Evapotranspiration and energy partitioning across a forest-shrub vegetation gradient in a subarctic, alpine catchment

As a result of altitude and latitude amplified climate change, widespread changes in vegetation composition, density and distribution have been observed across northern regions. Despite wide documentation of shrub proliferation and treeline advance, few field-based studies have evaluated the hydrological implications of these changes. Quantification of total evapotranspiration (ET) across a range of vegetation gradients is essential for predicting water yield, yet challenging in cold alpine catchments due to heterogeneous land cover, including both boreal forest and shrub taiga ecosystems. Here, we present six years of surface energy balance components and ET dynamics at three sites along an elevational gradient in a subarctic, alpine catchment near Whitehorse, Yukon Territory, Canada. These sites span a gradient of thermal and vegetation regimes, providing a space-for-time comparison for future ecosystem shifts: 1) a low-elevation boreal white spruce forest (~12-20 m), 2) a mid-elevation subalpine taiga comprised of tall, dense willow (Salix) and birch (Betula) shrubs (~1-3 m) and 3) a high-elevation subalpine taiga with short, sparse shrub cover (< 0.75 m) and moss, lichen, and bare rock. Eddy covariance instrumentation ran year-round at the forest and during the growing season at the two shrub sites. Total ET decreased and interannual variability increased with elevation, with mean May to September ET totals of 349 (±3) mm at the forest, 249 (±10) mm at the tall, dense shrub site, and 240 (± 26) mm at the short, sparse shrub site. Comparatively, over the same period, ET:R ratios were the highest and most variable at the forest (2.19 ± 0.37) and similar at the tall, dense shrub (1.22 ± 0.09) and short, sparse shrub (1.14 ± 0.05) sites. Our results suggest that advances in treeline will increase overall ET and lower interannual variability; however, the large growing season water deficit at the forest indicates strong reliance on soil moisture from late fall and snowmelt recharge. In contrast, ET was considerably less at the cooler higher elevation shrub sites, which exhibited similar ET losses over 6 years despite differences in shrub height and abundance. ET rates between the two shrub sites were similar throughout the year, except during the peak growing season. Greater interannual variability in ET at the short, sparse shrub site indicates the reduced influence of vegetation controls on ET. Results suggest that predicted changes in vegetation type and structure in northern regions will have a considerable impact on water partitioning and will vary in a complex way in response to changing precipitation timing, phase and magnitude, growing season length, and vegetation snow and rain interactions.


