Effects of interannual variability in snow accumulation on energy partitioning and surface energy exchange in a high-Arctic tundra ecosystem

Snow cover is one of the key factors controlling Arctic ecosystem functioning and productivity. In this study we assess the impact of strong interannual variability in snow accumulation during two subsequent years (2013-2014) on the 10 land-atmosphere interactions and surface energy exchange in two high-Arctic tundra ecosystems (wet fen and dry heath) in Zackenberg, Northeast Greenland. We observed that record-low snow cover during the winter 2012/13 resulted in strong response of the heath ecosystem towards low evaporative capacity and substantial surface heat loss by sensible heat fluxes (H) during the subsequent snow melt period and growing season. Above-average snow accumulation during the winter 2013/14 promoted summertime ground heat fluxes (G) and latent heat fluxes (LE) at the cost of H. At the fen ecosystem a 15 more muted response of LE, H and G was observed in response to the interannual variability in snow accumulation. Overall, the differences in flux partitioning and in the length of the snow melt periods and growing seasons during the two years had a strong impact on the total accumulation of the surface energy balance components. We suggest that in a changing climate with higher temperature and more precipitation the surface energy balance of this high-Arctic tundra ecosystem may experience a further increase in the interannual variability of energy accumulation, partitioning and redistribution. 20


Introduction
The presence or absence of snow exerts a strong impact on the land-atmosphere interactions and on the exchange of energy and mass.The high albedo of snow (Warren, 1982) reduces the amount of absorbed shortwave radiation at the surface which generally leads to a smaller magnitude of the surface energy balance components.The influence of the snow on the energy balance is most pronounced during spring when the commonly patchy distribution of snow causes strong spatial variations in 25 surface temperature and surface energy balance components (Chernov, 1988).The meltwater in Arctic soils contributes a considerable proportion of plant available water during summertime and as such, end-of-winter snow depth constitutes an important control of the summertime energy partitioning into sensible and latent heat fluxes (Langer et al., 2011).
The Cryosphere Discuss., doi :10.5194/tc-2016-5, 2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.The Cryosphere Discuss., doi :10.5194/tc-2016-5, 2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.snow melt periods and growing seasons in 2013 and 2014, and (2) to quantify and evaluate the driving factors of surface energy partitioning during the observation period at our two high-Arctic sites.

Study sites
The study sites are located in the valley Zackenbergdalen in Northeast Greenland near the Zackenberg Research Station 5 (74°30'N, 20°30'W) (Fig. 1a).The valley is surrounded by mountains to the west, north and east while the Young Sound and Tyrolerfjord form the valley boundary to the south.Vegetation is sparse and mainly found in the valley bottom and on the lower parts of the slopes.Cassiope heaths, Salix arctica snow-beds, grasslands and fens with sedges and grasses dominate in the lowlands while open Dryas sp.heaths and grasslands form the main plant communities on the slopes (Bay, 1998).10 We conducted measurements of surface energy balance components and meteorological variables in a wet fen and in a tundra heath, with a distance of approx.600 meters between the two measurement towers.The fen area can be divided into a continuous fen, with flat areas dominated by Eriophorum scheuchzeri, Carex stans and Duponita psilosantha, and a hummocky fen dominated by E. triste, S. arctica and Andromeda latifolia (Bay, 1998;Elberling et al., 2008).The tundra heath site is characterized by Cassiope tetragona, D. integrifolia, Vaccinium uliginosum and patches of mosses, E. 15 scheuchzeri and S. arctica (Lund et al., 2012).
Since August 1995 meteorological and environmental monitoring activities have been conducted by the Zackenberg Ecological Research Operations (ZERO), a part of the Greenland Ecosystem Monitoring (GEM) programme.Mean annual air temperature in Zackenberg (1996Zackenberg ( -2013) ) is -9.0°C, with an average span from -19.3°C in January to +6.3°C in July (Mylius et al., 2014).Annual precipitation is low (211 mm) (Mylius et al., 2014) and approx.85% consists of snow 20 precipitation (Hansen et al., 2008).Snow cover is unevenly distributed in the valley with large deposits on south-facing slopes as winds from the north (offshore) dominate during the winter (Soegaard et al., 2001).During the growing season winds from south-east (onshore) are dominating (Elberling et al., 2008).The area is located within the zone of continuous permafrost and active layer thicknesses at the end of the summer reach between 0.4 and 0.8 m within the valley (Christiansen, 2004;Pedersen et al., 2012).25 During the period 2000-2010 mean July air temperatures in the Zackenberg area increased by 0.18°C yr -1 and active layer thickened by 1.5 cm yr -1 (Lund et al., 2014).Based on the IPCC SRES A1B scenario (Nakićenović et al., 2000), local climate modelling for the region predict an increase in mean annual air temperature by 4.1°C for the period 2051-2080 compared to  with highest increase during winter (6.6°C) and spring (7.4°C) while precipitation over eastern Greenland is projected to increase by 60% (Stendel et al., 2007).30 The Cryosphere Discuss., doi :10.5194/tc-2016-5, 2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.

