Changes in Vegetation Distribution around the King Sejong Station in Antarctica: A Comprehensive Analysis of Time-Lapse Electrical Resistivity Data, Meteorological Data, and Vegetation Data Open Access

Geology in harsh environments like Antarctica features well-known active layer and permafrost. The active layer thaws in summer and refreezes in winter, whereas permafrost remains frozen year-round. The thickness of the active layer is closely linked to temperature fluctuations: in coastal Antarctica, where summer temperatures can reach 5°C and the active layer may be up to 3 m thick. In the Arctic regions, where summer temperatures reach 20°C, the active layers are extended to as much as 10 m thick [1]. These properties have been extensively studied in the past. Arctic Circle countries have conducted numerous collaborative studies on permafrost [2, 3]. The United States, Russia, and China are leading research on Antarctic permafrost at their respective research stations in Antarctica. In addition to polar regions, permafrost is being actively studied at high altitudes, such as the Alps and Himalayas, where it is relatively accessible. Recent research has increasingly linked these studies to global environmental changes [4-7]. Notably, advancements in drilling techniques in frozen environments and the accumulation of data on temperature and active layer variations have facilitated the development and implementation of theoretical models for permafrost and active layers [8].

Antarctica, known for its harsh environment, with 98% of the continent covered in ice and snow, is also home to flora and fauna adapted to these extreme conditions. Over recent decades, numerous countries have established research bases in Antarctica, and many researchers have studied its environment. South Korea established the King Sejong Station (KSS) in 1988 and has since investigated the distribution of vegetation, including mosses and lichens, as part of its environmental research. The vegetation has grown slowly, adapting to the low temperatures, strong winds, and heavy snowfall that characterize the Antarctic climate. Recently, however, global warming has enhanced photosynthesis and moisture, leading to vegetation distribution changes [9]. The changes in Antarctic vegetation are attributed not only to global warming but also to the effects of research station construction. In an environment with structures, snow cover changes during a snowstorm because the structures cause zonal variations in snow cover, typically resulting in zones of decreasing snow cover and zones of increasing snow cover [10]. Where snow cover increases, the distribution of vegetation is reduced relative to where it was before the snow cover increased [11]. The station buildings altered the local snowfall accumulation, causing localized environmental changes that affected vegetation. Specifically, such localized environmental changes have also been observed around the KSS. These changes in vegetation distribution are likely linked to changes in the active layer and permafrost, which require further investigation.

This study aimed to analyze changes in vegetation distribution due to local environmental changes caused by the buildings of the KSS in Antarctica by identifying changes in the subsurface characteristics. To achieve this, we employed time-lapse electrical resistivity survey, which is highly effective for characterizing frozen ground among geophysical survey techniques [12]. We conducted electrical resistivity surveys over 1 year to monitor temporal changes in the active layer and the upper frozen ground, a method also used for surveying landfills and seepage environments [13]. Kim et al. [14] used a dipole–dipole array with a 10-m electrode spacing to detect impermeable layers beneath the KSS, but they faced limitations in studying the active layers. To address these limitations, this study utilized both a dipole–dipole array and a Wenner array with very small electrode spacing.

In Section 2, we describe the active layer and permafrost to explain the characteristics of frozen regions. Section 3 analyzes the changes in the local environment and vegetation distribution around the KSS using meteorological and vegetation data. In Section 4, we present the results of electrical resistivity monitoring, examine the changes in the distribution of the active layer, and correlate these changes with the vegetation distribution. Finally, Section 5 summarizes and concludes the study.

