1
On the Response of the Aleutian Low to Greenhouse Warming
2
Bolan Gan1*, Lixin Wu1, Fan Jia2, Shujun Li1
3
Wenju Cai3,1, Hisashi Nakamura4, Michael Alexander5, and Art Miller6
4
1
5
Ocean Dynamic Process and Climate, Qingdao National Laboratory for Marine Science and
6
Technology
7
2
8
of Sciences
9
3
Physical Oceanography Laboratory/CIMST, Ocean University of China and Laboratory for
Key Laboratory of Ocean Circulation and Waves, Institute of Oceanology, Chinese Academy
CSIRO Marine and Atmosphere Research
10
4 Research
11
5
National Oceanic and Atmospheric Administration/Earth System Research Laboratory
12
6
Scripps Institution of Oceanography, University of California San Diego
Center for Advanced Science and Technology, University of Tokyo
13
14
15
16
Submitted to Journal of Climate
17
18
19
20
21
22
23
*Corresponding author: Dr. Bolan Gan, Physical Oceanography Laboratory, Ocean University
24
of China, 5 Yushan Road, Qingdao 266003, P. R. China. Email: gbl0203@ouc.edu.cn
1
25
Abstract
26
In this paper, both observations and Coupled Model Intercomparison Project Phase 5
27
(CMIP5) multi-models are used to investigate changes in the Aleutian Low in the past and
28
future climate. It is found that the intensity of the Aleutian Low, measured by the North Pacific
29
Index (NPI), has significantly strengthened by approximately 1.5 hPa on the average of five
30
different observation-based datasets in the 20th century. This observed centennial-trend of NPI,
31
however, is 2 times as large as the modeled counterpart for the CMIP5 multi-model ensemble
32
mean of the historical simulations. As the climate warms under the RCP8.5 scenario, CMIP5
33
models predict a robust northward intensification of the Aleutian Low, manifested in the
34
decreased NPI of approximately -1.3 hPa or -0.4 hPa per degree global surface warming, which
35
is 62% larger than the estimated internal variability of NPI (i.e., 0.8 hPa), and the central
36
low-pressure region expanded about 7 times as large as that in the 20th century. A set of
37
idealized experiments with an intermediate climate model further demonstrate that the
38
deepening of the Aleutian Low can be driven by an El-Niño-like warming pattern in the
39
tropical Pacific sea surface temperature (SST), with a reduction in the time-mean zonal SST
40
gradient, which overshadows the effect of the weakened land-ocean thermal contrast that tends
41
to dampen the Aleutian Low response to greenhouse forcing. This warming pattern is also
42
found to influence the winter precipitation responses to greenhouse warming in the North
43
Pacific Rim, likely owning to the Aleutian Low change.
44
2
45
1. Introduction
46
As one of the main centers of action in the atmospheric circulation over the Northern
47
Hemisphere, the Aleutian Low is a semi-permanent low-pressure system at the surface
48
centered over the Aleutian Islands chain (Fig. 1a). The pressure center is lowest in boreal
49
winter and nearly vanishing in summer. The Aleutian Low is associated with powerful
50
cyclones, which considerably affect formation of pack ice in the Bearing Sea (Overland and
51
Pease 1982) and the temperature extremes in the North Pacific Rim (Kenyon and Hegerl 2008).
52
It also closely links with upper-level teleconnections (Overland et al. 1999), such as the
53
Pacific-North American teleconnection (PNA) and the Arctic Oscillation, which are the
54
dominant modes of atmospheric circulation over the Northern Hemisphere. The Aleutian Low
55
thus plays a fundamental role in regulating the winter climate of the North Pacific and the
56
North American continent (e.g., Trenberth and Hurrell 1994; Deser et al. 2004). Furthermore,
57
changes in the intensity and position of the Aleutian Low substantially affect the North Pacific
58
oceanic gyres and upper-ocean temperature field, owing to the influence of wind stress curl and
59
thermal forcing (e.g., Seager et al. 2001; Kwon and Deser 2007; Pickart et al. 2009), which can
60
thereby alter marine biological resources (e.g., Wyllie-Echeverria and Wooster 1998; Benson
61
and Trites 2002) and fish stocks in the Northeast Pacific (e.g., Hollowed et al. 2001; Chavez et
62
al. 2003; Lehodey et al. 2006). Variations of the Aleutian Low are related not only intimately to
63
decadal climate variability over the North Pacific (e.g., Latif and Barnett 1994; Schneider and
64
Cornuelle 2005), but also remotely to the tropical interannual El Niño-Southern Oscillation
65
(ENSO; Alexander et al. 2002).
66
Observational evidence indicated that the Aleutian Low has strengthened since the late
67
1970s, based on the sea level pressure (SLP) reanalysis spanning the second half of the 20th
68
century (e.g., Gillett et al. 2003; Lu et al. 2004; Deser and Phillips 2009). The enhanced
69
basin-scale cyclonic flow associated with this stronger Aleutian Low brings colder and drier air
3
70
to the western North Pacific, and the reverse to the eastern North Pacific, resulting in
71
precipitation decreased over coastal Asia and increased over southern Alaska as well as the
72
southwestern United States (Deser et al. 2004; Honda et al. 2005). This multidecadal shift of
73
the Aleutian Low, corresponding to changes toward the positive phase of Northern Annular
74
Mode (e.g., Trenberth et al. 2007), is also found to largely contribute to the pattern of
75
wintertime surface-warming over the Northern Hemisphere continents during 1965-2000
76
(Wallace et al. 2012). However, no clear consensus on the relative importance of anthropogenic
77
warming and natural variability upon such dynamical contribution from the changing
78
atmospheric circulation to land surface warming, since most climate models are deficient in
79
reproducing the concurrent SLP variations (Gillett et al. 2005; Wallace et al. 2012). Beyond
80
that multidecadal enhancement of the Aleutian Low, its long-term changes over the 20th
81
century remains unknown.
82
Early modeling studies have mentioned an anomalously deepened Aleutian low-pressure
83
center under the greenhouse gas forcing (e.g., Meehl and Washington 1996; Boer et al. 2000;
84
Salathé 2006). However, the robust greenhouse warming signals in the changing Aleutian Low
85
is challenged, since the relatively large internal variability of extratropical atmosphere
86
introduces substantial uncertainty in the externally-forced change. Deser et al. (2012) and
87
Oshima et al. (2012) indicated that the internal climate variability primarily contributes to the
88
total uncertainty in the projected change of SLP during the first half of the 21st century. In
89
addition to the atmospheric internal variability, different responses of drivers of the Aleutian
90
Low variations to greenhouse forcing may lead to distinctions in the Aleutian Low response.
91
One of the important drivers is the tropical Pacific sea surface temperature (SST), with an
92
El-Niño-induced anomalous warming in the eastern tropical Pacific corresponding to a
93
strengthened Aleutian Low (Alexander et al. 2002). Climate models seem to suggest an
94
enhanced SST warming in the eastern tropical Pacific under greenhouse warming (e.g., An et al.
4
95
2012; Yeh et al. 2012), but reconstructed SST datasets spanning the 20th century demonstrate
96
diverse warming trends (Deser et al. 2010). Additionally, a weakening of the winter land-ocean
97
thermal contrast (LOTC) under greenhouse warming may result in a weakened Aleutian Low.