Measurements
Fluxes of sensible (H) and latent heat (LE) at the wet fen and the dry heath were measured by two eddy covariance systems.
At the wet fen a 3 m tower was equipped with a LI-7200 (LI-COR Inc., USA) enclosed-path gas analyser and a Gill HS (Gill Instruments Ltd, UK) 3D wind anemometer.Air was drawn at a rate of 15 L min -1 through a 1 m long tube (9 mm inner diameter).Data from both sensors was sampled at a rate of 20 Hz and fluxes were calculated using EddyPro software (LI-5 COR Inc., USA).Snow depth measurements were used to dynamically estimate sonic height above the snow layer.Air temperature (T a ), relative humidity (RH) and air pressure measured by external sensors was used in flux calculations.The gas analyser was calibrated based on manual calibrations using air with known CO 2 concentration and based on estimated H 2 O concentration from T a /RH measurements.Post-processing and quality checks follow standard procedures (Aubinet et al., 2012).10 At the dry heath a 3 m tower was equipped with a Gill R3 (Gill Instruments Ltd, UK) sonic anemometer and a LI-7000 (LI-COR Inc., USA) gas analyser.Air was drawn through 6.2 m of tubing (inner diameter: 1/8'') at a rate of 5.5 L min -1 to the sensor.More information on the eddy covariance system and data processing is provided by Lund et al. (2012Lund et al. ( , 2014)).At both locations soil temperature (T107, Campbell Scientific, USA) at various depths, soil heat flux (HFP01 Hukseflux, The Netherlands) and net radiation (CNR4 Kipp & Zonen, The Netherlands) were measured.Ancillary meteorological 15 parameters such as snow depth (SR50-45, Campbell Scientific, USA), air temperature and humidity (HMP 45D, Vaisala, Finland), radiation components (CNR1, Kipp & Zonen, The Netherlands) and precipitation (5915 x, OTT Hydromet GmbH, Germany) are provided from a nearby meteorological station operated by Asiaq -Greenland Survey.This station is located at the same heath, approx.150 m away from the heath eddy covariance tower.

Data analysis and derived parameters 20
The surface energy balance of the wet fen and the dry heath is described by: where R net is the net radiation, H is the sensible heat flux, LE is the latent heat flux, G is the ground heat flux at the soil or snow surface, and E melt is the energy flux used for snow melt.During the snow-free season E melt is zero.The left side of the equation represents the system's gain of energy and the right-hand side represents the losses of energy.Bowen ratio (H/LE) 25 and ratios of H/R net , LE/R net and G/R net are used to characterize relative magnitude of the heat transfer from the surface.
The net radiation balance (R net ) was defined as: where RS↓ and RS↑ are incoming and outgoing shortwave radiation, and RL↓ and RL↑ are upwelling and downwelling longwave radiation, respectively.Surface albedo was calculated as the quotient between RS↓ and RS↑.Missing values from 30 radiative components at the dry heath site were filled with measurements from the nearby Asiaq meteorological tower.
Ground heat flux at the soil or snow surface (G) was calculated by adding the storage flux in the layer above the heat flux plate (S) to the measured flux: where ∆T s /∆t is the change in soil or snow pack temperature (K) over time t (s), d is the heat flux plate installation depth (m) and C s is the soil or snow pack heat capacity (J m -3 K -1 ) defined as: where ρ b is the bulk density, C d is the dry soil heat capacity of 840 J kg -1 K -1 (Hanks and Ashcroft, 1980) or the heat capacity of ice (2102 J kg -1 K -1 ), θ v is the volumetric soil or snow pack water content (m 3 m -3 ), ρ w is the water density (1000 kg m -3 ) and C w is the water heat capacity (4186 J kg -1 K -1 ).For ρ b a value of 900 kg m -3 at the heath and 600 kg m -3 at the fen (Elberling et al., 2008) was used during the growing season while during the snow melt period in 2014 the density of the 10 snow pack was derived from in-situ measurements.Since no snow density measurements were performed during the snow melt period in 2013 and the soil surface was not completely snow-covered during that period, G in 2013 was not corrected for heat storage within the snow pack.
The aerodynamic resistance (r a , s m -1 ) determines the turbulent heat transfer from the surface and was defined as (Monteith and Unsworth, 2013) where u is the wind speed (m s -1 ) and u * is the friction velocity (m s -1 ).
Surface resistance (r s , s m -1 ), as a measure to quantify the stomatal control in the canopy on the turbulent fluxes, was calculated as (Shuttleworth, 2007) where ∆ is the slope of the saturated vapour pressure curve (Pa K -1 ), γ is the psychrometric constant (Pa K -1 ), β is the Bowen ratio, ρ is the air density (kg m-3) c p is the specific heat capacity of air at constant pressure (J kg -1 K -1 ), D is the atmospheric vapour pressure deficit (Pa) and A is the available energy for evaporation (R net -G, W m -2 ).
The decoupling coefficient (Ω) (Jarvis and McNaughton, 1986) expresses the degree of interaction between r a and r s : The decoupling coefficient varies from 0 to 1 where Ω close to zero indicates a strong coupling between the vegetation and the atmosphere, with vapour pressure deficit (VPD) being the main driver of LE, whereas Ω close to 1 suggest a decoupling of LE and VPD, with R net being the main driver for LE.
Priestley-Taylor coefficient (α) (Priestley and Taylor, 1972) was calculated as: Over ocean and saturated land surfaces the dimensionless α equals to 1.26 but fluctuates depending on surface structure and meteorological conditions.
Polar night was defined as the time period when the sun was below the free horizon.In Zackenberg, the polar night lasts from 10 November to 4 February (86 days).In our study, the subsequent pre-melt season marks the time period between the polar night and the first day of snow melt.In spring, a steady and constant decrease in daily average albedo and snow cover thickness until snow cover diminished was defined as the snow melt period.The beginning of the subsequent growing season was defined as the time when albedo was lower than 0.2.Positive daily average air temperature and top soil layer 5 temperature (2 cm depth), net radiation >0 W m -2 and albedo lower than 0.2 define the growing season.Gap-filling of the eddy covariance data was performed based on a look-up table approach (Falge et al., 2002;Reichstein et al., 2005) using the REddyProcWeb online tool (Max Planck Institute for Biogeochemistry).Due to very limited data availability at both locations during the second snow melt period in 2014 no gap-filling of H and LE was applied and turbulent fluxes of H and LE were excluded from the analysis during that period.10

