Introduction

African easterly waves (AEWs) are westward-propagating synoptic-scale disturbances that originate from central North Africa during boreal summer1,2. These waves develop and propagate along the African Easterly Jet (AEJ)1,3,4. Previous studies have linked AEW formation to the AEJ’s combined barotropic-baroclinic instability, characterized by a reversal in the meridional gradient of potential vorticity (PV)1,3,5,6 as well as pre-existing convection over East Africa7,8. Latent heating from moist convection and the radiative effects of Saharan dust also contribute to AEW growth and maintenance9,10,11,12,13,14,15. Given the significant influence of AEWs on regional and global weather patterns—such as organizing mesoscale convective systems (MCSs), modulating Sahel rainfall, mobilizing Saharan dust, and acting as precursors for Atlantic tropical cyclone (TC) development1,7,8,12,16,17,18,19,20,21,22,23—understanding how AEW characteristics (e.g., tracks) will change in a warmer future climate is crucial.

Previous studies have classified AEWs into two distinct types based on their periodicity: 3–5-day AEWs, which have wavelengths of 3000–5000 km with phase speeds of 8–12 m s⁻¹, and 6–9-day AEWs, which have longer wavelengths of 5000–6000 km and slower phase speeds of around 7 m s−¹ 1,2,8,16,24,25,26,27,28,29,30,31. The 3–5-day AEWs follow two main tracks over West Africa: one near 5°N and the other around 15°N, merging over the tropical Atlantic at about 17.5°N. In contrast, 6–9-day AEWs are located north of the AEJ, near 17.5°N, over both West Africa and the tropical Atlantic28. Although the northern and southern AEW tracks are distinct, they often represent different phases of the same wave4,17,32, and their impacts on regional and global climate vary. The southern track, which propagates near the Intertropical Convergence Zone (ITCZ), is typically associated with moist convection, has a greater influence on precipitation variability in West Africa, and is more efficient at undergoing tropical cyclogenesis16,25,33,34. Conversely, the northern track, located within the near-surface baroclinic zone, is characterized by dry convective processes35. These too can influence precipitation variability in the Sahel through the advection of moisture within the southerly flow of the vortex35, but they are less efficient at undergoing tropical cyclogenesis33,36. Additionally, since their track is near the southern edge of the Sahara, they play a crucial role in dust mobilization and transport across West Africa and the Atlantic37. In many cases, the northern and southern tracks of an AEW often merge to form a single structure, typically over the Atlantic Ocean38,39. These single-structured AEWs are often associated with the formation of TCs in the eastern Atlantic Ocean, contributing to more than half of such events40.

Previous studies have examined the impact of climate change on AEWs using global33,41,42,43 and regional44,45 climate model simulations. For example, using simulations from the fifth phase of the Coupled Model Intercomparison Project (CMIP5), Skinner and Diffenbaugh33 and Brannan and Martin43 reported a projected increase in the frequency and intensity of northern-track AEWs, whereas southern-track AEWs did not exhibit a robust response by the end of the twenty-first century. Similarly, Hannah and Aiyyer42, using a super-parameterized earth system model, found increased northern-track AEW activity due to enhanced baroclinicity and decreased southern-track activity, attributed to weakened temperature gradients. Martin and Thorncroft41 used CMIP5 and also observed reduced AEW activity in late spring and early summer but higher intensity from July to October, with southern-track responses being more uncertain due to poor model resolution of the Guinea Highlands. In contrast, Kebe et al.44, using a regional climate model (RCM, at 25 km resolution), projected a future decrease in AEW activity due to decreased barotropic and baroclinic instability. Additionally, they reported a strengthened and southward-shifted AEJ, attributed to a shift in the meridional temperature gradient (MTG). However, Bercos-Hickey and Patricola45 found a projected increase in AEW frequency and intensity using another RCM. These varying results highlight significant uncertainty in the future behavior of AEWs, likely due to differences in model physics, resolution, tracking methods, and climate scenarios.

Studies have documented significant improvements in climate models’ ability to capture key features of past and future climates, as seen in the progression from CMIP3 to CMIP546,47. The recent transition to CMIP648 includes updates such as better representation of physical processes, higher resolution, expansion of complexity by transitioning from General Circulation Models to Earth System Models, and new emission scenarios for future projections (i.e., shared socioeconomic pathways; SSP). However, it is unclear whether these updates will enhance the simulation of AEW activity and strengthen the reliability of projected changes. Therefore, this study builds on previous studies to investigate the future response of AEWs to global warming by leveraging CMIP6 simulations to address the following questions: (a) How well do CMIP6 models represent the characteristics of AEWs and their associated mechanisms? (b) How will AEW activity change in the future; will lower and higher emission scenarios lead to consistent or different results? (c) What are the dynamic and thermodynamic environments responsible for changes in AEW activity? A better understanding of AEW responses to global warming will provide valuable insights into the hydroclimate of West Africa and the broader Atlantic basin.

Results

Representation of AEW activity in CMIP6 present-day simulations

Realistic representation of the present-day climate is the basis for developing credible projections of future changes in global warming scenarios49,50. Following previous studies33,41,51,52,53,54, we use both the meridional wind variance (MWV) and eddy kinetic energy (EKE) approaches as proxies for AEW activity. These proxies are calculated for the 3–5-day and 6–9-day AEWs and are computed at 850 and 700 hPa to assess low- and mid-level AEW activity. First, we employ various descriptive statistical measures, including percentage bias, normalized root mean square error (NRMSE), pattern correlation coefficient (PCC), and Taylor skill score (TSS), to evaluate the accuracy of simulated AEW activity relative to ERA5. The results are presented in Figs. S1–S2 of the supplementary information. Generally, a low percentage bias and NRMSE, along with high PCC and TSS, are desirable.

