Embed Size (px)
Citation preview
1
Pacific Sea Surface Temperature and the Winter of 2014 1
2
Dennis L. Hartmann1 3
4
5
Corresponding Author: Dennis L. Hartmann, Department of Atmospheric Sciences, 6
Box 351640, University of Washington, Seattle, Washington 98195 7
8
Key Points: 9
• SST anomalies in the Pacific were a primary cause of the severe winter of 2014. 10
• Warm SST in the Northeast Pacific gives cold conditions in central North America. 11
• Fixed SST model experiments replicate observed teleconnections. 12
13
1 Department of Atmospheric Sciences, University of Washington, Seattle, Washington 98195
2
14
Abstract It is shown from historical data and from modeling experiments that a 15
proximate cause of the cold winter in North America in 2013-‐14 was the pattern of 16
sea surface temperature (SST) in the Pacific Ocean. Each of the three dominant 17
modes of SST variability in the Pacific is connected to the Tropics and has a strong 18
expression in extratropical SST and weather patterns. Beginning in the middle of 19
2013 the third mode of SST variability was two standard deviations positive and 20
remained so through the winter of 2013-‐14. This pattern is associated with high 21
pressure in the northeast Pacific and low pressure and low surface temperatures 22
over central North America. A large ensemble of model experiments with observed 23
SSTs confirms that SST anomalies contributed to the anomalous winter of 2014. 24
25
Index Terms: 3305 Climate change and variability; 3339 Ocean/Atmosphere 26
Interactions; 4513 Decadal Ocean Variability; 4572 Upper Ocean and Mixed-‐Layer 27
Processes 28
Keywords: Pacific Decadal Oscillation, North Pacific Mode, El Niño 29
30
3
1. Introduction 31
Much interest has focused on the very cold winter in the central and eastern US and 32
Canada in 2013-‐14 (hereafter the winter of 2014). It has been suggested that 33
melting of polar sea ice could be responsible for causing unusual weather events 34
([Francis and Vavrus, 2012] but see [Barnes, 2013]), perhaps through the 35
intermediacy of stratospheric warmings [Kim et al., 2014]. This would be 36
remarkable given the small area of the Arctic Ocean and its presence at the tail end 37
of the atmospheric and oceanic chain moving energy from the Tropics toward the 38
poles. Taking a different perspective, Ding et al. [2014] use models and 39
observations to show that at least some of the recent Arctic warming has been 40
forced by changes in the tropical Pacific. Here the coupled variability between the 41
Pacific Ocean and the global atmosphere is used to account in part for the unusual 42
winter of 2014. Using a combination of analysis of past data and a large ensemble of 43
fixed SST modeling experiments, a strong case can be made that variability of the 44
ocean-‐atmosphere system originating in the Tropics is the most likely “cause” of the 45
winter of 2014 anomalies. 46
47
The ocean provides the long-‐term memory for the climate system through its higher 48
heat capacity, slowly varying large-‐scale modes and long adjustment time to 49
changes in forcing. The El Niño Southern Oscillation (ENSO) phenomenon is a 50
coupled ocean-‐atmosphere phenomenon that is the dominant mode of interannual 51
variability globally, and has provided the opportunity to make useful forecasts of 52
seasonal weather and climate conditions a season or more in advance [Neelin et al., 53
4
1998; Rasmusson and Wallace, 1983]. ENSO has a related decadal variability, the 54
Pacific Decadal Oscillation (PDO), that was first defined in terms of North Pacific sea 55
surface temperature (SST) anomalies [Mantua and Hare, 2002; Zhang et al., 1997], 56
and is now thought to influence global mean temperature on decadal time scales 57
[Huber and Knutti, 2014; Kosaka and Xie, 2013; Trenberth and Fasullo, 2013]. 58
Interannual anomalies of SST are known to have a significant influence on climate 59
over land, and these relationships play an important role in seasonal climate 60
forecasting [Palmer and Anderson, 1994]. While the connection between tropical 61
SST anomalies and winter weather in the extratropics is well established, SST 62
anomalies in middle latitudes are driven by atmospheric anomalies and it is more 63
difficult to show an influence of extratropical SST anomalies in driving atmospheric 64
anomalies [Kushnir et al., 2002]. The extratropical SST anomalies associated with 65
ENSO are caused primarily by atmospheric anomalies propagating from the Tropics 66
[Alexander et al., 2002]. 67
68
In this paper we explore anomalies in monthly mean SST in the Pacific and their 69
relation to recent weather anomalies in an observational context. We first present a 70
decomposition of Pacific Ocean SST variability into three key modes, ENSO, which 71
has its strongest variability on time scales of 2 to 6 years, and two decadal modes 72
that are related to ENSO, but have a strong expression in the North Pacific, perhaps 73
