The impact of climate change on the frequency of occurrence of atmospheric circulation patterns can have a wide range of consequences ranging from weather extremes to the modification of pollutant transport. This study uses 11 CMIP6 global climate models (GCMs) to investigate the impact of future climate change on the frequency of occurrence of atmospheric circulation patterns, in Africa south of the equator. Here it is shown from the historical analysis that there are statistically significant trends in the frequency of occurrence of some of the classified circulation types (CTs) in the study region. Further, under the SSP5-8.5 and SSP2-4.5 emission scenarios, the historical CTs were reproduced, suggesting that future climate change will not constrain the existence of the CTs. However, for future emission scenarios, the ensemble of the GCMs projects notable changes in the spatial structure of the CTs and statistically significant trends in the frequency of occurrence of most of the CTs, towards the end of the 21st century. The intensity of the projected changes in the spatial structure and linear trends in the frequency of occurrence of the CTs are relatively stronger under the higher emission scenario. As regards changes in synoptic circulations in the study region, the ensemble of the GCMs project, (i) a positive trend in the frequency of occurrence of austral summer dominant CTs associated with atmospheric blocking of the Southern Hemisphere mid-latitude cyclones, adjacent to South Africa; (ii) alternating frequent periods of enhanced (suppressed) anticyclonic circulation at the western branch of the Mascarene high possibly due to a more positive phase of the Southern Annular Mode (warmer southwest Indian Ocean); (iii) possible weakening of the Angola low. The aforementioned changes can be expected to have direct impacts on the regional climates in the study region.

  • The obliquely rotated principal component analysis, applied in a fuzzy approach, is used to classify atmospheric circulation patterns using reanalysis data, and GCM data sets under the historical experiment and 2 future emission scenarios.

  • The individual GCMs faithfully simulate the circulation patterns as obtained from ERA5 reanalysis.

  • The climate models project an increase in the intensity and frequency of occurrence of atmospheric blocking, adjacent to southern Africa.

  • Climate models project an increase in the frequency of occurrence of two distinct summer circulation types, with one associated with stronger circulation and the other associated with weaker circulation at the western branch of the Mascarene high.

The climate system is sensitive to forcing from increased greenhouse gas emissions (Zeebe 2013). With increasing atmospheric greenhouse gas concentrations, the atmospheric composition is altered, and the resulting changes in the flow of energy imply an increase in global mean near-surface temperature (Arnell et al. 2019); hydroclimatic extremes (Giorgi et al. 2018); and changes in large-scale atmospheric circulation patterns (e.g., Zappa 2019), which is the focus of this study.

Atmospheric transport both at the global and regional scale is largely governed by the distinct patterns of atmospheric circulations. The responses of large-scale circulation patterns to climate change have been reported; for example, the poleward expansion of the descent region of the Hadley circulation (Byrne & Schneider 2016); the expansion of the subtropics (Lau & Kim 2015); the narrowing of the width of the Inter-Tropical Convergence Zone (ITCZ) (Hu & Fu 2007), among others. Furthermore, while the response of the El Niño Southern Oscillation (ENSO) to radiative heating is uncertain (e.g., Latif & Keenlyside 2009; Endris et al. 2019), Müller & Roeckner (2006) found that changing the state of ENSO with increased variability can significantly impact northern hemispheric mid-latitude circulation in future climate. The Southern Annular Mode (SAM), which is a major mode of variability that impacts synoptic circulations in southern Africa, is projected to undergo a shift toward positive polarity (Gillett & Thompson 2003), implying more frequent higher sea level pressure (SLP) over the mid-latitudes, especially during the austral summer season (Ding et al. 2012). These changes in large-scale circulation patterns in response to climate change have consequences on regional weather conditions and air quality (e.g., Ibebuchi & Paeth 2021). Given the vulnerability of the southern African region to climate change (Archer et al. 2018), such as an increase in drought in the subtropical domains of southern Africa (e.g., Lazenby et al. 2018; Maúre et al. 2018; Sousa et al. 2018; Spinonia et al. 2019) and changes in the track and intensity of tropical cyclones in the southwest Indian Ocean (e.g., Malherbe et al. 2013; Fitchett et al. 2016; Muthige et al. 2018), this study is dedicated to advance the understanding of how future climate change will affect the frequency of occurrence of circulation patterns in Africa south of the equator.

The climate of southern Africa is significantly modulated by atmospheric transport and sea surface temperature anomalies in the adjacent oceans – i.e., the South Indian Ocean, the South Atlantic Ocean, and the Southern Ocean. The anticyclonic circulations at the western branch of the Mascarene high drive southeasterly moisture fluxes from the southwest Indian Ocean into the southern African mainland (Cook 2000). Conversely, the anticyclonic circulation at the South Atlantic Ocean high-pressure drives moisture from the South Atlantic east coast offshore (Reason & Smart 2015). Both high-pressure systems are semi-permanent and migrate southward (northward) during austral summer (winter), in line with the seasonal migration of the ITCZ, and consequently impact sea surface temperature anomalies (Vigaud et al. 2009). Angola low through its cyclonic circulation advects moisture (from the Angola warm current and the tropical Indian Ocean) toward the eastern subtropical regions of southern Africa (Reason & Smart 2015). The convergence of southeasterly moisture fluxes driven by the Mascarene high and the Angola low is the main mechanism that forms the South Indian Ocean Convergence Zone, which is the major synoptic system that modulates the (austral summer) hydroclimate of southern Africa (Cook 2000; Lazenby et al. 2016). Furthermore, the band of westerlies associated with the mid-latitude cyclone, during its northward track allows cold fronts to sweep across the southwestern parts of South Africa (i.e., the regions with the Mediterranean type of climate) leading to rainfall formation over the regions (Reason & Rouault 2005; Ibebuchi 2021a) and pollutant dispersion (Ibebuchi & Paeth 2021). Thus, alterations in the synoptic circulations over southern Africa can have a significant impact on the weather and climate outlook of the region, as well as atmospheric pollutant concentration.

The modes of atmospheric circulations over the African domain, in addition to their reproducibility in climate models and linkages to surface variables, have been characterized by several studies (e.g., Moron et al. 1998, 2004, 2008a, 2008b, 2015; Pohl et al. 2005; Reason & Rouault 2005). It is documented in the aforementioned works of literature that the relationship between atmospheric circulations and the local climates of the African domains is expected and appropriate climate models can faithfully simulate the observed circulation patterns over the African domain. Over the East African domain, Souverijns et al. (2016) reported that the future changes in the frequency of occurrence of circulation patterns contribute to about 23% of the total change in precipitation. However, to the author's knowledge, no study has applied an ensemble of climate models to address the projected changes in the spatial structure and frequency of occurrence of circulation types (CTs) in Africa south of the equator under future climate change scenarios. Therefore, this study applies an ensemble of 11 CMIP6 global climate models (GCMs) to study the impact of climate change on the CTs, in terms of their (i) mean shape/reproducibility (ii) and frequency of occurrence, using a higher and lower future warming scenario (i.e., SSP5-8.5 and SSP2-4.5 emission scenarios). Hence, this study adds to improving the understanding of how climate change can impact the major synoptic circulations that modulate the climate of southern Africa and beyond.

The target region for the CT classification is 0–50.25°S and 5.75–55.25°E (Figure 1). It includes the adjacent oceans that are the principal sources of moisture to the landmasses. It also captures the major synoptic features that modulate atmospheric transport over the landmasses in Africa south of the equator. These synoptic features include the western branch of the Mascarene high, the Angola low, the Mozambique Channel trough, and the cross-equatorial northeast trade wind, among others. Figure 1 shows a schematic representation of the synoptic features during their typical active periods.
Figure 1

Idealized locations of synoptic rain-bearing systems in the study region during their active periods. CW: cross-equatorial northeast trade wind; MCT: Mozambique Channel Trough; SIOHP: western branch of the South Indian Ocean high-pressure (southeast moist winds deflected southward by the Madagascar topography penetrate southern Africa through its anticyclonic circulation); AGC: Agulhas warm current (it runs from 27°S to 40°S, retroflecting at about 21°E); CF: cold fronts (it moves from west to east when the mid-latitude cyclones track northward); SAOHP: South Atlantic Ocean high-pressure (it is associated with anticyclonic circulation driving moisture offshore); AC: Angola warm current; BC: Benguela cold current; ITCZ: Inter-tropical convergence Zone during its southward track in austral summer; KL: Kalahari low; AL: Angola low; SICZ: South Indian Ocean Convergence Zone.