For both the MWV and EKE approaches at 850 and 700 hPa, most CMIP6 models used here exhibit a high PCC (~0.9) and TSS (~0.7) for both the 3–5-day and 6–9 day-AEWs, with the exception of BCC-CSM2-MR, FGOALS-g3, and MIROC6, which show lower PCC (~0.6) and TSS (~0.5). For the 3–5-day AEWs (Fig. S1), five CMIP6 models (ACCESS-EM2, EC-Earth3, EC-Earth-CC, EC-Earth-Veg, and FGOALS-g3) considerably underestimate AEW activity, with percentage biases reaching ~ –50%. In contrast, BCC-CSM2-MR exhibits the highest positive percentage biases (~25%). Other models show moderate biases, aligning more closely with ERA5. NRMSE values are consistent across all models for both metrics and pressure levels. NRMSEs remain considerably low for all models (<0.5) except for FGOALS-g3 across all metrics and levels, and ACCESS-EM2, EC-Earth3, EC-Earth-CC, and EC-Earth-Veg for the MWV method at 850 hPa, where NRMSE exceeds 0.9. For the 6–9-day AEWs (Fig. S2), most CMIP6 models display a low percentage bias (<10%), except for BCC-CSM2-MR and CanESM5 (FGOALS-g3), which exhibit a positive (negative) bias of more than 20%. Only FGOALS-g3 has NRMSE >0.9, while other models generally exhibit values <0.5.

Figs. S3–S10 provide a clearer depiction of the variation in the representation of AEW activity by the CMIP6 models, in terms of both MWV (Figs. S3–S6) and EKE (Figs. S7–S10), both of which are spatially consistent. For instance, in the representation of the 3–5-day AEWs, some models (e.g., BCC-CSM2-MR and CanESM5) show higher magnitudes of AEW activity at both 850 hPa and 700 hPa compared to ERA5, while others (e.g., ACCESS-CM2 and FGOALS-g3) display relatively lower magnitudes. Overall, the largest biases relative to ERA5 occur mostly over land at the lower level and over the ocean at mid-level. We note that the biases reported herein have been a long-standing problem of global climate models33,41 and have not been significantly alleviated in CMIP6 models. Nevertheless, most of the models used in this study reasonably capture AEW activity.

Given that previous studies50,55 have noted that the multimodel ensemble mean (EnsMean) tends to reduce systematic biases in individual models by canceling out spatial errors, we further evaluate the performance of the CMIP6 EnsMean in accurately simulating the spatial distribution of present-day AEW activity. This assessment is made by comparing the results to ERA5 (Fig. 1 for 850 hPa and Fig. S11 for 700 hPa). Relative to ERA5, EnsMean considerably reproduces the present-day spatial distribution of both MWV and EKE for the 3–5-day and 6–9-day AEWs at both 850 hPa and 700 hPa. In the ERA5 reanalysis (Fig. 1a, c, e, g), the tracks of the 3–5-day (6–9-day) AEWs are evident at 850 hPa between 10°N and 25°N (15°N and 35°N) and west of 15°W and at the same latitudes at 700 hPa (Fig. S11a, c, e, g). EnsMean reasonably captures the location of maximum MWV and EKE, albeit with a lower magnitude (Figs. 1 b, d, f, h and S11b, d, f, h). Our analysis shows a clear distinction between 3–5-day and 6–9-day AEW tracks at both 850 and 700 hPa in many CMIP6 models (Figs. S3–S10) and EnsMean (Figs. 1 and S11). Previous studies28 demonstrated that 3–5-day AEWs typically follow two distinct tracks, one near 5°N and the other near 15°N, merging around 17.5°N over the tropical Atlantic. However, our analysis reveals that many CMIP6 models and EnsMean do not show a clear separation between these northern and southern tracks. This lack of distinction also noted in CMIP5 studies33, is often attributed to coarse model resolution or an overestimated connection between AEW circulation and convection, leading to excessively strong AEW activity within the ITCZ over West Africa. Nevertheless, our results suggest the presence of vertically deeper AEWs in the Sahel and the Guinea Coast in West Africa, considering that some CMIP6 models accurately simulate the AEW activity north and south of the AEJ. Despite the noticeable biases, which could potentially impact future projections, the CMIP6 EnsMean reasonably reproduces the spatial distribution of AEW activity. Thus, a discussion based on EnsMean is credible.

Fig. 1: Mean distribution of present-day African Easterly Wave activity.
figure 1

a–d JJAS mean meridional wind variability (shaded, m2s-2) and e–h JJAS mean EKE (shaded, m2s–2) at 850 hPa for 3–5-day (a, b, e, f) and 6–9-day (c, d, g, h) AEWs for the period 1985–2014. Mean averaged from ERA5 reanalysis (a, c, e, g) and CMIP6 EnsMean (b, d, f, h). The red contour line in a–h represents the −8 ms–1 zonal wind contour at 700 hPa during 1985–2014, from ERA5 and the CMIP6 EnsMean, indicating the location of the AEJ core.

