Tundra shrub expansion may amplify permafrost thaw by advancing snowmelt timing
Abstract
The overall spatial and temporal influence of shrub expansion on permafrost is largely unknown due to uncertainty in estimating the magnitude of many counteracting processes. For example, shrubs shade the ground during the snow-free season, which can reduce active layer thickness. At the same time, shrubs advance the timing of snowmelt when they protrude through the snow surface, thereby exposing the active layer to thawing earlier in spring. Here, we compare 3056 in situ frost table depth measurements split between mineral earth hummocks and organic inter-hummock zones across four dominant shrub–tundra vegetation types. Snow-free date, snow depth, hummock development, topography, and vegetation cover were compared to frost table depth measurements using a structural equation modeling approach that quantifies the direct and combined interacting influence of these variables. Areas of birch shrubs became snow free earlier regardless of snow depth or hillslope aspect because they protruded through the snow surface, leading to deeper hummock frost table depths. Projected increases in shrub height and extent combined with projected decreases in snowfall would lead to increased shrub protrusion across the Arctic, potentially deepening the active layer in areas where shrub protrusion advances the snow-free date.
Résumé
L’incidence spatiale et temporelle globale de l’expansion des arbustes sur le pergélisol est en grande partie inconnue en raison de l’incertitude entourant l’estimation de l’ampleur de nombreux processus de neutralisation. Par exemple, les arbustes ombragent le sol pendant la saison sans neige, ce qui peut réduire l’épaisseur de la couche active. En même temps, les arbustes accélèrent la fonte des neiges lorsqu’ils dépassent de la surface de la neige, exposant ainsi la couche active au dégel plus tôt au printemps. Ici, nous comparons 3056 mesures in situ de la profondeur de la table de gel–dégel réparties entre les hummocks de terre minérale et les zones inter-hummock organiques dans quatre types dominants de végétation arbustive–toundra. La date où il n’y a plus de neige, la profondeur de la neige, le développement d’hummocks, la topographie et le couvert végétal ont été comparés aux mesures de la profondeur de la table de gel–dégel à l’aide d’une approche de modélisation par équation structurelle qui quantifie l’incidence interactive directe et combinée de ces variables. Les zones de bouleaux arbustifs sont devenues dénudées de neige plus tôt, peu importe la profondeur de neige ou l’aspect de pente, parce que ceux-ci dépassaient à travers la surface de la neige, ce qui a mené à des profondeurs de la table gel–dégel plus importantes. Les augmentations prévues de la hauteur et de l’étendue des arbustes, avec les diminutions prévues des chutes de neige, entraîneraient une augmentation de la saillie des arbustes dans l’Arctique, ce qui pourrait rendre la couche active plus profonde dans les zones où la saillie des arbustes fait devancer la date à laquelle il n’y a plus de neige. [Traduit par la Rédaction]
Introduction
The progression of the frost table in permafrost areas to its maximum annual depth, the active layer (Harris et al. 1988), is controlled by climate at a broad scale and a multitude of other variables operating at smaller scales. These “micro-scale” variables include (micro)topography, snow depth and density, and vegetation type and density (Mackay 1980; Sturm et al. 2001; Bonnaventure and Lamoureux 2013; Fisher et al. 2016). Increasing air temperatures promote deeper thaw (Biskaborn et al. 2019); however, the spatially and temporally varying effects of micro-scale variables on ground heat flux mean permafrost thaw does not occur uniformly across the Arctic landscape (Loranty et al. 2018).
Mineral earth hummocks are common microtopographical features in permafrost regions (Schunke and Zoltai 1988) that are susceptible to degradation when the active layer deepens (Kokelj et al. 2007). Hummocks consist of mineral soil domes roughly 50–200 cm in diameter and 10–40 cm in height (Tarnocai and Zoltai 1978; Mackay 1980). Hummocks formed over thousands of years through cyclical upward frost heaving in their centres paired with subsidence at their perimeters, resulting in a locally deeper active layer beneath the centre of the hummocks (Mackay 1980). Hummocks are surrounded by well-defined, peat-dominated inter-hummock zones with a shallower active layer and act as areas of preferential flow (Quinton and Marsh 1999). When hummocks degrade as a result of active-layer deepening, they become flat and spread out radially, producing a deeper inter-hummock active layer due to the intruding mineral soils (Kokelj et al. 2007).
Shrubs are another main control on active layer thickness in tundra regions. Shrubs are increasing in height, density, and areal extent across the Arctic (Tape et al. 2006; Bjorkman et al. 2018), consequently altering permafrost conditions through multiple, counteracting mechanisms (Chapin et al. 2005; Myers-Smith et al. 2011). When shrubs typically buried by snow were experimentally removed, frost table depth deepened due to increased thermal load on the ground surface during the snow-free period (Blok et al. 2010; Nauta et al. 2015). In contrast, shrubs that protrude through the snow surface emit long-wave radiation and advance the timing of snowmelt, which large-scale modeling predicts will offset the shade provided by shrubs (Pomeroy et al. 2006; Marsh et al. 2010; Lawrence and Swenson 2011; Bonfils et al. 2012). However, shrubs protruding above the snow surface decrease surface wind speed, thereby decreasing the turbulent fluxes of heat and water (Endrizzi et al. 2011). In addition, patches of tall shrubs can trap blowing snow (Pomeroy et al. 1997), which in turn increases wintertime ground temperatures as deeper snow insulates the ground more effectively (Sturm et al. 2001; Hinkel and Nelson 2003; Myers-Smith and Hik 2013). Deeper snow also requires longer to melt and delays the beginning of active layer thaw (Wang et al. 2018).
Previous field studies that described the influence of shrub expansion on active layer thickness did not consider the combined influence of shrubs on snowmelt timing and snow redistribution. This study seeks to identify the relative influence of these variables on frost table depth while also evaluating how they influence each other, using a structural equation modeling approach. We used 3056 in situ frost table depth measurements taken from early June until late August 2015, split between mineral earth hummocks and organic inter-hummock zones across four dominant vegetation types and explored the effect of multiple known influences on frost table depth. These include hillslope angle and aspect, snow depth, hummock height, vegetation cover, and snow-free date.
