Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Landscape variation in soil carbon stocks and respiration in an Arctic tundra ecosystem, west Greenland

Landscape variation in soil carbon stocks and respiration in an Arctic tundra ecosystem, west... ARCTIC, ANTARCTIC, AND ALPINE RESEARCH 2018, VOL. 50, NO. 1, e1420283 (17 pages) https://doi.org/10.1080/15230430.2017.1420283 Landscape variation in soil carbon stocks and respiration in an Arctic tundra ecosystem, west Greenland a b Julia I. Bradley-Cook and Ross A. Virginia Department of Biological Sciences, Ecology and Evolutionary Biology Program, Dartmouth College, Hanover, New Hampshire, USA; Environmental Studies Program, Dartmouth College, Hanover, New Hampshire, USA ABSTRACT ARTICLE HISTORY Received 1 February 2017 The magnitude and acceleration of carbon dioxide emissions from warming Arctic tundra soil is Accepted 30 October 2017 an important part of the Region’s influence on the Earth’s climate system. We investigated the links between soil carbon stocks, soil organic matter decomposition, vegetation heterogeneity, KEYWORDS temperature, and environmental sensitivities in dwarf shrub tundra near Kangerlussuaq, Soil organic carbon; Greenland. We quantified carbon stocks of forty-two soil profiles using bulk density estimates landscape heterogeneity; based on previous studies in the region. The soil profiles were located within six vegetation tundra; soil respiration; soil types at nine study sites, distributed across an environmental gradient. We also monitored air temperature and soil temperature and measured in situ soil respiration to quantify variation in carbon flux between vegetation types. For spatial extrapolation, we created a high-resolution land cover classification map of the study area. Aside from a single soil profile taken from a fen soil −2 −2 (54.55 kg C m ;2.13kgNm ), the highest carbon stocks were found in wet grassland soils −2 (mean, 95% CI: 34.87 kg C m , [27.30, 44.55]). These same grassland soils also had the highest mid-growing-season soil respiration rates. Our estimation of soil carbon stocks and mid-grow- ing-season soil respiration measurements indicate that grassland soils are a “hot spot” for soil carbon storage and soil carbon dioxide efflux. Even though shrub, steppe, and mixed vegetation −2 had lower average soil carbon stocks (14.66 – 20.17 kg C m ), these vegetation types played an important role in carbon cycling at the landscape scale because they cover approximately 50 percent of the terrestrial landscape and store approximately 68 percent of the landscape soil organic carbon. The heterogeneous soil carbon stocks in this landscape may be sensitive to key environmental changes, such as shrub expansion and climate change. These environmental drivers could possibly result in a trend toward decreased soil carbon storage and increased release of greenhouse gases into the atmosphere. Introduction predictions to an aggregate response of the ecosystem 6 2 The tundra biome covers 7.5 × 10 km north of the at the landscape scale (Hinzman et al. 2013). Arctic tree line, a region that is undergoing rapid cli- Consideration of soil carbon processes at the land- mate and ecosystem change (Callaghan et al. 2005). The scape level introduces spatial heterogeneity and soils in this high-latitude ecosystem store an estimated dynamics of ecosystem properties (such as soil organic 1,300 Pg of carbon (Hugelius et al. 2014), which is carbon content) along with landscape characteristics approximately twice the carbon contained in the atmo- (e.g., elevation, topography), abiotic conditions (e.g., sphere. Climate and environmental impacts on these moisture), biotic factors (vegetation type), soil forma- soils could affect the global carbon cycle as a result of tion (e.g., time), and associated interactions among microbial release of stored soil carbon through decom- these variables (Jenny 1941). Previous studies on tundra position. Models predict that the response in decom- soils have identified soil temperature, soil moisture, position is based on molecular scale biokinetic disturbance, litter quality, permafrost, and microtopo- properties (Davidson and Janssens 2006; Sierra et al. graphy (Sullivan et al. 2008) as important controls on 2015), but it remains a challenge to link these soil carbon accumulation (Schmidt et al. 2011). CONTACT Julia I. Bradley-Cook julia.i.bradley-cook.gr@dartmouth.edu © 2018 The Author(s). Published by Taylor & Francis. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. e1420283-2 J. I. BRADLEY-COOK AND R. A. VIRGINIA Information about heterogeneity in carbon stocks belowground environmental conditions, such as soil needs to be combined with an understanding of temperature and moisture, which influence microbial variationin soilcarbonquality, temperature sensitiv- activity and soil carbon accumulation (Hudson, ity of decomposition processes, and environmental Henry, and Cornwell 2011;Ostle et al. 2009). controls on carbon cycling in order to best predict Shrub expansion into grassland is a trend that is how the carbon stocks will respond to landscape- widely observed throughout the Arctic (Frost and wide drivers of change. However, it is not well Epstein 2014; Myers-Smith and Hik 2013; Urban et al. understood how variations in soil carbon quality, 2014). Shrub expansion has been observed in west defined as the decomposability of carbon (Bosatta Greenland (Jørgensen, Meilby, and Kollmann 2013), and Ågren 1999), and temperature affect decomposi- but grazing from large herbivores has suppressed tion at the landscape scale. The carbon quality tem- shrub expansion (Post and Pedersen 2008) and slowed perature hypothesis predicts that the temperature the carbon cycling response to warming in the tundra sensitivity of decomposition increases with soil car- near Kangerlussuaq (Cahoon et al. 2011). Changes in bon recalcitrance as long as decomposition is not the extent and abundance of shrubs may also be locally constrained by environmental factors (Davidson and limited by low soil moistures (Myers-Smith et al. 2015). Janssens 2006). In support of the carbon quality Associations between vegetation and soil carbon are temperature hypothesis, Fierer, Colman, and valuable for prediction because, unlike subsurface char- Schimel (2006) found the temperature sensitivity of acteristics, vegetation can be detected remotely using decomposition to increase with soil carbon recalci- aerial and satellite imagery. Few studies have combined trance at sites across the continental United States. estimates of carbon stocks and temperature sensitivities However, conflicting results from a landscape analy- at the landscape scale to understand the landscape-level sis at the Konza Prairie Biological Station, USA, patterns of soil carbon storage and respiration found a wider range of temperature sensitivities and (Horwath Burnham and Sletten 2010; Hugelius and higher maximum sensitivity than observed at the Kuhry 2009) and, to our knowledge, no such study continental scale (Craine et al. 2009). Spatial hetero- has been undertaken in the Kangerlussuaq area in geneity in abiotic controls, carbon accumulation, and west Greenland. thermal sensitivities of decomposition should affect The objective of this study is to link landscape- carbon storage and response to environmental dri- level variation in temperature and vegetation cover to vers, such as climate change. Therefore, we need to soil carbon stocks and sensitivities to environmental understand landscape-level distribution of soil carbon change. We characterize variation in (1) soil tem- and variation in temperature sensitivity of decompo- perature, (2) soil carbon storage and soil chemistry, sition to improve spatially explicit predictions of the and (3) soil respiration by vegetation type across a effect of climate and environmental change on soil climate gradient in the tundra landscape near organic carbon (SOC) pools. Kangerlussuaq, Greenland. We hypothesized that Vegetation types also affect soil carbon storage soil temperature, soil carbon storage and soil chem- through organic matter input (i.e., litter, root exuda- istry, and soil respiration would vary by vegetation tion and turnover) and mediation of the below- type. Furthermore, we hypothesized that soil tem- ground environment. These belowground perature, carbon storage, and respiration would interactions define the stability of soil carbon and increase with distance from the Greenland Ice the thermal sensitivity of soil decomposition. Plant Sheet. By applying the results to a spatially explicit species and functional group can influence the quan- model, we aimed to identify the role of different land tity and quality of SOC (Creamer et al. 2011; cover classes in landscape-level soil carbon storage Hollingsworth et al. 2008;Ostle et al. 2009). For and carbon dioxide emissions. example, Arctic shrubs produce lignaceous biomass that tends to have high C:N mass ratios that, when compared to herbaceous species, result in a less Methods decomposable, lower quality resource for soil micro- Study Area bial communities (Chapin et al. 1996; Hooper and Vitousek 1998). Numerous studies in the Arctic have We conducted fieldwork at nine sites in the shrub shown that decomposition decreases with increasing tundra landscape near Kangerlussuaq, Greenland organic matter C:N ratios (Haddix et al. 2011; (Figure 1). The landscape was deglaciated approxi- Hobbie 1996; Thomsen et al. 2008). Vegetation func- mately 7,000 years ago (Levy et al. 2012), and is tional groups can also have distinct impacts on located at the margin of the current extent of the ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-3 Zone 1 Z1S2 Z1S1 Z1S3 Zone 2 Z2S1 Z2S2 Z2S3 Greenland Ice Sheet Zone 3 Z3S3 Z3S2 Z3S1 Figure 1. Map of the study area, with sites indicated over a WorldView2 satellite image from July 10, 2010. Inset map marks the study area (black rectangle) in west Greenland. Greenland Ice Sheet (GrIS). The landscape is covered Land Cover Classification by aeolian silt deposits above bedrock and glacial till (Dijkmans and Törnqvist 1991). Soils are humus-poor To create a land cover classification map of the study arctic brown soil in the soil order Gelisol (Jones et al. area near Kangerlussuaq, we conducted a multistage 2009). Soil erosion features, termed deflation patches, unsupervised classification of a WorldView2 image which are likely a result of strong winds from the (multispectral, 1.34 m resolution) taken on July 10, GrIS, are common on the landscape and are visually 2010. In ENVI Software (Harris Geospatial Solutions) distinct areas with low vegetative cover and produc- we ran a ISODATA land cover classification to obtain tivity (Heindel, Chipman, and Virginia 2015). In addi- twenty spectral classes with a 5 percent change thresh- tion to geophysical controls, climate and vegetation old and a maximum of twenty iterations. Drawing on shifts are likely to be key environmental drivers of field knowledge accumulated during the summers of biogeochemical cycling in this and other tundra eco- 2010–2013 and visual inspection of the satellite ima- systems. From 1973 to 1999, the average annual atmo- gery, we visually interpreted the output spectral classes spheric temperature observed at Kangerlussuaq was to determine land cover classes based on vegetation −5.7°C (DMI 2017). According to regional climate functional group. Six spectral classes contained pixels projections, by the years 2021–2050, atmospheric tem- of multiple land cover types, so we built a mask for perature is projected to increase by 2°C in summer each mixed class with Spatial Modeler in ERDAS and autumn seasons from historical seasonal averages Imagine to isolate the mixed classes. We then ran of 9.2°C and −4.9°C, respectively. Winter is projected ISODATA on each masked class with a maximum of to increase by 3°C from a historical mean of −19.2°C, eight classes and fifteen iterations. For classes that were while a 4°C increase from a historical mean of −7.7°C split between water and land after the second stage of is projected for the spring (DMI [Danish unsupervised classification, we chose to preserve the Meterological Institute] 2017; Stendel et al. 2007). integrity of terrestrial land cover. We merged the initial Annual precipitation in Kangerlussuaq is approxi- classified and the masked images using Spatial Modeler mately 250 mm (Mernild et al. 2015) and is projected and simplified the images into nine main classes: shrub, to increase 15 percent by 2021–2050 and 30–40 per- steppe, grassland, mixed vegetation, fen, eroded soil cent by 2051–2080 (Stendel et al. 2007). (ES), water, ice, and cloud (Table 1). The ES class e1420283-4 J. I. BRADLEY-COOK AND R. A. VIRGINIA includes deflation patches and exposed bedrock, but Table 1. Description of cover type used in land classification and field sampling. bedrock is less than 20 percent of unvegetated areas Cover Type Common Species Description (Heindel, Chipman, and Virginia 2015). To focus on Shrub Salix glauca Dominated by dwarf terrestrial classes, we calculated percent cover and total Betula nana shrub species area of shrub, steppe, grassland, mixed vegetation, fen, Rhododendron tomentosum and ES land cover types by multiplying the number of Mixed Betula nana Occurs mainly on damp vegetation Rhododendron soil on gently sloping pixels in each class by the pixel size. groenlandicum hillsides. Composed of a We assessed the accuracy of the land cover classifi- Epetrum nigrum mix of shrub, with forbs and graminoid cation by comparing land cover classes to 288 ground undergrowth control waypoints from field observations. These points Steppe Calamagrostis purpurascens Graminoid dominated were not used during the development of the classifica- Carex supina with scattered forbs, Agrostis mertensii common on south-facing tion. We used the ground control points to calculate an Kobresia myosuroides slopes error matrix and accuracy statistics for the terrestrial Potentilla arenosa classes in the land classification map. Eroded soil (ES) Biological soil crusts Common on ridgelines Dryas integrifolia and south-facing slopes, Silene acaulis where loess soil has been locally removed by wind Site Selection erosion: 10–20 percent vegetation cover* Study sites for field measurements and soil sample col- Grassland Poa pratensis Found in moist lection were selected from the land cover classification Erophorum angustifolum depressions, and are Calamagrostis lapponica dominated by graminoid map and ground observations. Three zones were estab- Carex bigelowii species with intermixed lished according to proximity to the margin of the GrIS, Campanula gieseckiana forbs Cerastium alpinum where Zone 1 borders the ice edge and Zone 3 is the Ranunculus hyperboreus farthest away, extending toward the fjord Kangerlussuaq Fen Carex sp. Found along the edge of (Figure 1). Within each zone we identified five to seven Eriophorum sp. streams and lakes, and Hippuris vulgaris includes shallow possible sites that each contained representative land lakebeds with ephemeral cover types (shrub, mixed vegetation, steppe, and ES). surface water Three sites were randomly selected from candidate sites. *Heindel, Chipman, and Virginia (2015). Grassland was included when present (a total of five sites). Fen samples were collected at one site, Zone 1 Site 2 (Z1S2). We measured air and soil temperature, Table 2. Sample collection and measurement across the study conducted vegetation surveys, and collected soil samples sites. Vegetation types are shrub (SH), steppe (ST), grasslands (GL), mixed vegetation (MX), eroded soil (ES), and fen (FN). at all sites, and took in situ soil respiration measurements No. Soil at seven sites (Table 2). Vegetation Temperature Surveys The terrestrial land cover classes used in the land Loggers by and Soil In Situ Soil Vegetation Class classification and sampling design (shrub, mixed vege- Air Samples Respiration Zone Site Temperature SH ST MX ES (2011) (2012) tation, steppe, grassland, fen, and ES; Table 1) are 1 Z1S1 Yes 3 2 3 3 SH, ST, GL, July 12 comparable to a Landsat-based vegetation classification MX, ES, FN created to assess caribou habitat (Tamstorf, Aastrup, Z1S2 Yes 2 1 3 3 SH, ST, MX, July 12 ES and Cuyler 2005). Z1S3 Yes 3 2 3 2 SH, ST, GL, N/A MX, ES 2 Z2S1 No 2 2 3 2 SH, ST, GL, July 13 Air and Soil Temperature MX, ES Z2S2 Yes 2 2 1 3 SH, ST, GL, July 13 MX, ES At each site, we monitored air and soil temperature Z2S3 Yes 1 2 1 3 SH, ST, GL, July 13 every four hours using Thermocron iButton loggers MX, ES 3 Z3S1 Yes 3 1 2 3 SH, ST, GL, N/A (Model DS 1921G, Embedded Data Systems®). To mea- MX, ES sure air temperature at each site and capture air tem- Z3S2 Yes 3 2 3 3 SH, ST, MX, July 14 ES perature patterns across the study area, iButton loggers Z3S3 Yes 3 2 2 2 SH, ST, MX, July 14 were installed in PVC capsules with a drilled hole to ES enable air exchange, attached to rebar at 30 cm height, and shaded under an aluminum roof in an area without types between July 17, 2011, and June 12, 2012. shrub cover from July 11, 2011, to August 22, 2012. Soil Within each vegetation type, we identified three soil temperature loggers were buried at 5 cm depth within temperature locations by randomly selecting a direction steppe, shrub, mixed vegetation, and ES land cover ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-5 (degrees from true north) and a number of steps from using standard methods (Sollins et al. 1999). the center point of a continuous patch of land cover. If Hydrochloric acid was added to samples from mineral a patch was smaller than approximately 8 m in dia- soils in vegetated areas to remove inorganic carbonates. meter, we distributed the temperature loggers among No reaction to the hydrochloric acid was visually more than one patch. We placed a total of twelve observed. loggers per site, and 108 across the entire study area. The data from twenty-five loggers were not included in In Situ Soil Respiration Measurements analyses because the loggers were either disturbed by wildlife or could not be located, but all vegetation