Introduction
Recent conclusions of Working Group I of the Intergovernmental Panel on Climate Change (IPCC; Reference SolomonSolomon and others, 2007) indicate that instrumental meteorological records show an overall warming around the globe, as well as changes in precipitation regimes. There is very high confidence now that a major part of this warming is due to anthropogenic emission of greenhouse gases (IPCC; Reference SolomonSolomon and others, 2007). The warming has triggered physical and biological changes in the Earth’s system, such as global sea-level rise, decrease of the Arctic sea-ice extent and thickness, generalized retreat and areal reduction of glaciers in the Arctic and in mountain regions around the world, areal decrease of winter snow cover in the Northern Hemisphere (Reference Lemke and SolomonLemke and others, 2007), earlier spring runoff due to enhanced snowmelt, enhanced glacial runoff, warming of lakes and upward and northward migration of biological species in the Northern Hemisphere (Reference Rosenzweig, Canziani, Parry and PalutikofRosenzweig and others, 2007).
A general trend of tropospheric warming and precipitation changes is also observed in Chile (e.g. Reference Aceituno, Fuenzalida, Rosenblüth, Mooney, Fuentes and KronbergAceituno and others, 1993; Reference Rosenblüth, Casassa and FuenzalidaRosenblüth and others, 1995, Reference Rosenblüth, Fuenzalida and Aceituno1997; Reference Carrasco, Casassa, Rivera, Casassa, Sepúlveda and SinclairCarrasco and others, 2002, Reference Carrasco, Casassa and Quintana2005; Reference QuintanaQuintana, 2004). A strong retreat and areal reduction of mountain glaciers has been documented in the Chilean Andes (e.g. Reference CasassaCasassa, 1995; Reference Rivera, Casassa, Acuña and LangeRivera and others, 2000, Reference Rivera, Benham, Casassa, Bamber and Dowdeswell2007; Reference Bown and RiveraBown and Rivera, 2007; Reference CasassaCasassa and others, 2008). A positive trend in the elevation of the equilibrium-line altitude (ELA), during the last quarter of the 20th century in central Chile, has been reported by Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) and Reference OsorioOsorio (2006).
Figure 1 summarizes available mean annual variations of frontal positions of several glaciers along the Chilean Andes (see Fig. 2 for geographical location). Most of the glaciers show a retreat of the frontal tongue, with maximum recession and thinning in Hielo Patagónico Norte and Sur (northern and southern Patagonian Icefields) (Reference Rignot, Rivera and CasassaRignot and others, 2003). The overall warming and changes in precipitation along the Andes are believed to be responsible for the observed glacier variations.
Mean annual frontal variation of glaciers along the western side of the Andes. Data are derived from aerial photography and satellite imagery acquired at different epochs (Reference Rivera, Anuña, Casassa and BownRivera and others, 2002). The grey horizontal bars indicate the latitudinal position of Hielo Patagónico Norte (HPN) and Hielo Patagónico Sur (HPS). The black curve is an exponential regression of all data points which shows an overall retreat of glaciers along the Chilean Andes.
Location of Chilean radiosonde stations (filled triangles).
Here we use meteorological information available from Chilean radiosonde stations in combination with data from surface stations as input within a simple empirical model to derive and analyze the behaviour of the ELA as a proxy for assessing glacier variations. This study extends the analysis carried out in the Andes of central Chile by Reference Carrasco, Casassa and QuintanaCarrasco and others (2005). In this work the annual average of the 0°C isotherm altitude derived from the vertical air-temperature profiles is used, along with precipitation records from surface stations, to determine the annual ELA and its variability at the four radiosonde sites available in Chile. These sites represent the northern (∼18–30° S), central (∼30–38° S), southern (∼38–46° S) and austral (∼46–55° S) regions of Chile, referred to as regions 1–4, respectively, in this work. An overview and an update of the near-surface air-temperature and precipitation pattern in Chile over the past 45–50 years are also given.