Study site
All observations were made within the Siksik Creek watershed (68.5°N, 133.75°W; Fig. 1), a 1 km2 catchment that flows into Trail Valley Creek, NT (Quinton and Marsh 1999). The site is located ca. 45 km north of Inuvik and ca. 80 km south of Tuktoyaktuk, at the northern edge of the boreal forest–tundra transition zone. Mean annual total precipitation in Inuvik from 1981 to 2010 was 240 mm of which 66% was snowfall, and mean annual air temperature was −8.2 °C (Environment and Climate Change Canada 2019). Snow cover commonly lasts for eight months of the year from October to May during which air temperatures remain below 0 °C and southwesterly winds are dominant. The total thickness of the ice-rich permafrost across the region spans from 350 to 500 m and is overlain by an active layer that varies between 0.5 and 0.8 m in thickness (Burn and Kokelj 2009). The topography of the region between Inuvik and Tuktoyaktuk is characterized by low rolling ice-rich morainal deposits interspersed with thermokarst lakes (Rampton 1987), although no lakes exist within the Siksik Creek watershed. Mineral earth hummocks are present throughout the watershed, surrounded by inter-hummock zones filled with 20–50 cm of peat (Fig. 2; Quinton and Marsh 1998). The bulk density of hummocks is one order of magnitude greater, on average, than inter-hummock zones, which are highly porous (Quinton et al. 2000).
Fig. 1.
Fig. 2.
The vegetation cover of the Siksik Creek watershed is broadly characterized as erect low-shrub tundra (S2 vegetation type as per Walker et al. 2005). We identified four unique vegetation cover classes within the Siksik Creek watershed, which occur in patches of tens to hundreds of metres across (Fig. 1). This type of heterogeneous vegetation cover is common in the treeless region between Inuvik and Tuktoyaktuk (Lantz et al. 2010). Over half of the watershed is covered by a mix of mostly shrub-free “open-tundra”, with maximum vegetation heights of 5–25 cm, which is predominantly covered by reindeer lichen (Cladonia rangiferina L.), Sphagnum moss (Sphagnum L.), tussock and non-tussock sedges (Carex L.), and Labrador tea (Rhododendron groenlandicum (Oeder) Kron & Judd). The second most common vegetation cover type consists of dwarf birch (Betula glandulosa Michx.) patches of 40–60 cm in height with open tundra species present beneath and are generally present on west-facing slopes. Similar, but taller (80–150 cm) alder (Alnus alnobetula (Ehrh.) K. Koch.) patches exist on east facing slopes and also contain open tundra species beneath. In stream channel areas, 150–250 cm alder and willow (Salix L.) grow through a Sphagnum-moss-dominated floor. Open tundra has virtually no canopy cover, whereas birch and channel areas have dense canopy cover with higher leaf area indexes (Marushchak et al. 2013). Individual alder shrubs have greater leaf area indexes than individual birch shrubs (Zwieback et al. 2019); however, alder shrub patches in Siksik Creek are “open” and less dense than birch shrub patches, resulting in a lower mean leaf area index when comparing patch to patch (Marsh et al. 2010; Lantz et al. 2013). Areas of shrub-covered tundra have cooler summer soil temperatures as a result of the shade they provide to the ground (Myers-Smith and Hik 2013). As a result of their denser canopy, birch shrub patches intercept 15%–30% of incoming rainfall, whereas alder shrub areas have little effective rainfall interception (Zwieback et al. 2019). Shrub patches are also characterized by larger evapotranspiration rates than tundra areas and typically have drier soils as a result (Bring et al. 2016). Areas of birch shrubs were observed to protrude through the snow in the Siksik Creek watershed (Fig. 3), whereas other areas of vegetation were mostly buried under the snow and were eventually exposed as the snow melted. The greater region between Inuvik and Tuktoyaktuk is experiencing alder and birch shrub expansion (Lantz et al. 2013).
Fig. 3.
Materials and methods
Frost table depth measurements
After snowmelt in early June 2015, frost table depth measurements began at eight transect and two grid locations within the Siksik Creek watershed (Fig. 1). Transects bisected stream channels to capture a wide range of micro-scale variables as efficiently as possible, whereas grids captured flatter areas. Measurements were made every 5 m along 100 m transects and one 150 m long transect, and two 15 m by 20 m grids, resulting in 20–30 hummock and 20–30 inter-hummock frost table depth measurements at each sampling location. Measurements were taken over the course of a two- to three-day period every one to three weeks. In total, eight rounds of sampling were completed between 11 June and 20 August 2015, resulting in 3056 total frost table depth measurements.
At sampling points along transects and grids, frost table depth in the nearest hummock and inter-hummock zone was measured by pushing a frost probe into the centre of each feature until submission against the frost table. Measurements were split between hummock and inter-hummock zones to control for the known differences in frost table depth between these two features. While leaving the frost probes in the hummock and inter-hummock zone, a construction level was placed perpendicular to the two probes, resting one end on the hummock. The point where the level intersected the inter-hummock frost probe was recorded (Fig. 2). Inter-hummock frost table depth was measured perpendicular to the aspect of the hillslope (e.g., if measurements were made on a north-facing hillslope, inter-hummock measurements were made to the east or the west of each hummock measurement) to minimize the influence of the hillslope on the topography of the frost table. Frost table depth measurements were not always performed at the exact same hummock and inter-hummock zone for each sampling location, but still within a 1 m radius of each measurement location. This was done to minimize soil disturbance that may affect the thawing beneath the measurement points by creating a preferential pathway for water to reach the frost table.
Some sampling points displayed only minimal hummock development, mostly occurring in stream channels which were typically saturated. If no discernible hummocks were present, areas of localized high and low elevation were selected to sample from instead. By sampling regardless of hummock presence or size, the natural variability of hummock development was incorporated into the dataset. If only well-defined hummocks were selected for sampling, then the information gained from the data could only be applied to areas with well-developed hummocks. The spatial distribution of hummocks across the watershed is not known and cannot be estimated using other variables and defining their state of development relies just on visual inspection. Therefore, the applicability of the knowledge gained from the data would be limited if only well-developed hummocks were measured.
