Climatic control of the surface mass balance of the Patagonian Iceﬁelds

. The Patagonian Iceﬁelds (Northern and Southern Patagonian Iceﬁelds (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) Iceﬁeld) are the largest ice masses in the Andes Cordillera. Despite its importance, little is known about the main mechanisms that underpin the interaction between these ice masses and climate. Furthermore, the nature of large-scale climatic control over the surface mass variations of the Patagonian Iceﬁelds still remains unclear. The main aim of this study is to understand the present-day climatic control of the surface mass balance (SMB) of the Patagonian Iceﬁelds at interannual timescales, especially considering large-scale processes. 5 We modeled the present-day (1980-2015) glacioclimatic surface conditions for the southern Andes Cordillera by statistically downscaling the output from a regional climate model (RegCMv4) from a 10 km spatial resolution to a 450 m resolution grid, and then using the downscaled ﬁelds as input for a simpliﬁed SMB model. Series of spatially averaged modeled ﬁelds over the Patagonian Iceﬁelds were used to derive regression and correlation maps against ﬁelds (cid:58) of (cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58)(cid:58) climate


Introduction
The Patagonian Icefields (Northern Patagonian Icefield (NPI) and Southern Patagonian Icefield (SPI)) are the most extensive ice bodies in the Andes Cordillera.Given their size, they play a significant role in modulating the local and regional environment, providing ecosystem processes such as climate regulation, gas regulation, and hydrologic cycles regulation, among others (Martínez-Harms and Gajardo, 2008;Dussaillant et al., 2012).Both icefields have been losing mass over the last decades (Rignot et al., 2003;Malz et al., 2018;Minowa et al., 2021), and recent evidence shows that they are the primary contributors to sea-level rise among all South American ice masses (Braun et al., 2019;Dussaillant et al., 2019).Overall, glaciers of the Southern Andes have contributed approximately 3.3 mm of sea-level rise between 1961 and 2016 (Zemp et al., 2019).Despite the importance of the Patagonian Icefields, little is known about the main mechanisms underpinning the interaction between these ice bodies and climate, especially the large-scale climate processes that determine their surface mass balance (SMB) at interannual timescales.This topic represents a significant issue for understanding the Patagonian Icefields' past, present, and future evolution and, more generally, the southern Andean cryosphere.
The Patagonian climate is controlled primarily by the strength of the westerly winds (Garreaud et al., 2013).Garreaud et al. (2013) find a high correlation between zonal wind and precipitation in western Patagonia at daily, monthly, and interannual timescales.They also find a seasonal correlation between zonal wind and temperature, indicating that windy summers tend to be colder than average and windy winters tend to be warmer than average.Consequently, modes of variability affecting the westerly flow impact the Patagonian climate profoundly, such as the Southern Annular Mode (SAM), the leading mode of extratropical Southern Hemisphere variability (Fogt and Marshall, 2020, for a review).This mode is characterized by an equivalent barotropic, zonally symmetric structure involving exchanges of mass between the mid and high latitudes with positive polarity associated with a strengthening and poleward shifting of the polar jet and negative polarity associated with a weakening and equatorward shift of the polar jet (Rogers and Van Loon, 1982;Thompson and Wallace, 2000).Additionally, subsidence and adiabatic warming occur in the troposphere on the equatorward side of the polar jet during the positive phase of SAM, while opposite temperature anomalies maintain during the negative phase (Fogt and Marshall, 2020).
Furthermore, Cai et al. (2020) found slightly different spatial patterns of precipitation anomalies for Central Pacific ENSO events and Eastern Pacific ENSO events (Capotondi et al., 2015;Timmermann et al., 2018).The diversity of spatial patterns and intensities of SST anomalies in the tropical Pacific Ocean among ENSO events results in different atmospheric circulation responses (Taschetto et al., 2020), which in turn would affect the linkage between ENSO and Patagonian climate.Thus, even though dryer :::: drier and warmer than normal conditions are expected during El Niño events, especially during summer, the net effect of ENSO on the Patagonian climate seems to depend on the specifics of each ENSO event.
The meteorological conditions over the Patagonian Icefields have a direct impact on the glaciological surface processes (e.g., snowfall, surface melting) that determine the gain (accumulation) and loss (ablation) of mass experienced by the glaciers.The SMB corresponds to the overall sum of the surface accumulation and surface ablation, i.e., the net surface gain of mass :::::: change :: of :::: mass :: at ::: the ::::::: surface, over a certain period of time.Processes affected by the specific glacier dynamics, such as calving, or basal conditions, such as basal melting, represent possible sources of additional gain and loss of mass and determine, along with the SMB, the total mass balance of the glaciers, which, in turn, modulates their geometry.Unlike the total mass balance or the glacier geometry, the SMB integrates the direct interplay between glaciers and climate and thus represents a suitable study variable for assessing climate-cryosphere interaction.
Similar to many mountain regions in the world, the Patagonian Icefields show a lack of in situ climatic and glaciologic measurements due to difficult access and harsh environmental conditions.There are few measurement stations and short records available (Fig. S1), and this scenario hinders ::::: which :::::: hinder a robust assessment of glacier response to current climate conditions.
Due to the inadequate observational network, various studies have tried to quantify the SMB of the Patagonian Icefields using different global gridded climate datasets (i.e., reanalysis), downscaling techniques (dynamical and statistical downscaling procedures), and SMB models of different complexity (Schaefer et al., 2013(Schaefer et al., , 2015;;Lenaerts et al., 2014;Mernild et al., 2017).Interestingly, all studies found positive trends in the SMB of the Patagonian Icefields and a positive SMB for the SPI.
Nonetheless, none of them assess ::::::: assesses the interannual variability of the SMB nor its relationship with local and large-scale atmospheric processes.
Motivated by improving our knowledge about the climate-cryosphere interplay, this paper links the annual anomalies in the SMB of the Patagonian Icefields with local, regional, and large-scale climate anomalies.The main goal of this work is to understand the present-day climatic control of the SMB of the Patagonian Icefields at interannual timescales, especially considering large-scale processes.Understanding the mechanisms behind year-to-year changes in the SMB is an essential requirement for deepening the comprehension of the climate processes responsible for past, present, and future trends of the SMB of the Patagonian Icefields and an important opportunity for future development of diagnostic and prognostic tools.
To achieve our goal, we :::: first simulate present-day glacioclimatic surface conditions for the southern Andes Cordillera using a simplified SMB model forced with a high-resolution regional climate model simulation. 2 Study Area, Data, and Methodology