Figure 1

Idealized locations of synoptic rain-bearing systems in the study region during their active periods. CW: cross-equatorial northeast trade wind; MCT: Mozambique Channel Trough; SIOHP: western branch of the South Indian Ocean high-pressure (southeast moist winds deflected southward by the Madagascar topography penetrate southern Africa through its anticyclonic circulation); AGC: Agulhas warm current (it runs from 27°S to 40°S, retroflecting at about 21°E); CF: cold fronts (it moves from west to east when the mid-latitude cyclones track northward); SAOHP: South Atlantic Ocean high-pressure (it is associated with anticyclonic circulation driving moisture offshore); AC: Angola warm current; BC: Benguela cold current; ITCZ: Inter-tropical convergence Zone during its southward track in austral summer; KL: Kalahari low; AL: Angola low; SICZ: South Indian Ocean Convergence Zone.

Close modal

Large parts of southern Africa receive the highest rainfall during the austral summer season when the southwest Indian Ocean is warmer. Hence, due to enhanced moisture uptake in the southwest Indian Ocean, the western branch of the Mascarene high through its anticyclonic circulation transports more moisture from the southwest Indian Ocean, which converges with moisture from the South Atlantic Ocean and the tropical Indian Ocean, over the southern African landmass. The mid-latitude cyclones impact rainfall variability in the regions with the Mediterranean type of climate (e.g., Western Cape). The ITCZ that relates well with rainfall variability in the tropical regions modulates the Angola low (Reason & Smart 2015). During the austral winter seasons, the subtropical anticyclone is located more northward over the landmasses, leading to large-scale subsidence and suppressed rainfall (Dedekind et al. 2016).

Data

Studies have used multiple variables that are related to atmospheric circulations in classifying circulation patterns over different regions (e.g., Moron et al. 2007, 2015; Saenz & Duran-Quesada 2015). In this study, due to the correlation between the climate variables, SLP was found sufficient in classifying CTs in the study domain. Moreover, according to Kidson (1997), SLP provides a good representation of synoptic-scale systems and explains the relationship between topography and low-level flow. Daily SLP data sets are obtained from ERA5 reanalysis (Hersbach et al. 2020) from 1979 to 2014 when there is an overlap with CMIP6 (Eyring et al. 2016) historical data sets.

The CTs from the ERA5 reanalysis data that are more consistent with the observed climate are used as a reference in evaluating the simulated CTs from the CMIP6 GCMs in the historical experiment. For analysis of the changes in the CTs under future climate change, SLP data sets are obtained from the SSP2-4.5 and SSP5-8.5 emission scenarios.

The ERA5 data are obtained at a horizontal resolution of 0.25° longitude and latitude. All the available CMIP6 GCMs are examined and ranked based on their applicability for circulation typing in the study region. The ranking of the GCMs is based on their capability to replicate the circulation patterns and their characteristics as obtained from ERA5 reanalysis under the historical, SSP2-4.5 and SSP5-8.5, experiments. GCMs that replicate the circulation patterns from ERA5 are pre-selected to be suitable for this particular study. Overall, 11 CMIP6 GCMs, outlined in Table 1, passed this test and thus are used in studying the impact of future climate change on the CTs. The GCMs were obtained at a common realization (i.e., r1i1p1f1).

Table 1

Congruence match between the PC loading vectors from the four retained components and the correlation vector that indexes the highest loading magnitude at a given PC

Data setHistoricalSSP2-4.5SSP5-8.5
ACCESS-ESM1-5 0.99; 1.00; 0.98; 0.90 1.00; 0.99; 0.99; 0.91 1.00; 1.00; 1.00; 0.90 
CanESM5 1.00; 0.99; 1.00; 0.92 0.99; 0.99; 1.00; 0.90 0.99; 1.00; 0.99; 0.91 
CESM2 1.00; 0.99; 0.99; 0.91 1.00; 0.99; 0.99; 0.90 1.00; 0.99; 1.00; 0.90 
EC-EARTH3-CC 1.00; 0.99; 0.99; 0.92 0.99; 1.00; 1.00; 0.93 1.00; 0.98; 1.00; 0.95 
FGOALS-g3 1.00; 1.00; 0.99; 0.92 0.99; 1.00; 0.99; 0.91 1.00; 0.99; 0.99; 0.90 
GFDL-ESM4 1.00; 0.99; 1.00; 0.90 1.00; 0.99; 1.00; 0.94 1.00; 1.00; 1.00; 0.90 
IPSL-CM6A-LR 1.00; 1.00; 0.98; 0.93 1.00; 1.00; 0.99; 0.90 1.00; 1.00; 0.99; 0.91 
MPI-ESM1-2-HR 1.00; 0.99; 0.99; 0.90 0.94; 1.00; 0.99; 0.90 1.00; 0.99; 1.00; 0.91 
MPI-ESM1-2-LR 1.00; 0.99; 0.99; 0.91 0.99; 1.00; 0.98; 0.96 1.00; 0.99; 0.99; 0.90 
TaiESM1 0.99; 0.99; 1.00; 0.91 1.00; 0.99; 0.99; 0.90 1.00; 1.00; 0.99; 0.90 
NorESM2-LM 1.00; 1.00; 0.99; 0.90 1.00; 0.99; 1.00; 0.93 1.00; 0.99; 1.00; 0.90 
ERA5 1.00; 1.00; 1.00; 0.94 – – 
Data setHistoricalSSP2-4.5SSP5-8.5
ACCESS-ESM1-5 0.99; 1.00; 0.98; 0.90 1.00; 0.99; 0.99; 0.91 1.00; 1.00; 1.00; 0.90 
CanESM5 1.00; 0.99; 1.00; 0.92 0.99; 0.99; 1.00; 0.90 0.99; 1.00; 0.99; 0.91 
CESM2 1.00; 0.99; 0.99; 0.91 1.00; 0.99; 0.99; 0.90 1.00; 0.99; 1.00; 0.90 
EC-EARTH3-CC 1.00; 0.99; 0.99; 0.92 0.99; 1.00; 1.00; 0.93 1.00; 0.98; 1.00; 0.95 
FGOALS-g3 1.00; 1.00; 0.99; 0.92 0.99; 1.00; 0.99; 0.91 1.00; 0.99; 0.99; 0.90 
GFDL-ESM4 1.00; 0.99; 1.00; 0.90 1.00; 0.99; 1.00; 0.94 1.00; 1.00; 1.00; 0.90 
IPSL-CM6A-LR 1.00; 1.00; 0.98; 0.93 1.00; 1.00; 0.99; 0.90 1.00; 1.00; 0.99; 0.91 
MPI-ESM1-2-HR 1.00; 0.99; 0.99; 0.90 0.94; 1.00; 0.99; 0.90 1.00; 0.99; 1.00; 0.91 
MPI-ESM1-2-LR 1.00; 0.99; 0.99; 0.91 0.99; 1.00; 0.98; 0.96 1.00; 0.99; 0.99; 0.90 
TaiESM1 0.99; 0.99; 1.00; 0.91 1.00; 0.99; 0.99; 0.90 1.00; 1.00; 0.99; 0.90 
NorESM2-LM 1.00; 1.00; 0.99; 0.90 1.00; 0.99; 1.00; 0.93 1.00; 0.99; 1.00; 0.90 
ERA5 1.00; 1.00; 1.00; 0.94 – – 

The congruence coefficients are arranged in descending order from the first to the fourth component. Further information on the GCMs is available from Eyring et al. (2016).

