Effect of ephemeral snow cover on the active layer thermal regime and thickness on CALM-S JGM site, James Ross Island, eastern Antarctic Peninsula

This study aims to assess the role of ephemeral snow cover on ground thermal regime and active layer thickness in two ground temperature measurement profiles on the Circumpolar Active Layer Monitoring Network – South (CALM-S) JGM site on James Ross Island, eastern Antarctic Peninsula during the high austral summer 2018. The snowstorm of 13–14 January 10 created a snowpack of recorded depth of up to 38 cm. The snowpack remained on the study site for 12 days in total and covered 46% of its area six days after the snowfall. It directly affected ground thermal regime in a study profile AWS-JGM while the AWS-CALM profile was snow-free. The thermal insulation effect of snow cover led to a decrease of mean summer ground temperatures on AWS-JGM by ca 0.5–0.7 °C. Summer thawing degree days at a depth of 5 cm decreased by ca 10% and active layer was ca 5–10 cm thinner when compared to previous snow-free summer seasons. Surveying by ground penetrating radar 15 revealed a general active layer thinning of up to 20% in those parts of the CALM-S which were covered by snow of > 20 cm depth for at least six days.


Introduction
Seasonal occurrence of snow cover, its thickness and spatial distribution significantly affect ground thermal regime as well as active layer refreezing in regions underlain by permafrost. It is estimated that a snow accumulation about 40 cm deep has the 20 maximum insulating effect (e.g., Zhang, 2005). Generally, thinner snow accumulation tends to have a cooling effect on the ground while a thicker and long-lasting snow cover can increase the ground temperature. The insulating effect of snow cover in permafrost conditions is the most prominent in winter due to maximum snow thickness, high surface albedo and the porosity of snow, and a large temperature difference between the atmosphere and the ground surface (Callaghan et al., 2011). Under these conditions, the solar energy absorbed by the snow surface is reduced, thermal conduction within the snowpack is low, 25 and the transfer of heat between the air and the soil surface is limited. The net influence of snow over the cold period amounts to an increase of mean ground temperatures and maximum thaw depths, but these effects vary greatly in space and time (Zhang, 2005). At the end of winter the insulating effect of snow decreases, and meltwater becomes the primary influence. During the summer period, the effect of snow on ground temperature and thaw depth is considered to be relatively small under permafrost conditions (Goodrich, 1982). The ground remains frozen underneath long-lasting snow patches but warms rapidly 30 in snow-free areas (Harris and Corte, 1992). The infiltration of meltwater increases the thermal conductivity of the ground, enhancing heat transfer from the ground surface to the permafrost table (Campbell et al., 1998). When ephemeral snow cover builds up during the summer, snow protects the ground from warm air and the active layer may eventually refreeze from the permafrost table upwards. However, this cooling effect is limited and its impact on the ground is controlled primarily by the duration of the snow cover, snow properties and a temperature difference between the air and the ground surface. The insulating 35 capacity of snow increases in late summer when the thaw plane tends to rise rapidly (Goodrich, 1982).
The snow cover effect on ground thermal regime or active layer thawing has been widely studied in the Arctic regions during the past decades (e.g., Zhang, 2005;Callahan et al., 2011;Johansson et al., 2013;Park et al., 2015). In contrast, the knowledge on snow-ground interactions in Antarctica is limited mostly to sites within the Antarctic Peninsula region (AP) with only a few studies from different parts of the continent. Generally, snowpack controls the ground thermal regime and active layer 40 thickness in Antarctica in different ways depending on its seasonal duration and prevailing depth. Only a negligible effect of winter snowpack on ground thermal regime was observed in dry and cold conditions of the north-eastern AP where snow cover in winter is irregular and usually thinner than 30 cm (Hrbáček et al., 2016). A ground temperature increase was observed in cases when snow persists for the majority of the winter with a thickness between 30 and 70 cm. Such conditions are characteristic for the study sites on the South Shetlands in the north-western AP (Oliva et al., 2017a;de Pablo et al., 2017;45 Ramos et al., 2017) and they were also observed in topographically similar sites on James Ross Island .
Importantly, ground warming during the thaw season can lead to permafrost degradation especially in areas on the border conditions between continuous and discontinuous permafrost of the north-western AP (Hrbáček et al., 2020).
At the same time, however, persistent snowpack can lead to active layer thinning and its effect can eventually promote permafrost aggradation. The main reason is a shortening of the thawing season (Ramos et al., 2017). Consequently, the heat 50 deficit reduces active layer thawing propagation. Such conditions were observed on the South Shetlands after 2010 (de Pablo et al., 2017;Ramos et al., 2017) and they were related to a recent climate cooling (Turner et al., 2016;Oliva et al., 2017b) and an associated increase in snow precipitation within the AP region (Carrasco and Cordero, 2020). Similar effect of long-lasting snow cover was reported from the Victoria Land where the presence of snow prevented the active layer from thawing in the summer completely (Guglielmin et al., 2014a). General cooling of the active layer and permafrost was further observed in 55 areas where snowpack exceeded 1 m, as reported by Guglielmin et al. (2014b) from Adelaide Island. According to Ramos et al. (2020), permanent accumulation of > 4.5 m can even lead to a complete insulation of the ground from atmosphere, causing permafrost aggradation and forming a specific subnival thermal regime of the ground.
This study brings a new perspective on the effect of snow on ground thermal regime and active layer thickness. We assess the thickness between the snow-covered and snow-free sites, and (iii) determine the effect of ephemeral snowpack on active layer thaw depth.

