Figure 2 shows the time series of the ice front displacement, averaged over the entire glacier width and for the sections in CT and BR separately. The glacier slightly advanced from 2000 to 2001 because of lack of seasonal retreat in the summer 2000/2001, but was otherwise fairly stable during the study period (Figure 2). Four ice-dam events were reported during this period (2003/04, 2005/06, 2007/08, and 2011/12), but no significant change was observed in the front position (Figure 2). Although there was no substantial long-term trend, ice front position showed a seasonal advance/retreat pattern. The glacier began to retreat in early summer (typically in December) and then turned to advance in mid autumn (typically in April) (Figure 2). The magnitude of the seasonal frontal variations was up to 200 m, which was greater than the variations in annual mean front position over the study period (120 m).

Figure 3A shows the most advanced and retreated glacier front over the study period, and annual cycles of the ice front variations constructed by stacking the data over the study period. On average, the glacier advanced from June to December (early summer), and retreated from December to April (Figure 3B). The ice front was relatively stable from April to June (Figure 3B). Glacier retreat occurs approximately twice as fast as advance, i.e., the rate of frontal position change from December to April was −0.7 m d⁻¹, whereas the rate from June to December was 0.3 m d⁻¹. The pattern of the seasonal variations was similar at the two ice fronts facing CT and BR, but the amplitude was two times greater in BR (±80 m) than in CT (±40 m) (Figures 3C,D).

The maximum speed (~4 m d⁻¹) was measured at around 1200 m a.s.l. (15 km from the front), approximately at the equilibrium line altitude (Figure 4). Surface speed decreased toward the ice front, except for the relatively fast following region within several kilometers near the front. Ice speed at 2 km from the front was 2.5 m d⁻¹ (Figure 4). The mean speed was greater at CT front (1.8 m d⁻¹) than that of BR (1.6 m d⁻¹) (Figure 4), which is consistent with the previous study (Stuefer et al., 2007). The ice surface speed showed no clear long-term trend over the study period (Figure 5). Nonetheless, linear trends of ice speeds were −3, −2.9 and −6.1 m a⁻² for the whole glacier, CT and BR, respectively (Figure 5). For the years with a sufficient amount of data (2000, 2001, 2005, 2010, and 2013), summer acceleration followed by deceleration in autumn was observed (Figure 5).

Clear seasonal variations in speed were observed in the region studied (Figures 4, 6). For example, the mean summer speed (December–February) 1.8 km from the ice front was 15% greater than the annual mean (Figure 6C). A similar speed-up was observed 0.9 km from CT (5%) (Figure 6A) and BR (11%) (Figure 6B). The seasonal speed variation is observable 5.2 km from the front (Figure 6D), but is unclear further up the glacier. The ice speeds obtained from satellite image analysis agree with those measured using the in-situ GPS survey (Table 2). If the daily mean speeds measured with GPS are then compared with the daily mean speeds from satellite images covering that day of the year (but not necessarily the same year), the root-mean-square error is 0.25 m d⁻¹ (Figure 6).

Figure 7 shows the displacement rate dL/dt, ice speed u_m, computed monthly frontal ablation rate c˙ + m˙, and monthly air and lake temperature. The annual mean frontal ablation rate was 1.3 ± 0.7 m d⁻¹ (± standard deviation). The frontal ablation rate varied from 0.13 ± 0.9 m d⁻¹ in July to 2.5 ± 0.4 m d⁻¹ in February, showing a large seasonal variation (Figure 7C). The monthly mean ice speed varied from 1.0 ± 0.13 m d⁻¹ in May to 1.5 ± 0.16 m d⁻¹ in November (Figure 7B). Therefore, the magnitude of the seasonal variations in the frontal ablation rate was approximately five times greater than that in the ice speed.

Monthly mean air temperature during the study period ranged from −1.9°C to 12°C with mean temperature of 6.6°C over the 15 years (Figure 8A). Linear regression of the annual mean temperature showed a warming trend of 0.059°C a⁻¹ (p < 0.03) over the 15 years (Figure 8B). More rapid warming was observed in winter (JJA) (0.091°C a⁻¹, p < 0.13), and this rate was four times greater than that in sumer (DJF) (0.023°C a⁻¹, p < 0.44) (Figure 8B). These trends are influenced by relatively cold temperatures observed from 2000 to 2002. For example, mean winter (JJA) temperature from 2000 to 2003 was 0.98–1.61°C below the mean over the 15 years. The warmest annual mean temperature was recorded in 2013 (0.49°C above the mean).