98
Therefore, it remains uncertain how robust the climatological-mean Aleutian Low responds to
99
greenhouse warming and what driving mechanisms are involved. Here, we investigated
100
long-term changes in the Aleutian Low over the 20th century using five observational SLP
101
datasets and CMIP5 multi-model historical simulations, and then projected changes in the
102
Aleutian Low is examined based on the 21st century climate change simulations under the
103
RCP8.5 scenario. The possible mechanism underlying is explored by a set of idealized
104
experiments conducted on an intermediate climate model.
105
The paper is organized as follows. Section 2 briefly describes the datasets, methods and an
106
intermediate climate model (CAM3.1-RGO). Section 3 characterizes changes in the Aleutian
107
Low intensity over the 20th century and its response to greenhouse warming. Section 4
108
presents the possible mechanism driving the Aleutian Low response that is demonstrated by
109
sensitivity experimental results. A summary and discussion is given in Section 5.
110
2. Datasets and methods
111
a. Observational NPI in the 20th Century
112
In this study, the intensity of the Aleutian Low is represented by the North Pacific Index
113
(NPI), which is defined as the area-weighted average of winter (December to February, DJF)
114
SLP over the region bounded by 30°‒65°N and 160°E‒140°W (Trenberth and Hurrell 1994).
115
Centennial-long NPIs were constructed from five historical observation-based SLP datasets.
116
The monthly NPI data is directly available at https://climatedataguide.ucar.edu/
117
climate-data/north-pacific-np-index-trenberth-and-hurrell-monthly-and-winter,
118
derived from the National Center for Atmospheric Research’ SLP dataset (NPI-NCAR;
119
Trenberth and Paolino 1980). The other four monthly SLP datasets include two reanalysis
which
is
5
120
datasets, i.e. the National Oceanic and Atmospheric Administration (NOAA)/Cooperative
121
Institute for Research in Environmental Sciences 20th Century Reanalysis version 2 (20CRv2;
122
2°×2° resolution; Compo et al. 2011) and the newly-developed European Centre for
123
Medium-Range Weather Forecasts (ECMWF) 20th Century Reanalysis (ERA20c; 1°×1°
124
resolution), and two reconstruction datasets, i.e. the NOAA Extended Reconstructed SLP
125
(ERSLP; 2°×2° resolution; Smith and Reynolds 2004) and the Met Office Hadley Centre's SLP
126
dataset version 2 (HadSLP2; 5°×5° resolution; Allan and Ansell 2006).
127
b. CMIP5 multi-model outputs
128
The simulated SLP, SST and surface air temperature (SAT) field were taken from 22
129
climate models (Table 1) that participate in the CMIP5 (Taylor et al. 2012), organized by the
130
Program for Climate Model Diagnosis and Intercomparison for the Intergovernmental Panel on
131
Climate Change Fifth Assessment Report. We analyzed three sets of simulations: the
132
pre-industrial control simulations that represent the natural variability of climate system with
133
atmospheric CO2 concentration fixed at 280-290 ppm, the historical simulations that
134
incorporate the anthropogenic and natural forcings from the observed atmospheric composition
135
changes in the 20th century, and the future climate change simulations under the RCP8.5
136
scenario that are characterized by an escalating radiative forcing throughout the 21st century
137
(reaching approximately 8.5 Wm−2 in 2100; equivalent to atmospheric CO2 concentration
138
exceeding 1370 ppm). All model outputs were interpolated to a common 1°×1° grid for SST
139
and 2°×2° grid for both SLP and SAT. In each model, the 50-yr-interval difference of variables
140
between RCP8.5 run (2050-2099) and historical run (1950-1999) is considered as the forced
141
response to greenhouse warming.
142
c. Estimation of trend and statistical significance
143
We estimated centennial trends with the Sen median slope (Sen 1968), which is much less
144
sensitive to outliers and skewed distributions than the conventional least-squares fit. The
6
145
corresponding statistical significance was assessed using the modified Mann-Kendall trend test,
146
a non-parametric method with accounting for the autocorrelation of time series (Hamed and
147
Rao 1998). We scaled all trends to the 100-yr change.
148
The statistical significance of correlation coefficient between the NPI and LOTC/SST
149
responses among the CMIP5 models was estimated using a non-parametric Monte Carlo test
150
(Livezey and Chen 1983). Specifically, the correlation coefficient was calculated 1000 times
151
using the original sequence of the LOTC/SST responses in the models with the NPI response
152
series randomly scrambled. Then the probability distribution function of these 1000 correlation
153
coefficients was constructed to rank the significance of the actual correlations.
154
d. Estimation of internal variability in the CMIP5 models
155
Following Collins et al. (2013), unforced internal variability of climate system in the
156
CMIP5 models is estimated using at least 500-year long (after spin-up period) outputs of
157
pre-industrial control simulations for each model. We calculated non-overlapping 20-year
158
means for each grid point, from which a quadratic fit as a function of time is subtracted to
159
eliminate model drift. Variability for each grid point was then estimated as the standard
160
deviation of that 20-year means, multiplied by the square root of 2 to account for the fact that
161
the variability of a difference in means is of interest. “This is by definition the standard
162
deviation of the difference between two independent 20-year averages having the same
163
variance and estimates the variation of that difference that would be expected due to unforced
164
internal variability”, as statement in the Collins et al. (2013). The median of those quantities
165
across all models was used as the estimated internal variability.
166
e. CAM3.1-RGO Model
167
We performed a set of idealized experiments with an intermediate climate model
168
(CAM3.1-RGO) that is a fully coupled system consisting of the Community Atmosphere
169
Model version 3 (CAM3.1; Collins et al. 2006) and a 1.5-layer reduced-gravity ocean (RGO)
7
170
model with flux corrections (Fang 2005; Jia and Wu 2013). The atmospheric component is part
171
of the Community Climate System Model version 3 (CCSM3) developed at the National
172
Center for Atmospheric Research (NCAR). It is based on a Eulerian spectral dynamical core,
173
with a T42 horizontal resolution and 26 vertical levels. The land surface processes in CAM3.1
174
are represented by a fully interactive land model, the Community Land Model version 3
175
(CLM3; Bonan et al. 2002). The oceanic component is an extended Zebiak-Cane type of
176
1.5-layer RGO model (Zebiak and Cane 1987; Clement et al. 1996), in which the active upper
177
layer is divided into a fixed depth mixed layer to simulate SST variation and a subsurface layer
178
to parameterize the entrained subsurface temperature through the multivariate linear
179
relationship with thermocline depth. This ocean model, containing variability off the equatorial
180
band, covers a global domain (80°S-80°N, 0°-360°) with 1° latitude by 2° longitude resolution,
181
which has been successfully used to study tropical oceanic processes and ENSO (e.g., Chiang
182
et al. 2008; Jia and Wu 2013).
183
3. Aleutian Low changes in the past and future climate
184
a. Long-term changes over the 20th century
185
First, we examined changes in the Aleutian Low intensity during the 20th century, based
186
on NPIs constructed from five different observation-based SLP datasets draw from
187
reconstructions and reanalysis. As shown in Fig. 1b, all the observed NPIs display consistent
188
and significant interannual-to-decadal fluctuations, e.g. decadal transitions in the 1940s and
189
1970s, except for the period of 1900-1910 in which NPIs are diverse from each other. Despite
190
this short-term variability, a long-term downward trend can be perceived over the 20th century.