Full size image

Projected changes in AEW activity

We examine the potential changes in AEW activity by the end of the 21st century (Figs. 2 and S12 for EnsMean changes at 850 hPa and 700 hPa, respectively). The projected changes in individual models are presented in Figs. S13–S20 (Figs. S21–S28) based on MWV (EKE). Under future warming, there is strong intermodel agreement among the CMIP6 models that AEWs along the Sahel-Sahara border (~20°N) will intensify. This projected change is robust, as at least 70% of the models agree on the sign of EnsMean change across most grid points (see hatching on the figures). For 3–5-day AEWs, the zone of maximum change at 850 hPa (Fig. 2a, c, e, g) is between 12°N and 26°N and 20°E and 30°W for SSP5–8.5, and between 16°N and 25°N and 5°E and 25°W for SSP2–4.5. In this region, EnsMean shows an average increase of up to 0.9 m²s⁻² under SSP2–4.5 and more than 1.5 m²s⁻² under SSP5–8.5 in the JJAS mean MWV (Fig. 2a, c) and EKE (Fig. 2e, g). Similar projected increases in MWV and EKE are observed in nearly the same region at 700 hPa (Fig. S12), although the region of robust projected increase extends southward to the Guinea Coast of West Africa (near 2°N) under SSP5–8.5 (Fig. S12c, g). We note that eight CMIP6 models, mostly from the same institution (EC-Earth3-CC, EC-Earth3, EC-Earth3-Veg, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, NESM3, NorESM3-LM), contribute to the southward increase in AEW activity evident in EnsMean (Figs. S18 and S26). This may be attributed to model interdependence, where models developed by the same or closely collaborating institutions tend to share similar physics and parameterization schemes56,57,58. Such interdependence can skew EnsMean; thus, caution is needed when interpreting such results, as the apparent consensus may not reflect independent lines of evidence but rather the overlap in model designs. Although model interdependence suggests a need for caution, the findings herein are robust and offer valuable insights into future AEW activity. Consistent with 3–5-day AEWs, we also find an intensification of 6–9-day AEWs (Figs. 2b, d, f, h and S12b, d, f, h), though the magnitude is 2–3 times lower, and the region of robust increase is confined mostly to land.

Fig. 2: Projected changes in African Easterly Wave activity.
figure 2

a–d CMIP6 EnsMean change in JJAS meridional wind variability (shaded, m2s-2) and e–h CMIP6 EnsMean change in JJAS EKE (shaded, m2s–2) at 850 hPa for 3–5-day (a, c, e, g) and 6–9-day (b, d, f, h) AEWs under (a, b, e, f) SSP2–4.5 and (c, d, g, h) SSP5–8.5. Mean change calculated between 2070–2099 and 1985–2014. The purple contour in each plot marks the -8 ms–2 zonal wind contour at 700 hPa during 2070–2099 and indicates the future location of the AEJ core. Hatching indicates grid points where at least 70% of ensemble members agree on the sign of the change in EnsMean.

Full size image

With the exception of SSP5–8.5 at 700 hPa, where the projected increase in 3–5-day AEWs by EnsMean extends to the Guinea Coast, the zone of enhanced AEW activity generally lies within and north of the maximum AEW activity observed during the present-day period (1985–2014) and projected future (2070–2099) location of the AEJ core. In general, at both 700 hPa and 850 hPa (except for 3–5-day AEWs at 700 hPa under SSP5–8.5), we find no robust changes in the EnsMean-projected AEW activity south of 12°N toward the Guinea Coast (i.e., on the cyclonic shear side of the AEJ). Previous studies33 have noted that AEW activity south of 12°N is likely complicated by the response of moist convective processes to enhanced greenhouse forcing. Indeed, considerable variations exist across individual models in the southern region. For instance, BCC-CSM2-MR and CESM2-WACCM project a future decrease in AEW activity across all levels and scenarios, while models such as EC-Earth3-CC, EC-Earth3, EC-Earth3-Veg, IPSL-CM6A-LR, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, and NESM3 (Figs. S13–S28) generally simulate future increases that extend to the Guinea Coast (i.e., south of 12°N). The inconsistency in the sign of projected changes in AEW activity south of 12°N (south of the AEJ) at 850 hPa and 700 hPa indicates higher uncertainty, and perhaps a lower likelihood of significant changes in AEW activity across the Guinea Coast of West Africa.

The relatively high-latitude location of the zone of maximum and robust increase in AEW activity (i.e., 12°N to 26°N) may suggest an influence from increased midlatitude trough activity59,60. To further assess this, we examine the changes in midlatitude troughs and their potential contribution to changes in AEW activity. Following Chang et al.61, we use MWV at 250 hPa as a proxy for midlatitude storm track activity. Our results indicate that future changes in midlatitude trough activity have minimal impact on the robust changes in AEW activity (Figs. 3 and S29). Specifically, under both the SSP2–4.5 and SSP5–8.5 scenarios, we find a projected decrease in summer mean variability of 3–5-day (Fig. 3) and 6–9-day (Fig. S29) AEWs at 250 hPa north of 30°N but south of 60°N, covering large parts of southern Europe and the subtropical Atlantic (Figs. 3a, c and S29a, c). In contrast, a robust increase is evident north of 60°N, suggesting that the mean location of midlatitude troughs may shift northward under global warming. A similar pattern of robust changes in MWV is observed at 850 hPa between 30°N and 60°N (Figs. 3b, d and S29b, d). However, these changes are generally weaker, with a maximum variance of –2.0 m²s⁻² at 850 hPa compared to more than –3.0 m²s⁻² at 250 hPa. Based on these findings, we conclude that changes in AEW activity drive most of the increased variability in MWV evident in the CMIP6 models. Similar results are found for EKE (Figure not shown).

Fig. 3: Projected changes in African Easterly Wave activity.
figure 3

CMIP6 EnsMean change in JJAS meridional wind variability (shaded, m2s-2) for 3–5-day AEWs at 250 hPa (a, c) and 850 hPa (b, d) under SSP2–4.5 (a, b) and SSP5–8.5 (c, d). Mean change calculated between 2070–2099 and 1985–2014 (shaded, m2s–2). Hatching indicates grid points where at least 70% of ensemble members agree on the sign of the change in EnsMean.