Figure 1 illustrates the relationship between stratigraphic properties and temperature in frozen regions, with the active layer at the top and the permafrost below. At the interface between the active layer and permafrost, water from the thawing active layer moves through the pores of the medium, with the depth of this movement increasing with higher summer temperatures [15]. An unfrozen layer exists below the permafrost, maintained above 0°C by geothermal heat [16]. References to permafrost date back to the 1830s when the Royal Geographical Society instructed an official in Hudson Bay about the Yakutsk region of Russia. Systematic scientific studies began during World War II with the construction of the Alaska Highway, which extended from northern British Columbia, Canada, to the Yukon Territory. The term “permafrost” originated from the use of “permanent frost” by S. W. Müller in 1943 [15].

The active layer is the layer above the permafrost that thaws in summer and freezes in winter and is influenced by atmospheric temperature, snow cover, and surface water distribution. As temperatures rise above 0°C in summer and fall below 0°C in winter, the active layer undergoes freezing and thawing, creating an environment conducive to vegetation growth. When snow accumulates on the ground surface above the active layer, it acts as an insulator, reducing thermal conductivity and thus reducing the impact of atmospheric temperature. These effects cause temperature of the active layer to rise at a slower rate than the ground surface without snow cover during temperature transitions from freezing to thawing [17]. The thickness of the active layer is primarily influenced by summer surface temperatures: cooler temperatures result in a thinner active layer. The average thickness ranged from 0.75 to 3.12 m, with maximum values between 1.0 and 8.55 m for each measurement from the European Permafrost Monitoring Network [18]. Since the 1990s, systematic studies of permafrost and active layers in response to global temperature changes have contributed to paleoclimate reconstructions [19].

The study area is around the KSS (⁠$62°13`28``S$⁠, $58°47`22``W$⁠), completed in February 1988 on the Barton Peninsula, King George Island, South Shetland Islands. The geological environment of the Barton Peninsula comprises andesite and marine sediments from the Quaternary-Holocene age (Figure 2). The underlying geology had a density of 2.2 g/cm³, while the densities of the major rocks, conglomerate, gabbro, and quartz diorite are 2.5, 2.6, and 2.7 g/cm³, respectively [14]. The terrain surrounding the station is flat with a gentle northwest slope, and the surface geology consists of a mixture of soil and large gravel partially covered with moss and lichen.

Figure 3 presents the atmospheric and soil temperatures, soil moisture saturation (percentage of content in unit volume) measured in 2020, and monthly average snowfall from 1988 to 2017 observed at the KSS. Antarctica, which has opposite seasons to the Northern Hemisphere, has the warmest air temperatures in January and February and the coldest in July and August. Figure 3(a) shows the variations in atmospheric temperature, with a summer peak at approximately 7°C in February. The temperature remains above 0°C until the end of March, drops below freezing in early April, and stays below freezing from May. Mid-July temperatures are around −5°C, reaching a low of −20°C in August, before rising again in September and exceeding 0°C by November. Figure 3(b) shows soil temperature and soil moisture data collected at 1-hour intervals from sensors at depths of 40 and 60 cm, located 25 m along the survey line shown in Figure 4. From late March to mid-June, the ground temperature hovered around 0°C, indicating continuous freezing, whereas soil moisture experienced six significant fluctuations, reflecting repeated freeze-thaw cycles. From mid-June to the end of November, ground temperatures remained consistently below 0°C, and soil moisture remained below 10%, indicating frozen ground. Thawing begins in December when ground temperatures rise above 0°C, causing a spike in soil moisture. Figure 3(c) illustrates the average monthly snowfall from 1988 to 2017. Snowfall increases significantly in April when temperatures fall below freezing, and the 5 months from June to October accounts for two-thirds of the average annual snowfall. Beginning in November, snowfall typically decreases sharply. Air temperature and snow cover were measured approximately 100 m southwest of the survey area. In addition, soil temperature and moisture measurements were taken in the survey area.

Figure 4 displays an aerial image from August 2020 around the KSS, including the survey area, where two buildings including the Atmospheric Glacier Observation Building are identified. The black dotted box in Figure 4 indicates an area with light snow cover, while the yellow dotted box in front of the station building indicates an area with relatively heavier snow cover. The red dotted line marks the 63-m long survey line. The average elevation of the study area is approximately 10 m, with the 0-m point of the survey line being about 2 m higher than the 63-m point. The difference in snowfall accumulation between the areas in the black and yellow boxes, despite their proximity, can be attributed to the influence of the station building.