Data Analysis
Radiosonde data
Daily upper-level meteorological observations started in Chile during the International Geophysical Year (IGY, 1957/58), when radiosonde stations were deployed at Antofagasta (23°26′ S, 70°26′ W; 135 m a.s.l.), Quintero (32°47′ S, 71°33′ W; 8 m a.s.l.) and Puerto Montt (41°26′ S, 73°07′ W; 81 m a.s.l.). In 1972 a fourth station was started at Punta Arenas (53°00′ S, 70°07′ W; 37 m a.s.l.). In 1999, the aerological station in Quintero was moved ∼100 km south, to its new location at Santo Domingo (33°38′ S, 71°18′ W; 77 ma.s.l.). Almost continuous time series of tropospheric and stratospheric temperatures are available since 1958 for the first three stations and since 1973 for Punta Arenas.
Upper-air temperatures above the boundary layer at 850 hPa for the four radiosonde stations show a warming over the 1958–2006 period as summarized in Table 1. A slight cooling is observed only in summer at Punta Arenas, although it is not statistically significant. A climate shift in 1976/77 can be observed at 850 hPa as well as at 700 and 500 hPa (Reference Boiser and AceitunoBoiser and Aceituno, 2006), revealing the horizontal and vertical dimension of the changes associated with the Pacific Decadal Oscillation (PDO) in the southeastern region of the Pacific Ocean along western South America.
Secular trend of the air temperature at 850 hPa (in °C (10 a)−1) for periods 1958–2006/1978–2006 for all stations, except Punta Arenas which only covers the 1975–2006 period
Surface temperature
Assuming a linear trend in the annual mean air temperature for 1901–2005, the IPCC Fourth Assessment Report (IPCC; Reference SolomonSolomon and others, 2007) indicates that an overall warming ranging from 0.2 to 1.1°C has occurred in Chile north of 38° S, while a cooling of 0.2–0.5°C took place in the region between 38 and 43°S. In fact, an overall cooling of 0.1–0.3°C (10 a)−1 has occurred along the western coast of central and southern South America (5–30° S and 35–55° S, respectively) during the period 1979–2005 (see fig. 3.9 of Reference Trenberth and SolomonTrenberth and others, 2007).
Figure 3 shows the secular trend of the near-surface minimum air temperature (in terms of its annual standardized anomalies) for 1961–2006 at several weather stations in Chile for summer (December–February) and winter (June–August). Minimum temperatures are selected because they better capture the long-term changes (Reference KingKing, 1994). The interannual variability was filtered out using double-exponential smoothing, as described by Reference Rosenblüth, Fuenzalida and AceitunoRosenblüth and others (1997) and Reference Carrasco, Casassa and QuintanaCarrasco and others (2005). As a result, the secular trends were obtained.
Anomalies of the seasonal mean of minimum near-surface air-temperature behaviour recorded at several weather stations in Chile. Solid (dashed) curves correspond to summer (winter) season, after applying an exponential filter to the annual values. Minus and plus signs indicate near-surface cooling and warming, respectively.
The secular trends for the period 1961–2006 show an overall warming north of 40° S that ranges between 0.06 and 0.35°C (10 a)−1 (0.12 and 0.38°C (10 a)−1) in summer (winter) and a slight warming south of 47° S. A cooling that fluctuates from 0.04 to 0.1°C (10 a)−1 occurred between 40 and 47° S (see Table 2 for details). Most of these tendencies can be attributed to the 1976/77 climate shift (Reference Giese, Urizar and FuckarGiese and others, 2002). In fact, individual station analysis (Table 2) revealed a slight negative trend (cooling) during the 1977–2006 period for most stations (although not statistically significant), which concurs in part with the Fourth IPCC Report (Reference Trenberth and SolomonTrenberth and others, 2007). This behaviour is associated with the PDO, which shows a phase change in 1976/77 (Reference Giese, Urizar and FuckarGiese and others, 2002).
