Generated using version 3.1 of the official AMS LATEX template

Impact of Antarctic ozone depletion and recovery on Southern1

Hemisphere precipitation, evaporation and extreme changes2

Ariaan Purich and Seok-Woo Son ∗

Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada

3

∗Corresponding author address: Seok-Woo Son, Department of Atmospheric and Oceanic Sciences, McGill

University, 805 Sherbrooke West, Montreal, Quebec, H3A 2K6, Canada.

E-mail: seok-woo.son@mcgill.ca

1

ABSTRACT4

The possible impact of Antarctic ozone depletion and recovery on Southern Hemisphere5

(SH) mean and extreme precipitation and evaporation is examined using multimodel output6

from the Climate Model Intercomparison Project 3 (CMIP3). By grouping models into four7

sets, those with and without ozone depletion in 20th century climate simulations and those8

with and without ozone recovery in 21st century climate simulations, and comparing their9

multimodel-mean trends, it is shown that Antarctic ozone forcings significantly modulate10

extratropical precipitation changes in austral summer. The impact on evaporation trends is11

however, minimal especially in 20th century climate simulations. In general, ozone depletion12

has increased precipitation in high-latitudes and decreased it in mid-latitudes, in agreement13

with the poleward displacement of the westerly jet and associated storm tracks by Antarc-14

tic ozone depletion. Although weaker, the opposite is also true for ozone recovery. These15

precipitation changes are primarily associated with changes in light precipitation (1–10 mm16

day−1). Contributions by very-light precipitation (0.1–1 mm day−1) and moderate-to-heavy17

precipitation (>10 mm day−1) are minor. Likewise, no systematic changes are found in18

extreme precipitation events, although extreme surface wind events are highly sensitive to19

ozone forcings. This result indicates that, while extratropical mean precipitation trends are20

significantly modulated by ozone-induced large-scale circulation changes, extreme precipita-21

tion changes are likely more sensitive to thermodynamic processes near the surface than to22

dynamical processes in the free atmosphere.23

1

1. Introduction24

Southern Hemisphere (SH) climate changes over the last few decades have been exten-25

sively documented in recent studies. They include an expansion of the Hadley cell (Seidel26

et al. 2008; Johanson and Fu 2009), a shift in atmospheric mass from high- to mid-latitudes27

(Thompson and Solomon 2002; Marshall 2003), a poleward displacement of the westerly jet28

and storm tracks (Thompson and Solomon 2002; Marshall 2003; Fyfe 2003), an increase in29

surface wind speeds over the Southern Ocean (Boning et al. 2008), anomalously dry condi-30

tions over southern South America, New Zealand and southern Australia, and anomalously31

wet conditions over much of Australia and South Africa (Gillett et al. 2006). A freshening32

and warming of the Southern Ocean (Wong et al. 1999; Gille 2002; Boning et al. 2008) and33

a significant warming of the Antarctic peninsula (Thompson and Solomon 2002) have also34

been observed. While some of these changes have been attributed to circulation changes35

induced by the increase in anthropogenic greenhouse gases (Fyfe et al. 1999; Kushner et al.36

2001; Cai et al. 2003), such changes in austral summer have also been influenced by Antarc-37

tic ozone depletion (Thompson and Solomon 2002; Shindell and Schmidt 2004; Arblaster38

and Meehl 2006; Perlwitz et al. 2008; Son et al. 2009, 2010; McLandress et al. 2011; Polvani39

et al. 2011; Kang et al. 2011). It is known that both increasing greenhouse gases, occurring40

year-round, and ozone depletion, which occurs most significantly in late spring and sum-41

mer, have driven SH extratropical circulation changes in a similar way, the cumulative effect42

resulting in more significant tropospheric climate change in austral summer than in other43

seasons. In the future, the effects of these two forcings are however predicted to oppose each44

other (Shindell and Schmidt 2004; Perlwitz et al. 2008; McLandress et al. 2011), as Antarctic45

ozone concentrations are anticipated to increase due to the implementation of the Montreal46

Protocol (Austin et al. 2010).47

While the surface climate impact of increasing greenhouse gases is relatively well un-48

derstood, our understanding of stratospheric ozone-related climate change at the surface,49

especially its mechanisms, is somewhat limited. In particular the impact of stratospheric50

2

ozone changes on the hydrological cycle in the SH is not well understood. A series of recent51

studies have shown that stratospheric ozone depletion has likely enhanced austral-summer52

precipitation changes in the subtropics and high-latitudes but reduced them in mid-latitudes,53

consistent with the poleward displacement of the westerly jet, or equivalently the positive54

trend in the Southern Annular Mode (SAM) index (Son et al. 2009; McLandress et al. 2011;55

Polvani et al. 2011; Kang et al. 2011). Although they are crucial for understanding salinity56

changes in the Southern Ocean, net hydrological changes, including evaporation, are not yet57

well understood. In addition and arguably more importantly, potential changes in extreme58

precipitation events are yet to be investigated. It is known that individual precipitation59

events are likely to get more intense as the climate warms (Emori and Brown 2005; Sun60

et al. 2007; O’Gorman and Schneider 2009). Previous studies suggest that it is predomi-61

nantly thermodynamics that control changes in extratropical extreme precipitation (Emori62

and Brown 2005; O’Gorman and Schneider 2009). Thus, it is questionable whether extreme63

precipitation events will respond to dynamical changes driven by the Antarctic ozone hole.64