Micro-scale variables
Frost table depth per vegetation cover class was compared to the sample date, hummock height, snow depth, snow-free date, hillslope angle, and aspect. Hummock height was considered an indicator of hummock development (Kokelj et al. 2007). Snow depth was derived by subtracting a bare-ground digital elevation model (DEM) captured by airborne light detection and ranging (LiDAR) (Hopkinson et al. 2009) from a DEM of the snow surface before snowmelt began (De Michele et al. 2016; Mann 2018). Each DEM contains a grid of 1 m by 1 m cells which hold an elevation value representing the height of the ground or snow above mean sea level. The snow surface DEM was generated using unmanned aerial system (UAS) imagery obtained by a fixed-wing eBee (SenseFly 2015) and structure-from-motion photogrammetry analysis in Pix4Dmapper (Pix4D SA 2016). Raw images taken by the UAS were geotagged and stitched using 12 ground control points placed throughout the Siksik Creek watershed and measured to an accuracy of 2 cm with a real-time kinematic global positioning system (GPS), to create an orthomosaic using Pix4Dmapper. UAS-derived snow depth from five pre-snowmelt flights were averaged together to create 1 m raster snow depth product for the entire watershed. As snowmelt had begun before UAS flights took place in 2015, images taken before snowmelt in 2016 were used to create the snow depth product. The spatial pattern of snow remains stable between years in tundra environments due to predominant wind directions and wide-spread snow redistribution and is therefore comparable between years for this application (Sturm and Wagner 2010). Measurements of snow depth taken in 2015 using a Magnaprobe (Sturm and Holmgren 2018), which records GPS positions while measuring snow depth, were compared to the 2016 UAS snow depth product. The mean snow depth of 2459 Magnaprobe measurements made in the Siksik watershed in 2015 was 69.6 ± 37.9 cm, whereas the mean snow depth from the 2016 UAS-derived snow depth map at the same points where Magnaprobe measurements were made was 72.5 ± 37.1 cm.
Snow-covered area was classified from UAS true colour imagery collected at multiple times over the course of the snowmelt period in 2015. Snowmelt first occurred on 21 April, and the UAS was flown on 28 April, 8 May, 11 May, 12 May, and 17 May 2015. Approximately 95% of the watershed was snow covered on 28 April, and some areas remained snow covered until the first week of June. Orthomosaics were then imported into ArcGIS version 10.5 (ESRI 2016), and snow-covered area was delineated using an iso-cluster unsupervised image classification with binary classes. Visual inspection of the classification showed that it was able to classify snow presence vs. snow absence accurately. Each frost table depth measurement location was assigned a snow-free date based on the amount of snow cover within a 1 m radius around each frost table depth measurement point. If >90% of the radius around the measurement location was snow free, it was classified as “snow-free”.
Vegetation classifications were delineated manually, as different vegetation cover classes were clearly distinguishable, using UAS true colour imagery of the Siksik watershed captured at ∼5 cm per pixel resolution. Four vegetation classes were delineated: alder shrub-dominated, birch shrub-dominated, open tundra, and channel shrub, as based on the classes outlined in the study site section. We chose these vegetation cover classes because each cover class encapsulates a variety of biotic and abiotic variables and offers a more simplified assessment of different vegetation covers, compared with measurements of numerous biotic variables in each vegetation cover.
Analysis of frost table depth and micro-scale variables
A piecewise structural equation model (SEM) using the R (R Core Team 2019, version 3.5.3) package piecewiseSEM (Lefcheck 2016, version 2.1.0) was used to quantify the effect of micro-scale variables on hummock and inter-hummock frost table depth. The SEM incorporates user-defined mixed effects models generated using the package nlme (Pinheiro et al. 2018, version 3.1-137). Structural equation models determine the effect of explanatory variables on both the dependent variable and other explanatory variables by calculating unstandardized and standardized path coefficients for each causal path of influence defined by the user. Unstandardized path coefficients represent the linear effect of one variable on another, in terms of the explanatory variable’s units and were used to compare how the magnitude of influence of each variable differed between hummock and inter-hummock frost table depths. Standardized path coefficients correct for natural differences in the range of measured variables and allow for the influence of different variables within each SEM to be compared; however, they are unitless and do not represent the direct effect of one variable on another. Path coefficients cannot be calculated for categorical variables (vegetation cover class); instead, the mean of each category is presented and tested for significant difference from all other categories.
The hummock and inter-hummock SEMs share identical paths for variables that do not influence frost table depth directly, as these measurements are identical for each hummock and inter-hummock measurement. Justification for the SEM paths specified in this analysis is provided in the Supplementary Material (S1).1 As part of the SEM development, a directed test of separation was performed on all unspecified paths to determine if any significant relationships existed between them. The test found that hummock height significantly correlated with snow depth. Based on our understanding of hummock development (Mackay 1980), there are no physical mechanisms which would cause areas of deeper snow to develop shorter earth hummocks. This relationship was identified as a correlated error in the SEM, and the path was removed from the analysis. The piecewise SEM approach was chosen over a variance-based approach because it facilitates the use of mixed-effects models, which integrate information about random variability caused by repeated measures across multiple measurement units, such as transects and grids.
Hummock and inter-hummock frost table depth were analyzed separately because we hypothesized that hummock and inter-hummock frost table depth may be influenced by different micro-scale variables, because of their different soil and hydrological characteristics. We expected explanatory variables that block incoming solar radiation to the ground to have less of an effect on inter-hummocks because of the insulative properties of peat, and that inter-hummocks would be more affected by variables that control their heat conductivity (e.g., soil moisture). Explanatory variables in each of the mixed effects models included hummock height, sample date, hillslope angle and aspect, snow depth, snow-free date, and vegetation cover class. The transect or grid where measurements were made was used as the random effect in the model. Fitted and residual model values were inspected visually and showed no violations of homoscedasticity or normality, and no transformations were applied to any variables. Significance for all tests was evaluated at α = 0.05 in R.
