What Drives the Intensification of Mesoscale Convective Systems over the West African Sahel under Climate Change?

HAL Id: hal-03002887

https://hal.archives-ouvertes.fr/hal-03002887

Submitted on 13 Nov 2020

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

What Drives the Intensification of Mesoscale Convective Systems over the West African Sahel under Climate

Change?

Rory Fitzpatrick, Douglas Parker, John Marsham, David Rowell, Francoise Guichard, Chris Taylor, Kerry Cook, Edward Vizy, Lawrence Jackson, Declan

Finney, et al.

To cite this version:

Rory Fitzpatrick, Douglas Parker, John Marsham, David Rowell, Francoise Guichard, et al.. What Drives the Intensification of Mesoscale Convective Systems over the West African Sahel under Cli- mate Change?. Journal of Climate, American Meteorological Society, 2020, 33 (8), pp.3151-3172.

�10.1175/JCLI-D-19-0380.1�. �hal-03002887�

1

1

What drives the intensification of mesoscale convective systems over the West African Sahel under Climate Change?

Rory G. J. Fitzpatrick

; Douglas J. Parker

; John H. Marsham

; David P. Rowell

; Francoise M. Guichard

; Chris M. Taylor

; Kerry H. Cook

; Edward K. Vizy

; Lawrence S. Jackson

; Declan Finney

; Julia Crook

; Rachel Stratton

; Simon Tucker

1. Institute for Climate and Atmospheric Sciences (ICAS), University of Leeds, Leeds, United Kingdom

2. Met Office, Exeter, United Kingdom

3. Centre National de Recherches Météorologiques (CNRM) - France 4. Centre for Ecology and Hydrology, Wallingford, United Kingdom 5. Jackson School of Geosciences, Austin, USA

Corresponding authors address: Rory Fitzpatrick, Jackson School of Geosciences, University of Texas at Austin, Austin TX, USA

E-mail: [email protected]

1

Abstract 2

Extreme rainfall is expected to increase under climate change carrying potential socio- 3

economic risks. However, the magnitude of increase is uncertain. Over recent decades, 4

extreme storms over the West African Sahel have increased in frequency with increased 5

vertical wind shear shown to be a cause. Drier mid-levels, stronger cold pools, and 6

increased storm organization have also been observed. Global models do not capture 7

the potential effects of lower-to-mid-tropospheric wind shear or cold pools on storm 8

organization since they parametrize convection. Here we use the first convection- 9

permitting simulations of African climate change to understand how changes in 10

thermodynamics and storm dynamics affect future extreme Sahelian rainfall. The 11

model, which simulates warming associated with Regional Climate Pathway 8.5 until 12

the end of the 21st Century, projects a 28% increase of the extreme rain rate of MCSs.

13

The Sahel moisture change on average follows Clausius-Clapeyron scaling, but has 14

regional heterogeneity. Rain rates scale with the product of time-of-storm total column 15

water (TCW) and in-storm vertical velocity. Additionally, pre-storm wind shear and 16

convective available potential energy both modulate in-storm vertical velocity. Although 17

wind shear affects cloud-top temperatures within our model, it has no direct correlation 18

with precipitation rates. In our model, projected future increase in TCW is the primary 19

explanation for increased rain rates. Finally, although colder cold pools are modelled in 20

the future climate, we see no significant change in near-surface winds, highlighting 21

avenues for future research on convection-permitting modelling of storm dynamics.

22

23

24

1. Introduction 25

The socio-economic risks to vulnerable people across West Africa due to climate 26

change are manifold (USAID 2017). Of particular interest are projections of the 27

frequency and intensity of extreme precipitation events. Mesoscale convective 28

systems (MCSs) are responsible for the majority of annual rainfall over the Sahel 29

(Laurent et al., 1998; Laing et al., 1999; Fink and Reiner 2003); understanding future 30

change of these events is pivotal to ascertaining future risk associated to rainfall.

31

The purpose of this study is to understand the processes by which climate 32

change can affect MCS precipitation rates, particularly the 99

percentile of surface 33

precipitation rates from MCSs (hereafter termed the extreme precipitation rate), over 34

the West African Sahel. We use a state-of-the-art regional climate model without any 35

active convection parameterization scheme to understand how variability on the 36

synoptic-to-climate timescale may affect MCS extreme precipitation rates.

37

The multi-decadal drought afflicting the Sahel in the 1970s and 80s has been 38

followed by a so-called “recovery period” (Nicholson 2005; Hagos and Cook 2008;

39

Lebel and Ali 2009; Lodoun et al., 2013; Evan et al., 2015; Sanogo et al., 2015) 40

during which average annual rainfall has returned to near-long-term mean levels 41

(Lélé and Lamb 2010; their Fig. 1). However, the nature of intra-seasonal rainfall 42

variability has changed over the Sahel during recent decades, with an observed 43

increase in the frequency of very rainy MCSs separated by longer periods of little to 44

no rainfall (Panthou et al., 2014, Taylor et al., 2017). More intense rain events carry 45

increased risk of flooding, with observational studies suggesting that flood events 46

over the Sahel have become more frequent over the past decades (Panthou et al., 47

2014; Nka et al., 2015, Panthou et al., 2018; Tazen et al., 2018; Wilcox et al., 2018).

48

Intense but temporarily sparse rain events can also deteriorate soil quality through 49

nutrient run off (Panagos et al., 2017) . 50

Taylor et al. (2017) observe a threefold increase in the frequency of early- 51

evening MCSs with mean cloud-top temperature < -75 °C over the Sahel since the 52

1980s. This increase is associated with an intensification of midday wind shear in the 53

vertical pressure column, henceforth wind shear, calculated as the difference 54

between low-level south-westerly/westerly winds and mid-level easterly winds, as 55

well as a drying of the Saharan air layer. Cold pools, and organized squall lines have 56

also been observed to intensify over the same timeframe, although the presence of 57

drier mid-level air and stronger cold pools may be coincidental and not dynamically 58

linked (c.f., James and Markowski 2010).

59

On the storm timescale, Taylor et al., (2017) found that ambient wind shear 60

intensity is significantly correlated with the mean cloud-top temperature of MCSs, 61

with subsequent observational work suggesting that ambient wind shear strength 62

may also influence the maximum precipitation rates of storms (Supplementary Fig.

63

S1). These findings are consistent with previous studies highlighting the role of wind 64

shear in organizing convection (Browning and Ludlam 1962; Moncrieff 1978, 1981;

65

Thorpe et al., 1982; Dudhia et al., 1987; Nicholls et al., 1988; Rotunno et al., 1988;

66

Mapes 1993; Houze Jr. 1993; Tao et al., 1995; Ferrier et al., 1996; LeMone et al.

