1

Large Scale Control on the Patagonia Climate 1 2

R. Garreaud1♦, P. Lopez2, M. Minvielle1 and M. Rojas1 3

(1) Department of Geophysics, Universidad de Chile, Santiago, CHILE 4 (2) Centro de Estudios Científicos del Sur (CECS), Valdivia, CHILE 5

Accepted in Journal of Climate 6 Final version (May 2012) 7

8 ♦ Corresponding author address: 9 Dr. René Garreaud, Departament of Geophysics, Universidad de Chile 10 Blanco Encalada 2002, Santiago, CHILE. E-mail: rgarreau@dgf.uchile.cl 11

12

2

Abstract 13

Patagonia, located in southern South America, is a vast and remote region 14

holding a rich variety of past environmental records but a small number of 15

meteorological stations. Precipitation over this region is mostly produced by 16

disturbances embedded in the westerly flow and is strongly modified by the 17

austral Andes. Uplift in the windward side leads to hyper-humid conditions 18

along the Pacific coast and the western slope of the Andes; in contrast, 19

downslope subsidence dries the eastern plains leading to arid, highly 20

evaporative conditions. 21

Here we investigate the dependence of Patagonia’s local climate (precipitation 22

and near surface air temperature) year-to-year variability on large-scale 23

circulation anomalies using results from a 30-year long high-resolution 24

numerical simulation. Variations of the low-level zonal wind account for a large 25

fraction of the rainfall variability at synoptic and interannual timescales. Zonal 26

wind also controls the amplitude of the air temperature annual cycle by 27

changing the intensity of the seasonally varying temperature advection. 28

We also investigate the main modes of year-to-year variability of the zonal flow 29

over southern South America. Year round there is a dipole between mid and 30

high latitudes. The node separating wind anomalies of opposite sign migrates 31

through the seasons, leading to a dipole over Patagonia during austral summer 32

and a monopole during winter. Reanalysis data also suggests that westerly flow 33

has mostly decreased over north-central Patagonia during the last four decades, 34

causing a drying trend to the west of the Andes, but exhibit a modest increase 35

over the southern tip of the continent. 36

3

1. Introduction 37

Patagonia is a large and diverse region in southern South America that extends 38

from about 40°S down to the southern tip the continent (55°S), including Tierra 39

del Fuego and the southernmost section of the Andes cordillera (Fig. 1). The 40

western (Chilean) Patagonia is the narrow (∼50-150 km), intricate strip of land 41

from the Pacific coast to the Andean crest that rises to about 1500 m ASL in 42

these latitudes. To the east of the ridge, the Argentinean Patagonia features low-43

lying (100-200 m ASL), steppe-like plains extending for several hundred of 44

kilometres to the Atlantic sea border. Situated in the core of the midlatitudes 45

and downstream of the vast southern Pacific, the Patagonia faces strong 46

westerlies throughout the year (Garreaud et al. 2009). The baroclinic waves 47

embedded in the main westerly flow are profoundly perturbed by the austral 48

Andes, leading to one of the most dramatic precipitation gradients on earth 49

(e.g., Smith and Evans 2005; Carrasco et al. 2002; see also Fig. 2a). Western 50

Patagonia features a temperate, hyper-humid climate with freezing levels 51

generally above 1 km ASL, a modest seasonal cycle and annual mean 52

precipitation in the 5.000-10.000 mm range (Fig. 2, see also Miller 1976). Such 53

high accumulations arise from the orographic enhancement of synoptic-scale 54

precipitation upstream of the mountains (e.g., Roe 2005) and support extensive 55

rain-forests and major rivers, numerous glaciers, the Northern and Southern 56

Patagonia Ice fields (NPI and SPI), and the Cordillera Darwin ice field (CDI). 57

4

The mean precipitation decreases to less than 300 mm/year just a few tens of km 58

downstream of the continental divide, leading to a rapid disappearance of 59

vegetation and a rain shadow that extends all the way to the Atlantic coast 60

where mean precipitation only reaches to 500-700 mm/year (Fig. 2, see also 61

Prohaska 1976; Paruelo et al. 1998; Carrasco et al. 2002). In addition to its 62

aridity, eastern Patagonia features a continental climate, a winter-to-summer 63

temperature amplitude of more than 10°C, and extremely windy, highly 64

evaporative conditions at the surface. 65

Climate variability in Patagonia is important on its own, as a key driver of local 66

changes in the cryosphere and biosphere (e.g., Schneider and Giess 2004; Lara et 67

al. 2005; Rasmussen et al. 2007), but also can shed light on the functioning of the 68

Southern Hemisphere (SH) extratropical circulation, as South America is the 69

only significant land-mass extending to the south of 45°S. For instance, the SH 70

annular mode and other hemispheric modes have a significant influence on 71

Patagonia’s precipitation and temperature (Gillet et al. 2006; Silvestri and Vera 72

2009; Garreaud et al. 2009; Quintana and Aceituno 2012), although a causal 73

relationship has not been proposed yet. Furthermore, as reviewed in Villalba et 74

al. (2009), the Patagonia region hosts a rich variety of environmental paleo-75

records, including ice-cores, glaciers, lake and marine sediments, and tree-rings, 76

thus offering an excellent opportunity to explore climates of the past in time 77

scales from the late-Holocene to the Last Glacial Maximum (LGM) and beyond 78

5

(e.g., Moreno 2004; Lamy et al. 2010; Blisniuk et al. 2005). There is also 79

mounting evidence of contemporaneous, spatially inhomogeneous trends in 80

precipitation (Quintana and Aceituno 2012; Aravena and Luckman 2009; see 81

section 4b) and subtle but widespread warming (Rosenbluth et al. 1997; 82

Rasmussen et al., 2007) affecting the regional cryosphere. A comprehensive 83

survey in Lopez et al. (2010) found that the majority of the Patagonia’s glaciers 84

have retreated, some of them quite dramatically, during the second part of the 85

XX century and NPI / SPI have shrunk considerably in the last decades 86

(Holmund and Fuenzalida 1995; Rignot et al., 2003; Aniya 2007). Nevertheless, a 87

few glaciers in SPI and CDI have advanced or remained stable (Lopez et al. 88

2010; Möller et al. 2007) 89

Like many other remote, sparsely populated regions Patagonia lacks of a 90

meteorological network with enough spatial density and record length to 91

properly describe climate variability and climate change. With only two regular 92

upper-air stations and less than 20 long-term surface stations (most of them 93

located near the coast, Fig. 1), researchers have pushed the limits of statistical 94

climatology to come up with a regional picture of the climate’s mean state and 95

variability or longer trends. To circumvent this problem, here we use output 96

6

from PRECIS-DGF, a high-resolution, limited-area simulation1 with lateral 97

boundaries forced by reanalysis data. The reanalysis is able to capture most of 98

the large-scale circulation variability and the regional model performs a 99

dynamical downscaling that is physically consistent with local geography 100

(detailed topography, land use, coastlines, etc.). As documented later, the 101

model shows a good agreement with the observations, especially resolving the 102

interannual variability. 103

The main goal of this work is to quantify and understand the relationship 104

between large-scale circulation (e.g., zonal wind aloft, U) and local climate 105

(precipitation, P, and surface air temperature, SAT) at interannual timescales, 106

under the premise that variations in the former accounts for most of the 107

fluctuations in the later. Indeed, using coarse-resolution datasets, Garreaud 108

(2007) found significant correlations between annual mean values of U at 850 109

hPa and collocated precipitation over the extratropics (particularly strong near 110

the austral Andes) and Rasmussen et al. (2007) investigated the influence of 111

upper-air conditions on the Patagonia icefield using a simple precipitation 112

model forced by reanalysis data. Owing to their nature, large-scale wind 113

anomalies exhibit significant spatial and temporal coherence, and they are well 114

1 An alternative dataset would be the recently released Climate Forecast System

Reanalysis (CFSR, Saha et al. 2010) at a horizontal resolution of 0.3°

(comparable with PRECIS-DGF).