The purpose of this study is to bridge the existing gap in understanding the relative65

contributions of anthropogenic greenhouse gas emissions and stratospheric ozone changes66

in forcing changes in the hydrological cycle. Multimodel output from the Climate Model67

Intercomparison Project 3 (CMIP3; Meehl et al. (2007)) are analysed. By grouping models68

into those with prescribed ozone depletion and recovery, and those without it, we show69

that Antarctic ozone forcings significantly affect seasonal-mean precipitation trends in the70

extratropics during austral summer, but play a minimal role in evaporation and extreme71

precipitation trends.72

2. Data and methods73

CMIP3 data from the 20th century climate simulations (20C3m) and 21st century climate74

simulations with the special report on emissions scenarios A1B forcing (A1B) are analysed.75

3

From all available models, the models which archived daily precipitation are first selected.76

For those models, evaporation is calculated from surface latent heat flux as outlined in Yu77

et al. (2008). Each model’s precipitation and evaporation climatologies are then compared78

with Global Precipitation Climatology Project version-2 (GPCP) precipitation (Adler et al.79

2003) and Objectively Analyzed Air-Sea Heat Fluxes version-3 (OAFlux) global ocean evap-80

oration data (Yu et al. 2008). Those models with significant biases1 are discarded and 1981

models are selected for the analyses as described in Table 1.82

All CMIP3 models have prescribed stratospheric ozone concentrations with a seasonal83

cycle. However, not all models have incorporated stratospheric ozone depletion in the latter84

part of the 20th century and recovery in the 21st century, as anthropogenic ozone forcings were85

not mandated in the CMIP3 (Meehl et al. 2007). Ten models prescribed ozone depletion and86

ozone recovery, whilst nine models simply used climatological ozone fields2. As such, models87

are grouped into four sets: those with and without ozone depletion in the 20th century, and88

those with and without ozone recovery in the 21st century. For each group, the multimodel-89

mean climatologies and trends are calculated for the fields of interest (precipitation and90

evaporation) over the 20th and 21st centuries. As in Son et al. (2009), climatologies and91

trends are first calculated for each ensemble member and averaged over all available ensemble92

members of a given model. The ensemble average of each model is interpolated onto a 4◦93

latitude by 4◦ longitude grid and averaged over all available models within a group. Hatching94

is used on trend maps to denote where the multimodel mean trend is greater than or equal95

to one standard deviation of the trends of different models within that group. By comparing96

the multimodel means of each group, the impact of Antarctic ozone forcings on hydrological97

climate changes is systematically examined. Although this approach does not necessarily98

1FGOALS1.0g (IAP, China) is discarded as 20th century precipitation in the high-latitude region is found

to be unreasonably higher than observations and all other models. GISS-ER (NASA, USA) is discarded as

1971–1999 daily precipitation appears to be erroneous across the extent of the SH.2Certain CMIP3 models prescribed ozone depletion in the 20th century but did not prescribe ozone

recovery in the 21st century, however for other reasons, such models were not included in this study.

4

reveal ozone-related surface climate changes, as each group comprises different models, it99

is known that trend differences resulting from different ozone forcings are likely larger than100

those associated with model-dependent internal variabilities (Son et al. 2009).101

Since daily data are archived only for selected decades in the A1B runs, long-term trends102

are estimated in this study using decadal differences. The 20th century change, reflecting the103

impact of ozone depletion, is defined by the difference between 1990–1999 and 1961–1970104

means. Likewise the 21st century change, reflecting the impact of ozone recovery, is defined105

by the difference between 2056–2065 and 1990–1999 means. Since decadal differences are106

qualitatively similar to the linear trends computed from monthly-mean data over 1960–1999107

and 2000–2079 (not shown), they are simply referred to as “trends” in this study. The108

possible changes in extreme precipitation events are examined by decomposing seasonal-109

mean precipitation trends into three regimes (Sun et al. 2007): very-light (0.1–1 mm day−1),110

light (1–10 mm day−1) and moderate-to-heavy (>10 mm day−1) precipitation changes. Five111

extreme precipitation indices are also examined. They are the sum of precipitation on all112

wet days divided by the number of wet days, the sum of rainfall on days exceeding the113

95th percentile threshold as determined for the base period of 1961–1990 (hereafter 95th114

percentile precipitation), the sum of rainfall on days exceeding the 99th percentile threshold115

as determined for the base period of 1961–1990, seasonal maximum one-day precipitation,116

and seasonal maximum five-day consecutive precipitation (ETCCDI/CRD 2009).117

3. Results118

Multimodel-mean trends of austral-summer (DJF) precipitation and evaporation are pre-119

sented in Fig. 1. Only the extratropics, poleward of 30◦S, are shown, as tropical and subtrop-120

ical trends are noisy and largely insignificant. In the 20th century the poleward displacement121

of storm tracks by increasing greenhouse gases causes a dipolar trend in precipitation (Yin122

