1. Introduction
Due to their substantial social and economic significance, understanding how regional monsoons are projected to change in the future is of great importance (e.g., Akinsanola and Zhou 2020; Raj et al. 2019; Sahastrabuddhe et al. 2023; Katzenberger et al. 2021). In the Intergovernmental Panel on Climate Change Sixth Assessment Report (IPCC AR6), it was shown that compared to other monsoon regions, West African monsoon (WAM) precipitation projections are much less certain, particularly, for long-term high emission scenarios (Masson-Delmotte et al. 2021 and references therein). Given the WAM supports millions of people and sustains a large agricultural sector in the region, understanding how WAM precipitation is likely to change in the future is of high importance and is key to providing suitable and well-informed adaptation policies (Cook and Vizy 2019; Raj et al. 2019).
To better understand this uncertainty in WAM precipitation projections, a variety of approaches have previously been taken. Monerie et al. (2020b) decomposed the Coupled Model Intercomparison Project (CMIP5 and CMIP6) intermodel precipitation uncertainty into dynamic and thermodynamic components. The thermodynamic component represents differences in moisture availability, which is expected to increase in a warmer climate (Chadwick et al. 2016), while the dynamic component represents the different atmospheric circulation responses across models. It was shown that much of the intermodel uncertainty results from the dynamic term, which Monerie et al. (2020b) linked to the atmospheric response to changes in sea surface temperatures (SSTs), particularly over the North Atlantic.
Understanding uncertainty in WAM projections caused by SST patterns is particularly challenging, since there is uncertainty caused by the different SST patterns projected by models as well as differing model responses to a given SST pattern change. Guilbert et al. (2024) employed metrics for both interhemispheric asymmetry in warming and tropical Pacific warming to explain uncertainty in WAM projections. They show that these SST patterns influence the WAM through shifts in the intertropical convergence zone (ITCZ) and Walker circulation, respectively. Monerie et al. (2023) investigated intermodel uncertainty with a “storylines” approach showing that both warming in the North Atlantic and the Mediterranean contribute to increased WAM precipitation in the future, consistent with Park et al. (2016). Park et al. (2015) investigated the influence of differential warming between NH tropics and extratropics. Giannini et al. (2013) developed an index whereby both past WAM variability and differing future projections in WAM precipitation could be understood through the difference between North Atlantic and tropical mean SSTs. They noted that the tropical mean SSTs play an important role in setting the vertical atmospheric stability across the tropics and setting the temperature threshold for deep convection (Johnson and Xie 2010). Provided sufficiently high temperature and moisture conditions are met in the lower troposphere and this threshold is exceeded, convection can occur. Since the Atlantic is a key moisture source for the WAM (Lélé et al. 2015; Gong and Eltahir 1996), the Atlantic SSTs play an important role in determining to what extent this convective threshold is met in the WAM region. Giannini et al. (2013) focus more specifically on the North Atlantic SSTs as being the most relevant to the WAM, finding strong correlations between monsoon precipitation and the difference between North Atlantic and tropical mean SSTs across a range of time scales and across different models. Bellomo et al. (2021) showed that variability in North Atlantic SSTs is related to differing projections of the AMOC, and the SSTs caused by this AMOC variability have large impacts on the latitudinal displacement of the ITCZ and the midlatitude jet.
Land–atmosphere coupling has also been found to be a key source of uncertainty. Dosio et al. (2020) took a different approach to understanding the key differences in models’ WAM projections under the RCP8.5 emission scenario. Regional climate models were used to downscale six CMIP5 GCMs, and these models were grouped into “wet” and “dry” classifications, where the wet models were those that projected an increase in West African precipitation and the dry models were those that projected a decrease. This grouping highlighted that a key difference between wet and dry models was associated with land–atmosphere coupling, with large discrepancies in soil moisture and evapotranspiration changes between the two groups. Mutton et al. (2022, 2024) demonstrated that a soil moisture–surface heat flux feedback over the Sahel can act to amplify any increases or decreases in WAM precipitation. It is possible that this feedback mechanism could be a key contributor to the model diversity highlighted by Dosio et al. (2020).
Previously, Mutton et al. (2022, 2024) employed a decomposition of the full coupled atmosphere–ocean model response to increasing CO2 into a number of direct and indirect drivers such as the direct radiative effect or a uniform ocean warming to better understand the mechanisms responsible for changes in WAM precipitation. Using this decomposition, it was shown that in response to the direct radiative effect of increased CO2, the WAM precipitation increases (Mutton et al. 2022; Gaetani et al. 2017; Chadwick et al. 2019; Biasutti 2019). This increase was shown to be caused by differential heating between the moist monsoon air mass and the drier desert air mass to the north, generating a weakening and a northward shift in the shallow meridional circulation, which climatologically tends to advect dry air in to the monsoon rainband at midlevels and inhibit precipitation (Shekhar and Boos 2017; Zhang et al. 2008). It was also shown that a local soil moisture–surface heat flux feedback over the Sahel acted to amplify circulation changes and determine the location of the intertropical discontinuity (Mutton et al. 2022). The mechanism discussed in Mutton et al. (2022) is summarized in Fig. 1a. Investigating the WAM response to a uniform ocean warming, shown to cause a decrease in WAM precipitation (Hill et al. 2017; Held et al. 2005; Gaetani et al. 2017; Biasutti 2013, 2019), Mutton et al. (2024) demonstrated that this decrease in WAM precipitation was caused by a strengthening of the 700-hPa moisture flux divergence associated with the shallow meridional circulation (due to the effect of both increased moisture gradients over the Sahel and a strengthening of the circulation itself), and a southward shift in the African Easterly Jet. Similar to the direct radiative effect mechanism discussed, a soil moisture–surface heat flux feedback over the Sahel also acts to amplify changes in circulation and precipitation. These mechanisms are summarized in Fig. 1b. Gaetani et al. (2017) also used a similar decomposition, suggesting that the spread in projections may be caused by the balance of the WAM response to the opposing signals from the direct radiative effect and a uniform ocean warming.
Diagram depicting the key processes responsible for (a) an increase in WAM precipitation in response to the direct radiative effect of increased CO2 and (b) a decrease in WAM precipitation in response to a uniform ocean warming (Mutton et al. 2022, 2024). Here, the dashed line around the box in (b) highlights the fact that this feature was predominantly present in an adjustment period over the days following an abrupt ocean warming. There is some evidence that the warming pattern mentioned is seen in the long-term steady-state response to ocean warming, particularly prior to the onset of the monsoon season, but more work is needed to investigate this further (Mutton et al. 2024).
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
Building on the mechanistic understanding provided by Mutton et al. (2022, 2024), as well as the SST pattern effect understanding provided by Giannini et al. (2013) and Guilbert et al. (2024), in this paper, we employ the decomposition used by Mutton et al. (2024) to better understand the different sources of uncertainty in the WAM precipitation response to increased CO2 across the CMIP6 ensemble. Using such a decomposition simplifies the response to increased CO2 and allows us to isolate the uncertainty in projections associated with different components of the response. Here, the direct radiative effect, the uniform ocean warming, and the patterned SST change are shown to be the largest source of uncertainty, and it is shown that the mechanisms found to be responsible for WAM precipitation changes in Mutton et al. (2022, 2024), in particular those associated with the shallow meridional circulation, may also contribute to the large spread in WAM projections. The uncertainty associated with the patterned SST change is investigated, using the indices developed by Giannini et al. (2013) and Guilbert et al. (2024), specifically the difference between North Atlantic and Tropical SSTs and the interhemispheric asymmetry in surface temperatures.
