Cliff retreat of permafrost coast in south‐west Baydaratskaya Bay, Kara Sea, during 2005–2016

Recent years of increasing air temperature in the Arctic have led to a significant increase in the rate of retreat of permafrost coast, which has threatened livelihoods and infrastructure in these areas. The Kara Sea hosts more than 25% of the total Arctic coastline. However, little is known about how coastal erosion in the Kara Sea may have changed through time, and the climatic and environmental drivers remain unclear. Here we study coastal dynamics along a 4‐km stretch of permafrost and sea‐ice‐affected coastline in south‐west Baydaratskaya Bay of the Kara Sea, western Siberia, between 2005 and 2016, by using handheld differential GPS mapping and satellite imagery. We identified temporal and spatial variations in the retreat rates, ranging between 1.0 (+0.1/−0.6) and 1.9 (+0.7/−1.3) m/yr over the studied coastline during 2005–2016. We also made ground temperature measurements, subsurface resistivity measurements and estimates of wave energy flux of wind‐driven ocean waves, to investigate the dominant climatic factors influencing the observed retreat rates through time. We found that wind‐driven wave activity during sea‐ice‐free days influences the magnitude of coastal retreat in the study area, while recent temperature rise has contributed less to enhancing coastal retreat during the study period. This suggests that the amount of eroded sediment and the associated release of nutrient to the nearshore zone are controlled by the magnitude of wave activity, which may influence infrastructure along the permafrost coast and marine ecosystems in the proximal ocean.

coasts. 6 However, little is known about how erosion of Arctic coasts may have changed through time, and the climatic and environmental drivers remain unclear. The rate of erosion of Arctic permafrost coasts currently averages 0.5 m/yr, 6 although this is known to be variable over a wide periods and for different coastal geomorphologies (e.g., ice-rich permafrost bluffs or barrier islands). 7,8 Along the Alaskan Beaufort Sea coast, mean annual rates of erosion range from 0.7 to 3.2 m/yr with maximum rates up to 17 m/yr. 9 In Muoskakh Island of the Laptev Sea, mean annual erosion during 1950-2014 ranged widely, between 2 and 25 m/yr. 10 In addition, increases in coastal erosion rates in recent decades have been reported along several Arctic coasts. For example, the mean annual rates along part of the Alaskan Beaufort Sea coast doubled between 1955 and 2007. 11 This acceleration is thought to be due to sea-ice loss that leaves the coast more exposed to wave action and storms. 12 In contrast, coastal accretion due to seaice decline has also been observed in the Alaskan Chukchi Sea and Canadian Beaufort Sea coasts. [13][14][15] Therefore, it is necessary to understand the coastal dynamics of permafrost coastal cliffs with regard to the factors (e.g., air temperature, solar radiation, ground ice content, sea-ice extent, wave action, storms, longshore sediment transport and major river discharge) that drive change in different Arctic coasts.
The Kara Sea coastline is longest among the ten Arctic Ocean sectors, 16 and comprises more than 25% of the total length of Arctic coasts. 7 Threats of coastal erosion to pipelines have encouraged an understanding of the processes of erosion in the Kara Sea. Studies have shown that mean annual rates of coastal erosion have ranged between 0.2 and 2.0 m/yr in most of the Kara Sea coasts (e.g., [17][18][19][20][21][22][23][24][25][26][27] ).
It has also been shown that thermodenudation and the action of slope processes alone can contribute up to 0.4 m/yr coastal retreat in the Kara Sea coasts. 19, 22 Yet, coastal dynamics along the Kara Sea coast are poorly reported 6

