1

Dynamical and thermodynamical modulations of future changes in landfalling 1

atmospheric rivers over North America 2

3

Yang Gao, Jian Lu, L. Ruby Leung, Qing Yang, Samson Hagos and Yun Qian 4

Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, 5

Richland, Washington, USA 6

Correspondence to: Dr. L. Ruby Leung (Ruby.Leung@pnnl.gov) 7

8

2

Abstract 9

This study examines the changes of landfalling atmospheric rivers (ARs) over the west coast of 10

North America in response to future warming using outputs from the Coupled Model 11

Intercomparison Project phase 5 (CMIP5). The result reveals a strikingly large increase of AR 12

days by the end of the 21st century in the RCP8.5 climate change scenario, with fractional 13

increases ranging between ~50% and 600%, depending on the seasons and landfall locations. 14

These increases are predominantly controlled by the super-Clausius-Clapeyron rate of increase 15

of atmospheric water vapor with warming, while changes of winds that transport moisture in the 16

ARs, or dynamical effect, mostly counter the thermodynamical effect of increasing water vapor, 17

limiting the increase of AR events in the future. The consistent negative effect of wind changes 18

on AR days during spring and fall can be linked to the robust poleward shift of the subtropical jet 19

in the North Pacific basin. 20

Keywords: Atmospheric rivers, CMIP5, global warming, Clausius-Clapeyron relation, jet 21

stream shift 22

23

3

1. Introduction 24

Atmospheric rivers (ARs) are narrow corridors of water vapor in the lower troposphere that 25

tranverse several thousand kilometers and transport over 90% of atmospheric moisture across 26

subtropical boundaries from their tropical source to midlatitude destinations globally [Zhu and 27

Newell, 1998]. In North Pacific, when atmospheric rivers make landfall at the coastal western 28

United States, they can bring about record-setting precipitations, causing water-related hazards 29

such as floods and mudslides [Dettinger, 2011; Lavers et al., 2011; Leung and Qian, 2009; 30

Neiman et al., 2011; Ralph et al., 2006]. ARs can also have benevolent impacts to society by 31

relieving drought impacts [Dettinger, 2013; Witze, 2015]. 32

Considerable efforts have been made to study the historical AR events in the Pacific and their 33

hydrological impacts to the western states of America [Dettinger, 2013; Jiang et al., 2014; Leung 34

and Qian, 2009; Neiman et al., 2008; Payne and Magnusdottir, 2014; Ralph and Dettinger, 2012; 35

Ralph et al., 2004]. Given that the water vapor holding capacity of the atmosphere increases 36

following the Clausius-Clapeyron relation [Held and Soden, 2006], the number of the ARs could 37

increase in an alarming rate as climate warms. Recently, [Lavers et al., 2013] examined ARs that 38

make landfall in the United Kindom in the simulations of CMIP5 and found that the frequency of 39

these landfalling ARs is projected to double by the end of this century under the RCP 8.5 climate 40

change scenario. Their results also suggested that the increase of ARs arises mainly from the 41

increase of moisture as climate warms, while the contribution from the change of winds that 42

transport moisture is hardly detectable. By analyzing the A2 scenario of CMIP3, Dettinger [2011] 43

found that the number and intensity of wintertime ARs making landfall in California remain 44

roughly unchanged by the end of the 21st century, but the years with many AR episodes and the 45

number of ARs with much-larger-than-historical water vapor transport is projected to increase. 46

4

More recently, [Warner et al., 2015] evaluated ARs along the North America west coast 47

simulated by 10 CMIP5 models under RCP8.5 and found the number of days of IVT above the 48

historical 99th

percentile increasing by almost three times by the end of the 21st century. However, 49

they did not detect ARs directly from the simulations, and their results for the 99th

percentile IVT 50

may be relevant mainly to the most intense ARs. 51

As the atmospheric circulation in the North Pacific and the related transient variability will be 52

significantly modified under global warming (e.g., [Barnes and Polvani, 2013; Gao et al., 2014; 53