Section 2 describes the data and methods used, and section 3 presents the results. The results have been split into four subsections where section 3a presents the decomposition used, section 3b investigates the intermodel spread associated with the direct radiative effect, section 3c investigates the intermodel spread associated with a uniform SST change, and section 3d investigates the intermodel spread associated with a patterned SST change. Finally, a discussion and conclusions are presented in section 4.
2. Data and methods
a. Experiment design
Two different experimental setups are used to isolate the different components of the response to the full forcing of increased CO2. This response to the full forcing of increased CO2 is defined as the difference between piControl and abrupt-4xCO2 experiments, where the piControl experiment is a coupled model experiment with preindustrial atmospheric constituents and the abrupt-4xCO2 experiment is a coupled model experiment where the CO2 concentrations have been abruptly quadrupled relative to preindustrial levels. The more accurate and comprehensive of the two decompositions used involves the CMIP6 piSST timeslice experiments. The piSST experiment is an atmosphere-only experiment that uses preindustrial atmospheric constituents and prescribed monthly varying SST from years 111 to 140 of the models own piControl simulation. piSST-4xCO2-rad is similar to piSST except that the CO2 concentrations seen by the radiation scheme have been multiplied by 4. Since this increase in CO2 is only seen by the radiation scheme, there are no changes due to the plant physiology response to increased CO2. The difference between piSST-4xCO2-rad and piSST gives the direct radiative effect. The piSST-4xCO2 experiment is identical to the piSST-4xCO2-rad experiment except the increase in CO2 is seen by both the vegetation and radiation scheme. Taking the difference between these two experiments provides the plant physiological effect, which captures the response to changes in plant stomata, which tend to close with increased CO2, as well as changes in leaf area index in some models (Chadwick et al. 2019). The piSST-pxK experiment is similar to the piSST experiment; however, here a uniform SST warming has been applied. The magnitude of this SST warming is calculated as the global climatological mean change in SSTs between years 111 and 140 of the model’s own piControl and abrupt-4xCO2 experiments. The difference between piSST-p4K and piSST provides the impact of a uniform ocean warming. The a4SST experiment is also similar to the piSST experiment but has monthly varying SSTs prescribed from years 111 to 140 of the model’s own abrupt-4xCO2 experiment. Taking the difference between a4SST and piSST-pxK provides the impact of a patterned SST change. a4SSTice is similar to the a4SST experiment but also has monthly varying sea ice prescribed from the models own abrupt-4xCO2 experiment. Comparing a4SST and a4SSTice provides the impact of changing sea ice. Finally, a4SSTice-4xCO2 is similar to the a4SSTice experiment but also has the CO2 concentrations set to four times preindustrial levels seen by both the vegetation and radiation scheme. Comparing a4SSTice-4xCO2 and piSST captures the atmosphere-only model’s full response to increased CO2. The experimental setup of these timeslice experiments is summarized in Table 1, and the different components of the full response are calculated as stated in Table 2 (Chadwick et al. 2017). The experimental setup is also described in Webb et al. (2017). This decomposition is discussed further in section 3a.
Description of experimental setup for piSST-based timeslice experiments used to decompose the abrupt-4xCO2 response. All experiments are run for 30 years. Four models have simulated these experiments.
Definition of components of full forcing of increased CO2 using timeslice experiments. Decomposition presented in Figures 3 and 4 and experimental setup described in Table 1.
Although these piSST timeslice experiments do give a comprehensive decomposition of the full response to increased CO2, across CMIP6, only four models have performed these simulations. Therefore, to capture the responses to different components of the full CO2 forcing in a larger ensemble of models, amip, amip-4xCO2, and amip-p4K experiments are used (Table 3). The amip experiment is an AGCM experiment forced by observed historical SSTs, sea ice, and atmospheric constituents between 1979 and 2014. The amip-4xCO2 uses an identical setup, only the CO2 concentrations are multiplied by 4, and this increase in CO2 is only seen by the radiation scheme, therefore causing no changes to plant physiology. In the case of amip-p4K, this also uses an identical setup to amip, and only a uniform warming of 4 K is applied to the SSTs. The impact of the direct radiative effect of increased CO2 is isolated by taking the difference between amip-4xCO2 and amip experiments, and the impact of a uniform ocean warming is calculated as the difference between amip-p4K and amip experiments. This decomposition has previously been used by Monerie et al. (2020a), Gaetani et al. (2017), and Mutton et al. (2022, 2024).
Description of experimental setup for amip experiments used. Each experiment runs from 1979 to 2014. Thirteen models have simulated these experiments.
Thirteen models have run the amip-based experiments shown in Table 3: CESM2, BCC-CM2-MR, CNRM-CM6-1, HadGEM3-GC31-LL, IPSL-CMA6A-LR, MRI-ESM2-0, CanESM5, MIROC6, GISS-E2-1-G, GFDL-CM4, TaiESM1, E3SM-1-0, and NorESM2-LM. Four models have run the timeslice experiments described in Table 1: HadGEM3-GC31-LL, IPSL-CM6A-LR, CESM2, and CNRM-CM6-1. In both sets of experiments, 1 ensemble member of each simulation was used. In figures where multimodel means have been calculated, data have been regridded onto a 1.85° longitude × 1.25° latitude grid. All model data used in this analysis are accessible from the ESGF CMIP6 site (Eyring et al. 2016).
b. Regions analyzed
Consistent with the analysis of Shekhar and Boos (2017) and Mutton et al. (2022, 2024), much of the analysis focuses on a cross section of the atmosphere, zonally averaging between 10°W and 25°E (Fig. 2, green lines). The WAM region (Fig. 2, dark blue box) is defined as a region between 10°W–25°E and 7°–15°N. This box captures much of the region where the June–August precipitation minus December–February precipitation is greater than 180 mm and the June–August precipitation accounts for over 35% of the total annual rainfall in GPCP observations between 1980 and 2010. These criteria are used by Wang and Ding (2006) to characterize monsoon regions. The cyan box and the orange lines are used to define northern tropical Atlantic SSTs and tropical mean surface temperatures, respectively, as used by Giannini et al. (2013) to calculate their index for capturing WAM variability in response to changing SST patterns.
Regions used for analysis; the green lines indicate the longitudinal bounds used for cross-sectional analysis and the dashed dark blue box indicates the WAM region. The dashed cyan box indicates the region used to characterize the North Atlantic SSTs, and the orange lines indicate the latitudinal bounds used to calculate the tropical mean SSTs.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
c. Definition of key quantities
1) Low-level atmospheric thickness
2) Horizontal divergences
Ideally, the analysis of horizontal moisture flux divergence would be performed using high temporal resolution (6 hourly) data (Seager et al. 2014). However, due to incomplete data availability in even daily data, monthly data have been used. To demonstrate the impact of using this coarser temporal resolution data, a comparison between monthly and daily data is provided in Fig. S1 in the online supplemental material. This shows that although this caveat should be taken into consideration, it would not likely alter the conclusions drawn from the results.