Polar night and pre-melt season
The pattern of snow accumulation during the polar night and pre-melt season differed strongly between the two winters (Fig. 1b).There was no distinct development of a closed snow cover or major events of snow accumulation during the first winter (2012/13) and by the end of the pre-melt season as little as 0.09 m of snow pack at the wet fen and 0.14 m at the dry heath 15 was present.The second winter (2013/14) showed a similar trend in the beginning of the winter with snow pack thickness of less than 0.1 m until mid-December 2013.After that, a snow pack of 0.8 m developed during two major events of snow fall (17-25 December 2013 and21-27 January 2014) and by the end of the second winter the snow pack reached a thickness of 1.04 m at the wet fen and 0.98 m at the dry heath (Fig. 2).
In the absence of solar radiation during the polar night thermal radiation (RL net ) is the sole driver of the surface energy 20 balance (Fig. 2).Daily average downwelling longwave radiation (RL↓) was 193 W m -2 during the first polar winter (2012/13) and 213 W m -2 during the second polar winter (2013/14).The corresponding values for the upwelling longwave radiation (RL↑) were 219 W m -2 and 237 W m -2 for the two polar winter periods.The differences in longwave radiation corresponded well with the differences in air and snow surface temperatures.The average air temperature, which mainly controls the downwelling longwave radiation, was -19.2°C and -15.2°C for the two consecutive polar winter periods and the 25 snow surface temperature together with cloudiness, which mainly controls emissivity, was -24.1°C and -19.2°C, respectively.Overall, average RL net was -27 W m -2 in 2012/13 and -24 W m -2 in 2013/14.
The two subsequent pre-melt seasons were characterized by a continuous increase in incoming solar radiation (RS↓) but with little impact on the air and surface temperatures (Fig. 2) due to high albedo.Actually, the lowest air temperatures (-33.7°C and -33.7°C) and snow surface temperatures (-39.8°C and -39.9°C) for both polar winters were reached during this period.

