latitude cyclone is more enhanced along the south coast of southern Africa (Fig. 1). The associated wind pattern along the south coast of southern Africa under CT1+ is westerly (Fig. 4), and thickness anomaly is positive over South Africa except for the western tips, noting further that the thickness anomaly is also positive in the central parts of southern Africa except for large parts of Mozambique and Madagascar (Fig. 2a). Regions with a positive thickness anomaly indicate warm advection and are generally associated with a positive temperature anomaly (Fig. 2b) as well as a negative relative humidity anomaly (Fig. 3).

Conversely, under CT1–, the South Atlantic ridge is enhanced along the south coast of southern Africa, driving cold air to the southeast into the southern African domain (Figs. 1, 4), and cyclonic activity is more enhanced in the southwest Indian Ocean compared to CT1+. Consistent with the cold air advection on the south coast of southern Africa, the thickness anomalies are negative over vast regions in South Africa (Fig. 2a). Under CT1–, the thickness anomalies are also negative over the central domains of southern Africa coincident with the advection of the cold air from the South Atlantic high-pressure system, except for parts of Mozambique and Madagascar given the enhanced oceanic low-pressure system (that is, convective activity) that results in warm air advection. As a result, regions with positive (negative) thickness anomalies are generally characterized by positive (negative) temperature anomalies and negative (positive) relative humidity anomalies (Figs. 2, 3). The analysis of CT1+/CT1– indicates that cold advection as a result of circulation at the subtropical ridge implies lower thickness and temperature values in the regions where the cold air penetrates; similarly, enhancement of convective (or cyclonic activity) in the southwest Indian Ocean, results in warmer temperatures at the coastal landmasses.

The CT2+ pattern further presents circulation variability associated with the enhancement of the anticyclonic circulation at the western branch of the Mascarene High and the enhancement of the mid-latitude cyclone south of South Africa (Fig. 1). Hence the wind patterns are southeast over the southwest Indian Ocean, driven by the Mascarene High, and westerly, along the south coast of southern Africa (Fig. 4). Given that cold air advection from the South Atlantic anticyclone is weakened along the south coast of southern Africa, the thickness anomaly is strongly positive at the south coast of southern Africa (Fig. 2a) as well as temperature anomaly (Fig. 2b). However, since the Mascarene High drives colder southeasterly winds into large parts of southern Africa, the thickness anomaly is negative and the temperature anomaly is negative over large parts of southern Africa. Consequently, the pattern of CT2+ brings a positive relative humidity anomaly over the central parts of southern Africa, but a negative relative humidity anomaly over parts of the southern tips that are anomalously warmer (Fig. 3). CT2$ - $ is characterized by opposing (and asymmetric) circulation features, relative to CT2+, thus bringing warmer (colder) temperatures over the central parts (southern tips) of southern Africa.

CT3+ is related to a positive SAM and CT3$ - $ is related to a negative SAM (Ibebuchi, 2021a). We recall that the SAM through its control of circulation in the mid-latitudes significantly impacts the regions in southern Africa with a Mediterranean type of climate, that is, the Western Cape Province located in the southwestern part of southern Africa. Based on the climatology of the regions in southern Africa with the Mediterranean type of climate, the enhancement of the mid-latitude cyclones along the south coast of southern Africa under CT3$ - $ (Fig. 1), and their associated westerly winds (Fig. 4) allow cold fronts to sweep across the southwestern parts of southern Africa. Thus, under CT3$ - $ a zonal dipole-like structure of thickness anomaly whereby the eastern (western) region of South Africa has positive (negative) thickness anomaly values (Fig. 2a). As a result, the temperature anomaly is positive (negative) in the eastern (western) region of South Africa (Fig. 2b). Also, the relative humidity anomaly is negative (positive) in the eastern (western) region of South Africa (Fig. 3), an indication of an atmosphere approaching saturation that eventually results in frontal rainfall in the western domain of South Africa. For the large parts in the central regions of southern Africa, Fig. 2 shows that under CT3$ -, $ thickness and temperature anomalies are positive, and relative humidity anomalies are symmetrically negative (Fig. 3). The reason is that while CT3$ - $ is dominant in winter seasons (Fig. A1), during its occurrence in DJF, the SST and low-level convergence are higher in the southwest Indian Ocean, and continental heating is also higher compared to austral winter (JJA). Thus, during DJF, the displacement of atmospheric blocking by the western branch of the Mascarene High, and the resultant weakening of cold advection, allows warmer ocean waters in the southwest Indian Ocean (Vigaud et al., 2009). Hence, during the occurrence of CT3$ - $ in DJF, warm air advection from the southwest Indian Ocean, driven by northeast trade winds (Fig. 4), leads to a warmer climate in large parts of southern Africa.