191
Stepping aside the inconsistency among the constructed NPIs in pre-1910, we evaluated the
192
long-term trends of NPIs as the Sen median slope from the start of 1911 to the end of the
193
available SLP data (i.e., 1911-2013 for NPI-NCAR, 1911-2011for 20CRv2, 1911-2009 for
194
ERA20c, 1911-1996 for ERSLP, and 1911-2003 for HadSLP2). Figure 1c shows that the
8
195
centennial-trend estimations of NPIs scaled to the 100-yr change are -1.5 hPa for NCAR-NPI,
196
-1.9 hPa for ERA20c, -1.2 hPa for HadSLP2 (all significant at the 95% confidence level), and
197
-1.4 hPa for 20CRv2 as well as -1.3 hPa for ERSLP (both significant at the 83% confidence
198
level). Therefore, NPI averaged over five observations has decreased by approximately -1.5
199
hPa throughout the 20th century, with a 95% confidence interval of -1.1 hPa to -1.9 hPa based
200
on a two-tailed Student’s t-test. This indicates that the Aleutian Low has intensified by
201
approximately 16% of the relative climatological-mean SLP gradient, defined as the minimum
202
SLP minus the area-averaged SLP in the aforementioned region (i.e., 9.3 hPa for the reference
203
period 1950-1999).
204
A natural question arises as to what the performance of CMIP5 climate models in
205
simulating the variability as well as long-term changes of NPI over the 20th century. Here we
206
examined the simulated NPIs constructed from 22 CMIP5 multi-model SLP outputs of the
207
historical simulations. Comparing Fig. 2a and Fig. 1a, the models can simulate the
208
climatological-mean Aleutian Low reasonably well, although the modeled minimum SLP is
209
slightly larger by 0.2 hPa and displaced southeastward by 2.5° latitude and 9.5° longitude
210
relative to that observed. However, the models generally lack the capability to reproduce the
211
variation of NPI in observation and most models underestimate the interannual standard
212
deviation of NPI. As shown in Fig. 2b, the multi-model ensemble mean (MMEM) time series
213
of historical NPI is substantially smoother than the observed NPI shown in Fig. 1b, with much
214
lower standard deviation than that of NCAR-NPI (0.7 hPa versus 3.1 hPa), owing to the large
215
inter-model spread (shading in Fig. 2b). Table 2 further indicates that 14 out of 22 CMIP5
216
models underestimate the interannual standard deviation of NPI by ranging from 4% to 27% of
217
the observed counterpart. Nevertheless, a statistically-significant downward trend of NPI can
218
be detected in the MMEM historical simulations, with the magnitude of 0.7 hPa per 100-yr that
219
is about half as much as the aforementioned one in observation (Fig. 2b). In fact, looking into
9
220
the performance of each model (Fig. 2c) indicates that 16 out of 22 CMIP5 models are
221
agreeable on the sign of downward centennial-trend, but diverse on the magnitude (a range of
222
0.1-3.3 hPa per 100-yr), such that only 5 models produce significant trends larger than or
223
comparable to the observation. Overall, the downward long-term trend of NPI in both
224
observations and the MMEM historical simulations may reflect an intensification of the
225
Aleutian low in response to anthropogenic warming over the past century. Such signal in
226
CMIP5 models, however, is confounded by the internal variability, since it is not highly robust
227
across all models. Indeed, we further evaluated the internal variability of NPI based on the
228
pre-industrial control simulations (Table 2), which shows that the internal variability
229
estimation is approximately 0.8 hPa, slightly larger than the centennial trend of MMEM NPI.
230
b. Future responses to greenhouse warming in the 21st century
231
To better understand how the Aleutian Low changes under external greenhouse forcing,
232
we examined the 50-yr-interval NPI/SLP difference between the future projections under the
233
RCP8.5 scenario (2050-2099) and the historical simulations (1950-1999) in 22 CMIP5 models,
234
which is considered as the response of climatological-mean Aleutian Low to greenhouse
235
warming. Overall, CMIP5 models predict a significant northward intensification of the
236
Aleutian Low with high model agreement in response to greenhouse warming during the 21st
237
century. As shown in Fig. 3a, more than 70% (16/22) of models agree on the sign of decreased
238
NPI, indicating a robust deeper-than-normal Aleutian Low. The MMEM of NPI changes shows
239
a decrease of approximately -1.3 hPa (a 95% confidence interval of -0.7 hPa to -1.9 hPa) or
240
approximately 15% of the relative climatological-mean SLP gradient in the historical
241
simulations, which is 62% larger than the estimated internal variability of NPI. This
242
50-yr-interval change of NPI is indeed almost identical to its long-term trend scaled to the
243
100-yr change across the 21st century (2006-2099), which is higher than the corresponding
244
variability in the MMEM RCP8.5 projections (i.e., interannual standard deviation of the
10
245
MMEM NPI is 0.8 hPa). Further normalizing the NPI change by the corresponding value of
246
global-mean surface temperature change for each model and then taking MMEM, it shows that
247
the scaled NPI change is approximately -0.4 hPa °C-1, with a 95% confidence interval of -0.2
248
hPa °C-1 to -0.6 hPa °C-1. In addition, comparing Fig. 3b and Fig. 2a, the central region of the
249
Aleutian Low (see the isobar of 999.0 hPa) in the 21st century expand about 7 times as large as
250
that in the 20th century, and the minimum SLP decreased by 3.3 hPa.
251
A robust northward intensification of the low-pressure system is clearly manifested in the
252
spatial pattern of winter SLP responses over the North Pacific. Figure 3c shows a
253
southeastward-tilted pattern of the estimated internal variability of winter SLP, with one center
254
of action of 1.6 hPa located over the Bering Sea and the other one of 1.5 hPa located over the
255
eastern North Pacific. Under the greenhouse forcing, the low-pressure system to north of 45°N
256
except over the western coast of North American displays a highly robust intensification, with
257
the maximum decrease of SLP reaching -5.5 hPa over the northern Bering Sea (stippling region
258
in Fig. 3d), which is about 3.5 times greater than the internal variability shown in Fig. 3c. This
259
is accompanied with a weak increase of SLP along a zonal band extending from Japan to the
260
central North Pacific. The MMEM SLP changes over the band of 30°-45°N and the
261
southeastern part of the North Pacific are found to be less robust, which could be resulted from
262
smaller changes in individual models than the internal variability and/or high-level
263
disagreement on the sign of change across models. Further inspection indicates, on average, a
264
larger contribution from small changes for regions to east of 160°W, with 47% of models
265
exhibiting small changes versus 35% of models disagreeable on the sign of MMEM change.
266
However, for the rest (i.e., region of 30°-45°N, 130°E-160°W), contributions from both of
267
them are identical, with 16% of models displaying both aspects and additional 23% of models
268
displaying either one aspect.
269
4. Possible mechanism driving the intensification of the Aleutian Low
11
270
in a warmer climate
271
a. Contrasting roles of the LOTC and tropical Pacific SST responses
272
The Aleutian Low is a surface manifestation of the atmospheric planetary-scale waves
273
forced partly by the LOTC in the wintertime (Terada and Hanzawa 1984). During winter, the
274
air over the North Pacific is much warmer than over the Eurasian and North American
275
continents, leading to the formation of the oceanic low-pressure system. Given this fact, we
276
investigated the relationship between responses of the Aleutian Low and the winter (DJF)
277
LOTC to greenhouse warming by comparing the NPI changes with the LOTC changes for 22
278
CMIP5 models. Here the LOTC was estimated as the difference between SAT over the North
279
Pacific
280
80°E-120°E)], based on the center of the “cold ocean-warm land” pattern induced by the
281
land-ocean distribution (Wallace et al. 1996).