because of their strong interaction with the winds there, and possibly also because 74
of the interaction with Kuroshio and Oyashio current systems. We next relate these 75
SST patterns to atmospheric variability and show that one of the decadal modes is 76
5
associated with atmospheric anomalies that look very much like those during the 77
winter of 2014. This mode of variability showed a positive anomaly of two standard 78
deviations beginning in the summer of 2013. To demonstrate a causality 79
relationship between this large anomaly in the SST mode and the winter of 2014, we 80
analyze a large ensemble of fixed-‐SST modeling experiments. Since these fixed SST 81
experiments produce anomalies with similar structure and amplitude to the 82
observed anomalies in 2014, we conclude that the SST variations played a key role 83
in producing the cold winter of 2014. 84
85
2. SST Variability: ENSO, PDO and Northeast Pacific Mode 86
87
The Pacific Decadal Oscillation (PDO) has classically been defined as the first 88
Empirical Orthogonal Function (EOF) of the anomalies of monthly mean SST 89
poleward of 20N in the Pacific Ocean [Mantua et al., 1997]. Here include the ocean 90
region from 30S-‐65N and 120E-‐105W, which includes the tropical and North Pacific. 91
We use the NOAA ERSST data set [Smith et al., 2008] to compute the EOF 92
decomposition after removing the seasonal cycle, the global mean and the linear 93
trend from the monthly data. We have obtained similar results with the HadISST 94
data set [Rayner et al., 2003] (not shown). Although the data set starts in 1854, we 95
focus on the period since 1900 when the data are more complete. The first EOF 96
expresses about 30% of the variance and two more modes that express significant 97
amounts of the variance by the North et al. [1982] criterion (S1). The second and 98
third modes are not distinguished from each other by their explained variance, and 99
6
their relative rankings will change for different sampling periods. For example, the 100
third mode for the period from 1900 becomes the second mode in the period since 101
1979. The remaining EOFs form a continuum of modes that are not distinguished 102
from each other by their explained variance. The results shown here are for all 12 103
months, but the basic structures are the same if winter or summer half years are 104
used instead. The structures and explained variances are very similar if we use 105
instead only the period since 1950, when the observational network had improved 106
further. The first three SST modes are shown in Fig. 1 as regressions of the SST 107
anomalies with the principal component time series of each mode. The first mode 108
captures much of the equatorial variance, but the second and third modes also have 109
significant connections to the Tropics. 110
111
In the supplementary materials we show that these three modes are related to each 112
other (S2). The second mode tends to become strongly positive after a warm ENSO 113
event. This expresses the fact that the extratropical signature of a warm event, 114
which gives cold SST in midlatitudes via the “atmospheric bridge”, tends to persist 115
longer than ENSO itself. So its structure is cold in midlatitude western Pacific and 116
also cold in the eastern equatorial Pacific. The third mode tends to be in its positive 117
phase prior to ENSO warm events and is closely related to the “footprinting” 118
mechanism [Vimont et al., 2001; Vimont et al., 2003]. During the satellite era since 119
1979, the third mode has been very strong, and it enters as the second mode of 120
global SST during that period (not shown). In the classical analysis of the PDO using 121
the region poleward of 20N [Mantua and Hare, 2002], the first mode is that 122
7
identified as the PDO, while the second mode is very similar to the third mode for 123
the larger domains employed here. Deser and Blackmon [1995] referred to the 124
second mode SST structure as the North Pacific Mode (NPM). The NPM mode is 125
very robust to the region used, appearing as the second mode for the Pacific region 126
north of 20N and one of two modes centered in the North Pacific for the region 127
north of 30S. 128
129
The PDO has a fairly strong connection to the tropical SST signal of ENSO, while the 130
NPM has much weaker connections to the equatorial region [Deser and Blackmon, 131
1995]. Both have very red temporal spectra and exhibit decadal variations. Bond et 132
al. [2003] noted that their second EOF of North Pacific SST (very similar to the NPM 133
in Fig. 1) was anomalously negative in 1999-‐2002, and emphasized that the PDO 134
alone is not sufficient to characterize the variability of the North Pacific. 135
136
The second and third modes in Fig. 1 may be related to the North Pacific Gyre 137
Oscillation [Ceballos et al., 2009; Chhak et al., 2009; Di Lorenzo et al., 2008] and to 138
