Drivers of tall shrub proliferation adjacent to the Dempster Highway, Northwest Territories, Canada

Arctic ecosystems are undergoing rapid changes as a result of climate warming and more frequent disturbances. Disturbances can have particularly large effects on high-latitude ecosystems when ecosystem structure and function is controlled by strong feedbacks between soil conditions, vegetation, and ground thermal regime. In this study we investigated the impact of road construction and maintenance on vegetation structure and biomass along the Dempster Highway where it crosses the Peel Plateau in the Northwest Territories. To explore drivers of tall shrub proliferation and to quantify shrub proliferation in this region of continuous permafrost, greyscale air photos (1975) and Quickbird satellite imagery (2008) were used to map landcover change within two 0.6 km2 belts next to the road and two 0.6 km2 belts 500 m away from the road. Maps showing areas where: 1) tall shrubs expanded, and 2) dwarf shrub tundra resisted invasion were then used to select field sites where a suite of biophysical variables were measured. Rapid tall shrub proliferation and greater biomass adjacent to the road indicate that disturbance can facilitate vegetation change in tundra environments. Our field data also suggests that increased shrub proliferation adjacent to the road was caused by greater soil moisture. Tall shrub proliferation adjacent to the road occurred at lower elevation sites characterized by wetter soils with thicker organic layers. Areas that resisted tall shrub encroachment were located at higher elevations and had drier soils with thin organic layers. Our observations also support previous work illustrating that tall shrub expansion next to the highway promotes strong positive feedbacks to ongoing shrub growth and proliferation.


Introduction
The structure and function of arctic ecosystems is changing in response to recent climate warming Recent research also shows that disturbances can transform Arctic vegetation. Disturbances such as thaw slumps, lake drainage, tundra fire, and frostheave all facilitate rapid shrub expansion in areas where shrubs were not previously dominant (Mackay and Burn 2002, Lantz et al 2009, 2010a, 2013, Frost et al 2013. Field studies of seismic lines, roads, and drilling mud sumps indicate that human-caused disturbances also stimulate tall shrub growth (Auerbach et al 1997, Johnstone and Kokelj 2008, Kemper and Macdonald 2009a, Gill et al 2014.
Evidence from plot-scale warming experiments Some evidence also indicates that the effect of temperature on tall shrub proliferation is mediated by soil moisture. Analysis of tall shrub growth rings and vegetation composition in permanent plots both show that increased shrub growth has been favoured at relatively wet sites (Elmendorf et al 2012b, Myers-Smith et al 2015). Tape et al (2006) also observed rapid tall shrub expansion in wet, high resource environments and snow depth manipulation experiments suggest that moisture facilitates shrub growth (Wahren et al 2005). Other studies also suggest that shrub proliferation at disturbed sites may also be mediated by changes to hydrology (Johnstone and Kokelj 2008, Naito and Cairns 2011b, Gill et al 2014, but additional field studies are required to test this hypothesis. Linear disturbances provide an excellent opportunity to study the edaphic factors that mediate tall shrub proliferation because they typically traverse a range of ground conditions. Historical images of the Dempster Highway from 1975 that precede its official opening to traffic in 1979 and Quickbird imagery from 2008 allow comparisons of vegetation structure before and after prolonged disturbance from road use. Inspection of these images suggests that patchy shrub proliferation is related to variation in hydrological changes along the highway corridor. To test the hypothesis that increases in tall shrub density are linked to hydrological changes following road construction, we compared current (2013) biophysical factors between areas where shrub density increased between 1975 and 2008 with areas where the vegetation structure did not change during this period.