67

1998; Mohr and Thorncroft 2006; Alfaro 2017).

68

Cloud-resolving model (CRM) studies have suggested that other ambient and 69

storm-relative variables are more important than wind shear in controlling MCS 70

precipitation rates. Takemi (2007; 2010; 2014) find that the vertical distribution of 71

high CAPE in the lower troposphere controls mean MCS precipitation rates to a

72

greater extent than lower/mid-tropospheric shear. However, Takemi (2010) also 73

found an inverse relationship between lower-tropospheric temperature lapse rate 74

and the maximum precipitation intensity of MCSs (their Fig. 14b), implying that 75

increased CAPE could be negatively correlated with maximum precipitation rates. A 76

similar result is found in Lucas et al., (2000), where enhanced wind shear positively 77

impacts total precipitation from MCSs, but negatively impacts maximum precipitation 78

rates. These results imply that storm-related drivers can be shown to positively or 79

negatively impact MCS intensity depending on the definition of storm intensity. It is 80

for this reason we primarily focus on extreme precipitation rates.

81

Finally, variability in moisture and ascent rates within MCSs have previously 82

been identified as influential in MCS precipitation rate variability. Increased inflow of 83

moist, buoyant air, and vertical ascent within convective updrafts increase maximum 84

precipitation rates (Alfaro 2017), as does enhanced lower-tropospheric relative 85

humidity (Lucas et al., 2000). Takemi (2014) hypothesized that increased 86

precipitable water available to a developing MCS will be positively correlated with 87

maximum precipitation rates. This theory is not explicitly tested by the author.

88

Rising temperatures under global warming have the potential to intensify 89

many of the MCS intensity drivers discussed above. Amplified warming of the 90

Sahara (Cook and Vizy 2015; Vizy and Cook 2017) will increase the meridional 91

temperature gradient across West Africa, leading to enhanced lower-to-mid- 92

tropospheric wind shear over the Sahel, in part through the enhancement of the 93

African Easterly Jet (Cook 1999). It is posited that a greater future meridional 94

temperature gradient over West Africa will lead to heavier precipitating MCSs under 95

climate change due to the strong water vapor feedback over the region (Dong and 96

Sutton, 2015; Evan et al., 2015). Through the Clausius-Clapeyron relationship, it is

expected that increased warming over the Sahel will lead to an increase in total 98

column water (TCW); this relationship has been identified previously within the 99

model used in this study (Kendon et al., 2019). Over the tropics, Takemi (2012a,b) 100

also find an increase in lower-tropospheric CAPE under global warming. In our 101

analysis, we examine the magnitude by which the above-identified MCS intensity 102

drivers increase in a potential future climate, and their subsequent dynamic effect on 103

MCS precipitation rates.

104

Coarse-resolution models with convection parameterization schemes, which 105

do not account for the role of wind shear, can struggle to model Sahelian storms and 106

their intensities accurately (Marsham et al., 2013). Convection-permitting models at 107

~4 km horizontal resolution have been shown to realistically simulate present day 108

monsoon flow, cold pool outflows (Marsham et al., 2013), convergence (Birch et al.

109

2014a,b), the diurnal cycle of rainfall (Vizy and Cook 2018a,b), and agricultural 110

decision-maker relevant monsoon metrics (Garcia-Carreras et al. 2015). Future 111

climate simulations of the West African climate at convection-permitting resolutions 112

therefore provide an opportunity to better understand changes in MCS dynamics 113

under global warming.

114

Section 2 describes the model used in this paper, with Section 3 explaining the 115

methods employed. Sections 4-7 provide results: first, we identify climatological 116

changes in the West African Monsoon between the current and future CP4-A climate 117

runs, then we analyze future changes in distributions of MCS precipitation intensity 118

relative to the current climate and the potential causes at different times of the day.

119

Section 8 gives conclusions and recommendations for future work.

120

2. Data

121

Within the Future Climate for Africa (FCFA) program, the Improving Model 122

Processes for African Climate (IMPALA) project has produced two 10-year model 123

simulations for the African continent at the convection-permitting horizontal grid 124

spacing of ~4.4 km. These simulations are referred to hereafter as “Convection 125

Permitting for Africa” (CP4-A; Stratton et al., 2018; Kendon et al., 2019). The CP4-A 126

simulations project the changes in African climate for the end of the 21

Century with 127

~4 K sea-surface temperature (SST) warming and ~5 K mean near-surface 128

temperature increase over the continent (Kendon et al., 2019). The model has 80 129

vertical levels with finer vertical resolution within the tropopause and boundary layer.

130

We provide a brief overview of the model and its quality below.

131

CP4-A is a regional model spanning longitudes 24°W – 56°E, and latitudes 45°S 132

– 39°N. The model is forced by lateral boundary conditions from an atmosphere-only 133

version the UK Met Office Global Climate Model (GCM) with ~25 km horizontal grid 134

spacing, and with prescribed daily SSTs. Land surface properties are initialized using 135

the most recent JULES land surface scheme (Walters et al., 2019), and allowed to 136

evolve freely over time. CP4-A uses an applied moisture conservation scheme 137

(Aranami 2015) and a three-dimensional blended boundary layer scheme (Boutle et 138

al., 2014). The large-scale cloud scheme is described in Smith (1990), and has been 139

used in other convection-permitting versions of the Met Office Unified Model. The 140

cloud scheme diagnoses liquid cloud fraction and condensed water when the grid- 141

box mean relative humidity exceeds a critical value. Ice water content is determined 142

by the microphysics scheme, with cloud fractions then diagnosed as in Abel (2017).

143

Other parameterization schemes are documented in Stratton et al. (2018, their Table 144

2). Both CP4-A, and the driving GCM, use the Even Newer Dynamics for General 145

Atmospheric Modelling of the Environment (ENDGame) dynamical core (Wood et al.

146

2014). The ENDGame dynamical core has been shown previously to improve the 147

representation of key processes over West Africa, particularly the seasonal 148

progression of monsoon rains (Vellinga et al. 2016; Fitzpatrick et al. 2016 Appendix 149

B).

150

In each climate respectively, the same SSTs are prescribed for the driving GCM, 151

CP4-A, and the parameterized model. Current climate SSTs are provided from the 152

Reynolds daily observational dataset (Reynolds et al., 2007). Future climate SSTs 153

are calculated first by quantifying the climatological change in SST values simulated 154

between 1975–2005 and 2085–2115 from a GCM run using the Coupled Model 155

Inter-comparison Project phase 5 Regional Climate Pathway 8.5 (RCP8.5). This 156

climatological SST change was calculated monthly, interpolated both spatially and 157

temporally, and added to the current climate values to produce future SSTs.