Results
Our results showed that hummock and inter-hummock zone frost table depths are influenced by different micro-scale variables. Overall, the strongest influence on hummock frost table depth after sampling date was hummock height (Fig. 4; standardized path coefficient = 0.13), and snow-free date (standardized path coefficient = −0.11). In turn, snow-free date was largely influenced by snow depth (standardized path coefficient = 0.44) and vegetation cover (Fig. 4). Snow depth had the smallest direct effect of any variable on hummock frost table depth (standardized path coefficient = −0.06), despite its ability to delay the snow-free date by 8.14 days per metre of snow. At points where frost table depth was measured, birch areas became snow free significantly earlier than other vegetation types (Fig. 4; p < 0.001). On average, birch areas (group “a”) became snow free six days earlier than alder or tundra (group “b”) and nine days earlier than channel shrub areas (group “c”). These earlier snow-free dates in birch areas corresponded with significantly deeper frost table depths compared with tundra areas. Alder areas had average snow depths 56% greater than tundra areas at points where frost table depth was measured, but the average snow-free date in alder was similar to that in tundra (Fig. 4).
Fig. 4.
The pattern of snow-free dates across vegetation cover types at points where frost table depth was measured did not match the pattern of snow-free dates observed from watershed-wide measurements of snow depth and snow-free date from UAS-measured values. When controlling for hillslope aspect and snow depth using the UAS data, alder areas became snow free latest, followed by channel, then tundra, and birch areas became snow free earliest (Fig. 5). Areas of deeper snow became snow free later, and southwest hillslope aspects became snow free earlier in each vegetation class. The average snow-free date of birch shrub areas was earlier than non-birch shrub areas regardless of the hillslope aspect or snow depth (Table 1).
Fig. 5.
Table 1.
Note: Positive numbers represent the average number of days birch shrub areas became snow free before non-birch shrub areas. A Welch two sample t-test, which is robust against unequal variances and sample sizes, was applied to test for significant differences between alder and birch/tundra areas (bold and italic = p >0.05, italic = p <0.05, bold = p <0.001). The median number of observations in each bin was 9932.
In inter-hummock zones, only sample date, hummock height, hillslope angle, and channel vegetation had a significant effect on frost table depth. Hummock height had a much stronger, negative influence (i.e., shallower frost table depth with greater hummock height) on inter-hummock frost table depth than on hummock frost table depth (unstandardized path coefficients = −0.64 and 0.24, respectively) and had an eight times larger effect on frost table depth than hillslope angle (standardized path coefficient = −0.39; p < 0.001 and 0.05, p = 0.02). The effect of hummock height on inter-hummock frost table depth was 70% as strong as sample date, whereas frost table depth measurements spanned two and a half months (standardized path coefficients = −0.39 and 0.55, respectively). Areas of channel vegetation had significantly deeper inter-hummock frost table depths, 6–9 cm deeper than alder, birch, and tundra areas. Alder, birch, and tundra areas did not significantly differ in average inter-hummock frost table depth. Plots of frost table depth across each micro-scale variable are shown in the Supplementary Material (S2).1
Discussion
Effects of snow–shrub interactions on hummock frost table depth
Here, we show that differences in hummock frost table depth across alder, birch, and tundra vegetation types are partially explained by shrubs’ influence on snow. Hummock height had the largest effect on hummock frost table depth after sample date, but snow-free date has a similarly strong influence on frost table depth based on the standardized path coefficients from the structural equation model (0.13 and −0.11, respectively). Snow-free date has been shown to control active layer thickness in large-scale modeling studies (Lawrence and Swenson 2011; Bonfils et al. 2012; Wang et al. 2018); however, the direct effect of snow-free date on frost table depth has not been reported at the plot scale previously. Our results indicate that the effect of shrub shading is not a dominant control on frost table depth at the Siksik Creek watershed, as areas of birch and channel shrubs had significantly deeper hummock frost table depths than tundra areas, whereas alder and tundra frost table depths did not differ significantly.
In birch areas, deeper frost table depths were partially explained by advanced snowmelt timing caused by their protrusion through the snowpack (Fig. 3). Although birch shrub areas had a significantly lower mean snow depth (which contributed to these points becoming snow free earlier, on average (Fig. 4)), birch areas became snow free significantly earlier than other vegetation cover types regardless of snow depth or hillslope aspect in most cases (Table 1). At points where frost table depth was measured, alder became snow free at the same time as tundra areas despite having 56% deeper snow. However, the 210 points where frost table depth was measured only represents 0.02% of all UAS snow observations. Utilizing all 894 523 UAS observations shows alder areas became snow free later across the watershed regardless of snow depth or hillslope aspect when compared with other vegetation cover classes (Fig. 5). Historical snow surveys completed near-yearly in the Siksik Creek watershed from 1991 to 2016 show that tundra and birch shrub areas have similar mean snow densities (0.226 g cm−3 for birch, 0.222 g cm−3 for tundra), whereas alder areas have a mean snow density of 0.340 g cm−3. Alders tend to grow on steeper slopes within a range of hillslope aspects where snowdrifts would form without alder shrub presence. Wind-packed drifts have denser snow than other areas and require more energy to melt. Therefore, although alder shrubs still increase energy input into the snow the same way birch shrubs do, denser snow in alder areas likely offset this effect and lead to later snow-free dates when compared with birch areas.
Despite growing in snow-drifting areas, snow depth in alder areas was greater than in birch and tundra areas at almost every given hillslope angle and aspect (Table 2). Snow depth was most amplified in areas where greatest numbers of alder tend to appear in the basin, at hillslope angles >10° and aspects from 0° to 135°. Even though snow drifts would form naturally in these areas without the presence of alder shrubs, snow depth in alder areas was greater in these ranges by 0.26–1.35 m, on average (Table 2).
Table 2.
