WAPUSK NATIONAL PARK

WNP was established in 1996 to protect a representative section of the northwestern HBL. WNP is underlain by continuous permafrost in northern sections and discontinuous permafrost in the far south end of the park. Vegetation spans the transition between Arctic tundra and boreal forest (Parks Canada, 2000). Wetlands cover approximately 80% of the surface area of the park, with fens and bogs located along the coast and polygonal peat plateaus located farther inland (Parks Canada, 2008). In this study, we focus on ponds located within the coastal fen of WNP because it is the ecotype selected by LSG for their breeding grounds (Fig. 1). The coastal fen region comprises approximately 11% of the park (Parks Canada, 2000). Within this area, vegetation includes mainly sedges (Carex aquatilis) and rushes (Scorpidium scorpoides and Scirpus caespitosus), and terrestrial plant cover is sparse. During the past ∼40 years, the LSG population has expanded dramatically throughout the coastal fen ecotype in population size, density, and geographic distribution (Figs. 1, 2). The LSG population typically arrives in the HBL in late May and begins the southward migration at the end of August (Jefferies et al., 2003). Terrestrial areas with the highest disturbance show evidence of substantial grubbing activity and vegetation loss, and contain abundant deposits of feces. However, based on our field observations, the influence of LSG within the mapped geographic limits of Figure 1 is patchy, and substantial areas remain relatively undisturbed.

STUDY SITES

The original intent of this study was to characterize seasonal hydrolimnological and biogeochemical behavior in coastal fen ponds. However, following conversations with wildlife ecologists at the Churchill Northern Studies Centre (CNSC) and assessment of our data and field observations, we recognized an opportunity to explore differences in hydrolimnological and biogeochemical conditions between a highly LSG-disturbed pond (WAP 20) in our data set and 15 other low-disturbance ponds (note that WAP 3, 10, 11, and 12 were removed from the data set because they desiccated during mid-summer of the sampling year (2010); Fig. 1). These ponds are all small (average area = 0.29 km²) and shallow (<0.5 m), and were selected under the assumption that they are representative of ponds within the coastal fen ecotype. Based on observations made during our field campaigns, ponds WAP 1–19 had little to no visual evidence of disturbance by geese within their catchments even though several are located within the mapped area of the LSG breeding grounds (Fig. 1). Catchments of these ponds contained few deposits of feces and only small, if any, amounts of grubbing (hereafter termed lowdisturbance coastal fen ponds, or LDCF ponds). In contrast, LSG have substantially disturbed the catchment of pond WAP 20 by depositing high densities of feces, establishing nests with eggs along the pond shoreline, and removing the terrestrial vegetation by grubbing (Fig. 2, parts c and d). WAP 20 is located in the area that has been longest disturbed by the LSG (e.g., Batt, 1997; Lacobelli and Jefferies, 1991; Abraham et al., 2005a, 2005b; Jefferies et al., 2006; Fig. 1).

CLIMATE

Meteorological conditions monitored at the Churchill airport show marked seasonal variations in temperature and precipitation (Fig. 3, parts a and b; Environment Canada, 2010). Based on 1971– 2000 climate normals, average annual temperature is -6.9 °C, and average annual precipitation is 431.6 mm (rain = 264.4 mm, snow = 167.2 mm). During the sampling season of 2010, temperature was similar to climate normals in May, June, and August but slightly elevated in July and September. Rainfall was lower in June but markedly higher in late August (Fig. 3). Snow accumulation during the preceding winter (2009–2010) was considerably lower than the climate normal.

FIGURE 2.

(a, b) Top photographs depict ponds from the coastal fen ecotype of Wapusk National Park with low disturbance from the lesser snow goose (LSG) population. (c, d) Bottom photographs are from pond WAP 20, which has evidence of high disturbance from the LSG population. (e) Graph was taken from data available in Abraham et al. (2005b) on the mid-winter index of LSG and Ross's geese from 1950 to 2003. The solid line is a 3-year running average.

FIGURE 3.

Comparison of average monthly (a) temperature and (b) precipitation at Churchill, Manitoba (1971–2000) (Environment Canada, 2010).

ISOTOPE HYDROLOGY

Pond water samples for analysis of water isotope composition were stored in tightly sealed 30-mL high density polyethylene bottles until analysis following standard methods (Epstein and Mayeda, 1953; Morrison et al., 2001) at the University of Waterloo Environmental Isotope Laboratory (UW-EIL). Isotopic compositions are reported as δ¹⁸O (‰) and δ²H (‰) relative to the Vienna Standard Mean Ocean Water (VSMOW) standard. δ¹⁸O and δ²H results are normalized to -55‰ and -428‰, respectively, for Standard Light Antarctic Precipitation (Coplen, 1996). The analytical uncertainty was ±0.2‰ and ±2.0‰ for δ¹⁸O and δ²H, respectively.