Figure 5 illustrates the effects of differential snow cover caused by the station buildings. These images present a time-lapse optical image of the survey transect area captured by a drone. In Figure 5(a), the red dashed line represents the survey line shown in Figure 4, and the blue box highlights the 25-m point along the line where temperature and humidity sensors were installed in depth of 0.4 and 0.6 m, respectively. The yellow triangles indicate the digging points marked at 10 and 45 m along the survey line, which were used to verify the actual subsurface conditions. Figure 5(a), taken in summer, shows a clear distinction between the optical properties of the two areas separated by the white dashed line. The left side of the image displays a greenish-brown coloration indicative of vegetation, whereas the right side shows no vegetation. Figures 5(b)–5(e) illustrate the snow cover changes in the study area in May, August, November, and December, respectively. The snow accumulation began in May and reached its peak in August, covering the entire study area. Notably, the right side of the image was consistently snowier than the left side (Figure 5(c)). In November (Figure 5(d)), the snow cover decreased during the thawing season, with the left side of the image revealing open ground first. Figure 5(e), taken in December, shows that the vegetated areas in Figure 5(a) are completely free of snow, while the nonvegetated areas remain covered. These images demonstrate that the snow cover over time is closely related to vegetation distribution.

To determine the cause of the localized changes in snow cover around the KSS, we analyzed wind observation data from the study area. Figure 6 shows wind rose diagrams of the study area. The average wind speed at the KSS over 30 years from 1988 to 2017 is approximately 8 m/s, with northwest winds shown in the wind rose diagram predominating and frequent, strong east winds are also observed (Figure 6(a)). During winter (April to October), east winds are notably stronger and coincide with heavier snowfall than in summer [20]. The wind observations for 2020 (Figure 6(b)) align with the 30-year average wind data, confirming northwest and east winds as the predominant wind direction. The predominance of east winds during winter is evident in the wind characteristics for winter 2020 in Figure 6(c), while northwest winds are the dominant wind direction in all seasons except winter, as depicted in Figure 6(d). Obstructions like facilities and buildings in the wind direction cause a rapid decrease in wind speed, creating vortices behind them and leading to significant snow accumulation on the opposite side of the wind direction [21]. This effect suggests that strong east winds carrying snow blow westward and are obstructed by the KSS buildings, resulting in high snow accumulation at the study site indicated by yellow dashed box in Figure 4.

We confirmed that the localized differences in snow cover caused by buildings during winter led to the vegetation changes identified in Figure 5. To more clearly quantify the changes in vegetation distribution, we used the Normalized Difference Vegetation Index (NDVI), which can be effectively obtained through remote sensing. Proposed by Rouse in the 1970s, NDVI quantifies vegetation by measuring the difference between near-infrared (NIR), which is strongly reflected by vegetation, and red light (RED), which is strongly absorbed. The primary wavelength band used is 860 $nm$⁠. NDVI is calculated using equation (1) [22, 23].

where NIR and RED represent the spectral reflectance in the near-infrared and red bands, respectively. More vigorous or denser vegetation reflects more NIR light and absorbs more RED than other wavelengths. NDVI values close to 1 indicate active vegetation growth, whereas values close to 0 indicate rocks and bare ground with minimal vegetation activity. NDVI has been used to study vegetation distribution in Antarctica [24]. Turner et al. [25] employed a multispectral sensor on an unmanned aerial vehicle (UAV) to capture high-resolution images of Antarctic vegetation, including small mosses and lichens. In this study, the NDVI values were obtained using a multispectral sensor mounted on a UAV. Figure 7 displays the NDVI values acquired in January 2018 at 0.25-m intervals along the survey line. NDVI values were not calculated beyond 43 m due to snow cover. The black solid line represents the raw observed values, while the blue solid line shows the moving average applied to identify trends. The NDVI values fluctuated along the survey line but remained relatively constant (above 0.25) until about 35 m, after which they showed a steady decrease. This trend aligns with the vegetated and nonvegetated areas identified in Figure 4(a). The decrease between 35 and 42 m on the survey line correlates with the transition zone from vegetated to nonvegetated areas observed in Figure 5(a).