Study Area
The study area comprises the Patagonian Icefields.They spread over a latitudinal band of 46-52º S and include the NPI and the SPI (Fig. 1).The NPI locates between 46º30' S and 47º30' S and covers a total ice area of 3953 km 2 (Rivera et al., 2007).
It elongates in the north-south direction with an axis near the 73º30' W, extending ∼100 km in length and 40-45 km in width (Aniya, 1988).It shows a steep topography with terrain elevation values increasing eastward in most parts of the icefield area, reaching the sea level at the west margin and a maximum of 3970 m a.s.l.: (above sea level) .at the summit of Mount San Valentín.Characteristic terrain elevation values are 1000 m a.s.l. for the west side and 1500 m a.s.l. for the east side (Warren and Sugden, 1993).The NPI is composed of 38 glaciers larger than 0.5 km 2 (Dussaillant et al., 2018).The SPI locates between 48º20' S and 51º30' S and covers a total ice area of 12514 km 2 (Casassa et al., 2014).It extends ∼350 km in length and generally 30-40 km in width, with the narrowest part only 8 km wide (Aniya et al., 1998).This icefield contains a central plateau lying between the 1400-2000 m a.s.l. with terrain elevation values decreasing southward.The SPI reaches its topographic maximum at Volcán Lautaro with a peak of 3607 m a.s.l.It is composed of 48 main outlet glaciers (Aniya et al., 1998). 135
To assess large scale patterns associated with SMB anomalies, data from several climatic variables for the period between 1980-2015 were taken from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA-Interim ) dataset with a grid spacing of 0.75º × 0.75º (Dee et al., 2011) including surface air temperature (SAT), zonal and meridional wind (u and v, respectively), mean sea level pressure (MSLP), geopotential height (Z), and SST.Also, outgoing longwave radiation (OLR) flux at the top of the atmosphere data was taken from the NOAA Climate Data Record (CDR) Program, with a grid spacing of 2.5º x 2.5º (Lee and NOAA CDR Program, 2011).