Neelin et al., 2013]). This will inevitablly exert an influence on the distribution of landfalling 54

ARs, but dynamical changes are arguably less certain than thermodynamical changes. In this 55

study, we will take it as our central task to use the full CMIP5 archive to i) document the 56

seasonal characteristics of landfalling ARs over the west coast and their changes under future 57

warming and ii) identify the dynamical and thermodynamical contributions to these changes. 58

2. Data and method 59

A total of 24 CMIP5 models are used in this study (see Table S1 for the list). The climate 60

change statistics of ARs are obtained by contrasting two 30-year periods, i.e., historical period 61

1975-2004 and future warming period 2070-2099 under RCP 8.5 [Moss et al., 2010; Vuuren et 62

al., 2011]. To evaluate the performance of the CMIP5 models in capturing the statistics of the 63

ARs, four reanalysis datasets are used. More details regarding the CMIP5 data and the four 64

reanalysis data were discussed in section 1 in the supplementary material. We adopt the 65

canonical IVT-based criteria for detecting ARs, while taking into consideration the coarse 66

vertical resolution of the CMIP5 data archive (only levels at 1000hPa, 850hPa, 700hPa, and 67

5

500hPa are archived). The IVT was calculated by vertically integrating the moisture transport 68

between 1000 hPa and 500 hPa pressure levels [Lavers et al., 2012; Warner et al., 2015] as 69

𝐼𝑉𝑇 = √(1

𝑔∫ 𝑞𝑢 𝑑𝑝

500

1000)

2

+ (1

𝑔∫ 𝑞𝑣 𝑑𝑝

500

1000)

2

, 70

where g is the gravitational acceleration, q is layer mean specific humidity, 𝑢 and 𝑣 are zonal and 71

meridional wind, respectively. With one reanalysis dataset (CFSR), we also confirm that IVT 72

calculated using only 4 vertical layers as with the CMIP5 models introduced negligible 73

uncertainty compared to using 16 layers from 1000 hPa to 500 hPa. 74

Following [Lavers and Villarini, 2013; Lavers et al., 2012], we first identify the ARs that make 75

landfall in the west coast of North America between 25°N to 60°N, as indicated by the colored 76

grid cells in Figure 1. Except Alaska and northern Canada (they belong to the same group), the 77

coastal grids are grouped into 7 bins from 25°N to 60°N, each spanning 5° in latitude. For each 78

model, we first compute the daily IVT for the 30-year historical period (denoted as 𝑉1𝑄1) and the 79

30-year future warming period (𝑉2𝑄2), respectively. On each day, the maximum IVT along the 80

west coast is recorded and the corresponding grid is identified; it is then evaluated against the 81

85th

percentile of the IVT of the bin group that the grid belongs to. The 85th

percentile threshold 82

is estimated using all the daily data for each of the 8 bins for each model separately. 83

If the IVT value of the identified grid exceeds the threshold, we search backward for the 84

maximum IVT values among the four adjacent grids (northwest/west/southwest/south) and 85

assess whether any of them also exceeds the 85th

percentile threshold of that bin. The upstream 86

search continues until none of the IVT values of the four upstream grids exceeds the threshold, 87

yielding the track of a prospective AR. If the track spans longer than 2000 km [Neiman et al., 88

6

2008; Ralph et al., 2004], we compute its mean vertically integrated water vapor (IWV) by 89

averaging over all the grids occupied by this track. If the equivalent precipitable water of the 90

resultant mean IWV is greater than 2 cm, an AR is detected and all the track grids are construed 91

to have an AR day [Ralph et al., 2004; Wick et al., 2013]. Similar to Lavers et al. [2012], no 92

width criterion is considered in our AR detections. Figure 1 shows an example of an AR track 93

that battered the northwest coast of the US on Dec 28, 1998 from ERA-Interim. 94

3. Results 95

3.1 AR climatology in CMIP5 simulations 96

To evaluate the credence of the CMIP5 models in capturing the basic statistics of the AR events, 97

we compare the number of AR days within the 8 coastal bins in each season from the CMIP5 98