Further, CT3+ is associated with increased SLP in the mid-latitudes but falls short of the closed and elongated South Atlantic anticyclone that results in cold air advection from the Southern Ocean to large parts of South Africa (cf. CT1– in Fig. 4). In this case (of CT3+), cold air advection to the eastern part of South Africa results from southeast winds driven by the western branch of the Mascarene High. Therefore, under CT3+ due to the cold air advection, the eastern part of South Africa has a negative thickness anomaly (Fig. 2a), negative temperature anomaly (Fig. 2b), and a positive relative humidity anomaly (Fig. 3). Since westerly winds and the associated cold fronts are suppressed south of South Africa, the western region of South Africa has a positive thickness anomaly (Fig. 2a), positive temperature anomaly (Fig. 2b), and negative relative humidity anomaly (Fig. 3). For the central parts of southern Africa, due to the enhanced cold air advection by southeast winds driven by the Mascarene High, a negative thickness and temperature anomaly prevails over the central regions (Fig. 2) coupled with positive relative humidity anomaly (Fig. 3).

For the other CTs, similar patterns of variations in the pressure systems and the associating cold (warm) air advection, reduces (increase) the thickness of the air layer, resulting in negative (positive) temperature anomalies in preferred regions. For example, CT5– reinforces the findings from CT1– that when the South Atlantic Ocean high pressure is closed, ridging at the southern coast of southern Africa, large parts of South Africa become colder due to enhanced cold air advection (Figs. 1–4); CT6+ depicts a very similar pattern but with stronger amplitude. The pattern of CT6– also reinforces previous findings that when southeast winds are enhanced by the circulation at the western branch of the Mascarene High then cold air advection to the eastern regions and large parts of southern Africa implies colder conditions in the study domain (Figs. 1–4).

Analysis of CT7+/CT7– illustrates the relationship to the positive (negative) SIOD (Ibebuchi, 2023a). Under CT7+/CT7– the southwest Indian Ocean is warmer (cooler), based on the enhanced cyclonic (anticyclonic) activity; in addition, the boundary layer in large parts of southern Africa is relatively unstable (stable) under CT7+ /CT7–. For CT7+, the pressure gradient between the South Atlantic high-pressure and the continental low results in cold air advection from the South Atlantic Ocean and the Southern Ocean to South Africa, but warm air advection from the (warmer) southwest Indian Ocean to the central regions of South Africa (Figs. 1, 2a, 4). Therefore, while colder temperatures are evident in the south and southwestern parts of South Africa, CT7+ brings warmer temperatures to the central domains of southern Africa (Figs. 2, 3). Conversely, in CT7– cold advection from the southwest Indian Ocean prevails over the central parts of southern Africa (Figs. 2a, 4) and when coupled with high-pressure anomaly in Fig. 1, leads to negative temperature anomalies over the central parts of southern Africa (Fig. 2b). But due to enhanced blocking of the mid-latitude cyclones at the southern coast of southern Africa, the activity of cold fronts that bring colder temperatures to the southwestern part of southern Africa is weakened so that the southwestern parts of southern Africa retain positive temperature anomalies.

The CT9– pattern is remarkable and results in widespread positive temperature anomalies in South Africa (Fig. 2b). This is because of the increasedmeridional pressure gradient between the low pressure in the Mozambique Channel and the western branch of the Mascarene High, which reduces the likelihood for colder southeast winds from the Mascarene High to penetrate into the eastern parts of South Africa (Figs. 1, 4), especially when coupled with the absence of the closed South Atlantic anticyclone along the southern coast of southern Africa. However, the CT9– pattern also implies that colder air from the western branch of the Mascarene High moves towards the low pressure in the Mozambique Channel, penetrating the central parts of South Africa and partly adjusting westward towards Madagascar (Fig. 4); this flow pattern brings negative thickness anomalies and colder temperatures to the central parts of southern Africa.