Insight into the causes of tall shrub proliferation is critical to our ability to forecast the nature and extent of Arctic vegetation change. Understanding the drivers of shrub proliferation is important for infrastructure management because shrub-snow feedbacks can increase ground temperatures and lead to permafrost thaw, which threatens terrain stability and increases the cost of infrastructure maintenance and repair (Gill et al 2014, O'Neill et al 2015. Understanding the factors that facilitate or constrain shrub proliferation is also significant because vegetation change can affect carbon storage, nutrient cycling, energy fluxes, hydrology, and ground thermal regime

Study area
This research was conducted in the northern portion of the Peel Plateau ecoregion, along a 14 km stretch of the Dempster Highway in the Northwest Territories (figure 1). This section of the highway is bounded to the west by the Richardson Mountains and by the Peel River valley to the east. The Dempster Highway was constructed between 1959 and 1979 and passes over continuous permafrost (Smith et al 2005, Tunnicliffe et al 2009, O'Neill et al 2015. This area is situated at the edge of the boreal forest in the taiga plains ecozone (Roots et al 2004) and has elevations that range from 150 to 600 m above sea level. Vegetation communities vary with elevation with black spruce forest transitioning into shrub-dominated communities at higher elevations (Roots et al 2004). In our study area, tall shrub tundra is dominated by patches of Alnus fruticosa ((Ruprecht) Nyman) that are between 40 and 400 cm tall. Rubus chamaemorus (L.), Betula glandulosa (Michx.), and Vaccinium spp (L.) are present in the understory. Dwarf shrub tundra in this location is characterized by vegetation less than 40 cm tall, which is dominated by Empetrum nigrum (L.), Ledum decumbens (Aiton), Rubus chamaemorus, Betula glandulosa, and Vaccinium spp. Tall and dwarf shrub tundra at our study site is analogous to erect shrub tundra in the circumpolar vegetation map .
The climate in this area is area is characterized by short cool summers and long cold winters. In Fort McPherson, the mean annual air temperature is −6.2°C, and the mean summer air temperature is 13.3°C. In this region, mean annual air temperatures have increased by 0.77°C per decade since the 1970s

Airphoto analysis
To map land cover change in the study area, greyscale aerial photos from 1975 were compared with pansharpened Quickbird imagery acquired in September, 2008. Quickbird imagery had a resolution of 0.6 m. Greyscale aerial photos from 1975 (1:15 000) were scanned at 1200 dpi, but had an effective pixel size of illustrate a transition from dwarf shrub tundra to tall shrub tundra (arrows). Images (c) and (d) demonstrate an increase in ponding adjacent to the Dempster Highway (arrows). Images (e) and (f) indicate a stable patch of dwarf shrub tundra (arrows). Images (g) and (h) show a stable patch of tall shrub tundra. Arrows in (g) and (h) show the same shrub patches in both time periods. 0.6 m, and were processed using the computer program Summit Evolution (version 6.4, DAT/EM Systems International, Alaska) to create soft copy stereo models. Vegetation mapping was completed inside 14 km belts on both sides of the road. Two of these were located adjacent to the road (road belts) and two were positioned away from the highway (control belts). The road belts extended 22 m past the toe of the road embankment. Control belts located 500 m away from the Dempster on both sides of the road were also 22 m wide. Both the control and road belts covered a total area of approximately 1.2 km 2 . In both sets of imagery, tall shrubs, dwarf shrubs, and water were mapped when their area exceeded 1 m 2 (figure 2). Mapping was undertaken by one person and was completed onscreen while viewing softcopy stereo (1975) or Quickbird images (2008). Absolute change in tall shrub, dwarf shrub, and water cover was calculated from the area of each land cover type in 1975 and 2008 as: Maps of land cover from each time period were also used to map areas of landscape change and stability (figure 2). This was accomplished by using the RIKS Map Comparison Toolkit (version 3.3, Netherlands Environmental Assessment agency, The Netherlands) to produce maps showing areas of stable tall shrub cover, stable dwarf shrub cover, and tall shrub expansion.

Field sampling
To contrast biotic and abiotic conditions beside the road in: (1) areas of dwarf shrub that transitioned to tall shrub with (2) areas of dwarf shrub that resisted invasion, we used maps of land cover change to select field sites and verified these locations in the field. To minimize the effects of mapping error we selected the largest possible areas that exhibited extensive (>1600 m 2 ) tall shrub proliferation. Stable dwarf shrub sites were located in polygons larger than 1850 m 2 that had resisted all tall shrub proliferation. Stable tall shrub sites were areas that remained dominated by tall shrubs from 1975 onwards, and were only included in vegetation community analysis. Field sites were separated from each other by at least 300 m and were distributed across the north and south sides of the road. Fifteen roadside field sites in each vegetation type were located between 11 and 14 m from the toe of the road embankment and consisted of three subplots (5 m 2 ) that were 3 m from the center of the site on 120°, 240°, and 360°bearings (n=135).