7

resolved by atmospheric reanalysis available for several decades now. In this 115

work we also investigate the main modes of year-to-year variability and 116

contemporaneous trend of the zonal flow over Patagonia, which exhibit 117

important seasonal variations, to infer their impacts on regional P and SAT. 118

Linking local climate variability with large-scale circulation anomalies has two 119

main applications. First and foremost, it allows a statistical downscaling of 120

large-scale signals to infer, interpret and extend regional changes in P and SAT, 121

a worthy effort considering the deficiencies in the observational network. 122

Secondly, it may help to upscale local climate signals assisting paleo-climate 123

studies. Indeed, a few studies have attempted to resolve changes in the 124

Southern Hemisphere Westerlies on the basis of paleo-records in Patagonia (e.g, 125

Lamy et al. 2010; Fletcher et al. 2012). 126

The rest of the paper is organized as follows. In section 2 we describe the 127

observational datasets as well as the regional and global climate model used in 128

this work. A customary validation of the regional climate model over southern 129

South America is included in that section. The co-variability between zonal flow 130

aloft and precipitation and surface air temperature over Patagonia is 131

documented in section 3. We begin by briefly exploring the relationship at the 132

scale of individual storms, and then address the year-to-year variation using 133

outputs from a regional simulation of the current climate. Once the strong 134

control of free-troposheric flow on P and SAT is established, we investigate the 135

8

main modes of the zonal flow interannual variability (section 4a) and trends 136

during the last decades (section 4b), which allow to infer their counterparts in 137

precipitation over Patagonia. Section 5 provides a summary of our main 138

findings. 139

2. Dataset 140

a. Meteorological observations and Reanalysis data 141

In this study we use monthly rainfall and near surface air temperature 142

(nominally at 2 m above ground) records from 31 stations in southern South 143

America (Fig. 1) from 1948 to 2010, obtained from the Global Historical Climate 144

Network (GHNC version 2; Peterson and Vose, 1997). This data provides a 145

station-based climatology and wind-precipitation relationship to compare 146

against model results. The record length is variable, but in each station we have 147

at least 10 years worth of data; missing values were not filled. 148

The large-scale state of the atmosphere was characterized using two 149

atmospheric reanalysis: the National Centers for Environmental Prediction / 150

National Center for Atmospheric Research Reanalysis (NNR; Kalnay et al. 1996) 151

and the European Centre for Medium Range Weather Forecast (ECMWF) 40 152

Years Reanalysis (ERA40; Upalla et al. 2005). ERA-40 was also used to force the 153

regional climate model (see below). Both reanalysis are available at a global, 154

regular 2.5°×2.5° lat-lon grid at standard pressure levels. The original time 155

interval is 6 hr, from which daily and monthly averages are calculated. The 156

9

NNR is available from 1948 to 2011 (routinely updated) and the ERA40 covers 157

the period from mid-1957 to mid-2002. Such record lengths allow their use for 158

studying interannual and interdecadal variability, but trends derived from 159

reanalysis must be used with caution because the number and type of ingested 160

data has varied over time. Indeed, Bromwich and Fogt (2004) report a 161

significant improvement in the skill of NNR and ERA-40 in the high and mid-162

latitudes of the SH after 1978, coincident with the beginning of the satellite-data 163

assimilation. 164

b. PRECIS-DGF simulations 165

Providing REgional Climate for Impact Studies (PRECIS) is a regional climate 166

model developed by the Hadley Centre in the UK (Jones et al., 2004) widely 167

used for studies on climate change (e.g. Marengo et al. 2009). Like any regional 168

model, PRECIS solves the atmospheric governing equations in an area limited 169

domain, forced by lateral, bottom and top boundary conditions (BC). We 170

performed a continuous 45-year long (1958-2001), 25-km grid-spacing 171

resolution simulation (referred to as PRECIS-DGF) over southern South 172

America (Fig. 1) using 6-hourly ERA-40 as lateral BC and observed sea-surface 173

temperature (the HadlSST dataset, Rayner et al., 2003) as bottom BC. Two-174

dimensional outputs (e.g., near surface air temperature and precipitation rate) 175

are available every 6 hours on a regular 0.25°×0.25° lat-lon grid; three-176

dimensional outputs (e.g., zonal wind and air temperature) have the same 177

10

horizontal and temporal resolution and were interpolated to standard pressure 178

levels (including 850 and 500 hPa). 179

Given its configuration, the ability of PRECIS-DGF to simulate the atmospheric 180

mean state and variability is critically dependent on the capacity of ERA-40 to 181

resolve the large-scale circulation over South American and the adjacent oceans, 182

so we decided to restrict our analysis to the period 1978-2001. A detailed 183

evaluation of the PRECIS-DGF modeling system is given in Fuenzalida et al. 184

(2007) and Garreaud and Falvey (2008). (The availability of these independent 185

verifications of the PRECIS-DGF performance over southern South America 186

was an important reason to chose the model output as the primary dataset for 187

the present analysis instead of the recently released CFSR dataset). Figures 2a,b 188

show the annual mean and seasonal amplitude of the precipitation over 189

southern South America using the PRECIS-DGF results. To facilitate 190

comparison, station-based long-term-mean values are superimposed on the 191

simulated field using the same color scale. The model is able to capture the 192

gradual north-to-south increase in precipitation along the Pacific coast and the 193

sharp gradient across the austral Andes. As commented before, there is a lack of 194

surface station in western Patagonia, but simulated mean accumulations in the 195

2.000-11.000 mm/year range are consistent with independent estimates based on 196

river-discharge (DGA 1987; Escobar et al. 1992), satellite-derived precipitation 197

offshore (Falvey and Garreaud 2005) and the mass balance over the NPI/SPI 198

11

(e.g., Rignot et al. 2003). The model, however, overestimates by a factor ~2 the 199

mean precipitation over high terrain at subtropical latitudes (30-35°S) where the 200

Andes cordillera exceeds 4000 m ASL. On the other hand, PRECIS-DGF 201

correctly simulates the annual cycle of precipitation (Fig. 2b) characterized by a 202

wintertime maximum at subtropical- and mid-latitudes and a conspicuous 203

minimum to the south of 47°S. Likewise, the annual mean and seasonal 204

amplitude of the surface air temperature are well simulated by the model over 205

Patagonia (Figs. 2c,d). 206

Since PRECIS-DGF was forced by realistic, time-varying lateral (ERA-40) and 207

bottom (HadISST) boundary conditions, the model also mimics precipitation 208

and temperature variability at daily to interannual timescales. As an example, 209

Figure 3 shows the observed and simulated annual mean time series of P and 210

SAT in Puerto Montt (northern Patagonia) and Punta Arenas (southern 211

Patagonia). A more complete analysis of the PRECIS-DGF ability to simulate 212

year-to-year precipitation changes in Patagonia is provided by Fuenzalida et al. 213

(2007) where they conclude that the PRECIS-DGF simulation realistically 214

represents the interannual variability as well as the annual and seasonal 215

precipitation trend over the 1978-2001 period. In the free troposphere, the 216

PRECIS-DGF simulation closely follows the ERA-40 circulation and properly 217

captures the westerly wind belt at midlatitudes and the upper-level jet stream. 218