Note: A Welch two sample t-test was applied to test for significant differences between alder and birch/tundra areas (bold and italic = p >0.05, italic = p <0.05, bold = p <0.001). Channel shrub areas are excluded as they are also tall and can therefore trap snow, whereas the purpose of this table is to show that alder shrub areas do increase snow depth. Alder shrubs increase snow depth the most at greater hillslope angles in the range of 0°–135° hillslope aspect. The median number of observations in each bin was 18 466, and the only bin with a p-value >0.05 had just 28 observations.
Previous work that found that removing Betula nana shrubs could deepen the active layer (Blok et al. 2010; Nauta et al. 2015) was conducted in areas where shrub heights were shorter than the relatively shallow, homogenous (25–35 cm) snow pack. Shrubs likely did not protrude through the snow pack at this site and may not have been able to initiate earlier snowmelt. Lafleur and Humphreys (2018) found little difference in snow-free date across three sites of varying shrub height, where mean shrub height and snow depth were closely matched (18.2–51.5 cm shrub height, S1, S2, and G4 vegetation types as per Walker et al. (2005), 29.4–57.1 cm snow depth). The similarity in snow-free date across different vegetation heights found by Lafleur and Humphreys (2018) may be caused by a combination of shallower snow depths and minimal shrub protrusion, as mean snow depth was greater than mean shrub height at each site in their study. The different pattern of results found in our study and others (Blok et al. 2010; Nauta et al. 2015) at shrub–tundra sites highlights the need to better understand interactions between shrub height, snow depth, shrub protrusion, and how they influence snow-free date. The opposing physical effects of shrubs on frost table depth can lead to either shallower or deeper frost table depths in different snow and shrub conditions (Fig. 6). In areas where shrub protrusion does not occur and snow depth is homogenous (Blok et al. 2010), shrubs decreased frost table depth through shrub shading. When shrub protrusion does not occur and snow depth is variable (Nauta et al. 2015), deeper frost table depths occurred in areas of deeper snow. Frost table depth changes in these studies may also be influenced by changes in transpiration and interception caused by the removal of the shrub canopy, which could alter soil moisture. The larger ranges of snow depth and shrub height in our study found that short protruding shrubs increase frost table depth by advancing snowmelt timing. Therefore, we suggest that with projected increases in shrub height, density, and areal extent (Loranty and Goetz 2012; Bjorkman et al. 2018) and decreases in snowfall (Bintanja and Andry 2017), shrub protrusion will increase across the Arctic, leading to earlier snow-free dates that contribute to deeper active layer depths through thawing at the top of the permafrost.
Fig. 6.
Influences on inter-hummock frost table depth, and unmeasured micro-scale variables
Inter-hummock frost table depth was dominantly controlled by hummock height, an indicator of overall hummock development. Previous research in this region has also shown that inter-hummock frost table depth is closely tied to hummock height, and that hummock height is an overall indication of hummock development (Kokelj et al. 2007). Previous work, however, did not consider the influence of snow or vegetation. Our inclusion of snow and vegetation showed that they do not have an impact on inter-hummock frost table depth. Deeper inter-hummock frost table depths in channel shrub areas were likely caused by a lack of hummock development as indicated by significantly lower hummock heights (Table 3). Hummock development was likely impeded by the fully saturated soils in channel shrub areas and not impeded through any mechanism caused by the presence of shrubs themselves. Deeper inter-hummock frost table depths were found in flatter areas (Fig. 4), which could be due to poorer drainage and more saturated soils that conduct heat more effectively. Future increases in active layer thickness will likely trigger hummock degradation, which would amplify active layer increases in the inter-hummock zone as mineral soils from hummocks invade inter-hummock zones (Kokelj et al. 2007). As inter-hummocks transition to mineral soils, variables that influence hummock frost table depth may become influential on inter-hummock frost-table depth. Although hummocks are observed widely across the Arctic (Schunke and Zoltai 1988), their presence or absence at finer scales is unknown, making it difficult to predict the impact their degradation will have.
Table 3.
Note: Difference between classes were tested using the Scheffé multiple comparison test (α = 0.05), which is suited for datasets with different numbers of observations (Scheffé 1953). Only channel vegetation areas had significantly different hummock heights. Superscript letters indicate classes that were significantly different from one another.
In this analysis, we did not measure soil properties (e.g., organic layer thickness, soil moisture, and moss cover), which can potentially influence frost table depth and correlate with measured variables, namely vegetation cover class. However, at a qualitative level, we expect these unmeasured properties in shrub-covered areas likely decrease overall heat flux into the ground compared with tundra areas and would not be confounded with areas that became snow free earlier. As described in the study site section, birch shrub-covered areas typically experience increased transpiration (Bring et al. 2016), intercept up to 30% of rainfall (Zwieback et al. 2019), and experience cooler soil temperatures as a result of shading (Myers-Smith and Hik 2013). This set of properties results in relatively lower soil moisture, which lowers the heat conductivity of the active layer and overall heat flux into the ground, compared with tundra areas. There were also no significant differences in hummock height which may explain differences between alder, tundra, and birch shrub hummock frost table depths (Table 3).
Conclusion
Extensive field measurements of frost table depth were taken over a three-month period across various shrub cover types (alder, birch, channel shrub, and shrub-free tundra-dominated areas) in a catchment containing mineral earth hummocks. Unmanned aerial system mapping allowed us to measure snow depth and snow-free date across the 1 km2 watershed. Our results quantify interactions among snow depth, shrub type, and snow-free date and their influence on frost table depth. Structural equation modeling showed that frost table depth in mineral hummocks and organic inter-hummock zones are controlled by different micro-scale variables. Hummock areas covered by low birch shrub vegetation had deeper frost table depths than tundra areas, likely due to the advancement of snowmelt timing caused by shrubs protruding through the snowpack (Fig. 5; Table 1). However, across the watershed, tall alder shrub areas were able to trap more snow than birch and tundra areas (Table 2), creating a deeper snowpack that delayed snow-free timing. This effect is counter-acted by the ability of alder shrubs to amplify snowmelt once the snow has melted enough to where they protrude through the snowpack. As a result, alder hummock areas did not have significantly different frost table depths when compared with tundra areas, despite the two areas experiencing different snow depth conditions. Inter-hummock areas were not affected by snow or vegetation and are dominantly controlled by hummock height, an indicator of hummock development. Importantly, our results show that low (birch) shrubs increased frost table depth compared with tundra areas at this study site, contradicting previous research that found that shrubs may preserve permafrost by shading the ground. Given projected increases in shrub height, density and areal extent, and decreasing snowfall, shrub protrusion through the snowpack would become more common, and could lead to increases in active layer thickness where shrubs did not protrude before. Further research is required to fully understand how the counteracting forces of shrubs on frost table depth (Fig. 6) will change and affect permafrost into the future at a circum-Arctic scale.