At each subplot, the percent cover of shrubs and trees was estimated inside a 5 m 2 quadrat. The percent cover of understory vegetation was estimated using a 0.625 m 2 quadrat randomly nested within the larger subplot. Gravimetric soil moisture was measured inside the 5 m 2 quadrat at each subplot by collecting a 250 cm 3 composite active layer sample. Soil samples were weighed to the nearest tenth of a gram, and then dried at 90°C for 48 h in an oven. Gravimetric soil moisture (percent) was calculated using the following formula presented in Auerbach et al (1997): Gravimetric soil moisture wet weight dry weight dry weight 100 . 3 A portion of the 250 cm 3 composite active layer sample was used to measure the soil pH of each subplot by vigorously mixing 10 mg of soil with 30 ml of deionized water. The soil mixture was left to stand for two hours before measuring soil pH with a pH meter (Oakton Model 510 pH meter, YSI Environmental 2006). Within each 5 m 2 subplot, six active layer measurements were acquired by pushing a graduated soil probe to the depth of refusal. In hummocky terrain measurements were restricted to hummock tops. A metal ruler was inserted into a small hole to measure litter and organic soil thickness at each 5 m 2 subplot.

Biomass
To estimate the biomass associated with the shrub canopy, we installed 24 transects parallel to the road at locations that were distinct from the field sites used to characterize biotic and abiotic conditions. Transect locations were selected using the vegetation maps of each belt. Twelve transects were located in tall shrub (n=6) and dwarf shrub vegetation (n=6) adjacent ) at tall shrub expansion and dwarf shrub sites adjacent to the Dempster Highway, and at tall shrub expansion sites >500 m from the highway. Bars show means for each site type, and error bars represent the 95% confidence interval of the mean. Three asterisks ( *** ) indicate that the contrast is significantly different (α=0.05). (b) Estimated dry Alnus fruticosa biomass (kg) in 1975 and 2008 at tall shrub expansion and dwarf shrub sites adjacent to the Dempster Highway and >500 m from the highway. Estimates from both time periods were calculated using average Alnus fruticosa biomass (kg m −2 ) at tall shrub expansion and dwarf shrub sites adjacent to the Dempster Highway, and at tall shrub expansion sites >500 m from the highway. Changes in biomass across the study area were obtained by multiplying plot-based estimates (kg m −2 ) with the area of tall shrub tundra in the belt adjacent to the Dempster Highway, and the belt more than 500 m from the highway.
to the Dempster Highway and twelve transects were located 500 m away from the highway in tall shrub (n=6) and dwarf shrub vegetation (n=6). Transects were 200 m long, and were located at least 300 m apart. Along each transect, three points were selected at random distances and the basal diameter of each Alnus fruticosa ramet was measured within a 5 m 2 quadrat centered on the transect point. Since no tall shrubs were observed at dwarf shrub transects within the control belt, this category was excluded from subsequent analysis. We measured a total of 435 Alnus ramets. On average, there were 34 ramets sampled in each subplot. To estimate Alnus fruticosa biomass, we used the generalized allometric equation derived by Berner et al (2015) relating Alnus basal diameter to dry biomass (kg m −2 ). To estimate Alnus dry biomass (kg) in tall shrub and dwarf shrub belts adjacent and away from the Dempster in 1975 and 2008, we multiplied our plot-scale estimates of Alnus dry biomass (kg m −2 ) with the area (m 2 ) of tall shrub tundra in each belt.