Snow melt season
At the beginning of the snow melt period in 2013 the vegetation on both sites was not completely snow-covered.Snow melt started on 13 May and lasted until 29 May (17 days) at the wet fen and until 30 May (18 days) at the dry heath.Snow ablation in 2014, however, started 7 days earlier compared to 2013 (6 May) but lasted 28 days longer at the dry heath site ( 27June) and 25 days longer at the wet fen site (23 June).Daily average air temperatures remained below 0°C for most of the 5 snow melt periods with average values of -1.8°C in 2013 and -1.3°C in 2014 (Fig. 3).High rates of snow ablation after 6 June 2014 coincided with daily average air temperatures >0°C which peaked on 16 June 2014 with 9.6°C.
During the snow melt period in 2013, albedo decreased gradually from about 0.76 in the beginning of the period down to 0.15 for the wet fen and 0.10 for the dry heath (Fig. 3a).The decrease in albedo happens simultaneously with a continuous increase in net radiation (R net ) (Fig. 3a).However, while daily average R net remained negative for most of the first half of the 10 snow melt period in 2014 (Fig. 3b), positive values dominated the entire snow melt period in 2013.Substantial differences in albedo cause these variations, especially in the beginning of the snow melt seasons when daily average albedo was >0.60.
During that period, average albedo at both sites was 0.71 in 2013 and 0.79 in 2014.These differences in the surface characteristics reduced average R net in 2014 by 4.1 W m -2 although incoming solar radiation (RS↓) was 87.0 W m -2 higher compared to 2013.By the end of the snow melt period in 2013, R net accumulated 76.1 MJ m -2 at the dry heath and 94.3 MJ 15 m -2 at the wet fen.The corresponding values for 2014 were 174.5 MJ m -2 and 202.9 MJ m -2 , respectively.Due to the lower albedo during the first snow melt season the soil experienced a faster spring warming compared to 2014.Top-soil layer temperatures (T surf ) at the dry heath increased by 0.41°C d -1 in 2013 and 0.28°C d -1 in 2014.At the wet fen, the warming of the soil was 0.35°C d -1 in 2013 and 0.28°C d -1 in 2014.
Daily average sensible heat flux (H) was small and negative (-4.1 W m -2 ) during the first ten days of the snow melt period in 20 2013 at the wet fen while it showed slightly positive values (4.3 W m -2 ) at the dry heath (Fig. 3a).The latent heat flux (LE) increased gradually up to about 18 W m -2 at the wet fen and up to 23 W m -2 at the dry heath (Fig. 3a).During this initial period the net radiation increased gradually up to about 58 W m -2 at the wet fen and 43 W m -2 at the dry heath.Over the entire snow melt period, H dominated over LE at the dry heath (Bowen ratio 1.7) with mean daily average H and LE of 25.0 W m -2 and 14.9 W m -2 , respectively, while at the wet fen LE dominated over H (Bowen ratio 0.5), with average H and LE of 25 7.6 W m -2 and 15.6 W m -2 (Table 2).By the end of the snow melt period H accumulated 32.5 MJ m -2 at the dry heath and 3.5 MJ m -2 at the wet fen.The corresponding values for LE were 22.7 MJ m -2 and 21.1 MJ m -2 , respectively.This corresponds to accumulated evaporation of 9.3 mm and 8.6 mm of evaporated water during the entire spring snow melt at the dry heath and the wet fen.
During both years the ground heat flux (G) at the wet fen and the dry heath used ~10-15% of the total energy supplied by 30 R net .The prolonged snow melt period during the second year had no impact on the magnitude of G.However, the total  2 MJ m -2 and 28.6 MJ m -2 .The average latent heat content of snow cover during the snow melt period in 2014 was 3.1 MJ m -2 which equals to 1.8% of R net at the dry heath and 1.5% of R net at the wet fen.

Season length and meteorological conditions 5
The growing season in 2013 lasted 101 days at the wet fen  and 100 days at the dry heath (31 May -8 September).Compared to 2013, the season in 2014 was 30 days shorter on the wet fen (24 June -4 September) and 31 days shorter at the dry heath (28 June -4 September).Mean July-August air temperatures (T a ) were similar in both years, reaching 6.5°C in the first year and 6.3°C in the second year, mean Ta over the entire growing seasons were lower in 2013 (5.3°C) compared to 2014 (6.1°C) (Fig. 4).Total amount of precipitation during both years reflect the dry summertime 10 conditions in Zackenberg with 80 mm in 2013 and 65.5 mm in 2014.In the first year, pronounced events of rainfall (>5 mm d -1 ) occurred at the end of the growing season while in the second year precipitation mainly fell shortly after snow melt and in late-August (Fig. 4).

Radiation balance
During the growing season in 2013, mean daily R net was slightly higher at the wet fen (114.2W m -2 ) compared to the dry 15 heath (111.4W m -2 ) (Table 3) while during July-August both sites showed similar R net (~93 W m -2 ).Compared to 2013, mean daily R net was slightly higher during July-August 2014, with 94.3 W m -2 at the wet fen and 98.6 W m -2 at the dry heath, but lower over the entire growing season, with 100.2 W m -2 and 98.9 W m -2 , respectively (Table 3, Fig. 5).By the end of the season in 2013, accumulated R net was 1006.6 MJ m -2 at the wet fen and 967.4 MJ m -2 at the dry heath compared to 632.2 MJ m -2 and 589.5 MJ m -2 in 2014.This increase of accumulated R net by 59% (fen) and 64% (heath) in 2013 relates to the earlier 20 onset of the growing season.By 24 June 2013, accumulated R net at the wet fen reached 406 MJ m -2 during the first 25 days of the growing season while in 2014, the growing season started on 24 June and the surface was thus still snow covered until that date.Similar values were observed at the dry heath (423 MJ m -2 ).
In 2013, mean surface albedo at both sites increased gradually during the growing season and reached its maximum towards the end of the season.Over the entire growing season albedo was higher at the wet fen (0.20) compared to the dry heath 25 (0.16).During a snow fall event on 30 August 2013 albedo increased up to 0.32.Similar trends in albedo were observed in 2014 but mean surface albedo over the entire growing season was lower compared to 2013 with 0.17 at the wet fen and 0.13 at the dry heath.The interannual differences are more pronounced during the period July-August when mean albedo was 0.20 (2013) and 0.16 (2014) at the wet fen and 0.16 (2013) and 0.12 (2014) at the dry heath.