fluctuations in the Kuroshio and Oyashio Currents [Frankignoul et al., 2011; Kwon et 139
al., 2010]. Both advection by ocean currents and propagation of Rossby waves are 140
slow in the North Pacific, so that dynamical anomalies can appear nearly stationary, 141
but an interpretation in terms of weather anomalies changing the heat content over 142
relatively large regions may be as applicable as an interpretation in terms of ocean 143
dynamics. Ultimately, both atmospheric forcing and the oceanic dynamic response 144
are important for the evolution of the North Pacific SST and surface height, however 145
8
(e.g. Miller et al. [1998], Vivier et al. [1999], Qiu [2000], Cummins and Freeland 146
[2007]). Local atmospheric forcing seems to be very important in the North Pacific 147
at least on seasonal and interannual time scales, and both local and remote 148
influences may be important [Alexander et al., 2002; Liu and Alexander, 2007]. 149
Smirnov et al [2014] conclude that internal ocean dynamics play little role in 150
interannual SST variability in the eastern Pacific in comparison to the western 151
Pacific where the Kuroshio and Oyashio currents may play a more important role. 152
153
The principal components of the first three EOFs are shown in Fig. 2 in units of 154
standard deviations for the period from 1979 to August 2014. These EOFs are 155
based on the Pacific north of 30S during the period from 1900 to 2014. During the 156
most recent period shown in Fig. 2, hints of a relationship between the three EOFs 157
during warming events can be seen. Both the second and third modes show large 158
amplitude variations associated with major warming events. Since the 1998 warm 159
ENSO event the principal components of the first two EOFs have been mostly 160
negative, which is a reflection of the tendency for cool ENSO and negative PDO since 161
then. After May 2013, however, the first two EOFs were near neutral amplitude, 162
while the third EOF (NPM) maintained a positive value near two standard 163
deviations above neutral. Unusually warm temperatures in the northeast Pacific 164
prevailed during this period. 165
166
The very high SST anomaly in the North Pacific that emerged in May of 2013 is 167
linked to reduced cooling of the ocean during the prior winter of 2013. The SST, 168
9
heat content, surface wind stress and surface heat flux anomalies for 2013 are well 169
illustrated in the State of the Climate – 2013 report [Blunden and Arndt, 2014], 170
particularly in Figs. 3.1 to 3.9 [Newlin and Gregg, 2014]. These show strongly 171
suppressed wind stresses, and positive anomalies of surface heat fluxes into the 172
North Pacific Ocean during 2013. It is important to better understand what caused 173
these anomalous fluxes in 2013 and to what extent the heat stored then influenced 174
the weather of the following winter. 175
176
3. Meteorological connections to Pacific SST anomalies 177
178
We have regressed the monthly principal components of the first three EOFs of SST 179
on the 500hPa height and lowest model level temperature from both the 20th 180
Century Reanalysis that starts in 1871 [Compo et al., 2011] and the NCEP/NCAR 181
Reanalysis that starts in 1948 [Kalnay et al., 1996]. We obtain similar results from 182
both reanalysis data sets and from shorter subsets of each data set. For the sake of 183
brevity we show the results for the NCEP/NCAR Reanalysis. 184
185
Since the focus of this paper is the influence of the third EOF on the winter of 2014, 186
we will show only the regression for that mode here, but correlation patterns for all 187
three modes are shown in the supplementary materials (S3). The top panel in Fig. 188
3 shows the anomaly of the 500hPa field for November to March of 2013-‐14, the 189
winter of 2014. It shows a ridge over the North Pacific and a deep trough over the 190
central North American centered just south of Hudson Bay. The northerly flow 191
10
between these two centers, which is roughly aligned with the Rocky Mountains, 192
brought cold Arctic air into central North America and gave unusually cold 193
conditions there. 194
195
The second panel in Fig. 3 shows the regression of the global 500hPa height field 196
onto the principal component of EOF3 (NPM) during months of November through 197
March based on the SST EOF from 1900 and the NCEP/NCAR height field starting in 198
1948. It shows the amplitude explained by a one standard deviation variation of the 199
principal component. The structure features a large high pressure over the 200
northern Pacific centered near the Aleutian Islands with a downstream low 201
centered near Hudson Bay. For comparison, the bottom panel of Fig. 3 shows the 202
regression of the global SST mode based on the period from 1979-‐2014 regressed 203
on the NCEP/NCAR 500hPa height anomalies for the same period. The same pattern 204
appears, but the low over Hudson Bay is better developed. Since the PC of NPM had 205