Based on meteorological and vegetation data, we concluded that the localized changes in snow cover caused by the construction of the KSS altered the distribution of vegetation. The area to the west of the building, heavily covered with snow, has become an inhospitable environment for vegetation. In contrast, areas where snow melts in summer provide favorable environments for vegetation growth. Additionally, the distribution of vegetation is likely related to changes in subsurface properties.

The presence of water in soil pores is crucial for vegetation distribution. Electrical resistivity is highly sensitive to fluids in these pores [26]. The electrical resistivity related to fluid in the pores can be expressed as equation (2):

Here, $ρ$ and $ρw$ represent the effective electrical resistivity of rock and pore water, respectively. $s$ denotes fluid saturation, and $a$⁠, $m$ and $n$ range from 0.5 to 2.5, 1.3 to 2.5, and $n≈2$⁠, respectively [27]. During the summer thawing season in frozen regions, fluid saturation increases, and electrical resistivity decreases. Conversely, in the winter freezing season, the fluid in the pores turns into ice, fluid saturation decreases, and the electrical resistivity increases. Typically, the electrical resistivity of frozen soil exceeds 1000 Ωm and varies with the medium, temperature, and amount of ice distributed in the pores [28]. To determine whether the change in the local snow cover around the station, resulting from the construction of the KSS, has shifted from a favorable to an unfavorable environment for vegetation, and if this shift is related to subsurface environmental changes, particularly, fluid distribution characteristics, we conducted time-lapse electrical resistivity survey. This geophysical technique, which is highly sensitive to fluids in pores, helps identifying subsurface changes.

For the electrical resistivity survey conducted in this study, we installed a total of 64 electrodes at 1-m intervals to achieve high resolution. The electrodes were embedded at a depth of 40 cm after removing some surface gravel and backfilling with surrounding soil to ensure good coupling with ground. This installation occurred in early 2019 as the surface began to fully thaw. To allow for an adequate stabilization period around the electrode area, a dipole–dipole resistivity survey was conducted in February 2020, 1 year after the electrodes were installed. Additionally, Wenner array data were collected seven times between March 2020 and December 2020. The electrodes and wires were fixed in the study area for the duration of the survey to maintain consistent positioning of the electrode set-up throughout the period.

Although the resolution in an electrical resistivity survey varies depending on the geological environment, the horizontal resolution generally equals the electrode spacing, and the vertical resolution is considered to be half of the electrode spacing. The survey employed a Wenner array, which is ideal for horizontal investigations, and a dipole–dipole array, which is well-suited for two-dimensional investigations (Figure 8; [29]).

To obtain subsurface information from electrical resistivity surveys, imaging the electrical resistivity distribution through appropriate data processing is essential. The most commonly used method for this purpose is the iterative inversion method, which minimizes the difference between the observed and theoretically calculated values. In this study, the commercial program Res2Dinv [30] was used to obtain reliable subsurface electrical resistivity information. The program employs an iterative reweighted smoothness-constrained least-squares optimization method using equation (3) [31].

where $ΔP$ is the model perturbation, $gi$ is the data misfit vector containing the difference between the logarithms of the field data and modeled data, and $ri-1$ is the model parameter vector from the previous iteration. $J$ represents the Jacobian matrix of the partial derivatives, and $W$ is the roughness filter [32]. The damping factor, $λ$⁠, determines the relative importance of minimizing model roughness. $Rd$ and $Rm$ matrices are introduced to modify the weights assigned to different elements of the data misfit and model roughness vectors. The subscript $i$ denotes the number of iterations, and the superscript $T$ denotes the conjugate. In this study, the $L1$-norm roughness filter was applied because the subsurface resistivity in the study area did not vary smoothly.