Table 2

Congruence match between the SLP composite anomaly patterns of the CTs from ERA5 and their counterparts as simulated by the individual GCMs

DataCT1+CT1−CT2+CT2−CT3+CT3−CT4+CT4−Average
ACCESS 0.98 0.97 0.98 0.97 0.98 0.98 0.97 0.97 0.98 
CanESM5 0.97 0.95 0.97 0.95 0.95 0.98 0.94 0.95 0.96 
CESM2 0.99 0.99 0.99 0.99 0.97 0.99 0.98 0.99 0.99 
EC-EARTH 0.99 0.98 0.96 0.99 0.97 0.98 0.98 0.97 0.98 
FGOALS 0.98 0.97 0.98 0.98 0.96 0.99 0.97 0.97 0.98 
GFDL 0.98 0.96 0.97 0.97 0.96 0.97 0.98 0.95 0.97 
IPSL 0.99 0.96 0.97 0.98 0.97 0.98 0.97 0.96 0.98 
MPI-ESM-HR 0.99 0.98 0.96 0.99 0.98 0.98 0.99 0.99 0.98 
MPI-ESM-LR 0.99 0.98 0.97 0.99 0.97 0.99 0.98 0.98 0.98 
TaiESM1 0.99 0.98 0.95 0.99 0.95 0.98 0.98 0.98 0.98 
NorESM2 0.99 0.98 0.98 0.99 0.98 0.99 0.93 0.96 0.98 
Ensemble median 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 
DataCT1+CT1−CT2+CT2−CT3+CT3−CT4+CT4−Average
ACCESS 0.98 0.97 0.98 0.97 0.98 0.98 0.97 0.97 0.98 
CanESM5 0.97 0.95 0.97 0.95 0.95 0.98 0.94 0.95 0.96 
CESM2 0.99 0.99 0.99 0.99 0.97 0.99 0.98 0.99 0.99 
EC-EARTH 0.99 0.98 0.96 0.99 0.97 0.98 0.98 0.97 0.98 
FGOALS 0.98 0.97 0.98 0.98 0.96 0.99 0.97 0.97 0.98 
GFDL 0.98 0.96 0.97 0.97 0.96 0.97 0.98 0.95 0.97 
IPSL 0.99 0.96 0.97 0.98 0.97 0.98 0.97 0.96 0.98 
MPI-ESM-HR 0.99 0.98 0.96 0.99 0.98 0.98 0.99 0.99 0.98 
MPI-ESM-LR 0.99 0.98 0.97 0.99 0.97 0.99 0.98 0.98 0.98 
TaiESM1 0.99 0.98 0.95 0.99 0.95 0.98 0.98 0.98 0.98 
NorESM2 0.99 0.98 0.98 0.99 0.98 0.99 0.93 0.96 0.98 
Ensemble median 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 

The Ensemble median is the median of the SLP composite patterns of the CTs classified from each of the individual GCMs.

Methodology

The classification of the CTs is completely eigenvector-based. The rotated principal component analysis (PCA) time decomposition technique (Compagnucci et al. 2001) is used to characterize the spatiotemporal SLP variability patterns in the study region, and then each day's SLP pattern (in the analysis period) is assigned to the classified SLP variability pattern(s) the day has the most correspondence to, using correlations as a similarity measure. Obliquely rotated PCA applied to the T-mode (i.e., column matrix contains the daily time series in the analysis period and row matrix contains the grid points in the study domain) matrix of SLP (Richman 1981, 1986; Zhang et al. 2012; Li et al. 2021a, 2021b; Ibebuchi 2022, 2023; Ibebuchi & Abu 2023) is applied in classifying the CTs. The correlation matrix contains the correlation coefficients between time series in the T-mode matrix. Singular value decomposition is used to factorize the correlation matrix resulting in the PC scores, eigenvalues, and eigenvectors. The PC scores are spatial variability patterns and the eigenvectors localize the spatial patterns in time (Compagnucci & Richman 2008). To make the eigenvectors responsive to rotation, they are multiplied by the square root of their corresponding eigenvalues, so that they become PC loadings. The loadings are rotated obliquely (to relax the orthogonality constraint on the PC scores, since the actual climate system is not orthogonal) and iteratively using the Promax rotation method (Hendrickson & White 1964). The Promax rotation method is an oblique simple structure rotation method that is applied in this work to maximize the number of near-zero PC loadings and fewer larger loadings, thus making the PC loadings unique, match the correlation patterns, and become more amendable to physical interpretations (Richman 1986). The Promax rotation method is applied at a power of 2 (i.e., the power that the Varimax solution is raised) and above, and keeping 2 components and above. The number of rotated components and Promax solution with the largest congruence matches (Equation (1)) between the PC loading vectors and the correlation vectors that the PC loadings are indexed to, is marked as the optimal number of components to retain, and an optimal Promax solution at which the rotated loadings fit the correlation pattern (Ibebuchi & Richman 2022).
formula
(1)
where I is the congruence coefficient, and X and Y are two distinct (PC loading) vectors. In the case of deciding the number of PCs to retain, X is the rotated PC loading and Y is the correlation vector (i.e., the date from the T-mode correlation matrix) that indexes the highest loading magnitude on that particular PC.

The absolute value of the loadings represents an important signal and thresholds within the range of 0.2–0.35 can be sufficient to separate noise (i.e., loadings that are close to zero) from the signal (Richman & Gong 1999; Ibebuchi & Richman 2022). Following its efficacy, in previous studies (e.g., Ibebuchi 2021a), ±0.2 is used in this study to assign a day to CTs. Each retained component gives two classes (i.e., clusters above or below the ±0.2 threshold) and the mean SLP of the days in a given class designates the spatial map of the CT. Since atmospheric circulation is a continuum (Compagnucci & Richman 2008), more than one CT occurring at a given time is realistic. This implies that as the simple structure simplification of the loadings permits, a day can have its loading magnitude greater than |0.2| under more than one retained component, and thus, more than one CT can be considered on a given day.

The decision that CTs from different data sets are truly counterparts is based on (i) a high congruence coefficient (>0.93) between the component scores and between the SLP composite anomaly patterns of the CTs; (ii) visual inspection of the SLP composite patterns; and (iii) replication of the frequency distribution of the corresponding CTs (from reanalysis). These criteria have to be in perfect agreement to conclude that two CTs from different data sets are counterparts. Also, the classification is applied to each of the SLP data sets from ERA5, and the individual selected GCMs, under each of the experiments.

The impact of future climate change is investigated on (i) the reproducibility and shapes of the SLP composite patterns; (ii) the annual cycle of the CTs; and (iii) the annual frequency of occurrence of the CTs.

Optimal number of retained components

Using the criterion of matching the PC loadings to the correlation patterns, an optimal number of four components are retained at a Promax solution of m = 4 (i.e., the power at which the Varimax solution is raised). Table 1 shows the output of the congruence match between the rotated PC loadings vectors and the correlation vectors, applied to the ERA5 classification and the classification from the individual GCMs. For some of the GCMs, the fourth component has congruence matches below 0.92. However, the ERA5 classification was used as the reference in deciding the optimal number of components to keep.

CTs in Africa south of the equator

Figure 2 shows the classified eight CTs. Each CT designates a pattern of atmospheric circulation in the study region. Based on their probability of occurrence, CT1+, CT2−, CT3+, and CT4− are relatively the dominant CTs (Supplementary Material, Fig. A1). They can be interpreted as the dominant states of the atmosphere. Furthermore, while the CTs can occur at any time of the year, they can also be relatively dominant at specific season(s). CT1+, CT2−, CT3−, and CT4+ are dominant during the colder seasons (i.e., austral winter, June to August (JJA), and the adjacent months). Conversely, CT1−, CT2+, CT3+, and CT4− are dominant during the warmer seasons (i.e., austral summer, December to February (DJF), and the adjacent months). For the austral summer dominant CTs, CT1− and CT3+ are characterized as blocking (BL) CTs, given that the subtropical anticyclone blocks the Southern Hemisphere mid-latitude cyclones, adjacent to the south of South Africa. The major difference between the structure of the summer BL CTs is that under CT3+, the southeast winds are driven by the enhanced anticyclonic circulation at the western branch of the Mascarene high (Supplementary Material, Fig. A3), while under CT1−, the ridging of the South Atlantic anticyclone, south of South Africa, drives southeast winds into the eastern subtropical regions (Supplementary Material, Fig. A2). Under CT2+, Figure 2 shows that the Mascarene high is enhanced at its western branch and appears to cause northward deformation of the mid-latitude cyclone (to move toward the south of South Africa). Hence, the southeast moisture fluxes driven by the Mascarene high are buffered by westerly moisture fluxes driven by the enhanced cyclonic system (Supplementary Material, Fig. A2). Under CT4−, cyclonic circulation (i.e., low-pressure system) is enhanced in the southwest Indian Ocean; the western branch of the Mascarene high is weak, and westerly moisture fluxes rather dominate over southern parts of the southwest Indian Ocean. The major sources of moisture to the landmasses during CT4− are northwest moisture fluxes from the tropical South Atlantic Ocean, driven by the Angola low, and the cross-equatorial northeast winds (Supplementary Material, Fig. A3).
Figure 2

SLP composite patterns of the CTs from ERA5 and ensemble median of the 11 GCMs under the historical analysis. The ensemble median is calculated as the median of the SLP composite patterns of the CTs simulated by each of the 11 GCMs.