a two standard deviation excursion in 2014, we should expect to see twice this 206
amplitude then. The observed composite has low height anomaly over Hudson Bay 207
of about 80 meters, while the regression in the bottom panel has a low height 208
anomaly of about 18 meters, or 36 meters for a two-‐sigma event. The regression is 209
thus off by about a factor of two in predicting the actual low anomaly over North 210
America. It is to be expected that a regression over a long time would 211
underestimate the anomaly of the winter of 2014, for which both the forcing from 212
the ocean and the natural variability of the midlatitude atmosphere must have 213
aligned to create a rare event. Monthly correlations between the height field and the 214
11
SST mode show stronger correlations when the height field precedes the SST 215
anomaly than vice versa (not shown). This is indicative of the strong role of 216
atmospheric variability driving the SST, but model experiments described in section 217
4 will show nonetheless that the SST does have an effect on the atmospheric 218
variability. 219
220
The statistical significance of the features in the regression plots shown here has 221
been assessed by determining whether the correlations associated with them are 222
different from zero at the 95% level. The degrees of freedom have been determined 223
using the approximation recommended by Bretherton et al. [1999]. Although the 224
SST indices are highly auto-‐correlated, the extratropical height anomalies are not, so 225
that for the 67-‐year record from 1948 and the 5 months of November to March, the 226
data used to perform the height regressions possesses about 250 degrees of 227
freedom, which is a large number. With this number of degrees of freedom, 228
relatively small correlations become significant, and the question shifts to whether 229
the fraction of variance explained is interesting or not. Thus the absolute values of 230
the correlation coefficients are of more interest than their proven statistical 231
significance. The correlation coefficient map corresponding to Fig. 3a has 232
correlations exceeding 0.4 for the two centers over the Pacific, but only exceeding 233
0.2 for the center over North America (S3). If we use the global SST mode or the 234
NPM mode for the period starting in 1979, then the centers are more similar, with 235
correlations exceeding 0.3 for all the major centers. The connection between 236
positive NPM and cold winters in eastern North America is stronger in the more 237
12
recent record. For comparison, the correlations associated with the ENSO mode 238
exceed 0.6 in the extratropics for SST EOFs computed from both the global and 239
Pacific regions. We therefore conclude that the winter of 2014 over North America 240
was favored by the prevailing SST anomaly pattern over the Pacific Ocean, but that 241
either the internal natural variability of the atmosphere, or perhaps a third actor, 242
also contributed. 243
244
Similar analysis suggests that warm North Pacific SST anomalies also affect summer 245
temperatures (see supplementary material (S4)), leading to cooler than normal 246
summer temperatures (NOAA Climate Prediction Center web page indicates cool 247
and wet conditions occurred in central North America during the summer of 2014). 248
Thus the warm anomaly in the NE Pacific could also have been a contributor to the 249
cool and moist summer of 2014 in the middle of North America. 250
251
Barlow et al. [2001] investigated the relationship between Pacific SST variations and 252
summertime hydrology over North America. Their analysis differs from ours in a 253
couple of respects; they used Pacific Ocean poleward 20S, instead of 30S, and they 254
rotated their EOFs before regression with North American Climate. They labeled 255
their first three modes an ENSO mode, a PDO mode and a North Pacific Mode. Their 256
North Pacific Mode is mostly the cold SST centered in the central and western 257
Pacific that is also seen weakly as part of the ENSO mode. The northwest Pacific 258
part of the ENSO EOF was reduced through rotation and this variability was 259
13
concentrated on their NPM. Their NPM mode is thus centered at about 40N and 260
150W and extends more strongly to the west than the NPM in Fig. 1. 261
262
Wang et al. [2013] explore the influence of SST modes on North American weather. 263
They estimate that the three leading modes they consider explain about 50% of the 264
total effect of SST on North American Climate, but the modes they considered in the 265
Pacific correspond roughly to the ENSO mode and the PDO mode (or Kuroshio 266
extension mode) and they do not consider the NPM described here. Since our 267
analysis shows that the second and third modes explain the same amount of 268
variance, any linear combination of the two is as valid a representation as either 269