types We measured in situ soil respiration (CO flux) with a at each site had at least one logger with a complete data portable Li-Cor 8100 (Lincoln, NE) infrared gas analy- record (Table 2). All temperature loggers were wrapped zer (IRGA) with a 20 cm survey chamber attached. At in parafilm and neoprene plastic for waterproofing. seven of the nine field sites, we installed three 20 cm diameter PVC collars at randomized locations within shrub, steppe, grassland, mixed vegetation, and ES Vegetation Surveys and Soil Sampling vegetation types. Following installation, collars sat for At each site, we identified a soil pit location at the at least twenty minutes to minimize the effect of phy- center of each vegetation type. We avoided vegetation sical disturbance on CO diffusion across the soil sur- boundaries to maximize the likelihood of collecting face. This time interval was selected based on sampling soils that have a long-lived association with the vegeta- logistics and a limited number of PVC collars that tion type of interest. Prior to disturbing the soil surface, precluded long-term set up in the sampling required we conducted a vegetation survey within a 0.5 m for this particular study. We measured the height of the quadrat at each soil pit location. We visually estimated collar above ground to calculate the volume of the percent cover of shrub, herbaceous, graminoid vegeta- headspace in each PVC ring. In a random order, we tion using a 0.5 m quadrant. recorded the CO flux with a two-minute observation We collected soil samples from a 50 cm soil profile after a twenty-second pre-purge. We collected a total of using visual classification to sample detectable hori- 106 measurements during a rainless three-day sampling zons. At some sites, the active layer was shallower period (July 12–14, 2012). Data were not collected at than 50 cm, so we sampled soil to the depth of frozen two sites because of logistical challenges and limited ground. We collected at least 75 g of soil from each field time: Z1S3 was too remote to access during the depth interval using a spoon that was cleaned between survey period, and Z3S1 was an outlier in elevation samples to minimize contamination, and stored sam- (354 m vs. 253 m and 264 m for the other Zone 3 sites). ples in separate sterile Whirl-pak® bags (Nasco, Fort Atkinson, WI). Soil samples were frozen and shipped Calculations and Data Analysis to the Environmental Measurements Lab, Dartmouth College (Hanover, NH) where they were kept at −20°C Temperature Analyses until processing. To test our assumption that air temperature increases extending away from the ice sheet, we compared mean annual temperature, growing season temperature, and win- Soil Analyses ter temperature between zones using a multivariate analysis Samples were thawed and sieved to isolate the less than of variance (MANOVA) followed by univariate analysis to 2 mm fraction of each soil horizon for laboratory test differences within each dependent variable. The annual analysis. Soil water content was estimated from a 10 g average was calculated from data recorded between July 11, soil sample that was dried at 95°C for 24 h, reweighed, 2011, and July 10, 2012. We defined growing season as the −1 and calculated as g water g dry soil. Soil pH was dates between leaf out and senescence, May 22 to August 7 measured using a 1:2 solution of soil:di-H O using a (Post and Forchhammer 2008;Postand Pedersen 2008), pH meter (Thermo Scientific, Orion 3 Star A111 pH and winter season as the cold season climate window, Benchtop, Waltham, MA). Electrical conductivity was November 28 to March 27 (Weatherspark 2015). The log- measured using a 1:5 solution of soil:di-H O using a ger foronesite(Z2S1)disappearedduring thewinter conductivity meter (Thermo Scientific, Orion 3 Star months, so the site was not included in air temperature Conductivity Benchtop, Waltham, MA). Carbon and comparisons. nitrogen content was measured on soils ground with a We compared the soil temperature environment mortar and pestle using a Carlo Erba NA-1500 elemen- using thermal sum for each logger during the measure- tal analyzer (Carlo Erba Instruments, Milan, Italy) ment period. Thermal sum is the difference between e1420283-6 J. I. BRADLEY-COOK AND R. A. VIRGINIA the recorded temperature above the baseline tempera- density to soil horizon and depth based on measure- ture of 0°C and the baseline, summed for the number ments from other studies that we conducted in the of days in the measurement period. We conducted a same area and vegetation types, with bulk densities of −3 model comparison of linear mixed effect models to test 0.25 g DW cm for organic soils (Bradley-Cook and −3 land cover type, zone, and the interaction between the Virginia 2016), 0.62 g DW cm for shallow mineral −3 two as predictor of soil thermal sum, with site as a soils (<10 cm depth), and 1.37 g DW cm for deeper random variable in the model. A second model com- mineral soils (10–20 cm; Petrenko et al. 2016). We parison was conducted to test land cover type, zone, estimated soil nitrogen pools with identical calculations and their interaction as predictors of thermal sum of from percent N. We calculated terrestrial soil carbon soils in vegetated land cover types, excluding the soil and nitrogen inventories at the landscape scale by mul- land cover type. We used Tukey’s HSD (α = 0.05) to tiplying soil carbon content by area in the land cover evaluate differences between means, and calculated classification map for each terrestrial class. marginal R , a measure of variance for mixed effect We used linear mixed effects models to test zone and models. vegetation type as predictor of soil carbon stocks of the full soil profile. We identified significant differences Soil Chemistry Ordination between vegetation types using Tukey’s HSD test in We conducted multivariate analysis to test whether soil the multcomp package in R. Carbon and nitrogen chemistry differed by vegetation type and zone. areal stocks were log-transformed to meet assumptions Measures of percent organic C, percent N, C:N, pH, of normality and homoscedasticity. We back-trans- EC, and soil water content were used in a partial formed the mean and the 95 percent confidence inter- redundancy analysis (RDA) to determine if these vari- val values (Hanlon and Larget 2011). ables differed by vegetation type and zone (vegan pack- age in R [Oksanen et al. 2015]). RDA provides a quantitative method of testing hypotheses in multidi- Results mensional datasets. Each soil sample is assigned scores Land Cover Classification on constrained axes of the predictor variables and unconstrained axes to account for the remaining var- The land cover classification contains nine land cover iance. Soil chemistry measures were used as response classes: shrub, steppe, grassland, mixed vegetation, variables. Vegetation type and zone were used as pre- eroded soil (ES), fen, water, fluvial sediment (outwash) dictor variables, with depth as a covariate in the model. and ice (Figure 2). Ground-based vegetation surveys at Eight soil samples were not included in the analysis 288 points reveal that the classification has an overall because the sample did not contain enough soil mass accuracy of 50 percent. Producer’s accuracy, which to conduct a full set of soil chemistry measurements. provides the probability that a pixel in the classification We analyzed a total of 239 soil samples. The response corresponds with the correct vegetation type, was as variables were standardized using the scale function to high as 84 percent in the ES class and as low as 17 reduce the influence of the magnitude of model vari- percent in the fen class (Table 3). User’s accuracy, or ables on the association between samples. A permuta- the probability that the cover type at a single point tions test with 5,000 permutations was used to corresponds with the land cover class on the map, determine if vegetation type and zone explained a sig- ranged from 20 percent for fen to 60 percent for grass- nificant portion of the variance in soil chemistry land. The most dominant land cover type of the terres- between samples. Adjusted R was calculated to parti- trial landscape was steppe (25%), and is closely followed tion variance between the explanatory and covariate by mixed vegetation (22%), ES (22%), and shrub (19%; variables (Borcard, Gillet, and Legendre 2011). Table 4). The least common land cover types were grassland (7%) and fen (5%; Table 4). Soil Carbon Pools and Landscape Storage We estimated the organic carbon pool for near-surface Air Temperature soil (0–20 cm depth) and the full active layer profile up to 50 cm depth (Equation 1): There was a statistically significant difference in air tem- 2 4 2 2 perature regime between the three zones (F =7.56, 1,6 SOC kg C m ¼%C  D  d  10 cm m (1) P =0.041; Wilk’s Λ = 0.1508) for the full measurement 3 1 10 kg g period, July 11, 2011, to August 22, 2012. Growing season where % C is SOC concentration, D is bulk density (g temperatures were significantly different by zone −3 soil × cm ), and d is depth (cm). We assigned bulk (F =9.335, P = 0.02). Zone 1 had a lower growing 1,6 ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-7 Greenland Ice Sheet Shrub Mixed Veg Outwash Steppe Eroded Soil Ice Grassland Fen Water Figure 2. Land cover classification map of the study area. Colors coincide with land cover classes, including the following vegetation classes: shrub, steppe, grassland, mixed vegetation, eroded soil, and fen. days, °Cd) than Zone 3 (mean = 1338.8 °Cd; Tukey’sHSD Table 3. Error matrix comparing land cover classes from satellite classification with ground-based vegetation observations. P = 0.0585; Figure 3). Thermal sum at Zone 2 Vegetation classes are shrub (SH), steppe (ST), grassland (GL), (mean = 1274.0 °Cd) did not differ from Zone 1 mixed vegetation (MX), eroded soil (ES), fen (FN), and water (W). (P = 0.191) or Zone 3 (P =0.770; Figure 3). Ground Control Data SH ST GL MX ES F W Row Total Classification by SH 21 7 1 17 3 0 0 49 Soil Temperature satellite image ST 12 34 28 8 1 0 65 GL 5 17 6 40 0 0 32 MX 21 8 1 35 33 0 71 Land cover type and zone explained a significant amount ES 2 7 0 0 47 00 56 of the variance in soil thermal sum (Table 5). ES had FN 4 0 0 0 1 1 06 W1 0 0 0 8 0 0 9 higher thermal sums than all other land cover classes at Column total 66 73 10 64 70 5 0 288 all zones (Figure 3). Thermal sums are reduced by more Producer accuracy User accuracy SH 43% SH 32% than 400 degree days in vegetated areas (Figure 3). Mean ST 52% ST 47% thermalsumsofsteppe soilshaveanarrowrange ofvalues GL 19% GL 60% MX 49% MX 55% across all zones (mean = 183°Cd, min = 171°Cd, ES 84% ES 67% max = 205°Cd). Shrub soils have lowest thermal sums FN 17% FN 20% Overall accuracy 50% in Zone 2 (Zone 1 = 298°Cd, Zone 2 = 94°Cd, Zone 3 = 395°Cd). Thermal sum in mixed vegetation decreases moving away from the ice sheet (Zone 1 = 308°Cd, Zone season temperature (mean = 9.0°C) than Zone 3 2 = 153°Cd Zone 3 = 119°Cd). (mean = 11.2°C; Tukey’sHSD P = 0.0586), but not com- pared to Zone 2 (mean = 10.9°C; Tukey’sHSD P =0.135). Variation in Soil Organic Carbon, Nitrogen, and Growing season temperatures of Zone 2 and Zone 3 were Chemistry not significantly different (Tukey’sHSD P =0.914). Annual temperatures did not differ by zone The partial RDA examined vegetation and zone as pre- (F =3.727, P = 0.102), with a mean annual air tempera- dictors of soil chemistry measurements. The permuta- 1,6 ture for all sites of −3.1°C. Winter temperatures also did tion test was significant (F = 22.763, P =0.001). The 6,231 not differ between zones (F =0.128, P = 0.733), with a first two constrained axes were both significant 1,6 mean temperature of −17.3°C. Thermal sum was signifi- (P < 0.01) and explain 33 percent of the variance, with cantly different by zone (F = 10.377, P = 0.018), with 25.4 percent attributed to RDA1 and 7.6 percent attrib- 1,6 lower thermal sums at Zone 1 (mean = 1084.1 degree uted to RDA2. The remaining constrained axes, RDA3 e1420283-8 J. I. BRADLEY-COOK AND R. A. VIRGINIA Table 4. Area, percent cover, carbon and nitrogen pools for each terrestrial land cover class. Area was extracted from the land cover classification. Pool sizes are mean values. Total C Total C Area 20-cm 50-cm −2 −2 2 Land Cover Class Mean C 20-cm (kg m ) Mean C 50-cm (kg m ) (km (%)) (Gg C (%)) (Gg C (%)) (A) Organic Carbon Shrub 6.44 14.67 16.13 (19%) 103.86 (16%) 236.59 (17%) Steppe 8.98 20.17 21.16 (25%) 189.91 (30%) 426.78 (31%) Grassland 19.07 34.87 5.60 (7%) 106.86 (17%) 195.35 (14%) Mixed veg 6.71 15.13 18.12 (22%) 121.66 (19%) 274.26 (20%) Eroded soil 0.46 0.70 18.63 (22%) 8.48 (1%) 13.02 (1%) Fen 24.36 54.55 4.22 (5%) 102.91 (16%) 230.44 (17%) TOTAL –– 83.86 633.68 1,376.45 (B) Total Nitrogen Shrub 0.32 0.75 16.13 (19%) 5.21 (13%) 12.13 (14%) Steppe 0.59 1.33 21.16 (25%) 12.54 (30%) 28.12 (31%) Grassland 1.30 2.30 5.60 (7%) 7.27 (18%) 12.87 (14%) Mixed veg 0.36 0.82 18.12 (22%) 6.54 (16%) 14.88 (17%) Eroded soil 0.04 0.06 18.63 (22%) 0.80 (2%) 1.12 (1%) Fen 2.13 4.85 4.22 (5%) 9 (22%) 20.49(23%) TOTAL –– 83.86 41.36 89.62 Figure 3. Thermal sums of (A) air and (B) soil temperatures at 5 cm depth in different vegetation types and in three zones. Data points mark site averages with error bars indicating ±1 standard error. Thermal sum of air temperature was calculated for one year (July 12, 2011–July 11, 2012), and soil temperature was calculated between July 17, 2011, and June 12, 2012. Shapes of soil thermal sums indicate vegetation cover (ES = eroded soil, ST = steppe, MX = mixed vegetation, SH = shrub). ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-9 Table 5. Comparison of models explaining soil thermal sums. The preferred model (ΔAICc < 2) contains an interaction between vegetation type and zone. Site was included as a random variable in all models. Marginal R of the best fitting model was 0.8116 (Veg = Vegetation). Rank Predictor Variable k AICc ΔAICc AICc Model Weight Log-likelihood Log(L) 1 Veg + Zone + Veg: Zone 14 1120.4 0.0 1 −543.1 2 Vegetation 6 1132.8 12.4 0 −559.9 3 Vegetation + Zone 8 1134.3 13.9 0 −558.2 4 Intercept 3 1232.7 112.4 0 −613.2 5 Zone 5 1235.8 115.4 0 −612.5 and RDA4, were not significant (P >0.05). The correla- Electrical conductivity varied substantially for all vegeta- tion variable, soil sample depth, explained 5.2 percent of tion types, and did not have a strong association with the variance. The remainder of the variance, 61.4 per- vegetation types or the other chemical measurements. cent, is unconstrained by vegetation type, zone, or depth. The permutation test showed that vegetation Soil C and N Storage by Vegetation Type explains a significant amount of the variation in soil chemistry measurements (F = 26.4, P = 0.0002). Soil carbon stocks varied among land cover types. In the 5,231 Zone did not explain a significant amount of the total prediction of soil carbon stocks, the best-ranked model variance in soil chemistry (F = 1.0, P = 0.38). (ΔAICc < 2) contained vegetation type as the only pre- 1,238 Total organic carbon, nitrogen, and soil water content dictor (Table 7). Grassland soils and the single fen sample were tightly correlated explanatory variables that aligned contained the highest mean carbon stocks (mean, 95% CI: −2 −2 with the RDA1 axis (Figure 4). Grassland and fen samples 34.87 kg C m , [27.30, 44.55] and 54.55 kg C m , had high values along this axis, indicating that the soils have respectively; Figure 5). Steppe (mean, 95% CI: 20.17 kg −2 high carbon, nitrogen, and soil water content. Grassland, Cm , [12.64, 32.18]), mixed vegetation (mean, 95% CI: −2 shrub, and mixed vegetation shared similar position on the 15.13 kg C m , [9.50, 24.11]), and shrub (mean, 95% CI: −2 RDA1 axis. C:N corresponds with RDA2, and shrub, 14.67 kg C m , [9.80, 21.95]) soil carbon storage were not steppe, and mixed vegetation vary along RDA2, with the significantly different from one another (Figure 5). ES highest C:N in shrub soils, the lowest values for steppe soils, areas had the lowest carbon storage (mean, 95% CI: −2 and mixed vegetation in an intermediate position. 