Figure 2

SLP composite patterns of the CTs from ERA5 and ensemble median of the 11 GCMs under the historical analysis. The ensemble median is calculated as the median of the SLP composite patterns of the CTs simulated by each of the 11 GCMs.

Close modal

For the winter dominant CTs, CT1+, CT2−, and CT3− have distinct synoptic structures but are similar because of the enhanced westerly fluxes over the mid-latitudes (Supplementary Material, Figs. A2 and A3). This is more expressed under CT3−: the mid-latitude cyclone is strongest, displacing the high-pressure over South Africa, and westerly winds dominate over the subtropical and mid-latitude regions. Furthermore, for all winter CTs, cyclonic activity is lowest in the southwest Indian Ocean (i.e., less convection). CT4+ is also a winter BL pattern but with the subtropical anticyclone more situated over the subtropical landmasses relative to the summer BL CTs. With the enhanced cyclonic activity in the southwest Indian Ocean, coupled with the convergence of moisture fluxes, the summer dominant CTs bring wetter conditions, and with enhanced subsidence and weaker moisture advection from the southwest Indian Ocean, the winter CTs bring drier conditions over southern Africa (Ibebuchi 2021a).

Capability of GCM ensemble to replicate the CTs

As highlighted in Ibebuchi (2021b), when the classification method is applied to the individual (appropriate) GCMs, the circulation patterns are replicated as obtained from reanalysis data. Given that GCMs are characterized by systematic biases and also have different responses to future climate change, in analyzing future changes in the CTs, the ensemble of GCMs listed in Table 1 is preferred over a single GCM in achieving a better and more robust projection. The representation of the ERA5 CTs and the characteristics of the CTs, in the historical experiments, are examined for the individual GCMs, and the ensemble median of the CTs as simulated by the individual GCMs. In the hindcast evaluation, the performance of the ensemble median is compared to using the individual GCMs. Table 2 shows that the selected GCMs replicate the SLP composite patterns of the CTs, with a congruence match generally greater than 0.92. The performances vary among the GCMs and the CTs. For example, CT4+ is the least represented under ACCESS GCM. On average, CESM2 best represents the SLP composite patterns of the CTs obtained from ERA5. The added value of the ensemble median is notable in Table 2, as the congruence match is 0.99 in all cases (i.e., an excellent match). Thus, from the hindcast evaluation, there is a better representation of the CTs in the ensemble median over any of the ensemble members. From Figure 2, though there are small differences in the structure of the isopleths, a cursory investigation also validates that the pattern configuration of the simulated CTs from the ensemble median is quite comparable to the CTs from ERA5.

Figure 3 shows the inter-annual variability in the amplitude (i.e., the PC loadings) of the classified CTs from ERA5, the simulated CTs from the individual GCMs, and the ensemble median (i.e., the middle value of the PC loadings from each of the 11 GCMs). Recall that the PC loadings are used in creating the CTs (i.e., both the composite maps and the frequency statistics of the CTs). Though the time sequence of the GCMs does not synchronize with the observed climate, nonetheless on average, the pattern configuration of the simulated PCs is expected to be comparable with that from ERA5. Similar to the SLP composite patterns of the CTs in Figure 2, the spread of the ensemble members is reduced by taking the ensemble median of the statistic, and it is interesting that the ensemble median from Figure 3 faithfully captures the phase shift in the PC loadings. For example, in PC1 and PC4, the shift to the negative phase toward the end of the analysis period is captured in the ensemble statistic (Figure 3). Thus, in the future projections, it is validated that using the ensemble median of the statistic reduces the uncertainty in the projections from the ensemble members and gives the most probable response of the CTs to future climate change, compared to that from any single contributing model. Similarly, the probability of occurrence of the CTs and the patterns of moisture circulations as obtained from ERA5 CTs are also well replicated in the ensemble median of the GCMs (cf. Supplementary Material, Figs. A1, A2, and A3).
Figure 3

Annual mean of the loadings of the four retained principal components from ERA5, the 11 GCMs, and the ensemble median of the 11 GCMs under the historical analysis. The ensemble median is calculated as the median of the annual mean of the loadings simulated by each of the 11 GCMs. The annual mean is calculated as the average of the daily PC loadings per year.

Figure 3

Annual mean of the loadings of the four retained principal components from ERA5, the 11 GCMs, and the ensemble median of the 11 GCMs under the historical analysis. The ensemble median is calculated as the median of the annual mean of the loadings simulated by each of the 11 GCMs. The annual mean is calculated as the average of the daily PC loadings per year.

Close modal

Projected changes in the CTs under future warming scenarios

Figure 4 and Table 3 show that the simulated CTs under the different GCM experiments considered in this work do not deviate from what is observed in the current climate using the reanalysis data as reference (congruence match of the SLP composite patterns >0.98 in all cases indicating an excellent match). The results suggest that the reproducibility of the CTs is not constrained by future climate change under the considered emission scenarios. However, there are observable changes in the isopleth of the composite patterns of the CTs under the future warming scenarios compared to the historical ones. Hence, while future climate change will not impact the existence/stationarity of the CTs, localized changes in geographical synoptic features associated with the CTs are plausible. To further investigate projected changes in the spatial structure of the CTs, Figure 5 shows the difference between the historical CTs and their counterparts under future climate change. It can be seen that the spatial structure of the composite patterns of the CTs changed in the future emission scenarios, relative to the historical patterns, and the change is more robust under the higher warming scenario. In Supplementary Material, Figs. A2 and A3, it can be seen from the austral summer dominant CTs that when the Angola low is well expressed (e.g., CT1−, CT2+, and CT4−), northwest winds are evident. The westerly winds typically transport the moisture feeding into the low to the subtropical parts of southern Africa. Figure 5 shows an anomalous increase in SLP over Angola under the future climate change scenarios, suggesting a weakening of the cyclonic system over Angola. The weakening of the Angola low can impact the rate at which westerly moisture fluxes converge with easterly moisture fluxes driven by the western branch of the Mascarene high. This can have a direct impact on the strength of the South Indian Ocean convergence Zone.
Table 3

Congruence match between the SLP composite anomaly patterns of the CTs from the ERA5 and the counterparts from the ensemble median SLP composite anomaly patterns of the simulated CTs under different emission scenarios

DataCT1+CT1−CT2+CT2−CT3+CT3−CT4+CT4−
    ERA5     
Historical 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 
SSP2-4.5 0.99 0.99 0.97 0.97 0.99 0.99 0.99 0.99 
SSP5-8.5 0.99 0.99 0.99 0.99 0.98 0.99 0.98 0.99 
    Historical     
SSP2-4.5 0.99 0.99 0.96 0.97 0.99 0.99 0.99 0.99 
SSP5-8.5 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 
DataCT1+CT1−CT2+CT2−CT3+CT3−CT4+CT4−
    ERA5     
Historical 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 
SSP2-4.5 0.99 0.99 0.97 0.97 0.99 0.99 0.99 0.99 
SSP5-8.5 0.99 0.99 0.99 0.99 0.98 0.99 0.98 0.99 
    Historical     
SSP2-4.5 0.99 0.99 0.96 0.97 0.99 0.99 0.99 0.99 
SSP5-8.5 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 
Figure 4

SLP composite patterns of the CTs from ERA5 and the ensemble median of the 11 GCMs under the historical analysis and SSP5-8.5 and SSP2-4.5 emission scenarios. The ensemble median is calculated as the median of SLP composite patterns of the CTs simulated by each of the 11 GCMs under each of the experiments.