mode alone. Since the third mode reaches large amplitude during the period of 270
interest, while the first and second are near neutral, it is meaningful to consider only 271
the third mode in detail here. 272
273
Wang et al [2014] associated the California drought of 2013-‐14 with an the anomaly 274
height pattern for NDJ, which is very similar in structure to Fig. 3 (top). They 275
focused particularly on the dipole pattern between the ridge in the Gulf of Alaska 276
and the trough centered just south of Hudson Bay. They found that this pattern was 277
correlated with the Niño4 SST index one year prior. In general outline this is very 278
consistent with our conclusions here, that the winter of 2014 pattern is associated 279
with a precursor mode of ENSO. They go on to conclude that this pattern is 280
becoming more prevalent in consequence of human-‐induced warming of the climate. 281
We have not investigated any connections to climate change in this paper, but in the 282
14
next section we will use models to show that the anomaly pattern over North 283
America can be driven by SST anomalies in models. 284
285
4. Fixed SST Experiments 286
So far it has been demonstrated that the structure that dominated the winter 287
weather in North America in 2014 is similar in shape and within a factor of two in 288
amplitude to the regression onto the SST pattern that dominated in that year. To 289
make a stronger case that the SST anomaly caused some important part of the 290
weather anomaly, rather than merely being associated with it, we turn to AMIP-‐style 291
experiment data provided by the NOAA-‐ESRL Physical Sciences Division, Boulder, 292
Colorado from their website at 293
http://www.esrl.noaa.gov/psd/repository/alias/facts/ . The experiments chosen 294
apply the observed radiative forcing and fix the observed SSTs under an 295
atmospheric model. Large ensembles of experiments starting in 1979 are 296
conducted with several different models of the atmosphere. Three model 297
ensembles are considered here; a 50-‐member ensemble with the ESRL-‐GFSv2 model, 298
a 30-‐member ensemble with the ECHAM5 model, and a 20-‐member ensemble with 299
the CAM4 model. We use the ensemble mean anomalies for each model to both 300
regress them onto the SST indices, and to compute the average anomalies for the 301
winter of 2014. To be consistent with our data analysis we remove the linear trend. 302
303
Figure 4 shows the ensemble mean 500hPa height anomaly for the winter of 2014 304
and the regression of the 500hPa height anomalies onto the NPM mode for the 305
15
ESRL-‐GFSv2 50-‐member ensemble mean. The predicted anomaly has similar key 306
features to the observed anomaly in Fig. 3a, with smaller amplitudes. The depth of 307
the low at about 35N in the central Pacific is well simulated, but the downstream 308
features over North America in the model ensemble mean are much attenuated 309
compared to the observed pattern. In future work we could look to see if a 310
significant number of the ensemble members do get the correct downstream 311
amplitudes, but that is a considerable data processing chore. The lower panel in Fig. 312
4 can be compared directly with the lower panel in Fig. 3, which shows the 313
regression of the global SST pattern onto the observed height anomalies for the 314
period from 1979 to 2014. The model pattern is similar in structure, but weaker in 315
amplitude compared to the observations. The other two models produce similar 316
results (S5). It is important to remember that fixed SST experiments do not capture 317
the full interaction between the atmosphere and ocean [Barsugli and Battisti, 1998], 318
and simple interpretations of ensembles often underestimate the confidence of 319
seasonal predictions [Eade et al., 2014]. For the 1979 to 2014 period there is little 320
difference between the regression on the global SST pattern, and the NPM pattern 321
for the Pacific north of 30S. Although the winter of 2014 was a two-‐sigma event in 322
terms of the NPM of SST, the regressions with NCEP/NCAR Reanalysis include 67 323
winters and are not very sensitive to the inclusion or not of the winter of 2014. 324
325
16
4. Discussion and Conclusion 326
327
Although many other influences on the weather during a particular month could 328
have contributed to the anomalies during the winter of 2013-‐14, it would appear 329
that the sudden increase in the NPM SST anomaly pattern in the North Pacific during 330
the middle of 2013 can be associated with the subsequent cold winter in the center 331
of North America. The reasons for this increase remain to be explored and its future 332
evolution will be interesting to watch. The stage was set for the very warm SST 333
anomaly of the 2014 winter by a reduction in heat extraction from the North Pacific 334
in the prior winter of 2013. The annual mean 500hPa anomaly shows a large high 335
pressure over the North Pacific centered at about 50N 150W (not shown). The 336