0.07 kg C m ,[0.46,1.07]; Figure 5). Neither pH nor electrical conductivity were tightly Mean nitrogen stocks in soils ranged from 0.06 to −2 correlated with the RDA axes (Figure 4). ES soils had a 4.85 kg N m . As in the prediction of soil carbon, vegeta- strong correlation with pH (Figure 4), with higher pH tion type was the only predictor term in the best model of measurements than the other vegetation types (Table 6). soil nitrogen stocks (Table 8). The highest average N stock Figure 4. Ordination biplot of the partial RDA of soil chemistry measurements using vegetation type and zone as the independent variables (Table 6). The centroid of each vegetation type is labeled for shrub (VegSH), mixed vegetation (VegHB), steppe (VegGR), grassland (VegGR2), eroded soil (VegDZ), and fen (VegLA). TOC is total organic carbon, Grav Moist is soil moisture, TN is total nitrogen, EC is electrical conductivity, and PH is pH. The biplot represents 31 percent of the total variance (25.4% on RDA1 and 7.6% on RDA2). e1420283-10 J. I. BRADLEY-COOK AND R. A. VIRGINIA Table 6. Mean and range of soil chemistry measurements collected from six different vegetation types: N is the number of soil −1 samples, %C is organic carbon content, %N is nitrogen content, SWC is soil water content (g H O g Soil ), and EC is electrical −1 conductivity (micro-siemens cm ). Land Cover Class N %C %N C:N pH EC SWC Eroded soil 50 0.2 (0.01–1.0) 0.02 (0–0.08) 13.1 (3.6–45.7) 7.6 (5.8–8.6) 51 (2–1648) 2.9 (0.09–7.05) Steppe 55 5.5 (0.2–21.6) 0.37 (0.01–1.38) 15.2 (8.5–25.3) 6.4 (5.8–7.6) 39 (4–228) 44.6 (2.2–138.5) Grassland 27 10.2 (0.5–27.5) 0.66 (0.04–1.63) 15.4 (12.3–19.8) 6.3 (5.1–7.4) 94 (7–594) 82.3 (3.61–381.4) Mixed vegetation 50 4.6 (0.2–12.9) 0.25 (0.02–0.66) 17.8 (8.5–29.1) 6.4 (5.4–7.2) 26.6 (3–314) 34 (4–80) Fen 5 22.7 (20.3–26.3) 1.99 (1.81–2.26) 11.4 (11.0–11.7) 6.3 (6.1–6.6) 135 (52–323) 189.1 (178–205) Shrub 60 4.0 (0.2–20.5) 0.21 (0.01–0.99) 19.7 (6.3–51.4) 6.3 (4.9–7.6) 25 (2–129) 23.4 (3.84–71.7) Table 7. Comparison of models explaining soil organic carbon (e.g., Z1S1-GL [grassland], Z3S2-ST [steppe], Z3S3-MX stocks. The top-ranked model, with vegetation as a main effect, [mixed vegetation], and Z3S3-ST [steppe]; Figure 7). is preferred. All other models, ΔAICc > 2. Steppe and mixed vegetation comprise the largest AICc Log- fraction of total landscape carbon (31% and 20%, Model likelihood Rank Predictor Variable k AICc ΔAICc Weight Log(L) respectively; Table 4). Fen and grassland make up 17 1 Vegetation 7 276.97 0 0.80 −129.8 percent and 14 percent, respectively, even though they 2 Vegetation + Zone 8 297.75 2.78 0.20 −129.7 cover the smallest fraction of the landscape area (5% 3 Veg + Zone + Veg: 12 286.90 9.93 0.01 −129.1 Zone and 7%, respectively; Table 4). An estimated 17 percent 4 Intercept 2 297.93 21.01 0.00 −146.8 of landscape carbon is stored in shrub soils. ES areas, 5 Zone 3 298.72 21.75 0.00 −146.0 which cover 22 percent of the landscape area, only store 1 percent of the total carbon in the landscape (Table 4). −2 was in the single fen soil profile (4.85 kg N m ; Figure 5), Nitrogen stocks in the terrestrial landscape are −2 and grasslands also had elevated N stocks (mean, 95% CI: 1.07 Gg N m in the full soil profile. Near-surface −2 2.30 kg N m , [1.60, 3.30]). Steppe (mean, 95% CI: 1.33 kg soils store 46 percent of total nitrogen, with an average −2 −2 Nm ,[0.76,2.32]), mixedvegetation(mean,95% CI: of 0.49 Gg N m in the top 20 cm. The largest nitrogen −2 0.82 kg N m , [0.50, 1.34]), and shrub (mean, 95% CI: stock is in steppe (31%), which is more than double −2 0.75 kg N m , [0.47, 1.19]) were not significantly different that of shrub (14%; Table 4). Fen soils store 23 percent (Figure 5). ES areas had the lowest nitrogen stocks of all of landscape nitrogen, which is disproportionately high −2 land cover types (mean, 95% CI: 0.06 kg N m , [0.04, when considering it covers only 5 percent of the land- 0.09]; Figure 5). scape (Table 4). ES only stores 1 percent of the total landscape nitrogen. Soil Carbon Respiration by Vegetation Type Discussion Soil carbon respiration rates during the mid-season dif- fered by vegetation type (Figure 6). The highest respiration Landscape Temperature Variation was observed in grassland vegetation, with an average rate Temperature regimes vary with proximity to the ice −2 −1 of 7.10 ± 0.60 μmols CO m s , which was significantly sheet. Within the study area, which captured a distance greater than all other vegetation types (P <0.05).Average of approximately 11 km from the GrIS margin, the soil respiration for steppe, shrub, and mixed vegetation influence of the GrIS on atmospheric temperatures is −2 −1 ranged from 3.11 to 3.42 μmols CO m s ,but did apparent during summer months, but does not have an not significantly differ from each other. In ES, average effect during the winter or at an annual scale. The 2°C respiration rate was an order of magnitude lower than cooler growing season temperatures near the margin of −2 −1 grasslands (0.69 ± 0.45 μmols CO m s ). the GrIS can have ecological consequences, including altered carbon cycling. Lower temperatures decrease net primary productivity and microbial activity, result- Landscape Soil Carbon and Nitrogen Estimates ing in a slower cycling of carbon (Chapin et al. 2009). The terrestrial landscape stores an average of 16.41 kg Regional climate models project that warming will be −2 Cm in the active layer soils to a depth of 50 cm. greatest during the winter season, and will be limited to Approximately 46 percent of this carbon is stored in approximately 2°C warming during the summer months the 0–20 cm depth increment, with an average of 7.4 kg because of a cooling effect from the GrIS (Stendel et al. −2 Cm for the top 20 cm of soil. Several soil profiles had 2007). Since proximity to the ice sheet does not affect elevated carbon content in the deepest sample collected winter temperature within the study area, we infer that ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-11 Figure 5. Soil (A) organic carbon and (B) nitrogen pools to a depth of 50 cm for soils from six vegetated land cover types in western Greenland (ES = eroded soil, ST = steppe, GL = grasslands, MX = mixed vegetation, FN = fen, SH = shrub). Horizontal tabs mark the mean of each vegetation type, with error bars that represent ±1 standard error of the mean. Gray data points are the carbon stocks from each soil pit. Different letters indicate significant differences in Tukey HSD post hoc test (P < 0.05). projected winter warming of 3°C by 2050 (Stendel et al. increase in temperature with distance away from the ice 2007) will be experienced equally across the landscape. sheet. The largest differences in thermal sums were The interactive effect of zone and land cover type on between ES and vegetated areas. Differences within vege- soil temperature supported our hypothesis that vegetation tation types were smaller. These results indicate that soil is an important mediator of the belowground abiotic temperature is associated with vegetation, and that vege- conditions. When vegetation is present, soil and atmo- tation could mediate the impact of climate change on spheric temperatures are decoupled to an extent. In the belowground abiotic conditions and biological response. minimally vegetated ES, the soil temperature is more Other studies have found vegetation to be an important tightly linked to air temperature as observed by the determinant of soil temperature (Aalto, Le Roux, and e1420283-12 J. I. BRADLEY-COOK AND R. A. VIRGINIA Table 8. Comparison of models explaining soil nitrogen stocks. likely because these features are commonly confined to The top-ranked model, with vegetation as a main effect, is small areas on the landscape and therefore have a higher preferred. All other models, ΔAICc > 2. likelihood of accumulating error in the accuracy assess- AICc Log- ment process. Improvements in the accuracy metrics of Model likelihood Rank Predictor Variable k AICc ΔAICc Weight Log(L) our classification might be gained by the addition of 1 Vegetation 7 54.38 0 0.75 −18.5 randomly generated ground control data points that are 2 Vegetation + Zone 8 56.68 2.31 0.24 −18.2 at least three meters away from a vegetation boundary. 3 Veg + Zone + Veg:Zone 12 62.08 7.71 0.02 −13.7 4 Zone 3 78.66 24.28 0.00 −36.0 The buffer around the data point should reduce the influ- 5 Intercept 2 78.83 24.45 0.00 −37.3 ence of instrument error. Heindel, Chipman, and Virginia (2015) also conducted a land cover classification of an overlapping, but larger, study area. Their land cover composition estimates were Luoto 2013; Migala et al. 2014) that can alter heat comparable to ours. We attribute discrepancies in direct exchange from warming (Hollister et al. 2006), which comparison of our estimates to the fact that their study adds uncertainty to modeling soil organic carbon area extended farther west than ours, thus included a response to climate change (Xiong et al. 2015). greater proportion of land where wind erosion features are less common. These landscape classification maps are useful tools for the scaling of ecological variables such as Heterogeneity in Land Cover Class soil carbon and nitrogen stocks and are reflective of the natural variability in the landscape. The high heterogeneity of land cover types at small spatial scales would decrease the overall accuracy of the land cover classification. Our ground control points were Soil Chemistry Variation within the nine field sites, which were chosen because they had representative land cover classes within close Multivariate analysis of the soil chemistry revealed that proximity to each other. As a result, many of the vegeta- carbon, nitrogen, and soil moisture are the constrained tion patches were small (size ranged from 3 m to 40 m in variables associated with most of the variation among diameter) and close to boundaries with other vegetation soil samples. Soil C, N, and moisture are highly corre- types. In these conditions, error from handheld GPS units lated in other arctic terrestrial systems (e.g., combined with any small georeferencing errors can easily Hollingsworth et al. 2008) and are likely a result of a propagate and result in a misalignment of ground control positive feedback between soil conditions and primary points and the land cover classification and reduced accu- productivity. Additionally, soil moisture can constrain racy metrics. The ES class, with a distinct spectral signal microbial decomposition when soils are near saturation from vegetation, had a producer accuracy of 84 percent, as a result of limited oxygen availability in the soil which we consider to be fairly high given the possibility of matrix. Grasslands and fen soils have high values of error propagation mentioned above. Grassland and fen C, N, and moisture. Meanwhile, steppe, shrub, and vegetation had the lowest producer accuracy (Table 3), mixed vegetation, which account for a majority of the Figure 6. Average soil respiration rates in five vegetation types during the growing season. Error bars are ±1 standard error around the mean. Different letters indicate significant difference in Tukey HSD post hoc test (P < 0.05). ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-13 Figure 7. Soil organic carbon content (%) for each soil profile. Symbols correspond to the vegetation types (ES = eroded soil, ST = steppe, GL = grasslands, MX = mixed vegetation, FN = fen, SH = shrub). Profiles are grouped by sites in each panel, with Zone 1 sites in the top row, Zone 2 sites in the middle row, and Zone 3 sites in the bottom row. Distribution of Soil Organic Carbon at a Landscape land cover, have similar position relative to the axis Scale describing the variation in C, N, and moisture, but are differentiated on the second axis, which is asso- The landscape-wide average carbon stock weighted by −2 ciated with the C:N ratio. These results support our area is 16.41 kg C m , which is within the estimate hypothesis and other studies that shrub soils have assigned to west Greenland in the regional assessments −2 higher C:N than graminoid-dominated soils, and the of soil organic carbon, 10–25 kg C m (Hugelius et al. mixed vegetation ratio reflects a mixture of C:N inputs 2013). This average carbon stock is greater than that of and places these soils at an intermediate position. high Arctic field sites near Thule, Greenland, where Higher C:N in shrub soils has also been found in striped patterned ground soil contains SOC content of −2 mineral soils in the study area (Petrenko et al. 2016). 9.4 kg C m (Horwath et al. 2008). Other prominent The same shrub soils also have reduced soil respiration Arctic field areas, such as Toolik Lake Field Station on potential and temperature sensitivity compared to adja- the north slope of Alaska and Svalbard, have higher −2 cent grassland soils, which may be a result of nutrient SOC stocks that fall in the 25–50 kg C m range limitation on decomposition from lower nitrogen (Hugelius et al. 2013). stocks in shrub soils (Bradley-Cook et al. 2016). ES Soil carbon content varied across the landscape, and have the lowest position on the axis associated with C, has statistically significant associations with land cover N, and moisture (RCA1), which indicate that these soils type. The highest carbon stocks of the vegetation classes have limited biological development. These soils also sampled at multiple sites are associated with grasslands, −2 have a higher pH than other vegetation types, high- with an average of 34.87 kg C m . Petrenko et al. lighting that they have a unique chemical profile that (2016) estimated mineral soil carbon stocks from grass- −2 could further limit enzymatic activity and microbial land soils to be 29 kg C m , which is probably slightly decomposition (Min et al. 2014). below our estimate because they focused on mineral e1420283-14 J. I. BRADLEY-COOK AND R. A. VIRGINIA soil C dynamics and did not include the surface organic vegetation have lower carbon stocks per unit area (i.e., kg −2 horizon in their analysis. Grassland vegetation types Cm values) than fen and grassland, they collectively store only comprise 7 percent of the total study area, but it most of the total soil organic carbon (i.e., kg C values) contains approximately 15 percent of the landscape soil because they comprise a majority of the total vegetation carbon stock. The elevated in situ respiration rates from cover. Thus, shrub, steppe, and mixed vegetation are bio- grassland soils correspond with laboratory measure- geochemically important partitions of the Kangerlussuaq ments of carbon mineralization in which grassland tundra landscape. soils have a 1.8 times higher soil respiration than shrub soils (Bradley-Cook et al. 2016). Together, these Landscape Carbon Storage Response to findings indicate that, in this landscape, grasslands are Environmental Change hot spots for mid-summer soil respiration and for soil carbon storage. The pattern in soil carbon storage sug- To investigate the possible trajectory of soil carbon stocks in gests that soil carbon inputs from plant production response to vegetation and climate drivers at the landscape have exceeded soil carbon losses from respiration over scale, we considered the impact of shrub expansion and a long time frame. Soil respiration rates, however, cap- climate warming scenarios on landscape soil carbon. Given ture instantaneous dynamics, so the strength of the the assumption that a vegetation shift is accompanied by a carbon sink may be different than in the past. It is corresponding shift in soil properties, shrub expansion and possible that recent atmospheric warming has elevated prolonged occupation into graminoid-dominated areas carbon losses relative to uptake and storage. (steppe and grassland) should result in a shift from the The single fen soil profile measured had a soil carbon high carbon, low C:N soils associated with grassland vege- −2 stock of 54.6 kg C m , which is the highest stock measured tation to the lower carbon, high C:N soils associated with ofallsoilpitsinthisstudy.Andersonetal. (2009)found shrub vegetation. One possible mechanism for this poten- lake sediments in the Kangerlussuaq region to contain large tial shift is that shrub expansion could increase winter soil −2 carbon stocks (an average of 42 kg C m ) and suggest that respiration and soil nitrogen mobilization as predicted by these low-lying, mostly saturated soils are the most carbon- the snow-shrub interaction hypothesis (Sturm, Holmgren, rich areas in the landscape. The high carbon content of and McFadden 2001;Sturm et al. 2005). Under this sce- these soils, combined with seasonally dynamic hydrology, nario, a shift to greater shrubiness would reduce the land- indicates that soils such as these could be an important scape soil carbon stock and increase soil C:N. However, it is contributor to the landscape dynamics of the CO and also possible that the general characteristics of graminoid CH exchange between the terrestrial ecosystem and the soils could be preserved under shrub expansion scenarios. atmosphere. Total landscape carbon storage will be a function of the net change in soil and vegetation carbon stocks. While the The most common vegetation types were ranked lowest vegetation carbon stock increases with shrub expansion, with respect to carbon stock per unit area. While the this component of Arctic terrestrial carbon stock is a average carbon stocks of steppe, shrub, and mixed vegeta- small percentage of soil carbon stocks (McGuire et al. tion are at least 40 percent lower than grassland, they 2009), which suggests that the net change will be driven comprise a substantial portion (approximately 68%) of by soil carbon response to climate change. Our data show total landscape storage because of their areal extent. that growing season soil respiration and soil moisture are Estimates of near-surface shrub carbon stocks (Table 5) lowerinshrub soils. If thesefactors constrainsoilcarbon are within the range of comparable measurements in the −2 efflux we could expect to see a legacy effect of graminoid study area, 3.07–7.67 kg C m (Bradley-Cook and Virginia soils that could contribute to SOC heterogeneity associated 2016) and to shrub tundra stocks measured in Siberia, 7.1– −2 with shrub vegetation under projected vegetation change 7.8 kg C m (Hugelius and Kuhry 2009). Many studies in scenarios. Arctic tundra systems, which were conducted in low-lying The carbon losses associated with shrub expansion could tundra, have focused on fen and bog hot spots instead of be compounded with a warming climate. Microbial decom- soils with lower carbon stocks, because these features are position most often increases with temperature, and the common and comprise a majority of the carbon stocks (e.g., rate of increase can vary between soils. Previous research on Hugelius and Kuhry 2009). However, in systems such as soils in this study area shows that the temperature sensitiv- Kangerlussuaq, characterized by steep terrain and semiarid ity of decomposition is higher in grasslands than shrub soils conditions, low-lying grassland and fen vegetation cover a (Q-10 =2.3, Q-10 = 1.8), and that soil moisture relatively small proportion of the area. From our landscape- grasslands shrub increases respiration in grassland soils but not shrub soils level analysis of total soil carbon storage by vegetation type, (Bradley-Cook et al. 2016). These known temperature we found that even though shrub, steppe, and mixed ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-15 sensitivities suggest that SOC decomposition rates will temperature sensitive than in grassland soils. accelerate with warming for both shrub and grassland However, vegetation mediation of air temperature soils, so increased soil respiration can be expected under adds uncertainty to the effect of climate change on any vegetation scenario. belowground temperature affects on carbon cycling. Climate change is a key driver of shrub expansion in These landscape-level carbon storage, turnover, and tundra ecosystems (Myers-Smith et al. 2011), so it is sensitivities indicate that climate drivers and vegetation likely that atmospheric warming and shrub expansion dynamics will likely lead to a loss of stored soil carbon will co-occur. However, as our findings indicate, vege- and an increase of greenhouse gases flux into the tation mediates the relationship between air and soil atmosphere. temperature, so atmospheric warming may not result in a concomitant increase in soil temperature. For Acknowledgments instance, shrub canopy shading can have a local cooling effect on soil temperatures (Jean and Payette 2014; We are deeply grateful to Courtney Hammond Wagner and Sturm, Holmgren, and McFadden 2001). We did not Ruth Heindel for superb field assistance, to Leehi Yona for observe differences in soil temperature between shrub helping with sample processing and analysis, and to Paul Zietz for laboratory support. Furthermore, we thank Amy Burzynski and graminoid vegetation, so it is possible that canopy and Jonathan Chipman for contributions to the mapping and shading does not have a sufficient effect on insulation spatial analysis, Thomas Kraft for guidance on the ordination to modify soil temperature in the low shrub tundra of analysis, Ruth Heindel for providing comments on the manu- this study area (Hollingsworth et al. 2008). It is also script, and all Virginia Lab members for moral support. We possible that other factors, such as aspect, insulation are thankful to CH2M Hill Polar Services for unfaltering logistical support in Kangerlussuaq that enabled productive from understory moss, or plant structural feature such fieldwork. This research was funded by an NSF grant (Award as leaf area index, have a stronger effect on soil tem- #: 0801490) to Ross Virginia, with additional support from the perature than small changes in solar insolation. In Institute of Arctic Studies, in the Dickey Center for summary, we posit that landscape soil carbon storage International Understanding at Dartmouth. is sensitive to both vegetation change and climate warming, with possible compounding impacts on soil carbon storage in a future in which these changes may References occur concurrently. Aalto, J., P. C. Le Roux, and M. Luoto. 