Figure 4

SLP composite patterns of the CTs from ERA5 and the ensemble median of the 11 GCMs under the historical analysis and SSP5-8.5 and SSP2-4.5 emission scenarios. The ensemble median is calculated as the median of SLP composite patterns of the CTs simulated by each of the 11 GCMs under each of the experiments.

Close modal
Figure 5

Difference between the SLP composite patterns of the historical CTs and their counterparts under the future emission scenarios (i.e., SLP composite of the CTs under future climate change minus the SLP composites of the same CTs under the historical analysis). Only values exceeding the 90% confidence limit based on the permutation test are plotted.

Figure 5

Difference between the SLP composite patterns of the historical CTs and their counterparts under the future emission scenarios (i.e., SLP composite of the CTs under future climate change minus the SLP composites of the same CTs under the historical analysis). Only values exceeding the 90% confidence limit based on the permutation test are plotted.

Close modal

An increase in the anti-circulation at the western branch of the Mascarene high is also evident in Figure 5 under CT3+. Under CT1−, CT4−, and CT2−, increased blocking adjacent to the south of South Africa is evident and the same goes for CT3−, associated with enhanced cyclonic activity over the mid-latitudes (cf. Figure 1 and Supplementary Material, Fig. A3). An increase in cyclonic activity adjacent to South Africa is only projected under CT2+, which can be a result of enhanced deformation of the cyclonic system by the stronger Mascarene high (cf. Figure 1). Under CT4+, which is a dominant winter BL CT, subsidence (i.e., positive SLP anomaly) is projected to increase under the future emission scenarios. In mostly all CTs, a negative SLP anomaly is equally projected in the tropical oceans. Moreover, under CT1− and CT4−, the projected increase in cyclonic circulation (i.e., negative SLP) is also evident over the southwest Indian Ocean. Again, these changes are more robust under the higher emission scenario.

Figure 6 shows the annual cycle of the ERA5 CTs and the annual cycle of the simulated CTs from the ensemble median of the statistic as simulated by the individual GCMs. From Figure 6, in the hindcast evaluation, while the GCM ensemble captures the structure of the annual cycle of the CTs (i.e., the dominant months/seasons of the CTs), the GCMs have difficulty in replicating the magnitude of the relative occurrence of some of the CTs as obtained from ERA5, in some specific months. For example, during October, there is a notable underestimation of the relative frequency of CT1− in the GCM ensemble relative to ERA5. A similar underestimation of the relative frequency of occurrence of the CTs can be seen during May and January in CT3−.
Figure 6

Annual cycle of the CTs from ERA5 and ensemble median of the 11 GCMs under the historical analysis and SSP5-8.5 and SSP2-4.5 emission scenarios. The ensemble median is calculated as the median of the annual cycle simulated by each of the 11 GCMs.

Figure 6

Annual cycle of the CTs from ERA5 and ensemble median of the 11 GCMs under the historical analysis and SSP5-8.5 and SSP2-4.5 emission scenarios. The ensemble median is calculated as the median of the annual cycle simulated by each of the 11 GCMs.

Close modal

Under the future emission scenarios, there are changes in the relative frequency of occurrence of the CTs at specific months, but the structure of the annual cycle (i.e., the dominant periods of the CTs) is the same and barely altered by future climate change. For example, under CT1− that is a summer BL CT, the relative frequency of occurrence under the emission scenarios increased from December to May (i.e., austral summer to autumn seasons). The months for the projected increase in CT3+, which is also a summer BL CT, are from May to November when the southwest Indian Ocean is relatively cooler. During the warmer months, CT4− associated with the enhanced cyclonic activity in the southwest Indian Ocean is projected to be more frequent (Figure 6), which can be a possible reason why the Mascarene high under CT3+ might experience periods with low occurrence during summer seasons. Also with the increase of CT3+ during the colder months, CT3− associated with the enhanced cyclonic activity adjacent to South Africa is projected to be less frequent during the colder months – under the emission scenarios – relative to the historical climate.

Figure 7 shows the annual frequency of occurrence of the CTs (i.e., the count of days with loadings greater than |0.2| under each retained PC, or in other words, when the CT occurred and can be associated with a signal). First, in the historical experiment, and in the emission scenarios (especially toward the end of the 21st century, when the change signal is highest), the frequency of occurrence of some of the CTs changed in the historical climate and is projected to change in the future emission scenarios. Second, from Figure 7, the response of the frequency of occurrence of the CTs to future climate change is different under the two emission scenarios, suggesting that warming magnitude influences to what extent the frequency distribution of some of the CTs is impacted. Figure 8 quantifies the difference in the projected frequency of occurrence of the CTs between the SSP5-8.5 and SSP2-4.5 emission scenarios. For example, toward the end of 21 century (i.e., from 2070), CT1+ that is the most frequent CT is projected to have an average decrease of about 6 days in the SSP5-8.5 emission scenario relative to the SSP2-4.5 emission scenario. Similarly, CT3+ that is the most frequent summer CT is projected to have an average increase of about 4 days toward the end of the 21st century under the SSP5-8.5 emission scenario relative to the SSP2-4.5 emission scenario.
Figure 7

Annual frequency of occurrence of the simulated CTs from the median of the 11 GCMs under the historical experiment and SSP5-8.5 and SSP2-4.5 emission scenarios. The ensemble median is calculated as the median of the annual frequency of occurrence simulated by each of the 11 GCMs under the historical, SSP5-8.5 and SSP2-4.5 scenarios, respectively, from 1979 to 2100. The vertical black line separates the historical time series (i.e., 1979–2014) and the time series from the emission scenarios (i.e., 2015–2100).

Figure 7

Annual frequency of occurrence of the simulated CTs from the median of the 11 GCMs under the historical experiment and SSP5-8.5 and SSP2-4.5 emission scenarios. The ensemble median is calculated as the median of the annual frequency of occurrence simulated by each of the 11 GCMs under the historical, SSP5-8.5 and SSP2-4.5 scenarios, respectively, from 1979 to 2100. The vertical black line separates the historical time series (i.e., 1979–2014) and the time series from the emission scenarios (i.e., 2015–2100).

Close modal
Figure 8

Difference between the frequency of occurrence of the simulated CTs from the SSP5-8.5 and SSP2-4.5 emission scenarios. The difference is calculated as the annual frequency of occurrence of the CT in the SSP5-8.5 scenario minus the annual frequency of occurrence of the same CT in the SSP2-4.5 scenario.

Figure 8

Difference between the frequency of occurrence of the simulated CTs from the SSP5-8.5 and SSP2-4.5 emission scenarios. The difference is calculated as the annual frequency of occurrence of the CT in the SSP5-8.5 scenario minus the annual frequency of occurrence of the same CT in the SSP2-4.5 scenario.

Close modal

Furthermore, using the modified Mann–Kendall test (Mann 1945; Kendall 1975; Yue & Wang 2004) that is both non-parametric and applicable to data that are serially correlated, Table 4 shows the test of statistical significance of linear trends in the annual frequency of occurrence of the CTs under all the considered experiments. There is a statistically significant decrease in the frequency of occurrence of CT1+ in the historical climate (both from ERA5 and the GCM ensemble). Also, toward the end of the 21st century under the SSP5-8.5 scenario, the frequency of occurrence of CT1+ is projected to decrease (Table 4). From Supplementary Material, Table A1, six of the ensemble members agree with ERA5 on the significant decreasing trend in the frequency of occurrence of CT1+ under the historical climate. Also, from Supplementary Material, Table A2, eight of the ensemble members agree on the significant projected decrease in the frequency of occurrence of CT1+. CT1− is a summer BL CT. Table 4 shows that the increase in the frequency of occurrence of CT1− is statistically significant both under the historical and SSP5-8.5 experiments. From Supplementary Material, Table A1, six of the ensemble members agree with ERA5 on the significant increased frequency of occurrence of CT1−; and eight of the ensemble members agree on the significant projected increase in the frequency of occurrence of CT1− (Supplementary Material, Table A2).