3. Results
The results of this analysis have been split into four sections. First, in section 3a, the WAM precipitation decomposition using the piSST-based and amip-based experiments is presented and the relative importance of each driver of WAM precipitation change to the intermodel spread in projections is shown. It is found that the largest contributions to intermodel spread come from the different responses to the direct radiative effect, the uniform ocean warming, and a patterned SST change. Sections 3b–d then investigate these components further applying the results of Mutton et al. (2022, 2024), Giannini et al. (2013), and Guilbert et al. (2024), to investigate whether the mechanisms previously shown to drive the WAM precipitation change in response to these drivers are also key to understanding the spread in model projections seen across a subset of CMIP6 models.
a. Timeslice decomposition
WAM precipitation change in response to a quadrupling of CO2 in the abrupt-4xCO2 experiment as well as the decomposition of this change using the piSST and amip-based experiments is shown in Fig. 3. As previously mentioned, the spread in the abrupt-4xCO2 response is substantial, with both large increases and large decreases seen across the 41-model CMIP6 ensemble (Fig. 3, boxplot). The large WAM precipitation decrease seen in some models is particularly policy relevant and would have substantial socioeconomic impacts were such a decrease to occur. This emphasizes the importance of improving projections in WAM precipitation.
WAM precipitation response to different components of the piSST-based timeslice decomposition of the full coupled CO2 response in four CMIP6 GCMs (see Tables 1 and 2 for the description of each component of the decomposition). The coupled response is defined as the difference between piControl and abrupt-4xCO2 experiments. WAM precipitation is defined as the area average precipitation over the blue box indicated in Fig. 1 between June and August. The dots alongside the direct radiative effect and uniform SST warming columns indicate the intermodel spread in the 13-model ensemble of amip-based experiments (see Table 3). Here, the different colored dots indicate whether it is in the stronger or weaker groups used later in sections 3b and 3c. The boxplot in the coupled model column indicates the intermodel spread in a 41-member ensemble of CMIP6 models. Lower and upper box boundaries indicate the interquartile range with central horizontal line indicating the median. Whiskers indicate the maximum and minimum values.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
Decomposing the abrupt-4xCO2 response into the different components demonstrates that the direct radiative effect of increased CO2 causes an increase in WAM precipitation, the uniform SST warming causes a decrease in WAM precipitation, and the response to a patterned SST change is uncertain, with a comparable spread across the four piSST models to that seen in response to the direct radiative effect or uniform SST warming. The spread in the response to sea ice melt or changing plant physiology is small compared to that of the direct radiative effect, uniform SST warming, and patterned SST changes, as is the spread in the residual term, which includes the nonlinearities in the decomposition. The direct radiative effect and uniform SST warming responses are also shown for the amip-based decomposition, and the associated abrupt-4xCO2 response for this ensemble is indicated. The increased ensemble size (13 models) compared to the 4-model ensemble for the piSST decomposition helps to capture a greater intermodel spread and also allows us to sample models spanning the entire range of the larger 41 model CMIP6 ensemble.
The spatial characteristics of the response to different components of the piSST decomposition are presented in Fig. 4. Consistent with Fig. 3, the largest components of the decomposition are the direct radiative effect, the uniform SST warming, and the patterned SST change, although the response to the patterned SST change is generally small in three of the four models and large and negative in CESM2. Again, the response to changes in sea ice and plants is shown to be small and not significant, and while the nonlinear term is also small compared to the direct radiative effect and uniform SST change, there are some regions where the small but negative anomalies are significant. Comparing the precipitation changes seen in the full AGCM response and the coupled abrupt-4xCO2 response, it can be seen that although there are some regions where small differences in the response emerge, broadly the piSST decomposition does a good job at capturing the coupled abrupt-4xCO2 response with spatial correlations between the two responses ranging from 0.93 to 0.99 across the four models. Given that the nonlinear term has very little intermodel spread, is small compared to the direct radiative effect, or uniform ocean warming response, and the piSST decomposition does a relatively good job of decomposing the abrupt-4xCO2 response, by better understanding the spread in the response to the direct radiative effect, uniform SST warming, and patterned SST change we are able to better understand the causes of intermodel spread in the wider CMIP6 model ensemble.
(top) Maps of JJA precipitation climatology and (all other panels) anomaly each of the different components of the piSST-based decomposition of the full forcing of increased CO2 (see Tables 1 and 2 for a description of the different experiments used). (ag),(ah),(ai),(aj) The coupled model response defined as the difference between abrupt-4xCO2 and piControl experiments. For all panels, stippling is used to indicate where the precipitation anomalies are not significant at the 95% confidence interval. Here, a t test is applied using the null hypothesis that the time mean of the precipitation anomalies in each location is zero.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
In the following sections, these components are analyzed individually using the larger (13 model) amip experiment–based ensemble to investigate the spread in the response to the direct radiative effect and uniform SST warming, and using the 41-member ensemble of abrupt-4xCO2 simulations to investigate the impact of different SST patterns.
b. Intermodel spread—Direct radiative effect
As discussed in section 1 and summarized in Fig. 1a, the direct radiative effect of increased CO2 causes an increase in WAM precipitation due to a northward shift and a weakening of the shallow meridional circulation, associated with large-scale temperature gradient changes, as well as the more local influence of a soil moisture feedback over the Sahel. Mutton et al. (2022) focused primarily on a single-model analysis of HadGEM2-A as well as the CMIP6 ensemble mean. Here, the 13-model ensemble of amip-4xCO2 experiments is analyzed and to identify key differences between models that project larger or smaller precipitation changes in response to the direct radiative effect, two groupings of models are used to capture the four weakest and four strongest responses (Fig. 3). The strongest 4 (S44xCO2) group consists of CNRM-CM6-1, HadGEM3-GC31-LL, IPSL-CM6A-LR, and MIROC6, while the weakest 4 (W44xCO2) group consists of GISS-E2-1-G, CESM2, NorESM2-LM, and MRI-ESM2-0.
In this analysis, zonal mean responses are investigated and to ensure that the results are not swayed by one particular model, the group means are calculated four times, leaving one model out each time. To test the significance of the difference between the response seen in the two groups, a t test is performed at each latitude, using the response over each time and longitude within the region to capture the spread in the response for each group mean. Performing this t test for every combination of the three model means from each group, we state a difference is only significance if the t test for every combination produces a p value below 1%. The analysis here focuses on several of the key processes identified by Mutton et al. (2022), investigating how these processes are simulated in the two model groups and therefore how they contribute to the uncertainty in WAM projections.
The precipitation climatology (amip) and direct radiative effect anomaly (amip-4xCO2–amip) in the S44xCO2 and W44xCO2 groups are presented in Fig. 5; there key difference between the two groups can be seen. In the climatology, the models with a stronger response tend to have a precipitation climatology shifted further south, with significantly more precipitation south of 7°N and significantly less precipitation north of 9°N. Comparing these distributions to the GPCP observations, the weaker models do a better job at capturing the northward extent of the rainband but simulate too little precipitation to the south. Comparing the precipitation response to the direct radiative effect, by construction the S44xCO2 models exhibit larger increases in precipitation. This increase is seen across almost the whole monsoon region, while the W44xCO2 models simulate a weaker increase and a dipole-type pattern with decreases in precipitation to the south and increases in precipitation to the north. Although previously Monerie et al. (2017, 2020b) and Yan et al. (2019) find that no relationship exists between climatological biases and future changes, with Monerie et al. (2020b) investigating the biases in coupled historical experiments and the response in the RCP85 scenario, these results show that at least in the case of the response to the direct radiative effect, there may be a relationship between a model’s climatology and its response.