Wind speed showed a large seasonal variations ranging from 1.8 m s⁻¹ in August to 6.6 m s⁻¹ in December (Figure 8D). Prevailing wind direction was west throughout the year (Figure 8E). Annual mean wind speed and direction from 1999 to 2013 were 4.1 ± 1.1 m s⁻¹ and 248 ± 8°, respectively. These results are in good agreement with previous study (Stuefer, 1999).

Near-surface lakewater temperature showed seasonal variations ranging from 3.8°C (August) to 10.3°C (February–March) (Figure 8C). The peaks in the monthly mean lakewater temperature lagged behind the air temperature by 1 month (Figures 8A,C). The water temperature might be influenced by the seasonal lake level variations (Figure 8C). However, the temperature change to be expected by the water level variations (~2 m) was only ~1.3°C in summer and less in winter, according to the vertical temperature profiles shown in Figure 9. Moreover, the temperature sensor was in deeper (i.e., colder) water in the summer, and this effect reduces the amplitude of the seasonal temperature variations. Therefore, the influence of the water level variations on the temperature record is not significant.

Figure 9 shows water temperature profiles obtained within 300 m from the glacier front (Figures 1D,E). Water temperature was higher in summer than in spring from the surface to the bottom. The mean temperature was 6°C in CT and 5.4°C in BR in summer, whereas 5.1°C in CT and 4.8°C in BR in spring (Figure 9). The magnitude of the seasonal temperature variation was slightly larger in CT (0.9°C) than BR (0.6°C). In summer, water temperature decreases from the surface (~8°C) to the bottom (5.5°C) in CT (Figure 9A) and the surface (~6.5°C) to the bottom (~5°C) in BR (Figure 9B). On the other hand, water temperature was more uniformly distributed in spring (Figure 9). This implies that lakewater in summer was more influenced by solar radiation than in spring. The upper 5 m mean temperature in CT (December; 7.1°C, October; 5.3°C) closely agreed with monthly mean near-surface temperature measured at the shore of CT (December; 7.2°C, October; 5.4°C) (Figure 9A). Further, CTD casts at other locations in the lakes showed similar profiles (Figure 9). Thus, it is likely that seasonal variations in the near-surface water temperature (Figure 8C) represent overall seasonal lakewater temperature variations. It should be noted that the water temperature in BR and CT was 2–3°C warmer than that which was measured in front of Glaciar Upsala, a glacier calving into another arm of Lago Argentino at about 70 km north of GPM (Sugiyama et al., 2016). This is because ice and meltwater discharge into the lake is much greater at Glaciar Upsala.

Seasonal variations in ice-front positions have previously been reported on tidewater glaciers in Greenland (e.g., Amundson et al., 2010; Howat et al., 2010; Carr et al., 2013; Moon et al., 2015) and Alaska (e.g., Ritchie et al., 2008; McNabb and Hock, 2014). These previous studies showed glacier advance from winter to early summer, and retreat from mid to late summer or early winter. The advance/retreat cycle observed in GPM was similar to the observations in Greenland and Alaska. GPM advanced from July to December (early summer) and retreated from December to April (mid autumn) (Figures 2, 3).

The magnitude of the seasonal variation in frontal ablation rate was five times greater than that in the ice speed, implying that the seasonal variations observed in the front position of GPM were contributed primarily by the frontal ablation rather than the ice speed (Figure 7). A potential driver of the seasonal variations in the frontal ablation is the air and/or lake temperature because these properties may affect calving and/or subaqueous melting rates. Monthly mean air temperature reaches seasonal maximum in January, whereas the warmest lakewater temperature occurs 2 months later in March (Figure 7D). The frontal ablation rate peaks in February–March (Figure 7C), and its correlation with lake temperature (r = 0.95, p < 10⁻⁵) is more significant than with air temperature (r = 0.86, p < 10⁻³) (Figures 10A,B). Additionally, we compared the frontal ablation with the ice speed, but the correlation is weak (r = 0.48, p < 0.12) (Figure 10C). These observations imply that lakewater temperature primarily controls the calving and/or subaqueous melting. The influence of air temperature cannot be excluded, but here we discuss further a more significant correlation between water temperature and frontal ablation.

