STUDY AREA

The study was conducted during the 2008 growing season near the Tundra Ecosystem Research Station at Daring Lake, Northwest Territories (NWT) (64°52′N, 111°34′W), located 300 km northeast of Yellowknife, NWT, and approximately 70 km north of the treeline. The region is classified as low-arctic shrub tundra, which is broken up by frequent eskers and hydrologic features. Local elevation ranges from 414 to 470 m above sea level. Soils in the region typically have a thin surface organic horizon in drier areas (less than 0.1 m) and a thicker organic horizon in wetter areas (up to 0.45 m), overlying a coarse mineral layer (Nobrega and Grogan, 2007, 2008). The area is underlain by continuous permafrost with an active layer typically ranging from 0.3 to 1.0 m. Climate in this region is characterized by long cold winters with a short growing season (mid-June to early September). The mean annual temperature ranges from −10.0 °C to −12.5 °C, and mean total annual precipitation ranges between 200 and 300 mm (Lafleur and Humphreys, 2008). Climate records for 1997–2008 from the Daring Lake weather station indicate growing season (June–August) daily mean temperature is 10.5 (±2.8) °C, with average precipitation of 31.2 (±24.0) mm month⁻¹ and daily mean solar radiation of 222.3 (±51.1) W m⁻².

CO₂ EXCHANGE MEASUREMENTS

An infrared gas analyzer (LICOR 840, Lincoln, Nebraska) and a custom-built, closed-flow portable chamber system were used to measure CO₂ and H₂O concentrations. The clear acrylic chamber (0.4 × 0.4 × 0.4 m  =  64 L) was fitted with a small fan to continually mix the air inside the chamber and 2 handles for ease of transport. Air from the chamber was drawn into the gas analyzer at a rate of 0.9 L minute⁻¹ by a pump (model UN89, KNF Neuberger, Trenton, New Jersey) and was returned to the chamber through a port at the top of the chamber. The lip of the chamber was lined with 5-mm-thick foam, and the chamber was held in place with 4 small clamps on aluminum collars (0.4 m × 0.4 m) that were inserted into the soil. Air temperature inside the chamber (T_cham) was measured with a 0.35 mm copper-constantan thermocouple mounted on the inside of the chamber. CO₂, H₂O concentrations, and T_cham were recorded every 2 s for 150 s with a CR21X datalogger (Campbell Scientific, Logan, Utah). The first 10 s of data were discarded due to the disturbance of the placement of the chamber. Sampling took place first under ambient light conditions, then under reduced incident sunlight, under three successive layers of netting, then in complete darkness using a thick black plastic shroud, for a total of 5 measurements per collar. Following each measurement the chamber was lifted and turned into the wind for ∼20 s to restore ambient conditions.

Net ecosystem CO₂ exchange (NEE) was calculated according to the following equation, after Shaver et al. (2007):

Samples taken in full darkness were assumed to represent ecosystem respiration (ER). ER values were subtracted from NEE values (samples taken under ambient light) to determine gross ecosystem productivity (GEP). Throughout this paper, the micrometeorological sign convention is used, where flux from atmosphere to tundra is negative and flux from the tundra to the atmosphere is positive.

Photosynthetic photon flux density (PPFD, µmol m⁻² s⁻¹) was detected with a quantum sensor (Kipp and Zonen PAR-LITE, Campbell Scientific, Logan, Utah). To prevent this sensor from blocking radiation entering the chamber, it was mounted on a small wooden platform ∼0.4 m in height placed within 1 m of the chamber and logged every 2 s during sampling on the CR21X datalogger. The platform and sensor were covered with shade cloths identical to those covering the chamber during the light reduction sampling. Additionally, because the Plexiglas chamber reduced PPFD reaching vegetation by ∼7%, measured PPFD values were reduced to reflect this shading effect. Light response curves were fitted to the data according to the following equation:

Supporting measurements included air temperature (T_air), which was measured with a 0.35 mm copper-constantan thermocouple inside a radiation shield, mounted on the same platform as the quantum sensor and logged every 2 s on the CR21X. Volumetric soil water content (VWC) was measured with a SMC probe (Hydrosense, Campbell Scientific Inc., Logan, Utah). To avoid repeated probing of the soil inside the collar, 4 soil moisture measurements were taken just outside the collar and the results were averaged. Copper-constantan thermocouples were inserted 5 cm deep into the soil in the center of each plot and soil temperature (T_soil ) was measured with a digital thermocouple thermometer (model HH501, Omega Engineering, Stamford, Connecticut). Prior to the start of sampling, all thermocouples employed in the study (n  =  40) were intercompared and no significant differences were found (p  =  0.886).

Forty plots were established in vegetation communities defined by the most abundant plant species or family and by soil moisture regime. Due to logistical restrictions, sample sizes were not equal for all groups. Collars were inserted into the soil by cutting a slot with a serrated knife in the soil around the outer edge of the collar, and pressing it 6–10 cm into the soil organic layer. CO₂ flux sampling did not begin until 2 days after collars were inserted to allow CO₂ released from disturbance of the soil to dissipate. Plots were sampled in a random order, with all plots sampled every 5–6 days, for a total of 8 sample sessions for each collar. CO₂ sampling occurred between 5 July and 12 August 2008, between 10:00 and 16:00 M.D.T., in all weather conditions except precipitation events. Vegetation groups (n  =  number of collars) were defined as follows (see Table 1 for further vegetation and soil characteristics):

TABLE 1

Soil characteristics of seven tundra vegetation groups at Daring Lake, Northwest Territories (NWT), from 5 July to 13 August, 2008. The minimum-maximum of all data points is presented in parentheses. N-values indicate the number of CO₂ sampling sessions. Common superscripts indicate no significant difference between groups (Tukeys HSD, p < 0.05). VWC  =  volumetric soil water content.

VEGETATION ANALYSIS

After the flux sampling was completed, all above-ground biomass, including the top green layer of moss, was harvested from the collars. Non-living matter was removed, vascular plant material (leaves and stems) was sorted by species and non-vascular plant material was divided in mosses or lichens. Biomass was dried in a plant press and later dried again at 100 °C for 48 hours before mass was determined to the nearest 0.001 g. To determine leaf area index (LAI (m[leaf]² m[ground]⁻²), additional vegetation plots (n  =  10) were harvested throughout the sampling period. All leaf material was harvested from a 0.4 m × 0.4 m plot and sorted by species. Leaf material was laid flat and traced on 1 mm² graph paper, total area per unit ground area was summed, and material was dried later in a 100 °C oven for 48 hours before mass was determined to the nearest 0.001 g. Three (3) samples of dry leaf material of the dominant species for each group (listed in Table 1) were powdered and combusted with tungsten in a Macro Elemental Analyzer (Elementar Americas, Inc, St Laurel, New Jersey) to determine nitrogen content by mass (g N g leaf⁻¹). Average species-specific nitrogen content values were calculated with specific leaf area (m[leaf]² g[leaf mass]⁻¹) and LAI (m[leaf]² m[ground]⁻²) to obtain total foliar nitrogen (TFN; [g N m(ground)⁻²]).

STATISTICAL METHODS

Statistical analyses for CO₂ flux were performed for T_cham and T_air, soil moisture, biomass, leaf N content, LAI, and plant community as independent variables. Differences in the normally distributed soil and vegetation characteristics data were determined with an ANOVA and Tukeys post-hoc test. CO₂ flux data were normalized with log transformations, and the results presented below have been back-transformed and to the original units. The unbalanced sampling design (different number of collars between groups) necessitated taking an average of the flux measurements. All flux data were analyzed for linear relationships with sampling date and no significant trends were found. Hence, data for each vegetation community were pooled and compared. Differences in flux between groups were determined with a generalized F-test (Weerahandi, 1995) due to heteroscedastic variance. Linear regressions were performed for all other analyses. GEP and NEE were estimated at 500 µmol m⁻² s⁻¹ PPFD using the fitted parameter estimates found in Equation 2 (above). All statistical analysis was performed using STATISTICA (Statsoft, 2007) software.

