KGI (Figure 1) is the largest of the South Shetland Islands, located 130 km at the northwestern tip of the Antarctic Peninsula (AP) (Falk et al., 2018), between 57° 35’ and 59° 02’ W, 61° 54’ and 62° 16’ S, with a length and width approximately of 79 and 27 km, respectively (Simões et al., 1999). About 90 % of the KGI is ice-covered (Simões et al., 1999), and influenced by maritime climatic conditions (Falk et al., 2018; Rückamp & Blindow, 2012; Ferron et al., 2004), with land-terminating and marine-terminating glaciers whose altitude varies between 0 to 700 m a.s.l. The average air temperature registered at Peninsula Fildes (Simões et al., 1999) was -2.8°C between 1947-1995, while at KGI was -2.5°C for the period 1948-2011 (Kejna et al., 2013). Temperatures rise above freezing in summer, and regularly during the spring and autumn months causing snowmelt. However, during the winter surface snowmelt is also observed occasionally (Rückamp et al., 2011).

Surface air temperature (2 m) monthly data from five stations located on KGI were used. The data series of the stations Bellingshausen (BEL), Jubany (JUB), and Great Wall (GRW) were obtained from the SCAR READER database available at https://legacy.bas.ac.uk/met/READER/surface/stationpt.html. Meanwhile, the data from Marsh or Frei (MAR) and Ferraz (FER) stations were downloaded from https://climatologia.meteochile.gob.cl/application/historicos/datosDescarga/950001 and http://antartica.cptec.inpe.br/, respectively. Monthly precipitation data series for Bellingshausen and Marsh stations were obtained from http://www.aari.aq/data/data.php?lang=1&station=0 and https://climatologia.meteochile.gob.cl/application/historicos/datosDescarga/9500, respectively (Table 1). We organized the data in glaciological years, where the glaciological year n is defined from September (year n) to August (year n+1). Seasonal periods were organized as follows: spring (September, October, and November -SON); summer (December, January, and February - DJF); autumn (March, April, and May -MAM); and winter (June, July, and August -JJA). The climatic variables were analysed for the period from 1980 to 2019. The temporal temperature series had less than 15 % of missing data and were well correlated between stations, for this reason the missing data of each station was completed by linear regressions with the regional vector. In case of precipitation series, the missing data was completed using linear regression between stations.

We used oceanic and atmospheric circulation indices to measure the influence of the ENSO variability: 1) Oceanic Niño Index (ONI) (available at https://psl.noaa.gov/data/correlation/oni.data), ONI is the running mean anomalies of three consecutive months of the Sea Surface Temperature (SST) in the region 3.4 of El Nino, and 2) Southern Oscillation Index (SOI) (available at https://psl.noaa.gov/data/correlation/soi.data), the SOI is one measure of the large-scale fluctuations in air pressure occurring between the western and eastern tropical Pacific during El Niño and La Niña episodes. These indices were selected based on previous research work that studied the influence of ONI (Walker & Gardner, 2017 and Paolo et al., 2018) and SOI (Clem & Fogt, 2013; Clem et al., 2016 and Santamaría-del-Ángel et al., 2021) on the climatic variability and glacier dynamic in the Antarctic. The indices were classified in 4 ranges to identify ENSO phases (Table 2). The teleconnection was analysed for the period from 1980 to 2019. We used the Pearson correlation coefficient at 95 % significance to identify the relationship between climatic variables and the indices. Subsequently, climate variables were normalized subtracting the mean and dividing by the standard deviation, and organized in ascending order of the indices to identify ENSO events influence.

Landsat satellite images (4-5, 7, and 8) with a spatial resolution of 30 m were used (Table 3) for the delimitation of glacier coverage. These images were downloaded from the GLOVIS portal of the United States Geological Survey (USGS), available at https://glovis.usgs.gov/app?fullscreen=0.