Study area 65
James Ross Island is located in the north-eastern sector of the AP (Fig. 1). The northern part of James Ross Island, Ulu Peninsula, constitutes the largest ice-free area in the entire AP. The mean annual air temperature in the lowland parts of JRI (<50 m asl) was -7.0 °C in the period 2006-2015 with a positive trend of annual temperatures but a cooling trend in the summer (DJF) months (Hrbáček and Uxa, 2020). The annual precipitation is estimated to about 300-700 mm w.e.y -1 (van Wessem et al., 2016, Palerme et al., 2017. Snow cover is distributed irregularly as indicated by perennial snowfields on lee-slopes and 70 snow patches around topographic obstacles. Snow thickness in the flat lowland areas is usually less than 20 cm, but can rarely exceed 40 cm (Hrbáček et al., 2016. The ice-free area is underlain by continuous permafrost with a modelled temperature between -4 °C and -8 °C (Obu et al., 2020). The active layer thickness strongly depends on lithology and varies between 50 and 120 cm (Hrbáček et al., 2019).

Figure 1 Regional setting (left) and the location of CALM-S JGM site (right)
The study site is located on the CALM-S JGM which is the part of a continental database (REF). The area of 80 m by 70 m was established in February 2014 following recommendations for CALM-S sites proposed by Guglielmin (2006). The area comprises diverse geological units, with a Holocene marine terrace covering ca 80% of the northern part of the site and 80 Cretaceous Whisky Bay Fm. forming the remaining 20% of the site (Fig. 2). The geological conditions lead to a variability of https://doi.org/10.5194/tc-2021-5 Preprint. Discussion started: 4 February 2021 c Author(s) 2021. CC BY 4.0 License. general ground physical characteristics and consequently drive the variability of active layer thaw depth which is about 20 cm thicker in the southern part underlain by the Cretaceous sediments (Hrbáček et al., 2017).