models against the four reanalysis datasets in Figure 2. Overall, the CMIP5 multi-model 99

ensemble mean (MME) tracks remarkably well the latitudinal and seasonal variations of AR 100

events, with an exception of the underestimation of the springtime number of AR days near the 101

southwest coast of the US (the bin between 35°N and 40°N). Consistent with previous studies, 102

landfalling ARs are more prevalent during fall and winter (e.g., [Neiman et al., 2008]). The peak 103

of the landfall shifts from the higher latitudes to the Californian coast from fall to winter; it 104

becomes smaller and retreats poleward more abruptly from the colder to warmer season. In 105

addition, there are also considerable AR events that make landfall in the Alaska coast in summer 106

(about 4/bin/season, see Neiman et al. [2008]). To further assess the fidelity of the CMIP5 107

models, we also composite the IVT, near surface velocity, and sea level pressure (SLP) based on 108

the AR events for CMIP5 multi-model ensemble (MME; Figure S1) and ERA-Interim (Figure 109

S2), and found consistent spatial distributions. 110

7

3.2 The change of ARs and thermodynamical and dynamical modulations 111

The impact of climate change on the statistics of landfalling ARs is elucidated in Figure 3 by 112

comparing the numbers of AR days estimated from the present-day simulations for 1975-2004 113

(black) with those from the future RCP 8.5 scenario for 2070-2099 (red), with the percentage 114

increase indicated on the top row of the numbers. The Student’s t test is used to assess whether 115

the MME differences are statistically significant, with the significant differences highlighted in 116

red. It is striking to see that the increases in all seasons over all the coastal areas of North 117

America are significant, and the west coast will experience a manyfold increase of AR days, 118

ranging from doubling (near California coast during winter and spring) to 6 times (along the 119

Alaskan coast in spring), depending on the seasons and locations. However, caution should be 120

used in the interpretation of the changes near the west coast of the US and Mexico during 121

summer, as AR events are very rare so the sample size may not yield a credible assessment. 122

In an attempt to separate the effect of wind changes or dynamical effects from that of increasing 123

moisture or thermodynamical effects in the projected increase of AR days, for each model, each 124

season, and each grid point, we rescale the present-day IVT by a factor of 𝑞𝑚2

𝑞𝑚1, where qm is the 125

thirty-year average of the IWV over the eastern Pacific basin (25°N to 55°N, 180°W to 130°W) 126

for the corresponding model and season, and subscript 1 and 2 indicate the present-day and 127

future episodes, respectively. The resultant IVT, referred to as 𝑉1�̅�2 symbolically, is used to 128

identify ARs in a conjured-up scenario in which the present day wind advects the moisture 129

scaled to have the same mean moisture of the future warmer climate. An example of the 130

probability distributions of historical, future and rescaled IVT is shown in Figure S3. We apply 131

the AR detection procedure to the rescaled data and the resulting statistics of ARs are plotted as 132

8

the blue lines in Figure 3. The difference between the blue line 𝑉1�̅�2 and the black solid line 133

(representing 𝑉1𝑄1) can be construed as the contribution of increasing water vapor in the future 134

climate to the total change of the AR frequency, and the corresponding percentage increases are 135

indicated by the numbers in the 2nd

row in Figure 3. Alternatively, one could also scale the future 136

scenario case (𝑉2𝑄2) back with the ratio of 𝑞𝑚1

𝑞𝑚2, and contrasting 𝑉2𝑄2 (red lines) with 𝑉2�̅�1 137

(orange lines) should result in the same thermodynamical effect. Quantitatively similar fractional 138

changes result from 𝑉2𝑄2 − 𝑉2�̅�1 compared to 𝑉1�̅�2 − 𝑉1𝑄1 (not shown). Through the rescaling 139

above, one may also infer the effect of the changing advection wind in the ARs, or the dynamical 140

modulation, by comparing 𝑉2�̅�1 against 𝑉1𝑄1, or 𝑉2𝑄2 against 𝑉1�̅�2. The percentage differences 141