2013. Vegetation mediates soil temperature and moisture in Arctic-Alpine environments. Arctic, Antarctic, and Alpine Research Conclusions 45:429–39. doi:10.1657/1938-4246-45.4.429. Anderson, N. J., W. D’andrea, and S. C. Fritz. 2009. Holocene The Kangerlussuaq tundra landscape has heteroge- carbon burial by lakes in SW Greenland. Global Change neous land cover and a growing season air temperature Biologic 15:2590–98. doi:10.1111/j.1365-2486.2009.01942.x. gradient extending from the GrIS. We estimate average Borcard, D., F. Gillet, and P. Legendre. 2011. Numerical −2 landscape carbon stock to be 16.41 kg C m and ecology with R, 292. New York: Springer. −2 nitrogen stock to be 1.07 kg N m . We found differ- Bosatta, E., and G. Ågren. 1999. Soil organic matter quality interpreted thermodynamically. Soil Biology & Biochemistry ences in carbon stocks and soil respiration by vegeta- 31:1889–91. tion type, with enhanced carbon storage and soil Bradley-Cook, J. I., C. L. Petrenko, A. J. Friedland, and R. A. respiration in grassland soils. Combined with the Virginia. 2016. Temperature sensitivity of mineral soil understanding that these soils are more sensitive to carbon decomposition in shrub and graminoid tundra, temperature and moisture, we conclude that grasslands west Greenland. Climate Change Responses 1–15. are highly sensitive hot spots of soil carbon storage and doi:10.1186/s40665-016-0016-1. Bradley-Cook, J. I., and R. A. Virginia. 2016. Soil carbon mid-growing season soil respiration within a shrub- storage, respiration potential, and organic matter quality dominated matrix. Even though shrub, steppe, and across an age and climate gradient in southwestern mixed vegetation have lower carbon stocks per unit Greenland. Polar Biology 39:1283–95. doi:10.1007/s00300- area, they are still important components of land- 015-1853-2. scape-level biogeochemical cycling because they cover Cahoon, S. M. P., P. F. Sullivan, E. Post, and J. M. Welker. 2011. Large herbivores limit CO2 uptake and suppress more than 50 percent of the area. Shrub expansion into carbon cycle responses to warming in West Greenland. graminoid-dominated areas could possibly result in Global Change Biologic 18:469–79. doi:10.1111/j.1365- lower landscape soil carbon storage in the long term. 2486.2011.02528.x. Warming could drive compounding losses of stored Callaghan, T. V., L. Bjorn, F. S. Chapin, Y. Chernov, T. carbon, although organic carbon may be more stable Christensen, B. Huntley, R. Ims, M. Johansson, D. Jolly, S. in shrub soils, where decomposition rates are less Jonasson, N. Matveyeva, W. Oechel, N. Panikov, and G. e1420283-16 J. I. BRADLEY-COOK AND R. A. VIRGINIA Shaver. 2005. Arctic tundra and polar desert ecosystems. In Black Spruce Ecosystems. Ecosystems 11:629–42. Arctic climate impact assessment, eds. C. Symon, L. Arris, and doi:10.1007/s10021-008-9147-y. B. Heal, 244–352. New York: Cambridge University Press. Hollister, R. D., P. J. Webber, F. E. Nelson, and C. E. Chapin, F., M. Bret-Harte, S. Hobbie, and H. Zhong. 1996. Tweedie. 2006. Soil thaw and temperature response to air Plant functional types as predictors of transient responses warming varies by plant community: Results from an of arctic vegetation to global change. Journal of Vegetation open-top chamber experiment in northern Alaska. Arctic, Science 7:347–58. Antarctic, and Alpine Research 38:206–15. Chapin, F., J. McFarland, A. D. McGuire, E. S. Euskirchen, R. Hooper, D., and P. Vitousek. 1998. Effects of plant composi- W. Ruess, and K. Kielland. 2009. The changing global tion and diversity on nutrient cycling. Ecological carbon cycle: Linking plant-soil carbon dynamics to global Monographs 68:121–49. consequences. Journal of Ecology 97:840–50. doi:10.1111/ Horwath Burnham, J., and R. Sletten. 2010. Spatial distribution j.1365-2745.2009.01529.x. of soil organic carbon in northwest Greenland and under- Craine, J., R. Spurr, K. Mclauchlan, and N. Fierer. 2009. estimates of high Arctic carbon stores. Global Biogeochemical Landscape-level variation in temperature sensitivity of Cycles 24:GB3012. doi:10.1029/2009GB003660. soil organic carbon decomposition. Soil Biology & Horwath, J. L., R. S. Sletten, B. Hagedorn, and B. Hallet. 2008. Biochemistry 42:373–75. doi:10.1016/j.soilbio.2009.10.024. Spatial and temporal distribution of soil organic carbon in Creamer, C. A., T. R. Filley, T. W. Boutton, S. Oleynik, and I. nonsorted striped patterned ground of the High Arctic. B. Kantola. 2011. Controls on soil carbon accumulation Journal Of Geophysical Research-Biogeosciences 113: during woody plant encroachment: Evidence from physical G03S07. doi:10.1029/2007JG000511. fractionation, soil respiration, and 13-C of respired CO2. Hudson, J. M. G., G. H. R. Henry, and W. K. Cornwell. 2011. Soil Biology and Biochemistry 43:1678–87. doi:10.1016/j. Taller and larger: Shifts in Arctic tundra leaf traits after soilbio.2011.04.013. 16 years of experimental warming. Global Change Biology Davidson, E. A., and I. A. Janssens. 2006. Temperature sensi- 17:1013–21. doi:10.1111/j.1365-2486.2010.02294.x. tivity of soil carbon decomposition and feedbacks to climate Hugelius, G., and P. Kuhry. 2009. Landscape partitioning and change. Nature 440:165–73. doi:10.1038/nature04514. environmental gradient analyses of soil organic carbon in a Dijkmans, J., and T. Törnqvist. 1991. Modern periglacial eolian permafrost environment. Global Biogeochemical Cycles 23: deposits and landforms in the Søndre Strømfjord area, West GB3006. doi:10.1029/2008GB003419. Greenland and their palaeoenvironmental implications. Hugelius, G., J. Strauss, S. Zubrzycki, J. W. Harden, E. A. G. Meddelelser Om Grønland. Geoscience 25:3–39. Schuur, C.-L. Ping, L. Schirrmeister, G. Grosse, G. J. DMI (Danish Meterological Institute). 2017. Klimanormaler for Michaelson, C. D. Koven, J. A. O’Donnell, B. Elberling, Grønland, Copenhagen, Denmark. Accessed June 4, 2017. U. Mishra, P. Camill, Z. Yu, J. Palmtag, and P. Kuhry. http://www.dmi.dk/groenland/arkiver/klimanormaler/. 2014. Estimated stocks of circumpolar permafrost carbon Fierer, N., B. Colman, and J. Schimel. 2006. Predicting the with quantified uncertainty ranges and identified data temperature dependence of microbial respiration in soil: A gaps. Biogeosciences 11:6573–93. doi:10.5194/bg-11-6573- continental-scale analysis. Global Biogeochemical Cycles 20: 2014-supplement. GB3026. doi:10.1029/2005GB002644. Hugelius, G., C. Tarnocai, G. Broll, J. G. Canadell, P. Kuhry, and Frost, G. V., and H. E. Epstein. 2014. Tall shrub and tree D. K. Swanson. 2013. The Northern circumpolar soil carbon expansion in Siberian tundra ecotones since the 1960s. database: Spatially distributed datasets of soil coverage and Global Change Biology 20:1264–77. doi:10.1111/gcb.12406. soil carbon storage in the northern permafrost regions. Earth Haddix, M. L., A. F. Plante, R. T. Conant, J. Six, J. M. Steinweg, Systems Sciences Data 5:3–13. doi:10.5194/essd-5-3-2013. K. Magrini-Bair, R. A. Drijber, S. J. Morris, and E. A. Paul. Jean, M., and S. Payette. 2014. Effect of vegetation cover on 2011. The role of soil characteristics on temperature sensi- the ground thermal regime of wooded and non-wooded tivity of soil organic matter. Soil Science Society Of America palsas. Permafrost and Periglacial Processes 25:281–94. Journal 75:56–68. doi:10.2136/sssaj2010.0118. doi:10.1002/ppp.1817. Hanlon, B., and B. Larget. 2011. Assumptions and transfor- Jenny, H. 1941. Factors of soil formation. New York: Dover mations. Madison, Wisconsin: Department of Statistics. Publications. Accessed May 20, 2017. http://www.stat.wisc.edu/~st571- Jones, A., V. Stolbovoy, C. Tarnocai, G. Broll, O. Spaargaren, 1/11-assumptions-2.pdf. and L. Montarella. 2009. Soil Atlas of the Northern Heindel, R. C., J. W. Chipman, and R. A. Virginia. 2015. The Circumpolar Region, 1. Luxembourg: European spatial distribution and ecological impacts of Aeolian soil Commission, Office for Official Publications of the erosion in Kangerlussuaq, West Greenland. Annals of the European Communities. Association of American Geographers 105:875–90. Jørgensen, R. H., H. Meilby, and J. Kollmann. 2013. Shrub doi:10.1080/00045608.2015.1059176. expansion in SW Greenland under modest regional warm- Hinzman, L. D., C. J. Deal, A. D. McGuire, S. H. Mernild, I. V. ing: Disentangling effects of human disturbance and graz- Polyakov, and J. E. Walsh. 2013. Trajectory of the Arctic as ing. Arctic, Antarctic, and Alpine Research 45:515–25. an integrated system. Ecological Applications 23:1837–68. doi:10.1657/1938-4246-45.4.515. Hobbie, S. 1996. Temperature and plant species control over Levy, L. B., M. A. Kelly, J. A. Howley, and R. A. Virginia. litter decomposition in Alaskan tundra. Ecological 2012. Age of the Ørkendalen moraines, Kangerlussuaq, Monographs 66:503–22. Greenland: Constraints on the extent of the southwestern Hollingsworth, T. N., E. A. G. Schuur, F. S. Chapin III, and margin of the Greenland Ice Sheet during the Holocene. M. D. Walker. 2008. Plant community composition as a Quaternary Science Reviews 52:1–5. doi:10.1016/j. predictor of regional soil carbon storage in Alaskan Boreal quascirev.2012.07.021. ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-17 McGuire, D. A., L. G. Anderson, T. R. Christensen, S. shrub tundra sites, western Greenland. Arctic Science Dallimore, L. Guo, D. J. Hayes, M. Heimann, T. D. 2:165–82. doi:10.1139/as-2015-0023. Lorenson, R. D. Macdonald, and N. Roulet. 2009. Post, E., and M. C. Forchhammer. 2008. Climate change Sensitivity of the carbon cycle in the Arctic to climate reduces reproductive success of an Arctic herbivore through change. Ecological Monographs 79:523–55. trophic mismatch. Philosophical Transactions of the Royal Mernild, S. H., E. Hanna, J. R. McConnell, M. Sigl, A. P. Society B 363:2367–73. doi:10.1098/rstb.2007.2207. Beckerman, J. C. Yde, J. Cappelen, J. K. Malmros, and K. Post, E., and C. Pedersen. 2008. Opposing plant community Steffen. 2015. Greenland precipitation trends in a long- responses to warming with and without herbivores. term instrumental climate context (1890–2012): Proceedings of the National Academy of Sciences 105:12353– Evaluation of coastal and ice core records. International 58. doi:10.1073/pnas.0802421105. Journal Climatology 35:303–20. doi:10.1002/joc.3986. Schmidt, M., M. Torn, S. Abiven, and T. Dittmar. 2011. Migala, K., B. Wojtun, W. Szymanski, and P. Muskala. 2014. Soil Persistence of soil organic matter as an ecosystem prop- moisture and temperature variation under different types of erty. Nature 478:49–56. doi:10.1038/nature10386. tundra vegetation during the growing season: A case study Sierra, C. A., S. E. Trumbore, E. A. Davidson, S. Vicca, and I. from the Fuglebekken catchment, SW Spitsbergen. Catena Janssens. 2015. Sensitivity of decomposition rates of soil 116:10–18. doi:10.1016/j.catena2013.12.007. organic matter with respect to simultaneous changes in Min, K., C. A. Lehmeier, F. Ballantyne, A. Tatarko, and S. A. temperature and moisture. Journal Advancement Model Billings. 2014. Differential effects of pH on temperature sen- Earth Systems 7:335–56. doi:10.1002/2014MS000358. sitivity of organic carbon and nitrogen decay. Soil Biology & Sollins, P., C. Glassman, E. A. Paul, C. Swanston, K. Lajtha, Biochemistry 76:193–200. doi:10.1016/j.soilbio.2014.05.021. and J. W. Heil. 1999. C and N Dry Combustion. In Myers-Smith, I. H., S. C. Elmendorf, P. S. A. Beck, M. Standard Soil Methods for Long-Term Ecological Research, Wilmking, M. Hallinger, D. Blok, K. D. Tape, S. A. eds. G. P. Robertson, D. C. Coleman, C. Bledsoe, and P. Rayback, M. Macias-Fauria, B. C. Forbes, J. D. M. Speed, Sollins, 92–95. Oxford: Oxford University Press. N. Boulanger-Lapointe, C. Rixen, E. Levesque, N. M. Stendel, M.,J.H.Christensen, G.Aoalgeirsdottir,N. Kliem,and Schmidt, C. Baittinger, A. J. Trant, L. Hermanutz, L. S. M. Drews. 2007. Regional climate change for Greenland and Collier, M. A. Dawes, T. C. Lantz, S. Weijers, R. H. surrounding seas. Danish Climate Centre Report 07-02, Jørgensen, A. Buchwal, A. Buras, A. T. Naito, V. Danish Meteorological Institute, Copenhagen, Denmark, 26. Ravolainen, G. Schaepman-Strub, J. A. Wheeler, S. Wipf, Accessed March 15, 2015. http://978-87-7478-547-7. K. C. Guay, D. S. Hik, and M. Vellend. 2015. Climate Sturm, M., J. Holmgren, and J. McFadden. 2001. Snow-shrub sensitivity of shrub growth across the tundra biome. interactions in Arctic tundra: A hypothesis with climatic Nature Climate Change. doi:10.1038/nclimate2697. implications. Journal of Climate 14:336–44. Myers-Smith, I. H., B. C. Forbes, M. Wilmking, M. Hallinger, Sturm, M., J. Schimel, G. Michaelson, and J. Welker. 2005. T. Lantz, D. Blok, K. D. Tape, M. Macias-Fauria, U. Sass- Winter biological processes could help convert arctic tun- Klaassen, E. Levesque, S. Boudreau, P. Ropars, L. dra to shrubland. BioScience 55:17–26. Hermanutz, A. Trant, L. S. Collier, S. Weijers, J. Rozema, Sullivan, P. F., S. J. T. Arens, R. A. Chimner, and J. M. S. A. Rayback, N. M. Schmidt, G. Schaepman-Strub, S. Welker. 2008. Temperature and microtopography interact Wipf, C. Rixen, C. B. Menard, S. Venn, S. Goetz, L. to control carbon cycling in a high arctic fen. Ecosystems Andreu-Hayles, S. Elmendorf, V. Ravolainen, J. Welker, 11:61–76. doi:10.1007/s10021-007-9107-y. P. Grogan, H. E. Epstein, and D. S. Hik. 2011. Shrub Tamstorf, M. P., P. J. Aastrup, and L. C. Cuyler. 2005. Modelling expansion in tundra ecosystems: Dynamics, impacts and critical caribou summer ranges in West Greenland. Polar research priorities. Environmental Research Letters Biology 28:714–24. doi:10.1007/s00300-005-0731-8. 6:045509. doi:10.1088/1748-9326/6/4/045509. Thomsen, I. K., B. M. Petersen, S. Bruun, L. S. Jensen, and B. Myers-Smith, I. H., and D. S. Hik. 2013. Shrub canopies T. Christensen. 2008. Estimating soil C loss potentials from influence soil temperatures but not nutrient dynamics: the C to N ratio. Soil Biology and Biochemistry 40:849–52. An experimental test of tundra snow-shrub interactions. doi:10.1016/j.soilbio.2007.10.002. Ecology and Evolution 3:3683–700. doi:10.1002/ece3.710. Urban, M., M. Forkel, J. Eberle, C. Huettich, C. Schmullius, and Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. M. Herold. 2014. Pan-Arctic climate and land cover trends Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. derived from multi-variate and multi-scale analyses (1981– H. Stevens, and H. Wagner 2015. vegan: Community ecol- 2012). Remote Sensing 6:2296–316. doi:10.3390/rs6032296. ogy package. http://CRAN.R-project.org/package=vegan. Weatherspark. 2015. Average Weather For Kangerlussuaq Ostle, N. J., P. Smith, R. Fisher, F. L. Woodward, J. B. Fisher, (Søndre Strømfjord), Greenland. Accessed August 6, J. U. Smith, D. Galbraith, P. Levy, P. Meir, N. P. 2015. https://weatherspark.com/averages/27554/ McNamara, and R. D. Bardgett. 2009. Integrating plant- Kangerlussuaq-S-ndre-Str-mfjord-Kitaa-Greenland. soil interactions into global carbon cycle models. Journal of Xiong, X., S. Grunwald, D. B. Myers, J. Kim, W. G. Harris, Ecology 97:851–63. doi:10.1111/j.1365-2745.2009.01547.x. and N. Bliznyuk. 2015. Assessing uncertainty in soil Petrenko, C. L., J. I. Bradley-Cook, E. M. Lacroix, A. J. organic carbon modeling across a highly heterogeneous Friedland, and R. A. Virginia. 2016. Comparison of carbon landscape. Geoderma 251–252:105–16. doi:10.1016/j. and nitrogen storage in mineral soils of graminoid and geoderma.2015.03.028. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png "Arctic, Antartic and Alpine Research" Taylor & Francis

Landscape variation in soil carbon stocks and respiration in an Arctic tundra ecosystem, west Greenland

Loading next page...
 
/lp/taylor-francis/landscape-variation-in-soil-carbon-stocks-and-respiration-in-an-arctic-hET0nyeKuJ

References (59)

Publisher
Taylor & Francis
Copyright
© 2018 The Author(s). Published by Taylor & Francis.