2005). This is evident in the models without ozone depletion, however trends are only weak.123

5

A similar pattern but with much stronger magnitude is found in the models with prescribed124

ozone depletion, indicating the combined effects of increasing greenhouse gases and ozone125

depletion on extratropical precipitation changes. The opposite is generally true in the 21st126

century: models with prescribed ozone recovery show relatively weaker precipitation trends127

than models with fixed ozone forcing. As shown in previous studies (Son et al. 2009; McLan-128

dress et al. 2011; Polvani et al. 2011), this sensitivity is observed only in austral summer.129

In contrast to the annular-like trends of precipitation, evaporation shows relatively weak130

trends in the extratropics, which lack organisation. The sensitivity of evaporation trends to131

ozone forcings is also weak, although there is a hint that models with ozone depletion have132

a weaker decreasing trend in high-latitude evaporation than those without ozone depletion133

presumably because of the acceleration of surface westerlies by ozone depletion. This result,134

combined with the findings related to precipitation trends, indicates that Antarctic ozone135

forcings modulate long-term trends of surface freshwater flux (or equivalently, precipitation136

minus evaporation) by primarily affecting precipitation trends. Figure 1 (third row) in137

particular suggests that Antarctic ozone depletion has likely contributed to the observed138

freshening of the Southern Ocean (e.g. Boning et al. 2008), but its recovery would have little139

impact on Southern Ocean freshening, as ozone-related precipitation changes in the 21st140

century are partly cancelled by evaporation changes.141

Figure 2 presents the relative contributions of very-light, light and moderate-to-heavy142

precipitation changes to the DJF-mean precipitation changes shown in Fig. 1. It is evident143

that mean precipitation trends are dominated by light precipitation trends in the models144

(second row), the precipitation regime which is somewhat overestimated in the CMIP3 mod-145

els (Dai 2006). Note that, although very-light precipitation events also show significant146

trends, their cumulative impact is very weak (shading interval in the first row is one tenth147

of that in the second row). More importantly, their sensitivity to ozone forcings is exactly148

opposite to those of seasonal-mean precipitation trends (compare the first rows of Fig. 1 and149

Fig. 2). This is somewhat surprising, but similar results, opposite trends between very-light150

6

and light precipitation events, are also found in the precipitation response to anthropogenic151

warming (e.g. Sun et al. 2007). For instance, as shown in the right-most column of Fig. 2,152

very-light and light precipitation trends in the 21st century with fixed ozone but with in-153

creasing greenhouse gases show opposite trends in the high-latitudes. It appears the similar154

excitation of the SAM by both increasing greenhouse gases and varying ozone affects the155

precipitation distribution. The mechanisms for this are however unknown: the response may156

be accounted for in terms of thermodynamic changes by increasing greenhouse gases (i.e. a157

warm atmosphere can hold more moisture, likely causing the probability distribution func-158

tion of precipitation events to shift away from very-light to light precipitation events), but159

is more perplexing in terms of ozone-forced dynamic changes. Further studies are needed.160

The trends in moderate-to-heavy precipitation events (Fig. 2 third row), which are quan-161

titatively similar to 95th percentile precipitation trends (fourth row), are relatively weak.162

Unlike trends in very-light and light precipitation, they show no dipole pattern: trends are163

predominantly positive across the extent of the SH, particularly in the 21st century, reflecting164

more intense and frequent extreme precipitation events with anthropogenic warming. More165

importantly, no notable sensitivity to ozone forcings is observed. Although not shown, a166

lack of sensitivity is also found in the other extreme precipitation indices described in the167

data and methods section. This suggests that Antarctic ozone forcings only play a minimal,168

if any, role in extreme precipitation changes. Here it should be noted that extreme precip-169

itation events are sensitive to model resolution. Comparisons of MIROC3.2 (hires, T106)170

with MIROC3.2 (medres, T42), and of CGCM3.1 (T63) with CGCM3.1 (T47), in fact show171

stronger trends in the higher resolution models (not shown). However, the fact that each172

model group contains models with a variety of resolutions and that multimodel-mean trends173

in extreme precipitation are quantitatively similar between models with time-varying ozone174

and models with fixed ozone, although the former have generally higher resolultion than the175

latter (Table 1), suggests that the conclusion that Antarctic ozone forcings play a minimal176

role in extreme precipitation changes holds.177

7

The above results suggest that Antarctic ozone forcings affect hydrological climate changes178

in the SH extratropics by modifying dynamics that in turn modify light precipitation trends.179

It may be questionable whether results are affected by model groups having different climate180

sensitivities: in fact models with varying ozone are known to have higher climate sensitiv-181

ity to increasing greenhouse gases (Miller et al. 2006). However, that in the 21st century182

models with varying ozone exhibit trends in mean precipitation consistent with ozone re-183

covery, suggests that models with varying ozone are not simply responding more strongly to184

increasing greenhouse gases, as the response to ozone recovery is in the opposite direction.185