of (𝑉2�̅�1 − 𝑉1𝑄1)/𝑉1𝑄1 are shown as the color-coded numbers in the 3rd row in Figure 3, with 142

the red numbers indicating significant differences at 95% confidence level. As a cross-validation, 143

we also compute the percentage changes of (𝑉2𝑄2 − 𝑉1�̅�2)/𝑉2𝑄2 and the result is qualitatively 144

consistent, but with non-negligible difference. 145

As explained in section 4 in the supplementary material, the condition for the rescaling to work 146

is that the rate of increase of the IVW in the ARs can be approximated by that of the seasonal 147

mean IVW, i.e., 𝑞𝑚2

𝑞𝑚1

𝑞1

𝑞2= 1. This holds true only approximately, as ARs are associated with 148

anomalously high moisture and strong low-level winds, which may respond to future warming 149

differently compared to the seasonal mean. But for most cases, the approximation has an error of 150

up to about 10% (see Table S2), which may be tolerable for the purpose of qualitative evaluation 151

for the relative contributions of dynamical and thermodynamical effects. For a sufficiently small 152

error, the dynamical ((𝑉2�̅�1 − 𝑉1𝑄1)/𝑉1𝑄1) and thermodynamical ((𝑉1�̅�2 − 𝑉1𝑄1)/𝑉1𝑄1) effects 153

from this rescaling exercise should add up to the total change. However, adding the 3rd

row to the 154

9

2nd

row does not result in the numbers in the 1st row in Figure 3. This is mainly because i) the 155

rescaling for the dynamical effect is very sensitive to the accuracy of the assumption 𝑞𝑚2

𝑞𝑚1

𝑞1

𝑞2= 1 156

(see the 2nd

term of equation S3) and ii) the co-variation component between the change of wind 157

and moisture is ignored in the rescaling approach. As such, the numbers listed in Figure 3 can 158

only be interpreted heuristically. Nevertheless, it is clear that water vapor increase plays an 159

overwhelmingly dominant role in the increase of AR days, while the dynamical effects are 160

ubiquitously negative or negligible for almost all seasons and over all latitudinal areas with only 161

one exception: a positive dynamical contribution of 39% to the increase of the spring time ARs 162

near the Alaskan coast. Statistically significant dynamical effects are found in the change of ARs 163

in spring and fall, with the largest dynamical reduction (by 46%) associated with the ARs that 164

influence the California-Oregon border in fall. This seasonal dependence of the dynamical 165

modulation is not by coincident; its connection to the underlying large-scale circulation will be 166

discussed later in this section. 167

Since this rescaling to estimate the dynamical effect has ignored the co-variation component, it 168

risks underestimating the dynamical modulation on the frequency of ARs. If viewing the co-169

variation component as being organized by the storm track dynamics so that it can be ascribed to 170

the dynamical effect, we can infer the total dynamical effect by taking the difference between the 171

first and second row of numbers. The result would be qualitatively similar (in seasonality and 172

latitudinal structure) to that from the direct rescaling, but with much larger magnitude. For 173

example, the total dynamical reduction on the ARs reaching the coast between 40°N and 50°N in 174

spring season amounts to >100% (compared to the modest numbers estimated from the direct 175

rescaling). In view of this ambiguity, a more rigorous approach to further quantify the dynamical 176

contribution to the AR change from changing atmospheric circulation is warranted. 177

10

The thermodynamical dominance on the increase of the ARs can also be illustrated by the 178

Clausius-Clapeyron (C-C) scaling following the approach of Lavers et al. [2013]. This is done by 179

rescaling the specific humidity by the C-C ratio of increase (7%/oC) based on the near surface 180

temperature increase averaged over the eastern Pacific basin (25°N to 55°N, 180°W to 130°W). 181

The resultant statistics of ARs (denoted by 𝑉1�̅�𝑐𝑐) is presented as the black dashed line in Figure 182