ISSN
1938-4246
eISSN
1523-0430
DOI
10.1080/15230430.2017.1420283
Publisher site
See Article on Publisher Site

Abstract

ARCTIC, ANTARCTIC, AND ALPINE RESEARCH 2018, VOL. 50, NO. 1, e1420283 (17 pages) https://doi.org/10.1080/15230430.2017.1420283 Landscape variation in soil carbon stocks and respiration in an Arctic tundra ecosystem, west Greenland a b Julia I. Bradley-Cook and Ross A. Virginia Department of Biological Sciences, Ecology and Evolutionary Biology Program, Dartmouth College, Hanover, New Hampshire, USA; Environmental Studies Program, Dartmouth College, Hanover, New Hampshire, USA ABSTRACT ARTICLE HISTORY Received 1 February 2017 The magnitude and acceleration of carbon dioxide emissions from warming Arctic tundra soil is Accepted 30 October 2017 an important part of the Region’s influence on the Earth’s climate system. We investigated the links between soil carbon stocks, soil organic matter decomposition, vegetation heterogeneity, KEYWORDS temperature, and environmental sensitivities in dwarf shrub tundra near Kangerlussuaq, Soil organic carbon; Greenland. We quantified carbon stocks of forty-two soil profiles using bulk density estimates landscape heterogeneity; based on previous studies in the region. The soil profiles were located within six vegetation tundra; soil respiration; soil types at nine study sites, distributed across an environmental gradient. We also monitored air temperature and soil temperature and measured in situ soil respiration to quantify variation in carbon flux between vegetation types. For spatial extrapolation, we created a high-resolution land cover classification map of the study area. Aside from a single soil profile taken from a fen soil −2 −2 (54.55 kg C m ;2.13kgNm ), the highest carbon stocks were found in wet grassland soils −2 (mean, 95% CI: 34.87 kg C m , [27.30, 44.55]). These same grassland soils also had the highest mid-growing-season soil respiration rates. Our estimation of soil carbon stocks and mid-grow- ing-season soil respiration measurements indicate that grassland soils are a “hot spot” for soil carbon storage and soil carbon dioxide efflux. Even though shrub, steppe, and mixed vegetation −2 had lower average soil carbon stocks (14.66 – 20.17 kg C m ), these vegetation types played an important role in carbon cycling at the landscape scale because they cover approximately 50 percent of the terrestrial landscape and store approximately 68 percent of the landscape soil organic carbon. The heterogeneous soil carbon stocks in this landscape may be sensitive to key environmental changes, such as shrub expansion and climate change. These environmental drivers could possibly result in a trend toward decreased soil carbon storage and increased release of greenhouse gases into the atmosphere. Introduction predictions to an aggregate response of the ecosystem 6 2 The tundra biome covers 7.5 × 10 km north of the at the landscape scale (Hinzman et al. 2013). Arctic tree line, a region that is undergoing rapid cli- Consideration of soil carbon processes at the land- mate and ecosystem change (Callaghan et al. 2005). The scape level introduces spatial heterogeneity and soils in this high-latitude ecosystem store an estimated dynamics of ecosystem properties (such as soil organic 1,300 Pg of carbon (Hugelius et al. 2014), which is carbon content) along with landscape characteristics approximately twice the carbon contained in the atmo- (e.g., elevation, topography), abiotic conditions (e.g., sphere. Climate and environmental impacts on these moisture), biotic factors (vegetation type), soil forma- soils could affect the global carbon cycle as a result of tion (e.g., time), and associated interactions among microbial release of stored soil carbon through decom- these variables (Jenny 1941). Previous studies on tundra position. Models predict that the response in decom- soils have identified soil temperature, soil moisture, position is based on molecular scale biokinetic disturbance, litter quality, permafrost, and microtopo- properties (Davidson and Janssens 2006; Sierra et al. graphy (Sullivan et al. 2008) as important controls on 2015), but it remains a challenge to link these soil carbon accumulation (Schmidt et al. 2011). CONTACT Julia I. Bradley-Cook julia.i.bradley-cook.gr@dartmouth.edu © 2018 The Author(s). Published by Taylor & Francis. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. e1420283-2 J. I. BRADLEY-COOK AND R. A. VIRGINIA Information about heterogeneity in carbon stocks belowground environmental conditions, such as soil needs to be combined with an understanding of temperature and moisture, which influence microbial variationin soilcarbonquality, temperature sensitiv- activity and soil carbon accumulation (Hudson, ity of decomposition processes, and environmental Henry, and Cornwell 2011;Ostle et al. 2009). controls on carbon cycling in order to best predict Shrub expansion into grassland is a trend that is how the carbon stocks will respond to landscape- widely observed throughout the Arctic (Frost and wide drivers of change. However, it is not well Epstein 2014; Myers-Smith and Hik 2013; Urban et al. understood how variations in soil carbon quality, 2014). Shrub expansion has been observed in west defined as the decomposability of carbon (Bosatta Greenland (Jørgensen, Meilby, and Kollmann 2013), and Ågren 1999), and temperature affect decomposi- but grazing from large herbivores has suppressed tion at the landscape scale. The carbon quality tem- shrub expansion (Post and Pedersen 2008) and slowed perature hypothesis predicts that the temperature the carbon cycling response to warming in the tundra sensitivity of decomposition increases with soil car- near Kangerlussuaq (Cahoon et al. 2011). Changes in bon recalcitrance as long as decomposition is not the extent and abundance of shrubs may also be locally constrained by environmental factors (Davidson and limited by low soil moistures (Myers-Smith et al. 2015). Janssens 2006). In support of the carbon quality Associations between vegetation and soil carbon are temperature hypothesis, Fierer, Colman, and valuable for prediction because, unlike subsurface char- Schimel (2006) found the temperature sensitivity of acteristics, vegetation can be detected remotely using decomposition to increase with soil carbon recalci- aerial and satellite imagery. Few studies have combined trance at sites across the continental United States. estimates of carbon stocks and temperature sensitivities However, conflicting results from a landscape analy- at the landscape scale to understand the landscape-level sis at the Konza Prairie Biological Station, USA, patterns of soil carbon storage and respiration found a wider range of temperature sensitivities and (Horwath Burnham and Sletten 2010; Hugelius and higher maximum sensitivity than observed at the Kuhry 2009) and, to our knowledge, no such study continental scale (Craine et al. 2009). Spatial hetero- has been undertaken in the Kangerlussuaq area in geneity in abiotic controls, carbon accumulation, and west Greenland. thermal sensitivities of decomposition should affect The objective of this study is to link landscape- carbon storage and response to environmental dri- level variation in temperature and vegetation cover to vers, such as climate change. Therefore, we need to soil carbon stocks and sensitivities to environmental understand landscape-level distribution of soil carbon change. We characterize variation in (1) soil tem- and variation in temperature sensitivity of decompo- perature, (2) soil carbon storage and soil chemistry, sition to improve spatially explicit predictions of the and (3) soil respiration by vegetation type across a effect of climate and environmental change on soil climate gradient in the tundra landscape near organic carbon (SOC) pools. Kangerlussuaq, Greenland. We hypothesized that Vegetation types also affect soil carbon storage soil temperature, soil carbon storage and soil chem- through organic matter input (i.e., litter, root exuda- istry, and soil respiration would vary by vegetation tion and turnover) and mediation of the below- type. Furthermore, we hypothesized that soil tem- ground environment. These belowground perature, carbon storage, and respiration would interactions define the stability of soil carbon and increase with distance from the Greenland Ice the thermal sensitivity of soil decomposition. Plant Sheet. By applying the results to a spatially explicit species and functional group can influence the quan- model, we aimed to identify the role of different land tity and quality of SOC (Creamer et al. 2011; cover classes in landscape-level soil carbon storage Hollingsworth et al. 2008;Ostle et al. 2009). For and carbon dioxide emissions. example, Arctic shrubs produce lignaceous biomass that tends to have high C:N mass ratios that, when compared to herbaceous species, result in a less Methods decomposable, lower quality resource for soil micro- Study Area bial communities (Chapin et al. 1996; Hooper and Vitousek 1998). Numerous studies in the Arctic have We conducted fieldwork at nine sites in the shrub shown that decomposition decreases with increasing tundra landscape near Kangerlussuaq, Greenland organic matter C:N ratios (Haddix et al. 2011; (Figure 1). The landscape was deglaciated approxi- Hobbie 1996; Thomsen et al. 2008). Vegetation func- mately 7,000 years ago (Levy et al. 2012), and is tional groups can also have distinct impacts on located at the margin of the current extent of the ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-3 Zone 1 Z1S2 Z1S1 Z1S3 Zone 2 Z2S1 Z2S2 Z2S3 Greenland Ice Sheet Zone 3 Z3S3 Z3S2 Z3S1 Figure 1. Map of the study area, with sites indicated over a WorldView2 satellite image from July 10, 2010. Inset map marks the study area (black rectangle) in west Greenland. Greenland Ice Sheet (GrIS). The landscape is covered Land Cover Classification by aeolian silt deposits above bedrock and glacial till (Dijkmans and Törnqvist 1991). Soils are humus-poor To create a land cover classification map of the study arctic brown soil in the soil order Gelisol (Jones et al. area near Kangerlussuaq, we conducted a multistage 2009). Soil erosion features, termed deflation patches, unsupervised classification of a WorldView2 image which are likely a result of strong winds from the (multispectral, 1.34 m resolution) taken on July 10, GrIS, are common on the landscape and are visually 2010. In ENVI Software (Harris Geospatial Solutions) distinct areas with low vegetative cover and produc- we ran a ISODATA land cover classification to obtain tivity (Heindel, Chipman, and Virginia 2015). In addi- twenty spectral classes with a 5 percent change thresh- tion to geophysical controls, climate and vegetation old and a maximum of twenty iterations. Drawing on shifts are likely to be key environmental drivers of field knowledge accumulated during the summers of biogeochemical cycling in this and other tundra eco- 2010–2013 and visual inspection of the satellite ima- systems. From 1973 to 1999, the average annual atmo- gery, we visually interpreted the output spectral classes spheric temperature observed at Kangerlussuaq was to determine land cover classes based on vegetation −5.7°C (DMI 2017). According to regional climate functional group. Six spectral classes contained pixels projections, by the years 2021–2050, atmospheric tem- of multiple land cover types, so we built a mask for perature is projected to increase by 2°C in summer each mixed class with Spatial Modeler in ERDAS and autumn seasons from historical seasonal averages Imagine to isolate the mixed classes. We then ran of 9.2°C and −4.9°C, respectively. Winter is projected ISODATA on each masked class with a maximum of to increase by 3°C from a historical mean of −19.2°C, eight classes and fifteen iterations. For classes that were while a 4°C increase from a historical mean of −7.7°C split between water and land after the second stage of is projected for the spring (DMI [Danish unsupervised classification, we chose to preserve the Meterological Institute] 2017; Stendel et al. 2007). integrity of terrestrial land cover. We merged the initial Annual precipitation in Kangerlussuaq is approxi- classified and the masked images using Spatial Modeler mately 250 mm (Mernild et al. 2015) and is projected and simplified the images into nine main classes: shrub, to increase 15 percent by 2021–2050 and 30–40 per- steppe, grassland, mixed vegetation, fen, eroded soil cent by 2051–2080 (Stendel et al. 2007). (ES), water, ice, and cloud (Table 1). The ES class e1420283-4 J. I. BRADLEY-COOK AND R. A. VIRGINIA includes deflation patches and exposed bedrock, but Table 1. Description of cover type used in land classification and field sampling. bedrock is less than 20 percent of unvegetated areas Cover Type Common Species Description (Heindel, Chipman, and Virginia 2015). To focus on Shrub Salix glauca Dominated by dwarf terrestrial classes, we calculated percent cover and total Betula nana shrub species area of shrub, steppe, grassland, mixed vegetation, fen, Rhododendron tomentosum and ES land cover types by multiplying the number of Mixed Betula nana Occurs mainly on damp vegetation Rhododendron soil on gently sloping pixels in each class by the pixel size. groenlandicum hillsides. Composed of a We assessed the accuracy of the land cover classifi- Epetrum nigrum mix of shrub, with forbs and graminoid cation by comparing land cover classes to 288 ground undergrowth control waypoints from field observations. These points Steppe Calamagrostis purpurascens Graminoid dominated were not used during the development of the classifica- Carex supina with scattered forbs, Agrostis mertensii common on south-facing tion. We used the ground control points to calculate an Kobresia myosuroides slopes error matrix and accuracy statistics for the terrestrial Potentilla arenosa classes in the land classification map. Eroded soil (ES) Biological soil crusts Common on ridgelines Dryas integrifolia and south-facing slopes, Silene acaulis where loess soil has been locally removed by wind Site Selection erosion: 10–20 percent vegetation cover* Study sites for field measurements and soil sample col- Grassland Poa pratensis Found in moist lection were selected from the land cover classification Erophorum angustifolum depressions, and are Calamagrostis lapponica dominated by graminoid map and ground observations. Three zones were estab- Carex bigelowii species with intermixed lished according to proximity to the margin of the GrIS, Campanula gieseckiana forbs Cerastium alpinum where Zone 1 borders the ice edge and Zone 3 is the Ranunculus hyperboreus farthest away, extending toward the fjord Kangerlussuaq Fen Carex sp. Found along the edge of (Figure 1). Within each zone we identified five to seven Eriophorum sp. streams and lakes, and Hippuris vulgaris includes shallow possible sites that each contained representative land lakebeds with ephemeral cover types (shrub, mixed vegetation, steppe, and ES). surface water Three sites were randomly selected from candidate sites. *Heindel, Chipman, and Virginia (2015). Grassland was included when present (a total of five sites). Fen samples were collected at one site, Zone 1 Site 2 (Z1S2). We measured air and soil temperature, Table 2. Sample collection and measurement across the study conducted vegetation surveys, and collected soil samples sites. Vegetation types are shrub (SH), steppe (ST), grasslands (GL), mixed vegetation (MX), eroded soil (ES), and fen (FN). at all sites, and took in situ soil respiration measurements No. Soil at seven sites (Table 2). Vegetation Temperature Surveys The terrestrial land cover classes used in the land Loggers by and Soil In Situ Soil Vegetation Class classification and sampling design (shrub, mixed vege- Air Samples Respiration Zone Site Temperature SH ST MX ES (2011) (2012) tation, steppe, grassland, fen, and ES; Table 1) are 1 Z1S1 Yes 3 2 3 3 SH, ST, GL, July 12 comparable to a Landsat-based vegetation classification MX, ES, FN created to assess caribou habitat (Tamstorf, Aastrup, Z1S2 Yes 2 1 3 3 SH, ST, MX, July 12 ES and Cuyler 2005). Z1S3 Yes 3 2 3 2 SH, ST, GL, N/A MX, ES 2 Z2S1 No 2 2 3 2 SH, ST, GL, July 13 Air and Soil Temperature MX, ES Z2S2 Yes 2 2 1 3 SH, ST, GL, July 13 MX, ES At each site, we monitored air and soil temperature Z2S3 Yes 1 2 1 3 SH, ST, GL, July 13 every four hours using Thermocron iButton loggers MX, ES 3 Z3S1 Yes 3 1 2 3 SH, ST, GL, N/A (Model DS 1921G, Embedded Data Systems®). To mea- MX, ES sure air temperature at each site and capture air tem- Z3S2 Yes 3 2 3 3 SH, ST, MX, July 14 ES perature patterns across the study area, iButton loggers Z3S3 Yes 3 2 2 2 SH, ST, MX, July 14 were installed in PVC capsules with a drilled hole to ES enable air exchange, attached to rebar at 30 cm height, and shaded under an aluminum roof in an area without types between July 17, 2011, and June 12, 2012. shrub cover from July 11, 2011, to August 22, 2012. Soil Within each vegetation type, we identified three soil temperature loggers were buried at 5 cm depth within temperature locations by randomly selecting a direction steppe, shrub, mixed vegetation, and ES land cover ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-5 (degrees from true north) and a number of steps from using standard methods (Sollins et al. 1999). the center point of a continuous patch of land cover. If Hydrochloric acid was added to samples from mineral a patch was smaller than approximately 8 m in dia- soils in vegetated areas to remove inorganic carbonates. meter, we distributed the temperature loggers among No reaction to the hydrochloric acid was visually more than one patch. We placed a total of twelve observed. loggers per site, and 108 across the entire study area. The data from twenty-five loggers were not included in In Situ Soil Respiration Measurements analyses because the loggers were either disturbed by wildlife or could not be located, but all vegetation types We measured in situ soil respiration (CO flux) with a at each site had at least one logger with a complete data portable Li-Cor 8100 (Lincoln, NE) infrared gas analy- record (Table 2). All temperature loggers were wrapped zer (IRGA) with a 20 cm survey chamber attached. At in parafilm and neoprene plastic for waterproofing. seven of the nine field sites, we installed three 20 cm diameter PVC collars at randomized locations within shrub, steppe, grassland, mixed vegetation, and ES Vegetation Surveys and Soil Sampling vegetation types. Following installation, collars sat for At each site, we identified a soil pit location at the at least twenty minutes to minimize the effect of phy- center of each vegetation type. We avoided vegetation sical disturbance on CO diffusion across the soil sur- boundaries to maximize the likelihood of collecting face. This time interval was selected based on sampling soils that have a long-lived association with the vegeta- logistics and a limited number of PVC collars that tion type of interest. Prior to disturbing the soil surface, precluded long-term set up in the sampling required we conducted a vegetation survey within a 0.5 m for this particular study. We measured the height of the quadrat at each soil pit location. We visually estimated collar above ground to calculate the volume of the percent cover of shrub, herbaceous, graminoid vegeta- headspace in each PVC ring. In a random order, we tion using a 0.5 m quadrant. recorded the CO flux with a two-minute observation We collected soil samples from a 50 cm soil profile after a twenty-second pre-purge. We collected a total of using visual classification to sample detectable hori- 106 measurements during a rainless three-day sampling zons. At some sites, the active layer was shallower period (July 12–14, 2012). Data were not collected at than 50 cm, so we sampled soil to the depth of frozen two sites because of logistical challenges and limited ground. We collected at least 75 g of soil from each field time: Z1S3 was too remote to access during the depth interval using a spoon that was cleaned between survey period, and Z3S1 was an outlier in elevation samples to minimize contamination, and stored sam- (354 m vs. 253 m and 264 m for the other Zone 3 sites). ples in separate sterile Whirl-pak® bags (Nasco, Fort Atkinson, WI). Soil samples were frozen and shipped Calculations and Data Analysis to the Environmental Measurements Lab, Dartmouth College (Hanover, NH) where they were kept at −20°C Temperature Analyses until processing. To test our assumption that air temperature increases extending away from the ice sheet, we compared mean annual temperature, growing season temperature, and win- Soil Analyses ter temperature between zones using a multivariate analysis Samples were thawed and sieved to isolate the less than of variance (MANOVA) followed by univariate analysis to 2 mm fraction of each soil horizon for laboratory test differences within each dependent variable. The annual analysis. Soil water content was estimated from a 10 g average was calculated from data recorded between July 11, soil sample that was dried at 95°C for 24 h, reweighed, 2011, and July 10, 2012. We defined growing season as the −1 and calculated as g water g dry soil. Soil pH was dates between leaf out and senescence, May 22 to August 7 measured using a 1:2 solution of soil:di-H O using a (Post and Forchhammer 2008;Postand Pedersen 2008), pH meter (Thermo Scientific, Orion 3 Star A111 pH and winter season as the cold season climate window, Benchtop, Waltham, MA). Electrical conductivity was November 28 to March 27 (Weatherspark 2015). The log- measured using a 1:5 solution of soil:di-H O using a ger foronesite(Z2S1)disappearedduring thewinter conductivity meter (Thermo Scientific, Orion 3 Star months, so the site was not included in air temperature Conductivity Benchtop, Waltham, MA). Carbon and comparisons. nitrogen content was measured on soils ground with a We compared the soil temperature environment mortar and pestle using a Carlo Erba NA-1500 elemen- using thermal sum for each logger during the measure- tal analyzer (Carlo Erba Instruments, Milan, Italy) ment period. Thermal sum is the difference between e1420283-6 J. I. BRADLEY-COOK AND R. A. VIRGINIA the recorded temperature above the baseline tempera- density to soil horizon and depth based on measure- ture of 0°C and the baseline, summed for the number ments from other studies that we conducted in the of days in the measurement period. We conducted a same area and vegetation types, with bulk densities of −3 model comparison of linear mixed effect models to test 0.25 g DW cm for organic soils (Bradley-Cook and −3 land cover type, zone, and the interaction between the Virginia 2016), 0.62 g DW cm for shallow mineral −3 two as predictor of soil thermal sum, with site as a soils (<10 cm depth), and 1.37 g DW cm for deeper random variable in the model. A second model com- mineral soils (10–20 cm; Petrenko et al. 2016). We parison was conducted to test land cover type, zone, estimated soil nitrogen pools with identical calculations and their interaction as predictors of thermal sum of from percent N. We calculated terrestrial soil carbon soils in vegetated land cover types, excluding the soil and nitrogen inventories at the landscape scale by mul- land cover type. We used Tukey’s HSD (α = 0.05) to tiplying soil carbon content by area in the land cover evaluate differences between means, and calculated classification map for each terrestrial class. marginal R , a measure of variance for mixed effect We used linear mixed effects models to test zone and models. vegetation type as predictor of soil carbon stocks of the full soil profile. We identified significant differences Soil Chemistry Ordination between vegetation types using Tukey’s HSD test in We conducted multivariate analysis to test whether soil the multcomp package in R. Carbon and nitrogen chemistry differed by vegetation type and zone. areal stocks were log-transformed to meet assumptions Measures of percent organic C, percent N, C:N, pH, of normality and homoscedasticity. We back-trans- EC, and soil water content were used in a partial formed the mean and the 95 percent confidence inter- redundancy analysis (RDA) to determine if these vari- val values (Hanlon and Larget 2011). ables differed by vegetation type and zone (vegan pack- age in R [Oksanen et al. 2015]). RDA provides a quantitative method of testing hypotheses in multidi- Results mensional datasets. Each soil sample is assigned scores Land Cover Classification on constrained axes of the predictor variables and unconstrained axes to account for the remaining var- The land cover classification contains nine land cover iance. Soil chemistry measures were used as response classes: shrub, steppe, grassland, mixed vegetation, variables. Vegetation type and zone were used as pre- eroded soil (ES), fen, water, fluvial sediment (outwash) dictor variables, with depth as a covariate in the model. and ice (Figure 2). Ground-based vegetation surveys at Eight soil samples were not included in the analysis 288 points reveal that the classification has an overall because the sample did not contain enough soil mass accuracy of 50 percent. Producer’s accuracy, which to conduct a full set of soil chemistry measurements. provides the probability that a pixel in the classification We analyzed a total of 239 soil samples. The response corresponds with the correct vegetation type, was as variables were standardized using the scale function to high as 84 percent in the ES class and as low as 17 reduce the influence of the magnitude of model vari- percent in the fen class (Table 3). User’s accuracy, or ables on the association between samples. A permuta- the probability that the cover type at a single point tions test with 5,000 permutations was used to corresponds with the land cover class on the map, determine if vegetation type and zone explained a sig- ranged from 20 percent for fen to 60 percent for grass- nificant portion of the variance in soil chemistry land. The most dominant land cover type of the terres- between samples. Adjusted R was calculated to parti- trial landscape was steppe (25%), and is closely followed tion variance between the explanatory and covariate by mixed vegetation (22%), ES (22%), and shrub (19%; variables (Borcard, Gillet, and Legendre 2011). Table 4). The least common land cover types were grassland (7%) and fen (5%; Table 4). Soil Carbon Pools and Landscape Storage We estimated the organic carbon pool for near-surface Air Temperature soil (0–20 cm depth) and the full active layer profile up to 50 cm depth (Equation 1): There was a statistically significant difference in air tem- 2 4 2 2 perature regime between the three zones (F =7.56, 1,6 SOC kg C m ¼%C  D  d  10 cm m (1) P =0.041; Wilk’s Λ = 0.1508) for the full measurement 3 1 10 kg g period, July 11, 2011, to August 22, 2012. Growing season where % C is SOC concentration, D is bulk density (g temperatures were significantly different by zone −3 soil × cm ), and d is depth (cm). We assigned bulk (F =9.335, P = 0.02). Zone 1 had a lower growing 1,6 ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-7 Greenland Ice Sheet Shrub Mixed Veg Outwash Steppe Eroded Soil Ice Grassland Fen Water Figure 2. Land cover classification map of the study area. Colors coincide with land cover classes, including the following vegetation classes: shrub, steppe, grassland, mixed vegetation, eroded soil, and fen. days, °Cd) than Zone 3 (mean = 1338.8 °Cd; Tukey’sHSD Table 3. Error matrix comparing land cover classes from satellite classification with ground-based vegetation observations. P = 0.0585; Figure 3). Thermal sum at Zone 2 Vegetation classes are shrub (SH), steppe (ST), grassland (GL), (mean = 1274.0 °Cd) did not differ from Zone 1 mixed vegetation (MX), eroded soil (ES), fen (FN), and water (W). (P = 0.191) or Zone 3 (P =0.770; Figure 3). Ground Control Data SH ST GL MX ES F W Row Total Classification by SH 21 7 1 17 3 0 0 49 Soil Temperature satellite image ST 12 34 28 8 1 0 65 GL 5 17 6 40 0 0 32 MX 21 8 1 35 33 0 71 Land cover type and zone explained a significant amount ES 2 7 0 0 47 00 56 of the variance in soil thermal sum (Table 5). ES had FN 4 0 0 0 1 1 06 W1 0 0 0 8 0 0 9 higher thermal sums than all other land cover classes at Column total 66 73 10 64 70 5 0 288 all zones (Figure 3). Thermal sums are reduced by more Producer accuracy User accuracy SH 43% SH 32% than 400 degree days in vegetated areas (Figure 3). Mean ST 52% ST 47% thermalsumsofsteppe soilshaveanarrowrange ofvalues GL 19% GL 60% MX 49% MX 55% across all zones (mean = 183°Cd, min = 171°Cd, ES 84% ES 67% max = 205°Cd). Shrub soils have lowest thermal sums FN 17% FN 20% Overall accuracy 50% in Zone 2 (Zone 1 = 298°Cd, Zone 2 = 94°Cd, Zone 3 = 395°Cd). Thermal sum in mixed vegetation decreases moving away from the ice sheet (Zone 1 = 308°Cd, Zone season temperature (mean = 9.0°C) than Zone 3 2 = 153°Cd Zone 3 = 119°Cd). (mean = 11.2°C; Tukey’sHSD P = 0.0586), but not com- pared to Zone 2 (mean = 10.9°C; Tukey’sHSD P =0.135). Variation in Soil Organic Carbon, Nitrogen, and Growing season temperatures of Zone 2 and Zone 3 were Chemistry not significantly different (Tukey’sHSD P =0.914). Annual temperatures did not differ by zone The partial RDA examined vegetation and zone as pre- (F =3.727, P = 0.102), with a mean annual air tempera- dictors of soil chemistry measurements. The permuta- 1,6 ture for all sites of −3.1°C. Winter temperatures also did tion test was significant (F = 22.763, P =0.001). The 6,231 not differ between zones (F =0.128, P = 0.733), with a first two constrained axes were both significant 1,6 mean temperature of −17.3°C. Thermal sum was signifi- (P < 0.01) and explain 33 percent of the variance, with cantly different by zone (F = 10.377, P = 0.018), with 25.4 percent attributed to RDA1 and 7.6 percent attrib- 1,6 lower thermal sums at Zone 1 (mean = 1084.1 degree uted to RDA2. The remaining constrained axes, RDA3 e1420283-8 J. I. BRADLEY-COOK AND R. A. VIRGINIA Table 4. Area, percent cover, carbon and nitrogen pools for each terrestrial land cover class. Area was extracted from the land cover classification. Pool sizes are mean values. Total C Total C Area 20-cm 50-cm −2 −2 2 Land Cover Class Mean C 20-cm (kg m ) Mean C 50-cm (kg m ) (km (%)) (Gg C (%)) (Gg C (%)) (A) Organic Carbon Shrub 6.44 14.67 16.13 (19%) 103.86 (16%) 236.59 (17%) Steppe 8.98 20.17 21.16 (25%) 189.91 (30%) 426.78 (31%) Grassland 19.07 34.87 5.60 (7%) 106.86 (17%) 195.35 (14%) Mixed veg 6.71 15.13 18.12 (22%) 121.66 (19%) 274.26 (20%) Eroded soil 0.46 0.70 18.63 (22%) 8.48 (1%) 13.02 (1%) Fen 24.36 54.55 4.22 (5%) 102.91 (16%) 230.44 (17%) TOTAL –– 83.86 633.68 1,376.45 (B) Total Nitrogen Shrub 0.32 0.75 16.13 (19%) 5.21 (13%) 12.13 (14%) Steppe 0.59 1.33 21.16 (25%) 12.54 (30%) 28.12 (31%) Grassland 1.30 2.30 5.60 (7%) 7.27 (18%) 12.87 (14%) Mixed veg 0.36 0.82 18.12 (22%) 6.54 (16%) 14.88 (17%) Eroded soil 0.04 0.06 18.63 (22%) 0.80 (2%) 1.12 (1%) Fen 2.13 4.85 4.22 (5%) 9 (22%) 20.49(23%) TOTAL –– 83.86 41.36 89.62 Figure 3. Thermal sums of (A) air and (B) soil temperatures at 5 cm depth in different vegetation types and in three zones. Data points mark site averages with error bars indicating ±1 standard error. Thermal sum of air temperature was calculated for one year (July 12, 2011–July 11, 2012), and soil temperature was calculated between July 17, 2011, and June 12, 2012. Shapes of soil thermal sums indicate vegetation cover (ES = eroded soil, ST = steppe, MX = mixed vegetation, SH = shrub). ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-9 Table 5. Comparison of models explaining soil thermal sums. The preferred model (ΔAICc < 2) contains an interaction between vegetation type and zone. Site was included as a random variable in all models. Marginal R of the best fitting model was 0.8116 (Veg = Vegetation). Rank Predictor Variable k AICc ΔAICc AICc Model Weight Log-likelihood Log(L) 1 Veg + Zone + Veg: Zone 14 1120.4 0.0 1 −543.1 2 Vegetation 6 1132.8 12.4 0 −559.9 3 Vegetation + Zone 8 1134.3 13.9 0 −558.2 4 Intercept 3 1232.7 112.4 0 −613.2 5 Zone 5 1235.8 115.4 0 −612.5 and RDA4, were not significant (P >0.05). The correla- Electrical conductivity varied substantially for all vegeta- tion variable, soil sample depth, explained 5.2 percent of tion types, and did not have a strong association with the variance. The remainder of the variance, 61.4 per- vegetation types or the other chemical measurements. cent, is unconstrained by vegetation type, zone, or depth. The permutation test showed that vegetation Soil C and N Storage by Vegetation Type explains a significant amount of the variation in soil chemistry measurements (F = 26.4, P = 0.0002). Soil carbon stocks varied among land cover types. In the 5,231 Zone did not explain a significant amount of the total prediction of soil carbon stocks, the best-ranked model variance in soil chemistry (F = 1.0, P = 0.38). (ΔAICc < 2) contained vegetation type as the only pre- 1,238 Total organic carbon, nitrogen, and soil water content dictor (Table 7). Grassland soils and the single fen sample were tightly correlated explanatory variables that aligned contained the highest mean carbon stocks (mean, 95% CI: −2 −2 with the RDA1 axis (Figure 4). Grassland and fen samples 34.87 kg C m , [27.30, 44.55] and 54.55 kg C m , had high values along this axis, indicating that the soils have respectively; Figure 5). Steppe (mean, 95% CI: 20.17 kg −2 high carbon, nitrogen, and soil water content. Grassland, Cm , [12.64, 32.18]), mixed vegetation (mean, 95% CI: −2 shrub, and mixed vegetation shared similar position on the 15.13 kg C m , [9.50, 24.11]), and shrub (mean, 95% CI: −2 RDA1 axis. C:N corresponds with RDA2, and shrub, 14.67 kg C m , [9.80, 21.95]) soil carbon storage were not steppe, and mixed vegetation vary along RDA2, with the significantly different from one another (Figure 5). ES highest C:N in shrub soils, the lowest values for steppe soils, areas had the lowest carbon storage (mean, 95% CI: −2 and mixed vegetation in an intermediate position. 