Nevertheless, results should be treated with caution, as they are based solely on simple mul-186

timodel averaging. One limitation of multimodel averaging, amongst others, is due to the187

fact that individual models have different climatologies (eg. Barnes and Hartmann 2010).188

For instance, the DJF-mean climatological jet, defined by the maximum westerly wind at189

925 hPa, varies from 43◦S to 56◦S amongst models (not shown). Models with fixed ozone190

forcing generally have a jet in lower latitudes (46.2◦S ± 2.9◦) compared to those with ozone191

depletion (49.8◦S ± 2.7◦) (as expected, the latter is closer to the real atmosphere (50◦S)).192

Since storm tracks are located on the poleward side of the westerly jet, the simple multi-193

model averaging used in this study, which ignores individual models’ bias, could over- or194

underestimate precipitation trends by averaging stormy regions with dry ones.195

Figure 3 shows seasonal-mean precipitation climatology and long-term trends for individ-196

ual models and multimodel means, shifted relative to the climatological jet. Only zonally-197

averaged fields are shown as precipitation trends are largely homogeneous in the zonal di-198

rection (Figs. 1, 2). The CMIP3 models reproduce DJF-mean precipitation reasonably well199

in the high-latitudes (-30–0◦ relative to the jet). However, they generally underestimate it200

in mid-latitudes (0–20◦ relative to the jet) and overestimate it in the subtropics and tropics201

(20–50◦ relative to the jet). Intermodel variation is particularly large in the subtropics and202

tropics, making any influence of ozone forcings difficult to distinguish. In fact, although a203

recent study by Kang et al. (2011) has shown that ozone depletion may have enhanced sub-204

8

tropical precipitation in the SH, no sensitivity of subtropical precipitation trends to ozone205

forcings is found in multimodel-mean trends (second row). A noticeable difference in subtrop-206

ical precipitation trends is only found in the 21st century (second row, right column). This is207

however unlikely to be associated with ozone recovery: it is, instead, thought to be a result208

of the intensification of moderate-to-heavy (fifth row, right column) tropical precipitation,209

caused by anthropogenic warming, amongst models with different tropical climatologies.210

Returning to extratropical precipitation trends, Fig. 3 confirms that ozone depletion211

tends to increase high-latitude precipitation trends but decrease mid-latitude precipitation212

trends by modulating light precipitation events. The role of ozone recovery is the reverse:213

decreasing (increasing) high- (mid-) latitude precipitation trends, relative to greenhouse gas214

induced trends alone. This sensitivity is statistically significant at the 99 % confidence level215

as summarised in Table 2. Note that, while the mid-latitude precipitation trend difference216

in the 20th century is not significant, the decrease in mid-latitude light precipitation by217

ozone depletion is statistically significant. Note also that although the sensitivity of high-218

latitude precipitation to ozone forcings is not significant in the 21st century in Table 2,219

it becomes significant if linear trends of monthly-mean precipitation, which are generally220

stronger than decadal differences, are used. The identical analyses are further performed for221

other seasons and no significant difference in multimodel-mean trends between the models222

with and without time-varying ozone forcings is found.223

4. Discussion224

The multimodel analyses, based on CMIP3 models, show that Antarctic ozone forcings225

significantly affect austral-summer precipitation trends in the SH extratropics. Their effects226

are primarily realised by changes in light precipitation events (1–10 mm day−1), with neg-227

ligible changes in extreme precipitation events attributable to ozone forcings. This lack of228

sensitivity of extreme precipitation events to ozone forcings is somewhat contradictory to the229

9

atmospheric circulation changes by ozone forcings: ozone depletion is known to strengthen230

extratropical westerlies, or equivalently to favour the positive polarity of the SAM (Son et al.231

2008, 2010; McLandress et al. 2011; Polvani et al. 2011). As shown in Fig. 4, trends of both232

seasonal-mean and frequency of occurence of 95th percentile surface westerlies are strength-233

ened by ozone depletion (second row). While mean tropospheric circulation changes are234

consistent with mean precipitation changes, no link is established between extreme events235

(compare first and second rows). In other words, strengthening of the 95th percentile wind236

by ozone depletion does not lead to strengthening of 95th percentile precipitation. This result237

suggests that thermodynamic effects are likely more important than dynamic effects in the238

extratropical extreme precipitation changes as discussed by Emori and Brown (2005) and239

O’Gorman and Schneider (2009). In fact, surface air temperature trends, which control wa-240

ter vapour content in the atmosphere, show a negligible sensitivity to ozone forcings (third241

row).242

As previously stated, the findings of this study are based soley on multimodel averaging243

and thus should be treated with care. Although multimodel averaging can reduce model244

biases, it does not allow direct attribution of SH surface climate changes to Antarctic ozone245

forcings. This approach may also underestimate surface climate responses to ozone forc-246

ings by averaging models with realistic climatologies and trends with those without them.247

More quantitative studies using climate model sensitivity tests (e.g. McLandress et al. 2011;248