Table 4

Linear trend from the modified Mann–Kendall test of statistical significance of linear trends in the annual frequency of occurrence of the CTs from ERA5, historical experiment, and future climate change scenarios

CTERA5 (1979–2014)Historical (1979–2014)SSP2-4.5 (2040–2100)SSP5-8.5 (2040–2100)
CT1+ −0.00* −0.00* 0.25 −0.00* 
CT1− +0.00* +0.01* 0.05 +0.00* 
CT2+ 0.82 0.53 0.44 0.102 
CT2− 0.58 0.12 0.39 0.339 
CT3+ 0.50 +0.00* 0.68 +0.00* 
CT3− 0.28 0.20 0.25 −0.00* 
CT4+ −0.00* 0.0* −0.00* −0.00* 
CT4− +0.00* +0.04* 0.40 +0.00* 
CTERA5 (1979–2014)Historical (1979–2014)SSP2-4.5 (2040–2100)SSP5-8.5 (2040–2100)
CT1+ −0.00* −0.00* 0.25 −0.00* 
CT1− +0.00* +0.01* 0.05 +0.00* 
CT2+ 0.82 0.53 0.44 0.102 
CT2− 0.58 0.12 0.39 0.339 
CT3+ 0.50 +0.00* 0.68 +0.00* 
CT3− 0.28 0.20 0.25 −0.00* 
CT4+ −0.00* 0.0* −0.00* −0.00* 
CT4− +0.00* +0.04* 0.40 +0.00* 

*Statistical significance at a 95% confidence level (i.e., p-value <0.05).Positive (negative) sign implies increasing (decreasing) trend.

Though some ensemble members predict significant changes in CT2+ and CT2− (Supplementary Material, Tables A1 and A2), based on the median statistic and the ERA5 CTs, no significant change was found in the frequency of occurrence of CT2+ and CT2− in all experiments considered here. The asymmetric CTs from PC3, i.e., CT3+/CT3−, associated with suppressed/enhanced westerly moisture fluxes in the Southern Hemisphere mid-latitudes (Supplementary Material, Fig. A2), were reported in Ibebuchi (2021a) to be significantly modulated by the anomalies of the SAM. CT3+(CT3−) is related to positive (negative) SAM. In the historical experiment, the ensemble median of the GCMs simulates a significant increase in the frequency of occurrence of CT3+ (Table 4), and nine of the ensemble members agree on the significant increase (Supplementary Material, Table A1). Under the SSP5-8.5 scenario, the frequency of occurrence of CT3+ (i.e., summer dominant BL CT) is projected to increase toward the end of the 21st century (Supplementary Material, Fig. A4), and nine of the ensemble members agree on the significant increase (Supplementary Material, Table A2). Conversely, the frequency of occurrence of CT3− is projected to significantly decrease under the SSP5-8.5 emission scenario, and nine of the ensemble members agree on the significant decrease (Supplementary Material, Table A2).

The frequency of occurrence of CT4− that is an austral summer CT associated with the enhanced cyclonic activity in the southwest Indian Ocean has a positive trend under the historical experiment – both from the GCM ensemble and ERA5 (Table 4). Five of the ensemble members agree with ERA5 on the significant increase in the frequency of CT4− (Supplementary Material, Table A1). Toward the end of the 21st century, the frequency of occurrence of CT4− is projected to significantly increase under the SSP5-8.5 scenario, and 10 of the ensemble members agree on the significant increase (Supplementary Material, Table A2). CT4+ is a winter BL CT. While from Figure 5, the magnitude of subsidence is projected to increase under CT4+, in the historical climate, the frequency of CT4+ has a negative significant trend both from ERA5 and the GCM ensemble (of which, five of the ensemble members agree with ERA5). The frequency of occurrence of CT4+ is projected to decrease under both the SSP5-8.5 scenario and the SSP2-4.5 scenario. All the ensemble members agree on the significant decrease in the frequency of occurrence of CT4+ under the SSP5-8.5 scenario (Supplementary Material, Table A2), and six of the ensemble members agree on the significant decrease under the SPP2-4.5 scenario (Supplementary Material, Table A3).

Several studies have indicated that future climate change can be expected to impact regional patterns of atmospheric circulations (Müller & Roeckner 2006; Hu & Fu 2007). This can have a wide range of consequences ranging from weather extremes to the modification of pollutant transport (Schlef et al. 2019; Marshall et al. 2020; Ibebuchi & Paeth 2021). Changes in the frequency distribution of regional atmospheric circulation patterns can be a function of variations in the leading modes of climate variability (e.g., ENSO and SAM) that can have a remote influence on the regional circulation patterns (Müller & Roeckner 2006). Studies have reported that anthropogenic climate change might alter the amplitude and frequency of ENSO and the SAM (Collins 2000; Wang & Cai 2013; Cai et al. 2021). Consequently, it can be expected that through the impact of anthropogenic climate change on ENSO and SAM, the related regional modes of atmospheric circulation will be impacted through a series of feedbacks.

Within the regional context of Africa, south of the equator, this study showed that while large-scale atmospheric circulation patterns are stationary, based on their existence regardless of the analyzed climate, their spatial structure and frequency of occurrence can be impacted by future climate change. Under the SSP5-8.5 emission scenario, the significant increase in the frequency of occurrence of summer BL CTs implies that the ensemble of the climate models projects a possible increase in the frequency of atmospheric blocking of the Southern Hemisphere mid-latitude cyclones, adjacent to South Africa. The shift of the SAM toward its positive phase during the austral summer season (Ding et al. 2012) might be implicated as a driver of the projected frequent increase in SLP, adjacent to South Africa. This is because positive SAM that correlates with the summer BL CTs (e.g., Ibebuchi 2021a) is associated with increased SLP in the mid-latitudes. Increased SLP variability at the Southern Hemisphere mid-latitudes can have a direct implication on the hydroclimate of the Western Cape (Reason & Rouault 2005) and the climatology of the Mascarene high (Morioka et al. 2015). Since rainfall formation and pollutant dispersion at parts of the Western Cape are associated with the band of westerlies linked to the northward track of the mid-latitude cyclones, it is thus projected that under the higher warming scenario, parts of the Western Cape become drier, coupled with higher pollutant concentrations (e.g., Reason & Rouault 2005; Ibebuchi 2021a; Ibebuchi & Paeth 2021).

Furthermore, increased anticyclonic circulation at the western branch of the Mascarene high was projected under CT3+ (Figure 5), implying an increase in southeasterly moisture fluxes advected into the southern African mainland. Nonetheless, the climate models also projected an increase in cyclonic activity in the southwest Indian Ocean and a weaker circulation at the western branch of the Mascarene high under CT4− (Figure 5), which was also projected to be more frequent under future climate change. Given that the CTs are continuous and their distinct signals can interact at a given time, the combined changes in CT3+ and CT4− suggest that under future climate change, the climatology of the western branch of the Mascarene high might be subjected to alternating periods of being anomalously stronger due to a more positive SAM that increases SLP at the Mascarene high – as reported by Morioka et al. (2015) – and periods of being anomalously weaker due to warmer southwest Indian Ocean.