Secular trends of the near-surface minimum airtemperature (°C (10 a)−1) after filtering out the interannual variability. Summer (winter) corresponds to December–February (June–August). The names of the stations located inland are bold; all the others are coastal stations
Precipitation
Precipitation in central and southern Chile is mainly of frontal origin with systems that develop in the Pacific Ocean and travel eastward. The north–south seasonal movement of the westerlies results in an annual precipitation cycle with maximum (minimum) amount falling in winter (summer) months (Fig. 4). Thus, >90% of the snowfall in the central Andes (between 30 and 38° S) occurs during April to September (Reference Falvey and GarreaudFalvey and Garreaud, 2007), while in the southern and austral Andes (including the Patagonian icefields) it snows year round with an annual cycle showing a slight maximum during the winter months. In the northern altiplano (north of 24° S), the precipitation (including snow) takes place during the summertime (from late November to April) being associated with the South American monsoon that brings moist air from the Amazon area towards the eastern side of the Andes (Reference Falvey and GarreaudFalvey and Garreaud, 2005).
Average monthly precipitation (mm a−1) at different locations from north to south along the Chilean Andes. All stations are located on the western side of the Andes, except for the southernmost plot where Torres del Paine, Fagnano and San Isidro are located east of the Andes. Station elevations are expressed in m a.s.l.
Figure 5 shows the variation of annual precipitation for all available stations (most of them deployed in the coastal and valley sectors) from 1950 to 2000. A lack of stations between 48 and 51° S precludes analysis in this region. No statistically significant changes are revealed for the central (30–38° S) and southernmost (austral, south of 52° S) parts of Chile, while a decrease in precipitation does occur in the southern part between 38 and 46° S, which is statistically significant at some stations.
Annual precipitation trend (in mm a−1) during the period 1950–2000 for 66 rain-gauge stations located between 30 and 55° S. Dots represent trend values in mm in the 50 year period, and the vertical lines across each dot are the confidence intervals calculated using the Monte Carlo test, with a significance level of 95% (Reference QuintanaQuintana, 2004).
The scarce precipitation records in the northern altiplano allow only a limited examination of the snowfall trends in this region. Figure 6 shows the long-term behaviour at Visviri (18°23′ S, 69°33′ W; 4070 m a.s.l.), which can be considered a reference station for the most northern region. An overall slight increase (not statistically significant) of 12 ± 2 mm (10 a)−1 occurred in precipitation at this station during the full 1962–2003 period, although a non-significant negative trend is observed between the late 1970s and the mid-1990s, followed by an increase in precipitation thereafter. At Huatacondo station, however, a slight not statistically significant negative trend is revealed (20°56′ S, 69°05′ W; 2450 m a.s.l.) although from the mid-1990s it shows an increase in precipitation as at Visviri station.
Standardized anomalies of the interannual mean (dashed curves) and secular trend (solid curves) obtained after applying an exponential filter for annual precipitation at Visviri/Huatacondo, Lagunitas, Punta Huano and Punta Arenas stations which, respectively, are locations close to the radiosonde stations Antofagasta, Quintero/Santo Domingo, Puerto Montt and Punta Arenas.
0°C isotherm and ELA
The altitude of the 0°C isotherm can be obtained from the daily vertical profile of the radiosonde data. This information was used to construct the monthly, seasonal and annual average at each of the three northern aerological stations from 1958 to 2006, and for Punta Arenas from 1975 to 2006. The interannual standardized anomalies and the secular trend obtained applying the exponential filter for the four aerological stations are shown in Figure 7. Standardization was performed taking the average value within the period subtracted from the annual value, and divided by the standard deviation within the period. Table 3 indicates the basic statistics at each site for three periods, 1958–2006, 1958–76 and 1978–2006, except for Punta Arenas where the data cover the 1975–2006 period only. An overall positive trend can be observed, which is statistically significant (at the 95% confidence level) in Antofagasta and Quintero/Santo Domingo. The climate shift of 1976/77 is also clearly revealed by the time series with a negative (cooling) trend before 1976 and a positive (warming) trend after 1977 (Reference Giese, Urizar and FuckarGiese and others, 2002). In fact, analysis of the mean altitudes of the 0°C isotherm indicates predominantly warmer conditions during the past 30 years, statistically significant (at the 95% level), compared to the period 1958– 76. The climate shift is associated with the PDO, which has a spatial climate pattern similar to the El Niño–Southern Oscillation (ENSO) but with a timescale of ∼20–30 years (Reference Zhang, Wallace and BattistiZhang and others, 1997). A cool phase of PDO prevailed from 1947 to 1976 while a warm phase dominated from 1976 to the late 1990s.