In summary, the PRECIS-DGF simulation provides a long (1978-2001), sub-219

12

daily, high-resolution (0.25°) dataset over southern South America that 220

reproduces the main features of the regional climate, climate variability and 221

synoptic-scale fluctuations. Because this dataset was created using a dynamical 222

model (instead of interpolating station data), it is also physically consistent and 223

therefore amenable to diagnose the relationship between the wind field aloft 224

and surface conditions. 225

3. Zonal wind control on precipitation and surface air temperature 226

a. Storm-scale behavior 227

Before we describe the year-to-year covariability of precipitation and wind aloft 228

it is instructive to briefly explore their association at the scale of individual 229

storms. To this effect, Fig. 4 shows the local (point-to-point) correlation map 230

between daily averages of precipitation (P) and the 850 hPa wind components 231

(U850, V850) from the PRECIS-DGF simulation, conditioned to presence of 232

precipitation (P > 1 mm/day). 1The synoptic-scale environment during storms 233

over southern South America typically features a midlatitude trough with its 234

axis just to the west of the Andes and northwesterly flow in the free 235

troposphere (e.g., Barrett et al. 2011; Moreno 2010). Nevertheless, the 236

association between wind and precipitation widely varies across the domain. 237

Over the south Pacific there is a moderate correlation between precipitation and 238

collocated wind aloft, with r(P,U850) ~ −r(P,V850) ~ +0.5, indicative of high 239

precipitation under strong synoptic forcing. Closer to the continent the 240

13

correlation with the zonal (meridional) flow increases (decreases) until reaching 241

the western slope of the Andes where r(P,U850) ~ +0.7 and r(P,V850) ~ -0.3. The 242

strong dependence of precipitation on zonal flow aloft is remarkably uniform 243

along extratropical Chile and results from the orographic enhancement of 244

precipitation, as mid-level flow normal to the Andes produces upslope flow 245

acting in concert with the synoptic-scale ascent (see Roe 2005 for a review). 246

These PRECIS results are also consistent with the analysis in Falvey and 247

Garreaud (2007) on the basis of daily rainfall station and radiosonde data in 248

central Chile (32-36°S). 249

The orographic enhancement ends sharply at the mountain ridge. Not only the 250

long-term-mean precipitation decreases by a factor ~10 from west to east (Fig. 251

2a), but r(P,U850) ~ −0.4 over much of the Argentinean Patagonia (Fig. 4). The 252

few precipitation events in this semiarid region are most often connected with 253

an incoming trough from the south Pacific, the same synoptic pattern that 254

favors rainfall to the west of the Andes. Under these conditions, however, the 255

amount of precipitation tends to be slightly larger in regions under weak 256

westerlies or even easterly flow. Foehn-like windstorms in western Argentina 257

(locally known as Zonda winds) are well documented around 32-36°S during 258

frontal episodes in south-central Chile (Seluchi et al., 2003). The same 259

phenomena is likely to take place farther south, in areas where strong westerlies 260

cross the Andes, resulting in vigorous downslope flow drying the lower 261

14

troposphere to the east of the ridge and thus locally reducing precipitation in 262

the Argentinean Patagonia. 263

b. Wind-Precipitation relationship 264

Figure 5a shows the local correlation between P and U850 using annual means 265

from PRECIS-DGF (both fields were detrended). We focus on the precipitation 266

dependence on the zonal flow because this component exhibits much larger 267

year-to-year variation than the meridional component in the SH (e.g., Trenberth 268

1981). Over southern South America, the interannual r(P,U850) map remains 269

similar when considering seasonal means (not shown). To the south of 40°S the 270

correlation pattern exhibits little dependence in the north-south direction with 271

positive values increasing from the South Pacific to a maximum along the 272

Chilean coast and the western slope of the Andes (r(P,U850) ~ 0.8), a sharp 273

transition just to the east of the mountain ridge and negative values over the 274

Argentinean Patagonia. Regardless of the latitude, the sign of r(P,U850) changes 275

in less than ∼100 km of the continental divide (Fig. 5b). Thus, at any zonal 276

transect over Patagonia, a year (or season) with stronger than average mid-level 277

westerly flow features increased precipitation to the west of the Andes and 278

decreased precipitation over the lowlands to the east. The marked west-east 279

precipitation gradient over Patagonia is always present but it is slightly less in 280

those years with weaker than average westerly flow aloft. 281

15

The scatter plot between monthly mean values of P and U850 in two boxes at 282

50°S further illustrates the strength and nature of their association (Fig. 6). A 283

linear relationship is evident for both sides of the Andes but with significant 284

more dispersion in the dry region to the east. We also constructed the local 285

r(P,ULEV) map varying LEV from 1000 to 200 hPa and verified that the overall 286

structure remains similar at different levels although the correlation values are 287

the most significant between 900 and 700 hPa and largest at 850 hPa, 288

approximately the Andean ridge level. 289

A strong, longitude-dependent linear relationship between P and U850 is also 290

evident in the interannual r(P,U850) map presented by Garreaud (2007) based on 291

NCEP-NCAR reanalysis winds and CMAP precipitation (their Fig. 2b), as well 292

as in Fig. 7 constructed using annual mean precipitation at the available 293

stations. The similarity with these observed patterns lends support to our 294

PRECIS-DGF results. Furthermore, the interannual r(P,U850) map is quite similar 295

to its daily counterpart (c.f. Figs. 4-5a), as the correlations at longer times 296

emerge from the aggregated effect of individual storms. The strong 297

covariability between zonal wind aloft and precipitation over western 298

Patagonia seems also to operate within the annual cycle, as the conspicuous 299

wintertime precipitation minimum over the austral Andes (south of 48°S, Fig. 300

2b) coincides with a seasonal weakening of the westerlies over this region. 301

302

16

c. Wind-Temperature relationship 303

We now explore the interannual relationship between the surface air 304

temperature2 (SAT) and low-level zonal flow. Over southern South America, 305

the annual mean correlations (not shown) are weak and smaller than their 306

precipitation counterparts (Fig. 5a). The temperature-wind relationship, 307

however, exhibits a marked annual cycle and r(SAT,U850) values can be as large 308

as r(P,U850) in individual seasons (Fig. 8). Also note that the correlation between 309

seasonal SAT and zonal wind maintains its sign across the Andes, contrasting 310

with the longitude-dependent r(P,U850) pattern, and its amplitude decreases 311

toward the tip of the continent. 312

During summer the correlation is mostly negative, with maximum values along 313

the Chilean coast (~ −0.8) decaying toward the east (Fig. 8a). Negative 314

correlations also dominate during fall, especially along the Chilean coast, but 315

with smaller values. During winter r(SAT,U850) is positive over much of 316

Patagonia with higher values to the east of the Andes (~ 0.8, Fig. 8c). This 317

pattern also appears in spring, but with smaller values. Let us consider a case in 318

which stronger than average westerly flow prevails year-round; according to 319

2 The correlations over the ocean are dubious, because SAT values there are strongly controlled by the underlying SST which is prescribed in PRECIS-DGF; in contrast sub-surface temperatures are calculated using a land-surface sub-model at each time step (including the varying effect of soil temperature and soil moisture) so that SAT over the continent is fully consistent with the atmospheric forcing, including radiative and turbulent heat fluxes as well as temperature advection.