158

Greenhouse gases are taken from the RCP8.5 scenario for 2100; however the same 159

ozone and aerosol climatology is used in both simulations. Future changes in 160

precipitation rates across West Africa between the two simulations can be attributed 161

to rising temperatures and increases in greenhouse gases.

162

CP4-A has no convection scheme of any kind, using just model dynamics to 163

explicitly represent convective clouds. We note the caveat that a model at 4.4 km 164

grid spacing with no convective parameterization scheme may not capture the 165

complexity of convective development simulated in higher resolution CRM studies, 166

as model resolution has been highlighted in the past as a key issue in correctly 167

simulating cloud processes (e.g. Lin et al., 2012). With that said, the decision to have 168

no active convection scheme allows for a clearer comparison between convection 169

permitting and parameterized models produced during IMPALA.

170

Kendon et al. (2019) find that CP4-A better recreates observed high precipitation 171

rates over Africa compared to the parameterized-convection counterpart model and 172

projects a larger future increase in intense precipitation rates (their Fig. 1). Berthou 173

et al. (2019) find that CP4-A accurately simulates the frequency of MCSs over the 174

AMMA-CATCH observational sites. Stratton et al. (2018) and Berthou et al. (2019) 175

also show that the seasonal transition of monsoon rains across West Africa is better 176

captured in CP4-A compared to its parameterized convection counterpart, as well as 177

a reduction of the persistent dry bias within CP4-A. However, (Crook et al., 2019) 178

highlight that CP4-A fails to simulate MCSs of comparable area to the largest ones 179

observed, or speeds comparable to the quickest systems.

180

Simulations at the temporal and spatial scale of CP4-A incur an extensive 181

computational expense. As such, there has been a need for compromise with 182

regards to model design and output. There exists only one model representation for 183

the current and future climates respectively. The authors stress that the findings of 184

this paper are based on a research model, with a particular set of boundary 185

conditions from one global model, and so cannot be taken as projections with a 186

certainty level attached, but rather as results from climate sensitivity study.

187

The authors also stress that the two, ten-year simulations are not considered as 188

historical and future climatologies. Decadal and multi-decadal variability of ambient 189

conditions across West Africa can influence our results. We explore the role of this 190

variability further in Section 6.1.

191

3. Methods 192

In this paper, we analyze the climatic drivers of changes in early-evening MCS 193

extreme precipitation rates (the 99

percentile of MCS precipitation rate) over the

194

Sahel. The calculation methods for extreme precipitation rate and other considered 195

MCS-relevant metrics are presented in Table 1. We define the Sahel as 15°W–15°E, 196

10–18°N (displayed in Fig. 3); the suitability of this region for both climates is 197

explored in Section 4.1. The codes employed in this paper are available at 198

http://doi.org/10.5281/zenodo.2560371 and http://doi.org/10.5281/zenodo.2560410.

199

MCSs are classified as any contiguous structure with an area of cold-cloud 200

(Outgoing Longwave Radiation [OLR] <= 167 Wm

) of at least 25,000 km

simulated 201

during July-September. Extreme precipitation rates are therefore calculated across a 202

region of at least 250 km

(~ 16 grid cells). MCS tracking is performed from storm 203

genesis to decay using the algorithm presented in Stein et al. (2015) by identifying 204

contiguous regions of cold cloud that partially overlap between hourly intervals.

205

Tracked storms are analyzed at 1800 UTC in Sections 5 and 6 as this is the time of 206

maximum modelled precipitation (see Section 4). Different times of MCS occurrence 207

have been evaluated in Section 7.

208

Tracking MCSs using OLR instead of precipitation allows for continuous 209

identification of large-scale organized systems even when MCS precipitation may be 210

spatially or temporarily intermittent (Klein et al., 2018; their Fig. 4a). We have 211

analyzed precipitation-tracked events, finding little difference in key results (not 212

shown). The choice of a fixed temporal analysis window means that MCSs may be 213

analyzed at times when they are not at their maximum intensity; however, we believe 214

that this decision is fair given the availability of data as well as the complementary 215

analysis in Sections 4 and 7.

216

In this paper, we set two further restrictions on MCSs identification with the following 217

rationales:

218

 All MCSs have an extreme rain rate at 1800 UTC greater than 10 mm/hr. This 219

restriction excludes non-precipitating cold cloud structures. The threshold of 220

10 mm/hr is arbitrary. Different thresholds (between 5 – 30 mm/hr) were 221

analyzed, with little difference in results.

222

 All MCSs must have mean ambient (1200 UTC – 1600 UTC) precipitation 223

rate below 1 mm/hr across the region where each MCS is located at 1800 224

UTC. This restriction removes MCSs where the midday atmosphere is 225

perturbed by precipitation. Use of this restriction reduces the number of 226

MCSs analyzed (only 26% of current climate storms and 27% of future 227

climate storms pass this criterion). Evaluation of the vertical profile of 228

horizontal winds, and CAPE, in the environment preceding the arrival of an 229

MCS requires an atmosphere not disturbed by the potential effects of prior 230

storms. Complementary analysis for all MCSs regardless of this restriction 231

shows similar findings to those given here (not shown).

232

There are 1020 current climate MCSs and 553 future climate MCSs that meet 233

both criteria. The fact that CP4-A projects fewer Sahelian MCSs in the future climate 234

fits with prior analysis (c.f., Riede et al., 2016). Berthou et al., (revised), and Kendon 235

et al., (2019) additionally show that CP4-A simulates fewer Sahelian rain events in 236

the future regardless of storm size.

237

All metrics listed in Table 1 are calculated over the location where each MCS 238

is present at 1800 UTC. “Ambient” conditions (e.g., MU-CAPE) are evaluated at 239

1200 UTC consistent with Taylor et al. (2017), while “internal” storm characteristics 240

(such as minimum omega), are evaluated at 1800 UTC. Our choice of evaluation 241

times allow us to examine the influence of pre-storm environment and the internal

242

highlight that our measure of ambient wind shear does not use fixed pressure levels 244

to evaluate the low-level south-westerly and mid-level easterly ambient wind 245

circulations. Rather, our shear measure captures the maximum difference between 246

these two circulations across the lower to mid-atmosphere. We consider this 247

modification of importance given that changes in surface pressure and the potential 248

vertical displacement of the AEJ are not variables we wish to explicitly consider as 249

potential drivers of storm intensification changes.