Full size image

Dynamic and thermodynamic factors contributing to future changes in AEW activity

Here, we assess the factors contributing to changes in AEW activity (i.e., EKE) by examining future changes in the underlying mechanisms. We first diagnose the Lorenz energy cycle62 and assess key energy conversion terms, including baroclinic energy and barotropic energy conversion, as well as diabatic heating. The primary energy source for eddies is available potential energy, which is converted into kinetic energy62. Both available potential energy and kinetic energy can be separated into their mean and eddy parts. Previous studies have shown that the northern-track AEWs, particularly along the Sahel-Sahara border (where this study identifies a higher likelihood of future changes), derive most of their energy from baroclinic processes. These processes are associated with the interaction of low-level MTG in the Sahel-Sahara region and potential vorticity gradients within the AEJ core. Projected AEW energetics are expected to partially reflect local surface air temperature changes due to increased radiative forcing. Enhanced potential temperature gradients lead to increased zonal available potential energy (ZAPE), fueling AEW development33,35,63,64. AEWs harness this reservoir of ZAPE by transferring temperature along the MTG and generating eddy available potential energy (EAPE). AEWs then grow by converting EAPE into EKE via baroclinic overturning. This continuous growth requires replenishing EAPE from the ZAPE reservoir63. Figures 4a, b and S30a, b show the EnsMean meridional cross-section of the baroclinic energy conversion term \((-\frac{R}{p}\overline{{\omega }^{{\prime} }{T}^{{\prime} }})\) for 3–5-day and 6–9-day AEWs under SSP5–8.5 and SSP2–4.5, respectively. For the 3–5-day AEWs, two distinct centers of positive baroclinic conversion values are evident in present-day simulations over West Africa (see contour line). The first is between 10°N and 22°N in the lower troposphere, spanning from the surface to 650 hPa. This center is closely associated with either the ascent of warm air or the descent of cold air, most coinciding with the Saharan heat-low region65. The second center is between 500 and 200 hPa from 0° to 16°N and is linked to latent heat release from AEW-induced convection, which drives the ascent of warm air at these levels. Similar regions of positive baroclinic energy conversion appear in the 6–9-day AEWs, though with reduced intensity compared to the 3–5-day AEWs. The pronounced baroclinic conversion in the upper troposphere for 3–5-day AEWs can be attributed to more intense convection, as simulated by the CMIP6 models. Under future warming (shading in Figs. 4a, b and S30a, b), both 3–5-day and 6–9-day AEWs show an increase in baroclinic energy conversion in both the lower and upper troposphere, although the 6–9-day waves remain less intense, consistent with their present-day patterns.

Fig. 4: Projected changes in energetic terms.
figure 4

Meridional height cross-sections of CMIP6 EnsMean change (shaded, m2s-2day-1) in the JJAS a, b baroclinic term \(({C}_{{pk}})\), c, d barotropic term \(({C}_{k})\), e, f diabatic heating term (GE), and g, h conversion term (CA) under SSP5–8.5, averaged between 20◦W and 10◦E. Contours represent present-day mean. Mean change calculated between 2070–2099 and 1985–2014. The left panel shows 3–5-day AEWs and the right panel shows 6–9-day AEWs. Stippling indicates grid points where at least 70% of ensemble members agree on the sign of the change in EnsMean. Values under topography are masked with gray shading.

Full size image

Our results consistently indicate that the projected conversion of ZAPE to EAPE is linked to future AEW development, as represented by negative t-v covariance (Fig. 5a, d) and negative meridional temperature gradient (MTG)-v covariance (Fig. S32). This conversion is intensified by the enhanced regional poleward MTG (Fig. S33). Notably, EnsMean reasonably reproduces the region of negative mean covariance of the filtered meridional wind and temperature over West Africa during the historical period (Fig. S31). Interestingly, both t-v covariance and MTG-v covariance become more negative in regions projected to experience increased AEW activity by the end of the 21st century, indicating a stronger temperature flux down the temperature gradient and greater EAPE for AEW growth (Figs. 5a, d and S32). Indeed, the MTG is projected to increase from the surface to the mid-troposphere (400 hPa), with the most significant increases occurring below 700 hPa over the Sahel-Sahara region, particularly between 14°N and 26 °N (Fig. S33). The projected intensification of the MTG is likely to have a substantial impact on the zonal atmospheric circulation patterns through the thermal wind relationship66,67, subsequently affecting the future position and strength of the AEJ.

Fig. 5: Projected changes in mean JJAS temperature–meridional wind covariance, sea level pressure and low-level wind, and vertical velocity.
figure 5

CMIP6 EnsMean change in JJAS a, d 3–5-day filtered temperature and meridional wind variance (t-v covariance) at 850 hPa, b, e sea level pressure and 850 hPa winds, and c, f vertical velocity. Mean change calculated between 2070–2099 and 1985–2014. Values under topography are masked with gray shading in (c, f). The mean vertical velocity change in the latitude-height cross-section in (c) is averaged between 15°W and 15°E. Hatching indicates grid points where at least 70% of GCMs agree on the sign of the change in EnsMean. Areas with statistically significant differences at the 95% level are marked with black stippling in (b, e).

Full size image

In general, we observe that warming in the Sahel-Sahara during the JJAS season at the end of the 21st century is notably greater than that in the Guinea Coast, with mean surface potential temperatures increasing by 5 °C to 6.6 °C (3.8 °C to 4.2 °C) north (south) of 15°N, as shown in Fig. 6a, e. This surface warming pattern, driven largely by both local and large-scale processes, aligns with the projected amplification of the MTG, particularly over the Sahel-Sahara, thereby intensifying the ZAPE. In fact, a previous study33,68 has indicated that surface temperatures by the end of the 21st century over southern Europe and the Mediterranean would increase in response to enhanced mid-troposphere anticyclonic circulation, large-scale subsidence, and reduced precipitation. These changes are associated with robust increases in surface shortwave downwelling radiation over much of Europe, the Mediterranean Sea, and northern Africa (Fig. 6c, g), while the Sahara warms preferentially in response to an anomalous increase in surface longwave downwelling radiation (Fig. 6d, h). Furthermore, the robust increases in precipitation also minimize low-level warming across the central Sahel, further enhancing the contrast in temperature between the Sahara and southern regions (Fig. 6b, f). Based on these results, we argue that the increased MTG over West Africa is partly responsible for providing the necessary energy for stronger AEW development under global warming by intensifying the AEJ (Fig. 2).