60
Changes in vegetation type and structure result in complex soil-plant-atmosphere feedbacks that alter how catchments collect, store and release water, which directly impacts the global climate cycle through the albedo feedback and energy partitioning (Bonan, 2008;Chapin et al., 2005;Lafleur and Humphreys, 2018;Sturm et al., 2005). For example, snow and rainfall accumulation and interception are strongly dependent upon vegetation type, height and density (Pomeroy et al., 2006;Sturm et al., 2001;Van Dijk et al., 2015;Zwieback et al., 2019). Leaf area controls rain and snow interception in 65 northern boreal forests (Hedstrom and Pomeroy, 1998), whereas increases in shrub height enhances snow trapping (Pomeroy et al., 2006). Shrubs can also play a substantial role in rainfall interception, although this process is dependent on plant height and density. For example, Zwieback et al. (2019) found a 15-30% reduction in effective rainfall under birch shrubs, but reduced interception under alder. Vegetation influences the surface energy balance, affecting the rate and timing of snowmelt and evapotranspiration (ET) (Brummer et al., 2012;Kasurinen et al., 2014;Lafleur and Humphreys, 2018). Tall 70 shrubs accelerate melt through the emission of long-wave radiation (Marsh et al., 2010;Pomeroy et al., 2006) despite This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 4 reducing turbulent fluxes (Endrizzi et al., 2010). Northern forests have distinct melt regimes from the dominance of shading and sheltering processes and often retain snow for longer periods, despite having an overall lower above-canopy albedo (Bonan, 2008). During the snow-free period, vegetation phenology controls energy partitioning and the ratio of evaporation to transpiration (Wang et al., 2014). Other widely identified impacts of northern vegetation change are to ground thermal and 75 moisture regimes (Domine et al., 2016;Myers-Smith and Hik, 2013;Paradis et al., 2016;Sturm et al., 2005;Wilcox et al. 2019), carbon and nutrient cycling (Lafleur and Humphreys, 2018) and terrain stability (Lara et al., 2016). These critical zone feedback processes are complicated by dynamic and changing vegetation covers, both through "greening" of northern ecosystems from climate change and increasing risk of fire (Lantz et al., 2010). While the literature on surface-atmosphere interactions has grown in recent decades, current understanding of these hydrological interactions along a vegetation gradient 80 does not: (1) represent alpine environments and (2) has been primarily focused on surface energy partitioning (i.e. Beringer et al., 2005). Therefore, the net impact of alpine shrub proliferation and treeline advance on the critical zone water balance remains unclear.
As northern ecosystems change, there is increasing pressure on water managers, First Nations and other stakeholders to 85 understand and adapt to future water resources. As river regimes shift from declines in the cryosphere and changes in weather and climate (IPCC, 2019), the additional pressures of transitioning ecosystems must be assessed as ET strongly influences water partitioning in relatively dry, northern ecosystems. However, predicting ET over large scales is challenging, particularly in cold, alpine environments where observations are sparse and landscapes are heterogeneous (Bring et al., 2016). The objective of this paper is to advance our understanding of surface energy partitioning and ET 90 dynamics across an ecological functional gradient within a heterogeneous, subarctic, alpine watershed. This provides a unique opportunity to compare the potential impacts of shifting ecosystem types within a local geographic area. Here, we present six years of observations of surface-atmosphere interactions from three subarctic sites across a thermal gradient and varying vegetation cover spanning boreal forest to sub-alpine taiga. Specifically, among a (1) white spruce forest, (2) tall, dense shrub and (3) short, sparse shrub sites, this paper compares seasonal and inter-annual surface energy partitioning ET 95 This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 5 dynamics. We subsequently highlight the role of shifting vegetation on the water balance across this gradient and discuss the hydrological implications of ecosystem change in northern alpine catchments.

Study Sites
Wolf Creek Research Basin (WCRB) is located in Yukon Territory, Canada (~60.5˚N), 15 km south of the city of 100 Whitehorse and within the southern Yukon River headwater region (Fig. 1). WCRB is located in the traditional territories of the Kwanlin Dün, Ta'an Kwäch'än Council and Carcross/Tagish First Nations. The drainage area of the basin is approximately 179 km 2 with elevation ranging between 660 and 2080 meters above sea level (masl) (Rasouli et al., 2019a).
The climate of the basin is subarctic continental (Köppen Dfc), and the long term mean annual air temperature is 0.2°C as measured at the Whitehorse Airport (706 masl, Environment Canada, 1981, and declines with elevation. The long-105 term Whitehorse precipitation average is 282 mm (Environment Canada, 1981, yet this increases with elevation and can reach > 400 mm in WCRB in certain years (Carey et al., 2013;Rasouli et al., 2019a) The basin is composed of three distinct ecozones: boreal forest (22%) at lower elevations, subalpine taiga (58%) and alpine tundra (20%) (Rasouli et al., 2019a) (Fig. 1). The geology is sedimentary sandstone, siltstone, limestone and conglomerate with thick stony till and glacial drift above the bedrock for most of the basin. Sporadic and discontinuous permafrost occurs at high elevations and on north-110 facing slopes above treeline (Lewkowicz and Endie, 2004).
Three study sites were selected along an altitudinal gradient to represent the dominant vegetation and are termed: Forest, Buckbrush and Sparse Shrub. Forest is located near the WCRB outlet, at an elevation of 743 masl ( Fig. A; Table 1). The canopy is composed primarily of White Spruce (Picea glauca), approximately 12-20 m in height. Buckbrush is a subalpine taiga site located at mid-elevation (1250 masl). Vegetation consists of willow (Salix) and dwarf birch (Betula nana) shrubs, 115 with mean heights of 1.24 ±0.33 and 0.83 ±0.45 m respectively. Sparse Shrub is a high-elevation (1425 masl) subalpine taiga site with shorter willow (Salix) and birch (Betula nana) cover and increased moss, lichen, and bare rock cover (Fig. 1). Mean height of willow and birch shrubs at Sparse Shrub is 0.55 ±0.3 m and 0.41 ±0.16 m respectively. Mean growing season leaf This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 6 area index (LAI) at Forest, Buckbrush and Sparse Shrub is 3.5, 2.5, and 1.7 m 2 /m 2 (Table 1). Soil profiles at each site consisted of an organic layer of varying thickness atop primarily coarse sand with spatially variable compositions of gravel 120 and silt.