250

In our analysis, we considered both zonal wind shear (i.e. maximum low-level 251

westerlies minus maximum mid-level easterlies) and wind shear magnitude (using 252

both the zonal and meridional horizontal winds) [Table 1]. We found little difference 253

in the influence of either metric on extreme precipitation rates. As such, we do not 254

show results for both metrics. Finally, due to an irreparable issue with CP4-A, 600 255

hPa model output is not available for the current climate storms in 1998 and some 256

storms in 1999.

257

258

4. Climatological differences in monsoon features over the Sahel.

259

For our analysis, a fixed geographical region has been selected for analysis 260

across climates. However, changes in extremes may arise from shifts of the summer 261

rain belt or other monsoon features. In this section, we consider whether the 262

latitudes 10–18°N offer a fair comparison of the West African Monsoon across 263

climates.

264

Figure 1 displays the zonally-averaged (across 15°W–15°E) latitudinal change 265

in measures of the monsoon state across climates. Mean daily precipitation 266

(indicative of the location of the monsoon rain belt) shows similar rainfall patterns

267

across climates with the maximum change in daily precipitation rate of 1 mm/day 268

(Fig. 1a). There is also evidence of a northwards shift of the northernmost extent of 269

monsoon rains. Proportionally, precipitation increases are greater farther north (~40 - 270

50% near 18°N). The ~1° northwards shift of monsoon rainfall limit across climates, 271

as well as the higher future precipitation rates over the Sahel, and a projected 272

reduction of precipitation rates near the Guinea Coast (5–7°N) is consistent with 273

previous studies (e.g., Vizy et al., 2013; Cook and Vizy 2015; Vizy and Cook 2017).

274

Mean midday zonal wind shear (Fig. 1b) increases predominantly due to 275

stronger low-level south-westerly flow across all latitudes of around 1-2 m/s or up to 276

40% (Figs. 1d-e) and a relatively smaller intensification of AEJ-level easterlies 277

across the northern Sahel of no more than 0.5 m/s (<5% across climates Fig. 1c).

278

There is a projected latitudinal broadening of the AEJ in the future, most pronounced 279

north of 13°N (Fig. 1c). However, the latitudinal position of peak zonal wind shear 280

and 650 hPa winds is the same in the current and future climate simulations.

281

Enhanced low level westerlies, indicative of the West African Westerly Jet diurnal 282

peaks (WAWJ, Pu and Cook 2010; 2012), and southerlies imply an enhanced 283

horizontal transport of moisture over the Sahel in the future.

284

Mean 850 hPa temperature (Fig. 1f) indicate a future increase in the 285

meridional temperature gradient across West Africa, with temperatures increasing 286

over the Sahara (6–7 K) more than over the Guinea Coast (5–6 K). This increased 287

meridional temperature gradient is consistent with the enhanced zonal wind shear 288

across the Sahel via their co-dependence due to the thermal wind balance (Fig. 1b;

289

Cook 1999; Parker et al. 2005; Cook and Vizy 2015). Specific humidity (Fig. 1g) is 290

much greater across the entire West African region in the future climate, with the 291

difference between the two climates maximized at ~12°N. We see little (< 3%)

change in lower-tropospheric relative humidity between climates across the Sahel 293

(Fig. 1h). Increased low-level southerly winds across the Guinea Coast, coupled with 294

higher specific humidity is associated with an increase in TCW over the Sahel (Fig.

295

1i). The location of maximum TCW does not change across climates (9–10°N), but 296

the greatest absolute increase occurs within the Sahel (~12–13°N).

297

The latitudinal pattern of each variable considered in Fig. 1 is consistent 298

across climates. We conclude that the region 10–18°N is a fair region over which to 299

consider MCSs in both climates.

300

Figure 2 shows the diurnal cycle of CP4-A precipitation (Fig. 2a) and lower-to- 301

mid-tropospheric wind shear magnitude (Fig. 2b), averaged over our analysis region.

302

In both climates, the maximum precipitation rate occurs during the late afternoon and 303

is minimized around 11 UTC. Precipitation increases at all times of day with climate 304

change, but percentage wise, the greatest increases occur prior to 1200 UTC. CP4- 305

A has a bias towards too much midday precipitation across the Sahel, however the 306

model is improved over parameterized counterpart models (Berthou et al., 2019).

307

Hourly rainfall rates increase by less than 20% in the future climate for all times after 308

1400 UTC. Kendon et al. (2019) show fewer short-lived evening storms in the future 309

climate within CP4-A, potentially explaining the relatively small absolute/percentage 310

increase in precipitation across the afternoon and evening period.

311

Observed wind shear in the lower-to-mid-troposphere is maximized overnight, 312

since this is the time of strongest south-westerly monsoon flow (Parker et al., 2005).

313

The diurnal cycle of zonal wind shear is consistent across climates within our 314

observation region (Fig. 2b). We see an increase throughout the day in the 315

percentage change of zonal wind shear across climates with a peak increase

316

occurring at 2100 UTC (~ 40%). The larger percentage increase at 1500 - 2100 UTC 317

is due in part to these times corresponding to the diurnal minimum in shear values.

318

Absolute increases in zonal wind shear range between about 3.5 m/s (at 1500 UTC) 319

and 4.8 m/s (at 2100 UTC) [Supplementary Figure S2]. Spatial changes in zonal 320

wind shear calculated between set levels (925 and 600 hPa) are provided in 321

Supplementary Fig. S3 and highlight that zonal wind shear increases more, in 322

absolute terms, within the eastern Sahel.

323

We decide that using 1800 UTC as our storm reference time is fair as 1800 324

UTC is shortly after the time of maximum area-averaged precipitation in both 325

climates and aligns with available instantaneous data for wind speed and 326

temperature. However, there exist regions where overnight rainfall provides the 327

majority of local precipitation over the Sahel (Mathon et al. 2002; Vizy and Cook 328

2018a). Drivers of MCS extreme precipitation rates at 0000 UTC and 0600 UTC 329

storms are discussed in Section 7.

330

We next consider changes in TCW and their relationship to near-surface 331

temperature changes across climates and compared them to the expected rate of 332

change in TCW per unit warming from the Clausius-Clapeyron relationship (i.e. 7%

333

per K warming). We use the differential rate of TCW change per Kelvin, r, expressed 334

as a percentage change, from O’Gorman and Muller (2010; their equation 2, our Fig.

