Active layer thickening and controls on interannual variability in the Nordic Arctic compared to the circum‐Arctic

Active layer probing in northern Sweden, northeast Greenland, and central Svalbard indicates active layer thickening has occurred at Circumpolar Active Layer Monitoring (CALM) sites with long‐term, continuous observations, since the sites were established at these locations in 1978, 1996, and 2000, respectively. The study areas exhibit a reverse latitudinal gradient in average active layer thickness (ALT), which is explained by site geomorphology and climate. Specifically, Svalbard has a more maritime climate and thus the thickest active layer of the study areas (average ALT = 99 cm, 2000–2018). The active layer is thinnest at the northern Sweden sites because it is primarily confined to superficial peat. Interannual variability in ALT is not synchronous across this Nordic Arctic region, but study sites in the same area respond similarly to local meteorology. ALT correlates positively with thawing degree days in Sweden and Greenland, as has been observed in other Arctic regions. However, ALT in Svalbard correlates with freezing degree days, where the maritime Arctic climate results in relatively high and variable winter air temperatures. The difference in annual ALT at adjacent sites is attributed to differences in snow cover and geomorphology. From 2000 to 2018, the average rate of active layer thickening at the Nordic Arctic CALM probing sites was 0.5 cm/yr. The average rate was 1 cm/yr for Nordic Arctic CALM database sites with significant trends, which includes a borehole in addition to probing sites. This range is in line with the circum‐Arctic average of 0.8 cm/yr from 2000 to 2018.

and northern Sweden ( Figure 1). These CALM sites have been operated consistently as part of different monitoring programs, resulting in 41 years of data in northern Sweden, 23 years in northeast Greenland, and 19 years in central Svalbard. Existing indepth analyses of these data series are over a decade old, and have never been compared in a regional study.
The Nordic region encompasses a continuum of permafrost types.
Isolated permafrost exists in palsas and peat hummocks in Iceland, Norway, Sweden, and Finland, 5,6 while permafrost is continuous in the periglacial landscapes of Svalbard and most of northern Greenland. [7][8][9][10] Mean annual ground temperatures (MAGTs) in permafrost in this region are higher than in other high-Arctic locations because of the climatic influence of the warm Norwegian Current, which extends into the West Spitsbergen Current in the eastern side of the Fram Strait. 11,12 However, Greenland has a relatively cold climate and thus thinner ALT due to more extensive sea ice and the cold East Greenland Current flowing south along Greenland's shelf. 13,14 The contrast of these currents and their influence results in a large climatic gradient across the Fram Strait. The purpose of the present study is to assess how this climatological setting, in addition to site geomorphology, affects ALT, and to determine the magnitude and variability in active layer thickening at the Nordic CALM probing sites with continuous, long-term records. These results are contextualized in the circum-Arctic through the analysis of ALT data from all Arctic sites in the CALM database. 15

| Study areas
There are six active CALM sites in the Nordic region; this study uses data from the five sites that have active, continuous mechanical probing data series of at least 10 years. These five sites are operated in fine-grained sediments or peat, allowing the probing technique to be used, and are located within three study areas. The utilized data series are 19-41 years long, allowing for the assessment of meteorological control on ALT trends and interannual variability. All data are available through the CALM database. 15

| Adventdalen, Svalbard
The UNISCALM site is located at 78 N in central Spitsbergen, Svalbard, in the Adventdalen sediment-infilled fjord-valley ( Figure 2). Permafrost is continuous in Svalbard's periglacial landscape, which exhibits several permafrost-controlled landforms including pingos and ice wedges. 9 Svalbard has a maritime Arctic climate, with winter air temperatures substantially higher than those at other Arctic sites of the same latitude. 16,17 Mean annual air temperature (MAAT) at the Svalbard Airport, located 10 km northwest of the UNISCALM site, was −5.9 C from 1971 to 2000, and mean annual precipitation during the same period was 196 mm. 18 In 2018, MAAT was −1.8 C and annual precipitation was 177 mm. 19 From 1971 to 2017, MAATs at Svalbard meteorological stations increased 3-5 C, with the largest seasonal temperature increase occurring during winter and the smallest during summer. 18 During this same period, precipitation has generally increased in autumn and winter, and decreased in spring and summer, although few of these trends are significant. 18 UNISCALM is located 10 m a.s.l. in a loess deposit on a flat terrace of the Adventelva river plain. 20 At this site, the loess is 2.5 m thick and consists of laminated silt and well-sorted fine-to very finegrained sand that has been bioturbated by plants. 21 The site has patchy vegetation consisting of dwarf willow, sedges, and mosses. 22 F I G U R E 1 The Nordic Arctic and Subarctic, with locations of the CALM grids studied in northeast Greenland, central Svalbard, and northern Sweden. Basemap from Jakobsson et al. 63  contains Quaternary glacial and periglacial landforms including moraines, meltwater plains, a raised delta, and large alluvial fans. [25][26][27] From 1996 to 2015, MAAT at Zackenberg was −9 C and mean annual precipitation was 218 mm. 28 From 1997 to 2014, the highest monthly mean air temperature was 6.3 C in July and the lowest was −19.8 C in February. 29 Thus, Zackenberg has a lower MAAT than central Spitsbergen, with colder winters and slightly cooler summers.
The ZEROCALM1 grid is 100 × 100 m with 10 m node spacing, resulting in 121 measurement points. ZEROCALM2 also has 10-m node spacing, but is 120 × 150 m to accommodate a snow patch, and thus has 208 measurement points. Thaw progression in the ZEROCALM grids has been monitored from late May or early June to