Instrumentation
Fluxes of latent heat (λvE), sensible heat (H) and CO2 were measured using the eddy covariance method at each site. Analysis presented here includes flux data year-round from 2017-2020 at Forest, and April to October from 2015-2020 at Buckbrush 125 and Sparse Shrub. Each tower was instrumented with a 3-axis sonic anemometer and infrared gas analyzer (IRGA). A closed path LICOR-7200RS was installed at Forest and an open-path LICOR-7500A at the two shrub sites (Buckbrush and Sparse Shrub). All signals were scanned at a frequency of 10 Hz and recorded on a Campbell Scientific datalogger. Thirty-minute fluxes were computed from high frequency signals with EddyPro v 6.2.2 using block averaging, planer fit coordinate rotation (Wilczak, Oncley, and Stage, 2001), and WindMaster Pro vertical wind velocity correction multipliers. Soil temperature and 130 moisture was measured at multiple depths at all sites from 2015 onwards, with soil water potential sensors (TEROS21) installed in June 2019 at all sites.

Data Processing, Gap-Filling and Energy Balance Closure
Open-path IRGA gas concentration density fluctuations were compensated using the Webb, Pearman, and Leuning correction 135 (Webb et al., 1980). Both H2O and CO2 densities were directly output by the closed-path IRGA. Time lags were detected and compensated through covariance maximization. Spike detection and removal were computed according to Vickers and Mahrt (1997). Analytic correction for high-pass filtering effects were computed following Moncrieff et al. (2004) for all sites.
Corrections for low-pass filtering effects were evaluated using methods by Moncrieff et al. (1997) for the open-path system and Horst (1997) for closed path. Fluxes were removed when turbulence was insufficient (friction velocity <0.1 m/s). Fluxes 140 of λvE and H were filled using artificial neural networks (ANNs) created with MATLAB (Mathworks Inc., Massachusetts, This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 7 USA) neural network toolbox. When data was available, λvE and H were modeled using net radiation, vapour pressure deficit (VPD), wind speed, downward shortwave radiation, and air temperature. Mean coefficients of determination for the gapfilled λvE data were 0.82 at Forest, 0.76 at Buckbrush, and 0.76 at Sparse Shrub (Table S1). Remaining gaps in data were filled using the mean diurnal variation method (Falge et al., 2001), replacing missing values with the mean for that time over the 145 previous and subsequent 7-day period. Missing data from May to September for each tower are summarized in Table S1 with mean values of 35, 30 and 41% missing at Forest, Buckbrush and Sparse Shrub respectively. Periods of missing data were primarily at night and during periods of rain. Mean energy balance closure over the study period was 0.86, 0.74, 0.80 at Forest, Buckbrush and Sparse Shrub respectively (Table S1). Canopy heat storage at Forest was estimated using iButtons placed along the height of the tower at 2 m intervals and was estimated as 1% of net radiation (Rn). 150

Climate
Climate results are confined from May to September for inter-comparison purposes and direct comparisons will be made from 2017 to 2020 when all sites were operational, unless otherwise stated. Average air temperature at the study sites decreased with increasing elevation, with mean May to September air temperatures of 11.1, 8.2, and 7.2°C at Forest, 155 Buckbrush and Sparse Shrub respectively ( Fig. 2; Table 2). The coolest year was 2020, when air temperatures were 0.8, 1.0, and 0.8°C lower than the mean at Forest, Buckbrush and Sparse Shrub respectively. The warmest year was 2019 when air temperatures 0.9, 1.0, and 1.2°C above the mean (Table 2). At all sites, mean monthly air temperature peaked in July ( This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 8 2021). At all sites, 2020 was the wettest with 325 mm rainfall at the shrub sites and 228 mm at Forest from May to 165 September. The driest year was 2018, with 132 and 182 mm received at the forest and shrub sites respectively.
Growing season length was defined as the number of days where net ecosystem exchange (NEE) was < 0 g C m 2 over a 24hour period. Between 2017 and 2020, mean growing season length was 175 (±15) days at Forest, 139 (±18) days at Buckbrush, and 107 (±15) days at Sparse Shrub. The shortest growing season was 2020, with 154, 127, and 89 days at Forest, Buckbrush and Sparse Shrub respectively 170