Polvani et al. 2011) are needed for better quantifying stratospheric ozone-related hydrological249

climate changes in the SH.250

Acknowledgments.251

We thank Jacques Derome for helpful discussion in the course of this research and the252

anonymous reviewers for helpful comments that improved this manuscript. We also thank253

the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and254

archiving the IPCC/AR4 model data, the JSC/CLIVAR Working Groups on Coupled Mod-255

10

eling (WGCM) and their Coupled Model Intercomparison Project (CMIP) and Climate256

Simulation Panel for organising the model data analysis activity, and the IPCC WG1 TSU257

for technical support. The IPCC Data Archive at Lawrence Livermore National Laboratory258

is supported by the Office of Sciences, U.S. Department of Energy. This study is funded259

by the Korea Polar Research Institute (KOPRI) grant under project PE 11010. The work260

by A.P. is in part supported by the Global Environmental and Climate Change Centre,261

Montreal, Quebec, Canada.262

11

263

REFERENCES264

Adler, R., et al., 2003: The Version-2 Global Precipitation Climatology Project (GPCP)265

monthly precipitation analysis (1979–present). Journal of Hydrometeorology, 4, 1147–1167.266

Arblaster, J. and G. Meehl, 2006: Contributions of external forcings to Southern Annular267

Mode trends. Journal of Climate, 19, 2896–2905.268

Austin, J., et al., 2010: Decline and recovery of total column ozone using a multi-269

model time series analysis. Journal of Geophysical Research, 115, D00M10, doi:10.1029/270

2010JD013857.271

Barnes, E. and D. Hartmann, 2010: Testing a theory for the effect of latitude on the per-272

sistence of eddy-driven jets using CMIP3 simulations. Geophysical Research Letters, 37,273

L15801, doi:10.1029/2010GL044144.274

Boning, C., A. Dispert, M. Visbeck, S. Rintoul, and F. Schwarzkopf, 2008: The response275

of the Antarctic Circumpolar Current to recent climate change. Nature Geoscience, 1,276

864–869, doi:10.1038/ngeo362.277

Cai, W., P. Whetton, and D. Karoly, 2003: The response of the Antarctic Oscillation to278

increasing and stabilized atmospheric CO2. Journal of Climate, 16, 1525–1538.279

Dai, A., 2006: Precipitation characteristics in eighteen coupled climate models. Journal of280

Climate, 19, 4605–4630.281

Emori, S. and S. Brown, 2005: Dynamic and thermodynamic changes in mean and extreme282

precipitation under changed climate. Geophysical Research Letters, 32, L17706, doi:10.283

1029/2005GL023272.284

12

ETCCDI/CRD, 2009: Climate change indices: Definitions of the 27 core indices.285

CCl/CLIVAR/JCOMM Expert Team (ET) on Climate Change Detection and Indices,286

available online from: http://cccma.seos.uvic.ca/ETCCDMI.287

Fyfe, J., 2003: Extratropical Southern Hemisphere cyclones: Harbingers of climate change.288

Journal of Climate, 16, 2802–2805.289

Fyfe, J., G. Boer, and G. Flato, 1999: The Arctic and Antarctic Oscillations and their290

projected changes under global warming. Geophysical Research Letters, 26, 1601–1604.291

Gille, S., 2002: Warming of the Southern Ocean Since the 1950s. Science, 295, 1275–1277,292

doi:10.1126/science.1065863.293

Gillett, N., T. Kell, and P. Jones, 2006: Regional climate impacts of the Southern Annular294

Mode. Geophysical Research Letters, 33, L23704, doi:10.1029/2006GL027721.295

Johanson, C. and Q. Fu, 2009: Hadley cell widening: Model simulations versus observations.296

Journal of Climate, 22, 2713–2725, doi:10.1175/2008JCLI2620.1.297

Kang, S., L. Polvani, J. C. Fyfe, and M. Sigmond, 2011: Impact of Polar Ozone Depletion298

on Subtropical Precipitation. Science, in press.299

Kushner, P., I. Held, and T. Delworth, 2001: Southern Hemisphere atmospheric circulation300

response to global warming. Journal of Climate, 14, 2238–2249.301

Marshall, G., 2003: Trends in the Southern Annular Mode from observations and reanalyses.302

Journal of Climate, 16, 4134–4143.303

McLandress, C., T. Shepherd, J. Scinocca, D. Plummer, M. Sigmond, A. Jonsson, and304

M. Reader, 2011: Separating the dynamical effects of climate change and ozone depletion:305

Part 2. Southern Hemisphere troposphere. Journal of Climate, 24, 1850–1868, doi:10.306

1175/2010JCLI3958.1.307

13

Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J.308

Stouffer, and K. E. Taylor, 2007: The WCRP CMIP3 multimodel dataset. Bulletin of the309

American Meteorological Society, 88, 1383–1394.310

Miller, R., G. Schmidt, and D. Shindell, 2006: Forced annular variations in the 20th century311

intergovernmental panel on climate change fourth assessment report models. Journal of312