Ocean
and
Asia
[i.e.,
SAT(40°N-60°N,
170°E-150°W)-SAT(40°N-60°N,
282
As shown in Fig. 4, all but one of the models projects a weakening of the LOTC,
283
exhibiting an MMEM decrease of -1.6°C, with a 95% confidence interval of -1.1°C to -2.1°C.
284
In addition, the majority of CMIP5 models (15 of 22) project the intensification of the Aleutian
285
Low coherently with the weakening of the thermal contrast, represented by decreases of the
286
NPI and LOTC, respectively. The changes in the NPI and LOTC are significantly correlated
287
among the CMIP5 models, with correlation coefficient of -0.47 significant at P=0.02 based on
288
a Monte Carlo test, suggesting that the models with larger decrease in the LOTC tend to have a
289
smaller decrease or even increase in the NPI. In other word, the weaker LOTC that arises in a
290
warmer climate dampens the response of Aleutian Low to greenhouse warming. Therefore, the
291
intensification of the Aleutian Low must involve other processes that dominate such effect of
292
the weakened LOTC.
293
As aforementioned, previous studies have highlighted the important role of the tropical
294
Pacific SST in modifying the Northern hemisphere winter atmospheric circulation during the
12
295
second half of the 20th century (e.g., Schneider et al. 2003; Lu et al. 2004; Deser and Phillips
296
2009). Particularly, an El Niño event acts to deepen the Aleutian Low through atmospheric
297
teleconnections (Alexander et al. 2002), which may operate on longer climate-change time
298
scales. Early works of Meehl and Washington (1996) and Boer et al. (2000) draw an
299
association between the deepened Aleutian low-pressure center and the enhanced warming in
300
the eastern tropical Pacific Ocean under greenhouse gas forcing. Based on these observations,
301
we explored the tropical Pacific SST changes in a warmer climate and its potential linkage with
302
the intensification of the Aleutian Low.
303
From Fig. 5a, it can be seen that the CMIP5 projections exhibit large SST warming in
304
winter over the eastern equatorial Pacific (EEP; 5°S–5°N, 90°–150°W), with a maximum
305
warming of 3.1oC predicted near the Galápagos Islands, which is higher than that over the
306
western equatorial Pacific (WEP; 5°S–5°N, 120°–170°E). This indicates a reduction in the
307
time-mean zonal SST gradient across the equatorial Pacific, which is referred as an
308
El-Niño-like warming (without implying a change in the El Niño variability). Notably, more
309
than two thirds (17/22) of the models produce an El-Niño-like warming pattern in the tropical
310
Pacific SST, as seen by the positive east-west warming differences across the equatorial Pacific
311
(i.e., SST warming area-averaged in the EEP minus that in the WEP) in the individual models
312
(Fig. 5b). The MMEM for this difference is 0.3°C, with a 95% confidence interval of 0.1°C to
313
0.5°C. We further compared the NPI changes with the east-west SST warming differences
314
projected in the CMIP5 models (Fig. 5b). It is found that the NPI changes are significantly
315
correlated with the tropical east-west SST warming differences among all the models, with a
316
correlation coefficient of -0.45 significant at P=0.04 based on a Monte Carlo test. This suggests
317
that an El-Niño-like warming in the tropical Pacific Ocean, which occurs as a result of
318
greenhouse warming, is likely to be an important driver in the intensification of the Aleutian
319
Low and will continue to do so as the climate warms further.
13
320
b. Results of idealized experiments
321
A recent study by Jia and Wu (2013) indicates that the response of the equatorial Pacific
322
SST to the doubled CO2 concentration (2CO2) forcing in the CAM3.1-RGO model strongly
323
resembles a robust El-Niño-like warming pattern. An additional examination shows that the
324
Aleutian Low is intensified in the 2CO2 run. This motivated us to conduct idealized
325
experiments with the CAM3.1-RGO model, to further verify the active role played by an
326
El-Niño-like warming of the tropical Pacific SST in the intensification of the Aleutian Low
327
under greenhouse forcing.
328
We first conducted a pair of multi-century integrations of CAM3.1-RGO fully-coupled
329
model: a 400-yr control run with the CO2 concentration fixed at 360 ppm
330
(CAM3.1-RGO-CTRL) and a 300-yr run under the quadrupled CO2 concentration (4CO2)
331
forcing
332
CAM3.1-RGO-CTRL. Then we performed a set of sensitivity experiments with CAM3.1
333
model stand-alone: 1) a 100-yr control run with the CO2 concentration fixed at 360 ppm
334
(CAM3.1-CTRL), in which CAM3.1 was forced by the monthly SST fields derived from the
335
last 200 years of CAM3.1-RGO-CTRL; 2) a 100-yr 4CO2 run with the CO2 concentration fixed
336
at 1460 ppm (CAM3.1-4CO2), in which CAM3.1 was forced by the monthly SST fields
337
derived from the last 200 years of CAM3.1-RGO-4CO2; 3) a 100-yr global uniform warming
338
run with the CO2 concentration fixed at 1460 ppm (CAM3.1-4CO2-GUW), in which CAM3.1
339
was forced by the monthly and spatially uniform SST increase obtained as the globally
340
averaged SST warming difference between CAM3.1-RGO-4CO2 and CAM3.1-RGO-CTRL
341
(over the last 200 years) added to the monthly SST field of CAM3.1-RGO-CTRL. Note that in
342
the CAM3.1-4CO2-GUW, the El-Niño-like warming pattern was removed. CAM3.1
343
reasonably simulated the Aleutian Low, albeit the NPI weaker than that observed by
344
approximately 3 hPa.
(CAM3.1-RGO-4CO2)
that
integrates
starting
at
the
100th
year
of
14
345
As for the response to quadrupled CO2 concentration forcing, the CAM3.1-RGO coupled
346
model produces a well-defined El-Niño-like warming in the tropical Pacific, with the east-west
347
warming difference of 1.5°C (Fig. 6a), and strengthening of the Aleutian Low as well as
348
weakening of the LOTC (similar to the results of CAM3.1 sensitivity run), as in the majority of
349
CMIP5 models. In the CAM3.1-4CO2 run, the Aleutian Low is significantly intensified
350
northward, with SLP change of -4.2 hPa over the northern Bering Sea (Fig. 6b), and the NPI
351
decreased by 1.4 hPa (Fig. 6d). However, when the El-Niño-like warming pattern was removed,
352
the Aleutian Low is considerably weakened (Fig. 6c), manifested as a substantial contraction of
353
the low-pressure system, with the maximum increase of SLP reaching 12 hPa over the eastern
354
North Pacific and the NPI increased by 4.1 hPa (Fig. 6d). In both CAM3.1-4CO2 and
355
CAM3.1-4CO2-GUW runs, the winter LOTC was significantly weakened, with the change of
356
-2.6 °C and -2.8 °C, respectively (Fig. 6e). Therefore, the results of idealized experiments
357
further confirm that an El-Niño-like warming in the tropical Pacific Ocean is a critical forcing
358
mechanism that causes the Aleutian Low to intensify under greenhouse warming.
359
The El-Niño-like warming pattern of tropical Pacific SST is supposed to force a planetary
360
wave train, akin to the positive phase of PNA, in the extratropical atmosphere through latent
361
heat release in tropical precipitation, which in turn intensifies the Aleutian Low (e.g., Trenberth
362
et al. 1998; Lu et al. 2004). Here we further examined the winter precipitation changes in the
363
CAM3.1 experiments to look into the influences of El-Niño-like SST warming (Fig. 7). The
364
winter precipitation pattern simulated in the CAM3.1-CTRL run (Fig. 7a) is largely similar to
365
the observed pattern [see Fig. 9 of Adler et al. (2003) for reference]. As for the tropical
366
precipitation change induced by the El-Niño-like SST warming, it is found that its spatial
367
distribution indeed resembles the typical precipitation anomalies associated with the warm
368
phase of ENSO (Dai and Wigley 2000), particularly with large increases over the central to
369
eastern equatorial Pacific Ocean and decreases over the maritime continent (Figs. 7b and d).