Standardized anomalies of the interannual mean (dashed curves) and secular trend (solid black curves) obtained after applying an exponential filter for annual anomalies of the altitude of the 0°C isotherm at radiosonde stations Antofagasta, Quintero/Santo Domingo, Puerto Montt and Punta Arenas. Grey curves show the secular trends before and after the 1976/77 climate shift (vertical line).
Altitude (ma.s.l.) and tendency (m (10 a)−1) of the 0°C isotherm for different periods
The ELA represents the altitude where the annually integrated glacier mass balance is zero. This is closely related to the air temperature and precipitation. Above the ELA, annual net accumulation takes place, while below the ELA annual net ablation occurs. A rise (lowering) of the ELA implies loss (gain) of glacial mass which results in a subsequent retreat (advance) of the glacier front after a certain response time which depends mainly on glacier accumulation and thickness (Reference Jóhannesson, Raymond and WaddingtonJόhannesson and others, 1989).
ELA data for the southern Andes are scarce, normally being measured by means of field observations and/or aerial and satellite imagery acquired at the end of the summer period. The sporadic nature of the ELA information generally precludes a time-series analysis. Here the ELA is reconstructed using a relationship between the precipitation and the altitude of the 0°C isotherm, based on a method first developed by Reference FoxFox (1993) and Reference CondomCondom (2002; see also Reference Condom, Coudrain, Sicart and ThéryCondom and others, 2007) and modified by Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) for central Chile. First, the concept of normalized snowline altitude (h snowline) is introduced, defined as the difference (in m) between the ELA or regional snowline altitude and the altitude of the mean annual 0°C isotherm (h 0°C):
Reference CondomCondom (2002) and Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) estimate the altitude of the 0°C isotherm (in m) from the vertical temperature gradient, which for the southern Andes is 0.65°C (100 m)−1 (Reference AceitunoAceituno, 1996). Thus,
where T is the mean annual temperature (°C) and z is the elevation (m) of the surface station used for the calculation. Reference FoxFox (1993) established an empirical relation linking the annual normalized snowline altitude (h snowline) and the annual precipitation P (mm a−1) measured at a nearby ground station:
Here, c 1 and c 2 are non-dimensional constants obtained by linear regression of the normalized snowline altitude (Equation (1)) and the logarithm of the precipitation (Reference FoxFox, 1993). For this purpose Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) used 0°C isotherm altitudes derived from ground stations, an estimated temperature lapse rate and real ELA data of glaciers. They determined values of c 1 = 3427 m and c 2 = −1148 m, which are well constrained for the tropical Andes, whereas they only rely on a few glacier data points in the southern Andes. Combining the three equations and considering precipitation data of nearby stations, Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) obtained:
The temperature and precipitation dataset used by Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) are those of Reference New, Lister, Hulme and MakinNew and others (2000) for the 1961–90 period, which have a spatial resolution of 10′ and cover the entire globe.
Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) derived a similar expression valid for central Chile (30–38° S) using the actual altitude of the 0°C isotherm (H0°C) obtained from radiosonde data at Quintero/Santo Domingo, precipitation for ∼260 ground stations extrapolated to 2500 m a.s.l. and real ELA information for eight sites between 27 and 39° S (Table 4). Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) obtained c 1 = 3104.9 m and c 2 = −920.24 m, and the expression results in:
Figure 8 shows the original north–south ELA profile reproduced from Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007), along with the altitude of the 0°C isotherm. Superposed are the ELAs calculated using the expressions of Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) and Reference Carrasco, Casassa and QuintanaCarrasco and others (2005), which correspond to Equations (4) and (5), respectively, computed at five latitudinal locations along the Chilean Andes: Visviri, Huata-condo, Quintero/Santo Domingo, Puerto Montt and Punta Arenas. Also shown in Figure 8 and Table 4 are ELA data available from remote-sensing studies and ground information, which are used to compare the different ELA models.