Already, there is high confidence that projected precipitation decreases in the western parts of southern Africa, coupled with an increase in the frequency of drought (Engelbrecht et al. 2015; Liu et al. 2018). Also, annual mean rainfall in the austral summer rainfall region is projected to decrease by 10–20%, coupled with an increase in the occurrence of consecutive dry days during the wet season under high emission scenarios (Kusangaya et al. 2014; Engelbrecht et al. 2015). The projected weaker Angola low under all the CTs (Figure 5) suggests that the rate at which northwesterly moisture fluxes are transported to the subtropical regions of southern Africa might be weakened. Also under CT4+, during the winter seasons, the magnitude of subsidence is projected to increase over the study region (Figure 5), which can result in drier conditions. Weakening of the Angola low, increase in subsidence, atmospheric blocking of the mid-latitude storms that bring rainfall to the southwestern regions, and (occasional) weakening of the Mascarene high can result in drier conditions over the Southern Africa region. Furthermore, while heavy precipitation in the southwestern parts of southern Africa is projected to decrease (e.g., Donat et al. 2019), it is also projected that heavy precipitation increases in the eastern parts of southern Africa under future warming scenarios (e.g., Li et al. 2021a, 2021b). The projected increase in (the magnitude and frequency of) convective/cyclonic activity in the southwest Indian Ocean under CT4− can result in heavy rainfall in the eastern regions when atmospheric conditions are favorable, due to enhanced moisture availability in the adjacent oceans of the eastern coastal regions, during the summer months.

This study examined the impact of future climate change on the atmospheric circulation patterns in Africa, south of the equator. The major conclusions of this study are:

  • The classified CTs are stationary based on their existence and reproducibility under future emission scenarios. However, the spatial structure of the CTs and so the strength of atmospheric circulation associated with the synoptic geographical features captured in the CTs can be impacted by climate change.

  • Under future climate change, there were marked changes in the monthly relative frequency of occurrence of the CTs; nonetheless, the dominant period of the CTs was barely affected. The annual frequency of occurrence of some of the CTs has started changing in the historical climate. Under future emission scenarios, the frequency of occurrence of some of the CTs is projected to change with different magnitudes dependent on the warming scenario. Specifically, the annual frequency of occurrence of summer blocking CTs is projected to increase under higher warming scenarios. Also, the frequency of the CT associated with the enhanced cyclonic activity in the southwest Indian Ocean and a weaker western branch of the Mascarene high (Figure 1) is projected to increase under the higher warming scenario.

The wider implication of this study is that under future climate change, (i) the poleward expansion of the subtropics over and adjacent to South Africa can be expected due to increased blocking of the mid-latitude cyclones by the subtropical high-pressure systems; (ii) southeast moisture fluxes driven by the western branch of the Mascarene high can be subjected to periods of being anomalous stronger (weaker) possibly due to a more positive SAM (warmer southwest Indian Ocean); (iii) the Angola low can be expected to weaken.

Finally, the major limitation of this study is that depending on the analyzed GCM, the simulated circulation patterns have different responses to future climate change (cf. Supplementary Material, Tables A1, A2, and A3), which is a source of uncertainty in the projected changes. In future work, an ensemble of high-resolution regional climate models will be used to study how changes in the atmospheric circulations (e.g., projected changes in moisture fluxes) under the classified CTs affect the spatiotemporal variability of a range of other climate variables in southern Africa.