Fig. 6: Projected changes in mean JJAS temperature, precipitation, and radiation.
figure 6

CMIP6 EnsMean change in JJAS a, e potential temperature (dashed lines), b, f precipitation, c, g surface downwelling shortwave radiation, and d, h surface downwelling longwave radiation under SSP2–4.5 (a–d) and SSP5–8.5 (e–h). Mean change calculated between 2070–2099 and 1985–2014. Shading in b, f represents the present-day mean JJAS potential temperature from CMIP6 EnsMean. Areas with statistically significant differences at the 95% level are marked with black stippling, and hatching indicates grid points where at least 70% of ensemble members agree on the sign of the change in EnsMean. Hatching is deliberately removed from (d, h) because all models agree on the sign of the change in EnsMean.

Full size image

Furthermore, barotropic energy conversion, \((-\overline{\overrightarrow{{V}_{H}^{{\prime} }}.(\overrightarrow{{V}^{{\prime} }}.\nabla )\overrightarrow{{V}_{H}}})\) is the secondary source of EKE, converting zonal kinetic energy into EKE. For 3–5-day AEWs in the present-day climatology, two primary regions of positive barotropic conversion are evident, spanning the upper and mid-to-lower troposphere (contours in Figs. 4c, d and S30c, d). The positive values in the upper troposphere, starting around 0°, are driven by the anticyclonic divergent flow from the ITCZ and the strong shear of the Tropical Easterly Jet (TEJ)65. The second region is in the middle troposphere, centered around 9°N, south of the AEJ. This region is associated with EKE production due to AEWs. In other words, the barotropic energy conversion that maintains AEWs occurs primarily south of the AEJ, where mostly 3–5-day AEWs occur. Meanwhile, on the northern flank of the AEJ, around 16°N, the barotropic conversion exhibits lower values, which become negative farther north toward the Sahara. This asymmetry on either side of the AEJ leads to the tilting of AEW axes. These results are consistent for the 6–9-day AEWs, although with a lower magnitude.

For both the 3–5-day and 6–9-day AEWs, a significant area of EKE sink (i.e., negative barotropic conversion) is observed north of 18°N below 700 hPa and north of 8°N above 700 hPa. The negative values are notably larger for the 6–9-day AEWs. This sink is associated with the shear in the subtropical jet54. Under future warming, we find a robust increase in the barotropic conversion term near the tropopause (shaded in Figs. 4c, d and S30c, d), extending from the upper to the lower troposphere, spanning from the equator at upper levels to around 20°N in the lower troposphere. Notably, there are regions of isolated projected decreases below 700 hPa, around 4°N to 10°N. More pronounced decreases are projected at mid-to-upper levels north of 22°N, with this pattern extending southward at upper levels. These results are consistent for both the 3–5-day and 6–9 day-AEWs, though the intensity is greater for the former. This pattern is consistent under the SSP2–4.5 scenario, although the magnitude of change is much lower.

The next term in the energy cycle is the generation term (GE; \(\frac{\gamma }{\bar{T}}\overline{{Q}_{1}^{{\prime} }{T}^{{\prime} }}\)), representing the source of EAPE via diabatic heating, as shown in Figs. 4e, f and S30e, f. According to Michaelides69, positive (negative) values of GE indicate the generation (destruction) of EAPE, corresponding to the warming (cooling) of warm regions or the cooling (warming) of cold regions at the same latitude. The meridional distribution of the EnsMean GE for both 3–5-day and 6–9-day AEWs is generally consistent, though the magnitude is higher for 3–5-day AEWs. This pattern of GE, particularly in the upper troposphere, is similar to that of the baroclinic conversion term, suggesting that the EAPE consumed by baroclinic processes is partially offset by diabatic heating. For both 3–5-day and 6–9-day AEWs, two regions of positive GE are observed in the present-day period (contours in Figs. 4e, f and S30e, f): one in the upper troposphere between 150 hPa and 500 hPa, centered around 8°N, and another below 700 hPa, near the core of the AEJ around 15°N. Under future warming (shaded areas in Figs. 4e, f and S30e, f), similar to the baroclinic conversion term, GE in the mid-to-upper troposphere centered around 8°N and in the mid-to-lower troposphere centered around 15°N is projected to increase in the future for both types of AEWs under SSP5–8.5. Additionally, a region of projected decrease in GE is evident between 500 hPa and 850 hPa south of 12°N. This pattern is consistent under the SSP2–4.5 scenario, although the magnitude of change is much lower.

The meridional cross–section of the conversion term (CA; \(-\frac{{c}_{p}\gamma }{\bar{T}}\overline{\overrightarrow{{V}_{H}^{{\prime} }}{T}^{{\prime} }}.{\nabla }_{H}\bar{T}\)), which represents the conversion of ZAPE to EAPE through eddy heat flux along the zonal mean temperature gradient, is shown in Figs. 4g, h and S30g, h. This conversion acts as an additional source of EAPE, complementing the baroclinic term. In the present-day period (see contour lines), unlike for the other terms, only one prominent region of conversion of ZAPE to EAPE (positive CA) is observed for both 3–5-day and 6–9-day AEWs, extending from 8°N to 22°N in the mid-to-lower troposphere. The magnitude of CA in this region is comparable to that of GE, suggesting a similar contribution to baroclinic conversion. This dominant contribution of CA likely results from thermal advection (see Eq. 5 in Methods) associated with the large-scale MTG. Under future warming (see shaded areas in Figs. 4g, h and S30g, h), this region of EAPE generation intensifies and extends up to around 600 hPa. The intensification is comparable for both 3–5-day and 6–9-day AEWs, though the magnitude is significantly higher under the SSP5–8.5 scenario compared to SSP2–4.5.