Geophysical Research, 111, D18101, doi:10.1029/2005JD006323.313

O’Gorman, P. and T. Schneider, 2009: The physical basis for increases in precipitation ex-314

tremes in simulations of 21st-century climate change. Proceedings of the National Academy315

of Science, 106, 14 773–14 777, doi:10.1073/pnas.0907610106.316

Perlwitz, J., S. Pawson, R. Fogt, J. Nielson, and W. Neff, 2008: Impact of stratospheric317

ozone hole recovery on Antarctic climate. Geophysical Research Letters, 35, L08714, doi:318

10.1029/2008GL033317.319

Polvani, L., D. Waugh, G. Correa, and S.-W. Son., 2011: Stratospheric ozone depletion: The320

main driver of 20th century atmospheric changes in the Southern Hemisphere. Journal of321

Climate, 24, 795–812, doi:10.1175/2010JCLI3772.1.322

Randall, D., et al., 2007: Climate models and their evaluation; in Climate Change 2007:323

The Physical Sciences Basis. Contribution of Working Group I to the Fourth Assessment324

Report of the Intergovernmental Panel on Climate Change. Tech. rep., Intergovernmental325

Panel on Climate Change, Cambridge, United Kingdom. [S. Solomon, D. Qin, M. Manning,326

Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)].327

Seidel, D., Q. Fu, W. Randall, and T. Reichler, 2008: Widening of the tropical belt in a328

changing climate. Nature Geoscience, 1, 21–24, doi:10.1038/ngeo.2007.38.329

Shindell, D. and G. Schmidt, 2004: Southern Hemisphere climate responses to ozone changes330

and greenhouse has increases. Geophysical Research Letters, 31, L18209, doi:10.1029/331

2004GL020724.332

14

Son, S., et al., 2010: Impact of stratospheric ozone on Southern Hemisphere circulation333

change: A multimodel assessment. Journal of Geophysical Research, 115, D00M07, doi:334

10.1029/2010JD014271.335

Son, S.-W., N. Tandon, M. Lorenzo, and D. Waugh, 2009: Ozone hole and South-336

ern Hemisphere climate change. Geophysical Research Letters, 36, L15705, doi:10.1029/337

2009GL038671.338

Son, S.-W., et al., 2008: The impact of stratospheric ozone recovery on the Southern Hemi-339

sphere westerly jet. Science, 320, 1486–1489, doi:10.1126/science.1155939.340

Sun, Y., S. Solomon, A. Dai, and R. Portmann, 2007: How often will it rain? Journal of341

Climate, 20, 4801–4818, doi:10.1175/JCLI4263.1.342

Thompson, D. and S. Solomon, 2002: Interpretation of recent Southern Hemisphere climate343

change. Science, 296, 895–899.344

Wong, A., N. Bindoff, and J. Church, 1999: Large-scale freshening of intermediate waters in345

the Pacific and Indian oceans. Nature, 400, 440–443.346

Yin, J., 2005: A consistent poleward shift of the storm tracks in simulations of 21st century347

climate. Geophysical Research Letters, 32, L18701, doi:10.1029/2005GL023684.348

Yu, L., X. Jin, and R. Weller, 2008: Multidecade Global Flux Datasets from the Objectively349

Analyzed Air-sea Fluxes (OAFlux) Project: Latent and Sensible Heat Fluxes, Ocean350

Evaporation, and Related Surface Meteorological Variables. OAFlux Project Technical351

Report OA-2008-01, available online from: http://oaflux.whoi.edu/documents.html,352

Woods Hole Oceanographic Institution, Woods Hole, USA.353

15

List of Tables354

1 Description of CMIP3 models used in this study. Details of each model are355

described in Randall et al. (2007). Resolutions refer to atmospheric resolution356

and horizontal resolution is approximate for spectral models, where “T” refers357

to triangular truncation. The number of ensemble members refers to those358

used in precipitation analyses. Brackets indicate where different ensemble359

members are used in evaporation analyses. 17360

2 Summary of differences in percentage change between the models with time-361

varying ozone forcings and those with fixed ozone forcing. Here percent-362

age change is defined as a decadal change normalised by long-term climatol-363

ogy, calculated for the high-latitude (averaged over 4–24◦ south of individual364

model’s climatological jet) and mid-latitude (averaged over 12–0◦ north of365

individual model’s climatological jet) regions. Only the values which are sta-366

tistically significant at the 99 % confidence level are shown. Significance tests367

are based on a Monte Carlo approach. This approach selects one group of368

ten models (eight for evaporation) and one group of nine models at random369

and calculates the percentage change difference between the means of the two370

groups. This is repeated 50,000 times to get a statistical distribution. The371

actual difference between the mean of the varying ozone group and the fixed372

ozone group is then compared with this statistical distribution at the 99 %373

confidence level. Although not shown, overall results are qualitatively similar374

to a two-sided Student’s t test at the 99 % confidence level. Only results from375

DJF are shown, as no significant values are found in other seasons. 18376

16

Table 1. Description of CMIP3 models used in this study. Details of each model aredescribed in Randall et al. (2007). Resolutions refer to atmospheric resolution and horizontalresolution is approximate for spectral models, where “T” refers to triangular truncation. Thenumber of ensemble members refers to those used in precipitation analyses. Brackets indicatewhere different ensemble members are used in evaporation analyses.