Temperature monitoring and data processing 90
There are two automatic weather stations (AWS) installed within the CALM-S JGM grid. The fully equipped site AWS-JGM is located in the part formed by the marine terrace. It provides data on air temperature at 2 m above ground measured by Pt100/8 sensor (accuracy +-0.15 °C) placed in a radiation shield, and ground temperature measured by Pt100/8 thermometers Climate data were analysed for the austral summer periods 2016/17 and 2017/18. Particularly, the following parameters were 100 calculated: a) Daily and seasonal mean air and ground temperatures at depths of 5, 50 and 75 cm; b) Thawing degree days of air and (TDDA) ground at a depth of 5 cm (TDD5) calculated as a sum of positive daily temperatures; c) Isothermal days at a depth of 5 cm set as an event with the maximum and minimum daily temperatures within the interval 105 0.5 °C to -0.5 °C (Guglielmin, 2006).
The daily variability of active layer thaw depth is defined as a maximum daily depth of 0 °C isotherm for the entire period of thawing season and the maximum seasonal active layer thickness is calculated accordingly.

ALT measurements
Mechanical probing and ground penetrating radar (GPR) soundings were used to determine seasonal changes of active layer 110 thickness. Thaw depths were measured on 31 January and 12 February during the summer period 2016/17 and on 20 January, 12 and 24 February in the 2017/18 summer period. The gridded sampling design with evenly distributed nodes at 10 m spacing was used for the probing, which was carried out at each of the grid nodes with a 1 cm diameter steel rod of 120 cm length.
GPR data were collected along nine parallel profiles with north-south orientation, four profiles ( Fig. 2) were selected for a detailed analysis. GPR sounding was carried out using a shielded 500 MHz antenna and RAMAC CU-II control unit (MALÅ 115 GeoScience, 2005). A wheel encoder was used to trig the measurements and control the distance along the profiles. The time window was set to 54.6 ns and scan spacing to 0.049 m. GPR data were processed using the REFLEXW software version 8.5 (Sandmeier, 2017). The depth axis of raw GPR profiles was converted from the time axis using the wave velocity determined from thaw depths probed at grid nodes and from the position of relevant reflector in GPR scans. The velocity obtained for each sampling date ranged between 0.072 and 0.094 m ns -1 increasing towards the late summer season with soil moisture (Fig.  120 3). The volumetric soil moisture was set as a mean value of 72 measurements recorded at each grid point of CALM-S JGM within 12 cm thick subsurface layer of the ground using the HydroSense II measuring device paired with the CS658 soil-water sensor (Campbell Scientific, Inc.).  velocities in active layer determined for the conversion of two-way travel times to depths.

Snow cover observations
A long-term monitoring of snow depths is provided at 2-hour interval using the Judd (accuracy +-2 cm) ultrasonic sensor

Snow cover at CALM-S JGM site in January 2018
Snowstorm occurred between 13 January 22:00 UTC and 14 January 4:00 UTC. A relatively short period of 6 hr was sufficient to create snow cover that reached a maximum depth of 38 cm at AWS-JGM site. The most rapid thinning of the snowpack was observed on the 16 January when the snow depth decreased by 8 cm as a result of relatively high air temperatures that exceeded 170 6.0 °C in maximum. Snow persisted until 24 January when the last shallow snowpack (3 cm) melted (Fig. 5).
The distribution and thickness of snow were surveyed six days after the snowstorm, when snow cover extent was reduced to 46% of the CALM-S JGM area. The ground plan of the snow patch was complex and snow depth variable (Fig. 5) due to the redistribution by strong wind during the snowstorm and irregular melting after the storm. Snow drift produced irregular pattern of elongated zastrugi features and interleaving troughs with the SW-NE orientation. Owing to the drift, snow depth ranged 175 from 20-30 cm on the ridges to less than 5 cm in troughs. While the maximum snow depths determined at probed profiles was 25 cm, the modelled mean snow depth was 10 cm. Within the snow patch, wind-blown ground appeared at seven places that increased rapidly in extent due to the enhanced snowmelt on the margins of bare ground.