| Abisko, northern Sweden
Abisko is located in northern Sweden in the Torneträsk region (68 N), where discontinuous and sporadic permafrost occurs. [32][33][34] Discontinuous permafrost is widespread in the mountains surrounding Abisko above 800-1000 m a.s.l., 35 while permafrost is sporadic at lower elevations, occurring in peat mires. 36 From 1985 to 2010, MAAT at Abisko was 0.02 C and mean annual precipitation was 330 mm. 37 Abisko has a continental, subarctic climate, meaning the coldest month has a mean temperature <0 C and the warmest month has a mean temperature >10 C.
ALTs have been reported from mechanical probing at multiple sites in the Abisko area, 33 and data from 11 sites are averaged into one annual ALT value for Abisko in the CALM database. Data from the individual Heliport (378 m a.s.l.) and Storflaket (383 m a.s.l.) sites are presented here, as these sites have standardized 100 × 100 m grids with 121 measurement points (10 m node spacing), and the longest and most continuous data series from the study region. The Storflaket site is an 1 km 2 peat plateau with Sphagnum peat 60-90 cm thick. 38 The Heliport site is a slightly smaller peat plateau.
The active layer was previously confined to the upper peat layer at both sites, 39 but recently includes part of the underlying fine-grained sediment. These grids are measured once annually around the time of maximum thaw depth, typically the third week in September. 33 Probing data exist for Heliport and Storflaket since 1978, but the 121-point grids were not established until 1994. From 1978 to 1986, probing was done in a transect, and from 1987 to 1993 measurements were made in a smaller 10 × 10 m grid. The two sites are located relatively close to the meteorological observatory of the Abisko Scientific Research Station.
2 | METHODS 2.1 | Active layer probing and Nordic CALM grid data ALT has been determined at all the sites using mechanical probing, according to the monitoring methods defined by the CALM program. 1 In this method, a steel rod is inserted perpendicular to the ground surface until complete resistance is met at the top of the frost table.
Thaw depth is defined as the depth the rod is inserted into the ground; measurements are recorded at the grid nodes. A grid thaw depth average was calculated from all grid point values for each day with thaw depth measurements. ALT is the maximum grid thaw depth average in any given year for the Zackenberg and Svalbard sites. For the sites at Abisko, ALT is assumed to be the thaw depth average in the grid during annual measurement in the second half of September, which is the period of maximum thaw in this area. 33 It should be acknowledged that the reliability of ALT data from probing can be impacted by operator strength, interference of buried clasts, and the potential difference between the frost table and the 0 C isotherm 40 ; however, these issues are not considered to be significant at the selected Nordic sites given site geomorphology and the authors' probing experience.

| CALM database and Arctic ALT trends
Annual ALT data from Arctic sites in the CALM database 15 were analyzed to compare the Nordic results to the entire Arctic. The CALM summary data span the years 1990-2018 and include both Arctic and lower-latitude sites in the Subarctic, Alps, and Tibetan plateau. In these data, ALT has been determined through mechanical probing, frost-tube measurements, or ground temperature interpolation from borehole measurements. There are three Svalbard sites in the CALM database that are not individually discussed in this paper (IT1, N1, and S1), as they did not fit the criteria of having active, continuous mechanical probing data series of at least 10 years. These sites are Ecogrid (IT1), a newer probing grid with only six years of data, Janssonhaugen (N1), a borehole in bedrock, and Kapp Linné (S1), an inactive probing site with noncontinuous data. Additionally, Abisko is listed as one site in the CALM database, but has been subdivided into two sites for the CALM summary analysis, given that the full Heliport and Storflaket grid data are available. Annual ALT values from all sites above the Arctic Circle (66 33 0 N) with at least 10 consecutive years of ALT data within the periods 1990-2018 (79 sites) and 2000-2018 (75 sites) were used to assess regional ALT variability and long-term trends. Shiklomanov et al. 41 also required 10 consecutive years of data for their determination of long-term, regional active layer trends. The Arctic sites that meet these criteria are located in nine different regions: the Alaska North Slope, Canada (which is not further subdivided in the CALM database), northwest Russia, western Siberia, central Siberia, northeast Siberia, central Svalbard, northeast Greenland, and northern Sweden. Reported regional averages are the mean of all annual ALT values of all sites within a region. The rate of change in ALT (cm/yr) was calculated for the Arctic sites using linear regression. Trends were considered significant for p-values ≤0.05.
CALM summary data processing and the computation of ALT trends and statistics were done in MATLAB version R2019b.42