15
370
Thus, the dynamics underlying the PNA wave train driven by ENSO is likely to operate on the
371
El-Niño-like SST warming modifying Aleutian Low. Furthermore, such SST warming pattern
372
is found to significantly affect precipitation change in the midlatitudes: a substantial decrease
373
of 20%-60% change over the coastal Asia and Bering Sea is accompanied by an increase of
374
20%-50% change over the eastern North Pacific, relative to the response to global uniform
375
warming (Fig. 7d). This could be understandable in the context of moisture flux changes driven
376
by the intensified basin-scale cyclonic flow associated with the deepening of the Aleutian Low.
377
Comparing Fig. 7b and c, precipitation change over the United States is also identified as a
378
greater decrease of 10% change and increase of 20% change over the western and central
379
region, respectively.
380
5. Summary and discussion
381
Based on analyses of CMIP5 multi-model historical simulations and RCP8.5 projections,
382
we find a robust northward intensification of the Aleutian low in response to greenhouse
383
warming, manifested in the NPI significantly decreased by approximately -1.3 hPa or
384
approximately -0.4 hPa per 1°C of global surface warming, which is 62% larger than the
385
estimated internal variability of NPI, and the central low-pressure region (i.e., the area of 999.0
386
hPa) expanded about 7 times as large as that in the 20th century. Spatially, the low-pressure
387
system to north of 45°N except over the western coast of North American displays a highly
388
robust intensification, with the maximum decrease of SLP reaching -5.5 hPa over the northern
389
Bering Sea, which is about 3.5 times greater than the internal variability. Results of the
390
idealized experiments further demonstrate that the deepening of the Aleutian Low can be
391
driven by an El-Niño-like warming in the tropical Pacific Ocean, with enhanced SST warming
392
in the eastern equatorial Pacific, which overshadows the effect of the weakened LOTC that
393
tends to dampen the Aleutian Low under greenhouse forcing. Salathé (2006) noted that
394
changes in the Aleutian Low under greenhouse warming are consistent with the northward shift
16
395
and intensification of storm tracks in the midlatitudes, which indicates an active role of storm
396
tracks in modulating Aleutian Low changes. However, the cause-and-effect relationship
397
between both seems hard to establish due to the eddy-mean flow interaction. Thus as a remote
398
forcing, the El-Niño-like SST warming in the tropical Pacific is of great importance to the
399
intensification of the Aleutian Low under greenhouse warming.
400
Although most CMIP5 models predict the deepened Aleutian low, the amplitude of its
401
deepening is subject to inter-model diversity and even a few models predict the opposite. We
402
examined the dominant pattern accounting for the large inter-model differences of the winter
403
SLP changes over the North Pacific by performing an inter-model Empirical Orthogonal
404
Function (EOF) analysis for 22 CMIP5 models. The first EOF mode, explaining 56% of total
405
variance, display a basin-scale pattern of the same polarity with the center of 2.4 hPa located
406
south of the Aleutian Islands (Fig. 8a). The normalized first principal component (PC1) is
407
highly correlated with the NPI changes among individual models with correlation coefficient
408
of -0.97 (Fig. 8b). Three models of highest (lowest) PC1 values that correspond to the
409
decreased (increased) NPI changes can be identified as CanESM2, CNRM-CM5, INM-CM4
410
(FGOALS-g2, IPSL-CM5A-MR, NorESM1-M). We further performed a composite analysis
411
of inter-model differences of tropical Pacific SST changes based on the above two groups of
412
models. Figure 9 shows that the models with strong intensification of the Aleutian Low
413
response feature a much more El-Niño-like SST warming in the equator region than the models
414
in opposite. This suggests that the tropical Pacific SST response to greenhouse warming is a
415
strong source of uncertainty in future projections of the Aleutian Low change. By analyzing the
416
wintertime midlatitude jet stream in the 21st-century projections of 17 CMIP3 models,
417
Delcambre et al. (2013) also found that ENSO-like mean winter SST changes explain 30% of
418
intermodel variance of upper-level zonal winds. Here the relative contributions of forcing
419
uncertainty, model uncertainty and internal climate variability to the total uncertainty need to
17
420
be further clarified, however, the latter two sources are suggested to be important based on the
421
CMIP3 models (Oshima et al. 2012).
422
As for the long-term trend of NPI in the 20th century, we find that it has decreased by
423
approximately -1.5 hPa on the average of five observations since 1911, which, however, is 2
424
times as large as the NPI trend in the MMEM of the historical simulations that is comparable to
425
the estimated internal variability of NPI. This could be in part due to the model deficiencies in
426
producing a significant El-Niño-like warming trend in the tropical Pacific SST, as inferred
427
from the weak increasing tendency of east-west SST gradient captured by most CMIP5 models
428
(Fig. 10). Other possible reasons include underestimations of the modeled atmospheric
429
teleconnection forced by the natural variability of the tropical Pacific SST, as suggested by
430
Furtado et al. (2011), and variations of midlatitude atmospheric planetary-scale waves, since
431
more than 60% of models undervalue the interannual standard deviation of NPI (recall Table 2).
432
Previous studies also have found that the simulated SLP trends in the Northern Hemisphere are
433
much lower in magnitude than its observed counterpart during the second half of the 20th
434
century, which is suggested to be attributable to an underestimation of the simulated SLP
435
response to external forcing and/or an inherent limitation in hindcasting the internal variability
436
of atmospheric circulation (Gillett et al. 2005; Wallace et al. 2012).
437
On the other hand, our results suggest that long-term changes in the Aleutian Low may be
438
indicative of the tropical Pacific SST changes expected under greenhouse warming. Various
439
historical SST reconstructions display either an El-Niño-like or a La-Niña-like warming
440
pattern in the tropical Pacific over the 20th century (Deser et al. 2010; Vecchi et al. 2008).
441
While it is difficult to distinguish which pattern is correct from the limited SST historical
442
reconstructions, the intensification of the Aleutian Low, which is more robust during the 20th
443
century, supports the notion of an El-Niño-like warming in the tropical Pacific Ocean. Previous
444
studies have suggested a long-term weakening of the Walker circulation over the 20th century,
18
445
also favouring an El-Niño-like warming in the tropical Pacific (Held and Soden 2006; Vecchi
446
et al. 2006). However, during recent three decades, the Aleutian Low is weakened significantly,
447
as seen by an increasing trend of NPI with the Sen median slope of 2.5 hPa/35yr (P=0.09) in
448
Fig. 11. This is closely related to a La-Niña-like SST cooling, as the east-west SST gradient
449
anomalies in the equatorial Pacific display a significant decreasing trend with the Sen median
450
slope of -1.0 °C/35yr (P=0.03) and the correlation coefficient between these two time series is
451
-0.56 significant at the 95% confidence level. Kosaka and Xie (2013) also indicated that a
452
decadal (2002-2012) La-Niña-like cooling in the tropical Pacific, concurrent with an
453
intensified Walker circulation, causes a weakened Aleutian Low, which is illustrated as a part
454
of the natural climate variability.