To select the multitemporal images, we prioritized those with minimal to no cloud cover over KGI, excluding images from winter months (June, July, and August) to avoid interference from seasonal snow cover that could disturb the identification of glacier front displacement. It was used Landsat Tier 2 images (good quality), accurate for land cover classification.

Glacier coverage in KGI was made through manual delimitation (Marghany, 2016; Baumhoer et al., 2018). The GLIMS Classification (Global Land Ice Measurements from Space) version 6 (RGI Consortium, 2017) was used to discriminate the glacier units in KGI. GLIMS Classification was spatially adjusted based on a TanDEM-X Digital Elevation Model (2016) with a spatial resolution of 90 m, available at https://geoservice.dlr.de/web/dataguide/tdm90/. The resulted adjusted GLIMS shapefile was overlapped with the temporal manual delimitation of the glacier fronts in KGI.

Table 4 shows the average air temperature and precipitation for the stations analysed in the study area. The mean temperatures for the complete period range from -5.6 °C to 1.8 °C, while the precipitation varies from 130.8 mm to 704.8 mm. The mean temperature between December and March was above 0°C, reaching up to 1.8°C (FER), with February being the warmest month of the year. Between April and November, the temperature was below 0°C, with temperatures that reached -5.6 °C (BEL).

The Table 5 shows the teleconnection analysis, where the annual and seasonal (SON) temperatures at the five weather stations are correlated negatively and significantly with the ONI. This means that air temperatures on KGI are higher during negative SST anomalies in the central Pacific. However, the behaviour was opposite (positive correlation) with the SOI. For the precipitation, only the BEL station showed a significant negative (positive) correlation during the annual and SON period with the ONI (SOI).

For the air temperature, MAR station showed a negative correlation during the annual period (-0.36) and SON (-0.39) with the ONI index. While SON register an opposite behaviour (0.31) with the SOI index. BEL station showed a negative correlation during the annual period and SON with a value up to -0.4 for the ONI index. However, positive correlations were found for the annual period (0.32) and SON (0.35) with the SOI index. FER station registered a negative (positive) correlation at annual scale and spring season up to -0.41 (0.31) and -0.46 (0.37) with the ONI (SOI) index. JUB station showed a negative correlation with the ONI index for the annual period and SON with values up to -0.41 and -0.42 respectively. These same periods exhibited a positive correlation with values of 0.34 (annual) and 0.37 (SON) with the SOI index. Like previous stations, GRW presented a negative correlation with the ONI index for the annual period (-0.38) and SON (-0.38), and a positive correlation with the SOI index for the annual period (0.29) and spring season (0.34).

For the precipitation, BEL station shows significant correlations with the ENSO indices. Specifically, a negative correlation for the ONI index at the annual scale (-0.43) and SON (-0.39), while the SOI index registers an opposite behaviour for both periods with a value up to 0.30. However, MAR station does not show significant correlations.

Table 6 shows the relationship between the ONI and SOI with the climatic variables (surface air temperature and precipitation) to characterize the ENSO events. The surface air temperature for all the stations shows that, at annual scale the cold phase (La Niña) of the ONI index increases significantly, exceeding the limit value. However, during the warm phase (El Niño) the ONI index shows an opposite behaviour, with the index decreasing. During SON (spring), the ONI and SOI index in the cold phase (La Niña) show an increase in surface air temperature. During the warm phase (El Niño), a temperature decrease is observed with values below -0.53 and -0.45 for the ONI and SOI index. In DJF, the cold phase for both indices show an increase in air surface temperature except for MAR station in SOI. Moreover, all the stations show a decrease registered in the ONI warm phase. A similar pattern was recorded with the SOI warm phase for JUB and GRW stations.