Figure 6 Seasonal evolution of active layer thaw depth and thawing degree days (TDD5) on AWS-JGM and AWS-CALM in the seasons 2016/17 and 2017/18. Dots indicate maximum thaw depths on both AWS sites. Grey rectangle indicates the period with compact snow cover detected on AWS-JGM.
The difference in ALT between the two subsequent summer seasons is clearly visible in Fig. 7 that also reveals relatively even thaw plane during the summer 2016/17 and its irregular course in January 2018. The largest interannual difference in thaw 200 depths coincides with those sections of the scanned profiles that were covered with thick continuous snowpack during the January 2018 measurement in profiles X20 and X40 (Fig. 7). Within these sections, the thaw plane is located 20 to 30 cm higher compared to the snow-free parts of the profiles. The thaw plane along the predominantly snow-free profile 70 in January https://doi.org/10.5194/tc-2021-5 Preprint. Discussion started: 4 February 2021 c Author(s) 2021. CC BY 4.0 License.
2018 is more or less parallel to the permafrost table observed in 2017 (Fig. 7).This is also the case of February measurements in all measured profiles. 205

Figure 7 Snow cover depth (grey polygons) in January 2018 and thaw depths in January and February 2017 and 2018 determined from GPR measurements at CALM-S JGM site.
Spatial distribution of thaw depth for the CALM-S JGM site during the late summer 2016/17 and 2017/18 is illustrated in Fig.   8. Thaw depths were more spatially uniform in February 2017 although two geologically distinct parts of the study area are 210 clearly visible. The bimodal distribution with prevailing lower ALT (< 80 cm) and thaw depths as high as 115 cm along the southern margin of the study area have already been observed over the years 2014-16. This large contrast in thaw depth has been tentatively attributed to different grain size distribution and moisture content in two sedimentary units within the study area (Hrbáček et al., 2017). The strongly bimodal pattern can be seen in the 2018 grid that shows even more pronounced difference in thaw depths between the two parts. The region of decreased ALT coincides with maximum snow depths in January 2018 that is also consistent with the difference between mean thaw depths over the period 2014-17 and 2018 grid (Fig. 8D).

Climate conditions in the study area in the context of previous years
Air temperatures in the study area during the summer periods 2016/17 and 2017/18 were below long-term (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) summer average reported for the AWS-JGM and nearby Abernethy Flats sites (Hrbáček and Uxa, 2020). Even though mean air temperature in the summer was the lowest in 2014/15, the year 2017/18 had the lowest TDDA (Table 2)  were the lowest in the season 2017/18. Noticeable decrease of the mean seasonal ground temperatures was recorded primarily on AWS-JGM. In 2017/18, the mean summer ground temperature on AWS-CALM was higher than on AWS-JGM.  Hrbáček et al., 2017. Regardless of relatively cold summer ground surface temperatures, the thickest active layer and the highest TDD5 on both AWS-JGM and AWS-CALM were observed in 2016/17 (Table 3). This was primarily caused by unusually warm spring months (September to November) in the north-eastern part of the AP (Turner et al., 2020). Such conditions favoured early 235 melting of the active layer in the beginning of October which started weeks earlier than usually (Hrbáček et al., 2017;Hrbáček and Uxa, 2020). Sandy soil texture together with lower moisture also supports warmer surficial conditions on AWS-JGM as documented by TDD5 increase by ca 28-50 °Cdays. However, TDD5 were 12°C days lower in 2017/18 on AWS-JGM than on AWS-CALM.