Figure 9 displays the 2D distribution of electrical resistivity obtained using a dipole–dipole array in February 2020, when the air temperature around the KSS was above 0°C. A low electrical resistivity zone, with values of 500 $Ωm$ or less, is identified between approximately 5 and 35 m and to a depth of about 3 m. The boundary between this low resistivity and the higher resistivity zones, marked by the white dashed line, is interpreted as the boundary between the active layer and permafrost. The electrical resistivity up to a depth of 3 m between a distance of 35 and 42 m shows a transition zone, where the electrical resistivity gradually increases with distance. Beyond 42 m of the survey line, the resistivity is consistently high, exceeding 1000 $Ωm$⁠, indicating the absence of low resistivity characteristics that can be interpreted as a nonactive layer after 42 m. After 35 m, the resistivity at depths deeper than 3 m exceeds 3000 $Ωm$⁠, with a significant difference in resistivity characteristics between the southeast and northwest zones, centered around 35 m.

The electrical resistivity cross-sections obtained using a dipole–dipole array characterized the distribution of electrical resistivity to a depth of about 9 m, allowing us to identify the distribution of the active layer and permafrost. To compare the distribution characteristics of vegetation with subsurface electrical resistivity, a more detailed shallow cross-section of the subsurface electrical resistivity distribution is required. Consequently, a Wenner array was employed to obtain this detailed cross-section (Figure 10). Additionally, the subsurface electrical resistivity characteristics were verified through direct digging conducted in February during the thaw season at distances of 10 and 45 m along the survey line (Figure 11). Figure 10(a) shows the electrical resistivity distribution in March. Very low resistivity, indicating the active layer, extends to a depth of about 3 m at 35 m along the survey line. These results are consistent with those of the dipole–dipole array shown in Figure 9. At depths of about 0.5 m, resistivity values were higher than those below, likely due to the presence of a significant amount of gravel mixed in the soil, as evidenced in Figure 11. Up to a distance of 35 m, corresponding to the vegetation zone, water infiltrated the ground from the melting snow, saturating the pores of the medium above the permafrost. Figure 11 illustrates that water began to accumulate at a depth of approximately 0.6 m at 10 m along the survey line. However, groundwater does not appear to have spread to the northwest from approximately 37 m of the survey line.

The active layer was about 3 m thick between 0 and 35 m along the survey line due to sufficient solar energy supply and water infiltration during the thaw season. Beyond 35 m of the survey line, the boundary between the active layer and permafrost was progressively lower, and the active layer does not develop. Figures 10(b), 10(c), and 10(d) show cross-sections of the electrical resistivity in April, June, and September, respectively, corresponding to the freezing season. Based on Figure 10(a), the electrical resistivity increases from the surface, with fluid in the pores appearing to be frozen to a depth of about 1 m by June and extending to about 2 m by September. Figures 10(e), 10(f), and 10(g) show electrical resistivity cross-sections during the early thawing season. During this season, the electrical resistivity values at shallow depths varied from 40 m along the survey line. The high resistivity zone (5000 $Ωm$⁠) from 0 to 37 m along the survey line significantly thinned over time, while the resistivity value remained high at 5000 $Ωm$ beyond 40 m until December. In February, when the thickness of the active layer was expected to be the deepest, the result of digging to a depth of 1 m at 45 m along the survey line revealed that the soil was drier than at the 10-m point along the survey line.