3 and its discrepancies (in terms of fractional difference) from the AR statistics based on scaling 183

by the actual future humidity (𝑉1�̅�2), i.e., (𝑉1�̅�2 − 𝑉1�̅�𝑐𝑐)/𝑉1�̅�2, are displayed in the fourth row 184

of numbers in Figure 3. It is interesting to note that the C-C scaling underestimates the actual 185

increase of the water vapor in the ARs throughout all seasons and for all coast areas of North 186

America. This appears to be consistent with the super-C-C increase of column-integrated water 187

vapor associated with the precipitation extreme simulated in an idealized aquaplanet model in Lu 188

et al. [2014]. However, its origin is less clear because ARs can draw moisture from multiple 189

pathways [Ryoo et al., 2015], so the C-C ratio and warming that influence AR moisture in the 190

future may deviate from the C-C ratio and the mean warming over the eastern Pacific that are 191

used in the C-C scaling. It suffices to say, however, that the C-C scaling largely explains the 192

CMIP5 ensemble mean change in AR days in the future. 193

3.3 The change of AR pathways 194

In view of the negative impact of the circulation changes on the occurrence of ARs, it would be 195

interesting to examine the related pattern of the AR pathways upstream in the North Pacific. To 196

this end, we first compute for each grid the frequency of occurrence of the ARs based on the 197

detected AR tracks for the present-day and future IVT, i.e., 𝑉1𝑄1 and 𝑉2𝑄2, respectively, as well 198

as the rescaled IVT 𝑉1�̅�2 and 𝑉2�̅�1. Then the fractional change of frequency through 199

operation(𝑉2𝑄2 − 𝑉1�̅�2)/𝑉2𝑄2 represents the dynamical modulation on the typical pathways of 200

11

the landfalling ARs at the west coast. The results for all four seasons are presented in Figure 4, 201

and these results, with larger sample sizes, are consistent with those from the operation of 202

((𝑉2�̅�1 − 𝑉1𝑄1)/𝑉1𝑄1) (not shown). Consistent with the negative dynamical effect on the AR 203

frequency at the landfall regions, the frequency of the AR pathways is also reduced by the wind 204

changes over the oceanic regions to the west of the North American coast. These reductions are 205

significant and robust for spring, summer and fall seasons. In winter, there is an increase of AR 206

pathways to the west of the Mexican coast, appearing to be guided by the enhanced westerly 207

wind there. However, this increase does not result in significant landfall. It is interesting to note 208

that the dynamically induced change of frequency is characterized by a southeast-northwest 209

dipole in both spring and fall, consistent with the northward and slightly westward shift of the 210

circulation pattern indicated by the composite SLP anomalies (contours in Figure 4a,c). 211

As the poleward shift of the subtropical high is most robust in spring and fall seasons [Simpson 212

et al., 2014] as the regional signal of the expansion of the Hadley circulation [Lu et al., 2007], 213

we underscore that a poleward shift of the mean circulation may also project on a weakening of 214

the ARs. This is because the poleward shift of the mean circulation structure will shift the origins 215

and pathways of the AR to higher latitudes, as ARs have been associated with wave breaking in 216

the storm track [Ryoo et al., 2013]. As a consequence the ARs may be weakened by the reduced 217

water vapor available from the cooler surface that provides the moisture for ARs near their 218

origins and/or along their pathways [Hagos et al., 2015]. Despite model consistency in projecting 219

negative impact of circulation changes on AR days, analysis discussed in section 5 of the 220

supplementary material suggests that uncertainty in projecting the meridional shift of the 221

subtropical high in the eastern basin of the Pacific contributes to considerable inter-model spread 222

in the change of the AR days during winter near the southern coast of the US. Hence the co-223

12

variation of moisture and wind changes that influence AR changes in a warmer climate should be 224

further investigated in the future. 225

Conclusions 226

We investigated the seasonal variations of ARs and their prospect of change in a warming 227

climate under the RCP8.5 scenario over the coastal regions of the western US and Canada, 228

including Alaska. Despite the relatively coarse horizontal resolutions of the CMIP5 climate 229

models to capture the full intensity of ARs, they do a very reasonable job in capturing the 230

observed seasonality in the frequency of occurrence of ARs and even the intensity of the 231

moisture transport therein. Under the future RCP8.5 climate change scenario, owing 232

predominantly to the increase of water vapor, the number of AR days increases ubiquitously for 233

all coastal regions and for all seasons, with magnitudes ranging from 50% to more than 6 times. 234