Statistical downscaling of the RegCMv4 output
The RegCMv4 DEM digital elevation model ::::: Digital :::::::: Elevation :::::: Model : (DEM) tends to underestimate the terrain elevation when compared with the SRTMv3 DEM, especially at higher elevations, both in the NPI and the SPI (Fig. S2 ::: S5).In order to avoid biases in near-surface temperature and precipitation due to elevation biases, we corrected the near-surface temperature and precipitation RegCMv4 model output accounting for the biases in the RegCMv4 DEM.
To do so, we first constructed a DEM resulting from the averaged SRTMv3 DEM at every five grid points (∼450 m spatial resolution) and used this as the default model DEM.This spatial resolution was selected after performing a sensitivity analysis (Carrasco-Escaff, 2021).We then statistically downscaled the RegCMv4 main surface atmospheric output (near-surface air temperature, precipitation, and surface downward solar radiation) from 10 km spatial resolution to the 450 m resolution grid.
Regarding the temporal resolution, each field remained at a 3 h resolution.
The statistically downscaling process started with the bilinear interpolation of the fields onto the RegCMv4 DEM to remap data from a 10 km resolution to a 450 m grid resolution of the default model DEM.Then, we performed altitudinal corrections for temperature and precipitation.In the case of temperature, we applied a constant lapse rate equal to the environmental lapse rate (6.5 ºC km −1 ).In this way, we computed the near-surface temperature every 3 h according to: where T is the downscaled near-surface temperature, T bil is the 450 m bilinearly interpolated RegCMv4 near-surface temperature, LR = 6.5 ºC km −1 is the lapse rate, z is the model reference DEM, and z bil is the 450 m bilinearly interpolated RegCMv4 DEM.In the case of precipitation, we used the following equation at every 3 h: where P is the statistically downscaled precipitation, P bil is the 450 m bilinearly interpolated RegCMv4 precipitation, and PG = 0.05% m − 1 is the precipitation gradient (as in Schaefer et al., 2013Schaefer et al., , 2015)).Finally, we downscaled the :::::::: remapped ::: the ::::::: original :::::::: RegCMv4 : surface downward solar radiation :: to ::: the ::: 450 : m ::: grid : by performing bilinear interpolationon the RegCMv4 original field.
The SMB model assigns one of three types of soil :::::: surface : to every grid cell: snow, firn, and ice.Each type of soil :::::: surface has a specific albedo (snow albedo is 0.85, firn albedo is 0.55, and ice albedo is 0.35), and they : .::: The :::::: albedo :::::: values were taken from Cuffey and Paterson (2010) ::: and :::::::::: correspond :: to ::: the :::::::::::: recommended :::::: values ::: for :::: fresh ::: dry :::::: snow, :::: clean :::: firn ::: and ::::: clean ::: ice.In the SMB model, every grid cell consists of a column of ice at the bottom, possibly followed by a column of firn and possibly by a column of snow.At every 3 h and for each grid cell, the SMB model calculates the accumulation added to the column of snow.Then, the SMB model computes the ablation, and the model simulates the melting of the (possible) snow, followed by the (possible) firn and the ice.Finally, the SMB is computed at every 3 h and for each grid cell according to: At the end of each summer ::: start ::: of :::: each :::::: autumn : season (1 April), the mass of firn in the firn column turns into ice, and the mass of snow in the snow column turns into firn.Initially, each grid cell consists of only a column of ice (infinitely deep), and the SMB model was forced with the downscaled climatological conditions obtained from the RegCMv4 for five years before feeding it with the true :::: actual : RegCMv4 downscaled fields.