JJA precipitation averaged between 10°W and 25°E (green dashed lines in Fig. 1) for (a) amip climatology and (b) amip-4xCO2–amip anomaly in the strongest 4 (S44xCO2) and weakest 4 (W44xCO2) models with respect to the magnitude of the direct radiative effect anomaly. Note that the four models in each group have been sampled four times, missing one model out each time. Therefore, each colored line represents an ensemble mean made up of three of the four models. Gray shading has been used to highlight where the difference between the two groups is not significant at the 1% confidence interval. (a) GPCP observed precipitation averaged between 1979 and 2019.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
In response to the direct radiative effect, the increase in precipitation has been shown to be associated with a northward shift and weakening of the shallow meridional circulation (Mutton et al. 2022). Changes in this shallow circulation can be characterized using the 700-hPa horizontal moisture flux divergence with a region of divergence between approximately 15° and 20°N. Any changes in moisture flux divergence can also be decomposed into dynamic and thermodynamic components (Fig. 6).
The JJA zonal mean of S44xCO2 and W44xCO2: (a) climatological and (b) anomalous (amip-4xCO2–amip) 700-hPa horizontal moisture flux divergence. The climatological horizontal divergence of moisture transport from the ERA5 reanalysis is also shown in (a). The total change in horizontal moisture flux divergence shown in (b) has also been decomposed into (c) a dynamic component and (d) a thermodynamic component. Similar to Fig. 5, the four models in each group have been sampled four times, missing one model out each time. Therefore, each colored line represents an ensemble mean made up of three of the four models. Gray shading has been used to highlight where the difference between the two groups is not significant at the 1% confidence interval.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
Climatologically, it can be seen that the S44xCO2 models have a stronger 700-hPa divergence associated with the shallow circulation located slightly further south compared to the W44xCO2 models (Fig. 6a). This is consistent with the climatological precipitation seen in the two groups, with the monsoon rainband in the S44xCO2 also being located southward of the W44xCO2 models. To compare the climatologies in the two groups with observations, horizontal moisture flux divergence has also been calculated using the ERA5 reanalysis (Fig. 6a). There the location and magnitude of the maximum divergence in ERA5 more closely match the W44xCO2.
In response to the direct radiative effect, the S44xCO2 models have a larger decrease in the shallow circulation strength, and while the distribution of anomalous horizontal moisture flux divergence in the W44xCO2 models is similar, the magnitude of these anomalies is smaller. Comparing the percentage change in total moisture flux divergence associated with the shallow circulation, a −22% change is seen in the S44xCO2 models and a −17% change is seen in the W44xCO2 models. Here, this total moisture flux divergence change has been estimated averaging between 8° and 26°N to account for both sides of the dipole seen in the direct radiative effect response, similar to Mutton et al. (2024). Decomposing the total change in horizontal moisture flux divergence into dynamic and thermodynamic components (Figs. 6c,d), the dynamic term contributes more to the weakening of the moisture flux divergence in both the W44xCO2 and S44xCO2 models. However, the thermodynamic term here does play a more important role compared to the HadGEM2-A analysis in Mutton et al. (2022). The results presented in Fig. 6 demonstrate that the weakening and northward shift of the shallow circulation is strongest in the models with the largest changes in precipitation.
Changes in the shallow meridional circulation have been linked to large-scale temperature gradient changes using LLAT (Shekhar and Boos 2017), where more warming over the drier desert to the north compared to the monsoon region leads to a northward shift and weakening of the circulation (Mutton et al. 2022). The climatological LLAT in the S44xCO2 and W44xCO2 models is shown in Fig. 7a, where both S44xCO2 and W44xCO2 models are relatively similar. In response to the direct radiative effect (Fig. 7b), the S44xCO2 and W44xCO2 models are again similar in the northern part of the domain. However, to the south, the W44xCO2 models demonstrate a much larger increase in the LLAT. This difference means that the large-scale gradient in the LLAT is larger for the S44xCO2 compared to the W44xCO2 models. The stronger LLAT gradient in the S44xCO2 models would be expected to lead to a larger change in the shallow circulation and therefore lead to a larger increase in precipitation, as are both seen.
(a) JJA, (b) LLAT, (c),(d) surface latent heat fluxes, (e),(f) surface sensible heat fluxes, and (g),(h) 2-m temperature in the S44xCO2 and W44xCO2 models zonally averaged between 10°W and 25°E in (a),(c),(e),(g) amip climatology and (b),(d),(f),(h) amip-4xCO2–amip anomaly. Black lines in (a), (c), (e), and (g) show the ERA5 reanalysis JJA climatology. Similar to Fig. 5, the four models in each group have been sampled four times, missing one model out each time. Therefore, each colored line represents an ensemble mean made up of three of the four models. Gray shading has been used to highlight where the difference between the two groups is not significant at the 1% confidence interval.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
Changes in the shallow meridional circulation have also been shown to be amplified by a soil moisture–surface heat flux feedback over the Sahel (Mutton et al. 2022). In response to the precipitation increase over the Sahel, the surface latent heat flux increases and the surface sensible heat flux decreases. This change in surface heat flux causes the near-surface temperatures to cool, acting to locally enhance the large-scale changes in the temperature gradient. These temperature changes lead to circulation changes, which positively feedback on the initial precipitation change. To demonstrate how this feedback is simulated in the S44xCO2 and W44xCO2 models, zonally averaged surface heat fluxes in the amip experiment and the amip-4xCO2–amip anomaly are shown in Figs. 7c–f, and air temperature 2 m above the surface is shown in Figs. 7g and 7h.
From Figs. 7d and 7f, it is clear that the soil moisture response is much more pronounced in the S44xCO2 compared to the W44xCO2, with a strong increase in latent heat flux and a decrease in sensible heat flux. Although the W44xCO2 models do capture increases in latent heat flux north of 15°N, the decrease in sensible heat is not evident. This result is consistent with the fact that the S44xCO2 models have a larger increase in precipitation. The W44xCO2 models also have both the climatological distribution and the direct radiative effect anomalies in surface heat fluxes shifted further northward compared to the S44xCO2 model (Figs. 7c–f). This is consistent with the results shown in Fig. 5 where the climatological precipitation distribution in the W44xCO2 models also extends further north.
The 2-m temperatures respond to these changes in sensible heat flux. Figures 7g and 7h indicate that both the S44xCO2 and W44xCO2 models exhibit similar large-scale climatological and anomalous patterns in 2-m temperature with the transition between the cooler monsoon region to the warmer Sahara shifted further north in the W44xCO2 models. In response to the direct radiative effect (Fig. 7h), both subsets of models exhibit similar large-scale warming patterns with more warming to the north compared to the south. The S44xCO2 models exhibit a larger increase in the large-scale temperature gradient with a smaller warming to the south and a larger warming to the north compared to the W44xCO2 models. Both S44xCO2 and W44xCO2 exhibit a local minimum in temperature anomalies associated with the reduction in sensible heat flux, and this acts to locally enhance the large-scale gradient in surface temperatures. The minimum in anomalous temperature seen around 12°–15°N is stronger in the S44xCO2 models, meaning that the anomalous temperature gradient to the north of this minimum is larger in the S44xCO2 models compared to the W44xCO2 models. This stronger minimum in temperature is consistent with the larger decrease in sensible heat flux in the S44xCO2 models.