It is especially noteworthy that the greatest increase in frequency occurs to the ARs landfalling in 235

the Alaskan coast in summer, reaching almost 13 days by the end of the 21st century (compared 236

to 4 days in the present climate). 237

A simple rescaling analysis is performed to tease out the relative importance of the changing 238

water vapor and the changing wind to the projected increase of ARs. Almost for all seasons and 239

all the coastal regions examined, the thermodynamical (water vapor) effect dominates the 240

dynamical (wind) one in the total increase, with significant negative dynamical contributions 241

detected for the west coast of the conterminous US and Canada for spring and fall and a 242

significant positive dynamical contribution to the ARs reaching Alaskan coast in spring. On the 243

other hand, no significant dynamical effect can be detected in the wintertime change of the ARs, 244

13

which may be a result of large inter-model uncertainties in the projections of winter time 245

circulation pattern over the eastern North Pacific among the CMIP5 models. 246

We caution that for the rescaling to work, an assumption has been made that the fractional 247

increase of the IWV in the ARs, which tends to occur at the tail of the IWV distribution, equals 248

to the fractional increase of the seasonal mean IWV. As our results suggest that this is not a very 249

accurate approximation, the rescaling approach for the role of the dynamics should be taken as 250

heuristic. Nevertheless, confidence can be assigned to the qualitative conclusions about the 251

dynamical modulations on the ARs. 252

253

Acknowledgments 254

This study was supported by the U.S. Department of Energy Office of Science Biological and 255

Environmental Research (BER) as part of the Regional and Global Climate Modeling program. 256

PNNL is operated for DOE by Battelle Memorial Institute under contract DE-AC05-76RL01830. 257

We acknowledge the World Climate Research Programme's Working Group on Coupled 258

Modelling, which is responsible for CMIP, and we thank the climate modeling groups (listed in 259

Table S1 of the supporting information) for producing and making available their model output. 260

261

References 262

Barnes, E. A., and L. Polvani (2013), Response of the midlatitude jets, and of their variability, to 263

increased greenhouse gases in the CMIP5 models, Journal of Climate, 26(18), 7117-7135. 264

Dettinger, M. D. (2011), Climate change, atmospheric rivers, and floods in California – A multimodel 265

analysis of storm frequency and magnitude changes, JAWRA Journal of the American Water 266

Resources Association, 47(3), 514-523. 267

14

Dettinger, M. D. (2013), Atmospheric rivers as drought busters on the U.S. west coast, Journal of 268

Hydrometeorology, 14(6), 1721-1732. 269

Gao, Y., L. R. Leung, J. Lu, Y. Liu, M. Huang, and Y. Qian (2014), Robust spring drying in the 270

southwestern U.S. and seasonal migration of wet/dry patterns in a warmer climate, Geophysical 271

Research Letters, 41(5), 2014GL059562. 272

Hagos, S., L. R. Leung, Q. Yang, C. Zhao, and J. Lu (2015), Resolution and dynamical core dependence 273

of atmospheric river frequency in global model simulations, Journal of Climate, 28(7), 2764-274

2776. 275

Held, I. M., and B. J. Soden (2006), Robust responses of the hydrological cycle to global warming, 276

Journal of Climate, 19(21), 5686-5699. 277

Jiang, T., K. J. Evans, Y. Deng, and X. Dong (2014), Intermediate frequency atmospheric disturbances: A 278

dynamical bridge connecting western U.S. extreme precipitation with East Asian cold surges, 279