Acknowledgements
The authors wish to acknowledge William Woodley who contributed to the collection of field data at the Trail Valley Creek Research Station. We acknowledge funding from the Natural Sciences and Engineering Research Council of Canada, Polar Knowledge Canada, ArcticNet, and the Canada Research Chair programs. Evan Wilcox’s field research was funded by the Northern Scientific Training Program. The research license (No. 15622) was administered by the Aurora Research Institute in Inuvik, Northwest Territories, Canada, and can be found at http://data.nwtresearch.com/Scientific/15622. The authors thank the three anonymous reviewers for their helpful comments which greatly improved the quality of this manuscript. The authors also thank colleagues Carolina Voigt, Gabriel Hould-Gosselin, and Helena Bergstedt for their comments and suggestions on previous versions of this manuscript.
Footnote
1
Supplementary material is available with the article through the journal Web site at Supplementary Material.
References
Bintanja R. and Andry O. 2017. Towards a rain-dominated Arctic. Nat. Clim. Change, 7: 263–267.
Biskaborn B.K., Smith S.L., Noetzli J., Matthes H., Vieira G., Streletskiy D.A., et al. 2019. Permafrost is warming at a global scale. Nat. Commun. 10: 1–11.
Bjorkman A.D., Myers-Smith I.H., Elmendorf S.C., Normand S., Rüger N., Beck P.S.A., et al. 2018. Plant functional trait change across a warming tundra biome. Nature, 562: 57–62.
Blok D., Heijmans M.M.P.D., Schaepman-Strub G., Kononov A. V., Maximov T.C., and Berendse F. 2010. Shrub expansion may reduce summer permafrost thaw in Siberian tundra. Glob. Chang. Biol. 16: 1296–1305.
Bonfils C.J.W., Phillips T.J., Lawrence D.M., Cameron-Smith P., Riley W.J., and Subin Z.M. 2012. On the influence of shrub height and expansion on northern high latitude climate. Environ. Res. Lett. 7: 015503:.
Bonnaventure P.P. and Lamoureux S.F. 2013. The active layer: a conceptual review of monitoring, modelling techniques and changes in a warming climate. Prog. Phys. Geogr. 37: 352–376.
Bring A., Fedorova I., Dibike Y., Hinzman L., Mård J., Mernild S.H., et al. 2016. Arctic terrestrial hydrology: a synthesis of processes, regional effects, and research challenges. J. Geophys. Res. Biogeosci. 121: 621–649.
Burn C.R. and Kokelj S.V. 2009. The environment and permafrost of the Mackenzie Delta area. Permafr. Periglac. Process. 20: 83–105.
Chapin F.S., Sturm M., Serreze M.C., McFadden J.P., Key J.R., Lloyd A.H., et al. 2005. Role of land-surface changes in Arctic summer warming. Science, 310: 657–660.
De Michele C., Avanzi F., Passoni D., Barzaghi R., Pinto L., Dosso P., et al. 2016. Using a fixed-wing UAS to map snow depth distribution: an evaluation at peak accumulation. Cryosphere, 10: 511–522.
Endrizzi S., Quinton W.L., and Marsh P. 2011. Modelling the spatial pattern of ground thaw in a small basin in the Arctic tundra. Cryosphere, 5: 367–400.
Environment and Climate Change Canada. 2019. Canadian climate normals. [Online]. Available from http://climate.weather.gc.ca/climate_normals/.
ESRI. 2016. ArcGIS Desktop version 10.5. Redlands, Calif., USA.
Fisher J.P., Estop-Aragonés C., Thierry A., Charman D.J., Wolfe S.A., Hartley I.P., et al. 2016. The influence of vegetation and soil characteristics on active-layer thickness of permafrost soils in boreal forest. Glob. Chang. Biol. 22: 3127–3140.
Harris, S.A., French, H.M., Heginbottom, J.A., Johnston, G.H., Ladanyi, B., Sego, D.C., and van Everdingen, R. 1988. Glossary of permafrost and related ground-ice terms. National Snow and Ice Data Center/World Data Center for Glaciology, Boulder, Colo., USA.
Hinkel K.M. and Nelson F.E. 2003. Spatial and temporal patterns of active layer thickness at Circumpolar Active Layer Monitoring (CALM) sites in northern Alaska, 1995–2000. J. Geophys. Res. 108: 1–13.
Hopkinson, C., Fox, A., Monette, S., Churchill, J., Crasto, N., and Chasmer, L. 2009. Mackenzie Delta lidar collaborative research data report. Applied Geomatics Research Group, Dartmouth, N.S., Canada.
Kokelj S.V., Burn C.R., and Tarnocai C. 2007. The structure and dynamics of earth hummocks in the subarctic forest near Inuvik, Northwest Territories, Canada. Arct. Antarct. Alp. Res. 39: 99–109.
Lafleur P.M. and Humphreys E.R. 2018. Tundra shrub effects on growing season energy and carbon dioxide exchange. Environ. Res. Lett. 13: 055001.
Lantz T.C., Gergel S.E., and Kokelj S.V. 2010. Spatial heterogeneity in the shrub tundra ecotone in the Mackenzie Delta region, Northwest Territories: implications for Arctic environmental change. Ecosystems, 13: 194–204.
Lantz T.C., Marsh P., and Kokelj S.V. 2013. Recent shrub proliferation in the Mackenzie Delta uplands and microclimatic implications. Ecosystems, 16: 47–59.