Furthermore, our results show that in addition to the increase in ZAPE and the energetic terms, low-level heating over the region at the end of the 21st century leads to a deepening of the Saharan Heat Low (SHL), strengthening of the monsoon flow, and increased convergence along the ITCZ (Fig. 5). The enhanced SHL, along with warmer surface air temperatures, maintains the region’s low static stability. This low static stability facilitates greater interaction between surface temperature gradients and the potential vorticity gradients at the AEJ level, resulting in stronger low-level AEWs north of the AEJ4,64,70. The pressure-driven increase in low-level westerly and southwesterly flow (Fig. 5b, e) yields greater convergence between the monsoon and Harmattan winds, leading to stronger vertical ascent along the intertropical discontinuity near 20°N over West Africa (Fig. 5c, f). We note that relative to ERA5, the CMIP6 models used herein accurately replicate a vertically deep region of ascent associated with the ITCZ near 8°N and a shallow region of ascent associated with the intertropical discontinuity near 17°N during the present-day period, indicating that the projected deepening of ascent near 20°N is not merely an amplification of model biases in simulating the large-scale overturning circulations in West Africa (Fig. S34). In fact, previous studies have shown that large-scale vertical ascent along the ITCZ contributes to the projected increases in AEW strength through vortex stretching35,40. Furthermore, the increase in deep vertical ascent along the intertropical discontinuity at the end of the 21st century suggests the presence of increased latent heating (Fig. 5c, f). This increase in latent heating aligns with substantial increases in moist monsoon flow and precipitation across the Sahel (except for drying over the far western Sahel) (Figs. 6b, f and 5b, e).

While AEWs north of the AEJ typically exhibit dry dynamics in the current climate35,53, the projected increase in moist convective processes in the Sahel implies an amplified role of such processes in AEW development along the Sahel-Sahara border in both SSP2–4.5 and SSP5–8.5. Given the significance of convectively driven diabatic heating in generating EAPE for AEW growth19,34, it is possible that the increased moisture within and north of the AEJ region also plays a pivotal role in enhancing AEW energy. Consistent with the increased AEW activity, most CMIP6 models show a slight strengthening of the AEJ at the end of the 21st century (Fig. S35). While there is no consistent shift in the jet’s location across the CMIP6 models, nearly all models project increased low-level westerly flow beneath the AEJ, particularly on the jet’s poleward side (Fig. S35). This resultant enhancement in vertical zonal wind shear may bolster the growth of AEWs71, helping to elucidate the substantial increases in AEW activity that occur north of the AEJ core in response to elevated forcing (Fig. 2).

Overall, the combined dynamic and thermodynamic responses to the pattern of low-level warming create an environment conducive for stronger AEWs north of the AEJ core over West Africa. Indeed, nearly all CMIP6 models simulate an increase in the occurrence of intense (mean plus one standard deviation) and extremely intense (mean plus two standard deviations) AEWs along the Sahel-Sahara border (Fig. 7). This includes a median seasonal increase of about 20–40% for intense AEWs and 50–80% for extremely intense AEWs under both SSP2–4.5 and SSP5–8.5.

Fig. 7: Projected change in percentage of intense and extreme intense AEWs.
figure 7

Boxplots of the change between 2070–2099 and 1985–2014 in the percentage of JJAS 3–5-day (a, c) and 6–9-day (b, d) AEWs that exceed the present-day period (1985–2014) one-SD and two-SD intensity thresholds at a reference location of (18°N, 0) under SSP2–4.5 (a, b) and SSP5–8.5 (c, d). Intensity thresholds are defined based on the magnitude of AEW meridional wind.

Full size image

Discussion

Previous studies have noted that AEWs exert significant influence on weather and climate by acting as precursor disturbances or preexisting cyclonic circulations for Atlantic tropical cyclone development, mobilizing and transporting Saharan dust within Africa and globally, and modulating rainfall across the Sahel during the West African monsoon. Hence, assessing how AEWs will change under continuous future warming is crucial for understanding future weather patterns and associated impacts in the region. In this study, we use an ensemble of CMIP6 models to investigate how AEW activity will change under global warming as well as the environmental factors driving these changes. We use both MWV and EKE as proxies for AEW activity. Analyses are conducted separately for 3–5-day and 6–9-day AEWs, with changes assessed under the SSP2–4.5 and SSP5–8.5 scenarios.

While CMIP6 models exhibit variability in simulating AEW activity, EnsMean provides a reasonable representation of the spatial distribution of MWV and EKE associated with AEWs. Our results show that AEW activity in the Sahel-Sahara will intensify in the future, suggesting a strengthening of the northern AEW track. The lack of clear separation between the northern and southern tracks of 3–5-day AEWs and the presence of vertically deeper AEWs in specific subregions highlight the complexities and challenges in accurately capturing AEW characteristics in climate models. Additionally, we find no robust future changes in the region occupied by the southern AEW track in EnsMean, which may be attributed to uncertainty arising from the conflicting projections of individual models in this region.

We diagnose four energetic terms related to the generation of EKE and EAPE to identify the primary energy source responsible for the future intensification of AEW activity. Baroclinic energy conversion is the leading term for the maintenance of AEWs. The next two crucial terms are barotropic energy conversion and generation term (GE) through diabatic heating. Our results indicate that under future warming, baroclinic conversion is the dominant source of future EKE for both types of AEWs, particularly near the tropopause. The projected increase in AEW activity is driven largely by an increase in ZAPE, resulting from stronger surface warming in the Sahel-Sahara region relative to the Guinea Coast (Fig. 6a, e). This uneven surface warming enhances the MTG, facilitating AEW growth via a stronger poleward baroclinic overturning process. Additionally, we find that barotropic conversion contributes to EKE generation primarily at the AEJ level, while energy generation due to diabatic heating is quantitatively similar, indicating that the GE supplements the EAPE utilized by baroclinic conversion. Notably, both the baroclinic overturning process and GE exhibit strong projected increases in the upper troposphere, suggesting enhanced latent heat release and convective heating in future climate scenarios. We emphasize that low-level heating at the end of the 21st century will deepen the SHL and enhance the monsoon flow, resulting in stronger vertical ascent along the intertropical discontinuity (Fig. 5c, f). This leads to enhanced convective processes in the Sahel-Sahara, shown in previous studies to strengthen AEWs north of the AEJ through increased latent heating. The convectively driven latent heating is expected to enhance moisture transport into the region by enhancing the monsoon and, consequently, precipitation across the Sahel.