Figure A2 shows that regions that experience a robust temperature decrease, i.e., the western parts in CT3– and CT6+, and the eastern parts in CT7– (Fig. 2b), are mostly associated with low specific humidity; and regions with a robust increase in temperatures are mostly associated with high specific humidity, for example, the southeastern parts under CT3– and CT6+ and the southern parts under CT9–. This is generally because a warmer (colder) atmosphere can hold more (less) moisture. However, there are exceptions to this, for example under CT1–, where the negative temperature anomaly in southern parts of southern Africa is associated with a positive specific humidity anomaly. This is because the moisture content over the landmass is also modulated by other factors, such as advection and moisture uptake over the oceans.

Previous studies linked the CTs in Fig. 1 to climatic modes of variability. CT3+/CT3– was found to be modulated by the SAM (Ibebuchi, 2021a) such that CT3+ (CT3–) is related to positive (negative) SAM. CT5+/CT5– was found to be modulated by the Niño-3.4 index (Ibebuchi, 2022a), such that CT5– (CT5+) is related to El Niño (La Niña). CT7+/CT7– was found to be modulated by the SIOD (Ibebuchi 2023a), such that CT7+ (CT7–) is related to the positive (negative) SIOD. CT9+/CT9– was equally found to be modulated by the IOD (Ibebuchi, 2023b) such that CT9+ (CT9–) is related to the negative (positive) IOD. The frequency of these CTs is modulated by these large-scale modes of variability and thus represents the regional-scale manifestations of the modes. That is the CTs are the physical mechanism through which these modes impact southern African temperatures.

In addition to the indirect control of DJF temperatures by the climatic modes through the CTs, as shown in Figs. 1 and 2, we also apply regression maps to specifically examine the direct association between these well-known climatic modes and temperature in southern Africa. Figure 5 shows this association to spatially heterogeneous with a dipole structure, especially over the southwestern and east-central parts of southern Africa. Figure 5 shows that El Niño is associated with significant temperature increases over the central parts of southern Africa. The positive phase of the IOD is associated with an increase in temperature over the northeastern, northwestern, and southwestern parts of southern Africa. The positive SAM is associated with temperature increases, mostly over the southwestern parts of southern Africa, while the SIOD associated with temperature increases over the central/northcentral parts of southern Africa.

Figure 5.  Regression map of the (a) Niño-3.4 index, (b) IOD index, (c) SAM index, and (d) SIOD index onto the annual mean temperature anomaly in southern Africa from 1979 to 2021. The stippling shows grid points that are not statistically significant at a 95% confidence level.

Liu et al. (2022) found that high-frequency atmospheric signals are crucial for improving the sub-seasonal predictability of precipitation in most land monsoon regions. The high-frequency intraseasonal variability was found to be responsible for a significant portion of the total intraseasonal variability and generally dominates the sub-seasonal predictability of various land monsoons. They further suggested that high-frequency variability is necessary for enhancing the predictability of climate variables such as precipitation and temperature at sub-seasonal time scales. The inclusion of high-frequency variability in the CTs created in this work with the application of rotated T-mode PCA to daily SLP (Ibebuchi, 2022a), in line with the recommendation of Liu et al. (2022), is necessary for the enhanced predictability of climate variables such as precipitation and temperature. This is because the multi-scale interaction between synoptic and inter-annual signals is connected by tropical and extratropical atmospheric signals affecting these CTs, on the intraseasonal time scales.