Available ELA data for the Chilean Andes from existing published sources. The ELA data are used in Figure 8 for comparing the different model predictions. The column labelled ** indicates whether the ELA data were used for calibrating the ELA model of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005)
ELA as calculated originally by Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007), from their figure 3. Also plotted are the ELA obtained using the expression of Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) using the annual mean air temperature (T) and geopotential height (z) at 850 hPa from radiosonde and precipitation data from the closest station, also calculated from ground-station data, as mentioned in the text. For the model of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005), the 0°C isotherm elevation was used in combination with precipitation data from ground stations, as described in the text. The latitudinal profile of the 0°C isotherm altitude is also plotted, based on radiosonde data for four stations, and a polynomial fit. Black diamonds are observed ELAs from existing data (Table 4).
In the northern region (region 1 in Fig. 8), the most complete precipitation time series available are for Visviri (17°37′ S, 69°30′ W; 4070 m a.s.l.) and Huatacondo (20°56′ S, 69°05′ W; 2450 m a.s.l.), and for air temperature it is data from Calama (22°29′ S, 68°54′ W; 2270 m a.s.l.). Applying Condom’s expression (Equation (4)), we obtained an ELA of 4678 ± 412 m using Visviri precipitation and 6565 ± 693 m using Huatacondo. Error estimates here represent one standard deviation.
The same exercise was conducted introducing into Equation (4) the mean annual air temperature and geopotential altitude observed at the 850 hPa level as temperature (T) and elevation (z) instead of the surface station data. In this case the results were 4713 ± 245 m for Visviri and 6495 ± 717 m for Huatacondo. The original calculation by Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) shows ELA values of 5800 m for Visviri and 4950 m for Huatacondo, both altitudes corresponding to the western side of the Andes. Now, applying the relationship of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) indicated in Equation (5), where the altitude of the 0°C isotherm is used, the results obtained for the ELA were 5423 ± 194 m for Visviri and 6850 ± 575 m for Huatacondo.
For the central, southern and austral regions (regions 2, 3 and 4, respectively) we used station data from Santiago (33°26′ S, 70°40′ W; 520 m a.s.l.) (region 2), Puerto Montt (region 3) and Punta Arenas (region 4) to calculate the ELA. Applying the relationship (Equation (4)) of Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) we obtained 3369 ± 257 m for region 1, 1350 ± 101 m for region 2 and 1409 ± 105 m for region 3. The original calculation by Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) indicates ELA values of 4150, 2250 and 600 m for regions 2, 3 and 4, respectively. Now, using the 850 hPa radiosonde data, the ELAs were 4061 ± 274, 1468 ± 149 and 1560 ± 123 m. Calculating the ELAs using the expression (Equation (5)) of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005), we found altitudes of 3954 ± 236, 2011 ± 172 and 991 ± 124 m for regions 2, 3 and 4, respectively.
The correlation coefficients between ELAs obtained applying the expressions of Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007) using radiosonde data as input values, and those obtained from the expression of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) using temperature from radiosonde data and precipitation from mountain areas, were statistically significant (at 95%) and above 0.95 in all regions. This suggests that even though there are differences in altitude from the calculated ELAs, the annual variability and tendencies are similar.
Inspection of Figure 8 shows that south of 27° S the model which best fits the available ELA data from existing studies is that of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) using the 0°C isotherm from radiosonde data and precipitation data extrapolated to 2500 m. Between 19 and 26° S there are virtually no glaciers along the Chilean Andes due to dry conditions in the so-called ‘arid diagonal’, with only a few rock glaciers, particularly on southern slopes. At the latitude of Huatacondo the highest mountains reach altitudes of ∼6500 m, and >6700 m a few hundred kilometres to the south. Similar to the regions south of 27° S, in the area between 19 and 26° S the model of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) is regarded as the best for simulating the ELA since it is the model that predicts the highest ELAs, with altitude values higher than the existing peaks, as expected. In the northernmost portion of region 1, the highest estimate of the ELA is provided by the original results of Reference Condom, Coudrain, Sicart and ThéryCondom and others (2007), who predict an ELA of 5800 m. This model of Condom and others is regarded as the best ELA prediction in the northernmost area since local glaciers, notably at Parinacota and Pomerape, have a relief ranging from ∼6300 to ∼5300 m on southern slopes and down to ∼5800 m on northern slopes.