WATER CHEMISTRY

Surface water samples and limnological measurements were collected from all the ponds during the three field visits. In situ measurements for pH were also conducted at approximately 10- to 15-cm depth using a YSI 600QS multimeter. Following collection, all water samples were stored at 4 °C until analysis. Immediately following collection, the pond water was passed through a screen (80-µm mesh) to remove large particles that can affect estimates of concentrations in the pond water. For determination of DIC and DOC, water was filtered within 12 hours of collection (cellulose acetate filters: 0.45-µm pore size, 47-mm diameter) and stored in the dark at 4 °C until analysis at Environment Canada's National Laboratory for Environmental Testing (Burlington, Ontario) following their standard methods (Environment Canada, 1994). Samples for determination of TP and TKN were acidified (held at 0.02% H₂SO₄) and analyzed at the Biogeochemistry Lab at the University of Waterloo (TKN = Bran Luebbe, Method No. G-189-097; TKP = Bran Luebbe, Method No. G-188-097; Seal Analytical, Seattle, U.S.A.). Chlorophyll a concentrations were determined by filtering measured volumes of water onto Whatman® GF/F filters at the CNSC, and the filters were stored frozen and in the dark until chlorophyll a concentration was determined using standard fluorometric techniques (Parsons and Strickland, 1963; SCOR/UNESCO, 1966) at the Aquatic Ecology Group Analytical Laboratory at the University of Waterloo.

CARBON ISOTOPE ANALYSES

Water samples were collected from all ponds during the three field visits in 2010 for the analysis of C isotope composition of DIC (δ¹³C_DIC) using 125 mL glass serum bottles, with rubber stoppers and syringe needles that were used to expel excess air. Samples were stored at 4 °C until analysis. Samples for the C isotope analysis of suspended particulate organic matter (δ¹³C_POM) were collected using multiple horizontal tows with a phytoplankton net (mesh size = 25 µm). Upon return to the field base (CNSC), samples were passed through a 63-µm mesh net to remove zooplankton and other large particles, filtered onto preashed Whatman® quartz filters (0.45-µm pore size), and dried at 60 °C for 24 hours in a drying oven. Filters were then exposed to 12 N HCl fumes in an air-tight vessel for 4 hours to remove carbonates, following the methods of Lorrain et al. (2003). Water samples were analyzed for δ¹³C_DIC and acidified filters were analyzed for δ¹³C_POM at the UW-EIL. Stable C isotope ratios are reported as δ¹³C (‰) relative to the Vienna-PeeDee Belemnite (VPDB) standard. The analytical uncertainty was ±0.05‰ and ±0.01‰ for δ¹³C_DIC and δ¹³C_POM, respectively. The C isotope fractionation factor was estimated from the difference between δ¹³C_DIC and δ¹³C_POM (i.e., Ĺ³C_DIC-POM; Fry 2006). To supplement C isotope measurements, dissolved CO₂ concentration and saturation values in the ponds were calculated based on methods outlined by Stumm and Morgan (1981) and Macrae et al. (2004), which determines the ratio of pond dissolved CO₂ concentration to equilibrium pond CO₂ concentration. A partial pressure of atmospheric CO₂ of 393 ppm was used for calculations (representative value for 2010 conditions; M. Macrae and R. Bourbonniere, unpublished data). Values of CO₂ saturation <1 indicate waters are under-saturated in dissolved CO₂, whereas values >1 indicate waters are over-saturated in dissolved CO₂.

NUMERICAL AND STATISTICAL ANALYSES

Principal components analysis (PCA) was used to explore patterns of variation in limnological and biogeochemical conditions among ponds over the course of the ice-free season. Only ponds that had a complete data set for all three sampling episodes were included in the PCA (i.e., ponds WAP 1 and 2 were not included in the PCA due to loss of the July δ¹³C_DIC sample). The median value for the sample scores of the LDCF ponds was added in the ordination plot to assess the typical pattern of seasonal change for the undisturbed ponds. The values for CO₂ saturation were added passively to the PCA to avoid over-representation, as it was calculated based on other variables included in the PCA. The PCA was performed using CANOCO version 4.5 software (ter Braak and Šmilauer, 2002).