Turbulent heat fluxes
The magnitude of the turbulent heat fluxes and the total amount of accumulated energy by H and LE revealed remarkable differences between the two years (Fig. 5, Fig. 6).Over the entire growing season in 2013, average fluxes of H and LE used 56% (62.6 W m -2 ) and 14% (16.0 W m -2 ), respectively, of R net at the dry heath.The corresponding values for the wet fen were 37% (41.5 W m -2 ) and 30% (25.2 W m -2 ), respectively (Table 3).This clear dominance of H over LE at both locations 5 is also reflected in the average Bowen ratio (H/LE) which reached a maximum of 3.9 at the dry heath and 1.6 at the wet fen.
During 2014, LE at the dry heath used 26% (29.3 W m -2 ) of R net (Table 3).This corresponds to an almost doubling of LE compared to 2013.Mean fluxes of H at the dry heath were 36.6 W m -2 and H used 37% of R net .Average Bowen ratio reached 1.3 which indicates a growing importance of LE at the dry heath compared to 2013.At the wet fen, mean daily LE was 23.3 W m -2 and LE used 25% of R net .The corresponding value for H was 26.6 W m -2 and H used 29% of R net .Average Bowen 10 ratio reached 1.1.
From the first day after snow melt until the last day of the growing season in 2013 accumulated H was 546.2 MJ m -2 at the dry heath and 365.7 MJ m -2 at the wet fen while accumulated LE was 139.9 MJ m -2 and 222.0 MJ m -2 , respectively (Fig. 6).
By the end of the growing season in 2014, however, accumulated H was 220.2 MJ m -2 at the dry heath and 167.8 MJ m -2 at the wet fen.The corresponding values for LE were 174.7 MJ m -2 and 146.9 MJ m -2 , respectively (Fig. 6).The observed 15 values of accumulated H at the dry heath during 2014 correspond to a decrease in accumulated H by 60% and an increase in accumulated LE by 24% compared to 2013.At the wet fen, accumulated H and LE in 2014 correspond to a decrease by 44% and 33%, respectively, compared to 2013.
The total accumulated evapotranspiration (ET) differed significantly between the seasons (Fig. 6).At the wet fen, total ET reached 91 mm during the growing season in 2013 and 48 mm in 2014.These values correspond to ~114% and ~73% of the 20 total precipitation during the growing seasons within the corresponding years.However, total ET at the dry heath remained below the growing season precipitation during the first growing season (57 mm, ~71%) and exceeded precipitation in the second growing season (80 mm, ~122%).

Controls of evapotranspiration
During both growing seasons, lower wind velocity at the dry heath compared to the wet fen resulted in slightly larger 25 aerodynamic resistances (r a ) at the latter site.Daily average r a showed no clear difference between the two years ranging between 75-82 s m -1 in both years at the wet fen.The corresponding values at the dry heath were between 107-120 s m -1 with slightly larger variability compared to the wet fen (Fig. 7).No significant changes during the course of both growing seasons indicate that r a was mainly dependent on atmospheric conditions.
The surface resistance (r s ) during both growing seasons, however, was characterized by large differences between the wet 30 fen and the dry heath and high daily fluctuations at the latter site (Fig. 7).Daily average r s at the dry heath ranged from 93 to 923 s m -1 in 2013 and from 64 to 428 s m -1 in 2014, with higher mean r s in 2013 (556 s m -1 ) compared to 2014 (247 s m -1 ).
During the first growing season r s showed a general increase with increasing air temperature (T a ), ranging from ~455 s m -1 when T a <0°C to ~804 s m -1 when T a >12°C.During the second growing season r s decreased from ~297 s m -1 when T a <0°C to ~210 s m -1 when T a was between 3-6°C, followed by an slight increase to 250 s m -1 when T a >12°C.However, no such trend was observed at the wet fen and the surface resistance had no pronounced daily fluctuations with daily averages ranging between 53 and 455 s m -1 .5 The McNaughton and Jarvis decoupling factor (Ω) expresses the degree of aerodynamic and radiative coupling between the vegetation and the atmosphere.The wet fen and the dry heath were characterized by mean daily Ω below or close to 0.5 during both growing seasons which indicates a relatively high contribution of vapour pressure deficit (VPD) on the control of ET (Table 3, Fig. 7).However, mean daily Ω during the first growing season was lower at the dry heath (0.30) compared to the wet fen (0.38) but showed a reverse trend during the second year with 0.52 compared to 0.48.The general decrease in Ω 10 over the course of the first season at the dry heath terminated around mid-August and Ω started to increase for the rest of the season.Similar behaviour for Ω was observed at the wet fen where Ω started to increase around late-August.The second growing season showed a general decrease in Ω at both sites.During both growing seasons Ω gradually decreased as r s increased.The Priestley-Taylor coefficient (α) showed large day-to-day variations, particularly at the wet fen, ranging between 0.05 and 1.06 in 2013 and 0.21 and 1.13 in 2014.The seasonal mean of α was 0.60 and 0.69 with a standard 15 deviation of 0.21 and 0.22 for the respective seasons.The day-to-day variation was less pronounced at the dry heath with seasonal means of 0.44 and 0.74 and standard variations of 0.18 and 0.17.(Table 3, Fig. 7).