Figure 9 shows the standardized anomalies of annual variability and secular trends of the ELA using the expression of Reference Carrasco, Casassa and QuintanaCarrasco and others (2005) with the radiosonde data and the precipitation data as described below. Table 5 summarizes statistics of mean altitude and tendency of the ELA for three periods, 1958–2006, 1958–76 and 1978–2006, except for Punta Arenas which is given only for the 1975–2006 period.
Standardized anomalies of the interannual mean (dashed curves) and secular trend (solid curves) obtained after applying an exponential filter of the derived ELAs at the radiosonde stations Antofagasta, Quintero/Santo Domingo, Puerto Montt and Punta Arenas. Grey lines show the secular trends before and after the 1976/77 climate shift (vertical line) and the year 2000.
Altitude (m a.s.l.) and tendency (m (10 a)−1) of the ELA for different periods after filtering the annual variability using an exponential filter
For the northern region the precipitation data correspond to Huatacondo station, which is regarded as representative for the altiplano in northern Chile. Our analysis shows that an ELA decrease took place until around 1976 and increased thereafter, with a more significant increase from 1985 to 1995. Linear regression over the period 1962–2003 indicates a positive trend of 68 ± 12 m (10 a)−1 .
Precipitation in central Chile is obtained from the mountain station Lagunitas (33°04′ S, 70°15′ W; 2765 ma.s.l.), which shows an overall positive trend of 27 ± 6 mm (10 a)−1 during the period 1958–2004, with a significant increase at the 95% level from 1958 to 1982 (170 ± 8 mm (10 a)−1), and a significant decline at the 95% level (92 ± 16 mm (10 a)−1) in precipitation after 1982. Based on the Lagunitas precipitation data and temperature data at 850 hPa for Quintero/ Santo Domingo, the ELA shows a negative trend (i.e. decreasing) from 1958 to the mid-1970s and a positive trend (i.e. increasing) thereafter. The linear ELA trend (not statistically significant) in the period 1958–2004 is 11 ± 3 m (10 a)−1.
In the southern region there is no precipitation record in mountain areas, so annual data from Punta Huano (41°08′ S, 72°17′ W; 200 ma.s.l.) were used for the ELA calculation, applying a correction factor of 1.5 for the orographic effect (Reference Adam, Clark, Lettenamier and WoodAdam and others, 2006). A statistically significant (at 95%) lineal decrease in precipitation of 51 ± 4 mm (10 a)−1 in the 1958–2004 period has taken place at this station. The precipitation decline in this region (Fig. 5), in addition to the increase in the 0°C isotherm altitude, results in a positive trend of the ELA after 1970. The overall linear rise of the ELA is 34 ± 2 m (10 a)−1 for the 1959–2004 period, which is statistically significant.
In the austral region, the annual precipitation record from Punta Arenas, located on the leeward side of the mountains, shows a linear trend of 17 ± 1 mm (10 a)−1 for the period 1958–2006, which is not statistically significant. The annual precipitation at this site was multiplied by a factor of 6.7 according to the ratio in the annual precipitation with respect to Faro Evangelistas station located on the coastal side, windward from the Andes, which is regarded as being more representative of precipitation of the mountain glacier areas (Reference Schneider, Glaser, Kilian, Santana, Butorovic and CasassaSchneider and others, 2003). The result is a significant (at the 95% level) increase in precipitation during the 1958–2006 period of 112 ± 4 mm (10 a)−1, with a corresponding ELA lowering of 8 ±2 m (10 a)−1 during the 1975–2006 period of available radiosonde data.
Comparing the values given in Tables 3 and 5 for regions 1, 2 and 3, it can be seen that the mean ELA values before the 1976/77 climate shift are lower than the mean obtained for the 1978–2006 period, with a statistically significant difference (at 95% level) between the 1958–76 average and the 1978–2006 average. Also, the linear tendencies change from a negative trend before the shift to a positive trend after the shift. These results indicate the influence of the PDO on the air temperature and on the ELA’s behaviour.
Discussion and Conclusions
According to surface air-temperature records from valley and coastal stations, a significant warming has occurred from 1961 onward in northern (region 1) and central (region 2) Chile. In the southern region (region 3) there is a slight cooling, with a non-significant warming in the austral region (region 4). In the lower troposphere, as revealed by the 850 hPa level of the radiosonde profiles, there is significant warming at all stations except Punta Arenas.