| STUDY AREA
Baydaratskaya Bay is a shallow gulf in western Siberia along the southwest margin of the Kara Sea ( Figure 1). The bay is thought to have been formed by glacial depression during the Late Pleistocene and subsequent submergence during the Holocene (e.g., 29 ). The Baydaratskaya Bay region is located in the continuous permafrost zone. 30 Ice starts to cover Baydaratskaya Bay between early October and early November, and becomes thickest generally by early May reaching up to 1.5 m. The bay is thus fully covered by continuous ice for around 7 months, and is ice-free for only 3-4 months. 31,32 Wind distribution observed at Ust-Kara station (see Figure 1 for location) shows that the prominent sector is north-west and north in summer and southwest and south in winter, 25,32 with an average wind speed of around 6 m/s during 2010-2016. Tidal range is 0.7-1.1 m although surges in autumn can reach 1.5-2.0 m. 25,32 Annual average temperature and annual precipitation are around −8°C and 300-400 mm, respectively. 32 The study area is located 2 km south-east of the mouth of the Oyu-Yakha River running between the islands of Trasayey and Levdiyev ( Figure 1). The Bovanenkovo-Uhta gas pipeline crosses Baydaratskaya Bay, transporting gas from the Yamal peninsula to Europe. The study area is located 0.5-4.5 km north-west of the Bovanenkovo-Uhta pipeline Cofferdam junction ( Figures 1 and 2 a,b).
In the south-east part of the study area, a low marine terrace with elevations of 3-6 m above the beach is developed along the coast for more than 4 km (e.g., Figure 3), extending from the Cofferdam junction. 33 This low marine terrace is composed of interbedded ice-rich silty clay to silty sand deposits (volumetric ice content of 50-70% in the upper 3 m) and is smoothly sloping. The low terrace decreases in height toward the north-west and then a laida (tidal flat) 1-2.5 m high is developed for 1.3 km, extending to 0.8 km south-east of the mouth of the Ngarka-Tambyakha River. 33 The higher part of the laida surface is located above the tide level and thus is flooded only due to seawater inflow during storm surges. The laida hosts numerous depressions with thaw lakes, several of which are being drained due to coastline retreat. Lake occupancy on the high laida exceeds 50%. The laida is composed of loams overlain by sands and pebbles. A high marine terrace 10-17 m high then extends for 1 km in the north-west part of the study area. 33 The high terrace is composed of silty sand interbedded with millimeter-thick peat layers and decimeter-thick ice layers (e.g., Figure 3). Total ice content in the cliff section is relatively high, ranging from 35% to 80%, due to massive ice inclusion and ice lenses. Slopes in the eastern part of the high terrace are generally steeper than those in the western high terrace (e.g., Figure 2c Table 1). An operator moved close to the cliff edge and placed the receiver exactly above the cliff for recording in a vertical position. The device antenna was differentially corrected to a reference station (66°53′46.502″E, 68°51′6.671″N; Figure 1). This allowed positional accuracies better than 0.2 m. Recordings were made every 0.5-1.0 m. We also used a geometrically corrected satellite image (DigitalGlobe product) taken Following the assessment of uncertainty as proposed previously (e.g., 15,34,35 ), the cumulative uncertainty in our retreat rates was   Table 1). The thermistors

| Resistivity measurement
The resistivity of frozen ice-rich material is significantly higher than unfrozen material due to the contrast in electrical properties between water and ice (e.g., 36 ). The transient electromagnetic method (TEM) was used at the Cofferdam site in the eastern study area (see

| Wave energy flux of wind-driven ocean waves
We estimated the wave energy flux (power per meter of wave crest) of wind-driven ocean waves along the south-west Baydaratskaya where ρ is the density of water (1000 kg/m 3 ), g is gravitational acceleration (9.82 m/s 2 ), H (m) is the significant wave height and T (s) is the peak wave period. 38 Because we did not observe a significant wave height and a peak wave period, we approximated the wave energy flux based on wind shear stress on the water surface. 39 The interfacial shear stress was calculated using daily

| Ground temperature
Ground temperature profiles were successfully obtained from bore-

| Resistivity cross-sections
At the Cofferdam site, we imaged resistivity cross-sections extending 50 m along three parallel lines perpendicular to the coastline by TEM measurements in September 2015 and September 2016 ( Figure 6).

| Wave energy flux of wind-driven ocean waves
We ranging from 16×10 9 to 24×10 9 J/m, as the number of ice-free days during this period did not vary considerably.