For the precipitation, only BEL had significant correlations with the ENSO indices. In the annual period, for the ONI cold phase, the precipitation decreases in BEL and MAR, which is not significant for the SOI cold phase. In the warm phase, BEL shows that precipitation decreases with ONI (-0.18) and SOI (-0.12). During SON the cold phase for both indices registered an increase in precipitation. Likewise, in DJF during the SOI cold phase, the precipitation decreases in both stations (BEL and MAR). Concerning the BEL station, the warm phase shows an opposite behaviour for the ONI and SOI.

In order to see the impact during the most characteristic years of both El Niño and La Niña, it is necessary to perform a detailed analysis at the station level. For this analysis, the ONI index and the BEL station were selected considering that the air temperature of BEL station has a similar pattern to the other stations in the study area, while in the case of precipitation, it is the only station that presented significant correlations with the analysed ENSO indices. Figure 2 shows the normalized data of air temperature and precipitation for the Annual and SON period, which are ordered following the increasing pattern of the ONI, likewise, the shaded areas represent in turquoise the cold phase (La Niña) and in yellow the warm phase (El Niño). Figure 2a for the annual temperature shows that out of 39 years analysed, 13 years (~30% of the period) exceeded the threshold of 0.5 standard deviations, which indicates an abnormal increase in temperature in those years, where six of the 13 years occurred within the cold phase and one within the warm phase of the ONI. From these results, we can infer that at annual scale level during La Niña events the temperature is more likely to increase, while during the warm phase, there could be (not always) a decrease in temperature. Figure 2b shows the air temperature for the SON period, in the cold phase two years exceed the threshold of 0.5, and seven years are below 0 (only two below -0.5). Likewise, during the cold phase, the mean (dotted line) is strongly influenced by the increase in temperature recorded in the years 2016-2017 and 2010-2011 but the largest number of events are below 0. During the warm phase, 13 years were recorded, of which four were positive and nine negatives. In the seven years where the ONI signal was stronger, the temperature showed a decrease and six of these years were below the threshold (-0.5), from which it is inferred that a decrease in temperature would be expected in the face of intense El Niño signals. Regarding annual precipitation (Figure 2c and 2d), we find that precipitation behaviour under different phases of ONI is variable with no clear relationships observed between these variables.

The predominant orientation of the glaciers in the KGI is south, southeast, north, and east, with a slope that varies between the range of 6° to 15°. In 1989, the glacier coverage of KGI was 1098.42 km² distributed among 73 glaciers (Figure 3). However, by 2020 the coverage was reduced to 989.45 km²; thus, in 31 years, 108.97 km² of glacier coverage has been lost, equivalent to a loss of 9.92% of its area with an average retreat rate of 3.52 km² · year^-1 for the study period (Table 7). Also, glacier retreat rates observed during the second half (2007-2015) of the study period show a slight deceleration (0.09 km² · year^-1) compared to the first half (1989-2006, 0.88 km² · year^-1) period. The highest rate of retreat occurred during 2005–2007, followed by 2001–2005. Many of the glaciers with an area of less than 5 km² (Figure 4a) located in the southern part of KGI have shown higher recession rates compared with others in the study area. It is important to mention that, five glaciers have shown significant retreat up to 12 km² (since 1989) such as Anna glacier (G-01), Polonia Piedmont glacier (G-02), Arctowski Icefield (G-03), Hektor Isfall Ouest (G-04) and Domeyco glacier (G-05) with 12.5 km², 9.36 km², 8.43 km², 8.0 km² and 5.17 km² glacier lost, respectively (Figure 3a). In addition, 34% of the glaciers in KGI have lost less than 10% of their glacial area, while 32% of the total glaciers lost between 10% and 20% (Figure 4b). Moreover, of the total of glaciers (73) on KGI, 37% had a mixed terminating, 42% were land-terminating and 21% were marine-terminating. Furthermore, of the total glacier area lost, 35% corresponds to marine-terminating glaciers, 16% corresponds to land-terminating glaciers, and 49% to mixed-terminating glaciers (Figure 3b). In addition, our findings show a recession glacier retreat observed since 2008 (Figure 5).