Model Group, country Horizontal res. Vertical res. 20C3m A1B(lat. × lon.) levels, top members members

Varying ozoneCCSM3.0 NCAR, USA T85 (1.4◦ × 1.4◦) 26, 2.2 hPa 4 (3) 5CSIRO-Mk3.0 CSIRO, Australia T63 (1.9◦ × 1.9◦) 18, 4.5 hPa 3 (2) 1CSIRO-Mk3.5d CSIRO, Australia T63 (1.9◦ × 1.9◦) 18, 4.5 hPa 3 1ECHAM5/MPI-OM MPI, Germany T63 (1.9◦ × 1.9◦) 31, 10 hPa 2 2GFDL-CM2.0 NOAA, USA 2.0◦ × 2.5◦ 24, 3 hPa 1 1GFDL-CM2.1 NOAA, USA 2.0◦ × 2.5◦ 24, 3 hPa 1 (0) 1 (0)INGV-SXG INGV, Italy T106 (1.1◦ × 1.1◦) 19, 10 hPa 1 1MIROC3.2(hires) CCSR, Japan T106 (1.1◦ × 1.1◦) 56, 40 km 1 1MIROC3.2(medres) CCSR, Japan T42 (2.8◦ × 2.8◦) 20, 30 km 2 3 (1)PCM1.1 NCAR, USA T42 (2.8◦ × 2.8◦) 26, 2.2 hPa 3 (0) 1 (0)Fixed ozoneBCCR-BCM2.0 BCCR, Norway T63 (1.9◦ × 1.9◦) 16, 25 hPa 1 1CGCM3.1(T47) CCCma, Canada T47 (2.8◦ × 2.8◦) 31, 1 hPa 5 3CGCM3.1(T63) CCCma, Canada T63 (1.9◦ × 1.9◦) 31, 1 hPa 1 1CNRM-CM3∗ CNRM, France T63 (1.9◦ × 1.9◦) 45, 0.05 hPa 1 1ECHO-G MIUB, Germ./Korea T30 (3.9◦ x 3.9◦) 19, 10 hPa 3 (1) 3 (1)GISS-AOM NASA, USA 3.0◦ × 4.0◦ 12, 10 hPa 1 1INM-CM3.0 INM, Russia 4.0◦ × 5.0◦ 21, 10 hPa 1 1IPSL-CM4 IPSL, France 2.5◦ × 3.7◦ 19, 4 hPa 2 1MRI-CGCM2.3.2 MRI, Japan T42 (2.8◦ × 2.8◦) 30, 0.4 hPa 5 (1) 5 (1)∗ Model documentation claims inclusion of ozone chemistry, however analysis of Antarctic polar-captemperature by Son et al. (2008) found no ozone impact in either 20C3m or A1B simulations.

17

Table 2. Summary of differences in percentage change between the models with time-varying ozone forcings and those with fixed ozone forcing. Here percentage change is definedas a decadal change normalised by long-term climatology, calculated for the high-latitude(averaged over 4–24◦ south of individual model’s climatological jet) and mid-latitude (av-eraged over 12–0◦ north of individual model’s climatological jet) regions. Only the valueswhich are statistically significant at the 99 % confidence level are shown. Significance testsare based on a Monte Carlo approach. This approach selects one group of ten models (eightfor evaporation) and one group of nine models at random and calculates the percentagechange difference between the means of the two groups. This is repeated 50,000 times toget a statistical distribution. The actual difference between the mean of the varying ozonegroup and the fixed ozone group is then compared with this statistical distribution at the99 % confidence level. Although not shown, overall results are qualitatively similar to atwo-sided Student’s t test at the 99 % confidence level. Only results from DJF are shown,as no significant values are found in other seasons.

High-latitudes Mid-latitudes20thC 21stC 20thC 21stCDJF DJF DJF DJF

Mean precipitation 3.1 % 2.6 %Very-light precipitation 3.5 %Light precipitation 4.2 % -4.1 %Moderate-to-heavy precip.

18

List of Figures377

1 Multimodel-mean trends of precipitation (first row), evaporation (second row)378

and precipitation minus evaporation (third row) in DJF. Trends are calcu-379

lated by differencing 1961–1970 from 1990–1999 averages, and 1990–1999 from380

2056–2065 averages, for 20th and 21st century trends, respectively. From left381

to right, multimodel-mean trends are shown for models with and without382

ozone depletion in the 20th century, and for models with and without ozone383

recovery in the 21st century respectively. Cool colours denote an increas-384

ing freshwater flux (increasing precipitation or decreasing evaporation) whilst385

warm colours denote a decreasing freshwater flux (decreasing precipitation or386

increasing evaporation). Hatched areas denote where the multimodel-mean387

trend is greater than or equal to one standard deviation. Contours show the388

climatology, with contour intervals of 100 mm season−1 for all panels. 21389

2 Multimodel-mean trends in very-light (0.1–1 mm day−1, first row), light (1–10390

mm day−1, second row), moderate-to-heavy (>10 mm day−1, third row), and391

95th percentile (fourth row) precipitation in DJF. Other details are the same392

as Fig. 1. Note that the colour scale is an order of magnitude less for very-light393

precipitation so that details can be seen. Contour intervals of climatologies394

are 20 mm season−1 for very-light precipitation panels and 50 mm season−1395

for all other panels. 22396

3 Zonal-mean precipitation climatology (first row) and total (second row), very-397

light (third row), light (fourth row), and moderate-to-heavy (fifth row) pre-398

cipitation trends in DJF, plotted as a function of jet-relative latitudes in the399