Large-scale climate indices
We determined the correlation between the modeled time series and several large-scale climate indices in order to assess the influence of the main modes of interannual variability on the SMB of the Patagonian Icefields.To characterize ENSO activity we used the monthly values of SST averaged over the Nino1+2 and the Nino3.4regions (see Fig. 4) obtained from the NOAA Climate Prediction Center (CPC, https://psl.noaa.gov/data/climateindices/list/).Additionally, we used the Central Pacific (CP) and Eastern Pacific (EP) ENSO indices to account for ENSO diversity (Kao and Yu, 2009;Yu et al., 2012).To obtain the spatial pattern and temporal index of EP ENSO, SST anomalies regressed with the Nino4 index are removed from total SST anomalies before performing Empirical Orthogonal Function (EOF) analysis.The same approach is used for computing the CP index but using the Nino1+2 index instead.To characterize SAM activity we used the AAO index obtained from NOAA CPC.We used monthly means of daily values, which are constructed by projecting the daily height anomalies at 700 hPa poleward 20º S onto the leading mode of EOF analysis of monthly mean 700 hPa height during the 1979-2000 period.:: the ::::: SMB ::::: time ::::: series :::::::: computed :::: from :::: each :::::::::: experiment :: at ::::: winter :::: and ::::::: summer ::::::::: timescales.
Regions used for the construction of climate indices.
Additionally, we constructed custom indices in order to assess the covariability between the modeled variables and some climatic variables averaged over specific regions of interest (Fig. 4).We spatially averaged the monthly values of the ERA-Interim geopotential height at 300 hPa and air temperature at 850 hPa in a box near the Drake Passage spanning the 68-53º S in latitude and 100-60º W in longitude (box R1 in Fig. 4).We did the same with the SST of the Pacific Ocean next to central Patagonia (box R2 in Fig. 4, at 52-46º S and 80-76º W) and with the zonal wind at 850 hPa impinging central Patagonia (box R3 in Fig. 4, at 52-46º S and 75.5-74.5ºW).We named these time series Z300 Drake, T850 Drake, SST-R2, and U850-R3, respectively.3 Results
Results show that the annual SMB is highly and positively correlated with the annual accumulation (r = 0.87 * ) and negatively correlated with the annual ablation (r = −0.69* ).During winter, the correlation between the SMB and the accumulation increases (r = 0.94 * ), while the correlation between SMB and the ablation decreases and becomes statistically non-significant.
First, we assessed the local-scale control over the SMB from the results of the sensitivity experiments described in Sect. 1.We use the squared correlation R 2 metric to measure the degree of dependence of the SMB interannual variability on a specific meteorological variable.A high (low) value of R 2 is interpreted as a low (high)
Precipitation exerts the primary control at annual and winter timescales, and the temperature does it at the summer timescale.
For simplicity, we analyze a year when the SMB is one standard deviation above the mean value(the analysis extends linearly to other cases).An : , ::::: which ::::::::::: corresponds :: to :: an : annual SMB anomaly of 537 mm w.e.::: (the ::::::: analysis ::::::: extends ::::::: linearly :: to :::: other :::::: cases).:::: Such :: a :::: year is associated with an annual accumulation anomaly of 351 mm w.e. and an annual ablation anomaly of -186 mm w.e.Regarding seasonal differences, years with relatively large SMB are associated with higher than average winter and summer accumulation values and lower than average summer ablation values.The winter accumulation anomaly is 1.15 times the summer accumulation anomaly but the summer ablation anomaly is more than 14 times the winter ablation anomaly.
As a result, when grouping contributions by process, the annual SMB anomalies are primarily explained by accumulation anomalies, while when grouping by season, the summer anomalies in the glaciological processes account for most of the annual SMB anomalies.
The same analysis (i.e., the SMB is one standard deviation above the mean value) yields an increase of winter precipitation of 212 mm w.e., whereas an almost null and not statistically significant variation in winter temperature.During summer, there is an increase of precipitation of 156 mm w.e. and a variation in temperature of -0.23 ºC.Thus, our results suggest that years with higher than average SMB are related to wetter than normal annual conditions and colder than normal summer conditions, while years with lower than average SMB are associated with the opposite.