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

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

Archer
E.
,
Engelbrecht
F.
,
Hänsler
A.
,
Landman
W.
,
Tadross
M.
,
Helmschrot
J.
,
2018
Seasonal prediction and regional climate projections for southern Africa
. In:
Climate Change and Adaptive Land Management in Southern Africa – Assessments, Changes, Challenges, and Solutions
(
Revermann
R.
,
Krewenka
K. M.
,
Schmiedel
U.
,
Olwoch
J. M.
,
Helmschrot
J.
&
Jürgens
N.
, eds).
Biodiversity & Ecology, 6, Klaus Hess Publishers
,
Göttingen
,
Windhoek
, pp.
14
21
.
doi:10.7809/b-e.00296
.
Arnell
N. W.
,
Lowe
J. A.
,
Challinor
A. J.
&
Osborn
T. J.
2019
Global and regional impacts of climate change at different levels of global temperature increase
.
Climatic Change
155
,
377
391
.
Byrne
M. P.
&
Schneider
T.
2016
Narrowing of the ITCZ in a warming climate: physical mechanisms
.
Geophysical Research Letters
43
,
350
357
.
Cai
W.
,
Santoso
A.
,
Collins
M.
,
Dewitte
B.
,
Karamperidou
C.
,
Kug
J.-S.
,
Lengaigne
M.
,
McPhaden
, M. J.
,
Stuecker
M. F.
,
Taschetto
A. S.
,
Timmermann
A.
,
Wu
L.
,
Yeh
S.-W.
,
Wang
G.
,
Ng
B.
,
Jia
F.
,
Yang
Y.
,
Ying
J.
,
Zheng
X.-T.
,
Bayr
T.
,
Brown
J. R.
,
Capotondi
A.
,
Cobb
K. M.
,
Gan
B.
,
Geng
T.
,
Ham
Y.-G.
,
Jin
F.-F.
,
Jo
H.-S.
,
Li
X.
,
Lin
X.
,
McGregor
S.
,
Park
J.-H.
,
Stein
K.
,
Yang
K.
,
Zhang
L.
&
Zhong
W.
2021
Changing El Niño–Southern Oscillation in a warming climate
.
Nature Reviews Earth & Environment
2
,
628
644
.
Compagnucci
R. H.
&
Richman
M. B.
2008
Can principal component analysis provide atmospheric circulation or teleconnection patterns?
International Journal of Climatology
28
,
703
726
.
Compagnucci
R. H.
,
Araneo
D.
&
Canziani
P. O.
2001
Principal sequence pattern analysis: a new approach to classifying the evolution of atmospheric systems
.
International Journal of Climatology
21
,
197
217
.
Ding
Q.
,
Steig
E. J.
,
Battisti
D. S.
&
Wallace
J. M.
2012
Influence of the Tropics on the Southern Annular Mode
.
J. Clim.
25
,
6330
6348
.
Donat
M. G.
,
Angélil
O.
&
Ukkola
A. M.
2019
Intensification of precipitation extremes in the world's humid and water-limited regions
.
Environmental Research Letters
14
,
065003
.
Endris
H. S.
,
Lennard
C.
,
Hewitson
B.
,
Dosio
A.
,
Nikulin
G.
&
Artan
G. A.
2019
Future changes in rainfall associated with ENSO, IOD and changes in the mean state over Eastern Africa
.
Climate Dynamics
52
,
2029
2053
.
Engelbrecht
F.
,
Adegoke
J.
,
Bopape
M.-J.
,
Naidoo
M.
,
Garland
R.
,
Thatcher
M.
,
McGregor
J.
,
Katzfey
J.
,
Werner
M.
&
Ichoku
C.
2015
Projections of rapidly rising surface temperatures over Africa under low mitigation
.
Environmental Research Letters
10
,
085004
.
Eyring
V.
,
Bony
S.
,
Meehl
G. A.
,
Senior
C. A.
,
Stevens
B.
,
Stouffer
R. J.
&
Taylor
K. E.
2016
Overview of the coupled model intercomparison project phase 6 (CMIP6) experimental design and organization
.
Geoscientific Model Development
9
,
1937
1958
.
Fitchett
J. M.
,
Grant
B.
&
Hoogendoorn
G.
2016
Climate change threats to two low-lying South African coastal towns: risks and perceptions
.
South African Journal of Science
112
(
n.5–6
),
1
9
.
Gillett
N. P.
&
Thompson
D. W. J.
2003
Simulation of recent Southern Hemisphere climate change
.
Science
302
,
273
275
.
Giorgi
F.
,
Coppola
E.
&
Raffaele
F.
2018
Threatening levels of cumulative stress due to hydroclimatic extremes in the 21st century
.
npj Climate and Atmospheric Science
1
,
18
.
Hendrickson
A. E.
&
White
P. O.
1964
Promax: a quick method to oblique simple structure
.
British Journal of Statistical Psychology
17
,
65
.
Hersbach
H.
,
Bell
B.
,
Berrisford
P.
,
Hirahara
S.
,
Horányi
A.
,
Muñoz-Sabater
J.
,
Nicolas
J.
,
Peubey
C.
,
Radu
R.
,
Schepers
D.
,
Simmons
A.
,
Soci
C.
,
Abdall
S.
,
Abellan
X.
,
Balsamo
G.
,
Bechtold
P.
,
Biaveti
G.
,
Bidlot
J.
,
Bonavita
M.
,
De Chiara
G.
,
Dahlgren
P.
,
Dee
D.
,
Diamantakis
M.
,
Dragani
R.
,
Flemming
J.
,
Forbes
R.
,
Fuentes
M.
,
Geer
A.
,
Haimberger
L.
,
Healy
S.
,
Hogan
R. J.
,
Hólm
E.
,
Janisková
M.
,
Keeley
S.
,
Laloyaux
P.
,
Lopez
P.
,
Lupu
C.
,
Radnoti
G.
,
de Rosnay
P.
,
Rozum
I.
,
Vamborg
F.
,
Villaume
S.
&
Thépaut
J.-N.
2020
The ERA5 global reanalysis
.
Quarterly Journal of the Royal Meteorological Society
146
,
1999
2204
.
Hu
Y.
&
Fu
Q.
2007
Observed poleward expansion of the Hadley circulation since 1979
.
Atmospheric Chemistry and Physics
7
,
5229
5236
.
Ibebuchi
C. C.
2021b
Circulation type analysis of regional hydrology: the added value in using CMIP6 over CMIP5 simulations as exemplified from the MPI-ESM-LR model
.
Journal of Water and Climate Change
3
,
1046
1055
.
Ibebuchi
C. C.
2022
On the representation of atmospheric circulation modes in regional climate models over Western Europe
.
International Journal of Climatology
.
https://doi.org/10.1002/joc.7807
.
Ibebuchi
C. C.
2023
Circulation patterns linked to the positive sub-tropical Indian ocean dipole
.
Advances in Atmospheric Sciences
40
,
110
128
.
Ibebuchi
C. C.
&
Abu
I.-O.
2023
Relationship between synoptic circulations and the spatial distributions of rainfall in Zimbabwe
.
AIMS Geosciences
9
,
1
15
.
Ibebuchi
C. C.
&
Richman
M. B.
2022
(Personal Discussion) Optimized Classification of Circulation Types Using the Rotated T-Mode Principal Component Analysis
.
Kendall
M.
1975
Rank Correlation Methods
.
Griffin
,
London
, p.
202
.
Kusangaya
S.
,
Warburton
M. L.
,
van Garderen E.
A.
&
Jewitt
G. P. W.
2014
Impacts of climate change on water resources in Southern Africa: a review
.
Physics and Chemistry of the Earth, Parts A/B/C
67-69
,
47
54
.
Latif
M.
&
Keenlyside
N. S.
2009
El Niño/Southern Oscillation response to global warming
.
Proceedings of the National Academy of Sciences
106
,
20578
20583
.
Lau
K. M.
&
Kim
K. M.
2015
Robust Hadley circulation changes and increasing global dryness due to CO2 warming from CMIP5 model projections
.
Proceedings of the National Academy of Sciences
112
,
3630
3635
.
Li
C.
,
Zwiers
F.
,
Zhang
X.
,
Li
G.
,
Sun
Y.
&
Wehner
M.
2021a
Changes in annual extremes of daily temperature and precipitation in CMIP6 models
.
Journal of Climate
34
,
3441
3460
.
Li
W.
,
Bi
X.
,
Sheng
L.
,
Luo
Y.
&
Sun
J.
2021b
Modulations of synoptic weather patterns on warm-sector heavy rainfall in south China: insights from high-density observations with principal component analysis
.
Frontiers in Earth Science
9
,
2296
6463
.
Liu
W.
,
Lim
W.
,
Sun
F.
,
Mitchell
D.
,
Wang
H.
,
Chen
D.
,
Bethke
I.
,
Shiogama
H.
&
Fischer
E.
2018
Global freshwater availability below Normal conditions and population impact under 1.5 and 7 2 °C stabilization scenarios
.
Geophysical Research Letters
45
,
9803
9813
.
Mann
H. B.
1945
Nonparametric tests against trend
.
Econometrica
13
,
245
259
.
Maúre
G.
,
Pinto
I.
,
Ndebele-Murisa
M.
,
Muthige
M.
,
Lennard
C.
,
Nikulin
G.
,
Dosio
A.
&
Meque
A.
2018
The Southern African climate under 1.5 °C and 2 °C of global warming as simulated by CORDEX regional climate models
.
Environmental Research Letters
13
,
065002
.
Morioka
Y.
,
Takaya
K.
,
Behera
S.
&
Masumoto
Y.
2015
Local SST impacts on the summertime mascarene high variability
.
Journal of Climate
28
,
678
694
.
Moron
V.
,
Robertson
A. W.
,
Ward
M. N.
&
Camberlin
P.
2007
Spatial coherence of tropical rainfall at regional scale
.
Journal of Climate
20
,
5244
5263
.
Moron
V.
,
Robertson
A. W.
,
Ward
M. N.
&
Ndiaye
O.
2008a
Weather types and rainfall over Senegal. part I: observational analysis
.
Journal of Climate
21
,
266
287
.
Moron
V.
,
Robertson
A. W.
,
Ward
M. N.
&
Ndiaye
O.
2008b
Weather types and rainfall in Senegal. Part II: downscaling of GCM simulations
.
Journal of Climate
21
,
288
307
.
Moron
V.
,
Robertson
A. W.
,
Qian
J.-H.
&
Ghil &
M.
2015
Weather types across the Maritime Continent: from the diurnal cycle to interannual variations
.
Frontiers in Environmental Science
2
,
https://doi.org/10.3389/fenvs.2014.00065
.
Müller
W. A.
&
Roeckner
E.
2006
ENSO impact on midlatitude circulation patterns in future climate change projections
.
Geophysical Research Letters
33
,
L05711
.
Muthige
M. S.
,
Malherbe
J.
,
Englebrecht
F. A.
,
Grab
S.
,
Beraki
A.
,
Maisha
T. R.
&
Van der Merwe
J.
2018
Projected changes in tropical cyclones over the South West Indian Ocean under different extents of global warming
.
Environmetal Research Letters
13
,
065019
.
Richman
M. B.
1986
Rotation of principal components
.
Journal of Climatology
6
,
293
335
.
Saenz
F.
&
Duran-Quesada
A. M.
2015
A climatology of low level wind regimes over Central America using a weather type classification approach
.
Frontiers in Earth Sciences
3
,
15
.
Sousa
P. M.
,
Blamey
R. C.
,
Reason
C. J. C.
,
Ramos
A. M.
&
Trigo
R. M.
2018
The ‘Day Zero’ Cape Town drought and the poleward migration of moisture corridors
.
Envirometal Ressearch Letters
13
,
124025
.
Souverijns
N.
,
Thiery
W.
,
Demuzere
M.
&
Van Lipzig
N. P. M.
2016
Drivers of future changes in East African precipitation
.
Envirometal Ressearch Letters
11
,
114011
.
Spinonia
J.
,
Barbosaa
P.
,
De Jager
A.
,
McCormick
N.
,
Naumann
G.
,
Vogt
J. V.
,
Magni
D.
,
Masante
D.
&
Mazzeschi
M.
2019
A new global database of meteorological drought events from 1951 to 2016
.
Journal of Hydrology: Regional Studies
22
,
100593
.
Yue
S.
&
Wang
C. Y.
2004
The Mann-Kendall test modified by effective sample size to detect trend in serially correlated hydrological series
.
Water Resources Management
18
,
201
218
.
Zeebe
R. E.
2013
Time-dependent climate sensitivity and the legacy of anthropogenic greenhouse gas emissions
.
Proceedings of the National Academy of Sciences
110
,
13739
13744
.
Zhang
J. P.
,
Zhu
T.
,
Zhang
Q. H.
,
Li
C. C.
,
Shu
H. L.
,
Ying
Y.
,
Dai
Z. P.
,
Wang
X.
,
Liu
X. Y.
,
Liang
A. M.
,
Shen
H. X.
&
Yi
B. Q.
2012
The impact of circulation patterns on regional transport pathways and air quality over Beijing and its surroundings
.
Atmospheric Chemistry and Physics
12
,
5031
5053
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).

Supplementary data