Lawrence D.M. and Swenson S.C. 2011. Permafrost response to increasing Arctic shrub abundance depends on the relative influence of shrubs on local soil cooling versus large-scale climate warming. Environ. Res. Lett. 6: 045504.
Lefcheck J.S. 2016. PiecewiseSEM: piecewise structural equation modelling in R for ecology, evolution, and systematics. Methods Ecol. Evol. 7: 573–579.
Loranty M.M. and Goetz S.J. 2012. Shrub expansion and climate feedbacks in Arctic tundra. Environ. Res. Lett. 7: 011005.
Loranty M.M., Abbott B.W., Blok D., Douglas T.A., Epstein H.E., Forbes B.C., et al. 2018. Reviews and syntheses: changing ecosystem influences on soil thermal regimes in northern high-latitude permafrost regions. Biogeosciences, 15: 5287–5313.
Mackay J.R. 1980. The origin of hummocks, western Arctic coast, Canada. Can. J. Earth Sci. 17: 996–1006.
Mann, P. 2018. Spatial and temporal variability of the snow environment in the Western Canadian Arctic. M.Sc. thesis, Wilfrid Laurier University, Waterloo, Ont., Canada.
Marsh P., Bartlett P., MacKay M., Pohl S., and Lantz T. 2010. Snowmelt energetics at a shrub tundra site in the western Canadian Arctic. Hydrol. Process. 24: 3603–3620.
Marushchak M.E., Kiepe I., Biasi C., Elsakov V., Friborg T., Johansson T., et al. 2013. Carbon dioxide balance of subarctic tundra from plot to regional scales. Biogeosciences, 10: 437–452.
Myers-Smith I.H. and Hik D.S. 2013. Shrub canopies influence soil temperatures but not nutrient dynamics: an experimental test of tundra snow-shrub interactions. Ecol. Evol. 3: 3683–3700.
Myers-Smith I.H., Forbes B.C., Wilmking M., Hallinger M., Lantz T., Blok D., et al. 2011. Shrub expansion in tundra ecosystems: dynamics, impacts and research priorities. Environ. Res. Lett. 6: 045509.
Nauta A.L., Heijmans M.M.P.D., Blok D., Limpens J., Elberling B., Gallagher A., et al. 2015. Permafrost collapse after shrub removal shifts tundra ecosystem to a methane source. Nat. Clim. Change, 5: 67–70.
Pinheiro, J.C., Bates, D., DebRoy, S., Sarkar, D., and Team, R.C. 2018. nlme: linear and nonlinear mixed effects models. [Online]. Available from https://cran.r-project.org/package=nlme.
Pix4D SA. 2016. Pix4Dmapper. Pix4D SA, Lausanne, Switzerland.
Pomeroy J.W., Marsh P., and Gray A.D.M. 1997. Application of a distributed blowing snow model to the Arctic. Hydrol. Process. 11: 1451–1464.
Pomeroy J.W., Bewley D.S., Essery R.L.H., Hedstrom N.R., Link T., Granger R.J., et al. 2006. Shrub tundra snowmelt. Hydrol. Process. 20: 923–941.
Quinton W.L. and Marsh P. 1998. The influence of mineral earth hummocks on subsurface drainage in the continuous permafrost zone. Permafr. Periglac. Process. 9(3): 213–228.
Quinton W.L. and Marsh P. 1999. A conceptual framework for runoff generation in a permafrost environment. Hydrol. Process. 13: 2563–2581.
Quinton W.L., Gray D.M., and Marsh P. 2000. Subsurface drainage from hummock-covered hillslopes in the Arctic tundra. J. Hydrol. 237: 113–125.
R Core Team. 2019. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. [Online]. Available from https://www.r-project.org/.
Rampton, V.N. 1987. Surficial Geology, Tuktoyaktuk Coastlands, District of Mackenzie, Northwest Territories. Geological Survey of Canada, Ottawa, Ont., Canada.
Scheffé H. 1953. A method for judging all contrasts in the analysis of variance. Biometrika, 40(4): 87–104.
Schunke, E., and Zoltai, S.C. 1988. Earth hummocks (Thufur). In Advances in periglacial geomorphology. Vol. 10. Edited by M.J. Clark. Wiley & Sons Ltd., Chichester, UK. pp. 231–245.
SenseFly. 2015. eBee Classic. SenseFly Parrot Group, Cheseaux-sur-Lausanne, Switzerland.
Sturm M. and Holmgren J. 2018. An automatic snow depth probe for field validation campaigns. Water Resour. Res. 54: 9695–9701.
Sturm M. and Wagner A.M. 2010. Using repeated patterns in snow distribution modeling: an Arctic example. Water Resour. Res. 46: 1–15.
Sturm M., Holmgren J., McFadden J.P., Liston G.E., Chapin F.S. III, and Racine C. 2001. Snow–shrub interactions in Arctic tundra: a hypothesis with climatic implications. J. Clim. 14: 336–344.
Tape K., Sturm M., and Racine C. 2006. The evidence for shrub expansion in Northern Alaska and the Pan-Arctic. Glob. Change Biol. 12: 686–702.
Tarnocai C.T. and Zoltai S. 1978. Earth hummocks of the Canadian Arctic and subarctic. Arct. Alp. Res. 10: 581–594.
Walker D.A., Reynolds M.K., Daniëls F.J.A., Einarsson E., Elvebakk A., Gould W.A., et al. 2005. The circumpolar Arctic vegetation map. J. Veg. Sci. 16: 267–282.
Wang Z., Kim Y., Seo H., Um M.-J., and Mao J. 2018. Permafrost response to vegetation greenness variation in the Arctic tundra through positive feedback in surface air temperature and snow cover. Environ. Res. Lett. 14: 044024.
Zwieback S., Chang Q., Marsh P., and Berg A. 2019. Shrub tundra ecohydrology: rainfall interception is a major component of the water balance. Environ. Res. Lett. 14: 055005.