Ground heat flux and soil properties
The ground heat flux (G) was the smallest flux in both ecosystems but due to the strong heat sink of permafrost G is a vital component of the surface energy budget in this high-latitude environment.The differences in snow cover between the two 20 years caused distinct differences in the soil water content during the two growing seasons, with strong impact on the partitioning of R net into G at the dry heath.Average soil moisture content at the dry heath of 19% in 2013 and 34% in 2014 restrained thermal conductivity more during 2013 than during 2014.Consequently, only 5% of R net (5.6 W m -2 ) was partitioned into G in 2013 while during 2014 G used 10% of R net (10.1 W m -2 ).At the wet fen, mean daily G used 12% of R net (10.8 W m -2 ) in 2013 while in 2014, G used 17% of R net (15.2 W m -2 ). 25 During the course of the two growing seasons G reached its maximum shortly after snow melt due to the strong thermal gradient between the soil surface and the underlying permafrost (Fig. 5).However, the low soil moisture content at the dry heath during the first growing season weakened the G signal after the snow melt.No clear seasonal trend in the development of G to R net was observed for both growing seasons and locations.By the end of the first growing season accumulated G was 48.7 MJ m -2 at the dry heath and 94.7 MJ m -2 at the wet fen (Fig. 6).Increased thermal conductivity due to higher soil 30

Total surface energy balance of the snow melt season and growing season
From the first day of the beginning of the snow melt period until the last day of the growing season in 2013 R net accumulated 1043.5 MJ m -2 at the dry heath and 1100.9MJ m -2 at the wet fen.The corresponding values for 2014 were 764.0 MJ m -2 and 5 835.1 MJ m -2 , respectively.Accumulated ground heat flux for the same time period was 56.7 MJ m -2 at the dry heath and 103.8 MJ m -2 at the wet fen in 2013 and 83.2 MJ m -2 and 125.3 MJ m -2 in 2014, respectively.These values correspond to a total amount of accumulated available energy at the surface (R net -G) of 986.8 MJ m -2 at the dry heath and 997.1 MJ m -2 at the wet fen in 2013.In 2014, R net -G was 683.4 MJ m -2 at the dry heath and 711.8 MJ m -2 at the wet fen which, compared to 2013, relates to a decrease in available energy at the surface by -31% and -29%, respectively.10 In 2013, the amount of total H loss was 578.7 MJ m -2 at the dry heath and 369.2 MJ m -2 at the wet fen while total LE loss was 162.6 MJ m -2 and 243.1 MJ m -2 , respectively.These values add up to a total energy consumption of H and LE (H+LE) of 741.3 MJ m -2 at the dry heath and 612.3 MJ m -2 at the wet fen.

Energy balance closure
The observed slopes of the regressions between available energy at the surface (R net -G) and the sum of the turbulent heat 15 fluxes (H+LE) serve as an indicator for the energy balance closure.Figure 8 shows the comparison between mean daily R net -G and mean daily H+LE for the two study sites.The observed slopes at the wet fen are 0.68 during both 2013 and 2014.