Soil Moisture
Soil moisture patterns were consistent among years and sites, with volumetric water content (VWC) greatest immediately after snowmelt followed by a gradual decline and intermittent responses to rain events throughout the year (Fig. 3). Using 2019 as an example when both VWC and suction were measured, VWC was consistently highest at Buckbrush throughout the upper 175 soil profile, followed by the Sparse Shrub and Forest soils (Fig. 3). There was considerable difference in VWC between the Forest and shrub sites as soils were consistently drier throughout the year at Forest and had muted responses to melt and rainfall. Soil matric potential sensors indicated that despite fluctuation in VWC with rainfall, soil tension at Buckbrush and Sparse Shrub never fell below -25 kPa, despite long periods without rain later in the year. In contrast, Forest soil tension was much lower, with near surface suctions falling below -30 kPa and 15 cm suctions below -90 kPa, while soil tensions at 30 cm 180 fell below -2500 kPa later in the summer, far below the average wilting point for plants (-1500kPa).

Albedo
The partitioning of available energy varied within and among years. During the middle growing season (June to August), albedo was similar among sites, although values increased with increasing elevation and decreasing vegetation height, with 185 mean mid-day (11:00-14:00) albedo of 0.09, 0.12 and 0.15 at the Forest, Buckbrush and Sparse Shrub, respectively (Fig. 4).
At Forest, albedo was relatively consistent, ranging from 0.08 in July to a maximum of 0.21 in December. In contrast, intra-This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 9 and inter-annual variability was large at the shrub sites, where peak mean daytime albedo during the snow-covered months was 0.50 at Buckbrush and 0.82 at Sparse Shrub. Although the dominant species cover at Buckbrush and Sparse Shrub were similar, Birch and Willow shrubs were on average 0.42 and 0.79 m taller at Buckbrush than Sparse Shrub. This increased net 190 radiation (Rn) during snowmelt and the initial snow-covered period in the fall (Fig. 4).

Energy Partitioning
The partitioning of energy among the study years showed marked yet consistent variability among sites (Fig. 4). On average, Rn peaked in June at Forest and May at both shrub sites, although there were subtle differences among sites, particularly the shrubs, where maximum Rn could occur in later summer months. All sites had a higher sensible heat flux (H) early in the 195 season and then shifted to more λvE dominated as the season progressed (Fig. 4). The transition from H to λvE occurred later in the year with increasing elevation, on average occurring on May 31, June 14 and June 29 at Forest, Buckbrush and Sparse Shrub respectively. From May to September, mean λvE/Rn at Forest was 0.47 (± 0.01), while λvE /Rn was similar between Buckbrush and Sparse Shrub with λvE /Rn of 0.31 (±0.02 and ±0.04) respectively (Table 2). Generally, variability in all interannual energy balance terms increased with elevation and reduced vegetation (Table 2

ET Dynamics
Annual growing season ET declined with increasing elevation, while interannual variability increased. Mean May to September ET totals were 349 (±3) mm, 249 (±10) mm, and 240 (±26) mm at Forest, Buckbrush and Sparse Shrub respectively  Table 2). At Forest, cumulative ET was consistent among the four years of study, regardless of growing season rainfall. 210 This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