ENVIRONMENTAL CONDITIONS

Daily temperature and daily solar radiation during June to August 2008 were close to the 10-year average (10.4 [±3.3 standard deviation] °C and 203.7 [±59.5] W m⁻², respectively). The same period was slightly drier than normal, with an average precipitation of 19.8 (±11.8) mm month⁻¹. During the sampling period (5 July to 12 August) the average daily temperature was 14.4 °C, with a maximum of 27.5 °C and a minimum of 2.5 °C. Total precipitation during the sampling period was 33.6 mm.

Across the sampling period soil moisture ranged from 4% VWC in dry heath to saturated (100% VWC) in the wet sedge, with wet tussock showing the greatest variability (Table 1). Mean soil moisture was significantly different between comparative groups (between dry and mesic birch, dry and wet tussock, mesic and dry heath) and soil moisture in wet sedge was significantly higher than all other groups except wet tussock (Table 1). Soil temperature in mesic birch and mesic heath was significantly lower than dry birch and dry heath (Table 1).

Total biomass was variable, ranging from 223 g dry mass m⁻² in the wet sedge plots to more than two and a half times greater (580 g dry mass m⁻² ) in the dry tussock plots (Table 2). Vascular biomass and LAI showed a similar trend, with the lowest values in wet sedge and highest in dry and mesic birch (Table 2). Moss biomass ranged widely across vegetation groups, from negligible in dry birch to a thick layer that was equivalent to the vascular biomass in dry and wet tussock and wet sedge (Table 2).

TABLE 2

Vegetation characteristics of seven tundra vegetation groups at Daring Lake, NWT, from destructive harvest on 13–15 August 2008. Values presented in parentheses are standard error of the mean. N-values indicate the number of collars in each group. Common superscripts indicate no significant difference between groups (Tukeys HSD, p < 0.05).

AMBIENT-LIGHT CO₂ EXCHANGE

Mean ambient-light GEP was greatest in mesic birch and lowest in the dry heath (Fig. 1). Similarly, the greatest NEE was also in mesic birch and lowest in dry heath and dry tussock (Fig. 1). Mean ER was greatest in wet tussock and lowest in wet sedge (Fig. 1). Comparisons of mean fluxes between hydrologically distinct groups with similar vegetation showed that ER was more variable within groups than GEP and NEE (Fig. 1). In the wet or mesic groups ER was significantly higher than in their dry counterparts (i.e., mesic birch ER was greater than dry birch, wet tussock ER was greater than dry tussock, and mesic heath ER was greater than dry heath) (Fig. 1). GEP and NEE in mesic birch were significantly higher than in dry birch. In wet tussock GEP was significantly higher than dry tussock, but NEE was not significantly different. Though GEP and NEE in mesic heath were larger than for dry heath, they were not significantly different (Fig. 1). Wet sedge plots were considered to be unique due to constantly saturated soil conditions and their distinct vegetative growth form, and thus were not compared with other groups.

FIGURE 1

Mean ambient-light carbon dioxide (CO₂) exchange for tundra vegetation groups for 5 July to 12 August 2008. Filled bars indicate gross ecosystem productivity, unfilled bars indicate ecosystem respiration, and hatched bars indicate net ecosystem CO₂ exchange (μmol m⁻² s⁻¹). Whiskers indicate standard deviations. Common superscripts indicate no significant difference between vegetation groups (n  =  230–249; Games-Howell test, p < 0.05).