Here, the key differences between models that simulate a strong precipitation response to the direct radiative effect and those that simulate a weak response have been investigated. It is demonstrated that the models with a larger precipitation response also exhibit more strongly the mechanisms described in Mutton et al. (2022), including the response of the shallow circulation, the large-scale temperature gradient changes, and the local soil moisture feedback, which locally enhances the large-scale temperature gradient anomalies.
c. Intermodel spread—Uniform ocean warming
In response to a uniform ocean warming, the WAM precipitation decreases due to enhanced humidity gradients over the Sahel, increased dry air advection into the monsoon rainband from the north at 700 hPa associated with the shallow meridional circulation, and similar to the direct radiative effect mechanisms discussed, a soil moisture–sensible heat flux feedback that acts to amplify any changes in rainfall through circulation changes (Mutton et al. 2024). These processes are summarized in Fig. 1b. In this section, the intermodel spread in projections in response to a uniform ocean warming across the 13 member amip-based ensemble is investigated. Similar to section 3b, a composite approach is taken, grouping models into the strongest 4 (S4p4K) and weakest 4 (W4p4K) model responses to the uniform SST warming. The S4p4K models are those that project the largest decrease in precipitation and the W4p4K models are those that project the smallest decrease. Here, the S4p4K models are MIROC6, BCC-CM2-MR, MRI-ESM2-0, and GISS-E2-1-G, and the W4p4K models are NorESM2-LM, TaiESM1, CanESM5, and CESM2. Note that these model groups are not the same as the W44xCO2 and S44xCO2 groups used in the previous section.
The climatological (amip) and anomalous (amip-p4K–amip) precipitation in the S4p4K and W4p4K groups is shown in Fig. 8. Climatologically, although the S4p4K models generally have slightly less precipitation compared to the W4p4K models, this difference is not significant. It can also be seen that unlike the W44xCO2 and S44xCO2 models, the northward extent of both the climatological monsoon rainband and the response to the uniform ocean warming in the S4p4K and W4p4K models is relatively similar.
JJA precipitation averaged between 10°W and 25°E (green dashed lines in Fig. 1) for (a) amip climatology and (b) amip-p4K–amip anomaly in the strongest 4 (S4p4K) and weakest 4 (W4p4K) models with respect to the magnitude of the precipitation response to a uniform ocean warming. Note that the four models in each group have been sampled four times, missing one model out each time. Therefore, each colored line represents an ensemble mean made up of three of the four models. Gray shading has been used to highlight where the difference between the two groups is not significant at the 1% confidence interval. (a) GPCP-observed precipitation averaged between 1979 and 2019.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
In response to a uniform ocean warming, the WAM precipitation reduces due to changes in moisture flux divergence associated with the shallow meridional circulation. Cross sections of 700-hPa horizontal moisture flux divergence climatology and anomaly are shown in Figs. 9a and 9b, with a decomposition of the anomalous moisture flux divergence into dynamic and thermodynamic terms presented in Figs. 9c and 9d. Climatologically, similar to the direct radiative effect, the S4p4K models have a stronger and more distinct shallow circulation. This can be seen in the region of divergence at approximately 17°N. In response to the uniform SST warming, the S4p4K models also have a greater strengthening of the moisture flux divergence associated with the shallow circulation.
JJA zonal mean of S4p4K and W4p4K: (a) climatological and (b) anomalous (amip-p4K–amip) 700-hPa horizontal moisture flux divergence. The climatological horizontal divergence of moisture transport from the ERA5 reanalysis is also shown in (a). The total change in horizontal moisture flux divergence shown in (b) has also been decomposed into (c) a dynamic component and (d) a thermodynamic component. Again, the four models in each group have been sampled four times, missing one model out each time. Therefore, each colored line represents an ensemble mean made up of three of the four models. Gray shading has been used to highlight where the difference between the two groups is not significant at the 1% confidence interval.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
Decomposing these changes in moisture flux divergence into dynamic and thermodynamic terms, in both the S4p4K and W4p4K models, the thermodynamic term is large and relatively consistent in magnitude between the two groups, with a slightly stronger response seen in the S4p4K models. The dynamic term, however, reveals a key difference between the two model groups. There, the S4p4K models exhibit a strengthening of the moisture flux divergence associated with changes in the circulation itself, whereas in the W4p4K models, only very small dynamic changes are seen. This would suggest that while in the S4p4K models, the influence of the shallow circulation changes comes from changes in moisture gradients and a strengthening of the circulation, in the W4p4K, only the influence of changing moisture gradients is seen, with little changes seen in the circulation itself.
The thermodynamic component of the horizontal moisture flux divergence is caused by a strengthening of the gradient in specific humidity across the Sahel (Mutton et al. 2024; Hill et al. 2017). In response to SST warming, the climatologically moister regions (such as the WAM region) have greater increases in moisture than the climatologically drier regions (such as the Sahara). This differential change in specific humidity between the WAM region and the Sahara leads to an increased gradient in specific humidity and enhances the efficiency with which northerly winds associated with the shallow circulation dry the monsoon rainband and inhibit precipitation. Climatological (amip) and anomalous (amip-p4K–amip) 700-hPa specific humidity zonally averaged between 10°W and 25°E is shown in Figs. 10a and 10b. Here, climatologically the S4p4K and W4p4K models have relatively similar specific humidity distributions with the S4p4K models having a slightly higher specific humidity across the transect. In response to the uniform SST warming, the S4p4K models exhibit a much larger increase in the specific humidity gradient compared to the W4p4K models, with a significantly larger increase in humidity south of 13°N. This stronger gradient in the S4p4K models explains why the thermodynamic term is larger for these models compared to the W4p4K models in Fig. 9.
(a) JJA, (b) 700-hPa specific humidity, (c),(d) surface latent heat fluxes, (e),(f) surface sensible heat fluxes, and (g),(h) 2-m temperature in the S4p4K and W4p4K models zonally averaged between 10°W and 25°E in (a),(c),(e),(g) amip climatology and (b),(d),(f),(h) amip-4xCO2–amip anomaly. Black lines in (a), (c), (e), and (g) show the ERA5 reanalysis JJA climatology. Again, the four models in each group have been sampled four times, missing one model out each time. Therefore, each colored line represents an ensemble mean made up of three of the four models. Gray shading has been used to highlight where the difference between the two groups is not significant at the 1% confidence interval.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
In Mutton et al. (2024), the changes in horizontal moisture flux divergence were shown to be reinforced by a soil moisture–surface heat flux feedback that acts to amplify changes in precipitation and circulation. As precipitation decreases, surface latent heat flux also decreases and sensible heat flux increases. These changes in sensible heat flux cause anomalous warming at the surface and generate circulation changes that act to positively feedback on the initial precipitation change. This mechanism is investigated in the S4p4K and W4p4K models (Figs. 10c–f). In response to a uniform ocean warming, the S4p4K models exhibit a much larger decrease in latent heat flux and a much larger increase in sensible heat flux compared to the W4p4K models.
The changes in surface heat flux influence the circulation patterns through changes to the atmospheric temperature. The 2-m air temperature averaged zonally between 10°W and 25°E in the amip climatology and the uniform SST anomaly is presented in Figs. 10g and 10h. Here, the S4p4K models tend to also have warmer surface temperatures over the whole transect, with particularly large differences seen over the Sahara in the amip climatology compared to the W4p4K. In response to the uniform SST warming, the S4p4K tends to have slightly weaker changes in the large-scale temperature gradients, with more warming over the monsoon region and slightly less warming to the north. The largest discrepancy in 2-m temperature between the S4p4K and W4p4K models in their response to the uniform SST warming can be seen at around 15°N, in the same location as the maximum in sensible heat flux seen in Figs. 10e and 10f. This suggests that in the S4p4K models, the changes in surface heat fluxes caused by precipitation anomalies are able to influence surface temperature patterns and therefore influence circulation changes. This may also be why the dynamic component of the horizontal moisture flux divergence in the W4p4K is small, since the surface heat flux feedback does not seem to act at all as strongly compared to the S4p4K.