The majority of the studies linking atmospheric circulations to the climate of southern Africa have focused on precipitation (Fauchereau et al., 2009; Jury, 2015; Barimalala et al., 2020). Other studies have examined the relationship between climate drivers and precipitation variability in southern Africa (Reason and Jagadheesha, 2005; Hart et al., 2013; Hoell et al., 2015). Indeed, the studies concur that the hydroclimate of southern Africa is modulated by large-scale circulations. However, little attention has been given to the relationship between synoptic to large-scale circulations in southern Africa and temperature anomalies in the southern African landmass. Among such studies, Gillet et al. (2006) investigated the relationship between the SAM and temperature in Southern Hemisphere landmasses. The authors found that positive SAM can be related to temperature increases over the southwestern parts of southern Africa, but temperature decreases over some northeastern parts. The results from Gillet et al. (2006) are consistent with the regression analysis between the annual temperature anomalies and the SAM index in this study (Fig. 5). Also, the circulations through which a positive SAM increases the temperature of the southwestern parts of southern Africa (i.e., CT3+) is manifested by a poleward shift of westerly winds, which in turn suppresses the passage of cold fronts over the southwestern parts of southern Africa. The positive IOD increases SST over the tropical western Indian Ocean (Saji et al., 1999) and alters atmospheric circulations over southern Africa (Manatsa et al., 2011). Our results indicate that the SST pattern of the positive IOD is linked to temperature increases over Madagascar and the northeastern, northwestern, and southwestern parts of southern Africa. Similarly, during El Niño and positive SIOD events, both associated with an SST increase in the southwest Indian Ocean, temperature increases are favored in some central parts of southern Africa. Also, El Niño and the positive SIOD appear to be related to temperature decreases over the southwestern parts of southern Africa possibly due to a weakening of high-pressure adjacent to South Africa.

A previous study by Ibebuchi (2022b) investigated the effects of climate change on the atmospheric circulation types in southern Africa and found that under a future climate change scenario, the frequency of occurrence, amplitude, and spatial configuration of the classified CTs in this work are expected to change. Consequently, the relationship between atmospheric circulations over southern Africa and temperature variability will be impacted by climate change. As documented in Ibebuchi (2022b), CMIP6 climate models projected summer periods of weaker circulation at the western branch of the Mascarene High due to warmer southwest Indian Ocean temperatures (i.e., CT4–) as well as summer periods of stronger circulation at the western branch of the Mascarene High due to a more positive SAM (i.e., CT3+). As shown in Fig. 2a, a more positive SAM, i.e., CT4+ (warmer southwest Indian Ocean i.e., CT3–) will imply warmer temperatures over the southwestern (southeastern) parts of southern Africa.

The climate of Southern Africa is influenced by a complex interplay of feedback between the atmosphere and surface processes. Changes in vegetation cover can impact the surface albedo, surface roughness, and evapotranspiration rates, which affect the temperature patterns in a region (e.g., Tran et al., 2017). For example, Clark and Arritt (1995) reported that the influence of increased vegetative cover can directly lead to increases in convective precipitation, by not only providing shade to reduce the conduction of heat into the soil (and thus increasing available heat energy in the atmosphere) but also by stimulating the extraction of soil moisture. Moreover, a study by Engelbrecht et al. (2015a) used a regional climate model to project future temperature changes in Africa, finding that (for southern Africa) as temperature increases, the soils become drier through enhanced evaporation; and this, in turn, impacts vegetation.

This study investigated the mechanisms through which synoptic circulations control atmospheric layer thickness, temperature, and relative and specific humidity in southern Africa, during the austral summer season. We also examined the impact of the SAM, IOD, ENSO, and SIOD on temperatures in southern Africa. Our results on the synoptic circulations that modulate summer temperature in southern Africa can be summarized as follows:

1. The temperature in southern Africa exhibits spatial heterogeneity under the classified CTs. Thus, the distinct regional climate zones within the southern African landmass contribute to defining the spatial variations of temperature anomalies under a given CT.

2. Generally, two asymmetric variabilities in the semi-permanent high-pressure system influence regional temperature variations in southern Africa. The first is evident when the South Atlantic Ocean high pressure is closed, and ridges along the southern coast of southern Africa; due to the ridging high pressure, this synoptic circulation pattern is associated with southeast winds impacting the southern parts of southern Africa and causes cold air advection and negative temperature anomalies in the southern tip of southern Africa. Conversely, when the South Atlantic High is weak on the southern coast of southern Africa, westerly winds dominate over the southern coast of southern Africa, the thickness anomaly value is positive, and the temperature anomaly is positive in the southern parts of southern Africa. Second, when the anticyclonic circulation along the western branch of the Mascarene High is stronger, there is an enhancement of cold air advection by southeast winds into large parts of southern Africa, reducing the atmospheric thickness, which results in negative temperature anomalies. Conversely, when the anticyclonic circulation at the western branch of the Mascarene High is weakened during the summer season, atmospheric blocking of the low-pressure system from the tropics is weakened as well, allowing for enhanced cyclonic/convective activity in the southwest Indian Ocean; the implication is an enhancement of warm air advection into parts of southern Africa, increased thickness, and positive temperature anomalies in parts of southern Africa. Hence when conditions are favorable, southeast winds from the semi-permanent high-pressure systems are mostly associated with colder temperatures in southern Africa.