Snow cover effect on ground thermal regime
Variations in summer air temperature and the length of the thaw season control ALT but snow cover may reduce it significantly 245 (e.g., Guglielmin, 2004;Zhang, 2005;Guglielmin et al., 2012). While a cooling effect of thick long-lasting snow patches has been reported from different regions including AP (Guglielmin et al., 2014b;Ramos et al., 2017;2020), the influence of thin snow cover that rapidly builds and melts during the warm period remains unknown. The ground thaw depths recorded in January 2018 at the CALM-S JGM site demonstrate a decrease in active layer thickness underneath ephemeral snow cover https://doi.org/10.5194/tc-2021-5 Preprint. Discussion started: 4 February 2021 c Author(s) 2021. CC BY 4.0 License.
with limited depth that protected the ground surface only for six days prior to ALT measurements. The initial maximum snow 250 depth of 38 cm falls within the range of 30-70 cm that is considered sufficient to isolate the ground from temperature variations (e.g., de Pablo et al., 2017;Kňažková et al., 2020). However, the maximum snow depth of 25 cm observed on the CALM-S JGM at the time of ALT measurements indicates much lower snow depth over most of the study area. Considering that ALT thinning was observed far beyond the margin of the snowpack, a mean snow depth of 10 cm has sufficient influence on the ground thermal regime. This value is significantly lower than the effective snow depth of 20 cm considered with respect to the 255 insulation effect of snow in AP region (de Pablo et al., 2016;Oliva et al., 2017a).
The largest impact of ephemeral snow cover on thaw depths in the study area can be seen underneath thick continuous snowpack in profiles X20 and X40 (Fig. 7), where active layer thinning reaches 20 to 30 cm. By contrast, a slight overall decrease (lower than 10 cm) in ALT was observed along profile X50 across irregularly thick snowpack with frequent holes and at predominantly snow-free profile X70 (Fig. 7. The difference in thaw depths between snow-covered and bare surface 260 decreases rapidly after the snowmelt. The snow cover disappeared from the study area on 24 January and measurements from 12 February reveal the unimodal distribution of ALT. The rapid restoration of a more or less uniform thaw plane results from the combination of relatively high air temperatures and an enhanced conduction of heat from the ground surface due to meltwater percolation (sensu de Pablo et al., 2014;Farzamian et al., 2020). The influence of ephemeral snow cover on thaw depths is apparent in the timing of the maximum ALT on snow-covered and bare sites. Whereas the maximum ALT on snow-265 affected AWS-JGM site was observed before the snowstorm event on 9 January, the ALT maximum on a snow-free site occurred on 18 February. The later culmination of the thaw depth at the snow-free site coincides well with the timing of the maximum ALT in previous summer seasons which is typically around mid-February (Hrbáček et al., 2017).
The ground temperatures recorded at the CALM-S JGM indicate that ephemeral snow cover may also affect thermal regime of the active layer. As Fig. 4 shows, the presence of snow cover with an initial depth of 38 cm eliminates diurnal temperature 270 fluctuations down to a depth of 50 cm and results in a substantial decrease in the mean daily ground temperatures. The mean monthly ground surface temperature in January is reduced by nearly one degree below the snow cover (with the depth decreasing from 38 to 12 cm over six days) compared to the bare surface (Tables 1 and 2). Although the net cooling effect of ephemeral snow over the summer season is limited, the number of isothermal days increases and TDD5 are reduced by ca 10% compared to snow-free summer seasons (Tables 1 and 3). This is in line with the reported insulating effect of snow patches 275 that persist over the warm season in permafrost conditions (de Pablo et al., 2017). Guglielmin et al. (2014b) and Oliva et al. (2017a) reported a cooling of the ground surface and a reduced magnitude of ground temperature fluctuations on Adelaide Island, western AP;and Livingston Island, north-western AP, respectively. De Pablo et al. (2017) observed an increase in the minimum ground surface temperatures below quasi-permanent snow patches on Livingston Island, north-western AP. The observed changes in recorded ground temperatures confirm that the presence of snow cover may be more important for the 280 active layer refreezing and thaw plane position than summer temperatures (Zhang, 2005;de Pablo et al., 2017;Ramos et al., 2017).