SH. Zonal-mean values are shown for individual models (varying ozone models400

in blue and fixed ozone models in red), multimodel averages (bold blue and401

red lines) and GPCP precipitation (bold black line). Both 20th century (left402

column) and 21st century (right column) simulations are shown. 23403

19

4 Zonal-mean precipitation (first row), surface zonal wind (second row) and404

surface air temperature (third row) trends in DJF, plotted as a function of405

jet-relative latitudes in the SH. Mean trends (left column) and frequency of406

occurence of 95th percentile events (right column) are shown. Only 20th cen-407

tury simulations are presented. Note that the frequency trends on the right408

are different from 95th percentile precipitation trends, as the latter is a cumu-409

lative quantity. Due to surface wind data availability, CCSM3.0, PCM1.1 and410

INM-CM3.0 are not included in these panels. The same colour convention as411

Fig. 3 is used. 24412

20

Fig. 1. Multimodel-mean trends of precipitation (first row), evaporation (second row) andprecipitation minus evaporation (third row) in DJF. Trends are calculated by differencing1961–1970 from 1990–1999 averages, and 1990–1999 from 2056–2065 averages, for 20th and21st century trends, respectively. From left to right, multimodel-mean trends are shown formodels with and without ozone depletion in the 20th century, and for models with and with-out ozone recovery in the 21st century respectively. Cool colours denote an increasing fresh-water flux (increasing precipitation or decreasing evaporation) whilst warm colours denotea decreasing freshwater flux (decreasing precipitation or increasing evaporation). Hatchedareas denote where the multimodel-mean trend is greater than or equal to one standarddeviation. Contours show the climatology, with contour intervals of 100 mm season−1 for allpanels.

21

Fig. 2. Multimodel-mean trends in very-light (0.1–1 mm day−1, first row), light (1–10 mmday−1, second row), moderate-to-heavy (>10 mm day−1, third row), and 95th percentile(fourth row) precipitation in DJF. Other details are the same as Fig. 1. Note that the colourscale is an order of magnitude less for very-light precipitation so that details can be seen.Contour intervals of climatologies are 20 mm season−1 for very-light precipitation panels and50 mm season−1 for all other panels.

22

0

200

400

600

800

clim

atol

ogy

(mm

)

20thC DJF

varying O3

fixed O3

GPCP

−10

−5

0

5

10

15

mea

n tr

end

−0.5

0

0.5

very

−lig

ht tr

end

−5

0

5

light

tren

d

−40 −30 −20 −10 0 10 20 30 40 50−10

−5

0

5

10

15

latitude relative to the jet (o)

mod

erat

e−to

−he

avy

tren

d

21stC DJF

−40 −30 −20 −10 0 10 20 30 40 50

latitude relative to the jet (o)

prec

ipita

tion

tren

ds (

mm

/ se

ason

/ de

cade

)

Fig. 3. Zonal-mean precipitation climatology (first row) and total (second row), very-light(third row), light (fourth row), and moderate-to-heavy (fifth row) precipitation trends inDJF, plotted as a function of jet-relative latitudes in the SH. Zonal-mean values are shown forindividual models (varying ozone models in blue and fixed ozone models in red), multimodelaverages (bold blue and red lines) and GPCP precipitation (bold black line). Both 20th

century (left column) and 21st century (right column) simulations are shown.

23

−10

−5

0

5

10

prec

ipita

tion

(mm

/ se

ason

/ de

cade

)

20thC DJF mean trends

varying O3

fixed O3

−1

−0.5

0

0.5

120thC DJF 95th percentile trends

−0.5

0

0.5

surf

ace

zona

l win

d(m

s−1

/ de

cade

)

−40 −30 −20 −10 0 10 20 30 40 50−1

−0.5

0

0.5

1

latitude relative to the jet (o)

(num

ber

of e

vent

s / s

easo

n / d

ecad

e)

−40 −30 −20 −10 0 10 20 30 40 50−0.5

0

0.5

surf

ace

air

tem

p.(K

/ de

cade

)

latitude relative to the jet (o)

Fig. 4. Zonal-mean precipitation (first row), surface zonal wind (second row) and surfaceair temperature (third row) trends in DJF, plotted as a function of jet-relative latitudesin the SH. Mean trends (left column) and frequency of occurence of 95th percentile events(right column) are shown. Only 20th century simulations are presented. Note that thefrequency trends on the right are different from 95th percentile precipitation trends, as thelatter is a cumulative quantity. Due to surface wind data availability, CCSM3.0, PCM1.1and INM-CM3.0 are not included in these panels. The same colour convention as Fig. 3 isused.

24