To further compare limnological and biogeochemical conditions between the highly disturbed WAP 20 and the LDCF ponds (WAP 1–19), the distribution of values for the LDCF ponds was compared using boxplots with values from WAP 20 for the three sampling periods. One-sample t-tests were used to determine if the mean value of each variable in the LDCF ponds differs from the value obtained at WAP 20. These tests were performed for each of the three sampling periods using SPSS version 20, and alpha was set at 0.05. Interquartile ranges (IQR) were also calculated using SPSS version 20 in order to provide information on the variability of data. Spearman Rank correlations were used to assess trends over time for pH and DIC.

ISOTOPE HYDROLOGY

During the ice-free season of 2010, seasonal variations in pond water balances did not differ substantially between WAP 20 and the LDCF ponds based on water isotope values (Fig. 4). Pond water isotope composition at WAP 20 plots in similar δ¹⁸O- δ²H space as the median values for the LDCF ponds during each of the corresponding three sampling periods in 2010 (June δ¹⁸O: -10.1‰ and -10.4‰, δ²H: -87.6‰ and -82.7‰; July δ¹⁸O: -6.2‰ and -6.2‰, δ²H: -66.7‰ and -67.1‰; September δ¹⁸O: -12.6‰ and -11.1‰, δ²H: -94.5‰ and -89.2‰; WAP 20 and LDCF, respectively). Pond waters were more isotopically depleted in the spring and fall, compared to summer values, and mostly plotted above the Local Evaporation Line (LEL), likely owing to the influence of rainfall. During the summer, waters became more isotopically enriched in the LDCF ponds and WAP 20 and exceeded δ_SSL, the terminal basin steady-state value, indicating marked mid-summer evaporation and pond-level drawdown. Substantial evaporation was noted throughout the coastal fen ecotype between June and July, and four ponds within our original study set (WAP 3, 10, 11, and 12) desiccated during this time.

FIGURE 4.

δ¹⁸O-δ²H graph showing the seasonal trajectory of median water isotope compositions for the low-disturbance coastal fen (LDCF) ponds (gray squares, dashed line) and WAP 20 (black triangles, solid line) with respect to the Global Meteoric Water Line (GMWL; Craig, 1961) and the Local Evaporation Line for the Churchill region as reported in Wolfe et al. (2011).

MULTIVARIATE ANALYSIS: LIMNOLOGY AND CARBON ISOTOPE MEASUREMENTS

Multivariate ordination by PCA was used to assess similarities and differences in seasonal patterns of variation in limnological conditions and C isotope biogeochemistry between WAP 20 and the LDCF ponds (Fig. 5). The first two PCA axes explained 59.9% of the total variation in the measured limnological variables and C biogeochemistry. Axis 1 explained 34.8% and separated sample scores mainly based on seasonal changes in C biogeochemistry (δ¹³C_DIC, DIC concentration), pH, and alkalinity. Axis 2 captured 25.1% of the variation and separated samples mainly based on variations in concentrations of nutrients (TP, TKN) and chlorophyll a. For the LDCF ponds, sample scores in June generally clustered in the lower left quadrant, associated with relatively high pH, low concentrations of DIC, DOC and TKN, and low values of alkalinity, CO₂ saturation, and δ¹³C_DIC. In June, limnological conditions of LDCF ponds differed markedly from those at WAP 20, based on clear separation of the sample score for WAP 20 from the median of the sample scores for the LDCF ponds. The sample score for WAP 20 was positioned centrally within the ordination plot, representing more moderate values of all the variables relative to the LDCF ponds. In July, even greater limnological differences occurred between WAP 20 and the LDCF ponds. At this time, WAP 20 possessed higher concentrations of nutrients (TP, TKN), chlorophyll a, and δ¹³C_POM than all but one of the LDCF ponds (WAP 8, which had high [13.9 µg L^-1] chlorophyll a values compared to other ponds). The highest amount of variability among LDCF ponds occurred in July, as illustrated by the wide dispersal of sample scores along PCA axes 1 and 2. By September, differences in the distribution of limnological and biogeochemical variables between WAP 20 and the LDCF ponds lessened considerably, as indicated by the relatively close positioning of sample scores for most of the ponds (including WAP 20) towards the lower right quadrant. Sample scores during this time were associated with lower nutrient concentrations, pH, and δ¹³C_POM, and higher DIC concentrations, alkalinity, CO₂ saturation, and δ¹³C_DIC compared to July. Overall, the position of the sample scores identified that both differences and similarities existed in the seasonal progression of limnological and biogeochemical conditions between WAP 20 and the LDCF ponds.