Regional-scale control over the SMB
To assess the regional scale control over the SMB, we first computed the regression of the annual SMB :::::::: anomalies : with the annual precipitation, near-surface temperature, and horizontal wind (at 10 m above ground level and 700 hPa).Results are shown in Fig. 6a :: 5a, c.For simplicity, we analyze the years when the SMB is above the mean value (the analysis extends linearly to other cases).Positive anomalies of annual SMB are associated with an intensification of the westerly winds impinging the Austral Andes, a regional cooling in the south of South America and over the Pacific Ocean adjacent to Patagonia, and an increase (decrease) of the precipitation to the west (east) of the Andean ridge.The cooling is stronger over the Pacific Ocean adjacent to central and north Patagonia and northeast of the Patagonian Icefields.The increase in precipitation reaches the highest values in central-western Patagonia, with a maximum over the Patagonian Icefields.Over the Pacific Ocean adjacent to Patagonia, the circulation acquires anticyclonic vorticity to the north and cyclonic vorticity to the south, both at the nearsurface level and at the 700 hPa pressure level.Some differences in horizontal wind anomalies are evident when comparing near-surface level and at 700 hPa pressure level due to the topographic blocking imposed by the Andes.
We also computed latitudinal profiles of regressions of the annual SMB with the mean annual fields of zonal wind, geopotential height, and air temperature at a longitude of 80 ºW.Results are shown in 6b ::: and maximizing near 50 ºS.At altitude, the positive zonal wind anomaly extends throughout the entire troposphere and reaches its maximum around 300 hPa, between cores of high and low anomalous geopotential height, in a region where the pressure gradient is maximum.In turn, these cores of anomalous geopotential height are located in regions where the magnitude of the temperature gradient is maximum, resembling a thermal wind balance.Interestingly, the anomalous cold region below the core of low anomalous geopotential height extends to the lower troposphere and comprises the latitudinal band where the Patag- onian Icefields are located.This suggests that during years of relatively high SMB, the reinforcement of the westerlies wind impinging Patagonia and the temperature anomaly observed in the Patagonian Icefields could be linked to the same mechanism.
Regarding seasonal differences, years with SMB above the average shows :::: show a stronger circulation and a more pronounced precipitation change during winter than summer (Figs. 7, 8 : 6 :::: and : 7).Also, these years are associated with a pronounced summer cooling over the south of South America and the adjacent Pacific Ocean (Fig. 8c :: 7c), while correlations with winter near-surface temperature are virtually null.The latitudinal profiles also show a stronger reinforcement of the westerly winds during winter than summer (Fig. 7b :: 6b), associated with more pronounced cores of anomalous geopotential height (Fig. 7d :: 6d).Nonetheless, during summer, the low-pressure structure appears displaced northward, especially in the lower troposphere (Fig. 8d :: 7d).A more intense cooling of the anomalous cold region tends to concentrate in the lower troposphere (1000 to 700 hPa), which could explain the summer cooling observed along the Patagonian Icefields.