We note that the changing AEW activity under global warming could have substantial implications for the regional climate of the Sahelian region of West Africa. Given the link between AEWs and MCSs that cause extreme flooding in the Sahel, our results suggest an increase in the intensity and/or frequency of such events in a warmer climate. Additionally, the projected increase in AEW activity holds significant implications for future Saharan dust transport, given that the region with the highest projected increase in AEW activity lies over prolific dust sources within the western Sahara (Fig. 2). Strong winds associated with a northern-track AEW can transport dry Saharan air downstream, either inhibiting TC formation entirely or delaying tropical cyclogenesis until the wave reaches more favorable environmental conditions farther west in the Atlantic basin72,73, where sea surface temperatures are warmer74. Despite the increasing AEW activity, its impact on the tropical North Atlantic climate remains uncertain. The relationship between AEWs and TCs has been questioned, as recent studies suggest that suppressing AEWs does not substantially alter basin-wide seasonal Atlantic TC frequency but does influence TC intensity, genesis time, and location75,76. These studies concluded that the presence or absence of AEWs could modulate the large-scale environment conducive for TC genesis, although there has been considerable evidence suggesting that northern-track waves can influence tropical cyclogenesis. Therefore, the potential future impact of the increased frequency of intense AEWs on TC development remains uncertain. Nonetheless, our findings highlight potential future changes in AEW activity, which could have significant implications for regional weather and climate variability, including changes in weather patterns, climate dynamics, and regional climate variability in West Africa. Some of the findings presented here are consistent with those of previous studies33,41 using CMIP5, emphasizing the robustness of AEW responses to future warming across successive generations of coupled climate models. Lastly, the analysis presented here focuses on the summer monsoon period (JJAS), which traditionally and climatologically coincides with the peak months of AEW activity. However, a recent study77 indicates that AEWs, along with other easterly waves, now occur year-round. To better capture the full spectrum of their impacts, future research should broaden its focus to include AEW activity beyond the traditional JJAS period.

Methods

In this study, we use climate simulations from the present day (1985–2014) and the end of the 21st century (2070–2099) from 24 CMIP6 models48. For each model, we use the first realization (‘r1i1p1f1’) for the present-day and future projection periods. The future projection is based on two SSP, including an intermediate greenhouse gas (GHG) emission scenario (SSP2–4.5) and an extreme scenario with very high GHG emissions (SSP5–8.5). We select only models that deliver complete outputs for all meteorological variables used for all the analyses presented herein. For all variables, we investigate the response to future warming by comparing the present-day period in the historical simulation with projections for the end of the 21st century. We focus on the end of the century, where studies have shown that forced changes are relatively larger than internal variability.

To evaluate the accuracy of the CMIP6-simulated AEWs, we compare the model results with the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis (ERA5)78 at a spatial resolution of 0.25° × 0.25°. Broad spatial assessment and several statistical metrics are used to evaluate the models, including percentage bias, NRMSE, PCC, and TSS, which have been used extensively in our previous studies (see Eqs. 1–4 in Akinsanola et al.50). Furthermore, we employ the multimodel ensemble mean (referred to hereafter as EnsMean) to address systematic biases due to model differences50,79. We regrid all datasets to the lowest model horizontal resolution (2.81° × 2.81°) using an area-conserving remapping procedure, which is implemented in the Climate Data Operators (https://code.zmaw.de/projects/cdo) to produce EnsMean. Furthermore, following Akinsanola et al.80, we estimate statistical robustness, particularly for analyses of spatial changes, by highlighting grid points where at least 70% of CMIP6 models agree on the sign of the change in EnsMean. In addition, a Student’s t-test is used to identify the grid points where changes are statistically significant (at the 95% confidence level). We focus only on the monsoon period of June-July-August-September (JJAS), corresponding to the season when AEWs are active.

Tracking algorithm43,81,82, MWV, and EKE approaches33,41,51,52 are the most common methods for detecting AEWs. Tracking algorithms, while effective, are highly sensitive to parameter tuning, require high temporal resolution (six-hourly or higher), demand substantial computational resources, and may overlook weaker AEWs17,52. These limitations make tracking algorithms more suitable for single-model analysis, where such fine-tuning and resource allocation are more feasible. In contrast, the MWV and EKE approaches provide a more computationally efficient alternative and offer a clear, cohesive spatial representation of average AEW activity33,41,51,52,53,54,83. These methods are better suited for analyzing long-term climate datasets and large model ensembles33,83, making them more practical for the current study.

In this study, we use both the MWV and EKE methods. For the variance method, we detect AEWs using the meridional component of wind at 700 hPa and 850 hPa. We isolate variations in the flow associated with the passing of AEWs by band-pass filtering84 the daily time series of the JJAS meridional wind (each year) to identify AEWs of 3–5-day and 6–9-day periods. Afterward, we compute the variance of the filtered time series at each grid point to attain a proxy of mean AEW activity. Furthermore, we identify the number of AEWs during JJAS by counting the instances in which the filtered daily mean meridional wind at the 850 hPa level exceeds 1 ms–1. This counting is done at a reference point north of the EnsMean AEJ core in both the present-day and end-of-the-21st-century periods (18°N, 0°E; Fig. 1). We consider each instance a separate AEW event. In cases where the filtered meridional wind exceeds 1 ms–1 on consecutive days, we consider only the day with the highest meridional wind value to prevent duplication of AEW events. To evaluate changes in the average number of intense AEWs, we calculate a distribution of AEW event intensities based on the 850 hPa meridional wind magnitude associated with each AEW at (18°N, 0°E) during the present-day period (i.e., 1985 to 2014). Following the approach of Skinner and Diffenbaugh33, for each model, we define an AEW as intense if its meridional wind exceeds the mean-plus-one-standard-deviation threshold derived from the present-day period’s distribution. From these, we designate an AEW as extremely intense if its meridional wind exceeds the mean-plus-two-standard-deviations threshold from the present-day period’s distribution. Subsequently, we compute the percentage difference in the number of AEWs that exceed intensity thresholds in each model’s present-day and end-of-the-21st-century periods.