17

Fig. 8 that would result in a milder winter and a cooler summer, thus reducing 320

the amplitude of the temperature annual cycle (seasonality) over Patagonia. In 321

contrast, weaker than average westerlies year round result in a colder winter 322

and a warmer summer, thus increasing the temperature seasonality. 323

The winter-to-summer sign reversal of r(SAT,U850) reflects changes in the low-324

level energy balance over Patagonia. From spring to summer, insolation tends 325

to warm the land very effectively and hence to increase SAT throughout 326

sensible heat fluxes. In contrast, SST over the adjacent south Pacific experience 327

little change and so does the SAT over ocean. Thus, an increase in low-level 328

westerly flow enhances cold air advection over the continent, especially in 329

western Patagonia, leading to a cooler summer season. A similar mechanism 330

explains the mostly positive correlations during winter, when an increase in 331

westerly flow enhances advection of warm, maritime air over the continent, 332

leading to a milder winter. The key role of the temperature advection in setting 333

r(SAT,U850) underlines the control of the low-level zonal flow upon SAT 334

seasonality rather than annual mean SAT values. Indeed, the surface air 335

temperature at any given location in Patagonia is also highly correlated (r ≥ 336

0.85, not shown) with the regional-average free tropospheric air temperature 337

(e.g., T850), which is in turn determined by large-scale processes that may occur 338

independent of changes in zonal wind intensity. 339

340

18

4. Zonal wind patterns 341

The correlation maps presented in the previous section links local interannual 342

variations of annual mean precipitation and surface air temperature seasonality 343

with collocated 850-hPa zonal flow anomalies. Nevertheless, the zonal wind in 344

the free troposphere does not vary independently from one point to another 345

and organized structures in the large-scale flow will imprint the climate 346

conditions over Patagonia. In this section we explore these structures at 347

different timescales and analyze their impacts on regional P and SAT. 348

a. Interannual variability 349

To study the main modes of year-to-year zonal flow variability over southern 350

South America we applied an EOF analysis to the 850 hPa zonal wind 351

anomalies in the domain 35°-65°S and 90-60°W (Fig. 9). To isolate the 352

interannual variability we detrended the series by removing a linear fit with 353

time at each grid point. The EOF was performed using the NCEP-NCAR 354

reanalysis (NNR) and ECMWF reanalysis (ERA40) between 1978 and 2001. The 355

spatial pattern (EOF1) and loading factor (PC1) of the leading mode are quite 356

similar between the two reanalysis, indicative of its robust character. In sake of 357

brevity we only show results from NNR. The zonal wind field was projected 358

upon the PC to obtain the global signature associated to each mode. 359

The leading mode accounts for about 40-50% of the variance in each month and 360

it is well separated from the second mode considering the statistical rule of 361

19

North et al. (1982). Examination of the monthly EOFs, however, reveals 362

substantial differences in EOF1 between summer and winter months. During 363

summer (DJF), there are strong, opposite correlations in two zonally elongated 364

bands extending across the whole SH (Fig. 9a). The bands are centered at about 365

40°S and 60°S with its node coincident with the axis of the surface westerly belt. 366

The circumpolar pattern in Fig. 9a is very similar to the Antarctic Annular 367

Mode (AAO, e.g., Gong and Wang 1999; Thompson and Wallace 2000; Marshall 368

2003) and the correlation between PC1 and the seasonal averaged AAO index 369

reaches +0.7. Thus, we interpret the zonal wind variability over Patagonia 370

during austral summer as part of the hemispheric AAO mode, characterized by 371

a latitudinal vacillation of the tropospheric-deep westerly wind maxima around 372

50°S. 373

The EOF1 during winter-spring also exhibits opposite correlations between mid- 374

and high-latitudes (Fig. 9b) but the high values are restricted to southern South 375

America and the adjacent Pacific, and PC1 is not well correlated with the 376

seasonal AAO index ( r ∼ 0.2). This result3 is consistent with the less prominent 377

3 Understanding the seasonal change in the leading mode is beyond the scope

of this paper. Recall, however, that over the Pacific and during austral summer,

there is one single upper-level zonal wind maxima (i.e., subtropical jet merged

with the polar front jet) at ∼50°S, the same latitude of the surface westerly wind

belt. In contrast, during austral winter the subtropical jet migrates to about 30°S

and the surface westerlies relax substantially but their maximum remains in

midlatitudes. We speculate that such marked seasonal differences in the

20

role of the SH annular mode in austral winter documented by Matthewman 378

and Magnusdottir (2011). On the other hand, there are high correlations off the 379

subtropical Chilean coast, where zonal wind anomalies are out-of-phase with 380

those farther to the south. We also projected the sea level pressure on PC1 (not 381

shown) to better detect the subtropical forcing of EOF1. Thus, winter-to-winter 382

zonal wind anomalies over Patagonia are associated with pressure changes in 383

the austral Pacific as well as with fluctuations in the positioning and intensity of 384

the subtropical Pacific anticyclone. When the subtropical anticyclone extends 385

abnormally south the zonal wind strengthens over southern South America; an 386

anticyclone more confined into the subtropics is associated with relaxed zonal 387

flow at midlatitudes. 388

Of particular relevance for our work is the seasonal displacement of the node 389

(the 0-line) in EOF1 over the continent most readily seen in a time-latitude 390

section (Fig. 10a). From December to April, zonal wind anomalies (enhanced or 391

relaxed westerlies) are out-of-phase between north-central Patagonia and Tierra 392

del Fuego. In contrast, from May to November the node is located at about 40°S 393

so the sign of the zonal wind anomalies is uniform over much of southern 394

South America. These results are confirmed by the correlation of U850 at each 395

month between one point in central Patagonia and the other in the tip of the 396 background flow may determine the preferred modes of interannual variability

in each season, even though the mean field was subtracted before the EOF

analysis.

21

continent: moderate positive correlations from June to November and negative 397

correlations the rest of the year (not shown). Considering the P-U850 correlations 398

over the western sector of Patagonia, we expect a tendency for a dipole (with its 399

node around 50°S) in summer precipitation variability and a monopole in 400

winter-spring precipitation variability. The temperature interannual variability 401

is less influenced by the seasonal migration of the node in EOF1 since 402

r(SAT,U850) is highest in central Patagonia, where the U850 anomalies are 403

coherent year round. 404

The differences in the geographical span of the seasonal EOF1 are also relevant 405

when attempting to upscale environmental conditions over Patagonia. 406

Precipitation anomalies in summer can be linked with circumpolar anomalies in 407

the low-level (and probably tropospheric-deep) zonal flow at mid- and high-408

latitudes, which in turn are well correlated with the hemispheric-scale AAO. In 409

contrast, precipitation anomalies in winter can only be linked with low-level 410

westerly wind anomalies in the southeast Pacific and are less correlated with 411

the hemispheric AAO, but they also indicate changes in the position and 412

intensity of the subtropical Pacific anticyclone. 413

b. Contemporaneous trends 414

Let us now describe the trends in zonal wind over the last few decades when 415

reanalysis are available. Figure 11 shows the linear trend in the annual mean 416