10
The Sparse Shrub site had the greatest year-to-year variability compared to Buckbrush, where cumulative ET was relatively more consistent (Fig. 5). From 2015-2020 at the shrub sites, total growing season ET was highest in the 2016, with cumulative ET higher at Sparse Shrub (285 mm) than Buckbrush (261 mm). This year had slightly higher mean May to September air temperature and rainfall (Table 2). Conversely, in the driest year (2018), where the shrub sites only received 182 mm of rainfall, we observed the lowest amount of ET at the Sparse Shrub (192 mm), while ET at Buckbrush was close to the 215 interannual mean of 236 mm. Although 2020 was very wet, receiving 62 mm and 97 mm more rainfall than average from May to September, cumulative ET losses at all sites were close to the mean over the study years. Ratios of May to September ET to rainfall (ET:R) ( Table 2) were the highest and most variable at Forest with mean ET:R of 2.19 (±0.37) compared to 1.22 (±0.09) and 1.14 (±0.05) at Buckbrush and Sparse Shrub respectively. The decrease in variability with increasing vegetation cover is also apparent seasonally. 220 The overall pattern of ET was more dynamic on a seasonal basis at the shrub sites, reflecting plant growth and senescence, with notable declines in ET rates between August and September (Fig. 5, 6). Comparatively, daily ET at Forest was more consistent between months, increasing in spring with a gradual decline in September. Daily ET peaked in July at all sites, with 2017-2020 mean July ET rates of 3.05 (±0.17), 2.48 (±0.09), and 2.13 (±0.31) mm/day (Fig.6). The largest difference between 225 mean monthly ET rates between Forest and the two shrub sites occurred in June (Forest rates 1.0 mm/day greater than Buckbrush, and 0.9 mm/day greater than Sparse Shrub), while the largest difference between Buckbrush and Sparse shrub occurred in July during the peak growing season (Buckbrush ET rates greater by 0.35 mm/day). In all other months, mean ET rates at the shrub sites were similar, with differences less than 0.17 mm/day. Moving 14-day mean daily ET rates (Fig. 6) show larger variability at the Sparse Shrub site than Buckbrush or Forest, particularly early in the season. Variability in ET 230 rates among years was lowest in the late growing season at all sites (early August). This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

Influence of changing vegetation cover on energy partitioning
Small decreases in albedo will have a considerable positive impact on atmospheric temperatures, even with modest shifts in 235 the landscape (Bonfils et al., 2012;Chapin et al., 2005). In all seasons, albedo at the Forest was lower than the Buckbrush and Sparse Shrub sites, and the difference between winter and summer was low (Fig. 4); a characteristic of boreal forests (Betts et al., 1997). Mean mid-day snow-free albedo of Buckbrush (0.12) was slightly lower than Sparse Shrub (0.15), which was similar to shrub and shrub taiga sites reported at a number of northern sites Eugster et al., 2000;Lafleur and Humphreys, 2018;Williamson et al., 2016). Unlike Williamson et al. (2016), who suggested an increase in 240 albedo along a successional gradient in Yukon Territory, we observed a more widely reported decline in albedo with increased abundance and height of shrubs during the snow-free season.
Important feedbacks from future vegetation shifts on energy partitioning occurs during winter and the melt/accumulation season (Bonfils et al., 2012). The influence of shrub height and emergence during melt on snow accumulation, albedo and the 245 energy balance has been well documented in northern environments (Bewley et al., 2010;Marsh et al., 2010;Menard et al., 2012;Pomeroy et al., 2006). When considering early season surface energy balances in varying shrub cover, the abundance of exposed branches controls albedo, which is dependent on the amount of snow, wind redistribution, the plasticity of woody stems and importantly, shrub height . For low-stature shrubs (<0.6 m), Lafleur and Humphreys (2018) found no difference in late winter/spring albedo. Here however, differences in peak-winter albedo between 250 shrub sites were large (0.5 at Buckbrush and 0.82 at Sparse Shrub), similar to results of Sturm et al. (2005) who reported midwinter albedo of 0.85 for snow-covered shrubs and 0.6 when exposed. While the interplay between snow burial and shrub height varies regionally across the pan-arctic, in all cases, shifts in vegetation that act to increase shrub height and abundance will lower albedo, as well the encroachment of forests, with global radiative impacts considering the current rate of ecosystem change. 255 This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