335

3a):

336

r=100 x

∆T

(1)

337

Where ΔT is the mean 1.5 meter temperature difference across climates, and r

is 338

the fractional rate of TCW change per Kelvin, calculated as:

339

𝑟

=

𝑐𝑢𝑟𝑟𝑒𝑛𝑡

(2)

340

Figure 3 also displays spatial maps of projected TCW, 1.5-meter temperature, 341

and near surface circulation changes across climates (Figs. 3b and 3c) and r as 342

defined in equation 1 (Fig. 3d).

343

Figure 3a shows that climatological 1.5 meter temperatures increase by 4-5 K 344

near the Guinea Coast (0-5°N), and by over 6 K across the Sahara (north of 20°N).

345

CP4-A simulates zonally-averaged r > 7% K

from 10-20°N implying super-Clausius- 346

Clapeyron scaling of TCW occurs across our analysis region (although the scaling is 347

zonally heterogeneous as seen in Fig. 3d).

348

Consistent with Vizy et al. (2013), we see projected enhancement of the 349

WAWJ across Senegal and southern Mali, and increased south-westerly monsoon 350

flow across Benin, Burkina Faso and Nigeria. Both these circulation changes would 351

be expected to increase low-level moisture across the eastern Sahel. Accordingly, 352

there is an apparent south-westerly gradient in the percentage TCW increase 353

between the current and future climate (Fig. 3b). Across much of the eastern Sahel, 354

there is >50% more moisture in the future climate. Towards Senegal, the increase is 355

30-40%. Changes in surface temperature across climates appear relatively more 356

zonally homogeneous across West Africa (Fig. 3c).

357

CP4-A projects super-Clausius-Clapeyron increases in r projected across the 358

eastern Sahel, as well as Guinea, Liberia and Sierra Leone (Fig. 3d). This high 359

increase in TCW per K warming is of particular interest over the Niger/Nigeria 360

border, as this is a location where many observed MCSs originate during the boreal 361

summer (Vizy and Cook 2018a, b) and occur in the current climate CP4-A 362

simulation. Sub-Clausius-Clapeyron r values over Senegal and Mauritania are due to

363

strengthened northerlies from the increased low-level cyclonic circulation about the 364

deepened trough providing warming and drying in over this region. The zonal 365

contrast in simulated r presented in Fig. 3d is consistent with projected CP4-A 366

increases in extreme precipitation across the Sahel (Kendon et al., 2019), currently 367

observed spatial heterogeneity in precipitation trends (Panthou et al., 2018), and 368

spatial patterns of seasonal precipitation change from other future climate studies 369

using coarser-resolution models (e.g., James et al., 2014; their Fig. 1).

370

In conclusion, we find an intensification of climatological precipitation across 371

much of our study region present at all times of day. There is an increase in zonal 372

and horizontal wind shear in the future climate, primarily associated with a 373

strengthening of the low-level monsoon flow and the WAWJ, with a relatively smaller 374

intensification of the AEJ apparent. Zonally-averaged low-level specific humidity, and 375

TCW increase in the future. There is a spatial heterogeneity to the increase in 376

available moisture for MCSs relative to localized heating, which is spatially 377

consistent with prior studies highlighting zonal contrasts in current and future 378

precipitation trends over the Sahel.

379

380

5. Changes in storm intensity and drivers within and across climates 381

5.1 Vertical profile of storm environment: Dependence on intensity and climate 382

We next investigate whether selection of fixed pressure levels for comparison 383

of storm drivers across climates is valid, or whether considerations for vertical 384

displacement of storm intensity drivers must be taken into account. Figure 4 displays 385

the vertical profile of mean midday zonal wind speed (Fig. 4a), 1800 UTC minimum

386

omega (Fig. 4b), and relative humidity at 1200 UTC (Fig. 4c) and 1800 UTC (Fig. 4d) 387

for storms with the highest and lowest extreme precipitation rates in each climate.

388

At 925 hPa, we see stronger composite low-level westerlies preceding storms 389

with the lowest 25% of extreme precipitation rates in the current climate simulations 390

(Fig. 4a, ~ 1m/s difference). Stronger composite ambient westerly winds are 391

simulated at this level in the future climate compared to the current climate, but we 392

do not see a difference between ambient westerlies preceding MCSs with the 393

highest and lowest extreme precipitation rates.

394

More intense easterlies at 650 hPa precede MCSs with higher extreme 395

precipitation rates in the current climate with easterlies approximately 1 m/s more 396

intense. At higher levels, current climate ambient easterlies are weaker preceding 397

MCSs with the highest extreme precipitation rates, We also note that composite 650 398

hPa easterlies preceding both subsets of current climate MCSs are actually weaker 399

than those preceding the all-storm composite at 650 hPa (not shown). In the future 400

climate, we see enhanced mid- and upper tropospheric easterly flow preceding both 401

subsets of storms, with the strongest composite easterlies from 650 hPa upwards 402

simulated prior to MCSs with the lowest extreme precipitation rates. In both climates, 403

Fig. 4a implies that enhanced ambient zonal-wind shear may not be present prior to 404

systems with higher extreme precipitation rates. As an aside, when storm intensity is 405

measured using mean OLR, we do see a clear pattern towards more intense mid- 406

level easterly winds preceding colder-topped MCSs (Supplementary Fig. S4a). For 407

the remainder of our analysis, set pressure ranges across which to evaluate lower- 408

tropospheric westerlies (925 - 800 hPa), and mid-tropospheric easterlies (700 - 500 409

hPa) for both climates are considered fair.

410

In both climates, MCSs with higher extreme precipitation rates have higher 411

composite minimum omega at the time of storm throughout the mid- and upper 412

troposphere compared to those with lower extreme precipitation rates (Fig. 4b). For 413

storms with the highest extreme precipitation rates in either climate, composite 414

omega values are similar between climates below 600 hPa with stronger omega 415

simulated higher than 600 hPa in the future implying greater in-storm ascent rates.

416

The opposite is true for events with the lowest extreme precipitation rates, as current 417

climate storms are simulated to have more intense in-storm omega. Figure 4b 418

implies relatively more extremes (both strong and weak) of in-storm ascent rates 419

simulated for future climate MCSs. Through the rest of this article, we use minimum 420

omega above 600 hPa for each MCS as the indicator of maximum speed of 421

ascending air in MCS convective cores (whilst appreciating that the grid-resolution of 422

our model reduces the precision of this measurement).