relative roles of surface heat exchange and Ekman transport on this changed heat 337
loss can be estimated in future work. 338
339
Midlatitude seasonal weather and climate anomalies receive a large contribution 340
from internal atmospheric variability that is unrelated to any interactions with the 341
SST (e.g. Deser et al. 2012). The effect of ocean heat anomalies and associated SST 342
anomalies like those shown in Fig. 1 on the seasonal weather patterns can thus be 343
negated by other variability in any particular month, and the SST anomalies 344
themselves are driven in large part by weather variations. Nonetheless, the 345
systematic connection between NPM SST anomalies and winter weather in the past, 346
the extreme amplitude of the SST anomalies in 2013-‐14, and the similarly of the 347
2014 winter anomalies to the historical pattern, all suggest that the warm SST in the 348
17
North Pacific could have made a significant contribution to the cold winter in central 349
North America in 2013-‐14. Fixed SST simulations with three models confirm the 350
causal relationship between the NPM mode and winter weather in North America. 351
352
353
Aknowledgments: The SST data were obtained from the NOAA ERSST data set at 354
http://www.esrl.noaa.gov/psd/data/gridded/data.noaa.ersst.html. The altimeter 355
products were produced by Ssalto/Duacs and distributed by Aviso, with support from 356
Cnes (http://www.aviso.altimetry.fr/duacs/). 20th Century Reanalysis and NCEP/NCAR 357
Reanalysis products were obtained from the NOAA web sites 358
http://www.esrl.noaa.gov/psd/data/20thC_Rean/ and 359
http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html. AMIP simulations 360
from the NOAA web site http://www.esrl.noaa.gov/psd/repository/alias/facts/ 361
were crucial to this analysis. This work was supported by NSF grant AGS-‐0960497. 362
The author gratefully acknowledges colleagues, Nick Bond, Paulo Ceppi, Greg 363
Johnson, Todd Mitchell, LuAnne Thompson, and Mike Wallace, whose generous 364
advice and comments greatly improved this contribution. The comments of two 365
anonymous reviewers of an earlier version of this work were extremely helpful. 366
367
368
References: 369
370
18
Alexander, M. A., I. Blade, M. Newman, J. R. Lanzante, N. C. Lau, and J. D. Scott (2002), 371
The atmospheric bridge: The influence of ENSO teleconnections on air-‐sea 372
interaction over the global oceans, J. Climate, 15(16), 2205-‐2231. 373
Barlow, M., S. Nigam, and E. H. Berbery (2001), ENSO, Pacific decadal variability, and 374
US summertime precipitation, drought, and stream flow, J. Climate, 14(2001), 375
2105-‐2128. 376
Barnes, E. A. (2013), Revisiting the evidence linking Arctic amplification to extreme 377
weather in midlatitudes, Geophys. Rese. Lett., 40(17), 4734-‐4739. 378
Barsugli, J. J., and D. S. Battisti (1998), The basic effects of atmosphere-‐ocean 379
thermal coupling on midlatitude variability, J. Atmos. Sci., 55(4), 477-‐493. 380
Blunden, J., and D. S. Arndt (2014), State of the Climate in 2013, Bull. Amer. Meteorol. 381
Soc., 95(7), S1-‐S257. 382
Bond, N. A., J. E. Overland, M. Spillane, and P. Stabeno (2003), Recent shifts in the 383
state of the North Pacific, Geophys. Res. Lett., 30(23), 384
doi:10.1029/2003gl018597. 385
Bretherton, C. S., M. Widmann, V. P. Dymnikov, J. M. Wallace, and I. Blade (1999), The 386
effective number of spatial degrees of freedom of a time-‐varying field, J. 387
Climate, 12(JUL 1999), 1990-‐2009. 388
Ceballos, L. I., E. Di Lorenzo, C. D. Hoyos, N. Schneider, and B. Taguchi (2009), North 389
Pacific Gyre Oscillation Synchronizes Climate Fluctuations in the Eastern and 390
Western Boundary Systems, J. Climate, 22(19), 5163-‐5174, 391
doi:5110.1175/2009jcli2848.5161. 392
19
Chhak, K. C., E. Di Lorenzo, N. Schneider, and P. F. Cummins (2009), Forcing of Low-‐393
Frequency Ocean Variability in the Northeast Pacific, J. Climate, 22(5), 1255-‐394
1276, doi:1210.1175/2008jcli2639.1251. 395
Compo, G. P., et al. (2011), The Twentieth Century Reanalysis Project, Quart. J. Royal 396
Meteorol. Soc., 137(654), 1-‐28, doi; 10.1002/qj.1776. 397
Cummins, P. F., and H. J. Freeland (2007), Variability of the North Pacific current and 398
its bifurcation, Progr. Oceanography, 75(2), 253-‐265, doi: 399
210.1016/j.pocean.2007.1008.1006. 400
Deser, C., and M. L. Blackmon (1995), On the relationship between tropical and 401
North Pacific Sea-‐Surface Temperature Variations, J. Climate, 8(6), 1677-‐402
1680. 403
Deser, C., A. Phillips, V. Bourdette, and H. Y. Teng (2012), Uncertainty in climate 404
change projections: the role of internal variability, Clim. Dyn., 38(3-‐4), 527-‐405
546, doi:510.1007/s00382-‐00010-‐00977-‐x. 406
Di Lorenzo, E., et al. (2008), North Pacific Gyre Oscillation links ocean climate and 407
ecosystem change, Geophys. Res. Lett., 35(8), doi: 10.1029/2007gl032838. 408