| Meteorological data
Daily mean air temperatures were used for the determination of freezing and thawing degree days. Daily mean air temperatures for Svalbard Airport, located 10 km northwest of the UNISCALM site, were obtained from eKlima, the online meteorological database hosted by the Norwegian Meteorological Institute. 19 where Z is ALT, TDD is the thawing degree days, and E is the edaphic factor, which is a scaling parameter dependent on site conditions. 1,44 Thawing degree days were calculated by summing positive daily mean air temperatures during each thawing season. The start and end dates of thawing seasons were defined by the local minimum and maximum of the summation curve of daily air temperatures from each calendar year. 45    covers. 31,48 In 2018, both ZEROCALM sites had minimum or nearminimum ALT; this is explained by an unusually large amount of snow and extraordinarily late snow melt at Zackenberg that year. 48 Åkerman and Johansson 33 found that ALT responded to snow depth at five of nine sites in the Abisko region, but this correlation was not as strong as that between ALT and TDD. Given that snow cover is consistently thin at UNISCALM, 23 snow depth is not thought to be a relevant forcing factor of ALT at this site. Comparable, regular snow measurements are not available for all of the Nordic sites, and thus it was not possible to calculate and report regression statistics for ALT and snow depth.

| Relationship between ALT, TDD, and FDD in the Arctic
ALT is usually assumed to be controlled by conditions during the thawing season, and thus is often positively correlated with TDD on the basis of the Stefan equation (Equation 1). Positive correlation between ALT and ffiffiffiffiffiffiffiffiffiffi ffi TDD p has been observed in Alaska, 49,50 Canada,51 and at Abisko. 33

| Circum-Arctic trends in ALT
There is consensus that ALT has increased in the Northern Hemisphere since the 1990s, with the exception of some specific sites,

| CONCLUSIONS
ALT in the studied Nordic Arctic CALM probing sites is controlled by a combination of geomorphology and climate. Although it is the most southern area in this study, the active layer in northern Sweden is the thinnest because the active layer is mostly confined to superficial peat. The active layer in central Svalbard is the thickest, compared to both the Nordic-and the circum-Arctic, due to the area's maritime Arctic climate, in contrast to the continental climates in Greenland and northern Scandinavia. Interannual variability in ALT is not consistent across the Nordic Arctic region; this is attributed to the substantial climate differences between the study areas and the large climatic gradient between central Svalbard and northeast Greenland, in addition to site-specific geomorphological controls. Differences in the magnitude of ALT between grids in the same area also arise from site-specific geomorphological controls, specifically snow dynamics and peat characteristics (at peatland sites).
Climate warming has caused active layer thickening at all three Nordic Arctic study areas. Seasonality in climate change has been identified as explaining differences in the relationship between ALT and degree days between the three study areas. Specifically, increased summer air temperatures are responsible for increased ALT in the more continental climates of northern Sweden and northeast Greenland, while significantly increasing and variable winter air temperatures impact ALT in maritime central Svalbard. The ability of winter air temperatures to impact ALT is dependent on thin winter snow cover and the resulting coupling between winter air and active layer temperatures. We have found that ALT in central Svalbard correlates with FDD, but not with TDD. Here, there is relatively thin snow cover in the large valleys due to significant wind-driven snow redistribution.
These findings add to our understanding of active layer sensitivity to climate, and show that ALT is not always correlated with TDD.
Additionally, the impact of climate on ALT is superimposed on critical site-specific controls, such as the presence of peat and the timing and thickness of snow cover. Making accurate regional assessments of ALT trends is thus complicated by these unique site aspects, but also by varying lengths and inconsistencies in time series. However, there is consensus that Arctic ALT has been increasing, and we have found that the average rate of active layer thickening in the Nordic region is of the same magnitude as the average rate across the Arctic.
cal data were accessed through eKlima, a product of the Norwegian Meteorological Institute. Abisko meteorological data were provided by the Swedish Polar Research Secretariat. The authors would also like to thank the two anonymous reviewers who took the time to critically read the manuscript and whose comments led to significant improvements.

Data Availability Statement
The data that support the findings of this study are openly available in online repositories.