GIS analysis
To examine associations between biophysical factors and roadside tall shrub proliferation at a broader-scale, we compared maps of vegetation change with biophysical parameters derived from a LiDAR DEM (2013) of the Peel Plateau. The LiDAR DEM had a horizontal resolution of 1m and a vertical resolution of <1 m. We used ArcGIS and this DEM to calculate elevation, aspect, slope, and area solar radiation (ASR). ASR is calculated based on latitude, elevation, and surface orientation. A topographical wetness index (TWI) raster was also calculated in ArcGIS using the following formula provided by Sörensen et al (2006), where flow accumulation is based on the local upslope area that flows into each cell: To characterize elevation, aspect, slope, ASR, and TWI, we used ArcGIS to select 1000 random points in areas beside the road mapped as stable dwarf shrub and tall shrub expansion (n=2000). To reduce the likelihood of mapping error, the following constraints were applied to point selection: (1) random points were allocated to areas of tall shrub expansion and  To test for significant differences in biomass and abiotic and biotic conditions at tall shrub expansion and stable dwarf shrub sites, we used the GLIMMIX procedure in SAS (version 9.3) to create linear mixed models (SAS Institute, Cary, NC, USA). In models for abiotic and biotic conditions, site type (tall shrub expansion, stable dwarf shrub) was included as a fixed factor, and site and subplot were treated as random factors. In biomass models, transect identity (roadside tall shrub, control tall shrub, roadside dwarf shrub) was treated as a fixed factor, and transect number was treated as a random factor. In both cases the Kenward-Roger approximation was used to estimate the degrees of freedom in these models (Kenward and Roger 1997).
To explore the interrelationships among biotic and abiotic variables measured in the field, we used the statistical program R to perform a principal components analysis (R Core Team 2013). A correlation matrix was selected because abiotic factors were measured on different scales. To assess the significance of variable loadings on PC 1 and PC 2, we used R to perform 1000 permutations of a bootstrapped sample (Peres-Neto et al 2003).

Results
Disturbance associated with the construction and maintenance of the Dempster Highway has caused vegetation change adjacent to the road between 1975 and 2008 ( figure 3). Shrub proliferation was more extensive adjacent to the Dempster, where the absolute increase in tall shrub cover was 42% ( figure 3(a)) and the relative increase in tall shrub cover was 525% ( figure 3(b)). In areas more than 500 m from the road, the absolute increase was only 5% (figure 3(a)), and the relative increase in tall shrub cover was 34% ( figure 3(b)). The road had a significant impact on hydrology and large ponds were frequently visible on the modern images (figure 2). The absolute increase in the cover of water adjacent to the road was 3.4%. More than 500 m from the road, the absolute increase in the cover of water was negligible. Tall shrub expansion and increases in the cover of water were accompanied by concomitant decreases in the area of dwarf shrub ( figure 3(a)). Although the area of dwarf shrub tundra decreased from 1975 to 2008, large patches of stable dwarf shrub tundra persisted adjacent to the Dempster (figure 2).
Average tall shrub biomass adjacent to the road was close to 13 times greater per square meter than tall shrub biomass in the control belt (figure 4, F 2,15 =17.36, p<0.001 ). Average tall shrub biomass (kg m −2 ) at dwarf shrub sites adjacent to the road and tall shrub sites away from the road was not significantly different (figure 4(a)). By combining our air photo estimates of the change in areal cover of shrubs with the allometric measurements made in the field, we estimate that increases in tall shrub biomass next to the road were 106 times greater than increases in shrub biomass 500 m away from the road ( figure 4(b)). From 1975 to 2008, tall shrub biomass at the Dempster increased by a factor of 81.2, whereas at the control belt, tall shrub biomass only increased by a factor of 1.3 ( figure 4(b)).