Ding, Q. H., J. M. Wallace, D. S. Battisti, E. J. Steig, A. J. E. Gallant, H. J. Kim, and L. Geng 409
(2014), Tropical forcing of the recent rapid Arctic warming in northeastern 410
Canada and Greenland, Nature, 509(7499), doi 10.1038/nature13260. 411
Eade, R., D. Smith, A. Scaife, E. Wallace, N. Dunstone, L. Hermanson, and N. Robinson 412
(2014), Do seasonal-‐to-‐decadal climate predictions underestimate the 413
predictability of the real world?, Geophys. Res. Lett., 41(15), 5620-‐5628. 414
20
Francis, J. A., and S. J. Vavrus (2012), Evidence linking Arctic amplification to 415
extreme weather in mid-‐latitudes, Geophys. Res. Lett., 39. 416
Frankignoul, C., N. Sennechael, Y. O. Kwon, and M. A. Alexander (2011), Influence of 417
the Meridional Shifts of the Kuroshio and the Oyashio Extensions on the 418
Atmospheric Circulation, J. Climate, 24(3), 762-‐777. 419
Huber, M., and R. Knutti (2014), Natural variability, radiative forcing and climate 420
response in the recent hiatus reconciled, Nature Geosci, 7(9), 651-‐656, 421
doi:610.1038/ngeo2228. 422
Kalnay, E., et al. (1996), The NCEP/NCAR 40-‐year Reanalysis Project, Bull. Amer. 423
Meteor. Soc., 77(March), 437-‐471. 424
Kim, B.-‐M., S.-‐W. Son, S.-‐K. Min, J.-‐H. Jeong, S.-‐J. Kim, X. Zhang, T. Shim, and J.-‐H. Yoon 425
(2014), Weakening of the stratospheric polar vortex by Arctic sea-‐ice loss, 426
Nature Com., 5, doi:10.1038/ncomms5646. 427
Kosaka, Y., and S. P. Xie (2013), Recent global-‐warming hiatus tied to equatorial 428
Pacific surface cooling, Nature, 501(7467), doi 10.1038/nature12534. 429
Kushnir, Y., W. A. Robinson, I. Blade, N. M. J. Hall, S. Peng, and R. Sutton (2002), 430
Atmospheric GCM response to extratropical SST anomalies: Synthesis and 431
evaluation, J. Climate, 15(16), 2233-‐2256. 432
Kwon, Y. O., M. A. Alexander, N. A. Bond, C. Frankignoul, H. Nakamura, B. Qiu, and L. 433
Thompson (2010), Role of the Gulf Stream and Kuroshio-‐Oyashio Systems in 434
Large-‐Scale Atmosphere-‐Ocean Interaction: A Review, J. Climate, 23(12), 435
3249-‐3281. 436
21
Liu, Z. Y., and M. Alexander (2007), Atmospheric bridge, oceanic tunnel, and global 437
climatic teleconnections, Rev. Geophys., 45(2), doi 10.1029/2005RG000172. 438
Mantua, N. J., and S. R. Hare (2002), The Pacific decadal oscillation, J. Oceanography, 439
58(1), 35-‐44. 440
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis (1997), A Pacific 441
interdecadal climate oscillation with impacts on salmon production, Bull. 442
Amer. Meteorolog. Soc., 78(6), 1069-‐1079. 443
Miller, A. J., D. R. Cayan, and W. B. White (1998), A westward-‐intensified decadal 444
change in the North Pacific thermocline and gyre-‐scale circulation, J, Climate, 445
11(12), 3112-‐3127. 446
Neelin, J. D., D. S. Battisti, A. C. Hirst, F.-‐J. Fei, Y. Wakata, T. Yamagata, and S. E. Zebiak 447
(1998), ENSO theory, J. Geophys. Res., 103(C7), 14261-‐14290. 448
Newlin, M. L., and M. C. Gregg (2014), Global Oceans [in “State of the Climate in 449
2013”], Bull. Amer. Meteor. Soc., 95(7), S51–S78. 450
North, G. T., Bell, R. Cahalan, and F. Moeng (1982), Sampling errors in the estimation 451
of empirical orthogonal functions, Mon. Wea. Rev., 110, 699-‐706. 452
Palmer, T. N., and D. L. T. Anderson (1994), The prospects for seasonal forecasting-‐a 453
review paper, Quart. J. Roy. Meteorol. Soc., 120(518), 755-‐793. 454
Qiu, B. (2000), Interannual variability of the Kuroshio Extension system and its 455
impact on the wintertime SST field, J. Phys. Ocean., 30(6), 1486-‐1502. 456
Rasmusson, E. M., and J. M. Wallace (1983), Meteorological aspects of the El 457
Niño/Southern Oscillation., Science, 222, 1195-‐1202. 458
22
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. 459
C. Kent, and A. Kaplan (2003), Global analyses of sea surface temperature, sea 460
ice, and night marine air temperature since the late nineteenth century, J. 461
Geophys. Re.-Atmos., 108(D14), doi10.1029/2002jd002670. 462
Smirnov, D., M. Newman, and M. A. Alexander (2014), Investigating the Role of 463
Ocean-‐Atmosphere Coupling in the North Pacific Ocean, J. Climate, 27(2), 464
592-‐606. 465
Smith, T. M., R. W. Reynolds, T. C. Peterson, and J. Lawrimore (2008), Improvements 466
to NOAA's historical merged land-‐ocean surface temperature analysis (1880-‐467
2006), J. Climate, 21(10), 2283-‐2296. 468
Trenberth, K. E., and J. T. Fasullo (2013), An apparent hiatus in global warming?, 469
Earth's Future, 1(1), 19-‐32. 470
Vimont, D. J., D. S. Battisti, and A. C. Hirst (2001), Footprinting: A seasonal 471
connection between the tropics and mid-‐latitudes, Geophys. Res. Lett., 28(20), 472
3923-‐3926. 473
Vimont, D. J., J. M. Wallace, and D. S. Battisti (2003), The seasonal footprinting 474
mechanism in the Pacific: Implications for ENSO, J. Climate, 16(16), 2668-‐475
2675. 476
Vivier, F., K. A. Kelly, and L. Thompson (1999), Contributions of wind forcing, waves, 477
and surface heating to sea surface height observations in the Pacific Ocean, 478