| Coastal retreat in Baydaratskaya Bay
Numerous studies have looked at the coastal dynamics in the Kara Sea coasts since the 1960s based on field measurements and analyses of satellite and aerial images (e.g., [17][18][19][20][21][22][23][24][25][26][27][28]   0.2 m/yr (e.g., 18,21 ) or were in fact not retreating (e.g., Figure 1). 7 The Gulf of Yenisei coast was also reported to be retreating relatively slowly, between 0.2 m/yr and 0.4 m/yr. 21,22 The Gulf of Ob coast was eroding at moderate rates for the Kara Sea coasts, with mean erosion rates of 0.33-0.7 m/yr. 21,22 In contrast, the Western Yamal   19 We also found geomorphological variations in coastal retreat rates during the study period. For example, the low- This fast retreat might be due to that the lowelevation laida was readily exposed to the activity of strong waves that enhanced erosion. We therefore discuss below some of the climatic drivers that may influence the observed cliff retreat rates.

| Influence of temperature rise and snowfall days
Thermal insulation with snow accumulation and temperature rise has been reported to result in the retreat of permafrost coasts. In the Mackenzie Delta area of north-west Canada, multiple retrogressive thaw slumping is enhanced once landslides occur, because less vegetation leads to higher temperatures and thermal insulation associated with snow accumulation in winter. 45 In Muostakh Island of the central Laptev Sea, the mean annual rate of coastal thermo-erosion correlates strongly with mean air temperature in summer. 46   during the period. These findings therefore suggest that temperature is not a primary driver that enhances yearly coastal retreat in the southwest Baydaratskaya Bay coast, although a more definitive correlation cannot be studied without further ground temperature monitoring.

| Influence of wind-driven ocean wave activity
Having ruled out that the observed cliff retreat has changed primarily based on temperature variations during the study period, we evaluated the relationship between cliff retreat rates and total wind-driven ocean wave energy. The role of wind activity on erosional processes along the permafrost coast has not been clear. Leont'yev 48 was the first to ascribe that wind-driven wave could be the primary driver determining coastline retreat in the Kara Sea coast. Vasiliev et al. 19  wind-wave energy received during the ice-free periods of interest is higher (e.g., Pearson correlation coefficient between mean annual retreat rate and mean monthly wind-wave energy is r = 0.99). This suggests that northerly wind-driven wave activity during ice-free days is important in determining the magnitude of yearly coastal retreat. Wave activity may therefore lead to higher coastal retreat rates along the south-west coast than the south-east coast of Baydaratskaya Bay.
Moreover, the difference between Q 3/4 and Q 1/4 of the observed cliff retreat rate over the entire studied coast in a given period is larger (i.e., the retreat rate over the entire studied coast is more variable) when the received monthly wind-wave energy during ice-free days is higher ( Figure 8), suggesting that stronger wave activity makes the magnitude of coastal retreat within the studied coast spatially more variable. This is probably due to different morphology and lithology among the studied coastal sections that differ in their resistance to wave activity.
Further continuous monitoring and observations are necessary to verify this hypothesis, as the larger variation may also result from differences in the wind-wave energy for the study periods.

| Release of sediment and organic carbon into the ocean
Erosional processes in the permafrost coast result in release of large amounts of sediment and organic matter to the nearshore zone. 3 Bay, as discussed above, the amounts of sediment and organic carbon released are expected to vary with the magnitude of wave activity.
We first quantify the sediment volume released to the nearshore zone associated with coastal erosion during 2005-2016 using the profile of cliff height 33 22 Therefore, the release of eroded sediment to the nearshore zone may influence the marine ecosystem of Baydaratskaya Bay through increasing ocean turbidity, acidity and organic carbon supply (e.g., 4 ), and is particularly important in summer when river discharge is low (e.g., 52 ). Although it is difficult to quantify its impact on the marine ecosystem, the quantities of sediment and nutrient supplied to the ocean are considered to have changed with the magnitude of wind-driven ocean wave activity during sea-ice-free days. The quantification of sediment volume together with further continuous monitoring may help to assess the long-term risk to livelihoods and infrastructure along the Kara Sea coast. We conclude that wind-driven ocean wave activity during the sea-ice-free period influenced the magnitude of retreat, whereas recent temperature increases contributed less to enhancing coastal retreat during the period. This suggests that wave activity has controlled the amount of eroded sediment and organic matter released from the permafrost coast to the nearshore zone, perhaps influencing infrastructure along the coast and the marine ecosystem in Baydaratskaya Bay. We also hypothesize that the stronger wave activity has increased variability in the magnitude of coastal retreat within the same coastal section, and this needs to be examined through further continuous monitoring.