0.07 kg C m ,[0.46,1.07]; Figure 5). Neither pH nor electrical conductivity were tightly Mean nitrogen stocks in soils ranged from 0.06 to −2 correlated with the RDA axes (Figure 4). ES soils had a 4.85 kg N m . As in the prediction of soil carbon, vegeta- strong correlation with pH (Figure 4), with higher pH tion type was the only predictor term in the best model of measurements than the other vegetation types (Table 6). soil nitrogen stocks (Table 8). The highest average N stock Figure 4. Ordination biplot of the partial RDA of soil chemistry measurements using vegetation type and zone as the independent variables (Table 6). The centroid of each vegetation type is labeled for shrub (VegSH), mixed vegetation (VegHB), steppe (VegGR), grassland (VegGR2), eroded soil (VegDZ), and fen (VegLA). TOC is total organic carbon, Grav Moist is soil moisture, TN is total nitrogen, EC is electrical conductivity, and PH is pH. The biplot represents 31 percent of the total variance (25.4% on RDA1 and 7.6% on RDA2). e1420283-10 J. I. BRADLEY-COOK AND R. A. VIRGINIA Table 6. Mean and range of soil chemistry measurements collected from six different vegetation types: N is the number of soil −1 samples, %C is organic carbon content, %N is nitrogen content, SWC is soil water content (g H O g Soil ), and EC is electrical −1 conductivity (micro-siemens cm ). Land Cover Class N %C %N C:N pH EC SWC Eroded soil 50 0.2 (0.01–1.0) 0.02 (0–0.08) 13.1 (3.6–45.7) 7.6 (5.8–8.6) 51 (2–1648) 2.9 (0.09–7.05) Steppe 55 5.5 (0.2–21.6) 0.37 (0.01–1.38) 15.2 (8.5–25.3) 6.4 (5.8–7.6) 39 (4–228) 44.6 (2.2–138.5) Grassland 27 10.2 (0.5–27.5) 0.66 (0.04–1.63) 15.4 (12.3–19.8) 6.3 (5.1–7.4) 94 (7–594) 82.3 (3.61–381.4) Mixed vegetation 50 4.6 (0.2–12.9) 0.25 (0.02–0.66) 17.8 (8.5–29.1) 6.4 (5.4–7.2) 26.6 (3–314) 34 (4–80) Fen 5 22.7 (20.3–26.3) 1.99 (1.81–2.26) 11.4 (11.0–11.7) 6.3 (6.1–6.6) 135 (52–323) 189.1 (178–205) Shrub 60 4.0 (0.2–20.5) 0.21 (0.01–0.99) 19.7 (6.3–51.4) 6.3 (4.9–7.6) 25 (2–129) 23.4 (3.84–71.7) Table 7. Comparison of models explaining soil organic carbon (e.g., Z1S1-GL [grassland], Z3S2-ST [steppe], Z3S3-MX stocks. The top-ranked model, with vegetation as a main effect, [mixed vegetation], and Z3S3-ST [steppe]; Figure 7). is preferred. All other models, ΔAICc > 2. Steppe and mixed vegetation comprise the largest AICc Log- fraction of total landscape carbon (31% and 20%, Model likelihood Rank Predictor Variable k AICc ΔAICc Weight Log(L) respectively; Table 4). Fen and grassland make up 17 1 Vegetation 7 276.97 0 0.80 −129.8 percent and 14 percent, respectively, even though they 2 Vegetation + Zone 8 297.75 2.78 0.20 −129.7 cover the smallest fraction of the landscape area (5% 3 Veg + Zone + Veg: 12 286.90 9.93 0.01 −129.1 Zone and 7%, respectively; Table 4). An estimated 17 percent 4 Intercept 2 297.93 21.01 0.00 −146.8 of landscape carbon is stored in shrub soils. ES areas, 5 Zone 3 298.72 21.75 0.00 −146.0 which cover 22 percent of the landscape area, only store 1 percent of the total carbon in the landscape (Table 4). −2 was in the single fen soil profile (4.85 kg N m ; Figure 5), Nitrogen stocks in the terrestrial landscape are −2 and grasslands also had elevated N stocks (mean, 95% CI: 1.07 Gg N m in the full soil profile. Near-surface −2 2.30 kg N m , [1.60, 3.30]). Steppe (mean, 95% CI: 1.33 kg soils store 46 percent of total nitrogen, with an average −2 −2 Nm ,[0.76,2.32]), mixedvegetation(mean,95% CI: of 0.49 Gg N m in the top 20 cm. The largest nitrogen −2 0.82 kg N m , [0.50, 1.34]), and shrub (mean, 95% CI: stock is in steppe (31%), which is more than double −2 0.75 kg N m , [0.47, 1.19]) were not significantly different that of shrub (14%; Table 4). Fen soils store 23 percent (Figure 5). ES areas had the lowest nitrogen stocks of all of landscape nitrogen, which is disproportionately high −2 land cover types (mean, 95% CI: 0.06 kg N m , [0.04, when considering it covers only 5 percent of the land- 0.09]; Figure 5). scape (Table 4). ES only stores 1 percent of the total landscape nitrogen. Soil Carbon Respiration by Vegetation Type Discussion Soil carbon respiration rates during the mid-season dif- fered by vegetation type (Figure 6). The highest respiration Landscape Temperature Variation was observed in grassland vegetation, with an average rate Temperature regimes vary with proximity to the ice −2 −1 of 7.10 ± 0.60 μmols CO m s , which was significantly sheet. Within the study area, which captured a distance greater than all other vegetation types (P <0.05).Average of approximately 11 km from the GrIS margin, the soil respiration for steppe, shrub, and mixed vegetation influence of the GrIS on atmospheric temperatures is −2 −1 ranged from 3.11 to 3.42 μmols CO m s ,but did apparent during summer months, but does not have an not significantly differ from each other. In ES, average effect during the winter or at an annual scale. The 2°C respiration rate was an order of magnitude lower than cooler growing season temperatures near the margin of −2 −1 grasslands (0.69 ± 0.45 μmols CO m s ). the GrIS can have ecological consequences, including altered carbon cycling. Lower temperatures decrease net primary productivity and microbial activity, result- Landscape Soil Carbon and Nitrogen Estimates ing in a slower cycling of carbon (Chapin et al. 2009). The terrestrial landscape stores an average of 16.41 kg Regional climate models project that warming will be −2 Cm in the active layer soils to a depth of 50 cm. greatest during the winter season, and will be limited to Approximately 46 percent of this carbon is stored in approximately 2°C warming during the summer months the 0–20 cm depth increment, with an average of 7.4 kg because of a cooling effect from the GrIS (Stendel et al. −2 Cm for the top 20 cm of soil. Several soil profiles had 2007). Since proximity to the ice sheet does not affect elevated carbon content in the deepest sample collected winter temperature within the study area, we infer that ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-11 Figure 5. Soil (A) organic carbon and (B) nitrogen pools to a depth of 50 cm for soils from six vegetated land cover types in western Greenland (ES = eroded soil, ST = steppe, GL = grasslands, MX = mixed vegetation, FN = fen, SH = shrub). Horizontal tabs mark the mean of each vegetation type, with error bars that represent ±1 standard error of the mean. Gray data points are the carbon stocks from each soil pit. Different letters indicate significant differences in Tukey HSD post hoc test (P < 0.05). projected winter warming of 3°C by 2050 (Stendel et al. increase in temperature with distance away from the ice 2007) will be experienced equally across the landscape. sheet. The largest differences in thermal sums were The interactive effect of zone and land cover type on between ES and vegetated areas. Differences within vege- soil temperature supported our hypothesis that vegetation tation types were smaller. These results indicate that soil is an important mediator of the belowground abiotic temperature is associated with vegetation, and that vege- conditions. When vegetation is present, soil and atmo- tation could mediate the impact of climate change on spheric temperatures are decoupled to an extent. In the belowground abiotic conditions and biological response. minimally vegetated ES, the soil temperature is more Other studies have found vegetation to be an important tightly linked to air temperature as observed by the determinant of soil temperature (Aalto, Le Roux, and e1420283-12 J. I. BRADLEY-COOK AND R. A. VIRGINIA Table 8. Comparison of models explaining soil nitrogen stocks. likely because these features are commonly confined to The top-ranked model, with vegetation as a main effect, is small areas on the landscape and therefore have a higher preferred. All other models, ΔAICc > 2. likelihood of accumulating error in the accuracy assess- AICc Log- ment process. Improvements in the accuracy metrics of Model likelihood Rank Predictor Variable k AICc ΔAICc Weight Log(L) our classification might be gained by the addition of 1 Vegetation 7 54.38 0 0.75 −18.5 randomly generated ground control data points that are 2 Vegetation + Zone 8 56.68 2.31 0.24 −18.2 at least three meters away from a vegetation boundary. 3 Veg + Zone + Veg:Zone 12 62.08 7.71 0.02 −13.7 4 Zone 3 78.66 24.28 0.00 −36.0 The buffer around the data point should reduce the influ- 5 Intercept 2 78.83 24.45 0.00 −37.3 ence of instrument error. Heindel, Chipman, and Virginia (2015) also conducted a land cover classification of an overlapping, but larger, study area. Their land cover composition estimates were Luoto 2013; Migala et al. 2014) that can alter heat comparable to ours. We attribute discrepancies in direct exchange from warming (Hollister et al. 2006), which comparison of our estimates to the fact that their study adds uncertainty to modeling soil organic carbon area extended farther west than ours, thus included a response to climate change (Xiong et al. 2015). greater proportion of land where wind erosion features are less common. These landscape classification maps are useful tools for the scaling of ecological variables such as Heterogeneity in Land Cover Class soil carbon and nitrogen stocks and are reflective of the natural variability in the landscape. The high heterogeneity of land cover types at small spatial scales would decrease the overall accuracy of the land cover classification. Our ground control points were Soil Chemistry Variation within the nine field sites, which were chosen because they had representative land cover classes within close Multivariate analysis of the soil chemistry revealed that proximity to each other. As a result, many of the vegeta- carbon, nitrogen, and soil moisture are the constrained tion patches were small (size ranged from 3 m to 40 m in variables associated with most of the variation among diameter) and close to boundaries with other vegetation soil samples. Soil C, N, and moisture are highly corre- types. In these conditions, error from handheld GPS units lated in other arctic terrestrial systems (e.g., combined with any small georeferencing errors can easily Hollingsworth et al. 2008) and are likely a result of a propagate and result in a misalignment of ground control positive feedback between soil conditions and primary points and the land cover classification and reduced accu- productivity. Additionally, soil moisture can constrain racy metrics. The ES class, with a distinct spectral signal microbial decomposition when soils are near saturation from vegetation, had a producer accuracy of 84 percent, as a result of limited oxygen availability in the soil which we consider to be fairly high given the possibility of matrix. Grasslands and fen soils have high values of error propagation mentioned above. Grassland and fen C, N, and moisture. Meanwhile, steppe, shrub, and vegetation had the lowest producer accuracy (Table 3), mixed vegetation, which account for a majority of the Figure 6. Average soil respiration rates in five vegetation types during the growing season. Error bars are ±1 standard error around the mean. Different letters indicate significant difference in Tukey HSD post hoc test (P < 0.05). ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-13 Figure 7. Soil organic carbon content (%) for each soil profile. Symbols correspond to the vegetation types (ES = eroded soil, ST = steppe, GL = grasslands, MX = mixed vegetation, FN = fen, SH = shrub). Profiles are grouped by sites in each panel, with Zone 1 sites in the top row, Zone 2 sites in the middle row, and Zone 3 sites in the bottom row. Distribution of Soil Organic Carbon at a Landscape land cover, have similar position relative to the axis Scale describing the variation in C, N, and moisture, but are differentiated on the second axis, which is asso- The landscape-wide average carbon stock weighted by −2 ciated with the C:N ratio. These results support our area is 16.41 kg C m , which is within the estimate hypothesis and other studies that shrub soils have assigned to west Greenland in the regional assessments −2 higher C:N than graminoid-dominated soils, and the of soil organic carbon, 10–25 kg C m (Hugelius et al. mixed vegetation ratio reflects a mixture of C:N inputs 2013). This average carbon stock is greater than that of and places these soils at an intermediate position. high Arctic field sites near Thule, Greenland, where Higher C:N in shrub soils has also been found in striped patterned ground soil contains SOC content of −2 mineral soils in the study area (Petrenko et al. 2016). 9.4 kg C m (Horwath et al. 2008). Other prominent The same shrub soils also have reduced soil respiration Arctic field areas, such as Toolik Lake Field Station on potential and temperature sensitivity compared to adja- the north slope of Alaska and Svalbard, have higher −2 cent grassland soils, which may be a result of nutrient SOC stocks that fall in the 25–50 kg C m range limitation on decomposition from lower nitrogen (Hugelius et al. 2013). stocks in shrub soils (Bradley-Cook et al. 2016). ES Soil carbon content varied across the landscape, and have the lowest position on the axis associated with C, has statistically significant associations with land cover N, and moisture (RCA1), which indicate that these soils type. The highest carbon stocks of the vegetation classes have limited biological development. These soils also sampled at multiple sites are associated with grasslands, −2 have a higher pH than other vegetation types, high- with an average of 34.87 kg C m . Petrenko et al. lighting that they have a unique chemical profile that (2016) estimated mineral soil carbon stocks from grass- −2 could further limit enzymatic activity and microbial land soils to be 29 kg C m , which is probably slightly decomposition (Min et al. 2014). below our estimate because they focused on mineral e1420283-14 J. I. BRADLEY-COOK AND R. A. VIRGINIA soil C dynamics and did not include the surface organic vegetation have lower carbon stocks per unit area (i.e., kg −2 horizon in their analysis. Grassland vegetation types Cm values) than fen and grassland, they collectively store only comprise 7 percent of the total study area, but it most of the total soil organic carbon (i.e., kg C values) contains approximately 15 percent of the landscape soil because they comprise a majority of the total vegetation carbon stock. The elevated in situ respiration rates from cover. Thus, shrub, steppe, and mixed vegetation are bio- grassland soils correspond with laboratory measure- geochemically important partitions of the Kangerlussuaq ments of carbon mineralization in which grassland tundra landscape. soils have a 1.8 times higher soil respiration than shrub soils (Bradley-Cook et al. 2016). Together, these Landscape Carbon Storage Response to findings indicate that, in this landscape, grasslands are Environmental Change hot spots for mid-summer soil respiration and for soil carbon storage. The pattern in soil carbon storage sug- To investigate the possible trajectory of soil carbon stocks in gests that soil carbon inputs from plant production response to vegetation and climate drivers at the landscape have exceeded soil carbon losses from respiration over scale, we considered the impact of shrub expansion and a long time frame. Soil respiration rates, however, cap- climate warming scenarios on landscape soil carbon. Given ture instantaneous dynamics, so the strength of the the assumption that a vegetation shift is accompanied by a carbon sink may be different than in the past. It is corresponding shift in soil properties, shrub expansion and possible that recent atmospheric warming has elevated prolonged occupation into graminoid-dominated areas carbon losses relative to uptake and storage. (steppe and grassland) should result in a shift from the The single fen soil profile measured had a soil carbon high carbon, low C:N soils associated with grassland vege- −2 stock of 54.6 kg C m , which is the highest stock measured tation to the lower carbon, high C:N soils associated with ofallsoilpitsinthisstudy.Andersonetal. (2009)found shrub vegetation. One possible mechanism for this poten- lake sediments in the Kangerlussuaq region to contain large tial shift is that shrub expansion could increase winter soil −2 carbon stocks (an average of 42 kg C m ) and suggest that respiration and soil nitrogen mobilization as predicted by these low-lying, mostly saturated soils are the most carbon- the snow-shrub interaction hypothesis (Sturm, Holmgren, rich areas in the landscape. The high carbon content of and McFadden 2001;Sturm et al. 2005). Under this sce- these soils, combined with seasonally dynamic hydrology, nario, a shift to greater shrubiness would reduce the land- indicates that soils such as these could be an important scape soil carbon stock and increase soil C:N. However, it is contributor to the landscape dynamics of the CO and also possible that the general characteristics of graminoid CH exchange between the terrestrial ecosystem and the soils could be preserved under shrub expansion scenarios. atmosphere. Total landscape carbon storage will be a function of the net change in soil and vegetation carbon stocks. While the The most common vegetation types were ranked lowest vegetation carbon stock increases with shrub expansion, with respect to carbon stock per unit area. While the this component of Arctic terrestrial carbon stock is a average carbon stocks of steppe, shrub, and mixed vegeta- small percentage of soil carbon stocks (McGuire et al. tion are at least 40 percent lower than grassland, they 2009), which suggests that the net change will be driven comprise a substantial portion (approximately 68%) of by soil carbon response to climate change. Our data show total landscape storage because of their areal extent. that growing season soil respiration and soil moisture are Estimates of near-surface shrub carbon stocks (Table 5) lowerinshrub soils. If thesefactors constrainsoilcarbon are within the range of comparable measurements in the −2 efflux we could expect to see a legacy effect of graminoid study area, 3.07–7.67 kg C m (Bradley-Cook and Virginia soils that could contribute to SOC heterogeneity associated 2016) and to shrub tundra stocks measured in Siberia, 7.1– −2 with shrub vegetation under projected vegetation change 7.8 kg C m (Hugelius and Kuhry 2009). Many studies in scenarios. Arctic tundra systems, which were conducted in low-lying The carbon losses associated with shrub expansion could tundra, have focused on fen and bog hot spots instead of be compounded with a warming climate. Microbial decom- soils with lower carbon stocks, because these features are position most often increases with temperature, and the common and comprise a majority of the carbon stocks (e.g., rate of increase can vary between soils. Previous research on Hugelius and Kuhry 2009). However, in systems such as soils in this study area shows that the temperature sensitiv- Kangerlussuaq, characterized by steep terrain and semiarid ity of decomposition is higher in grasslands than shrub soils conditions, low-lying grassland and fen vegetation cover a (Q-10 =2.3, Q-10 = 1.8), and that soil moisture relatively small proportion of the area. From our landscape- grasslands shrub increases respiration in grassland soils but not shrub soils level analysis of total soil carbon storage by vegetation type, (Bradley-Cook et al. 2016). These known temperature we found that even though shrub, steppe, and mixed ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-15 sensitivities suggest that SOC decomposition rates will temperature sensitive than in grassland soils. accelerate with warming for both shrub and grassland However, vegetation mediation of air temperature soils, so increased soil respiration can be expected under adds uncertainty to the effect of climate change on any vegetation scenario. belowground temperature affects on carbon cycling. Climate change is a key driver of shrub expansion in These landscape-level carbon storage, turnover, and tundra ecosystems (Myers-Smith et al. 2011), so it is sensitivities indicate that climate drivers and vegetation likely that atmospheric warming and shrub expansion dynamics will likely lead to a loss of stored soil carbon will co-occur. However, as our findings indicate, vege- and an increase of greenhouse gases flux into the tation mediates the relationship between air and soil atmosphere. temperature, so atmospheric warming may not result in a concomitant increase in soil temperature. For Acknowledgments instance, shrub canopy shading can have a local cooling effect on soil temperatures (Jean and Payette 2014; We are deeply grateful to Courtney Hammond Wagner and Sturm, Holmgren, and McFadden 2001). We did not Ruth Heindel for superb field assistance, to Leehi Yona for observe differences in soil temperature between shrub helping with sample processing and analysis, and to Paul Zietz for laboratory support. Furthermore, we thank Amy Burzynski and graminoid vegetation, so it is possible that canopy and Jonathan Chipman for contributions to the mapping and shading does not have a sufficient effect on insulation spatial analysis, Thomas Kraft for guidance on the ordination to modify soil temperature in the low shrub tundra of analysis, Ruth Heindel for providing comments on the manu- this study area (Hollingsworth et al. 2008). It is also script, and all Virginia Lab members for moral support. We possible that other factors, such as aspect, insulation are thankful to CH2M Hill Polar Services for unfaltering logistical support in Kangerlussuaq that enabled productive from understory moss, or plant structural feature such fieldwork. This research was funded by an NSF grant (Award as leaf area index, have a stronger effect on soil tem- #: 0801490) to Ross Virginia, with additional support from the perature than small changes in solar insolation. In Institute of Arctic Studies, in the Dickey Center for summary, we posit that landscape soil carbon storage International Understanding at Dartmouth. is sensitive to both vegetation change and climate warming, with possible compounding impacts on soil carbon storage in a future in which these changes may References occur concurrently. Aalto, J., P. C. Le Roux, and M. Luoto. 