850-hPa zonal wind using ERA-40 and NNR for the period 1968-2001. Both 417

22

reanalysis show a band of increasing westerlies over the Southern Oceans 418

around Antarctica (with the largest amplitude in the Atlantic and Indian 419

sectors) and decreasing winds poleward. This dipole in the zonal wind trend is 420

consistent with a tendency of the AAO toward its positive phase (lower 421

pressure at high latitudes), especially in spring-summer, detected in several 422

studies (e.g., Thompson and Solomon 2002; Gillet and Thompson 2003). At 423

midlatitudes there is a tendency toward weaker westerlies, although the signal 424

there is less coherent between the reanalysis and exhibits more dependence on 425

longitude. 426

To focus on the changes over southern South America, the linear trend for each 427

month was averaged between 75-65°W and presented as a time-latitude section 428

in Fig. 10b. Consistent with the hemispheric, annual mean trends, decreasing 429

(increasing) westerlies prevails around 45°S (60°S). Both reanalysis indicate a 430

significant reduction of the westerlies over north-central Patagonia throughout 431

the year, but with a larger amplitude during winter and spring. On the other 432

hand, the high-latitude area of increasing westerlies reaches Tierra del Fuego 433

most of the year but leads to significant trends during austral summer only. 434

Such a winter monople and a summer dipole in the zonal wind trend is similar 435

to the seasonal march in the leading mode of interannual variability (cf. Fig. 9a). 436

Linear trends of U850 derived from radiosonde observations at Puerto Montt 437

23

(42°S, 73°W) and Punta Arenas (53°S, 71°W) are generally consistent with the 438

previous analysis lending credibility to the reanalysis-based results. 439

We can use the interannual correlations obtained in section 3b to infer long-440

term precipitation changes in Patagonia that are congruent with the zonal wind 441

trend, calculated as ∆P* = β⋅∆U850, where β is the slope of the linear fit and ∆U850 442

is the reanalysis mean (average of NNR and ERA40) zonal wind change 443

between 1968 and 2001 interpolated to each PRECIS-DGF grid box. The slope β 444

has the same pattern as the P-U850 correlation coefficient (c.f., Fig. 5a); the coarse 445

reanalysis trend ∆U850 has little dependence on longitude but it varies more in 446

latitude (c.f., Fig. 10b). The results are shown in Fig. 12 for annual 447

accumulations. To the west of the Andean ridge there is a 300-800 mm/decade 448

decrease in precipitation over north-central Patagonia and a ∼200 mm/decade 449

increase to the south of 50°S. Changes in eastern Patagonia are too small and 450

not significant. The previous changes per decade represent about ±20% of the 451

long-term annual means. These results are in good agreement with a trend 452

analysis based on station data by Carrasco et al. (2002), Aravena and Luckman 453

(2010), Lopez et al. (2008) and Quintana and Aceituno (2012). Such agreement 454

further indicates that wind-driven changes account for most of the total 455

precipitation changes in western Patagonia. The slight increase in precipitation 456

in the western side of Tierra del Fuego is also compatible with a moderate 457

24

glacial thickening in the centre of the Gran Campo Nevado ice cap (53°S, 73°W) 458

during the period 1984-2000 found by Möller et al. (2007). 459

5. Concluding remarks 460

Interannual variability of precipitation (P) and near-surface air temperature 461

(SAT) over Patagonia is important on its own, since it controls changes in the 462

regional cryosphere and biosphere, but it also sheds light on a broader context, 463

since southern South America is the only significant land mass in the SH 464

midlatitudes and the region holds a rich variety of past environmental records. 465

Nevertheless, the number of meteorological stations in this vast and diverse 466

region is clearly insufficient to describe its mean climate, interannual variability 467

and contemporaneous climate change. This problem motivated a central 468

objective of this work, in which we try to link local climate variability with 469

large-scale circulation anomalies. The local climate fluctuations are properly 470

described by the results of a high-resolution (25 km horizontal grid spacing) 471

regional numerical model (PRECIS) forced by realistic boundary conditions 472

(ERA40) in the period 1978-2001. The large-scale circulation is described using 473

two reanalysis datasets: ERA-40 and NCEP/NCAR. 474

By performing a local (point-to-point) linear correlation analysis between the 475

seasonal values of zonal flow at 850 hPa (U850, representative of the low-level 476

circulation) and P and SAT, we reach the following conclusions: 477

25

• To the south of 40°S, U850 is strongly and positively (negatively) 478

correlated with P to the west (east) of the Andes ridge in all seasons. This 479

result is in line with previous findings and our own analysis based on 480

station data. It indicates that, at any latitude in western (eastern) 481

Patagonia, a season with stronger westerlies will augment (decrease) the 482

local precipitation. 483

• The P-U850 seasonal correlation stems from their association at synoptic 484

scale, and is particularly strong because of the mechanical effect of the 485

Andes (windward uplift and leeside subsidence). The P-U850 even seems 486

to operate within the annual cycle, as the conspicuous wintertime 487

precipitation minimum over the austral Andes (south of 48°S, Fig. 2b) 488

coincides with a seasonal weakening of the westerlies over this region. 489

• Over much of Patagonia, U850 and SAT are positively correlated during 490

winter but negatively correlated during summer. The sign change of the 491

correlation is borne in the seasonal variation of the low-level air 492

temperature advection from the Pacific Ocean into the continent. Thus, if 493

stronger (weaker) than normal westerlies prevail year round that will 494

result in a decreased (increased) amplitude of the local air temperature 495

annual cycle. 496

26

• The annual mean surface air temperature depends on other factors as 497

well (e.g., SST and air temperature aloft) so that, in general, it is not 498

possible to diagnose the mean SAT on the basis of U850 alone. 499

The correlation analysis links local interannual variability of P and SAT with 500

collocated U850 anomalies. The latter field, however, exhibits a large-scale 501

organization that was also examined in this work by performing EOF and 502

linear-trend analyses of the zonal flow over southern South America and the 503

adjacent oceans. Our main findings are: 504

• The year-to-year fluctuations of the seasonal mean flow typically exhibits 505

an out-of-phase behavior between the mid- and high-latitudes. The node 506

separating wind anomalies of opposite sign migrates through the 507

seasons, leading to a dipole over Patagonia during austral summer and a 508

monopole during winter. 509

• Coupled with our correlation analysis, the previous result has a major 510

implication in the precipitation variability over western Patagonia: 511

winter-to-winter anomalies tend to be coherent (either positive or 512

negative) over this region, but summer-to-summer anomalies tend to 513

exhibit a dipole with its node around 50°S. 514

• We also found that Precipitation anomalies over Patagonia during 515

summer can be linked with circumpolar anomalies zonal flow at mid- 516

and high-latitudes, and are well correlated with the hemispheric-scale 517

27

AAO. In contrast, precipitation anomalies in winter can only be linked 518

with low-level westerly wind anomalies in the southeast Pacific and are 519

less correlated with the hemispheric AAO. (Indeed, winter anomalies of 520

the SH flow is dominated by the wave-number-three mode rather than 521

by a zonally symmetric AAO pattern). 522

• Considering the period 1968-2001, both reanalysis exhibit a reduction of 523

the westerlies over north-central Patagonia throughout the year, but with 524

a larger amplitude during winter and spring. On the other hand, the 525

high-latitude area of increasing westerlies reaches Tierra del Fuego most 526

of the year but leads to significant trends during austral summer only. 527

• Again, coupling the previous finding with the correlation analysis over 528

western Patagonia indicates a 300-800 mm/decade decrease in 529

precipitation over north-central Patagonia and a 200-300 mm/decade 530

increase to the south of 50°S. These results are qualitatively in agreement 531

with recent trends estimates based on station data. 532

533

Acknowledgements 534

We acknowledge FONDECYT grants 1110169 (RG), 1080606 (MR) and 535

1090791 (PL) and 3110120 (MM). We are grateful of constructive 536

comments two anonymous reviewers and Dr. Dave Rahn. 537

538

28

References 539

Aniya, M., 2007. Glacier variations of Hielo Patagonico Norte, Chile, for 540

1944/45-2004/05. Bulletin of Glaciological Research, 24, 59 - 70. 541

Aravena, J. C. and B. H. Luckman, 2009: Spatio-temporal rainfall patterns in 542

Southern South America. International Journal of Climatology, 29, 2106-2120. 543

Barrett, B. S., D. B. Krieger, and C. P. Barlow, 2011: Multiday Circulation and 544