Assessment of the results in a wider regional context of Arctic/mountain regions
The previous studies on JRI considered the winter snowpack as too thin and temporally unstable to create a sufficient insulation layer on the ground (Hrbáček et al., 2016). This is mostly because of dynamic weather conditions during winter and prevailing 285 strong southern winds which remove the snow from flat surfaces resulting in its irregular distribution within the terrain . In specific conditions of long-lasting snow accumulations, a snowpack more than 30 cm deep causes isothermal regime of ground temperature during the winter months .
Most of the studies dealing with a relationships between snow and the ground thermal regime in Antarctica are located in the South Shetlands in the north-western AP. The oceanic climate makes this part of AP one of the wettest regions in Antarctica 290 with precipitation rates of more than 1000 mm yr -1 (Carrasco and Cordero, 2020). This results in high annual net snow accumulation and snow cover depth, with mean values of about 0.5 m (de Pablo et al., 2017). Such snowpack can cause a significant warming of the ground compared to windswept areas (Oliva et al., 2017a). As the northern AP is located in border conditions between continuous and discontinuous permafrost (Bockheim et al., 2013;Obu et al., 2020), regular and thick snowpack in the winter months can lead to permafrost degradation (Hrbáček et al., 2020). On the other hand, mean annual air 295 temperature has decreased at a statistically significant rate since the beginning of the 20 th century, with a most rapid cooling during the summer season (Turner et al., 2016, Oliva et al., 2017b. This was reflected in an increase in the number of snowfall events on the north-western side of the AP since the mid-2010s during the summer period (Carrasco and Cordero, 2020).
Despite the fact that the trends in the annual total of precipitation from extreme precipitation events over 1979-2016 are small (Turner et al., 2019), positive trends in the precipitation total and summer snowfall events may result in a more frequent 300 accumulation of ephemeral snow cover with a temporary cooling effect.
This study brings new insight into the functioning of the snow cover in relation to active layer thickness. So far, the active layer thinning was attributed mostly to long-lasting snow accumulation which prevented active layer thawing during the early warm season. Therefore, its thinning was directly related to the shortening of the thawing season (Ramos et al., 2017;de Pablo et al., 2018). Consequently, the long-term active layer thinning rate was more likely connected with an increased duration of 305 the snow cover, rather than with the climate variability during summer months (Ramos et al., 2017). In other Antarctic regions, like Victoria Land, snowpack can persist for an entire summer and completely prevent active layer thawing (Guglielmin et al., 2014a). However, the ephemeral snowpack's role in summer has not been reported yet from neither Antarctic nor Arctic regions. The influence of summer snowfalls and related snowpack on the thermal regime and thickness of the active layer has been reported from alpine areas (Hanson and Hoelzle, 2004;Magnin et al., 2017;Zhao et al., 2018), but snow and ground 310 conditions in the mountains differ strongly from those in high latitudes (Hoelzle and Gruber, 2008). Usually, the period of high summer is characterized by a low snowfall intensity and predominantly snow-free ground surface. Therefore, snowfall events leading to an accumulation of more than 30 cm deep snow cover are scarce.

Conclusions
This study brings a new perspective on the effect of snow cover on active layer thermal regime and thickness in the Antarctic 315 Peninsula region. The recent research deals mostly with the role of winter or permanent snow cover, whereas the role of ephemeral snow occurring during the high summer was not documented at all in Antarctica. This study shows that even shortterm presence of a relatively thick snow cover during the high summer can significantly affect active layer thermal regime and thawing propagation. Considering the observations from 2017/18 within the context of the snow-free seasons with similar air temperature conditions, the ephemeral snow cover considerably affected the ground thermal regime and active layer thickness. 320 The magnitude of the snow insulation effect was controlled by the snow depth and duration of snowpack. The observed snowpack with an initial snow depth of 38 cm lasted for 12 days during the high summer season. Below the ephemeral snow cover, the diurnal temperature fluctuations were limited down to a depth of 50 cm, thaw depths decreased by nearly 20%, the maximum active layer thickness was observed almost six weeks prior to the maximum on a snow-free site, and the mean monthly ground temperature was reduced by nearly one degree compared to the snow-free surface. The number of isothermal 325 days increased and TDD5 were reduced by ca 10% in the presence of snow cover.

Data availability
All data are available upon request to FH.

Author contribution:
FH and ZH designed the research hypothesis, wrote the manuscript and participated in fieldwork. MK and JS participated in 330 fieldwork.

Competing interests
The authors declare that they have no conflict of interest.