Snow cover and surface energy budget 20
The disappearance of the snow cover and the onset of the growing season coincide with the time period when incoming solar radiation (RS↓) is at its annual peak and the surface (soil or snow) receives high irradiance from RS↓.During the snow melt period most of R net is used for warming and melting of the snow pack and therefore not available for atmospheric warming through sensible heat fluxes.This pattern was reflected in the average air temperature of the two snow melt periods which showed only small differences between 2013 and 2014 although the snow melt period in 2014 was much longer compared to 25 2013.The combination of snow layer thickness, snow physical properties such as density or grain size (Warren, 1982), and the fraction of exposed dark underlying surface control the snow and soil surface albedo, with direct impact on R net .Our results showed that low and partly absent snow cover in the winter 2012/13 resulted in a short snow melt period dominated by low albedo of the snow-soil surface and increased average R net .Simultaneously, sensible heat fluxes dominated at the dry heath while latent heat fluxes dominated at the wet fen.Relatively thick snow cover and a prolonged snow melt period in The Cryosphere Discuss., doi:10.5194/tc-2016-5,2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.
2014 increased surface albedo and limited average R net .Further, the total energy accumulated by R net during the snow melt period was higher in 2014 compared to 2013.
During the snow melt period, only a small fraction of R net can be used by plants for growth and development as the snow shields off most of the incoming solar radiation and sustains relatively low surface temperatures (Walker et al., 2001).
Consequently, a prolonged period of snow cover and snow melt at our study sites in 2014 delayed the onset of the growing 5 season and limited the amount of total R net available for plant metabolism.In coherence with that, the magnitude of the surface energy balance components and the partitioning of R net into H, LE and G continued to be influenced by the presence of the snow.An earlier disappearance of the snow in 2013, however, was related to an earlier increase in the magnitude of R net at the surface which facilitated earlier plant development and growth, soil warming and permafrost active layer development through G, evapotranspirative heat loss through LE, and atmospheric warming through H. 10The disappearance of the snow and related increase in surface heating mark the transition into a convective summer-type precipitation regime (Callaghan et al., 2011b) which initiates a rapid and distinct change in both plant metabolism and surface energy balance.Ongoing and predicted climate change, however, promotes an increase in atmospheric moisture, winter snowfall over land areas (Rawlins et al., 2010), and interannual variability in the amount of snowfall (Callaghan et al., 2005;Kattsov et al., 2005;Stocker et al., 2013) while higher temperatures stimulate an earlier snow melt and transition to 15 convective precipitation patterns (Groisman et al., 1994).We observed that the impact of this interannual variability in snow cover and snow melt on the seasonal surface energy budget is strongly connected to the storage of meltwater in the soil and its evaporation and transpiration over the subsequent growing season.A higher proportion of soil moisture, combined with a high atmospheric moisture demand, generally stimulates ET.In our study, the impact of the availability of soil moisture from snow melt and its loss through ET on the surface energy budget was most pronounced at the heath.Here we observed that 20 low soil moisture content in 2013 amplified H at the cost of LE and G while increased soil moisture in 2014 favoured LE and G at the cost of H. Further, the Bowen ratio of the heath during the same year showed that energy partitioning into H and LE was similar to the wet fen.In contrast, the wet fen showed attenuated behaviour of LE, ET, H and G to the variability in snow meltwater as the fen receives its moisture supply mostly from groundwater close to the surface.
Negative water balances during the growing season, with ET exceeding precipitation, are common in high-latitude 25 ecosystems (Woo et al., 1992).At the wet fen, our observations showed negative water balances during the growing season 2013 but the loss of water through ET was compensated by moisture supply from groundwater.Consequently, the partitioning of R net into LE showed only small differences between the two growing seasons.At the dry heath, the growing season in 2013 ended with a positive water balance.However, this was related to both pronounced precipitation at the end of the season and to low rates of ET due to declining R net .During most of the season snow meltwater was the only supplier of 30 soil moisture and the small amount of snow meltwater was evaporated relatively soon after snow melt.Consequently, for most parts of the remaining season the soil was not able to supply moisture for ET.This resulted in low LE, relatively low soil thermal conductivity and G, and in a clear dominance of H.During the growing season in 2014, the water balance of the The Cryosphere Discuss., doi :10.5194/tc-2016-5, 2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.
heath was negative over the entire season but the soil experienced greater saturation from snow meltwater storage which was reflected in the partitioning of the surface energy balance towards a greater share of LE and G on R net .
Interannual variability of snow cover and length of the growing season is not only reflected in the partitioning of the surface energy balance components.More important, total accumulated R net increases with increased length of the snow-free period.
Further, any increase in the length of the snow-free season or in summer temperatures is manifested in a general increase in 5ET, resulting in negative water balances over the snow-free season if precipitation shows no increase (Eugster et al., 2000).
Our results showed that with the onset of the snow melt period and growing season during the part of the year when incoming solar radiation is at its peak, an earlier onset of the growing season of approx.four weeks resulted in dramatic differences in accumulated energy fluxes at the end of the growing season.Summarizing, the increased amount of accumulated R net contributed to increased ET and surficial heat loss from LE, soil and permafrost warming through G, and 10 atmospheric warming through H.

Energy balance closure and system performance
The observed values of energy balance closure lie in the range of other energy balance closure terms reported from various field studies in Arctic environments (Westerman et al., 2009;Langer et al., 2011;Liljedahl et al., 2011) and are in good agreement with earlier long-term observations from the dry heath (Lund et al., 2014).Studies on the closure problem of eddy 15 covariance measurements highlight the landscape heterogeneity as a driving factor for the lack of closure (Stoy et al., 2013) but also stress the importance of measurement scales (Foken, 2008).Due to the striking small-scale heterogeneity of soil moisture availability and vegetation cover in Arctic landscapes, measurements of point-scale G and radiative components may not represent the large source area of eddy covariance H and LE.However, at our sites a flux footprint analysis at the dry heath revealed that fluxes of H and LE on average originate from the Cassiope heath (Lund et al., 2012) while the 20 highest contribution of fluxes at the wet fen originates from the continuous fen area (Tagesson et al., 2012).Large variability in soil heat capacity due to spatial variation in soil moisture may account for underrepresentation of the stored energy between the heat flux plates and the soil surface within the eddy covariance flux footprint area (Leuning et al., 2012).