Journal of Geophysical Research: Atmospheres, 119(7), 2013JD021209. 280

Lavers, D. A., and G. Villarini (2013), The nexus between atmospheric rivers and extreme precipitation 281

across Europe, Geophysical Research Letters, 40(12), 3259-3264. 282

Lavers, D. A., G. Villarini, R. P. Allan, E. F. Wood, and A. J. Wade (2012), The detection of atmospheric 283

rivers in atmospheric reanalyses and their links to British winter floods and the large-scale 284

climatic circulation, Journal of Geophysical Research: Atmospheres, 117(D20), D20106. 285

Lavers, D. A., R. P. Allan, E. F. Wood, G. Villarini, D. J. Brayshaw, and A. J. Wade (2011), Winter 286

floods in Britain are connected to atmospheric rivers, Geophysical Research Letters, 38(23), 287

L23803. 288

Lavers, D. A., R. P. Allan, G. Villarini, B. Lloyd-Hughes, D. J. Brayshaw, and A. J. Wade (2013), Future 289

changes in atmospheric rivers and their implications for winter flooding in Britain, Environmental 290

Research Letters, 8(3), 034010. 291

15

Leung, L. R., and Y. Qian (2009), Atmospheric rivers induced heavy precipitation and flooding in the 292

western U.S. simulated by the WRF regional climate model, Geophysical Research Letters, 36(3), 293

L03820. 294

Lu, J., G. A. Vecchi, and T. Reichler (2007), Expansion of the Hadley cell under global warming, 295

Geophysical Research Letters, 34(6), L06805. 296

Lu, J., L. Ruby Leung, Q. Yang, G. Chen, W. D. Collins, F. Li, Z. Jason Hou, and X. Feng (2014), The 297

robust dynamical contribution to precipitation extremes in idealized warming simulations across 298

model resolutions, Geophysical Research Letters, 41(8), 2014GL059532. 299

Moss, R. H., et al. (2010), The next generation of scenarios for climate change research and assessment, 300

Nature, 463, 747-756. 301

Neelin, J. D., B. Langenbrunner, J. E. Meyerson, A. Hall, and N. Berg (2013), California winter 302

precipitation change under global warming in the coupled model intercomparison project phase 5 303

ensemble, Journal of Climate, 26(17), 6238-6256. 304

Neiman, P. J., F. M. Ralph, G. A. Wick, J. D. Lundquist, and M. D. Dettinger (2008), Meteorological 305

characteristics and overland precipitation impacts of atmospheric rivers affecting the west coast 306

of North America based on eight years of SSM/I satellite observations, Journal of 307

Hydrometeorology, 9(1), 22-47. 308

Neiman, P. J., L. J. Schick, F. M. Ralph, M. Hughes, and G. A. Wick (2011), Flooding in western 309

washington: the connection to atmospheric rivers, Journal of Hydrometeorology, 12(6), 1337-310

1358. 311

Payne, A. E., and G. Magnusdottir (2014), Dynamics of landfalling atmospheric rivers over the north 312

Pacific in 30 years of MERRA reanalysis, Journal of Climate, 27(18), 7133-7150. 313

Ralph, F. M., and M. D. Dettinger (2012), Historical and national perspectives on extreme west coast 314

precipitation associated with atmospheric rivers during December 2010, Bulletin of the American 315

Meteorological Society, 93(6), 783-790. 316

16

Ralph, F. M., P. J. Neiman, and G. A. Wick (2004), Satellite and CALJET aircraft observations of 317

atmospheric rivers over the eastern north Pacific Ocean during the winter of 1997/98, Monthly 318

Weather Review, 132(7), 1721-1745. 319

Ralph, F. M., P. J. Neiman, G. A. Wick, S. I. Gutman, M. D. Dettinger, D. R. Cayan, and A. B. White 320

(2006), Flooding on California's Russian River: role of atmospheric rivers, Geophysical Research 321

Letters, 33(13), L13801. 322

Ryoo, J.-M., D. E. Waliser, D. W. Waugh, S. Wong, E. J. Fetzer, and I. Fung (2015), Classification of 323

atmospheric river events on the U.S. West Coast using a trajectory model, Journal of Geophysical 324

Research: Atmospheres, 2014JD022023. 325

Ryoo, J.-M., Y. Kaspi, D. W. Waugh, G. N. Kiladis, D. E. Waliser, E. J. Fetzer, and J. Kim (2013), 326