Large-scale control over the SMB
To assess the large-scale control over the SMB, we computed the regression fields of several climatic variables (at annual, winter, and summer timescales) onto the annual time series of the spatially averaged field of SMB ::::: SMB ::::::: anomaly :::: time ::::: series.
Years with SMB above the average are characterized by the presence of an anomalous low-pressure center located around the Drake Passage (hereafter Drake low) with a longitudinal extension from the northeastern Amundsen Sea and northeastern Antarctic Peninsula (∼120ºW to 50ºW), and a latitudinal extension from the west Antarctic coast to the southern tip of South America (Fig. 9a :: 8a).Around the Drake low, anomalous high-pressure centers are established over the subtropical South Pacific, extending towards the Amundsen Sea and the South Atlantic.
The Drake low is associated with an anomalous cyclonic circulation established around the Drake Passage (Fig. 9b :: 8b).
A strengthening of the annual zonal winds in the latitudinal band comprising the Patagonian Icefields and the longitudinal band comprising the 60-120 ºW is observed, while a weakening of the zonal wind is exhibited southward.Furthermore, an intensification of the trade winds is also observed over the central equatorial Pacific, with magnitudes comparable to the ones exhibited by the westerly winds impinging the Patagonian Icefields.
Regarding SST anomalies :::: (Fig. ::: 8a), positive anomalies of annual SMB are associated with a surface cooling off the coast of Patagonia, in accordance with the regional-scale analysis (see Fig. 6 : 5).A large-scale cooling is observed around the centraleastern equatorial Pacific and the west coast and southern tip of South America, resembling an Eastern Pacific La Niña-like pattern (Fig. S5 ::: S10).Nonetheless, this pattern is latitudinally asymmetric with respect to the equator, and the strongest SST correlations are associated with off-equatorial tongues off the coast of South America.In addition, beneath the anomalous cold tongue around the equatorial Pacific, an anomalous warm tongue emerges from the western equatorial Pacific towards the subtropical South Pacific.
Years with SMB above the average are associated with positive OLR anomalies over the central equatorial Pacific (i.e., decreased convective activity) and negative OLR anomalies over the western equatorial Pacific (Fig. 9c :: 8c), consistent with the SST patterns.These OLR anomalies are accompanied by anomalies of geopotential height at 300 hPa that account for both (i)     High correlation values are found between the Z300 Drake index and the modeled time series.For instance, there is a strong and statistically significant correlation between Z300 Drake and the SMB time series at annual (r = −0.65 * ), winter (r = 540 −0.66 * ), and summer (r = −0.54* ) timescales.Furthermore, the Z300 Drake index is highly correlated with the precipitation time series at all timescales, while there is a statistically significant correlation with the temperature time series only in summer (r = 0.42 * ).Additionally, the T850 Drake index shows a strong correlation with the SMB time series at all timescales and a  significance ::: level :: of :::: 5%.high correlation with the near-surface temperature time series during summer (r = 0.63 * ).The highest correlation between the SMB time series and the set of indices explored is maintained with the U850-R3 index, at annual (r = 0.71 * ), winter (r = 0.85 * ), and summer (r = 0.59 * ) timescales.
We found only weak correlations between the SMB and atmospheric modes of variability, such as the El Niño-Southern Oscillation (ENSO) and the Southern Annular Mode (SAM), implying little dependency between these modes and the SMB of the Patagonian Icefields (Table 5 : 2).We highlight that this result characterizes the present-day long-term  linear relationship between the annual variability of atmospheric modes and the SMB.One single event may profoundly impact the mass balance of the Patagonian Icefields (see for example Gómez et al. (2022)), which agrees with our results as long as the long-term linear relationship maintains :::::: remains : weak.Our results suggest that the low (high) pressure anomalies located over the Amundsen-Bellingshausen Sea during Central Pacific La Niña (El Niño) events (see Yuan et al. (2018)  This means that for the SAM index the two observed processes tend to cancel each other out in developing SMB anomalies.
This study does not assess the teleconnections that potentially trigger the Drake low.However, we speculate that the origin of this pressure feature might be associated with tropical forcing due to the decreased convective activity over the central equatorial Pacific and increased convective activity over the western equatorial Pacific :::::::::::::::::::::::::::::::::::::: (e.g., Hoskins and Karoly, 1981;Karoly, 1989) observed during years of relatively high SMB (Fig 9c :: 8c).As the SMB correlates better to the Eastern Pacific ENSO index than the Central Pacific ENSO index (Table 5 : 2), we argue that the establishment of the Drake low would be highly sensitive to the specific location of SST anomalies in the tropical Pacific.The low correlation between the Eastern Pacific ENSO index and the SMB could be a consequence of similar considerations : in ::: the ::::: sense : that only certain eastern Pacific SST warming and cooling events could activate an anomalous pressure center near the Drake Passage.Additionally, we conjecture that the summer cooling of the Patagonian Icefields during years of relatively high SMB is mainly associated with the thermodynamics of the Drake low, as exposed previously, and not with the eastern Pacific SST cooling.This seems reasonable since nearly 40% of the variance of the summer temperature over the Patagonian icefields is explained by the lower tropospheric temperature near the Drake Passage (Table 5 : 2).
-Concerning the large-scale conditions, years of relatively high SMB ::: are characterized by the establishment of an anomalous low-pressure :::::::::: low-pressure : center near the Drake Passage, the Drake low, that induces an anomalous cyclonic circulation accompanied with :: by : enhanced westerlies impinging the Patagonian Icefields.The Drake low is thermodynamically maintained by a core of cold air that cools the Patagonian Icefields during summer.Years with lower than average SMB are associated with the opposite conditions.
-We found little dependency between the interannual SBM ::::: SMB variability of the Patagonian Icefields and main atmospheric modes of variabilities such as SAM and warm and cold ENSO phases.

Figure 1 .
Figure 1.(a) Satellite image of Northern Patagonian Icefield and Southern Patagonian Icefield taken by the MODIS sensor on board the NASA's TERRA satellite on February 19, 2011.(b) Terrain elevation (m a.s.l.) of southern South America obtained from the digital elevation model ETOPO1 with 1 minute of arc resolution.The black box spans the area of panel (a).(c) Schematic of the main features of largescale circulation near the Patagonian Icefields.The red polygon indicates the spatial domain used for running the RegCMv4 present-climate simulations.The black box spans the area of panel (b).