Precipitation Climatology during Winter Rain Events of Differing Intensities in 545

Central Chile. Journal of Hydrometeorology, 12, 1071-1085. 546

Blisniuk, P. M., L. A. Stern, C. P. Chamberlain, B. Idleman, and P. K. Zeitler, 547

2005: Climatic and ecologic changes during Miocene surface uplift in the 548

Southern Patagonian Andes. Earth and Planetary Science Letters, 230, 125-142. 549

Braconnot, P., B. Otto-Bliesner, S. Harrison, S. Joussaume, J. Y. Peterchmitt, A. 550

Abe-Ouchi, M. Crucifix, E. Driesschaert, T. Fichefet, and C. Hewitt, 2007: 551

Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial 552

Maximum–Part 2: feedbacks with emphasis on the location of the ITCZ and 553

mid-and high latitudes heat budget. Climate of the Past, 3, 279-296. 554

Bromwich, D. H. and R. L. Fogt, 2004: Strong Trends in the Skill of the ERA-40 555

and NCEP-NCAR Reanalyses in the High and Midlatitudes of the Southern 556

Hemisphere, 1958-2001*. Journal of Climate, 17, 4603-4619. 557

Carrasco, J., G. Casassa and A. Rivera, 2002: Meteorological and climatological 558

aspects of the Southern Patagonia Ice Cap. In "Ice fields scientific task force", 559

New York, Kluwer Academic, pp. 29-41. 560

DGA (Direccion General de Aguas, Chile) 1987: National Water Balance. DGA 561

Technical report, Santiago, Chile. 60 pp. 562

29

Escobar, F., Vidal, F. and Garin, C., 1992. Water balance in the Patagonia 563

Icefield. In: R. Naruse (Editor), Glaciological Researches in Patagonia, 1990. 564

Japanese Society of Snow and Ice, pp. 109 – 119. 565

Falvey, M. and R. Garreaud, 2007: Wintertime precipitation episodes in central 566

Chile: Associated meteorological conditions and orographic influences. Journal 567

of Hydrometeorology, 8, 171-193. 568

Falvey, M. and R. D. Garreaud, 2009: Regional cooling in a warming world: 569

Recent temperature trends in the southeast Pacific and along the west coast of 570

subtropical South America (1979–2006). Journal of Geophysical Research, 114, 571

D04102. 572

Fletcher, M. S. and P. I. Moreno, 2012: Have the Southern Westerlies changed in 573

a zonally symmetric manner over the last 14,000 years? A hemisphere-wide 574

take on a controversial problem. Quaternary International, 253, 32-46. 575

Fuenzalida H, Aceituno P, Falvey M, Garreaud R, Rojas M, Sanchez R. 2007. 576

Study on Climate Variability for Chile during the 21st century. Technical Report 577

prepared for the National Environmental Committee. In Spanish. Available on-578

line at http://www.dgf.uchile.cl/PRECIS (August 2007). 579

Garreaud, R. D., 2007: Precipitation and circulation covariability in the 580

extratropics. Journal of Climate, 20, 4789-4797. 581

Garreaud, R. and M. Falvey, 2009: The coastal winds off western subtropical 582

South America in future climate scenarios. Int. J. of Climatology, 29, 543-554. doi: 583

10.1002/joc.1716 584

Garreaud, R. D., M. Vuille, R. Compagnucci, and J. Marengo, 2009: Present-day 585

South American climate. Palaeogeography Palaeoclimatology Palaeoecology, 281, 586

180-195. 587

Gillett, N. P. and D. W. J. Thompson, 2003: Simulation of recent Southern 588

Hemisphere climate change. Science, 302, 273. 589

30

Gillett, N., T. Kell, and P. Jones, 2006: Regional climate impacts of the Southern 590

Annular Mode. Geophys. Res. Lett, 33, L23704. 591

Gong, D. and S. Wang, 1999: Definition of Antarctic oscillation index. 592

Geophysical Research Letters, 26, 459-462. 593

Holmlund, P. and Fuenzalida, H., 1995. Anomalous glacier responses to 20 th 594

century climatic changes in the Darwin Cordillera, southern Chile. Journal of 595

Glaciology, 41, 465 - 473. 596

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. 597

Iredell, S. Saha, G. White, J. Woollen, Y. Zhu, A. Leetmaa, R. Reynolds, M. 598

Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K. C. Mo, C. Ropelewski, J. 599

Wang, R. Jenne, and D. Joseph, 1996: The NCEP/NCAR 40-Year Reanalysis 600

Project. Bulletin of the American Meteorological Society, 77, 437-471. 601

Lamy, F., R. Kilian, H. W. Arz, J. P. Francois, J. Kaiser, M. Prange, and T. 602

Steinke, 2010: Holocene changes in the position and intensity of the southern 603

westerly wind belt. Nature Geoscience. 604

Lara, A., R. Villalba, A. Wolodarsky-Franke, J. C. Aravena, B. H. Luckman, and 605

E. Cuq, 2005: Spatial and temporal variation in Nothofagus pumilio growth at 606

tree line along its latitudinal range (35 40′–55 S) in the Chilean Andes. Journal of 607

Biogeography, 32, 879-893. 608

Lopez, P., K. Najib, J. Carrasco and G. Casassa, 2008: Analysis of meteorological 609

data from NCEP/NCAR reanalysis, radiosonde data and surface stations, and 610

its relation with glacier behaviour in Patagonia during the last 50 years. 4th 611

Alexander von Humboldt Intl. Conference. Santiago, Chile, 24-28 Nov 2008. 612

Pp73. 613

Lopez, P., P. Chevallier, V. Favier, B. Pouyaud, F. Ordenes, and J. Oerlemans, 614

2010: A regional view of fluctuations in glacier length in southern South 615

America. Global and Planetary Change, 71, 85-108. 616

31

Marengo, J., R. Jones, L. M. Alves, and M. Valverde, 2009: Future change of 617

temperature and precipitation extremes in South America as derived from the 618

PRECIS regional climate modeling system. International Journal of Climatology, 619

29, 2241-2255. 620

Marshall, G. J., 2003: Trends in the Southern Annular Mode from observations 621

and reanalyses. Journal of Climate, 16, 4134-4143. 622

Marti, O., P. Braconnot, J. L. Dufresne, J. Bellier, R. Benshila, S. Bony, P. 623

Brockmann, P. Cadule, A. Caubel, and F. Codron, 2010: Key features of the IPSL 624

ocean atmosphere model and its sensitivity to atmospheric resolution. Climate 625

Dynamics, 34, 1-26. 626

Matthewman, J. and G. Magnusdottir, 2011: Clarifying ambiguity in 627

intraseasonal Southern Hemisphere climate modes during austral winter. 628

Accepted in J. Geoph. Res. Atmos., Dec. 2011. 629

Miller, A., 1976: The climate of Chile. In: Schwerdtgefer, W. (Ed) World survey of 630

climatology. Vol. 12, Climates of Central and South America, 113-145. Elsevier, 631

Amsterdan. 632

Moller, M., C. Schneider, and R. Kilian, 2007: Glacier change and climate forcing 633

in recent decades at Gran Campo Nevado, southernmost Patagonia. Annals of 634

Glaciology, 46, 136-144. 635

Moreno, P. I., 2004: Millennial-scale climate variability in northwest Patagonia 636

over the last 15 000 yr. Journal of Quaternary Science, 19, 35-47. 637

North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling Errors in 638

the Estimation of Empirical Orthogonal Functions. Monthly Weather Review, 110, 639