At sites where tall shrub expansion occurred, the vegetation was significantly different from stable dwarf shrub sites (R ANOSIM =0.89, p<0.001) (figure S1, table S1). Tall shrub expansion sites were characterized by greater cover of A. fruticosa and litter, and lower cover of R. chamaemorus, L. decumbens, E. nigrum, and V. vitis-idaea than stable dwarf shrub sites. The vegetation at stable tall shrub sites differed significantly from stable dwarf shrub sites (R ANOSIM =0.93, p<0.001). This difference was driven by greater cover of A. fruticosa, Salix spp, and litter and lower cover of R. chamaemorus and ericaceous shrubs at stable tall shrub sites. Vegetation community  1985 1, 1998 composition at stable tall shrub tundra and tall shrub tundra expansion sites was nearly indistinguishable (R ANOSIM =0.12, p<0.001), with the main differences being greater Salix spp. cover at stable tall shrub sites, and greater A. fruticosa cover in tall shrub expansion sites (table S1). Biotic and abiotic response variables measured adjacent to the road in 2013 varied between site types ( figure 5). The principle component analysis shows that higher soil moisture, greater shrub height and thicker organic soils were strongly associated with tall shrub expansion sites and correlated with each other. Deeper active layers were strongly associated with stable dwarf shrub sites and were negatively correlated with greater shrub height and thick organic soil and to a lesser degree, soil moisture (figure 5). Comparisons of field-measured biotic and abiotic response variables revealed significant differences between sites types (table 1, figure S2). The average elevation of dwarf shrub sites was 46 m higher than tall shrub expansion sites, but the average embankment height did not differ (table 1, figure S2(a)). Soil conditions also showed significant differences between site types. Tall shrub expansion sites were 2.3 times wetter and had organic soils horizons close to twice as thick as stable dwarf shrub sites (table 1, figures S2(c), S2(d)). Active layer thickness at tall shrub expansion sites was significantly lower when compared with stable dwarf shrub sites (table 1, figure S2(e)), but soil pH and average litter depth were similar between sites types (table 1, figures S2(f) and S2(g)). Maximum shrub height was 7 times greater at tall shrub expansion sites than stable dwarf shrub sites, but maximum understory height did not differ between site types (table 1, figure S2(h)).
Abiotic variables derived from GIS revealed differences between stable dwarf and tall shrub expansion sites (table 2, figure S3). Stable dwarf shrub sites occurred at higher elevations than tall shrub expansion sites beside the road (table 2, figure S3(a)) and the topographic wetness (TWI) index was significantly higher at tall shrub expansion sites than stable dwarf shrub sites (table 2, figures S3(b)). The incidence of ASR was higher at tall shrub expansion sites (table 2, figure S3(c)), but slope was not significantly different between site types (table 2, figure S3(d)).

Discussion
Rapid tall shrub proliferation next to the Dempster Highway suggests that abiotic changes associated with road construction and maintenance have intensified the effects of a warming climate on shrub growth and reproduction. Increases in shrub cover of approximately 0. strongly indicate that pan-arctic tall shrub expansion has been driven by increases in air temperatures. Rapid temperature increases in Canada's western subarctic have been reported (IPCC 2013) and it is likely that the direct and indirect effects of warmer air temperatures have contributed to the shrub expansion we observed at sites on the Peel Plateau. Large increases in the biomass of tall shrubs adjacent to the road between 1975 and 2008 show that abiotic changes associated with disturbance drive significantly larger changes in tall shrub productivity than the effects of warmer air temperatures alone. This clearly indicates that tundra disturbances can have a disproportionately large impact on ecosystem configuration relative to the size of the disturbance-affected area (Lantz et al 2010, Frost et al 2013. Greater shrub proliferation and biomass adjacent to the road was likely caused by the effects of elevated soil moisture on Alnus establishment and growth. We observed higher gravimetric soil moisture readings at tall shrub expansion sites next to the road, and TWI values along the Dempster indicated that tall shrub expansion sites had higher soil moisture compared to stable dwarf shrub sites. Our air photo analysis also revealed increases in standing water next to the road between 1975 and 2008, suggesting that changes to soil moisture regimes were pronounced adjacent to the Dempster. In contrast, minute increases in standing water at sites away from the road during this same time period corresponded with much smaller changes to tall shrub proliferation and biomass away from the road. Generally, patches of stable dwarf shrub tundra persisted at the crest of elevated ridges along the plateau, where soils were drier and vegetation was more exposed to snow scouring from winter winds (Blok et al 2015). The idea that dry soils limit tall shrub proliferation is supported by our observation that shrub patches in 1975 were constrained to drainages and water tracks. Alnus fruticosa is known to favour mesic to moist soil conditions since higher soil moisture allows for increased rates of N mineralization and accelerated shrub growth rates (Furlow 1979, Hendrickson et al 1982, Binkley et al 1994, Myers-Smith et al 2015. Blowing snow that accumulates in drainages and water tracks can also protect shrubs from wind and desiccation damage (Ropars et al 2015, Swanson 2015). Our field evidence that tall shrub proliferation is facilitated by elevated soil moisture is consistent with Tape et al's (2006) observation that shrub expansion in Alaska occurred preferentially in wet, high resource environments.