UNIVARIATE ANALYSES: WATER-CHEMISTRY VARIABLES

Use of boxplots (Fig. 6) and one-sample t-tests (Table 1) allowed more detailed assessment of the similarities and differences in the values and seasonal patterns of variation in water chemistry variables between WAP 20 and the LDCF ponds. Generally, the boxplots illustrate that concentrations of TKN, TP, DOC (except September), and chlorophyll a at WAP 20 were within the range of values at the LDCF ponds (Fig. 6, parts a–d). However, two interesting patterns were observed. First, concentrations of TKN and DOC at WAP 20 were near the upper range of the LDCF ponds, but chlorophyll a concentrations at WAP 20 were not elevated. In fact, chlorophyll a was near the lower range for the LDCF ponds in June and September of 2010. Second, WAP 20 and the LDCF ponds had different seasonal patterns of variation in TP concentration. Additional details are presented below.

FIGURE 5.

Principal components analysis (PCA) ordination biplot comparing limnological and biogeochemical conditions in WAP 20 (triangles) and the LDCF ponds (small circles) from the three sample periods in 2010. Vectors with solid lines represent active variables in the PCA, whereas vectors with dashed lines represent variables added passively. Lines connect sample scores for WAP 20 (triangles and solid line) and the median for the 15 LDCF ponds (squares, dashed lined) to assess seasonal patterns of change in the limnological variables with and without extensive lesser snow goose activity, respectively.

In June, the range of TKN concentrations in the LDCF ponds was relatively small (IQR: 0.4 mg L^-1), and the median concentration (0.54 mg L^-1) was low (Fig. 6, part a). At WAP 20, TKN concentration was 1.01 mg L^-1, which approached the 90th percentile of the LDCF ponds and differs significantly from the mean of the LDCF ponds (Table 1). Between June and July, TKN concentration increased (median = 1.32 mg L^-1) in the LDCF ponds and variability rose as shown by a larger interquartile range (0.8 mg L^-1). TKN concentration also rose at WAP 20 (2.28 mg L^-1), to a value slightly above the 75th percentile of the LDCF ponds. By September, TKN concentrations declined, the IQR remained similar (0.9 mg L^-1), and the concentration at WAP 20 (0.69 mg L^-1) was close to the median of the LDCF ponds (0.64 mg L^-1).

Seasonal patterns of DOC concentration were similar to the patterns for TKN concentration for both WAP 20 and the LDCF ponds, with lower concentrations in June followed by increased values in July and declines between July and September (Fig. 6, part b). However, mean DOC concentrations for the LDCF ponds were significantly lower than the values at WAP 20 for all three sampling periods (Table 1; Fig. 6, part b). In June, both WAP 20 (11.0 mg L^-1) and the LDCF ponds (median: 6.0 mg L^-1) had low concentrations compared to the rest of the season. Variability of DOC concentrations among the LDCF ponds was small in June (IQR: 4.7 mg L^-1), and the concentration at WAP 20 was close to the 90th percentile of the LDCF ponds. Increased concentrations were observed for both WAP 20 (21.6 mg L^-1) and the LDCF ponds (median 11.1 mg L^-1) in July. Also in July, a larger range of concentrations was observed for the LDCF ponds (IQR: 6.4 mg L^-1), and the DOC concentration at WAP 20 again approached the 90th percentile of the LDCF ponds. In September, DOC concentrations declined to values similar to those in June (13.6 mg L^-1 and 9.8 mg L^-1; WAP 20 and LDCF medians, respectively). Additionally, the LDCF ponds had a narrower range of DOC concentrations in September (IQR: 1.8 mg L^-1), and the concentration at WAP 20 was slightly above the maximum value of the LDCF ponds.

TABLE 1

Results of one-sample t-tests, which compare the mean value of water chemistry variables from the low-disturbance coastal fen (LDCF) ponds with the value obtained at WAP 20. Comparisons are presented for all three sampling episodes (June, July, September).

In contrast to TKN and DOC, water-column concentrations of chlorophyll a were not elevated in WAP 20 compared to the LDCF ponds throughout the ice-free season (Fig. 6, part c). In June and September, chlorophyll a concentrations were low in the LDCF ponds (median: 1.9 µg L^-1 and 1.6 µg L^-1, respectively) and spanned a relatively narrow range (June IQR: 1.8 µg L^-1; September IQR: 2.2 µg L^-1). At these times, chlorophyll a concentrations at WAP 20 (1.1 µg L^-1 and 0.7 µg L^-1, respectively) were near the 10th percentile and significantly lower than the mean of the LDCF ponds (Table 1). Maximum chlorophyll a values occurred in July for both WAP 20 and the LDCF ponds, and the value at WAP 20 was close to the median value for the LDCF ponds (3.2 µg L^-1 and 3.0 µg L^-1, respectively), while the IQR remained similar (2.2 µg L^-1).