A time-lapse electrical resistivity survey conducted in 2020 confirmed that the active layer, characterized by very low resistivity values at the KSS, extends well up to approximately 35 m along the survey line. The development of this active layer explains why vegetation is distributed in the area. The electrical resistivity of the active layer is related to the air temperature. In May, when the air temperature drops below 0°C, the electrical resistivity of the active layer increases. Conversely, in November, when the air temperature rises above 0°C, the electrical resistivity of the active layer decreases. These features, validated by optical imagery and vegetation data, persisted throughout the thawing and freezing seasons in June. Beyond the transition zone, where the electrical resistivity gradually increases, the active layer does not develop until the end of the survey line, regardless of the season. This indicates that the distribution of vegetation is closely linked to the development of the active layer.

In this study, we integrated meteorological, vegetation, and time-lapse electrical resistivity data to assess the impact of local changes in snowfall accumulation around the KSS on vegetation distribution. Data were collected from both vegetation and nonvegetation zones, with each dataset meticulously analyzed.

From the surface to a depth of 0.5 m, a zone of relatively high electrical resistivity was observed, which was attributable to a significant quantity of gravel mixed with the soil. The 0–35 m sections along the survey line experience minimal snow accumulation during winter, fostering robust growth of surface mosses and lichens, as evidenced by the highest NDVI values. This zone also experiences minimal snow accumulation in winter. At the 10-m point along the survey line, where digging was conducted, a significant amount of water was found to have been ponded at a depth of 0.6 m, indicating that it could supply moisture to the vegetation. The electrical resistivity data also revealed an active layer of below 500 $Ωm$⁠, estimated to be approximately 3 m thick.

From 35 to 42 m along the survey line, which was identified as a transition zone, vegetation decreased and NDVI values dropped rapidly. The electrical resistivity data indicate that groundwater migrating through pores in the active layer at depths of 1–3 m could not spread beyond this zone. During the freezing season, when temperatures drop below freezing, the upper part of the active layer begins to freeze, resulting in a high electrical resistivity zone, as observed in the electrical resistivity data from March to September. The active layer did not extend beyond the transition zone, even in summer. The lack of active layers in this zone is likely due to high snow accumulation caused by the station buildings, which reduces pore fluid, and air temperatures during the thawing season. Notably, the amount of snow cover between the two zones significantly affects the amount of solar energy absorbed and reflected by the ground surface. Zones with more vegetation, which absorbed solar energy in early November, began to thaw faster than zones with less vegetation, which did not absorb solar energy until late December. These observations indicate that changes in vegetation are due to variations in active layer development. Figure 12 shows the subsurface structure derived from meteorological and time-lapse electrical resistivity data. As depicted in Figure 12, the change in vegetation distribution is interpreted as a change in the distribution of active layers caused by differences in snowfall accumulation induced by the station building between the two zones.

In this study, we used meteorological and vegetation data to understand how changes in snow cover caused by the construction of the KSS affect vegetation distribution. Additionally, by identifying changes in the distribution of electrical resistivity in the subsurface through a year-long time-lapse electrical resistivity survey, we were able to identify the changes in the active layer where vegetation could grow. However, to more precisely delineate the boundary between the active layer and permafrost, it is necessary to conduct multigeophysical surveys, including multifrequency ground-penetrating radar and high-frequency seismic surveys.

The data of this study are available in the manuscript. Requests to access the datasets can be directed to the corresponding authors Joohan Lee and Won-Ki Kim.

The authors declare that they have no conflicts of interest.

This work was supported by Korea Institute of Marine Science & Technology Promotion (KIMST) grant funded by the Ministry of Oceans and Fisheries (KIMST RS-2021-KS211509), the Korea Polar Research Institute project PE24080, and by Global - Learning & Academic research institution for Master’s·PhD students, and Postdocs (LAMP) Program of the National Research Foundation of Korea (NRF) grant funded by the Ministry of Education (No. RS-2024-00445180).

We would like to thank the overwinter members of the King Sejong Station for their help and cooperation in obtaining the data.