699-706. 640

Paruelo, J. M., A. Beltran, E. Jobbagy, O. E. Sala, and R. A. Golluscio, 1998: The 641

climate of Patagonia: general patterns and controls on biotic processes. Ecología 642

Austral, 8, 85-101. 643

32

Peterson, T. C. and R. S. Vose, 1997: An Overview of the Global Historical 644

Climatology Network Temperature Database. Bulletin of the American 645

Meteorological Society, 78, 2837-2849. 646

Prohaska, F., 1976: The climate of Argentina, Paraguay and Uruguay. In: 647

Schwerdtgefer, W. (Ed) World survey of climatology. Vol. 12, Climates of Central 648

and South America, 57-69. Elsevier, Amsterdan. 649

Quintana, J. and P. Aceituno, 2012: Changes in the rainfall regime along the 650

extratropical west coast of South America (Chile). Atmosfera, 25, 1-22 651

Rasmussen, L., H. Conway, and C. Raymond, 2007: Influence of upper air 652

conditions on the Patagonia icefields. Global and Planetary Change, 59, 203-216. 653

Rayner, N., D. Parker, E. Horton, C. Folland, L. Alexander, D. Rowell, E. Kent, 654

and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and 655

night marine air temperature since the late nineteenth century. J. Geophys. Res, 656

108, 4407. 657

Rignot, E., A. Rivera, and G. Casassa, 2003: Contribution of the Patagonia 658

Icefields of South America to sea level rise. Science, 302, 434. 659

Roe, G. H., 2005: Orographic precipitation. Annu. Rev. Earth Planet. Sci., 33, 645-660

671. 661

Rojas, M., P. Moreno, M. Kageyama, M. Crucifix, C. Hewitt, A. Abe-Ouchi, R. 662

Ohgaito, E. C. Brady, and P. Hope, 2009: The Southern Westerlies during the 663

last glacial maximum in PMIP2 simulations. Climate Dynamics, 32, 525-548. 664

Rosenblüth, B., H. Fuenzalida and P. Aceituno, 1997: Recent temperature 665

variations in southern South America. Intl. Journal of Climatology, 17, 67-85. 666

Saha, S., S. Moorthi, H. L. Pan, X. Wu, J. Wang, S. Nadiga, P. Tripp, R. Kistler, J. 667

Woollen, and D. Behringer, 2010: The NCEP climate forecast system reanalysis. 668

Bulletin of the American Meteorological Society, 91, 1015-1057. 669

33

Schneider, C. and D. Gies, 2004: Effects of El Niño–southern oscillation on 670

southernmost South America precipitation at 53° S revealed from NCEP–NCAR 671

reanalyses and weather station data. International Journal of Climatology, 24, 1057-672

1076. 673

Seluchi, M. E., F. A. Norte, P. Satyamurty, and S. C. Chou, 2003: Analysis of 674

three situations of the Foehn Effect over the Andes (Zonda Wind) using the 675

ETA-CPTEC Regional Model. Weather and forecasting, 18, 481-501. 676

Silvestri, G. E. and C. S. Vera, 2003: Antarctic Oscillation signal on precipitation 677

anomalies over southeastern South America. Geophys. Res. Lett, 30, 2115. 678

Smith, R. B. and J. P. Evans, 2007: Orographic precipitation and water vapor 679

fractionation over the southern Andes. Journal of Hydrometeorology, 8, 3-19. 680

Thompson, D. W. J. and J. M. Wallace, 2000: Annular Modes in the Extratropical 681

Circulation. Part I: Month-to-Month Variability*. Journal of Climate, 13, 1000-682

1016. 683

Thompson, D. W. J. and S. Solomon, 2002: Interpretation of recent Southern 684

Hemisphere climate change. Science, 296, 895. 685

Trenberth, K. E., 1981: Interannual Variability of the Southern Hemisphere 500 686

mb Flow: Regional Characteristics. Monthly Weather Review, 109, 127-136. 687

Uppala, S. M., P. Kållberg, A. Simmons, U. Andrae, V. Bechtold, M. Fiorino, J. 688

Gibson, J. Haseler, A. Hernandez, and G. Kelly, 2005: The ERA-40 re-analysis. 689

Quarterly Journal of the Royal Meteorological Society, 131, 2961-3012. 690

Villalba, R., M. Grosjeanb, T. Kiefer, 2009: Long-term multi-proxy climate 691

reconstructions and dynamics in South America (LOTRED-SA): State of the art 692

and perspectives. Palaeogeography, Palaeoclimatology, Palaeoecology, 281, 175-179 693

694

34

Figure captions 695

Figure 1. Topographic map of southern South America (color scale in m ASL), 696

highlighting key features in the Patagonia Region. Black circles indicate the 697

location of the meteorological stations used in this work (precipitation and 698

surface air temperature). White circles indicate the location of the regular 699

upper-air (radiosonde) stations: Puerto Montt and Punta Arenas. The upper 700

inset shows the PRECIS-DGF domain used in this work. 701

Figure 2. Comparison of the PRECIS-DGF climatology (shaded) against station 702

observations (circles filled with the same color scale). The model climatology is 703

computed using the period 1978-2001; the station climatology is calculated 704

using all available monthly data during the second half of the 20th century. 705

Dashed line indicates the ridge of the Andes. (a) Annual mean precipitation; 706

note the logarithmic color scale. (b) Summer (DJF) minus winter (JJA) mean 707

precipitation. (c) Annual mean near surface air temperature (1.5 m in the model; 708

approximately 2m in observations). (d) Summer (DJF) minus winter (JJA) mean 709

near surface air temperature. 710

Figure 3. Time series of observed (grey circles) and PRECIS-DGF (black circles) 711

annual mean near-surface air temperature (upper panel) and precipitation 712

(lower panel) in Puerto Montt (northwestern Patagonia; 41.5°S-72.8°W) and 713

Punta Arenas (southeastern Patagonia, 53.2°S-70.9°W). 714

Figure 4. Local (point-to-point) correlation map between daily precipitation (P) 715

and 850 hPa zonal and meridional wind components (U850; V850) using 716

PRECIS-DGF results from 1980-1990. At each grid point the correlation was 717

calculated for the sample of days with P>1 mm. Colors indicate the P-U850 718

correlation. Vectors are constructed using r(P,U850) and r(P,V850) (scale at the 719

bottom) and only shown where their absolute value exceed 0.3. 720

35

Figure 5. (a) Local (point-to-point) correlation between annual mean 850 hPa 721

zonal wind and precipitation (r(U850,P)) using PRECIS-DGF results from 1978-722

2001. For display purposes the correlation values are shown over the model 723

topography. (b) Longitudinal profile of terrain elevation (shaded area, scale at 724

left), long-term-mean annual precipitation (black line, scale at left in mm/year) 725

and the r(U850,P) correlation, averaged between 42°-52°S. The profiles were 726

constructed every 0.5° of latitude and then composited taken the longitude of 727

highest elevation as a horizontal reference. 728

Figure 6. Scatter plot between monthly precipitation and 850 hPa zonal wind in 729

western and eastern Patagonia. Here we used PRECIS-DGF values from 1978-730

2001 in two 2°x2° boxes centered at 50°S-75°W and 50°S-70°W. The 731

precipitation values were standardized dividing by their respective long term 732

means. 733

Figure 7. Correlation coefficient between annual mean precipitation and 734

collocated 850 hPa zonal wind (from NNR). Precipitation data from surface 735

observations (at least 15 years of data). Closed (open) circles indicate 736

correlations significant (not significant) at the 99% confidence level. Red circles: 737

negative correlation; cyan circles: positive correlation. 738

Figure 8. Local (point-to-point) correlation between seasonal mean 850 hPa zonal 739

wind and surface air temperature, using PRECIS-DGF results from 1978-2001. 740

For display purposes the correlation values are displayed over a 3D 741

topography. The r(U850, SAT) values over ocean are slightly masked since SAT 742

there is strongly controlled by the sea surface temperature which is prescribed 743

in PRECIS-DGF. 744

Figure 9. Leading mode of the 850 hPa zonal (U850) wind interannual 745

variability over southern South America and the adjacent oceans in austral (a) 746

summer and (b) winter, using NCEP-NCAR reanalysis data. The EOF analysis 747

was performed for each month in the area outlined by the solid rectangle (35°-748