Conclusions
In this study we documented the effects of interannual variability in snow accumulation on the surface energy balance of a 25 high-Arctic tundra ecosystem in Northeast Greenland during two subsequent years (2013)(2014)            Tables Table 1: Summary of polar night and pre-melt season.
List of mean daily radiation components and environmental characteristics during the polar night seasons and pre-melt seasons in 2012/13 and 2013/14.The Cryosphere Discuss., doi :10.5194/tc-2016-5, 2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.
amount of energy supplied to the ground was larger in 2014 than in 2013.At the dry heath, average G of 5.1 W m -2 in 2013 and 5.1 W m -2 in 2014 resulted in a total energy consumption of 8.0 MJ m -2 in the first year and 20.5 MJ m -2 in the second The Cryosphere Discuss., doi:10.5194/tc-2016-5,2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.year.Similar behaviour was observed at the wet fen where average G of 8.4 W m -2 in 2013 and 7.2 W m -2 in 2014 added up to a total energy consumption of 10.
moisture content in 2014 compared to 2013 amplified the total amount of accumulated G at the dry heath in the latter year although prolonged snow cover reduced the length of the second growing season.By the end of the second growing season The Cryosphere Discuss., doi:10.5194/tc-2016-5,2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.accumulated G was 60.1 MJ m -2 at the dry heath (+23% compared to 2013) while at the wet fen, no clear difference in accumulated G was observed (95.6 MJ m -2 ).

Figure 2 :
Figure 2: Polar night and pre-melt seasons of the winters 2012/13 and 2013/14.Characteristics of snow cover development, air temperature, net longwave radiation (RL net ) and net shortwave radiation (SW net ) during the polar night periods and subsequent pre-melt seasons of the winters 2012/13 and 2013/14.5

Figure 3a : 5 Dry
Figure 3a: Snow melt period 2013.Development of mean daily snow cover thickness, albedo, net radiation (R net ), sensible heat fluxes (H), latent heat fluxes (LE), ground heat fluxes (G), air temperature (T a ), and soil temperature at 2 cm depth (T s ) at the wet fen (left) and dry heath (right) during the snow melt period in 2013 and first 5 days of the growing season in 2013.5

Figure 3b : 5 Wet
Figure 3b: Snow melt period 2014.Development of mean daily snow cover thickness, albedo, net radiation (R net ), sensible heat fluxes (H), latent heat fluxes (LE), ground heat fluxes (G), air temperature (T a ), and soil temperature at 2 cm depth (T s ) at the wet fen (left) and dry heath (right) during the snow melt period in 2014 and first 5 days of the growing season in 2014.5

Figure 4 :
Figure 4: Growing seasons 2013 and 2014.Development of mean daily air temperature, precipitation, incoming solar radiation (RS↓) and vapour pressure deficit (VPD) during the growing seasons in 2013 and 2014.

Figure 5 :
Figure 5: Growing season energy fluxes.Mean daily net radiation (R net ), sensible heat flux (H), latent heat flux (LE) and ground heat flux (G) at the wet fen and dry heath during the growing seasons in 2013 (left) and 2014 (right).

Figure 6a : 5 Figure 6b :
Figure 6a: Development of internal energy.Accumulated ground heat (G), latent heat (LE) and sensible heat (H) at the wet fen (left) and the dry heath (right) during the growing seasons in 2013 and 2014.

Figure 7 :
Figure 7: Omega, alpha, aerodynamic resistance and surface resistance.Mean daily (empty circles) and 5-day running mean (solid line) of McNaughton & Jarvis decoupling factor (Omega), Priestley and Taylor Alpha-value (Alpha), aerodynamic resistance (r a ) and surface resistance (r s ) at the wet fen (left) and dry heath (right) during the growing seasons in 2013 and 2014.5 ., doi:10.5194/tc-2016-5,2016   Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.

Figure 8 :
Figure 8: Energy balance closure.Mean daily available energy (R net -G) and turbulent heat fluxes (H+LE) at the wet fen (upper) and the dry heath (lower) during the observation periods in 2013 and 2014.

Table 2 :
Summary of snow melt period.List of mean daily surface energy balance and environmental characteristics during the snow melt period in 2013 and 2014 at the wet fen and dry heath site.The Cryosphere Discuss., doi:10.5194/tc-2016-5,2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.

Table 3 :
Summary of growing season.List of mean daily surface energy balance and environmental characteristics during the growing seasons in 2013 and 2014 at the wet fen and dry heath site.The Cryosphere Discuss., doi:10.5194/tc-2016-5,2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.Measured at 10 cm depth, 2 Wind speed, 3 Friction velocity The Cryosphere Discuss., doi:10.5194/tc-2016-5,2016 Manuscript under review for journal The Cryosphere Published: 25 February 2016 c Author(s) 2016.CC-BY 3.0 License.