LIGHT-ADJUSTED CO₂ EXCHANGE

GEP estimated at 500 µmol m⁻² s⁻¹ PPFD (GEP₅₀₀), ranged from −2.5 in wet tussock to −0.9 µmol m⁻² s⁻¹ in dry heath. NEE estimated at 500 µmol m⁻² s⁻¹ PPFD (NEE₅₀₀), ranged from −1.3 in mesic birch to −0.38 µmol m⁻² s⁻¹ in dry heath (Table 3). Neither NEE₅₀₀ or GEP₅₀₀ showed a significant relationship with leaf area or with soil moisture. However, when fluxes were adjusted for leaf area across the range of soil moisture there was a significant negative linear relationship (Fig. 2), indicating that CO₂ uptake per unit leaf was greater with higher soil moisture.

FIGURE 2

Linear regressions between mean soil volumetric water content (%) and (a) gross ecosystem productivity at 500 μmol m⁻² s⁻¹ of light (F_1,6  =  15.59, R²  =  0.757, p  =  0.01), and (b) net ecosystem CO₂ exchange at 500 μmol m⁻² s⁻¹ of light (F_1,6  =  36.35, R²  =  0.855, p < 0.01) for all vegetation groups. CO₂ exchange values were adjusted for leaf area (μmol m[leaf]⁻² s⁻¹). Whiskers indicate standard error of the mean.

TABLE 3

Parameter estimates of the rectangular hyperbola light response curve between gross ecosystem productivity and light availability using non-linear, least-squares regressions with the Quasi-Newton method (custom loss set as [(OBS-PRED)² + (P_max > 0)*100 + (α < 0)*100] and start values set at 0.1). NEE₅₀₀ and GEP₅₀₀ indicate gross ecosystem productivity and net ecosystem CO₂ exchange (μmol m⁻² s⁻¹) at 500 μmol m⁻² s⁻¹ light, and P_max indicates the maximum rate of photosynthesis.

There was a significant negative linear relationship between TFN and soil moisture (F_1,6  =  15.37, R²  =  0.755, p  =  0.01; Fig. 3a) and a significant positive relationship between leaf N concentration and soil moisture (F_1,6  =  6.55, R²  =  0.466, p  =  0.05; Fig. 3b). Thus, despite the higher concentration of N in foliage at the upper end of the soil moisture gradient, the total N in foliage was low, because the quantity of foliage was small. When fluxes were adjusted for TFN and compared across the range of soil moisture, it was found that GEP₅₀₀ showed a significant negative linear relationship (F_1,6  =  16.22, R²  =  0.764, p  =  0.01; Fig. 3c), but no trend was found for NEE₅₀₀ (F_1,6  =  1.06, p  =  0.34; Fig. 3d).

FIGURE 3

Linear regressions between mean soil volumetric water content (%) and (a) mean total foliar nitrogen (TFN) (g N m[ground]⁻²); (b) mean leaf N concentration (g N m[leaf]⁻²); (c) GEP₅₀₀ adjusted for mean TFN (μmol s⁻¹ g N⁻¹); and (d) NEE₅₀₀ adjusted for mean TFN (μmol s⁻¹ g N⁻¹), for all vegetation groups. Whiskers indicate standard error of the mean.

TEMPERATURE AND MOISTURE EFFECTS ON RESPIRATION

T_air and T_cham were significantly correlated for all samples (n  =  242, r  =  0.778, p < 0.001) and even more strongly correlated for ER measurements (samples taken in darkness) (n  =  224, r  =  0.852, p < 0.001); thus air temperature was not treated as a significant variable because of this co-linearity. ER dependence on temperature was found only in the dry and mesic groups. This suggests an interactive effect between temperature and moisture up to a threshold level. Once soils were saturated (85% VWC or higher), no significant temperature effects were found. There was a significant exponential relationship between T_cham and ER in dry soil types (dry heath: F_1,43  =  12.516, R²  =  0.220, p  =  0.02; dry birch: F_1,17  =  14.318, R²  =  0.252, p  =  0.02; dry tussock: F_1,31  =  8.622, R²  =  0.129, p  =  0.04). The effects of T_soil were also found in mesic soil moisture conditions (mesic birch: F_1,34  =  6.564, R²  =  0.153, p  =  0.018; mesic heath: F_1,21  =  7.31, R²  =  0.122, p  =  0.037).