423

We see a clear separation between composite RH profiles for MCSs with the 424

largest extreme precipitation rates and lowest extreme precipitation rates for a given 425

climate (Fig. 4c for 1200 UTC, Fig. 4d for 1800 UTC). In either climate, higher 1200 426

UTC relative humidity values are simulated below about 550 hPa preceding storms 427

with higher extreme precipitation rates (Fig. 4c), with higher 1800 UTC RH values 428

simulated throughout the vertical column (Fig. 4d). Across climates, we see little 429

change in 1200 UTC composite low-level RH for either MCS subset, but greater mid- 430

to upper tropospheric humidity preceding MCSs with the highest precipitation rates in 431

the future climate. At the time of storm, there is a consistent simulated drying of the 432

lower- to mid-troposphere simulated for future climate storms (~800 – 600 hPa) for 433

both respective MCS populations. For the rest of this paper, we evaluate 1200 UTC 434

RH in order to focus on the influence ambient conditions have on extreme

435

precipitation rates, and select 700 hPa as a representative pressure level for 436

analysis of mid-tropospheric RH.

437

Figure 5 shows the skew-t plot of composite profiles preceding current climate 438

(blue lines) and future climate (red lines) storms; Table 2 provides relevant statistics 439

from these soundings. Future storms are preceded by higher equilibrium and lifting 440

condensation levels, and greater ambient bulk CAPE and bulk CIN, which are 441

expected to favor more intense deep convection. The higher CIN environment seen 442

in the future climate suggests that stronger triggers (such as cold pools) are required 443

to initiate deep convection in this environment, consistent with the greater 444

preponderance of nocturnal propagating systems, as compared with scattered 445

evening cumulonimbus.

446

447

5.2 Character of extreme storms and pre-storm environment in current and 448

future climate 449

Figure 6 compares populations of current climate and future climate 450

ambient/time-of-storm metrics. Changes between climates are statistically significant 451

beyond the 90-99

percentile for all metrics apart from 1800 UTC minimum omega 452

(Fig. 6h). Figure 6a shows a 28% increase of mean extreme precipitation rates under 453

climate change. Future climate storms are on average deeper, with higher rain rates, 454

but have warmer cloud tops (Fig. 6g) due to a warmer troposphere (Fig. 5).

455

We find intensifications in mean 1200 UTC wind shear (Figs. 6b), 1800 UTC 456

mean TCW (Fig. 6c), and ambient bulk CAPE of the most unstable air parcel (Fig.

457

6e) as well as drier ambient mid-level air (Fig. 6f). Relative to the large increase in 458

mean 1800 UTC TCW across climates (41%), we see a smaller, but significant

459

increase in the 1800 UTC – 1200 UTC TCW anomaly across climates (21%, Fig.

460

6d). 1-hour total precipitation accumulation increases in the future climate (Fig. 6j), 461

as well as the temperature deficit of cold pool outflows (Fig. 6l). The future change in 462

1-hour precipitation accumulation is not linked to larger storms in the future; in fact 463

there are fewer large storms in the future climate (Fig. 6k). Figure 6 highlights a 464

significant shift in the CP4-A future climate scenario towards conditions conducive to 465

more intense rain events.

466

467

6. Storm-level dynamics and intensity drivers in the current and future CP4-A 468

simulations 469

6.1 What controls extreme precipitation rates in CP4-A?

470

Figure 7 shows the relationships between different ambient and internal 471

variables and modelled extreme precipitation rates for both climates. There is no 472

statistically significant (at the 95% confidence interval) correlation between future 473

climate ambient wind shear and extreme precipitation rates, with a weak yet 474

significant negative correlation between the same metrics seen for the present 475

climate (Fig. 7a). The relationship between the vertical profile of lower-to-mid- 476

tropospheric ambient winds and extreme precipitation rates is complex, with 477

observational and modelling studies highlighting different potential relationships (c.f., 478

Takemi 2014 and Taylor et al., 2017). The correlations shown in Fig. 7a agree with 479

prior studies performed using CRMs (e.g. Lucas et al., 2000; Takemi 2010; 2014), 480

but disagree with Supplementary Fig. S1.

481

Time of storm TCW and vertical velocity exhibit the strongest control on 482

extreme precipitation in either climate (Figs 7b, 7d respectively), with the product of

483

these two variables, an approximation for water uplift in the most intense convective 484

core, also significantly scaling with extreme precipitation rates (not shown).

485

Additionally, Taylor et al., (2017) hypothesized that more intense systems supply 486

increased moisture. Figure 7c agrees with this notion with good agreement between 487

the correlation of extreme precipitation rate and TCW anomaly at 1800 UTC across 488

climates. However, this model result could be coincidental, with intense storms 489

forming in large-scale environments that generate increases of TCW between 12 490

and 18 UTC for unrelated reasons.

491

Colder-topped MCSs have higher extreme precipitation rates in both climates 492

consistent with expectations (Fig. 7e). Lower MU-CAPE (Fig. 7f) and more humid 493

ambient mid-levels (Fig. 7g) are significantly correlated with higher extreme rain 494

rates. We stress that MU-CAPE as defined in Table 1 is not directly comparable to 495

the depth of high CAPE air within the lower-troposphere analyzed in Takemi (2010;

496

2014). It is important also to note that the processes through which increased MU- 497

CAPE can affect precipitation may be crudely modelled at 4.4 km horizontal 498

resolution, given the potential influence of cloud-mixing and microphysical 499

processes. Other measures of CAPE, including for surface based parcels and the 500

mean parcel CAPE in the lowest 100 hPa of the atmosphere, have also been 501

evaluated, with little difference in their relationship with precipitation rates.

502

Storms with higher extreme precipitation rates in either climate output greater 503

quantities of rainfall during 1 hour across the entire MCS area (Fig. 7h), implying 504

these systems potentially have broad impacts over the regions they are present.

505

Finally, there is no significant correlation between cold pool intensity and extreme 506

precipitation rate in either climate (Fig. 7i). Although intense storms can be 507

associated with strong cold pools, ice hydrometeor melting, sublimation, and

508

evaporation help generate cold pools, which can correspond to minimal precipitation 509

at the surface.

510

As discussed in Section 2, our two 10-year simulations are each affected by 511

the models own internal natural variability. We note however that the natural 512

variability in each run is constrained by the lower and lateral boundary conditions. In 513

order to quantify the potential influence of long-timescale variability on ambient TCW, 514

we have compared the change in June-September TCW from a historical period 515

(1950-1999) to the end of the 21

Century (2070-2099) within a 4-member GCM 516

ensemble using a single model (HadGEM2-ES) forced with different initial conditions 517

following RCP8.5. HadGEM2-ES is employed as it provides the best comparison to 518

CP4-A given similarities in model design.