2013. Vegetation mediates soil temperature and moisture in Arctic-Alpine environments. Arctic, Antarctic, and Alpine Research Conclusions 45:429–39. doi:10.1657/1938-4246-45.4.429. Anderson, N. J., W. D’andrea, and S. C. Fritz. 2009. Holocene The Kangerlussuaq tundra landscape has heteroge- carbon burial by lakes in SW Greenland. Global Change neous land cover and a growing season air temperature Biologic 15:2590–98. doi:10.1111/j.1365-2486.2009.01942.x. gradient extending from the GrIS. We estimate average Borcard, D., F. Gillet, and P. Legendre. 2011. Numerical −2 landscape carbon stock to be 16.41 kg C m and ecology with R, 292. New York: Springer. −2 nitrogen stock to be 1.07 kg N m . We found differ- Bosatta, E., and G. Ågren. 1999. Soil organic matter quality interpreted thermodynamically. Soil Biology & Biochemistry ences in carbon stocks and soil respiration by vegeta- 31:1889–91. tion type, with enhanced carbon storage and soil Bradley-Cook, J. I., C. L. Petrenko, A. J. Friedland, and R. A. respiration in grassland soils. Combined with the Virginia. 2016. Temperature sensitivity of mineral soil understanding that these soils are more sensitive to carbon decomposition in shrub and graminoid tundra, temperature and moisture, we conclude that grasslands west Greenland. Climate Change Responses 1–15. are highly sensitive hot spots of soil carbon storage and doi:10.1186/s40665-016-0016-1. Bradley-Cook, J. I., and R. A. Virginia. 2016. Soil carbon mid-growing season soil respiration within a shrub- storage, respiration potential, and organic matter quality dominated matrix. Even though shrub, steppe, and across an age and climate gradient in southwestern mixed vegetation have lower carbon stocks per unit Greenland. Polar Biology 39:1283–95. doi:10.1007/s00300- area, they are still important components of land- 015-1853-2. scape-level biogeochemical cycling because they cover Cahoon, S. M. P., P. F. Sullivan, E. Post, and J. M. Welker. 2011. Large herbivores limit CO2 uptake and suppress more than 50 percent of the area. Shrub expansion into carbon cycle responses to warming in West Greenland. graminoid-dominated areas could possibly result in Global Change Biologic 18:469–79. doi:10.1111/j.1365- lower landscape soil carbon storage in the long term. 2486.2011.02528.x. Warming could drive compounding losses of stored Callaghan, T. V., L. Bjorn, F. S. Chapin, Y. Chernov, T. carbon, although organic carbon may be more stable Christensen, B. Huntley, R. Ims, M. Johansson, D. Jolly, S. in shrub soils, where decomposition rates are less Jonasson, N. Matveyeva, W. Oechel, N. Panikov, and G. e1420283-16 J. I. BRADLEY-COOK AND R. A. VIRGINIA Shaver. 2005. Arctic tundra and polar desert ecosystems. In Black Spruce Ecosystems. Ecosystems 11:629–42. Arctic climate impact assessment, eds. C. Symon, L. Arris, and doi:10.1007/s10021-008-9147-y. B. Heal, 244–352. New York: Cambridge University Press. Hollister, R. D., P. J. Webber, F. E. Nelson, and C. E. Chapin, F., M. Bret-Harte, S. Hobbie, and H. Zhong. 1996. Tweedie. 2006. Soil thaw and temperature response to air Plant functional types as predictors of transient responses warming varies by plant community: Results from an of arctic vegetation to global change. Journal of Vegetation open-top chamber experiment in northern Alaska. Arctic, Science 7:347–58. Antarctic, and Alpine Research 38:206–15. Chapin, F., J. McFarland, A. D. McGuire, E. S. Euskirchen, R. Hooper, D., and P. Vitousek. 1998. Effects of plant composi- W. Ruess, and K. Kielland. 2009. The changing global tion and diversity on nutrient cycling. Ecological carbon cycle: Linking plant-soil carbon dynamics to global Monographs 68:121–49. consequences. Journal of Ecology 97:840–50. doi:10.1111/ Horwath Burnham, J., and R. Sletten. 2010. Spatial distribution j.1365-2745.2009.01529.x. of soil organic carbon in northwest Greenland and under- Craine, J., R. Spurr, K. Mclauchlan, and N. Fierer. 2009. estimates of high Arctic carbon stores. Global Biogeochemical Landscape-level variation in temperature sensitivity of Cycles 24:GB3012. doi:10.1029/2009GB003660. soil organic carbon decomposition. Soil Biology & Horwath, J. L., R. S. Sletten, B. Hagedorn, and B. Hallet. 2008. Biochemistry 42:373–75. doi:10.1016/j.soilbio.2009.10.024. Spatial and temporal distribution of soil organic carbon in Creamer, C. A., T. R. Filley, T. W. Boutton, S. Oleynik, and I. nonsorted striped patterned ground of the High Arctic. B. Kantola. 2011. Controls on soil carbon accumulation Journal Of Geophysical Research-Biogeosciences 113: during woody plant encroachment: Evidence from physical G03S07. doi:10.1029/2007JG000511. fractionation, soil respiration, and 13-C of respired CO2. Hudson, J. M. G., G. H. R. Henry, and W. K. Cornwell. 2011. Soil Biology and Biochemistry 43:1678–87. doi:10.1016/j. Taller and larger: Shifts in Arctic tundra leaf traits after soilbio.2011.04.013. 16 years of experimental warming. Global Change Biology Davidson, E. A., and I. A. Janssens. 2006. Temperature sensi- 17:1013–21. doi:10.1111/j.1365-2486.2010.02294.x. tivity of soil carbon decomposition and feedbacks to climate Hugelius, G., and P. Kuhry. 2009. Landscape partitioning and change. Nature 440:165–73. doi:10.1038/nature04514. environmental gradient analyses of soil organic carbon in a Dijkmans, J., and T. Törnqvist. 1991. Modern periglacial eolian permafrost environment. Global Biogeochemical Cycles 23: deposits and landforms in the Søndre Strømfjord area, West GB3006. doi:10.1029/2008GB003419. Greenland and their palaeoenvironmental implications. Hugelius, G., J. Strauss, S. Zubrzycki, J. W. Harden, E. A. G. Meddelelser Om Grønland. Geoscience 25:3–39. Schuur, C.-L. Ping, L. Schirrmeister, G. Grosse, G. J. DMI (Danish Meterological Institute). 2017. Klimanormaler for Michaelson, C. D. Koven, J. A. O’Donnell, B. Elberling, Grønland, Copenhagen, Denmark. Accessed June 4, 2017. U. Mishra, P. Camill, Z. Yu, J. Palmtag, and P. Kuhry. http://www.dmi.dk/groenland/arkiver/klimanormaler/. 2014. Estimated stocks of circumpolar permafrost carbon Fierer, N., B. Colman, and J. Schimel. 2006. Predicting the with quantified uncertainty ranges and identified data temperature dependence of microbial respiration in soil: A gaps. Biogeosciences 11:6573–93. doi:10.5194/bg-11-6573- continental-scale analysis. Global Biogeochemical Cycles 20: 2014-supplement. GB3026. doi:10.1029/2005GB002644. Hugelius, G., C. Tarnocai, G. Broll, J. G. Canadell, P. Kuhry, and Frost, G. V., and H. E. Epstein. 2014. Tall shrub and tree D. K. Swanson. 2013. The Northern circumpolar soil carbon expansion in Siberian tundra ecotones since the 1960s. database: Spatially distributed datasets of soil coverage and Global Change Biology 20:1264–77. doi:10.1111/gcb.12406. soil carbon storage in the northern permafrost regions. Earth Haddix, M. L., A. F. Plante, R. T. Conant, J. Six, J. M. Steinweg, Systems Sciences Data 5:3–13. doi:10.5194/essd-5-3-2013. K. Magrini-Bair, R. A. Drijber, S. J. Morris, and E. A. Paul. Jean, M., and S. Payette. 2014. Effect of vegetation cover on 2011. The role of soil characteristics on temperature sensi- the ground thermal regime of wooded and non-wooded tivity of soil organic matter. Soil Science Society Of America palsas. Permafrost and Periglacial Processes 25:281–94. Journal 75:56–68. doi:10.2136/sssaj2010.0118. doi:10.1002/ppp.1817. Hanlon, B., and B. Larget. 2011. Assumptions and transfor- Jenny, H. 1941. Factors of soil formation. New York: Dover mations. Madison, Wisconsin: Department of Statistics. Publications. Accessed May 20, 2017. http://www.stat.wisc.edu/~st571- Jones, A., V. Stolbovoy, C. Tarnocai, G. Broll, O. Spaargaren, 1/11-assumptions-2.pdf. and L. Montarella. 2009. Soil Atlas of the Northern Heindel, R. C., J. W. Chipman, and R. A. Virginia. 2015. The Circumpolar Region, 1. Luxembourg: European spatial distribution and ecological impacts of Aeolian soil Commission, Office for Official Publications of the erosion in Kangerlussuaq, West Greenland. Annals of the European Communities. Association of American Geographers 105:875–90. Jørgensen, R. H., H. Meilby, and J. Kollmann. 2013. Shrub doi:10.1080/00045608.2015.1059176. expansion in SW Greenland under modest regional warm- Hinzman, L. D., C. J. Deal, A. D. McGuire, S. H. Mernild, I. V. ing: Disentangling effects of human disturbance and graz- Polyakov, and J. E. Walsh. 2013. Trajectory of the Arctic as ing. Arctic, Antarctic, and Alpine Research 45:515–25. an integrated system. Ecological Applications 23:1837–68. doi:10.1657/1938-4246-45.4.515. Hobbie, S. 1996. Temperature and plant species control over Levy, L. B., M. A. Kelly, J. A. Howley, and R. A. Virginia. litter decomposition in Alaskan tundra. Ecological 2012. Age of the Ørkendalen moraines, Kangerlussuaq, Monographs 66:503–22. Greenland: Constraints on the extent of the southwestern Hollingsworth, T. N., E. A. G. Schuur, F. S. Chapin III, and margin of the Greenland Ice Sheet during the Holocene. M. D. Walker. 2008. Plant community composition as a Quaternary Science Reviews 52:1–5. doi:10.1016/j. predictor of regional soil carbon storage in Alaskan Boreal quascirev.2012.07.021. ARCTIC, ANTARCTIC, AND ALPINE RESEARCH e1420283-17 McGuire, D. A., L. G. Anderson, T. R. Christensen, S. shrub tundra sites, western Greenland. Arctic Science Dallimore, L. Guo, D. J. Hayes, M. Heimann, T. D. 2:165–82. doi:10.1139/as-2015-0023. Lorenson, R. D. Macdonald, and N. Roulet. 2009. Post, E., and M. C. Forchhammer. 2008. Climate change Sensitivity of the carbon cycle in the Arctic to climate reduces reproductive success of an Arctic herbivore through change. Ecological Monographs 79:523–55. trophic mismatch. Philosophical Transactions of the Royal Mernild, S. H., E. Hanna, J. R. McConnell, M. Sigl, A. P. Society B 363:2367–73. doi:10.1098/rstb.2007.2207. Beckerman, J. C. Yde, J. Cappelen, J. K. Malmros, and K. Post, E., and C. Pedersen. 2008. Opposing plant community Steffen. 2015. Greenland precipitation trends in a long- responses to warming with and without herbivores. term instrumental climate context (1890–2012): Proceedings of the National Academy of Sciences 105:12353– Evaluation of coastal and ice core records. International 58. doi:10.1073/pnas.0802421105. Journal Climatology 35:303–20. doi:10.1002/joc.3986. Schmidt, M., M. Torn, S. Abiven, and T. Dittmar. 2011. Migala, K., B. Wojtun, W. Szymanski, and P. Muskala. 2014. Soil Persistence of soil organic matter as an ecosystem prop- moisture and temperature variation under different types of erty. Nature 478:49–56. doi:10.1038/nature10386. tundra vegetation during the growing season: A case study Sierra, C. A., S. E. Trumbore, E. A. Davidson, S. Vicca, and I. from the Fuglebekken catchment, SW Spitsbergen. Catena Janssens. 2015. Sensitivity of decomposition rates of soil 116:10–18. doi:10.1016/j.catena2013.12.007. organic matter with respect to simultaneous changes in Min, K., C. A. Lehmeier, F. Ballantyne, A. Tatarko, and S. A. temperature and moisture. Journal Advancement Model Billings. 2014. Differential effects of pH on temperature sen- Earth Systems 7:335–56. doi:10.1002/2014MS000358. sitivity of organic carbon and nitrogen decay. Soil Biology & Sollins, P., C. Glassman, E. A. Paul, C. Swanston, K. Lajtha, Biochemistry 76:193–200. doi:10.1016/j.soilbio.2014.05.021. and J. W. Heil. 1999. C and N Dry Combustion. In Myers-Smith, I. H., S. C. Elmendorf, P. S. A. Beck, M. Standard Soil Methods for Long-Term Ecological Research, Wilmking, M. Hallinger, D. Blok, K. D. Tape, S. A. eds. G. P. Robertson, D. C. Coleman, C. Bledsoe, and P. Rayback, M. Macias-Fauria, B. C. Forbes, J. D. M. Speed, Sollins, 92–95. Oxford: Oxford University Press. N. Boulanger-Lapointe, C. Rixen, E. Levesque, N. M. Stendel, M.,J.H.Christensen, G.Aoalgeirsdottir,N. Kliem,and Schmidt, C. Baittinger, A. J. Trant, L. Hermanutz, L. S. M. Drews. 2007. Regional climate change for Greenland and Collier, M. A. Dawes, T. C. Lantz, S. Weijers, R. H. surrounding seas. Danish Climate Centre Report 07-02, Jørgensen, A. Buchwal, A. Buras, A. T. Naito, V. Danish Meteorological Institute, Copenhagen, Denmark, 26. Ravolainen, G. Schaepman-Strub, J. A. Wheeler, S. Wipf, Accessed March 15, 2015. http://978-87-7478-547-7. K. C. Guay, D. S. Hik, and M. Vellend. 2015. Climate Sturm, M., J. Holmgren, and J. McFadden. 2001. Snow-shrub sensitivity of shrub growth across the tundra biome. interactions in Arctic tundra: A hypothesis with climatic Nature Climate Change. doi:10.1038/nclimate2697. implications. Journal of Climate 14:336–44. Myers-Smith, I. H., B. C. Forbes, M. Wilmking, M. Hallinger, Sturm, M., J. Schimel, G. Michaelson, and J. Welker. 2005. T. Lantz, D. Blok, K. D. Tape, M. Macias-Fauria, U. Sass- Winter biological processes could help convert arctic tun- Klaassen, E. Levesque, S. Boudreau, P. Ropars, L. dra to shrubland. BioScience 55:17–26. Hermanutz, A. Trant, L. S. Collier, S. Weijers, J. Rozema, Sullivan, P. F., S. J. T. Arens, R. A. Chimner, and J. M. S. A. Rayback, N. M. Schmidt, G. Schaepman-Strub, S. Welker. 2008. Temperature and microtopography interact Wipf, C. Rixen, C. B. Menard, S. Venn, S. Goetz, L. to control carbon cycling in a high arctic fen. Ecosystems Andreu-Hayles, S. Elmendorf, V. Ravolainen, J. Welker, 11:61–76. doi:10.1007/s10021-007-9107-y. P. Grogan, H. E. Epstein, and D. S. Hik. 2011. Shrub Tamstorf, M. P., P. J. Aastrup, and L. C. Cuyler. 2005. Modelling expansion in tundra ecosystems: Dynamics, impacts and critical caribou summer ranges in West Greenland. Polar research priorities. Environmental Research Letters Biology 28:714–24. doi:10.1007/s00300-005-0731-8. 6:045509. doi:10.1088/1748-9326/6/4/045509. Thomsen, I. K., B. M. Petersen, S. Bruun, L. S. Jensen, and B. Myers-Smith, I. H., and D. S. Hik. 2013. Shrub canopies T. Christensen. 2008. Estimating soil C loss potentials from influence soil temperatures but not nutrient dynamics: the C to N ratio. Soil Biology and Biochemistry 40:849–52. An experimental test of tundra snow-shrub interactions. doi:10.1016/j.soilbio.2007.10.002. Ecology and Evolution 3:3683–700. doi:10.1002/ece3.710. Urban, M., M. Forkel, J. Eberle, C. Huettich, C. Schmullius, and Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. M. Herold. 2014. Pan-Arctic climate and land cover trends Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. derived from multi-variate and multi-scale analyses (1981– H. Stevens, and H. Wagner 2015. vegan: Community ecol- 2012). Remote Sensing 6:2296–316. doi:10.3390/rs6032296. ogy package. http://CRAN.R-project.org/package=vegan. Weatherspark. 2015. Average Weather For Kangerlussuaq Ostle, N. J., P. Smith, R. Fisher, F. L. Woodward, J. B. Fisher, (Søndre Strømfjord), Greenland. Accessed August 6, J. U. Smith, D. Galbraith, P. Levy, P. Meir, N. P. 2015. https://weatherspark.com/averages/27554/ McNamara, and R. D. Bardgett. 2009. Integrating plant- Kangerlussuaq-S-ndre-Str-mfjord-Kitaa-Greenland. soil interactions into global carbon cycle models. Journal of Xiong, X., S. Grunwald, D. B. Myers, J. Kim, W. G. Harris, Ecology 97:851–63. doi:10.1111/j.1365-2745.2009.01547.x. and N. Bliznyuk. 2015. Assessing uncertainty in soil Petrenko, C. L., J. I. Bradley-Cook, E. M. Lacroix, A. J. organic carbon modeling across a highly heterogeneous Friedland, and R. A. Virginia. 2016. Comparison of carbon landscape. Geoderma 251–252:105–16. doi:10.1016/j. and nitrogen storage in mineral soils of graminoid and geoderma.2015.03.028.

Journal

"Arctic, Antartic and Alpine Research"Taylor & Francis

Published: Jan 1, 2018

Keywords: Soil organic carbon; landscape heterogeneity; tundra; soil respiration; soil temperature

There are no references for this article.