For the LDCF ponds, relatively low TP concentrations in June (median: 4.0 µg L^-1; IQR: 9.0 µg L^-1) were followed by an increase in July (median: 12.0 µg L^-1), with a marked increase in range (from <0.5 µg L^-1 to 69.0 µg L^-1; IQR: 18.0 µg L^-1; Fig. 6, part d). In September, TP concentrations declined in the LDCF ponds (median: <0.5 µg L^-1), and there was a much narrower range of values (IQR: 0.0 µg L^-1). In contrast, TP concentration decreased from June to July (13.0 µg L^-1 and 10.0 µg L^-1, respectively) at WAP 20, corresponding to a decline from the 75th percentile of the LDCF ponds in June to the 25th percentile in July, and then declined more substantially between July and September (to <0.5 µg L^-1, a value similar to the median of the LDCF ponds). Although seasonal patterns in TP differed, one-sample t-tests indicate a significant difference only in June between the mean of the LDCF ponds and the value for WAP 20 (Table 1).

UNIVARIATE ANALYSIS: VARIABLES RELATED TO CARBON ISOTOPE BEHAVIOR

In contrast to broadly similar patterns of seasonal variation in pond-water concentrations of TKN, DOC, and chlorophyll a, seasonal behavior of variables associated with C biogeochemistry differed markedly between WAP 20 and the median of the LDCF ponds. For example, seasonal trajectories of change in DIC concentration differed between WAP 20 and the LDCF ponds (Fig. 6, part e). Although values at WAP 20 were within the range of the LDCF ponds for all three sampling periods, WAP 20 was the only pond where DIC concentrations did not increase between June and July, as demonstrated by L^-1 (r_s= 0.697, p = 1.870 × 10–5, d.f. = 28). In June, WAP 20 had a higher DIC concentration than the 90th percentile of the LDCF ponds (WAP 20: 20.1 mg L^-1 and LDCF median: 13.2 mg L^-1; IQR: 6.9 mg L^-1) and this value is significantly different from the mean of the LDCF ponds (Table 1). In July, the median DIC concentration in the LDCF ponds increased by ∼7.8 mg L^-1 (range 3.6–15.6 mg L^-1; IQR: 8.4 mg L^-1), whereas the DIC concentration remained relatively constant (19.8 mg L^-1) at WAP 20 and fell within the 25th and 50th percentile of values for the LDCF ponds. Between July and September, both the range (IQR: 6.7 mg L^-1) of DIC concentrations and the median value (21.3 mg L^-1) remained similar at the LDCF ponds. In contrast, DIC concentration increased at WAP 20 between July and September, to 25.8 mg L^-1, a value near the 75th percentile of the LDCF ponds.

From June to July, pH declined in all the LDCF ponds that are supported by a significant negative trend (r_s (45) = -0.805, p = 2.671 × 10–11), but in July, pH was significantly higher in WAP 20 compared to the mean for the LDCF ponds (Fig. 6, part f; Table 1). In June, both WAP 20 and the LDCF ponds possessed similar pH values (9.2 and 9.1; WAP 20 and LDCF median, respectively). Between June and July, pH declined only slightly at WAP 20 (9.2 to 9.0, respectively), whereas median pH declined more strongly in the LDCF ponds (9.2 to 8.5). In the LDCF ponds, the range of pH values in July declined to below the 25th percentile of the values measured in June. In the LDCF ponds, pH continued to decline between July and September (median 8.2), with the 75th percentile value in September falling below the 25th percentile from July. At WAP 20, pH declined to 8.4 in September, which was slightly above the 75th percentile of the LDCF ponds. The variability of pH in the LDCF ponds was similar in June, July, and September (IQR: 0.2, 0.2, and 0.3, respectively).

FIGURE 6.

Boxplots depicting seasonal changes in (a) TKN concentration, (b) DOC concentration, (c) Chlorophyll a concentration, (d) TP concentration, (e) DIC concentration, (f) pH, (g) δ¹³C_DIC, (h) δ¹³C_POM, (i) Ĺ³C_DIC-POM, and (j) CO₂ saturation. The boxes identify the 25th percentile, median value, and 75th percentiles for the low disturbance coastal fen (LDCF) ponds. The whisker bars represent the 10th and 90th percentile, the solid black circles represent the maximum and minimum values observed for the LDCF ponds, and asterisks represent outliers. Black squares represent the mean seasonal value for the LDCF ponds. Black triangles joined by the solid line represent the values for WAP 20.