The air-temperature behaviour is strongly forced in all regions by the climate shift that took place in the mid-1970s, mainly in the central and southern regions. Data analysis from surface stations indicates a slight cooling observed at coastal surface stations (mainly in summer) between Antofagasta and Coyhaique (Table 2) after 1976/77. According to radiosonde air-temperature data at 850 hPa, the cooling is restricted to the near-surface troposphere and probably within the boundary layer, below 800–1000 m.
An evaluation of the air temperatures obtained from the radiosonde data at 850 hPa for the 1958–2006 period indicates a significant increase (95% level) of 0.22 ± 0.01°C (10 a)−1 in Antofagasta, 0.15 ± 0.10°C (10 a)−1 in Quintero/Santo Domingo and 0.12 ± 0.10°C (10 a)−1 in Puerto Montt (not statistically significant). Although the annual warming at Puerto Montt is not significant, in winter (summer) there is a significant warming of 0.14°C (0.13°C).
There is an overall positive trend in ELA in regions 1, 2 and 3 for the study period 1958–2006. In region 4 there is no trend in the ELA for the same period. Before the 1976/77 climate shift the ELAs show a statistically significant (at 95%) negative linear trend. By contrast, after this date, the ELAs present a significant positive trend until 1998–99 in the central and southern regions (regions 2 and 3), and a nonsignificant positive trend in the northern region (region 1). In the austral region (region 4) there is a slight ELA decrease after 1982. Both central and southern regions show a significant ELA decline between 1998 and 2000 (Fig. 9) and thereafter a significant positive trend took place.
As reported by Reference Rivera, Acuña, Casassa and KnightRivera and others (2006), glaciers in the central region (region 2) showed a significant ice-mass reduction with an accelerating trend in recent decades (Fig. 1). In this region, the annual mass balance is closely related to the ENSO. During the warm El Niño (cold La Niña) phase, winter precipitation is likely to be higher (lower) than normal and consequently the mass balance is likely to be positive (negative). This has been recorded by the systematic measurements at Glaciar Echaurren Norte (33°35′ S, 70°08′ W) by the Dirección General de Aguas (Reference Escobar, Casassa and PozoEscobar and others, 1995). Although mass-balance data from this glacier show a net positive balance for the 1973–93 period (Reference Escobar, Casassa and PozoEscobar and others, 1995), updated analysis reveals a clearly negative trend (Reference Casassa, Rivera, Schwikowski and KnightCasassa and others, 2006) after 1988. The overall glacier retreat and thinning in the central Chilean Andes concurs with the overall ELA increase found in our analysis.
The ELA behaviour in the southern region concurs with the maximum decrease of the ice-mass area in the most recently analyzed period observed by Reference Rivera, Bown, Casassa, Acuña and ClaveroRivera and others (2005), who reported a loss of 0.45 km2 a−1 between 1987 and 2003 at glaciers on Volcán Mocho-Coshuenco (39°55′ S, 72°02′ W). This enhanced mass loss in recent years is confirmed by Reference Bown and RiveraBown and Rivera (2007) who found an areal change in Glaciar Casa Pangue (41°08′ S, 71°52′ W) ranging from −0.19 km2 for the 1961–81 period to −0.47 km2 for the 1981–98 period, and a frontal variation of −10 to −51 m a−1 for the respective periods.