Supplementary Material
File (as-2018-0028suppla.pdf)
- Download
- 494.51 KB
Information & Authors
Information
Published In
Arctic Science
Volume 5 • Number 4 • December 2019
Pages: 202 - 217
History
Received: 26 October 2018
Accepted: 9 August 2019
Accepted manuscript online: 22 August 2019
Version of record online: 22 August 2019
Copyright
© 2019 The Author(s). This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Key Words
Mots-clés
Authors
Metrics & Citations
Metrics
Other Metrics
Citations
Cite As
Evan J.Wilcox, DawnKeim, Tylerde Jong, BrandenWalker, OliverSonnentag, Anastasia E.Sniderhan, PhilipMann, and PhilipMarsh. 2019. Tundra shrub expansion may amplify permafrost thaw by advancing snowmelt timing. Arctic Science.
5(4): 202-217. https://doi.org/10.1139/as-2018-0028
Export Citations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.
Cited by
1. Investigating ten years of warming and enhanced snow depth on nutrient availability and greenhouse gas fluxes in a High Arctic ecosystem
2. Transpiration and evaporative partitioning at a boreal forest and shrub taiga site in a subarctic alpine catchment, Yukon territory, Canada
3. Change Detection Applications in the Earth Sciences Using UAS-Based Sensing: A Review and Future Opportunities
4. Mechanisms and Impacts of Earth System Tipping Elements
5. Estimating snow cover from high-resolution satellite imagery by thresholding blue wavelengths
6. Active layer variability and change in the Mackenzie Valley, Northwest Territories between 1991-2014: An ecoregional assessment
7. Sensitivity of seasonal air temperature and precipitation, and onset of snowmelt, to Arctic Dipole modes across the Taiga Plains, Northwest Territories, Canada
8. Vegetation type is an important predictor of the arctic summer land surface energy budget
9. Monitoring 13 years of drastic catchment change and the hydroecological responses of a drained thermokarst lake
10. Thermodynamic basis for the demarcation of Arctic and alpine treelines
11. Winters are changing: snow effects on Arctic and alpine tundra ecosystems1
12. The International Tundra Experiment (ITEX): 30 years of research on tundra ecosystems1
13. Scale‐dependent responses of understory vegetation to the physical structure of undisturbed tundra shrub patches
14. Large herbivores on permafrost— a pilot study of grazing impacts on permafrost soil carbon storage in northeastern Siberia
15. Sensitivity of Erosion‐Rate in Permafrost Landscapes to Changing Climatic and Environmental Conditions Based on Lake Sediments From Northwestern Alaska
16. PermaBN: A Bayesian Network framework to help predict permafrost thaw in the Arctic
17. Thermal bridging by Arctic shrubs
18. Permafrost cooled in winter by thermal bridging through snow-covered shrub branches
19. Shrubification along Pipeline Corridors in Permafrost Regions
20. Comparison of Soil Nutrient Supply Patterns among Full and Drained Beaver Ponds and Undisturbed Peat in a Rocky Mountain Fen
21. Standardized monitoring of permafrost thaw: a user-friendly, multiparameter protocol1
22. Airborne SnowSAR data at X and Ku bands over boreal forest, alpine and tundra snow cover
23. Potential of X-band polarimetric synthetic aperture radar co-polar phase difference for arctic snow depth estimation
24. Impact of measured and simulated tundra snowpack properties on heat transfer
25. Tundra vegetation change and impacts on permafrost
26. Assessing the influence of lake and watershed attributes on snowmelt bypass at thermokarst lakes
27. Application of L-band SAR for mapping tundra shrub biomass, leaf area index, and rainfall interception
28. Recent greening driven by species-specific shrub growth characteristics in Nunatsiavut, Labrador, Canada
29. Evapotranspiration and energy partitioning across a forest-shrub vegetation gradient in a subarctic, alpine catchment
30. Modelled Soil Temperature Sensitivity to Variable Snow and Vegetation Conditions in Low-Relief Coastal Mountains, Nunatsiavut and NunatuKavut, Labrador
31. Assessing Pathways of Climate Change Effects in SpaDES: An Application to Boreal Landbirds of Northwest Territories Canada
32. Accuracy assessment of late winter snow depth mapping for tundra environments using Structure-from-Motion photogrammetry1
33. The Canadian Federation of Earth Sciences Scientific Statement on Climate Change – Its Impacts in Canada, and the Critical Role of Earth Scientists in Mitigation and Adaptation
34. Rapid shrub expansion in a subarctic mountain basin revealed by repeat airborne LiDAR
35. Seasonal thaw and landscape position determine foliar functional traits and whole‐plant water use in tall shrubs on the low arctic tundra
36. Biophysical controls of increased tundra productivity in the western Canadian Arctic
37. Rapid Ecosystem Change at the Southern Limit of the Canadian Arctic, Torngat Mountains National Park
38. Does tall vegetation warm or cool the ground surface? Constraining the ground thermal impacts of upright vegetation in northern environments
39. Arctic tundra shrubification: a review of mechanisms and impacts on ecosystem carbon balance
40. Integrating Arctic Plant Functional Types in a Land Surface Model Using Above‐ and Belowground Field Observations
41. Summary and synthesis of Changing Cold Regions Network (CCRN) research in the interior of western Canada – Part 2: Future change in cryosphere, vegetation, and hydrology
42. A low-cost method for monitoring snow characteristics at remote field sites
43. Snow water equivalent measurement in the Arctic based on cosmic ray neutron attenuation
44. CliGAN: A Structurally Sensitive Convolutional Neural Network Model for Statistical Downscaling of Precipitation from Multi-Model Ensembles
45. Advancing Field-Based GNSS Surveying for Validation of Remotely Sensed Water Surface Elevation Products
46. Early Warning from Space for a Few Key Tipping Points in Physical, Biological, and Social-Ecological Systems
47. Riverbank erosion in cold environments: Review and outlook
48. Nitrous oxide emissions from permafrost-affected soils
49. Feasibility of tundra vegetation height retrieval from Sentinel-1 and Sentinel-2 data
50. Linking tundra vegetation, snow, soil temperature, and permafrost