A relationship between frontal ablation and ocean temperature in front of glaciers was reported in Svalbard (Luckman et al., 2015) Based on satellite image analyses and mooring observations of ocean temperature, a linear relationship between the frontal ablation rate and ocean temperature was proposed for Kronebreen and Tunabreen, tidewater glaciers in Svalbard.

T is ocean temperature at a depth between 20 and 60 m, and the coefficient k was 1.015 m d⁻¹ °C⁻¹ for Kronebreen glacier and 0.35 m d⁻¹ °C⁻¹ for Tunabreen. These results were obtained from tidewater glaciers, thus care should be taken when they are compared to our data, which is from a freshwater calving glacier. Nevertheless, Kronebreen and Tunabreen have some similarities with GPM in terms of ice speed (1–4 m d⁻¹), width of the ice front (~4 km) and the frontal ablation rate (0–8 m d⁻¹). Thus, we applied our data to Equation (4) and compared the linear coefficient k. In the case of GPM, k = 0.5 m d⁻¹ °C⁻¹ was obtained from Figure 10B. An important difference from the relationship in Svalbard is that ablation approaches zero at water temperatures above 0°C. For example, frontal ablation rate of GPM is very small (0.14 m d⁻¹) in July when lakewater temperature was 4.9°C, and such a small ablation rate causes advance of the glacier in austral winter.

The linear relationship between the lakewater temperature and frontal ablation suggests that melting at/under waterline strongly influences on the frontal ablation of GPM. A possible mechanism connecting the water temperature and frontal ablation is the formation of thermal notch, an erosional notch on the vertical ice face along the waterline. In GPM, this notch formation is observed to have a depth of ~1–2 m, and is observed throughout the year (Figure 11). Thermal notch formation was proposed as a triggering mechanism of calving, and supporting evidence was reported in freshwater calving glaciers in New Zealand (Kirkbride and Warren, 1997; Röhl, 2006), and Patagonia (Haresign and Warren, 2005). Röhl (2006) observed the notch formation at the front of Tasman Glacier, and concluded that the rate of erosion was controlled by water temperature and circulation, cliff geometry, debris supply and water-level fluctuation. In front of GPM, the warmest water was observed at the lake surface (~8°C) (Figure 9). Presumably, seasonal temperature variations of the near-surface water control the rate of notch formation at the front of GPM, which results in seasonal changes in the calving rate. Furthermore, Patagonia is characterized by consistently strong westerly winds and the wind speed is enhanced particularly in summer (Stuefer, 1999; Lenaerts et al., 2014). This is consistent with our AWS data (Figures 8D,E). High waves due to the strong wind and the relatively warm lake surface may enhance waterline melting.

Another process linking the lake temperature with frontal ablation is subaqueous melting. Recent studies report that the subaqueous melt rate at the front of tidewater glaciers is substantially larger than previously assumed (e.g., Motyka et al., 2003; Bartholomaus et al., 2013). Our CTD data show that seasonal water temperature change occurs not only near the surface, but extends to the bottom of the lakes (Figure 9). Therefore, the water temperature variations have a potential to cause the frontal ablation rate variation by changing subaqueous melt rate. Subaqueous melting of icebergs in the ocean has been studied using laboratory experiments (Weeks and Campbell, 1973; Neshyba and Josberger, 1980; Russel-Head, 1980). These studies proposed equations for subaqueous melting by considering water temperature as the most important variable. In front of GPM, depth averaged temperature varies from 5.1°C in CT and 4.8°C in BR in October to 6.0°C in CT and 5.4°C in BR in December (Figure 9). According to the previously proposed equations by Weeks and Campbell (1973), Neshyba and Josberger (1980), and Russel-Head (1980), the observed seasonal temperature difference cause ~10–30% changes in subaqueous melt rate.