36

65°S, 90-60°W) and the resulting principal component (time-series) was 749

projected on the global U850 field. The summer (winter) map was constructed 750

by averaging the D-J-F (J-J-A) values at each grid box. 751

Figure 10. (a) Latitudinal-seasonal variation of the leading mode of the 850 hPa 752

zonal wind (U850) interannual variability averaged between 75-65°W. The EOF 753

analysis was performed for each month in the area outlined in Fig. 10. (b) 754

Linear trends (1968-2001) in U850 averaged between 75-65°W. The gray area to 755

the right shows the mean topography. Results are based on NCEP-NCAR 756

reanalysis. 757

Figure 11. Linear trends in the annual mean zonal wind at 850 hPa (U850) using 758

the (a) ERA-40 and (b) NCEP-NCAR reanalysis. Shading indicates the change 759

between 1968 and 2001 of a linear least squares trend fit calculated at each grid-760

box. The thick (thin) black contour outlined areas where the annual mean U850 761

exceeds 15 m/s (10 m/s). 762

Figure 12. Annual mean precipitation changes in the period 1968-2001 763

congruent with the reanalysis (NNR & ERA40) zonal wind changes (see text for 764

details). (a) Absolute changes in mm/decade. (b) Absolute changes 765

[mm/decade] standardized by the long-term-mean annual precipitation 766

(derived from PRECIS-DGF). The relative changes are not shown in areas 767

where the annual mean precipitation is less than 300 mm/year. 768

37

Figure 1. Topographic map of southern South America (color scale in m ASL), highlighting key 769 features in the Patagonia Region. Black circles indicate the location of the meteorological 770 stations used in this work (precipitation and surface air temperature). White circles indicate the 771 location of the regular upper-air (radiosonde) stations: Puerto Montt and Punta Arenas. The 772 upper inset shows the PRECIS-DGF domain used in this work. 773

38

Figure 2. Comparison of the PRECIS-DGF climatology (shaded) against station observations 774 (circles filled with the same color scale). The model climatology is computed using the period 775 1978-2001; the station climatology is calculated using all available monthly data during the 776 second half of the 20th century. Dashed line indicates the ridge of the Andes. (a) Annual mean 777 precipitation; note the logarithmic color scale. (b) Summer (DJF) minus winter (JJA) mean 778 precipitation. (c) Annual mean near surface air temperature (1.5 m in the model; approximately 779 2m in observations). (d) Summer (DJF) minus winter (JJA) mean near surface air temperature. 780

39

Figure 3. Time series of observed (grey circles) and PRECIS-DGF (black circles) annual mean 781 near-surface air temperature (upper panel) and precipitation (lower panel) in Puerto Montt 782 (northwestern Patagonia; 41.5°S-72.8°W) and Punta Arenas (southeastern Patagonia, 53.2°S-783 70.9°W). 784

40

Figure 4. Local (point-to-point) correlation map between daily precipitation (P) and 850 hPa 785 zonal and meridional wind components (U850; V850) using PRECIS-DGF results from 1980-786 1990. At each grid point the correlation was calculated for the sample of days with P>1 mm. 787 Colors indicate the P-U850 correlation. Vectors are constructed using r(P,U850) and r(P,V850) 788 (scale at the bottom) and only shown where their absolute value exceed 0.3. 789

41

790 Figure 5. (a) Local (point-to-point) correlation between annual mean 850 hPa zonal wind and 791 precipitation (r(U850,P)) using PRECIS-DGF results from 1978-2001. For display purposes the 792 correlation values are shown over the model topography. (b) Longitudinal profile of terrain 793 elevation (shaded area, scale at left), long-term-mean annual precipitation (black line, scale at 794 left in mm/year) and the r(U850,P) correlation, averaged between 42°-52°S. The profiles were 795 constructed every 0.5° of latitude and then composited taken the longitude of highest elevation 796 as a horizontal reference. 797

42

798 Figure 6. Scatter plot between monthly precipitation and 850 hPa zonal wind in western and 799 eastern Patagonia. Here we used PRECIS-DGF values from 1978-2001 in two 2°x2° boxes 800 centered at 50°S-75°W and 50°S-70°W. The precipitation values were standardized dividing by 801 their respective long term means. 802

43

Figure 7. Correlation coefficient between annual mean precipitation and collocated 850 hPa zonal 803 wind (from NNR). Precipitation data from surface observations (at least 15 years of data). 804 Closed (open) circles indicate correlations significant (not significant) at the 99% confidence 805 level. Red circles: negative correlation; cyan circles: positive correlation. 806 807

44

Figure 8. Local (point-to-point) correlation between seasonal mean 850 hPa zonal wind and 808 surface air temperature, using PRECIS-DGF results from 1978-2001. For display purposes the 809 correlation values are displayed over a 3D topography. The r(U850, SAT) values over ocean are 810 slightly masked since SAT there is strongly controlled by the sea surface temperature which is 811 prescribed in PRECIS-DGF. 812

45

Figure 9. Leading mode of the 850 hPa zonal (U850) wind interannual variability over southern 813 South America and the adjacent oceans in austral (a) summer and (b) winter, using NCEP-814 NCAR reanalysis data. The EOF analysis was performed for each month in the area outlined by 815 the solid rectangle (35°-65°S, 90-60°W) and the resulting principal component (time-series) was 816 projected on the global U850 field. The summer (winter) map was constructed by averaging the 817 D-J-F (J-J-A) values at each grid box. 818

46

819 Figure 10. (a) Latitudinal-seasonal variation of the leading mode of the 850 hPa zonal wind 820 (U850) interannual variability averaged between 75-65°W. The EOF analysis was performed for 821 each month in the area outlined in Fig. 10. (b) Linear trends (1968-2001) in U850 averaged 822 between 75-65°W. The gray area to the left shows the mean topography. Results are based on 823 NCEP-NCAR reanalysis. 824

47

Figure 11. Linear trends in the annual mean zonal wind at 850 hPa (U850) using the (a) ERA-40 825 and (b) NCEP-NCAR reanalysis. Shading indicates the change between 1968 and 2001 of a 826 linear least squares trend fit calculated at each grid-box. The thick (thin) black contour outlined 827 areas where the annual mean U850 exceeds 15 m/s (10 m/s). 828

48

Figure 12. Annual mean precipitation changes in the period 1968-2001 congruent with the 829 reanalysis (NNR & ERA40) zonal wind changes (see text for details). (a) Absolute changes in 830 mm/decade. (b) Absolute changes [mm/decade] standardized by the long-term-mean annual 831 precipitation (derived from PRECIS-DGF). The relative changes are not shown in areas where 832 the annual mean precipitation is less than 300 mm/year. 833