Finally, the EKE is calculated as follows:

$${{{\rm{EKE}}}}=\frac{{u}^{{{\prime} }^{2}}+{v}^{{{\prime} }^{2}}}{2}$$
(1)

where u and v represent the zonal and meridional winds, respectively. The primes indicate band-pass filtered anomalies of the daily wind field. We have filtered u and v for 3–5- and 6–9-day periods.

To better understand the factors contributing to changes in EKE, we employ the energetics analyses originally formulated by Lorenz62 for the general circulation of the atmosphere and later modified for a limited area63,65 by incorporating energy transport at the boundaries. AEWs can be regarded as eddy disturbances propagating along the zonal flow, with energy conversions classified into contributions from the zonal mean state and eddies65. The primary source of energy for eddies is the available potential energy, which is subsequently converted into kinetic energy62. Both available potential and kinetic energy can be decomposed into their mean and eddy components. The generation of EKE in AEWs is driven primarily by baroclinic and barotropic energy conversions, as well as diabatic heating processes85. These energy conversions are described by the Lorenz energy cycle, which governs the interaction between EKE (KE) and available potential energy in an open system and can be expressed as follows:

$$\frac{\partial {K}_{E}}{\partial t}={C}_{k}+{C}_{{pk}}-{D}_{E}+{K}_{{EB}}+{\phi }_{{EB}}$$
(2)
$$\frac{\partial {A}_{E}}{\partial t}={C}_{A}-{C}_{{pk}}+{G}_{E}+{A}_{{EB}}$$
(3)

where \({K}_{E}\) is the average EKE; \({A}_{E}\) is the EAPE; and \({C}_{k}\) is barotropic energy conversion, which captures the conversion of zonal to EKE through zonal (u) and meridional (v) wind shears. Also, \({C}_{{pk}}\) represents the baroclinic energy conversion term, accounting for the conversion of \({A}_{E}\) to \({K}_{E}\) through vertical overturning. Note that the reappearance of this term with opposite sign in Eq. 3 indicates that the consumed \({A}_{E}\) is converted to \({K}_{E}\), or vice versa. \({D}_{E}\) represents the dissipation of energy due to friction, whereas \({K}_{{EB}}\) and \({A}_{{EB}}\) are the boundary fluxes of \({K}_{E}\) and available potential energy, respectively, advecting into and out of the region of interest. \({\phi }_{{EB}}\) is the boundary pressure work done by the eddies.

The conversion term of ZAPE to \({A}_{E}\) by means of the eddy heat flux along the zonal mean temperature gradient is given by \({C}_{A}\) in Eq. 3. The generation of \({A}_{E}\) by diabatic heating is given as \({G}_{E.}\) A positive value of \({G}_{E}\) (generation of \({A}_{E}\)) results from heating in warmer regions or cooling in colder regions at the same latitude69. In contrast, a negative value of \({G}_{E}\) (destruction of \({A}_{E}\)) is due to cooling in warmer regions or heating in colder regions at the same latitude.

The energy conversion terms are calculated as

$${C}_{k}= -\overline{\overrightarrow{{V}_{H}^{{\prime} }}.(\overrightarrow{{V}^{{\prime} }}.\nabla )\overrightarrow{{V}_{H}}}$$
(4)
$${C}_{{pk}}= -\frac{R}{p}\overline{{\omega }^{{\prime} }{T}^{{\prime} }}$$
(5)
$${C}_{A}= -\frac{{c}_{p}\gamma }{\bar{T}}\overline{\overrightarrow{{V}_{H}^{{\prime} }}{T}^{{\prime} }}.{\nabla }_{H}\bar{T}$$
(6)
$${G}_{E}= \frac{\gamma }{\bar{T}}\overline{{Q}_{1}^{{\prime} }{T}^{{\prime} }}$$
(7)

where \(T\) denotes temperature, \(\omega =\frac{\partial p}{\partial t}\) is vertical velocity, \({V}_{H}\) is horizontal wind speed, and p is pressure. Also, for \(\gamma =\frac{{\varGamma }_{d}}{{\varGamma }_{d}-\varGamma }\), \({\varGamma }_{d}\) and \(\varGamma\) are the dry adiabatic and observed lapse rates, respectively; \({C}_{p}\) is the heat capacity at constant pressure; \(R\) is the dry air gas constant; and \({Q}_{1}\) is the apparent heat source.

The apparent heat source is calculated as:

$${Q}_{1}=\frac{{c}_{p}T}{\theta }\left(\frac{\partial \theta }{\partial t}+u\frac{\partial \theta }{\partial \phi }+v\frac{\partial \theta }{\partial \lambda }+\omega \frac{\partial \theta }{\partial p}\right)$$
(8)

where \(\theta\) is the potential temperature, and u and v are zonal and meridional velocities. Primes in these equations are calculated using the two band-pass filters. Following Hsieh and Cook65, higher-order terms (i.e., those involving triple products of perturbations) in Eqs. 1 and 2 are neglected in the computation. The positive and negative values in these figures represent gain and loss of EKE or EAPE, respectively.