In addition to the relationship with lakewater temperature, the frontal ablation also correlates with air temperature, suggesting a possible influence of surface melting on frontal ablation (Figure 10A). In tidewater glaciers, subaqueous melting is enhanced by subglacial meltwater discharge because meltwater plume entrains warm water and supplies heat for ice melting. However, unlike in seawater, glacier meltwater is not buoyant in lakewater, and thus upwelling of subglacial discharge and entrainment of warm bottom water is not likely in GPM. Possibly, meltwater input into cracks and crevasses reduces the fracture strength of ice (Liu and Miller, 1979) and enhances subaerial calving (Benn et al., 2007). Since glacier surface is heavily crevassed near the front of GPM, weakening of ice in summer is one of the likely mechanism of seasonal variations in frontal ablation rate.

Interestingly, the amplitude of the seasonal frontal variation was greater in BR (±80 m) than in CT (±40 m) (Figure 3). Since the amplitude of the ice speed variation is similar in BR and CT, the more significant frontal variations in BR was presumably due to the variations in the frontal ablation (Figure 6). Because the magnitude of the seasonal lakewater temperature variation is greater in CT (0.9°C between October and December) than in BR (0.6°C) (Figure 9), water temperature cannot explain the difference. A possible interpretation of more significant seasonal variations of frontal ablation in BR is more heavily crevassed surface condition. Because of the divergent ice flow regime in the vicinity of BR (Figure 4), large and deep crevasses are formed near the calving front in BR (Figure 3A). Our hypothesis is that frontal ablation in BR is more sensitive to the waterline melting and surface meltwater production because of the heavily fractured ice structure near the front. Another interpretation is the influence of bed geometry on the ice front position. Frontal variations of BR are the most significant near the center of the calving front and our bathymetry data shows that the lake bottom is elevated in this region (Figure 3A). This elevated bed geometry may enhance seasonal advance of the ice front, causing the frontal variation to have a greater magnitude.

The frontal position of GPM has been stable since 1920, unlike most other calving glaciers in Patagonia (Skvarca and Naruse, 1997). Previous studies attributed the stability of GPM to the large up-glacier ice flux due to relatively large accumulation area (e.g., Stuefer et al., 2007; Minowa et al., 2015) and elevated bedrock topography near the glacier front (e.g., Aniya and Skvarca, 1992; Skvarca and Naruse, 1997). Although our data show a generally stable front position since 1999, the glacier showed advance from 2000 to 2004 and small retreat after 2008 (Figure 2). These fluctuations were related with an annual temperature anomaly (Figure 7). In summer 2000/2001, the glacier retreated only a little and began to advance already in February 2001, which resulted in an advanced front position in the following (2001/2002) summer season (Figure 8F). The small summer retreat in 2000/2001 coincided with the cold winter (JJA), with temperature −1.6°C below the 15-year mean winter temperature (Figure 8B). The low temperature condition continued until the following winter in 2001 (Figure 8B). Presumably, the calving and subaqueous melt were reduced under the influence of cold air and lakewater conditions. In summer 2008/2009, the ice front in BR retreated at a greater rate than in other years, and this led to a relatively retreated front positions after 2009 (Figure 2C). The large retreat coincided with a relatively warm air and lake temperatures in summer 2008/2009 and winter 2009 (Figures 8B,D). According to these observations, the ongoing and future atmospheric and lakewater warming should affect the frontal position change. The GPM air temperature record shows a warming trend of 0.059°C a⁻¹ from 1999 and 2013 (Figures 8A,B). This trend is consistent with the previously reported observations in Patagonia. Ibarzábal y Donángelo et al. (1996) found a warming of 0.3°C from 1940–1990 (0.006°C a⁻¹) recorded at a weather station in El Calafate (50°30′S; 72°26′W), located ~70 km east to the GPM. Reanalysis data at the grid point of 50°S and 75°W showed warming of 0.5°C between 1960 and 1999 (0.013°C a⁻¹) at 850 hPa (Rasmussen et al., 2007). Moreover, O'Reilly et al. (2015) reported rapid warming of global lake surface temperature in summer (global mean 0.34°C a⁻¹) between 1985 and 2009. Lakewater and atmosphere warming have the potential to increase the frontal ablation rate as demonstrated in this study (Figure 10). If frontal ablation increases and the ice front retreats from the current position, constrained by bed geometry, GPM would change rapidly as has been observed in other calving glaciers in Patagonia. As well as the atmospheric warming, it is crucial to monitor the change in lakewater temperature to understand the future evolution of calving glaciers in Patagonia.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.