When ER was adjusted with residual analysis to remove the effects of temperature mentioned above, there was a non-linear trend across the range of soil moisture (Fig. 4). The impact of vegetation type on ER was evident, particularly in tussock groups. ER in wet tussock was higher and more variable than ER in wet sedge around 90% VWC, and ER in dry tussock was more variable than other groups in the same moisture range (Fig. 4).

FIGURE 4

Temperature-adjusted ecosystem respiration (μmol m⁻² s⁻¹) and soil moisture (%VWC) for all vegetation groups (Spearmans r  =  0.45, p  =  0.038).

CO₂ EXCHANGE BETWEEN VEGETATION GROUPS

Comparisons among vegetation groups of similar biomass and species composition, but different soil moisture regimes, showed that while GEP and ER were generally larger in wet and mesic tundra types, mean NEE was significantly different between mesic and dry plots for only one vegetation type (birch; Fig. 1). Higher productivity in the wet or mesic groups as compared to the dry groups may be partly attributed to the presence of the thick moss layer, which was not present in the dry plots. Although this was not evident from visual inspection, the dry groups are found in conditions that are drier than optional for these specific communities, which can frequently lead to water stress (Illeris et al., 2004). Further, dry groups tend to be found at higher topographic positions where snow cover is lower and vegetation is exposed to the wind, leading to foliar damage during winter, which can decrease summer productivity (Shaver and Chapin, 1991). The limited variability in NEE between groups is due to the covariance of GEP and ER. Increased ER has been associated with stimulated GEP under both greenhouse warming and fertilization treatments (Shaver et al., 1998; Boelman et al., 2003) and is often attributed to the large contribution of plant respiration to ER at peak season (Johnson et al., 2000). Others have suggested that increased biomass is correlated with a higher rate of litter production, leading to a higher heterotrophic respiration component of ER (Boelman et al., 2003).

The pattern in wet sedge CO₂ exchange was anomalous compared to the other groups. Wet sedge exhibited a high mean NEE due to a high GEP and relatively low ER, similar to findings of Shaver et al. (2007) and Nobrega and Grogan (2008). The high GEP per unit leaf area (Fig. 2a) may be explained by the high leaf N concentration (Fig. 3a). The low ER is expected in saturated soils because CO₂ production is limited by low oxygen concentration in the soil (Oechel and Billings, 1992).

CO₂ EXCHANGE ACROSS THE SOIL MOISTURE GRADIENT

The effects of both soil moisture and vegetation community can be seen in the trends of CO₂ exchange across all tundra vegetation communities. Across groups, NEE and GEP were both significantly related to soil moisture only after adjusting for leaf area, with NEE showing somewhat greater explained variance (Fig. 2). This finding is consistent with other studies that found moisture was important in explaining the variation in CO₂ exchange after accounting for differences in LAI (Street et al., 2007) or biomass (Shaver et al., 1996; Welker et al., 2004). Such a result likely underscores by the fact that the distribution of vegetation types is ultimately a response to long-term differences in soil moisture driven by topographic differences, which control establishment and persistence of tundra communities. This gives rise to the observed close associations in soil moisture and community both at the local level (Walker et al., 1994) and across the arctic region (Walker, 2000). However, carbon flux responds to the vegetative expression in each of these communities (translated as either LAI or biomass) and these differences must be accounted for before flux-soil moisture relationships emerge.