12
The influence of vegetation on energy partitioning of turbulent fluxes has highlighted differences between shrub and forest species across a range of environments (Eaton et al;Eugster et al., 2000;Kasurinen et al., 2014). With respect to shrubs, extent and wetness play a significant role, whereas for forests the role of leaf area index, meteorological conditions, and physiological regulation by vegetation and the distinction between conifer and deciduous species has been highlighted 260 (Brümmer et al., 2012, Kasurinen et al., 2014. While synthesis studies provide important information on partitioning controls, generalizations occur due to the large variability in climate across northern biomes. There have been few studies comparing vegetation assemblages in the same hydroclimatological regime over several years, with the most similar study that of Beringer et al. (2005) who compared surface energy partitioning over one growing season at nearby sites in Alaska among tundra, shrubs and boreal forests. They reported that the greatest increases in H was between tall shrubs and forest, but not between 265 tundra and tall shrubs, which is similar to our sites in that there was less difference in the magnitude and timing of energy partitioning among the shrub and Forest sites. While mean H was greater at the Forest, the relative partitioning of turbulent energy favoured λvE in all years (Table 2). This is in contrast to Beringer et al. (2005) who report higher partitioning of available energy to H, with a Bowen ratio () of 1.22 from June to August at their white spruce site in one year compared to our average of 0.94 over four years. At Forest, there was a gradual decline and stabilization in  throughout the growing 270 season, where in contrast,  at the shrub sites was much more variable on an annual basis and controlled by vegetation phenology (Fig. S1), with higher  occurring in the shoulder season when vegetation was not transpiring. While mean May to September  was similar between Buckbrush and Sparse Shrub (1.32 ±0.13 and 1.34 ±0.26) ( Table 2) among years, seasonal variability in  was larger at Buckbrush than Sparse Shrub, with higher  in May and lower  in the mid-growing season .
Whereas there is equivocal evidence as to how height and abundance affects λvE and total water loss (Lafleur and Humphreys,275 2018), our results suggest encroachment of woody vegetation will likely result in an increased  early in the year, and lower  in mid-growing season. However, the change in net radiation, seasonality of vegetation growth, nature of precipitation delivery, and physiological response will all control total evaporative losses, despite shifts in ratios of turbulent fluxes. The hydrological implications for this, discussed below, are that transitions in vegetation cover in WCRB will likely enhance water loss through an increase in ET despite increases in H. 280 This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

Hydrological Implications of Vegetation Change
While research indicates that increased terrestrial productivity in northern environments will increase ET (Bring et al., 2016;Rawlins et al., 2010;Zhang et al., 2009), there is limited empirical data as to how treeline advance and shrub height and abundance affect ecosystem ET and other components of the hydrological cycle. Our results are similar to those that showed negligible differences in ET between shrubs of different heights (Chapin et al., 2000;Lafleur and Humphreys, 2018;McFadden 285 et al., 2003). However, while seasonal totals were similar, there were distinct processes operating that distinguished the two shrub sites. Diurnal monthly traces of ET and net ecosystem exchange (NEE) (Fig. 7) suggest that during July, increased ET at Buckbrush is driven by enhanced transpiration as reflected by the greater NEE compared with Sparse Shrub, which has more modest mid-season NEE. However, during May, when photosynthesis/transpiration was low, the Sparse Shrub site has greater ET, most likely driven by enhanced direct soil evaporation as there is much more exposed bare ground. These offsetting 290 processes result in annual ET totals being similar, yet the reduced influence of vegetation at the Sparse Shrub site increases ET variability. During wet years, ET is similar between Buckbrush and Sparse Shrub, yet during dry years such as 2018, the Sparse Shrub is notably lower. While shrubs can effectively utilize available soil moisture, sustaining high rates at Buckbrush where vegetation coverage is dense, bare ground evaporation at Sparse Shrub becomes suppressed during dry periods, lowering overall ET. Of note, the length of growing season did not influence total ET between the two shrub sites (Fig. 3), again 295 supporting the role of bare ground evaporation at the Sparse Shrub site.
Although warmer temperatures and a longer growing season partly explain greater total ET at Forest than the shrub sites ( Fig.   7), physiological and ecosystem processes help explain enhanced water loss and low ET variability. Both Forest and Buckbrush have relatively low variability in ET rates among years, particularly later in the growing season (July/August) 300 relative to the Sparse Shrub site (Fig. 6), highlighting a mediating effect of vegetation during the prime growing season.
Monthly diurnal traces of ET and NEE from the three sites indicate that as soon as soils thaw in May, transpiration/photosynthesis begins at Forest, whereas at the shrub sites, ET/NEE in May is negligible before bud-burst in late May or early June (Fig. 7). Monthly mean ET rates at the Forest site are similar in June, July and August, while at the two shrub sites, the greatest difference in mean monthly daily ET occurs in July. Similar timing of senescence between both shrub 305 This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 14 sites is apparent in lower ET rates and declines in carbon uptake. Physiologically, the forest is transpiring/photosynthesizing throughout the growing season at consistent rates, and while there are declines related to periods of low rainfall, the forest effectively uses available water. The role of vegetation in making ET a conservative process has been outlined in theoretical (Jarvis and McNaughton, 1986;Rodriguez-Iturbe and Porporato, 2007), and field studies (Oishi et al., 2010 Phillips andOren, 2001;Roberts, 1983). Therefore, we assert that treeline advance and shrub proliferation will result in 310 more stable, higher ET rates that are less responsive to inter-annual variation in precipitation compared to areas with shorter, less dense shrub cover. However, the total increase in ET from shrub infilling will be muted and increases primarily in the mid-growing season.
The nature of evaporative fluxes coupled with the timing, magnitude and phase of precipitation delivery has important 315 implications for watershed scale hydrological processes. A critical implication of the higher ET at Forest is the increased ET:R and considerable water deficit from May to September (Fig. 5). This indicates the site is heavily reliant on snowmelt recharge and potentially late season rains to sustain high ET (ET:R = 2.19) rates compared with the shrub sites where there is less evident moisture limitation (ET:R =1.20 and 1.14 at Buckbrush and Sparse Shrub respectively). In a synthesis of northern watersheds, Tetzlaff et al. (2013) suggest that with climate change, catchments will move towards increasing aridity 320 as ET increases at a rate greater than precipitation. However, in this region, total precipitation is expected to increase with enhanced rainfall contributions (IPCC, 2019;Tetzlaff et al., 2013). Climate and vegetation change modelling by Rasouli et al. (2019b) in WCRB suggest that at higher elevations, increased in snow trapping due to higher stature shrubs will likely offset the decrease in snow water equivalent due to climate change. This compensation does not, however, apply to boreal forest covers at low elevations. Considering the large growing season water deficit and drier soils, decreases in snow input early in 325 the season or more periods of extreme weather (and potentially drought) could push the Forest into a situation of moisture stress. The growth of white spruce forests in areas of low precipitation and rapid warming is known to be limited due to water deficits, and future climate change may cause higher tree mortality in these populations in high latitudes of North America (Hynes and Hamann, 2020).