The results shown here indicate that the models that exhibit both a stronger enhancement of the meridional gradient in specific humidity, the presence of the soil moisture–surface heat flux feedback, and stronger changes in moisture flux divergence associated with the shallow circulation, also simulate a stronger precipitation decrease in response to a uniform ocean warming. This provides valuable information that can inform our understanding on the large spread seen in the coupled model response to increased CO2.
d. Intermodel spread—Patterned SST change
From Fig. 3, it can be seen that the patterned SST change also contributes substantially to the intermodel spread in WAM precipitation projections. In this section, the 4-model ensemble of piSST-based timeslice experiments are used to investigate the precipitation changes in response to patterned SST changes, and to explore sources of intermodel spread more robustly, the larger 41-model ensemble of abrupt-4xCO2 simulations is used to apply the metric developed by Giannini et al. (2013) and Guilbert et al. (2024).
The 2-m temperature climatology (piSST) and anomaly associated with a patterned SST change (a4SST–piSST-pxK) is plotted in Fig. 11. Here, the variations in SST projections across different models can be seen. Generally, models project a relative warming of the equatorial Atlantic and in the Gulf of Guinea and a relative cooling of the subtropical Atlantic (a pattern particularly prominent in the Northern Hemisphere in CESM2), consistent with Xie et al. (2010), Liu et al. (2005), Leloup and Clement (2009), and Vecchi and Soden (2007). IPSL-CM6A-LR and HadGEM3-GC31-LL both have substantial warming over the North Atlantic and the equatorial Atlantic warming extends furthest north in CNRM-CM6-1.
JJA maps of 2-m temperature piSST climatology (lines) and a4SST–piSST-pxK anomaly associated with a patterned SST change in (a) CESM2, (b) HadGEM3-GC31-LL, (c) CNRM-CM6-1, and (d) IPSL-CM6A-LR.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
The precipitation response to these SST pattern changes is presented in Figs. 4m–p. Given each model is being forced by different SST patterns, it is not surprising that the precipitation response to the patterned SST change is highly variable between models. The response over the ocean is generally for the precipitation to increase over the ocean regions that warm and to decrease over the ocean regions that cool. This is consistent with the warmer get wetter hypothesis (Xie et al. 2010). CESM2 shows this pattern clearly where a strong precipitation dipole is seen between the central Atlantic warming region and the subtropical cooling region. HadGEM3-GC31-LL also exhibits a similar response, only less extreme compared to CESM2, since the SST pattern changes are smaller. CNRM-CM6-1 does not exhibit the precipitation dipole evident in CESM2 and HadGEM3-GC31-LL and instead simulates two precipitation anomaly maxima over the central Atlantic. These maxima are located over two regions of anomalously warm SST indicated in Fig. 11. Inland, three of the four models project only relatively small precipitation changes in response to the patterned SST change. The exception to this trend is CESM2, where the precipitation dipole over the central Atlantic extends inland causing a large precipitation decrease over the WAM region and a precipitation increase over central Africa. The precipitation dipole evident in CESM2 and HadGEM3-GC31-LL suggests a southward shift in the oceanic intertropical convergence zone (ITCZ).
Giannini et al. (2013) suggested that the difference between north subtropical Atlantic SSTs and tropical mean SSTs is a key metric in capturing WAM precipitation variability (see Fig. 2 for the regions used), and Guilbert et al. (2024) explored the influence of interhemispheric temperature gradients on WAM uncertainty. We explore the relevance of these two metrics to the uncertainty of WAM projections within the piSST timeslice experiments, and since there are only four models within this ensemble, we also investigate the influence of these different SST patterns within the larger CMIP6 ensemble of coupled abrupt-4xCO2 experiments. This enables an estimation of the intermodel spread in the coupled models that is associated with patterned changes to the SSTs. The changes in these indices are plotted against anomalous WAM precipitation for both a4SST–piSST-pxK and abrupt-4xCO2–piControl experiments (Fig. 12). The correlation between the anomalous WAM precipitation and the anomalous Giannini et al. (2013) SST and Guilbert et al. (2024) indices are 0.67 and 0.74, respectively, suggesting a relatively strong influence of both North Atlantic SST patterns and interhemispheric temperature gradients on the WAM precipitation. It can also be seen that the SST pattern metrics and WAM precipitation responses seen in the piSST-based experiments (a4SST–piSST-pxK) are consistent with the distribution seen in response to the full forcing of increased CO2 (i.e., abrupt-4xCO2–piControl). The remainder of the intermodel spread in the precipitation anomaly is likely associated with the direct radiative effect and the uniform SST warming, together with any other SST pattern influences not correlated with the SST indices used [e.g., the influence of the Mediterranean (Park et al. 2016)].
Scatterplot of anomalous WAM precipitation against (a) North Atlantic–tropical mean SSTs and (b) Northern Hemisphere–Southern Hemisphere mean surface temperatures. Forty-one-model ensemble of abrupt-4xCO2–piControl is shown in black dots, and anomalies from the a4SST–piSST-pxK are also shown.
Citation: Journal of Climate 38, 13; 10.1175/JCLI-D-24-0506.1
4. Discussion and conclusions
The key sources of uncertainty in the WAM precipitation response to increased CO2 in the CMIP6 coupled models are found to be the direct radiative effect, the impact of a uniform SST warming, and the impact of a patterned SST change.
The results highlight that in the case of the direct radiative effect and the uniform SST warming, the models that more strongly produce the mechanisms presented in Mutton et al. (2022, 2024) also project the strongest changes in precipitation. By grouping the models into the strongest and weakest 4 by their precipitation anomaly, it is possible to show that in response to the direct radiative effect, the models that exhibit a more pronounced weakening and northward shift in the shallow circulation, a larger surface heat flux response, and a greater increase in the large-scale temperature gradient between the WAM region and the Sahara also project a larger increase in precipitation. For the response to a uniform SST warming, the models that project a larger strengthening in the shallow circulation, a stronger surface heat flux response, and a greater increase in the large-scale gradient in specific humidity between the WAM region and the Sahara also project a larger decrease in precipitation. Therefore, focusing more work on improving the representation of the shallow circulation and land surface coupling in models, and on understanding their responses to climate change, could lead to valuable improvements to our projections of future WAM precipitation under increased CO2. The influence of the shallow circulation shown here is consistent with the findings of Monerie et al. (2020b) who showed that dynamic changes in precipitation contributed more to the intermodel spread compared to thermodynamic changes, and the importance of the soil moisture–surface heat flux feedback seen is consistent with the results of Dosio et al. (2020) who identified different surface heat flux responses as a key difference between models that projected increases or decreases in WAM precipitation.