Total foliar nitrogen reflects the interaction between plant community composition and nitrogen constraints on growth because it is the product of leaf area and mass and leaf N concentration. Because LAI had a weak significant (negative) relationship with soil moisture (F_1,6  =  5.213, p < 0.10; data not shown), the negative TFN-soil moisture relationship (Fig. 3a) suggests that the nitrogen constraints on growth (represented by the distribution of LAI) affect TFN more strongly than the relative differences in N uptake capacity (represented by leaf N concentration). Effects of physiological differences in N-uptake between vegetation groups can be seen in Figure 3b, where high leaf N concentrations are seen at high moisture levels. The wet tussock and wet sedge community are dominated by E. vaginatum and Carex spp., which have a greater N uptake capacity than most arctic species (McKane et al., 2002), allowing them to exist in N-poor environments. While it has been suggested that TFN could be used to predict NEE through the strong correlation between LAI and TFN (van Wijk et al., 2005; Campioli et al., 2009), our results produced a significant linear relationship between GEP per unit N, but not with NEE per unit N, due to the confounding effects of ER.

The temperature-ER relationship in arctic ecosystems has been the subject of considerable debate in the literature. Temperature is the greatest control on ER at a very broad scale (Raich and Schlesinger, 1992). Within a given vegetation community, greenhouses and other warming manipulations have led to increased soil temperature and increased respiration (e.g. Grogan and Chapin, 2000; Dorrepaal et al., 2009). However, the temperature-ER relationship is not straightforward, as moisture conditions play a large role in determining the temperature sensitivity of ER (Illeris et al., 2004, Kwon et al., 2006). In some cases substrate quality has been found to be of greater importance than temperature in predicting ER (Flanagan and Van Cleve, 1983; Hobbie, 1996). We found significant soil temperature-ER relationships only in mesic moisture conditions. This was surprising because in general the mesic groups exhibited a low mean and range in T_soil (Table 1), whereas it would be expected that relationships would be revealed in ecosystems that exhibited a greater range in T_soil. Similar to findings here, Nobrega and Grogan (2008) found that the birch ecosystems with the lowest soil temperature had the highest ER and soil respiration, indicating that temperature is not the primary control in comparisons across the range of soil moisture conditions.

The findings of this study emphasize the importance of strong moisture control at the highest and lowest soil moisture conditions. Soil respiration in saturated ecosystems has been shown to be insensitive to increased temperature (Johnson et al., 1996, 2000), likely due to oxygen limitation. In dry arctic ecosystems, lack of moisture has been found to be a greater limiting factor than temperature on ER (Illeris and Jonasson, 1999; Welker et al., 2000) because water stress limits microbial growth and root respiration (Illeris et al., 2003). These findings are consistent with those from other dry ecosystems where the primary control on variation in ER was moisture input, and no relationships between ER and soil temperature were found (e.g., Chimner and Welker, 2005; Knapp et al., 2002).

While soil moisture and temperature together create a non-linear trend in ER across the landscape (Fig. 4), other factors contribute to the variability seen in this relationship. Vegetation community plays a role in determining ER, through control on the quality of litter inputs to the soil (Hobbie, 1996; Christensen et al., 1999) and because of vegetation structure. In particular, tussock tundra demonstrated the most variable ER, especially wet tussock at high soil moisture content. Johnson et al. (1996) found no significant difference between ER in tussock and intertussock spaces, although within the intertussock group, ER was reduced in saturated conditions versus ER at field capacity. In a comparison of hummock and hollows, Sullivan et al. (2008) found that ER was much lower in hollows due to limitations imposed by the water table. Variations in ER in wet tussock may be at least partly influenced by the highly variable soil moisture, which may be due to the vertical structure of the tussock (which can reach heights of 0.5 m). Soil moisture may temporarily increase after a precipitation event, but this moisture quickly drains due to gravitational flow. While dry tussock did not exhibit the same variability in soil moisture, the deep layer of organic matter and the high root density found under tussocks (Chapin et al., 1979) may allow the vegetation greater access to the soil carbon, leading to increased ER.