Measurements of the C isotope composition of DIC revealed unique seasonal C behavior in WAP 20 compared to the LDCF ponds (Fig. 6, part g). In June, the δ¹³C_DIC value at WAP 20 was similar to the median value for the LDCF ponds (-3.9 and -4.0‰, respectively). At this time, the LDCF ponds spanned a relatively broad range of δ¹³C_DIC values, (-7.6 to -0.3‰; IQR: 3.0‰). In July and September, there was a large divergence in the seasonal pattern of change in δ¹³C_DIC between WAP 20 and the LDCF ponds, and the mean for the LDCF ponds differed significantly from the value at WAP 20 (Table 1). Median δ¹³C_DIC values of the LDCF ponds increased throughout the ice-free season (July: -3.2‰ and September: -2.5‰). All but three of the LDCF ponds showed this pattern. Those three LDCF ponds showed only small decreases between June and July (0.4 to 1.1‰). In contrast, the δ¹³C_DIC of WAP 20 decreased sharply between June and July (to -7.3‰). The δ¹³C_DIC increased between July and September (to -6.1‰) as was observed in the LDCF ponds. In July, the interquartile range of δ¹³C_DIC values in the LDCF ponds was more narrow (IQR: 2.0‰), but the overall range was large due to occurrence of ponds with more extreme minimum (WAP 13: -6.8‰) and maximum (WAP 9: 0.1‰) values. Despite the large range, the July δ¹³C_DIC value for WAP 20 fell below the minimum value for the LDCF ponds. And, δ¹³C_DIC at WAP 20 remained below the range of the LDCF ponds in September (IQR: 2.6‰).

Seasonal patterns for δ¹³C_POM were similar at WAP 20 and the LDCF ponds, but larger changes occurred between sampling episodes for WAP 20 (Fig. 6, part h). In June, δ¹³C_POM at WAP 20 fell within the range of LDCF ponds (-23.5 and -25.4‰, respectively). Between June and July, δ¹³C_POM increased by 3.3‰ to -20.3‰ in WAP 20. In contrast, the median value of the LDCF ponds increased only slightly. Also, the range of values for the LDCF ponds was wider in July (IQR: 3.4‰) than in June (IQR: 2.4‰), and δ¹³C_POM at WAP 20 was near the 90th percentile of the LDCF ponds. The mean for the LDCF ponds was significantly lower than the value at WAP 20 in July (Table 1). In September, δ¹³C_POM decreased substantially (to -32.1‰) in WAP 20, whereas the decline was more muted at the LCDF ponds (median: -27.0‰). In September, the range of δ¹³C_POM values declined in the LDCF ponds (IQR: 1.8‰), and δ¹³C_POM at WAP 20 fell well below the minimum value of the LDCF ponds.

The patterns of seasonal change in C isotope fractionation (Ĺ³C_DIC-POM; Fig. 6, part i) also differed markedly between WAP 20 and the LDCF ponds. C isotope fractionation did not change appreciably for most of the LDCF ponds during the 2010 sampling season, as evidenced by the large amount of overlap among interquartile ranges in June, July, and September, and the comparable amount of variation (3.0, 4.6, and 3.3‰, respectively) and median values (21.8, 23.5, and 24.0‰, respectively). In contrast, Ĺ³C_DIC-POM declined significantly in July at WAP 20 (June: 19.7‰, July: 13.0‰, and September: 25.9‰; Table 1) to well below the minimum value in July at the LDCF ponds.

In June, all ponds were under-saturated for CO₂ with little variability (CO₂ saturation = 0.2 and 0.1 for WAP 20 and median LDCF ponds, respectively; IQR: 0.1; Fig. 6, part j). In July, all of the LDCF ponds had increased values for CO₂ saturation, and the majority (9 out of 15) were supersaturated in CO₂ (median value 1.3). The range of values also increased (0.5 at WAP 13, to 5.7 at WAP 2; IQR: 0.8). In contrast, WAP 20 remained under-saturated in CO₂ in July and its value was significantly different from the mean of the LDCF ponds (Table 1) with a CO₂ saturation value (0.3) similar to that measured in June. In September, the majority (11 out of 15) of the LDCF ponds were supersaturated in CO₂ (median value 1.4; IQR: 1.4) and CO₂ saturation in WAP 20 increased to a value similar to the LDCF ponds (1.3).