330
This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

15
We found the primary hydrological difference between shrub sites was during the mid-growing season (Fig. 7), although on an annual timescale over six years there does not appear to be a large difference in ET when shrubs increase in height and density. This may change as warming alters the timing of snowmelt and growing season length, resulting in a longer "midseason" difference between shrub covers of varying height. The greater influence is the enhanced snow trapping associated with shrub growth (Pomeroy et al., 2006;Rasouli et al., 2019b), which will enhance recharge and runoff generation, 335 particularly during freshet.

Considerations for predicting future hydrologic regimes in northern regions
Despite accelerating research in the north and a new dominance of remote sensing and model-based outcomes at the expense 340 of empirical data, there remains a nascent understanding of how integrated environmental changes will alter the storage and cycling of water in cold regions. While hydrological changes are most often ascribed to warmer air temperatures and changes in frozen ground status at larger scales (Walvoord and Kurylyk, 2016), the influence of changing vegetation communities on hydrological storage and flux remains uncertain as there have been few studies showing only modest influences. There has been considerable research documenting rapid vegetation change, particularly the advance of tree-dominated communities into 345 tundra ecosystems (Danby and Hik, 2007;Conway and Danby, 2014;Harsch et al. 2009;Lloyd and Fastie, 2003) and the proliferation of tall shrubs into low tundra (Myers-Smith and Hik, 2017;Tape et al., 2006;. Feedbacks from shrub expansion include enhanced snow trapping, warming soils, and enhanced nutrient availability, all of which enhance further shrub growth and potential for tree establishment (Myers-Smith et al., 2011;Sturm et al., 2001). The timescales of these changes must be considered, however, in future hydrological projections as treeline advance and infilling is slow (Danby 350 and Hik, 2007) compared with rapid increases in shrub extent, density and height (Myers-Smith et al., 2011;Tape et al., 2006). These changes are also highly dependent on permafrost thaw and disturbances such as fire, which are expected to continue increasing in rate in coming decades (Lantz et al., 2010a). This will, to varying degrees, start new successional shifts depending on the severity and may result in more deciduous forest rather than boreal conifer species. This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