Increased soil moisture and altered drainage patterns beside the road was likely caused by increased snow accumulation next to the road embankment. Research in other areas of the subarctic and arctic has shown that obstructions, such as trees, roads, and snow fences promote snow drift formation and can significantly increase maximum winter snow depth (Hiemstra et al 2002, Hinkel and Hurd 2006, Burn et al 2009, Fortier et al 2011. Deeper snow increases localized spring run-off and likely elevates soil moisture (Wahren et al 2005, Hinkel andHurd 2006). Deeper snow pack also insulates the ground, reduces winter cooling, and can promote permafrost degradation and thicker active layers As well as providing larger inputs of high quality litter, the proliferation of Alnus, a minerotrophic shrub, has been associated with favourable conditions for organic soil development (Zasada 1986, Buckeridge et al 2010, Frost et al 2013. Thicker organic soils affect ground thermal properties (Dyrness 1982), and a larger shrub canopy enhances road dust interception, which increases nutrient availability and promotes shrub growth (Gill et al 2014).
Despite reports of warmer ground temperatures at tall shrub sites adjacent to the Dempster (Gill et al 2014), we found that active layer thickness was significantly reduced at roadside tall shrub expansion sites. It is likely that summer shading by the tall shrub canopy, thick organic soil layers, and increased soil moisture reduced summer ground thaw and limited active layer development (Blok et al 2010). However, Gill et al's (2014) observation of elevated permafrost temperatures beneath shrub canopies beside the road, indicates that winter conditions have a larger impact on ground thermal regime than summer processes (Romanovsky and Osterkamp 2000, Sturm et al 2001, Palmer et al 2012, Gill et al 2014. Additional research on the impact of tall shrub proliferation on the balance between winter and summer heat flux is needed, but the results from our study, as well as work by O'Neill et al (2015) and Gill et al (2014) indicate that continued shrub growth has the potential to facilitate permafrost degradation and compromise the structural integrity of the Dempster Highway.
Although our study area was relatively small, we believe that the Peel Plateau is representative of erect shrub tundra across the Arctic. This terrain type makes up about 18% of the arctic and includes large areas of: the Seward Peninsula, the Tuktoyaktuk Coastlands, the Hudson's Bay Lowlands and eastern Russia . Our findings suggest that disturbances affecting hydrological conditions and snow accumulation patterns in this terrain type will have strong impacts on vegetation structure and biomass. Our observations from the Peel Plateau also suggest that temperature-induced shrub proliferation in other areas of erect shrub tundra may have been mediated by soil moisture. Fine and broad-scale change detection studies show that shrub proliferation has been patchy (Tape et al 2006, Bhatt et al 2010, Fraser et al 2014, and it is possible that this pattern of shrub proliferation is related to spatial variation in soil moisture. Shrub dendrochronology (Myers-Smith et al 2015), plot-based studies (Elmendorf et al 2012b), modelling studies Cairns 2011a, 2011b), and air photo analysis (Tape et al 2006) also suggest that rapid shrub growth preferentially occurs in mesicmoist areas.

Conclusions
(1) Construction and maintenance of the Dempster Highway has intensified the effects of a warming climate on tall shrub growth and proliferation.
(2) Rapid tall shrub proliferation adjacent to the Dempster was facilitated by increases in soil moisture.
(3) Greater tall shrub biomass adjacent to the Dempster shows that disturbances can have a disproportionately large impact on ecosystem configuration relative to the size of the disturbed area.