CARBON ISOTOPE DYNAMICS

At the 15 ponds with little to no LSG activity, δ¹³C_DIC values increased over the course of the ice-free season (Fig. 6, part g). As summer progressed, aquatic primary productivity increased, which caused an increase in the δ¹³C_DIC due to preferential use of ¹²C by algae during photosynthesis (e.g., Quay et al., 1986; Keeley and Sandquist, 1992; Wachniew and Różański, 1997). Thus, the increase in δ¹³C_DIC values that occurred between June and July in the LDCF ponds likely reflects an increase in primary productivity under conditions where C supply exceeded C demand. This is supported by the increase in DIC concentrations that occurred between June and July in the LDCF ponds, which indicate ample supply of DIC to meet demands by photosynthesis. And, it is consistent with the modest rise in phytoplankton biomass in the ponds between June and July, as estimated by pond water chlorophyll a concentration. The source of the DIC to the ponds is not clear, but likely came from decomposition in the pond sediments, which would have been warmer in July compared to June. The import of DIC from the catchment is unlikely as the ponds were drying and no surface inflow was observed at that time. Additionally, C isotope fractionation values were consistently around -20‰ throughout the ice-free season in the LDCF ponds, which are expected when there is sufficient dissolved CO₂ to support aquatic photosynthesis (e.g., Rau, 1978; Herczeg and Fairbanks, 1987; Bade et al., 2004).

In sharp contrast to the rise in δ¹³C_DIC values in the LDCF ponds between June and July, driven by seasonally increasing aquatic productivity under conditions of adequate C supply, δ¹³C_DIC declined sharply between June and July at WAP 20 (Fig. 6, part g). This feature suggests substantially different behavior of dissolved C cycling at the pond affected by marked LSG activity. Such deviation from the characteristic seasonal pattern of 13C-enrichment of DIC under conditions of increasing productivity can occur in certain situations. For example, elevated supply of soil-derived isotopically depleted DIC from the catchment can lower δ¹³C_DIC values (Wachniew and Różański, 1997). Also, increased net respiration of particulate organic carbon can result in increased DIC concentrations, decreased pH, and consequently a decline in δ¹³C_DIC (Quay et al., 1986; Wachniew and Różański, 1997). However, elevated supply of soil-derived isotopically depleted DIC was unlikely at WAP 20, because as stated above, the ponds were drying and no surface inflow was observed at that time. We note that approximately 90% of rainfall in July occurred prior to July 11 and is unlikely to account for the decline in δ¹³C_DIC as this would have probably influenced all of the ponds. Additionally, because pond water levels were lower in July, they were likely not receiving substantial input from permafrost thaw. Indeed, 4 of our 20 study ponds desiccated in July. Increased net respiration is also an unlikely explanation as this would result in a corresponding increase in DIC concentration, which was not observed in this pond (Fig. 6, part e).

We propose an alternative mechanism to explain the sharp decline in δ¹³C_DIC values between June and July at WAP 20. Under conditions of high algal production, high C demand, and high pH, previous studies have identified that strong kinetic C isotope fractionation can occur during CO₂ invasion from the atmosphere, which causes a decline in δ¹³C_DIC (Herczeg and Fairbanks, 1987; Takahashi et al., 1990; Bade et al., 2004). Under these conditions, C isotope fractionation will decline substantially from ∼20‰ (Rau, 1978; Herczeg and Fairbanks, 1987), as was observed in WAP 20 (13.0%; Fig. 6, part i). This process, previously termed “chemically enhanced CO₂ invasion,” was initially described by Herczeg and Fairbanks (1987) in their isotope study of Mohonk Lake during a cyanobacterial bloom. This phenomenon has also been observed in other highly productive temperate lakes and ponds (e.g., Emerson, 1975; Takahashi et al., 1990, Bade et al., 2004) and represents a plausible explanation for the δ¹³C_DIC decline that occurred between June and July at WAP 20. We propose that increased photosynthesis due to the LSG activity generated substantial C demand that exceeded dissolved CO₂ availability, which subsequently led to strong kinetic C isotope fractionation during CO₂ invasion. The increased C uptake is likely a consequence of high photosynthetic demand by benthic algae (not measured in our study) rather than phytoplankton, because phytoplankton biomass (as chlorophyll a concentration) was similar in ponds with and without the waterfowl activity and increased only modestly between June and July (see also next section for further details). Also, macrophytes are extremely sparse in WAP 20, and cannot account for the C demand. This description of C behavior implies that WAP 20 was a net C sink during mid-summer, which is further supported by low values of CO₂ saturation in July indicative of CO₂ under-saturation (Fig. 6, part j). In contrast, the majority (9 out of 15) of the LDCF ponds (and their median value) were super-saturated with CO₂ at this time and all of the LDCF ponds had CO₂ saturation values higher than WAP 20. The super-saturated CO₂ values in the LDCF ponds may be a consequence of less intense algal productivity, sediment degassing, or higher rates of respiration than in WAP 20. We attribute the rise in δ¹³C_DIC in September at WAP 20 (Fig. 6, part g) to the reduction in the intensity of this process due to lower C demand as temperatures cooled, which is consistent with the increase in C isotope fractionation (Fig. 6, part i). High amounts of rainfall and subsequent runoff of comparably isotopically enriched DIC derived from carbonate dissolution may have also contributed to the rise in δ¹³C_DIC in September at WAP 20 and the LDCF ponds.