J.Geophys. Res.-Oceans, 104(C9), 20767-‐20788. 479
23
Wang, F. Y., Z. Y. Liu, and M. Notaro (2013), Extracting the Dominant SST Modes 480
Impacting North America's Observed Climate, J. Climate, 26(15), 5434-‐5452, 481
doi: 5410.1175/jcli-‐d-‐5412-‐00583.00581. 482
Wang, S. Y., L. Hipps, R. R. Gillies, and J.-‐H. Yoon (2014), Probable causes of the 483
abnormal ridge accompanying the 2013–2014 California drought: ENSO 484
precursor and anthropogenic warming footprint, Geophys. Rese. Lett., 41(9), 485
doi:10.1002/2014GL059748. 486
Zhang, Y., J. M. Wallace, and D. S. Battisti (1997), ENSO-‐like interdecadal variability: 487
1900-‐93, J. Climate, 10(5), 1004-‐1020. 488
489
490
491
492
24
Figure Captions: 493 494 Fig. 1 The regressions of monthly mean anomalies of global SST for the first three 495
EOFS of SST for the ocean areas included in the region from 30S-‐65N, 120E-‐105W 496
and the period from January 1900 to July 2014. Contour interval is 0.1˚C. Positive 497
values are red and negative values are blue, the zero contour is white. Robinson 498
Projection from 60S-‐60N. 499
500
Fig. 2 Time series of the Principal Components of the first three EOFs of monthly 501
mean Pacific SST poleward of 30S as shown in Fig. 1. Only values from January 1979 502
to August 2014 are shown. Units are standard deviations. Values are offset by four 503
standard deviations for clearer viewing. 504
505
Fig. 3 (top) The observed anomaly in 500hPa height for November 2013 to March 506
2014 (contour interval is 10m, positive is red and negative is blue and zero contour 507
is white). (middle) Regression of the 500hPa height anomalies for the months of 508
November to March for the years from 1948-‐2014 onto the principal component 509
time series for the third EOF of SST for the Pacific Ocean north of 30S determined 510
from the SST record between 1900 and 2014. (contour interval is 3m) (Bottom) 511
Same as middle panel, except that the height data is limited to 1979-‐2014 and 512
regressed on the principal component of the second EOF of the global SST 513
determined from the SST record between 1979 and 2014. 514
515
25
Fig. 4 (top) The November 2013 to March 2014 anomaly of the 500hPa height for 516
the 50-‐member AMIP ensemble for the ESRL-‐GFSv2 model. (contour interval is 10m 517
and zero contour is white). (bottom) Regression of 500hPa height anomalies from 518
the 50-‐member AMIP ensemble for the ESRL-‐GFSv2 model regressed onto the 519
principal component time series for the Pacific SST based on the period from 1900 520
to 2014. (contour interval is 3m). 521
522 523
26
524 Figures to go with: “Pacific Sea Surface Temperature and the Winter of 2014” 525
526
527 528
529 Fig. 1 The regressions of monthly mean anomalies of global SST onto the first three 530 EOFS of SST for the ocean areas included in the region from 30S-‐65N, 120E-‐105W 531 and the period from January 1900 to July 2014. Contour interval is 0.1˚C. Positive 532 values are red and negative values are blue, the zero contour is white. Robinson 533 Projection from 60S-‐60N. 534 535
EOF 1
EOF 2
EOF 3
27
536 537
538 Fig. 2 Time series of the Principal Components of the first three EOFs of monthly 539 mean Pacific SST poleward of 30S as shown in Fig. 1. Only values from January 1979 540 to August 2014 are shown. Units are standard deviations. Values are offset by four 541 standard deviations for clearer viewing. 542 543
1980 1985 1990 1995 2000 2005 2010 2015−8
−6
−4
−2
0
2
4
6
8
Year
Stan
dard
Dev
iation
s
PC1
PC3
PC2
28
544 545 Fig. 3 (top) The observed anomaly in 500hPa height for November 2013 to March 546 2014 (contour interval is 10m, positive is red and negative is blue and zero contour 547 is white). (middle) Regression of the 500hPa height anomalies for the months of 548 November to March for the years from 1948-‐2014 onto the principal component 549 time series for the third EOF of SST for the Pacific Ocean north of 30S determined 550 from the SST record between 1900 and 2014. (contour interval is 3m) (Bottom) 551 Same as middle panel, except that the height data is limited to 1979-‐2014 and 552 regressed on the principal component of the second EOF of the global SST 553 determined from the SST record between 1979 and 2014. 554 555 556
Observed Anomaly Nov-March 2013-14
Pacific Mode-3 Regression Nov-March 1948-2014
Global Mode-2 Regression Nov-March 1979-2014
29
557
558 559 Fig. 4 (top) The November 2013 to March 2014 anomaly of the 500hPa height for 560 the 50-‐member AMIP ensemble for the ESRL-‐GFSv2 model. (contour interval is 10m 561 and zero contour is white). (bottom) Regression of 500hPa height anomalies from 562 the 50-‐member AMIP ensemble for the ESRL-‐GFSv2 model regressed onto the 563 principal component time series for the Pacific SST based on the period from 1900 564 to 2014. (contour interval is 3m). 565 566 567 568 569
2014 Winter Anomaly - ESRL-GFSv2 Ensemble
Regression onto EOF3 of SST - ESRL-GFSv2 Ensemble