455
Furthermore, recent study indicates that subtropical western boundary currents are
456
accelerating, which leads to the formation of oceanic hotspots over the subarctic frontal zones
457
(Wu et al. 2012). In conjunction with the enhancement of the summertime subtropical Pacific
458
High (Li et al. 2012), the intensification of the wintertime Aleutian Low may provide an
459
important driving mechanism underpinning the acceleration of the subtropical gyre circulation
460
as the climate warms.
461
462
463
Acknowledgments
464
We acknowledge the WCRP's Working Group on Coupled Modelling, which is responsible for
465
CMIP, and we thank the climate modeling groups for producing and making available their
466
model output. This work was supported by China National Global Change Major Research
467
Project (2013CB956201) and China National Science Foundation Key Project (41130859).
19
468
References
469
Adler, R. F., et al., 2003: The version-2 global precipitation climatology project (GPCP) monthly
470
precipitation analysis (1979-present). J. Hydrometeorol., 4, 1147–1167.
471
Alexander, M. A., I. Bladé, M. Newman, J. R. Lanzante, N. C. Lau, and J. D. Scott, 2002: the
472
atmospheric bridge: the influence of ENSO teleconnections on air–sea interaction over the
473
global oceans. J. Climate, 15, 2205-2231.
474
475
Allan, R., and T. Ansell, 2006: A New Globally Complete Monthly Historical Gridded Mean Sea
Level Pressure Dataset (HadSLP2): 1850-2004. J. Climate, 19, 5816–5842.
476
An, S.-I., J.-W. Kim, S.-H. Im, B.-M. Kim, and J.-H. Park, 2012: Recent and future sea surface
477
temperature trends in tropical pacific warm pool and cold tongue regions. Clim. Dyn., 39,
478
1373-1383.
479
480
Benson, A. J., and A. W. Trites, 2002: Ecological effects of regime shifts in the Bering Sea and
eastern North Pacific Ocean. Fish Fish., 3, 95–113.
481
Boer, G. J., G. Flato, and D. Ramsden, 2000: A transient climate change simulation with
482
greenhouse gas and aerosol forcing: projected climate to the twenty-first century. Clim. Dyn.,
483
16, 427-450.
484
485
486
487
488
489
490
491
492
Bonan, G. B., et al., 2002: The land surface climatology of the Community Land Model coupled to
the NCAR Community Climate Model. J. Climate, 15, 3123–3149.
Chavez, F. P., J. Ryan, S. E. Lluch-Cota, C. M. Ñiquen, 2003: From anchovies to sardines and back:
multidecadal change in the Pacific Ocean. Science, 299, 217– 221.
Chiang, J. C. H., Y. Fang, and P. Chang, 2008: Interhemispheric thermal gradient and tropical
Pacific climate. Geophys. Res. Lett., 35, L14704.
Clement, A. C., R. Seager, M. A. Cane, and S. E. Zebiak, 1996: An ocean dynamical thermostat. J.
Climate, 9, 2190–2196.
Collins, M. R., et al., 2013: Long-term Climate Change: Projections, Commitments and
20
493
Irreversibility, in Climate Change 2013: The Physical Science Basis. Contribution of Working
494
Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change,
495
edited by T. F. Stocker et al., chap. 12, pp. 1029–1136, Cambridge Univ. Press, Cambridge, U.
496
K. and New York, USA.
497
498
499
500
501
502
Collins, W. D., et al., 2006: The formulation and atmospheric simulation of the Community
Atmospheric Model: CAM3. J. Climate, 19, 2144–2161.
Compo, G. P., et al., 2011: The Twentieth Century Reanalysis project. Q. J. R. Meteorol. Soc., 137,
1–28.
Dai, A., and T. M. L. Wigley, 2000: Global patterns of ENSO-induced precipitation. Geophys. Res.
Lett., 27, 1283-1286.
503
Delcambre, S. C., D. J. Lorenz, D. J. Vimont, and J. E. Martin, 2013: Diagnosing northern
504
hemisphere jet portrayal in 17 CMIP3 global climate models: twenty-first-century projections.
505
J. Climate, 26, 4930-4946.
506
507
Deser, C, A. S. Phillips, and M. A. Alexander, 2010: Twentieth century tropical sea surface
temperature trends revisited. Geophys. Res. Lett., 37, L10701.
508
Deser, C., A. Phillips, and J. W. Hurrell, 2004: Pacific interdecadal climate variability: Linkages
509
between the Tropics and North Pacific during boreal winter since 1900. J. Climate, 17,
510
3109–3124.
511
512
Deser, C., A. Phillips, V. Bourdette, and H. Teng, 2012: Uncertainty in climate change projections:
the role of internal variability. Clim. Dyn., 38, 527-546.
513
Deser, C., and A. S. Phillips, 2009: Atmospheric circulation trends, 1950–2000: the relative roles
514
of sea surface temperature forcing and direct atmospheric radiative forcing. J. Climate, 22,
515
396–413.
516
517
Fang, Y., 2005: A coupled model study of the remote influence of ENSO on tropical Atlantic SST
variability. Ph.D. thesis, Texas A&M University, 93 pp.
21
518
519
520
521
522
523
524
525
526
527
528
529
530
531
Furtado, J. C., E. D. Lorenzo, N. Schneider, and N. A. Bond, 2011: North pacific decadal
variability and climate change in the IPCC AR4 models. J. Climate, 24, 3049-3067.
Gillett, N. P., F. W. Zwiers, A. J. Weaver, and P. A. Stott, 2003: Detection of human influence on
sea level pressure. Nature, 422, 292–294.
Gillett, N. P., R. J. Allan, and T. J. Ansell, 2005: Detection of external influence on sea level
pressure with a multi-model ensemble. Geophys. Res. Lett., 32, L19714.
Hamed, K. H., and A. R. Rao, 1998: A modified Mann-Kendall trend test for autocorrelated data. J.
Hydrol., 204, 182–196.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming.
J. Climate, 19, 5686–5699.
Hollowed, A. B., S. R. Hare, and W. S. Wooster, 2001: Pacific Basin climate variability and
patterns of northeast Pacific marine fish production. Prog. Oceanogr., 49, 257–282.
Honda, M., S. Yamane, and H. Nakamura, 2005: Impacts of the Aleutian-Icelandic low seesaw on
the surface climate during the twentieth century. J. Climate, 18, 2793–2802.
532
Jia, F., and L. Wu, 2013: A study of response of the equatorial Pacific SST to doubled-CO2 forcing
533
in the coupled CAM-1.5 layer reduced-gravity ocean model. J. Phys. Oceanogr., 43,
534
1288–1300.
535
536
537
538
539
540
541
542
Kenyon, J., and G. C. Hegerl, 2008: Influence of modes of climate variability on global
temperature extremes. J. Climate, 21, 3872–3889.
Kosaka, Y., and S.-P. Xie, 2013: Recent global-warming hiatus tied to equatorial Pacific surface
cooling. Nature, 501, 403–407.
Kwon, Y., and C. Deser, 2007: North Pacific decadal variability in the Community Climate System
Model version 2. J. Climate, 20: 2416–2433.
Latif, M., and T. P. Barnett, 1994: Causes of decadal climate variability over the North Pacific and
North America. Science, 266, 634–637.
22
543
Lehodey P, et al., 2006: Climate variability, fish and fisheries. J. Climate, 19, 5009–5030.
544
Li, W., L. Li, M. Ting, and Y. Liu, 2012: Intensification of Northern Hemisphere subtropical highs
545
546
547
in a warming climate, Nature Geoscience, 5, 830–834.
Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance and its determination by Monte
Carlo techniques. Mon. Wea. Rev., 111, 46–59.
548
Lu, J., R. J. Greatbatch, and K. A. Peterson, 2004: Trend in Northern hemisphere winter
549
atmospheric circulation during the last half of the twentieth century. J. Climate, 17,
550
3745–3760.
551
552
Meehl, G. A., and W. M. Washington, 1996: El Niño-like climate change in a model with increased
atmospheric CO2 concentrations. Nature, 382, 56-60.
553
Oshima, K., Y. Tanimoto, S.-P. Xie, 2012: Regional patterns of wintertime SLP change over the
554
North Pacific and their uncertainty in CMIP3 multi-model projections. J. Meteorol. Soc. Jpn.,
555
90, 385–396.
556
557
558
559
Overland, J. E., and C. H. Pease, 1982: Cyclone climatology of the Bering Sea and its relation to
sea ice extent. Mon. Wea. Rev., 110, 5–13.
Overland, J. E., J. M. Adams, and N. A. Bond, 1999: Decadal variability of the Aleutian Low and
its relation to high-latitude circulation. J. Climate, 12, 1542–1548.
560
Pickart, R. S., G. W. K. Moore, A. M. Macdonald, I. A. Renfrew, J. E. Walsh, and W. S. Kessler,
561
2009: Seasonal evolution of Aleutian low pressure systems: implications for the North Pacific
562
sub-polar circulation. J. Phys. Oceanogr., 39, 1317–1339.
563
564
565
566
567
Salathé, E. P., Jr., 2006: Influences of a shift in North Pacific storm tracks on western North
American precipitation under global warming. Geophys. Res. Lett., 33, L19820.
Schneider, N., and B. D. Cornuelle, 2010: The forcing of the pacific decadal oscillation. J. Climate,
18, 4355-4373.
Seager, R., Y. Kushnir, N. H. Naik, M. A. Cane, and J. Miller, 2001: Wind-driven shifts in the
23
568
latitude of the Kuroshio–Oyashio Extension and generation of SST anomalies on decadal
569
timescales. J. Climate, 14, 4249–4265.
570
571
Sen, P. K., 1968: Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc.,
63, 1379–1389.
572
Smith, T. M., and R. W. Reynolds, 2004: Reconstruction of Monthly Mean Oceanic Sea Level
573
Pressure Based on COADS and Station Data (1854-1997). J. Oceanic. Atmos. Tech., 21,
574
1272–1282.
575
Smith, T. M., Reynolds R. W., T. C. Peterson, and J. Lawrimore, 2008: Improvements to NOAA’s
576
historical merged land–ocean surface temperature analysis (1880–2006). J. Climate, 21,
577
2283–2296.
578
579
580
581
582
583
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment
design. Bull. Amer. Meteor. Soc., 93, 485–498.
Trenberth, K. E., and D. A. Paolino, 1980: The Northern Hemisphere sea level pressure data set:
Trends, errors, and discontinuities. Mon. Wea. Rev., 108, 855–872.
Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific.
Clim. Dyn., 9, 303–319.
584
Trenberth, K. E., et al., 2007: Observations: Surface and atmospheric climate change, in Climate
585
Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth
586
Assessment of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al.,
587
chap. 3, pp. 235–336, Cambridge Univ. Press, Cambridge, U. K.
588
Trenberth, K. E., G. W. Branstator, D. Karoly, A. Kumar, N. Lau, and C. Ropelewski, 1998:
589
Progress during TOGA in understanding and modeling global teleconnections associated with
590
tropical sea surface temperatures. J. Geophys. Res., 103, 14291–14324.
591
592
Vecchi, G. A., A. Clement, and B. J. Soden, 2008: Examining the tropical Pacific's response to
global warming. Eos Trans AGU, 89, 81–83.
24
593
Vecchi, G. A., B. J. Soden, A. T. Wittenberg, I. M. Held, L. Ants, and M. J. Harrison, 2006:
594
Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature,
595
441, 73–76.
596
Wallace, J. M., Q. Fu, B. V. Smoliak, P. Lin, and C. M. Johanson, 2012: Simulated versus observed
597
patterns of warming over the extratropical Northern Hemisphere continents during the cold
598
season. Proc. Natl. Acad. Sci. USA, 109, 14337–14342.
599
600
601
602
603
604
605
Wu, L., et al., 2012: Enhanced warming over the global subtropical western boundary currents.
Nature Clim. Change, 2, 161–166.
Wyllie-Echeverria, T., and W. S. Wooster, 1998: Year-to-year variations in Bering Sea ice cover
and some consequences for fish distributions. Fish Oceanogr. 7, 159–170.
Yeh, S.-W., Y.-G. Ham, and J.-Y. Lee, 2012: Changes in the tropical pacific SST trend from CMIP3
to CMIP5 and its implication of ENSO. J. Climate, 25, 7764-7771.
Zebiak, S. E., and M. A. Cane, 1987: A model ENSO. Mon. Wea. Rev., 115, 2262–2278.
606
25
607
608
Table 1. List of 22 CMIP5 models analyzed in this study.
Model Name
Institute
Country
Commonwealth Scientific and Industrial
Research Organization/Bureau of Meteorology
Beijing Climate Center, China Meteorological
Administration
Canadian Centre for Climate Modelling and
Analysis
Australia
CCSM4
National Center for Atmospheric Research
United States
CNRM-CM5
Météo-France/Centre National de Recherches
Météorologiques
Commonwealth Scientific and Industrial
Research Organization/Queensland Climate
Change Centre of Excellence
Institute of Atmospheric Physics, Chinese
Academy of Sciences
France
Australia
National Oceanic and Atmospheric
Administration/Geophysical Fluid Dynamics
Laboratory
United States
National Aeronautics and Space
Administration/Goddard Institute for Space
Studies
United States
Met Office Hadley Centre
United
Kingdom
Institute for Numerical Mathematics
Russia
Institute Pierre Simon Laplace
France
ACCESS1.0
BCC-CSM1.1
CanESM2
CSIRO-Mk3.6.0
FGOALS-g2
China
Canada
China
GFDL-CM3
GFDL-ESM2G
GFDL-ESM2M
GISS-E2-H
GISS-E2-R
HadGEM2-CC
HadGEM2-ES
INM-CM4
IPSL-CM5A-MR
IPSL-CM5B-LR
MIROC-ESM
University of Tokyo, Atmosphere and Ocean
Research Institute; National Institute for
Japan
Environmental Studies; Japan Agency for Marine
Earth Science and Technology
MPI-ESM-LR
Max Planck Institute for Meteorology
Germany
MRI-CGCM3
Meteorological Research Institute
Japan
NorESM1-M
Norwegian Climate Centre
Norway
MPI-ESM-MR
26
609
Table 2. Normalized standard deviation (dimensionless) and estimated internal variability of
610
NPI in 22 CMIP5 models. Here the interannual standard deviation of NPI in each model was
611
normalized by the observed counterpart averaged over the NPIs in Fig.1 (i.e., 3.04 hPa).