3. Other variabilities that modulate summer temperature anomalies in southern Africa and interfere with circulations in the high-pressure systems are the formation of continental lows and a trough in the Mozambique Channel. Continental lows increase boundary layer instability, thickness, and summer temperature anomalies. Further, the strengthening of the Mozambique Channel trough coupled with a weak South Atlantic anticyclone at the southern coast of southern Africa can be implicated to cause widespread warming over South Africa. This is because the aforementioned circulation pattern (i.e., CT9$ - $) increases the pressure gradient between Mascarene High and the low pressure in the Mozambique Channel so that cold air advection to South Africa is significantly limited.

4. Overall, at the synoptic scale, summer temperature anomalies in Madagascar are modulated by cold advection from a pronounced Mascarene High and warm advection resulting from a weaker Mascarene High and warmer southwest Indian Ocean waters.

5. Climate drivers such as SAM, ENSO, IOD, and the SIOD impact temperatures over southern Africa. El Niño and positive SIOD are associated with temperature increases over the central parts of southern Africa. Positive SAM and positive IOD are associated with a temperature increase over the southwestern parts of southern Africa, while additionally positive IOD is linked to a temperature increase over the northwestern and northeastern parts of southern Africa.

Conflict of interest: There are no conflicts of interest in this paper.

Funding statement: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Author's Contribution: All authors worked together on all aspects of the manuscript.

Ethics approval: No human subject is involved in this study and Figures belong to the authors. The paper is also not under consideration in any Journal. There is also no conflict of interest in this paper.

Consent to participate: No human research is used. The authors' consent for this paper to be considered.

Consent for publication: The authors' consent for this paper to be published.

APPENDIX

Figure A1.  Annual cycle of the CTs in Fig. 1. The y-axis is the relative frequency of occurrence of the CTs and the x-axis is the calendar months.

Figure A2.  Standardized composite anomaly maps of 850-hPa specific humidity during DJF for the CTs in Fig. 1.

Classification of circulation types using the obliquely rotated T-mode PCA

The classification of the CTs is completely eigenvector-based (Ibebuchi and Richman, 2023). It involves the application of an obliquely rotated PCA to the T-mode matrix (variable or column matrix is time series and row matrix is grid points) of the z-score standardized SLP field. The SLP field is standardized to give equal weight to all days in the analysis period. Singular value decomposition is applied to the correlation matrix, containing the correlation between SLP observations at each time in the analysis period, to obtain the PC scores, eigenvalues, and eigenvectors. The PC scores capture the spatial variability patterns, and the eigenvectors localize the spatial patterns in time. To make the eigenvectors responsive to rotation and to become correlations between the PC scores and the standardized SLP field, the eigenvectors are multiplied by the square root of their corresponding eigenvalues so that they become PC loadings that can be longer than a unit length. To enhance the physical interpretability of the PC loadings, they were rotated obliquely using Promax at a power of 2. The oblique rotation simplifies the structure of the PCs by maximizing the number of near-zero loadings, so that a unique time series with large loading magnitudes are clustered under a given PC. Given that we desire to analyze both the dominant and (rare) patterns associated with extremes, of which the latter is often located in the higher-order PCs, we decide on the optimal number of PCs to retain and rotate by iteratively increasing the number of PCs until the spatial pattern of the CTs created from the next added PC is the least unique from the already retained PCs. Since each PC contains asymmetric patterns separated by the sign of the PC loadings, following its efficacy in previous studies (Ibebuchi and Richman, 2023), $ \pm $0.2 is used in this study to separate PC loadings in the signal range from PC loadings in the noise range. Introducing this threshold allows for a day can be classified under more than one PC pattern so long as the PC has signal magnitude >$ \left|0.2\right| $ on that day. Hence each retained PC gives two asymmetric classes (i.e., clusters above or below the $ \pm $0.2 threshold) and the SLP mean of the days in a given class is the CT.