16
To our knowledge, there has been no documentation of surface energy and evaporative fluxes across a northern alpine 355 watershed over multiple years. From an ET perspective, increased forest abundance and shrub proliferation will lead to changing soil moisture regimes and runoff timing and volume. Specifically, increases in ET will reduce runoff generation, particularly in the growing season. For WCRB, most of the streamflow is generated in the alpine and subalpine zone, and upward expansions of trees would reduce total catchment yield. While increase shrubs may trap more snow and provide more water during freshet, mid-season flows are expected to become less variable and decline with increasing shrub abundance. 360 Perhaps the greatest challenge in understanding the role of vegetation change on catchment hydrology is the role of precipitation. Changes from snow to rain, enhanced periods of drought or intense precipitation, all of which are predicted for pan-Arctic ecosystems (IPCC, 2019), will have profound effects on runoff regimes. While ET is a large flux, increased vegetation will likely reduce its variability, emphasizing the role of future precipitation uncertainty in watershed scale 365 hydrological processes. The role of interception on total ET is an area for additional research to evaluate the relative impact of plant structure and physiology. While Zweibeck (2019) suggested that shrub interception reduced below canopy rainfall by 15-30% for shrub species, the role of rainfall interception has not been directly evaluated in this environment. We have a much greater understanding on how vegetation influences snow processes than rainfall processes in cold environments, but the large LAI and relatively open canopy suggests that rainfall interception at the Forest is considerable. 370

Conclusion
The impact of vegetation change on energy and water cycling in the pan-Arctic is an area of intense research, with literature syntheses (Bring et al., 2016;Wrona et al., 2016), numerical model outputs (Rawlins et al., 2010), and remote sensing (Zhang et al., 2009) suggesting ET will increase from increased air temperature, VPD, precipitation, a longer growing season and an expansion of shrubs and trees. There remains, however, few field-based studies that directly evaluate these 375 processes in northern ecosystems. In this multi-year study, we highlight differences and similarities in the timing and magnitude of energy and mass fluxes among boreal forest and shrub taiga ecosystems in an alpine environment where elevation controls temperature and moisture regimes. Results indicate that the greatest change in water and energy fluxes will This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. 17 occur from changes in treeline compared with shrubification across all seasons as they have distinct regimes. ET was greatest at the Forest site compared with the shrub sites, and varied little over 4 years despite differences in meteorology and a 380 considerable growing season water deficit. In contrast, ET was considerably less at the cooler higher elevation shrub sites, which exhibited similar ET losses over 6 years despite differences in shrub height and abundance. While the observation of limited differences in ET has been reported before for a single year, our results indicate that ET variance is much greater at the short, sparse shrub site compared with the tall, dense shrub site due the reduced influence of vegetation controls on total ET. While there are multiple interacting processes, our results support the assertion that a greening north will have lower albedos 385 and higher ET losses. How this affects catchment response will be strongly influenced by precipitation timing, phase and magnitude, growing season length, and vegetation snow and rain interactions.

390
Data will be made available to reviewers upon request and made publicly available upon acceptance of this manuscript. We are willing to provide all data and code to reviewers of this manuscript and upon request in HESS-D. However, as the study site is located on the traditional territories of three Canadian First Nations, provisions under the Ownership, Control, Access and Possession guidelines must be adhered to (see http://fnigc.ca/ocap-training/ for information). Upon acceptance of this manuscript, a DOI will be created for the final submission and the data stored in https://www.frdr-dfdr.ca/ where previous data 395 from Wolf Creek Research Basin has been deposited. Figure 2. 7-day moving mean daily temperatures for Forest (top), Buckbrush (middle) and Sparse Shrub (bottom) for all available years of data. Shaded green regions show mean growing season. The start of the growing season was defined as the first of at least 600 three consecutive days with net ecosystem exchange (NEE) less than 0 g C m -2 and the end of the growing season was defined as the first day where NEE was greater than 0 g C m-2 .

Buckbrush (middle), and Sparse Shrub (right) in 2019.
This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.

630
This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.  This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.  This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed. This manuscript has been submitted for publication in Journal of Hydrology. Please note that this manuscript has not been peer-reviewed.