519

Historic climatological TCW values over the Sahel range from 30.3 – 30.6 kg 520

m

across ensemble members; future values range from 46.3 – 47.1 kg m

. The 521

difference in TCW across climates (~ 16 kg m

) is comparable to the difference in 522

TCW simulated in CP4-A (Fig. 7b). Inter-ensemble variability in each climate is much 523

lower than the change projected across climates, implying that the TCW differences 524

projected in CP4-A are likely associated with climate change and not a product of 525

random variability.

526

6.2 Modelled interactions of storm intensity drivers in CP4-A 527

Figure 8 shows the relationships between selected ambient and time of storm 528

metrics. Figure 8a shows a significant negative correlation between midday wind 529

shear and mean time-of-storm OLR in both climates. The correlation coefficients are 530

of similar magnitude to that found in observations (Taylor et al., 2017; their 531

coefficient: -0.347), suggesting that CP4-A captures the observed influence of

532

ambient wind shear on MCS organization. The role of shear in determining the 533

structure and organization of MCSs has also been identified in observational studies 534

and studies using CRMs (e.g., Newton 1960; Browning and Ludlam 1962; Rotunno 535

et al., 1988; Houze Jr 2004; Takemi 2007). We additionally note that, for a given 536

temperature and humidity profile, there may exist a vertical wind profile which 537

optimizes MCS organization (Takemi 2007). Across climates, the horizontal winds, 538

humidity and temperature significantly change in CP4-A (Fig. 1) implying that the 539

future climate ambient environment may be more conducive to MCS organization 540

once a storm is generated.

541

Within the range of simulated wind shear values, we see a weak, but 542

significant correlation between ambient shear and in-storm minimum omega (Fig. 8b) 543

which agrees with prior studies (e.g., Fovell and Ogura, 1989; Lucas et al., 2000). In 544

each climate, ambient wind shear is also correlated with 1200 UTC MU-CAPE (Fig.

545

8c). This link may be co-fluctuation not causation, may be due to intense shear in the 546

lower troposphere inhibiting convective initiation, or may be due to the presence of 547

enhanced low-level convergence in high shear environments. Ambient wind shear is 548

either significantly negatively correlated (in the current climate: -0.227 correlation co- 549

efficient), or not correlated (in the future climate: 0.003 correlation coefficient) with 550

time-of-storm TCW (Fig. 8d).

551

Both climates simulate stronger wind shear preceding MCSs with more 552

intense cold pools, however this relationship is only statistically significant in the 553

current climate (Fig. 8e). Past studies have highlighted the importance of wind shear 554

in controlling cold pool strength (Thorpe et al., 1982; Rotunno et al., 1998), but this 555

relationship is complex (Parker 1996). Prior research has also highlighted ambiguity 556

with the role of other ambient conditions such as dry mid-levels in intensifying

557

convective cold pools (James and Markowski 2010). Detailed analysis of cold pool 558

simulations for CP4-A storms is beyond the scope of this paper. However, our 559

analysis implies that the factors required to sustain a strong cold pool are likely to be 560

prohibitive to higher extreme precipitation rates in CP4-A.

561

Figure 8f implies that systems preceded by higher ambient TCW have 562

statistically significant lower 12 to 18 UTC TCW increases in both climates. As in Fig.

563

7c, the correlation co-efficient between these two variables is consistent across 564

climates suggesting that the fundamental behavior of MCSs with regards to available 565

moisture does not change under global warming. In both climates we also note that 566

higher time-of-storm TCW is significantly correlated with increased mid-tropospheric 567

RH with similar correlation coefficients (~0.65) found for each climate (Fig. 8g). The 568

maximum vertical velocity of MCSs is significantly correlated with mean OLR in both 569

climates (Fig. 8h). Finally, the maximum buoyancy of rising air parcels simulated for 570

each MCS is closely tied to the uplift velocity (Fig. 8i) implying a realistic relationship 571

between buoyancy and upwards motion is simulated within CP4-A.

572

Figures 7 and 8 suggest that current- and future-climate storms simulated 573

within CP4-A behave similarly. Ambient wind difference exhibits control on the 574

organization of developing systems, with higher wind difference associated with 575

colder MCS cloud-tops, potentially through the ability of lower-to-mid-tropospheric 576

wind shear to control the verticality of ascending air within a system and the build-up 577

of CAPE. Higher TCW and ascent rates at the time of storm are closely associated 578

with greater extreme precipitation rates, with the product of these two variables 579

showing good correlation with extreme precipitation rates. Although it may be 580

expected that time-of-storm TCW correlates with precipitation metrics (as 581

precipitation can only fall if there is precipitable water available, and precipitation

actually contributes a small fraction of the TCW), it is of interest that TCW is the 583

strongest control of the selected potential drivers of extreme rain rates within CP4-A 584

both in each climate, and across climates.

585

Stronger ambient wind shear is associated with stronger cold pools in the 586

current climate (Fig. 8e), thus encourages gregarious development of MCSs, and 587

stronger ascent rates in both climates, implying faster uplift of moisture. However, 588

wind shear is also associated with lower TCW and higher MU-CAPE, and therefore 589

may also negatively impact extreme precipitation rate. Our results imply that there is 590

no simplistic direct relationship between ambient wind shear and extreme 591

precipitation rates in CP4-A for either climate.

592

Under global warming, the atmosphere becomes more conducive to higher 593

extreme precipitation rates for early evening MCSs, however less favorable overall to 594

MCS genesis (Berthou et al., revised). With regards to changes in extreme 595

precipitation rates, the key climatic difference is the 41% increase in time-of-storm 596

TCW, which is attributable to both an increase in ambient (i.e. 12 UTC) TCW, and 597

greater 12-to-18 UTC TCW. This increase in the afternoon increase in TCW may be 598

generated by developing MCSs, may simply reflect the greater availability of TCW in 599

the future climate which has a diurnal variation, or may reflect a change in that 600

diurnal cycle (we do not investigate further here). From Fig. 3, we note that the 601

eastern Sahel is projected to have super-Clausius-Clapeyron scaling of TCW with 602

temperature; accordingly this region sees the largest frequency of future climate 603

MCSs (not shown).

604

Figures 1, and 3 highlighted zonal and meridional heterogeneity in the 605

projected change of several storm intensity drivers, notably TCW and CAPE. The

606

analysis presented in Figures 7 and 8 has been reproduced for the western Sahel 607

(15 – 0°W – Supplementary Figs. S5 – S6), and the eastern Sahel (0 – 15°E – 608

Supplementary Figs. S7 – S8), as well as over more constrained latitudinal bounds 609

(11 – 12°N – Supplementary Figs. S9 – S10, and 16 – 17°N – Supplementary Figs.