Figure 2 .
Figure 2. (a) Patagonian Icefields together with their glacier divides and type of terminus.(b) The RegCMv4 grid (∼10 km spatial resolution), the model land use (grid box colors) and the NPI and SPI outlines (blue contours).(c) Comparison of annual mean temperature time series (box at 46-52º S and 72.5-74.5ºW) using RegCMv4 data and CR2MET data.(d) Same as (c) but for accumulated annual precipitation.

Figure 5 .
Figure 5. Regional correlation and linear regression maps of the annual (April to March) time series of the spatially averaged field of SMB with fields obtained from the RegCMv4 simulation data and ERA-Interim reanalysis.a) Regression with annual fields of horizontal wind at 700 hPa (vectors in ms −1 /std.dev.) and correlation with accumulated precipitation (colors).Fields were obtained from the RegCMv4 data.b) Latitudinal and atmospheric profile of the regression with annual field of zonal wind (contours in ms −1 /std.dev.) for a transect at 80 ºW.Negative regression values are shaded, and the Andes topography within the latitudinal band of the Patagonian Icefields is shown with blue lines.Fields were obtained from the ERA-Interim reanalysis.c) Regression with annual fields of horizontal wind at 10 m above ground level (vectors in ms −1 /std.dev.) and correlation with the mean near-surface air temperature (colors).Fields were obtained from the RegCMv4 data.d) Latitudinal and atmospheric profile of the regression with annual field of geopotential height (contours in gpm/std.dev.) and temperature (colors in ºC/std.dev.) for a transect at 80 ºW.Negative regression values are shaded, and the ::: The Andes topography within the latitudinal band of the Patagonian Icefields is shown with blue lines.Fields were obtained from the ERA-Interim reanalysis.

Figure 6 .
Figure6.The same as in Fig.6: 5 : but for the winter (April to September).

Figure 7 .
Figure7.The same as in Fig.6: 5 : but for the summer (October to March).

3. 6 Figure 8 .
Figure 8. Large-scale correlation and linear regression maps of the annual (April to March) time series of the spatially averaged field of SMB with fields obtained from the Era-Interim reanalysis.a) Regression with the annual field of mean sea level pressure (contours in hPa/std.dev.) and correlation with the annual field of sea surface temperature (colors over ocean) and near-surface air temperature (colors over land).b)Regression with the annual field of horizontal wind at 850 hPa (vectors in ms −1 /std.dev.) and correlation with the annual field of zonal wind at 850 hPa (colors).c) Regression with the annual field of geopotential height at 300 hPa (contours in gpm/std.dev.) and correlation with the annual field of outgoing longwave radiation (colors).

AnnualFigure 9 .
Figure9.The same as in Fig.9: 8 : but for the winter (April to September).

AnnualFigure 10 .
Figure10.The same as in Fig.9: 8 but for the summer (October to March).
::::::: anomalies :: of ::: the :::::: annual ::::: SMB ::: are :::::::: associated :::: with : an intensification of the westerly winds impinging the Patagonian Icefields and an increase of the precipitation in western Patagonia accompanied with relatively dry dryer :: by ::::: drier conditions to the east of the Andes ridge.Higher than average SMB years indicate a :: A regional decrease in summer near-surface temperatures , while no or little winter : is :::::::: observed ::::: during :::::::: summer, ::::: while ::: null ::: or :::: little temperature changes are evident .Years with relatively low SMB show ::::: during :::::: winter.::::::: Negative ::::::::: anomalies :: of ::: the :::::: annual ::::: SMB ::: are ::::::::: associated :::: Further work is required to understand the low annual correlation between EP ENSO index and the SMB of the Patagonian Icefields.This research study gives new insights for understanding the complex interplay between the present-day climate processes and local-scale cryospheric processes in the southern Andean Cordillera.Low dependence of the Patagonian Icefields' SMB on main atmospheric modes of variability suggests a poor ability of ENSO and SAM indexes to reproduce the past and future interannual variability of the SMB.Instead, this study highlights the Drake Passage as a key region capable of reproducing the interannual variability of the SMB since it explains the linkage between large-scale processes and the SMB behavior reasonably.Finally, findings from local-scale assessment facilitate the diagnostic of SMB anomalies in terms of precipitation, near-surface air temperature, and surface downward solar radiation anomalies, providing a conceptual framework useful for future research in the area.