In the austral region (∼46–55° S) an ELA lowering of 8 ± 1 m (10 a)−1 in the period 1975–2006 is modelled in this study due to a slight increase in precipitation and a decrease in the 0°C isotherm altitude. An ELA lowering is not evident in this region based on the extensive observations of glacier-front positions and glacier thinning available, for example, for the Patagonian icefields (e.g. Reference Rignot, Rivera and CasassaRignot and others, 2003), Gran Campo Nevado (Reference Schneider, Glaser, Kilian, Santana, Butorovic and CasassaSchneider and others, 2003), northern Cordillera Darwin (Reference Holmlund and FuenzalidaHolmlund and Fuenzalida, 1995; Reference Porter and SantanaPorter and Santana, 2003) and eastern Cordillera Darwin (Reference Strelin and IturraspeStrelin and Iturraspe, 2007), all of which show strong and increased glacier wasting in recent years. Only in the southwestern Cordillera Darwin are many glaciers growing (Reference Holmlund and FuenzalidaHolmlund and Fuenzalida, 1995), with advancing fronts at least until January 2007 according to a field visit made to the area by one of the authors (G.C.). Therefore, the combination of the radiosonde data at Punta Arenas and the precipitation data at Faro Evangelistas is either not representative for the prevailing glacier conditions in Patagonia and Tierra del Fuego, or perhaps the Fox–Condom parameterization of the ELA performed in this study is not valid for this particular region.
A lack of precipitation trends in western Patagonia for the last 40 years is suggested by analysis of upper-air US National Centers for Environmental Prediction–US National Center for Atmospheric Research (NCEP–NCAR) data (Reference Rasmussen, Conway and RaymondRasmussen and others, 2007), although a positive trend has been detected in the Southern Annular Mode (Reference Marshall, Stott, Turner, Connolley, King and Lachlan-CopeMarshall and others, 2004) during the last few decades which should translate into an intensification of the westerlies and an accompanying precipitation increase. There seems to be a non-negligible warming signal in Patagonia in recent decades based on the NCEP–NCAR analysis of Reference Rasmussen, Conway and RaymondRasmussen and others (2007) and data from Argentine stations (Reference Ibarzabal, Donángelo, Hoffman and NaruseIbarzabal and others, 1996). In a study of Gran Campo Nevado (53° S), Reference Möller, Schneider and KilianMoeller and others (2007) detected an average thinning of glaciers greater than 1 m a−1 at low elevation and a small but significant thickening at higher elevations, resulting in an overall net thinning, which they explain as a result of a combined increase in precipitation and warming.
In the northern (∼18–30° S) and central (∼30–38° S) parts of the Andes in Chile, there is no significant change in precipitation after the 1976/77 climate shift; at the same time a significant (at 95%) positive trend of the 0°C isotherm (Table 3) has occurred, indicating that the ELA rise in these regions is likely to be due to warming only. In the southern region (∼38–46° S) there has been a significant decrease in precipitation (Fig. 4), with a significant winter and summer increase in the 0°C isotherm, suggesting that the ELA rise is due both to precipitation decline and warming.
Our radiosonde data for the southern region indicate that the mid-elevation surface-warming signal in the Andes is not reflected by the slight cooling trend observed at sea level. The same conclusion was drawn by Reference Vuille and BradleyVuille and Bradley (2000) for the tropical Andes, where there is significant warming at high altitudes but slight cooling in the lower troposphere. A lower-troposphere warming which increases with altitude is predicted by general circulation models under scenarios of enhanced concentration of greenhouse gases (Reference Bradley, Vuille, Diaz and VergaraBradley and others, 2006). Therefore we should expect that glaciers located in the high Andes will be especially affected by future warming.
This study has shown that except for the northernmost area of Chile (∼18° S) and in the Patagonia–Tierra del Fuego region (∼46–55° S), the radiosonde data along with precipitation records of surface stations can be useful in reconstructing a time series of ELA information for application in glaciological studies. The ELA model can be applied to project the effect of predicted climate changes on the future condition of glaciers, which are important sources of water and can also present increased hazards under warming scenarios.
Acknowledgements
This study was funded by Comisiόn Nacional de Investigaciόn Científica y Tecnolόgica de Chile through FONDECYT Projects 1040989 and 1061269, and Proyecto Anillo, PBCT ACT-19. Centro de Estudios Científicos (CECS) is funded by the Chilean Government through the Millennium Science Initiative and the Centers of Excellence Base Financing Program of CONICYT (Comision Nacional de Investigacion Cientifica y Technologica de Chile). CECS is also supported by a group of private companies which at present includes Antofagasta Minerals, Arauco, Empresas CMPC, Indura, Naviera Ultragas and Telefόnica del Sur. Figure 3 was provided by J. Quintana of the Direcciόn Meteorolόgica de Chile. F. Ordenes from CECS provided help with the figures.