610

S11 – S12). We draw similar conclusions to those presented above for MCSs within 611

either the west or east Sahel as well as those present across 11 – 12N. For MCSs 612

found over 16 – 17N, we find that their behavior is very similar to the all-storm 613

analysis (compare Figs. 8 and Supplementary Fig. S12), however, ambient CAPE 614

and mid-level RH do not appear to be significant drivers of extreme precipitation rate 615

variance in either climate for these events (Supplementary Fig. S11). Although the 616

interaction of storm processes remains similar throughout the Sahel region studied, 617

latitudinal variations in humidity and available energy can affect the relationship 618

between these metrics and extreme precipitation rates.

619

620

7. Change in correlation of storm drivers for different intensity measures, 621

and at 0000 and 0600 UTC 622

In Section 6, we consider drivers of extreme precipitation rates of early evening 623

MCSs across the Sahel. Here, we extend our analysis to consider drivers of mean 624

precipitation rate change across climates, as well as drivers of extreme precipitation 625

rate changes for nocturnal MCSs. This analysis is necessary, as prior work has 626

suggested that mean MCS precipitation rates are controlled by ambient/internal 627

drivers different from those which affect extreme precipitation rates (Takemi 2010), 628

and different storm intensity drivers are present at different times of day (Vizy and 629

Cook 2018a).

630

Figure 9 displays analogous information to Fig. 7, except for mean precipitation 631

rates. We note the caveat that Fig. 9 does not account for changes in MCS area 632

across climates.

633

The drivers of extreme precipitation rate, and mean precipitation rate variability 634

within the current climate and future climate simulations are broadly similar. For both 635

measures of early-evening precipitation, increased ambient wind shear does not 636

have a significantly positive effect [panel (a) of both figures]. Increased time-of-storm 637

TCW and 12-to-18 UTC TCW increases show the strongest correlation with mean 638

precipitation rates across both the current and future climates (Figs. 9b-c). MCSs 639

with higher mean precipitation rates have significantly stronger minimum omega (Fig.

640

9d) and significantly colder mean OLR (Fig. 9e), consistent with Fig. 7. Figure 9f 641

implies that increased CAPE is significantly negatively correlated with mean 642

precipitation rates within both climates. This result is seemingly at odds with Takemi 643

(2014) who highlight the importance of increased depth of high CAPE in the lower- 644

troposphere. However, the effect of latitudinal variations in CAPE cannot be 645

discounted as a cause for this difference (as seen in Section 6), nor can differences 646

in methodology for evaluating CAPE. A more thorough investigation of the spatial 647

and vertical distribution of CAPE in the lower troposphere simulated in CP4-A prior to 648

MCSs of different intensities is beyond the scope of this study, but will form the basis 649

of future work.

650

Drivers of extreme precipitation rates of MCSs at 0000 UTC (Fig. 10) are similar 651

to those found for 1800 UTC extreme precipitation rates. Note that for 0000 UTC 652

MCSs, we consider ambient conditions at 1800 UTC, and apply criteria akin to those 653

for 1800 UTC MCSs with adjusted time windows. The quantity of ambient TCW (Fig.

654

10b), the 18-to-00 UTC increase in TCW (Fig. 10c), and the minimum omega within

655

each MCS (Fig. 10d) are the predominant controls on extreme precipitation rate 656

variability at this time.

657

At 0600 UTC, there is a large decline in the number of future climate MCSs 658

simulated that meet our pre-set areal threshold (approximately 400 MCSs) and both 659

our pre-set precipitation criteria (29 MCSs). Figure 11 therefore shows all MCSs of at 660

least 25,000 km

that are preceded by ambient mean precipitation not exceeding 1 661

cm/hour rather than 1 mm/hour. Using this relaxed restriction allows for 354 storms 662

to be evaluated for the future climate, and 1307 to be evaluated for the current 663

climate.

664

Although different MCS triggering mechanisms have been highlighted for early 665

evening and nocturnal systems in observations (Vizy and Cook 2018a), the degree 666

to which our a priori drivers affect extreme precipitation rates are reasonably 667

consistent between 0600 UTC (Fig. 11) and 1800 UTC (Fig. 7). The one exception to 668

this finding is the weaker link between ambient CAPE and 0600 UTC extreme 669

precipitation rates (Fig. 11f) compared to earlier MCS analysis times. Nevertheless, 670

we conclude from Figs. 7, 10, and 11 that CP4-A highlights the particular importance 671

of ambient TCW, time-of-storm omega and RH for MCS extreme precipitation rate 672

variability across the early evening and night-time.

673

674

8. Conclusions and Discussion 675

Using output from the first convection-permitting model of pan-African climate 676

changes we have evaluated the processes through which global warming within a 677

realistic future climate scenario can impact extreme precipitation rates of Sahelian 678

MCSs, as well as the modelled drivers of change on the storm timescale. Our work is

679

considered a climate sensitivity experiment, and the quality of projected changes is 680

not assessed here. We further do not provide an evaluation of modelled precipitation 681

against present observations as this has been assessed elsewhere (Stratton et al., 682

2018; Berthou et al., 2019). Results presented primarily focus on early evening 683

storms (1800 UTC) across the whole Sahel (15°W – 15°E, 10 – 18°N). Overnight 684

precipitation systems as well as potential longitudinal and latitudinal variation in 685

findings have been presented, with little difference in key results.

686

We find a 28% increase in the mean extreme precipitation rate of early 687

evening MCSs by the end of the 21

Century, and an increase in precipitation rates 688

at all times of day despite fewer MCSs in the future climate. We primarily link this 689

increase in extreme precipitation rate to increases in TCW, which scales at close to 690

Clausius-Clapeyron scaling (although the increases have zonal heterogeneity). The 691

greatest TCW increases are simulated across Nigeria, Niger and Lake Chad, which 692

corresponds to the location of most early-evening storms with particularly high 693

extreme rain rates in the future climate. We also note that there are fewer MCSs 694

simulated in the future climate during the early evening and overnight based on our 695

areal threshold. Prior analysis of CP4-A output has also shown that, despite a 696

projected increase in precipitation rates at all times of day, future rainfall over the 697

Sahel is provided by fewer, and on average smaller systems compared to current 698

climate precipitation (Kendon et al., 2019; Berthou et al., revised).

699

Figure 12 provides a more complete overview of modelled processes by 700

which global warming affect user-relevant storm metrics across the Sahel within 701

CP4-A. Figure 12 does not consider potential climate change effects due to changes 702

in aerosols or microphysics, and as our focus is uniquely on how global and regional

703