Normalized
STD
0.96
Internal
variability (hPa)
BCC-CSM1.1
1.00
0.76
CanESM2
1.12
0.81
CCSM4
1.50
0.68
CNRM-CM5
0.83
0.67
CSIRO-Mk3.6.0
1.05
0.87
FGOALS-g2
0.89
0.93
GFDL-CM3
0.93
1.11
GFDL-ESM2G
1.19
0.91
GFDL-ESM2M
1.28
1.11
GISS-E2-H
0.82
0.72
GISS-E2-R
0.83
0.76
HadGEM2-CC
0.86
0.68
HadGEM2-ES
0.78
1.18
INM-CM4
0.73
0.73
IPSL-CM5A-MR
1.24
1.08
IPSL-CM5B-LR
0.81
0.70
MIROC-ESM
0.82
0.67
MPI-ESM-LR
0.90
0.75
MPI-ESM-MR
0.87
0.72
MRI-CGCM3
0.80
0.89
NorESM1-M
1.37
0.80
Model Name
ACCESS1.0
0.82
612
27
(a)
(b)
(c)
613
614
FIG. 1. (a) The climatological mean of winter (DJF) SLP from 1950 to 1999 based on the 20CRv2
615
reanalysis data. The cross marks the minimum SLP of 998.6 hPa and the dashed line denotes the isobar
616
of 999.0 hPa. (b) The observed NPI relative to the mean over the period of 1950-1999 derived from five
617
historical SLP datasets, except for 1950-1996 used for the ERSLP dataset. The individual long-term
618
trend calculated from 1911 to the end of the available SLP data is indicated by dashed line. (c) The
619
centennial trends and corresponding confidence levels of the observed NPIs in historical datasets,
620
estimated by the Sen median slope and the modified Mann-Kendall trend test, respectively. Note that
621
the negative NPI indicates a relatively stronger Aleutian Low.
622
28
623
(a)
(b)
(c)
624
625
FIG. 2. (a) As in Fig. 1a, but for the CMIP5 MMEM of winter SLP in the historical run. The cross marks
626
the minimum SLP of 998.8 hPa and the dashed line denotes the isobar of 999.0 hPa. (b) The MMEM of
627
NPI in the historical run (dot-solid line) and the inter-model spread (shading) estimated as the
628
inter-model standard deviation, superimposed on the long-term trend (dashed line), with respect to the
629
1950-1999 reference period. (c) As in Fig. 1c, but for the simulated NPIs in 22 CMIP5 models as well as
630
the MMEM time series of NPI, estimated on 1900-2004.
631
29
632
(a)
(b)
(c)
(d)
633
634
FIG. 3. (a) NPI changes in 22 CMIP5 models between the 50-yr RCP8.5 run (2050-2099) and historical
635
run (1950-1999). Asterisks (error bar) denote the 95% confidence level (interval) based on a two-tailed
636
Student’s t-test. (b) The MMEM of winter (DJF) SLP in the RCP8.5 run. The cross marks the minimum
637
SLP of 995.5 hPa and the dashed line denotes the isobar of 997.0 and 999.0 hPa. (c) The estimated
638
internal variability of winter (DJF) SLP based on the CMIP5 pre-industrial control experiments. (d) The
639
MMEM of winter SLP change between the 50-yr RCP8.5 run and historical run. Stippling indicates
640
regions where the MMEM change is greater than two standard deviations of internal variability and
641
where at least 85% (19/22) of the models agree on the sign of change. Hatching indicates regions where
642
the MMEM change is in the range of one to two standard deviations of internal variability. The rest are
643
regions where the MMEM change is less than one standard deviation of internal variability.
644
30
645
646
647
FIG. 4. Scatter diagram of changes in the NPI versus changes in the winter (DJF) LOTC over the North
648
Pacific Ocean and the Asian continent in the CMIP5 models. LOTC and NPI are derived from the
649
50-yr-interval difference between RCP8.5 run (2050-2099) and historical run (1950-1999). A negative
650
651
LOTC and
NPI indicate a weakening of the LOTC and a strengthening of the Aleutian Low,
respectively. The solid line denotes the linear regression.
652
31
653
(a)
(b)
654
655
FIG. 5. (a) The MMEM of the tropical Pacific SST change between the 50-yr RCP8.5 run (2050-2099)
656
and historical run (1950-1999) in winter (DJF). Note that all models show significant change. (b)
657
Scatter diagram of changes in the NPI versus changes in the equatorial Pacific east-west SST gradient
658
between the 50-yr RCP8.5 run and historical run. A positive value of the SST difference indicates an
659
El-Niño-like response in a warmer climate. The solid line denotes the linear regression.
32
660
(a)
(b)
(c)
(d)
(e)
661
662
FIG. 6. (a) The climatological-mean tropical Pacific SST change between the winter (DJF)
663
equilibrium-states (the last 50-yr) of the CAM3.1-RGO-4CO2 and CAM3.1-RGO-CTRL. Note that
664
SST change at each grid is significant at the 95% confidence level based on a two-tailed Student’s t-test.
665
The climatological-mean winter SLP changes based on the equilibrium-states (the last 50-yr) difference
666
in terms of (b) CAM3.1-4CO2 minus CAM3.1-CTRL and (c) CAM3.1-4CO2-GUW minus
667
CAM3.1-CTRL. Stippling indicates regions where SLP changes significant at the 95% confidence level.
668
The corresponding 50-yr changes in the winter (d) NPI and (e) LOTC. Blue and red bars represent
669
CAM3.1-4CO2 minus CAM3.1-CTRL and CAM3.1-4CO2-GUW minus CAM3.1-CTRL, respectively.
670
Asterisks denote the 95% confidence level.
671
33
672
(a)
(b)
(c)
(d)
673
674
FIG. 7. (a) The winter (DJF) precipitation pattern simulated in the CAM3.1-CTRL run. (b) The
675
climatological-mean (the last 50-yr) winter precipitation changes in percentage for CAM3.1-4CO2
676
relative to CAM3.1-CTRL. (c) As in (b), but for CAM3.1-4CO2-GUW. (d) The difference between (b)
677
and (c). Contour interval in (b-d) is 20% change. Stippling indicates regions where precipitation
678
changes significant at the 95% confidence level.
679
34
680
(a)
(b)
681
682
FIG. 8. (a) The first inter-model EOF mode of the winter (DJF) SLP change (hPa) between the 50-yr
683
RCP8.5 run (2050-2099) and historical runs (1950-1999). This pattern is corresponding to one standard
684
deviation of the inter-model spread, i.e., the normalized PC1 shown as blue line in (b). The yellow line
685
in (b) indicates the NPI change between the 50-yr RCP8.5 run and historical run.
686
35
687
(a)
(b)
(c)
688
689
FIG. 9. (a) Composite of the tropical Pacific SST change between the 50-yr RCP8.5 run (2050-2099)
690
and historical run (1950-1999) in winter, corresponding to the three models of highest PC1 values
691
shown in Fig. 8b. (b) Same as (a), but for the three models of lowest PC1 values. (c) The difference
692
between (a) and (b).
693
36
694
695
696
FIG. 10. The centennial trends of the equatorial Pacific east-west SST gradient anomalies (i.e., EEP
697
minus WEP SST anomalies) in winter (DJF) in the historical run of 22 CMIP5 models. The trend and its
698
statistical significance are estimated with the Sen median slope and the modified Mann-Kendall trend,
699
respectively.
700
37
701
702
703
FIG. 11. The NPI (blue solid line) and the equatorial Pacific east-west SST gradient anomalies (i.e.,
704
EEP minus WEP SST anomalies; red solid line) in winter (DJF) during 1979-2013. The blue (red)
705
dashed line denotes long-term trend of NPI (east-west SST gradient). SST is taken from the Extended
706
Reconstructed Sea Surface Temperature version 3b dataset (Smith et al. 2008).
38