The intermodel spread in the precipitation response to a patterned SST change has also been investigated. In the timeslice experiments, all models exhibit large precipitation anomalies over the ocean with the warmer SSTs associated with increases in precipitation and cooler SSTs associated with decreases in precipitation. Inland however, only one model (CESM2) produces large precipitation anomalies over the WAM region. This model also exhibits the largest cooling of the North Atlantic SSTs, which has been shown to cause a decrease in WAM precipitation (Giannini et al. 2013). Since only four models produced the timeslice analysis, the intermodel spread in projections associated with a patterned SST change is investigated further in an ensemble of coupled models using the piControl and abrupt-4xCO2 experiments. Using indices developed by Giannini et al. (2013) and Guilbert et al. (2024), it is shown that changes in WAM precipitation are moderately well correlated (correlation coefficient of 0.67 and 0.74, respectively) with these indices. This result suggests that the correct simulation of SSTs in climate models is of great importance if accurate climate projections of WAM precipitation are to be produced.
Although this analysis does give useful insight into the key sources of uncertainty in the coupled model response to increased CO2, there are a few limitations. The first of these limitations is regarding the ensemble size of both the amip-based decomposition (13 models) and the piSST-based timeslice decomposition (4 models). The small ensemble size makes it difficult to find robust and significant correlations when investigating the intermodel spread. This limitation is particularly present when analyzing the patterned SST change using the timeslice experiments. The use of the coupled models in this analysis is useful. However, given that in the abrupt-4xCO2 experiment, the influence of the direct radiative effect and the uniform SST warming are also present, the analysis would be improved if a similarly large ensemble could be used with the timeslice experiment setup, where only the impact of the patterned SST change is investigated.
The analysis presented here also highlights the areas for future work. One result not fully explored is the possible relationship between a model climatological bias and its response to different components of the response to increased CO2. Although previously Monerie et al. (2017, 2020b) found no relationship exists between climatological biases and future changes, this was investigated in the coupled model response in the RCP85 scenario. Therefore, the possibility that the response to different components of the full forcing to increased CO2 (e.g., the direct radiative effect or a uniform ocean warming) may be constrained by climatological biases cannot be ruled out.
As well as larger ensembles providing a useful path for further analysis, new experimental designs could also help improve our understanding of changes in WAM precipitation. For example, the SST patterns from each coupled model could be used to prescribe SSTs in a single-model analysis to investigate how different potential SST pattern changes would effect the WAM in the absence of the direct radiative effect or a uniform ocean warming, similar to Zhang et al. (2018). This could help separate the uncertainty in projections sourced from the different SST patterns from different model responses to the same SST pattern change. In addition to this, another useful path for future work would be to try to understand whether the models’ WAM response to a certain SST pattern is consistent with the real world. This could be done by finding periods in observations where SST patterns have looked like those seen in the models’ response to increased CO2 and looking to see if a similar precipitation change was observed in response to those SSTs. Given SST patterns may also vary strongly with internal variability, utilizing larger ensembles to disentangle the influence of internal variability and the forced SST pattern change would also be useful.
Acknowledgments.
Harry Mutton was supported by the Met Office Hadley Centre Climate Programme funded by DSIT. Matthew Collins and F. Hugo Lambert were supported by NERC NE/S004645/1. Chris Taylor was supported by the African Monsoon Multidisciplinary Analysis-2050 project (Grant NE/M020428/1). Ruth Geen was supported by the Natural Environment Research Council (Grant NE/X014827/1). Thanks to Brian Medeiros for producing the CESM2 piSST-based timeslice experiments used in the analysis, and thanks also to the anonymous reviewers for their helpful comments and suggestions that have greatly helped to improve this manuscript.
Data availability statement.
Data used in this analysis consist of model simulations performed for CMIP6 Eyring et al. (2016) and can be accessed from the ESGF CEDA data node at https://esgf-index1.ceda.ac.uk/search/cmip6-ceda/.
REFERENCES
Akinsanola, A. A., and W. Zhou, 2020: Understanding the variability of West African summer monsoon rainfall: Contrasting tropospheric features and monsoon index. Atmosphere, 11, 309, https://doi.org/10.3390/atmos11030309.
Bellomo, K., M. Angeloni, S. Corti, and J. V Hardenberg, 2021: Future climate change shaped by inter-model differences in Atlantic meridional overturning circulation response. Nat. Commun., 12, 3659, https://doi.org/10.1038/s41467-021-24015-w.
Biasutti, M., 2013: Forced Sahel rainfall trends in the CMIP5 archive. J. Geophys. Res. Atmos., 118, 1613–1623, https://doi.org/10.1002/jgrd.50206.
Biasutti, M., 2019: Rainfall trends in the African Sahel: Characteristics, processes, and causes. Wiley Interdiscip. Rev.: Climate Change, 10, e591, https://doi.org/10.1002/wcc.591.
Chadwick, R., D. Ackerley, T. Ogura, and D. Dommenget, 2019: Separating the influences of land warming, the direct CO2 effect, the plant physiological effect, and SST warming on regional precipitation changes. J. Geophys. Res. Atmos., 124, 624–640, https://doi.org/10.1029/2018JD029423.
Chadwick, R., H. Douville, and C. B. Skinner, 2017: Timeslice experiments for understanding regional climate projections: Applications to the tropical hydrological cycle and European winter circulation. Climate Dyn., 49, 3011–3029, https://doi.org/10.1007/s00382-016-3488-6.
Chadwick, R., P. Good, and K. Willett, 2016: A simple moisture advection model of specific humidity change over land in response to SST warming. J. Climate, 29, 7613–7632, https://doi.org/10.1175/JCLI-D-16-0241.1.
Cook, K. H., and E. K. Vizy, 2019: Contemporary climate change of the African monsoon systems. Curr. Climate Change Rep., 5, 145–159, https://doi.org/10.1007/s40641-019-00130-1.
Dosio, A., and Coauthors, 2020: A tale of two futures: Contrasting scenarios of future precipitation for West Africa from an ensemble of regional climate models. Environ. Res. Lett., 15, 064007, https://doi.org/10.1088/1748-9326/ab7fde.
Endo, H., and A. Kitoh, 2014: Thermodynamic and dynamic effects on regional monsoon rainfall changes in a warmer climate. Geophys. Res. Lett., 41, 1704–1711, https://doi.org/10.1002/2013GL059158.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Gaetani, M., C. Flamant, S. Bastin, S. Janicot, C. Lavaysse, F. Hourdin, P. Braconnot, and S. Bony, 2017: West African monsoon dynamics and precipitation: The competition between global SST warming and CO2 increase in CMIP5 idealized simulations. Climate Dyn., 48, 1353–1373, https://doi.org/10.1007/s00382-016-3146-z.
Giannini, A., S. Salack, T. Lodoun, A. Ali, A. T. Gaye, and O. Ndiaye, 2013: A unifying view of climate change in the Sahel linking intra-seasonal, interannual and longer time scales. Environ. Res. Lett., 8, 024010, https://doi.org/10.1088/1748-9326/8/2/024010.
Gong, C., and E. Eltahir, 1996: Sources of moisture for rainfall in West Africa. Water Resour. Res., 32, 3115–3121, https://doi.org/10.1029/96WR01940.
Guilbert, M., P. Terray, J. Mignot, L. Ollier, and G. Gastineau, 2024: Interhemispheric temperature gradient and equatorial Pacific SSTs drive Sahel monsoon uncertainties under global warming. J. Climate, 37, 1033–1052, https://doi.org/10.1175/JCLI-D-23-0162.1.