Impact of Rossby Wave Breaking on U.S. West Coast Winter Precipitation during ENSO Events, 327

Journal of Climate, 26(17), 6360-6382. 328

Simpson, I. R., T. A. Shaw, and R. Seager (2014), A diagnosis of the seasonally and longitudinally 329

varying midlatitude circulation response to global warming, Journal of the Atmospheric Sciences, 330

71(7), 2489-2515. 331

Vuuren, D., et al. (2011), The representative concentration pathways: an overview, Climatic Change, 332

109(1-2), 5-31. 333

Warner, M. D., C. F. Mass, and E. P. Salathé (2015), Changes in winter atmospheric rivers along the 334

North American west coast in CMIP5 climate models, Journal of Hydrometeorology, 16(1), 118-335

128. 336

Wick, G. A., P. J. Neiman, and F. M. Ralph (2013), Description and validation of an automated objective 337

technique for identification and characterization of the integrated water vapor signature of 338

atmospheric rivers, Geoscience and Remote Sensing, IEEE Transactions on, 51(4), 2166-2176. 339

Witze, A. (2015), California study targets rivers in the sky, Nature, 517(7535), 424-425. 340

Zhu, Y., and R. E. Newell (1998), A proposed algorithm for moisture fluxes from atmospheric rivers, 341

Monthly Weather Review, 126(3), 725-735. 342

17

Figures 343

344

Figure 1. An example of IVT distribution associated with an atmospheric river event on Dec 28 1998 from ERA-345

Interim reanalysis data. Only IVT greater than 400 kg m-1

s-1

(close to 85th

percentile of historical period) is shown. 346

The black contours indicate the daily total precipitation on that day, with a minimum contour value of 4 mm, at 2 347

mm interval. The color-coded squares are the grids used to detect landfalling ARs, and grids with the same color 348

belong to the same latitude bin. 349

350

351

352

353

354

355

18

356

Figure 2. Seasonal variations of the number of atmospheric river days in the eight latitudinal bins from 25° to 60°N 357

estimated from each of the CMIP5 models (gray) and their MME mean (blue line) during 1975-2004, and from four 358

reanalysis data sets: CFSR (solid red), ERA-INTERIM (solid green), MERRA (dashed red), and NCEP1 (dashed 359

green) during 1979-2004. 360

19

361

Present: 𝑉1𝑄1 RCP8.5: 𝑉2𝑄2 𝑉2�̅�1 𝑉1�̅�2 362

Figure 3. Number of AR days by seasons for 8 latitudinal bins along the x-axis for the present day (1975-2004, 363

black) and future climate under RCP8.5 (2070-2099, red) simulated by the CMIP5 MME. Also shown are the 364

number of AR days determined by rescaling the future IVT (𝑉2�̅�1, orange), rescaling the present-day IVT 365

(𝑉1�̅�2, blue), and rescaling the present-day IVT with the C-C rate (dashed black). The shading indicates one 366

standard deviation of the inter-model spread. In each panel, the numbers in the first row indicate the percentage 367

change of AR days in future compared to present; the second row shows the effect of increasing moisture; the third 368

row shows the effect of wind changes; the fourth row indicates the differences between the rescalings using future 369

mean water vapor versus the C-C ratio. Red numbers are statistically significant at 95% level. 370

20

371

Figure 4. Spatial distribution of the MME mean dynamical modulation on AR frequency (in percentage) determined 372

from the future IVT versus present day IVT scaled by the ratio of future to present seasonal mean water vapor. Only 373

grid points with landfalling ARs detected in at least 10 CMIP5 models and more than 70% of these models agreeing 374

on the sign of the dynamical change are displayed. Overlaid are the composite sea level pressure differences 375

between the AR days determined from the future and scaled IVT, with thick dark gray contours for positive values 376

and light thin gray contours for negative values at contour intervals of 40 Pa. Similar differences for winds are 377

shown by the vectors. 378

379

380

381

382

383