Held, I. M., T. L. Delworth, J. Lu, K. L. Findell, and T. R. Knutson, 2005: Simulation of Sahel drought in the 20th and 21st centuries. Proc. Natl. Acad. Sci. USA, 102, 17 891–17 896, https://doi.org/10.1073/pnas.0509057102.
Hill, S. A., Y. Ming, I. M. Held, and M. Zhao, 2017: A moist static energy budget–based analysis of the Sahel rainfall response to uniform oceanic warming. J. Climate, 30, 5637–5660, https://doi.org/10.1175/JCLI-D-16-0785.1.
Johnson, N. C., and S.-P. Xie, 2010: Changes in the sea surface temperature threshold for tropical convection. Nat. Geosci., 3, 842–845, https://doi.org/10.1038/ngeo1008.
Katzenberger, A., J. Schewe, J. Pongratz, and A. Levermann, 2021: Robust increase of Indian monsoon rainfall and its variability under future warming in CMIP6 models. Earth Syst. Dyn., 12, 367–386, https://doi.org/10.5194/esd-12-367-2021.
Lavaysse, C., C. Flamant, S. Janicot, D. J. Parker, J.-P. Lafore, B. Sultan, and J. Pelon, 2009: Seasonal evolution of the West African heat low: A climatological perspective. Climate Dyn., 33, 313–330, https://doi.org/10.1007/s00382-009-0553-4.
Lélé, M. I., L. M. Leslie, and P. J. Lamb, 2015: Analysis of low-level atmospheric moisture transport associated with the West African monsoon. J. Climate, 28, 4414–4430, https://doi.org/10.1175/JCLI-D-14-00746.1.
Leloup, J., and A. Clement, 2009: Why is there a minimum in projected warming in the tropical North Atlantic Ocean? Geophys. Res. Lett., 36, L14802, https://doi.org/10.1029/2009GL038609.
Liu, Z., S. Vavrus, F. He, N. Wen, and Y. Zhong, 2005: Rethinking tropical ocean response to global warming: The enhanced equatorial warming. J. Climate, 18, 4684–4700, https://doi.org/10.1175/JCLI3579.1.
Masson-Delmotte, V., and Coauthors, 2021: Climate Change 2021: The Physical Science Basis. Cambridge University Press, 2392 pp., https://doi.org/10.1017/9781009157896.
Monerie, P.-A., E. Sanchez‐Gomez, and J. Boé, 2017: On the range of future Sahel precipitation projections and the selection of a sub-sample of CMIP5 models for impact studies. Climate Dyn., 48, 2751–2770, https://doi.org/10.1007/s00382-016-3236-y.
Monerie, P.-A., E. Sanchez‐Gomez, M. Gaetani, E. Mohino, and B. Dong, 2020a: Future evolution of the Sahel precipitation zonal contrast in CESM1. Climate Dyn., 55, 2801–2821, https://doi.org/10.1007/s00382-020-05417-w.
Monerie, P.-A., C. M. Wainwright, M. Sidibe, and A. A. Akinsanola, 2020b: Model uncertainties in climate change impacts on Sahel precipitation in ensembles of CMIP5 and CMIP6 simulations. Climate Dyn., 55, 1385–1401, https://doi.org/10.1007/s00382-020-05332-0.
Monerie, P.-A., M. Biasutti, J. Mignot, E. Mohino, B. Pohl, and G. Zappa, 2023: Storylines of Sahel precipitation change: Roles of the North Atlantic and Euro‐Mediterranean temperature. J. Geophys. Res. Atmos., 128, e2023JD038712, https://doi.org/10.1029/2023JD038712.
Mutton, H., R. Chadwick, M. Collins, F. H. Lambert, R. Geen, A. Todd, and C. M. Taylor, 2022: The impact of the direct radiative effect of increased CO2 on the West African monsoon. J. Climate, 35, 2441–2458, https://doi.org/10.1175/JCLI-D-21-0340.1.
Mutton, H., R. Chadwick, M. Collins, F. H. Lambert, C. M. Taylor, R. Geen, and A. Todd, 2024: The impact of a uniform ocean warming on the West African monsoon. Climate Dyn., 62, 103–122, https://doi.org/10.1007/s00382-023-06898-1.
Park, J.-Y., J. Bader, and D. Matei, 2015: Northern-hemispheric differential warming is the key to understanding the discrepancies in the projected Sahel rainfall. Nat. Commun., 6, 5985, https://doi.org/10.1038/ncomms6985.
Park, J.-Y., J. Bader, and D. Matei, 2016: Anthropogenic Mediterranean warming essential driver for present and future Sahel rainfall. Nat. Climate Change, 6, 941–945, https://doi.org/10.1038/nclimate3065.
Raj, J., H. K. Bangalath, and G. Stenchikov, 2019: West African monsoon: Current state and future projections in a high-resolution AGCM. Climate Dyn., 52, 6441–6461, https://doi.org/10.1007/s00382-018-4522-7.
Sahastrabuddhe, R., S. A. Ghausi, J. Joseph, and S. Ghosh, 2023: Indian summer monsoon rainfall in a changing climate: A review. J. Water Climate Change, 14, 1061–1088, https://doi.org/10.2166/wcc.2023.127.
Seager, R., H. Liu, N. Henderson, I. Simpson, C. Kelley, T. Shaw, Y. Kushnir, and M. Ting, 2014: Causes of increasing aridification of the Mediterranean region in response to rising greenhouse gases. J. Climate, 27, 4655–4676, https://doi.org/10.1175/JCLI-D-13-00446.1.
Shekhar, R., and W. R. Boos, 2017: Weakening and shifting of the Saharan shallow meridional circulation during wet years of the West African monsoon. J. Climate, 30, 7399–7422, https://doi.org/10.1175/JCLI-D-16-0696.1.
Vecchi, G. A., and B. J. Soden, 2007: Effect of remote sea surface temperature change on tropical cyclone potential intensity. Nature, 450, 1066–1070, https://doi.org/10.1038/nature06423.
Wang, B., and Q. Ding, 2006: Changes in global monsoon precipitation over the past 56 years. Geophys. Res. Lett., 33, L06711, https://doi.org/10.1029/2005GL025347.
Webb, M. J., and Coauthors, 2017: The Cloud Feedback Model Intercomparison Project (CFMIP) contribution to CMIP6. Geosci. Model Dev., 10, 359–384, https://doi.org/10.5194/gmd-10-359-2017.
Xie, S.-P., C. Deser, G. A. Vecchi, J. Ma, H. Teng, and A. T. Wittenberg, 2010: Global warming pattern formation: Sea surface temperature and rainfall. J. Climate, 23, 966–986, https://doi.org/10.1175/2009JCLI3329.1.
Yan, Y., R. Lu, and C. Li, 2019: Relationship between the future projections of Sahel rainfall and the simulation biases of present South Asian and western North Pacific rainfall in summer. J. Climate, 32, 1327–1343, https://doi.org/10.1175/JCLI-D-17-0846.1.
Zhang, G., D. S. Nolan, C. D. Thorncroft, and H. Nguyen, 2008: Shallow meridional circulations in the tropical atmosphere. J. Climate, 21, 3453–3470, https://doi.org/10.1175/2007JCLI1870.1.
Zhang, H., Y. Zhao, A. Moise, H. Ye, R. Colman, G. Roff, and M. Zhao, 2018: On the influence of simulated SST warming on rainfall projections in the Indo-Pacific domain: An AGCM study. Climate Dyn., 50, 1373–1391, https://doi.org/10.1007/s00382-017-3690-1.