Further support for the interpretation of C isotope dynamics from the δ¹³C_DIC patterns is evident in the values of δ¹³C_POM and their seasonal trends (Fig. 6, part h). Both WAP 20 and the LDCF ponds depicted similar seasonal patterns of δ¹³C_POM, with increases between June and July, and declines between July and September. This pattern, however, was more striking at WAP 20, with both a larger increase between June and July and larger decline between July and September compared to the LDCF ponds. The C isotope composition of particulate organic matter is dependent on the concentration of dissolved CO₂, the degree of productivity within a pond, and the source of C (O'Leary, 1981; Allen and Spence, 1981; Keeley and Sandquist, 1992). In June, δ¹³C_POM values were approximately -25‰ in both WAP 20 and the LDCF ponds, which are typically observed when there is high availability of dissolved CO₂ allowing ample isotopic fractionation to occur during photosynthesis (Keeley and Sandquist, 1992). By July, a large increase in δ¹³C_POM occurred at WAP 20, whereas this value increased only slightly at the LDCF ponds. The larger increase in δ¹³C_POM at WAP 20 can be attributed to higher C demand as evidenced by reduced C isotope fractionation (Fig. 6, part i). In contrast, ample supply of dissolved CO₂ at the LDCF ponds resulted in only slight increases in δ¹³C_POM. Alternatively, the increase in δ¹³C_POM at WAP 20 between June and July may be due to bicarbonate uptake. Under conditions of low concentrations of DIC and high pH, as was the case for WAP 20, algae may switch from using dissolved CO₂ as their main form of C to bicarbonate (Deuser et al., 1968; Emrich et al., 1970; Allen and Spence, 1981). Bicarbonate is isotopically heavier in C, which would lead to 13C-enrichment of POM. By September, δ¹³C_POM values decreased in WAP 20 likely due to lower C demand, increased inputs of DIC due to large precipitation events, and/or a switch back to uptake of dissolved CO₂ to support photosynthesis.

POND HYDROLOGY AND NUTRIENT DYNAMICS

In contrast to the striking differences in the seasonal behavior of the C isotope composition of dissolved inorganic and particulate organic fractions between the LSG-disturbed pond WAP 20 and LDCF ponds unaffected by LSG population expansion, seasonal patterns of variation in basin hydrology (assessed using δ¹⁸O, δ²H) and concentrations of major nutrients (except TP) and phytoplankton biomass (as chlorophyll a) were broadly comparable at all ponds. This finding suggests that the changes in basin hydrology are strongly regulated by a common external driver such as seasonal variations in meteorological conditions, rather than differences in the amount of disturbance by LSG populations. In northern landscapes, intense seasonality of meteorological conditions can exert strong influence on hydrology of tundra ponds, and affect seasonal patterns of change in nutrients and aquatic productivity (Sheath, 1986; Lougheed et al., 2011; White et al., 2014). For example in the HBL, snowmelt in spring increases connectivity of ponds with the surrounding catchment, and higher hydrologic inflow increases the transfer of allochthonous materials to ponds (e.g., Macrae et al., 2004; Wolfe et al., 2011; White et al., 2014). As the ice-free season progresses, these allochthonous inputs supplied in spring may lead to higher nutrient concentrations and productivity during summer. Indeed, Macrae (1998) observed that snowmelt runoff supplied nutrients to ponds. Then, hydrologic connectivity may decline as a result of drier conditions that increase evaporation and reduce overland flow (Macrae, 1998; Macrae et al., 2004), which can lead to lake-level drawdown and raise nutrient concentrations and productivity in ponds of the HBL (Macrae et al., 2004; White et al., 2014). Conversely, intense episodes of rain in late summer may result in subsequent dilution of nutrients and ions (Fig. 3; White et al., 2014). Such seasonal shifts in meteorological conditions appear to have exerted strong influence on pond water balance, nutrient concentrations, and phytoplankton biomass in all of the study ponds. For example, increasing pond-water isotope values between June and July indicate strong influence of evaporation (Fig. 4). These changes were concomitant with increases in concentrations of TKN, DOC, and chlorophyll a in the water column of all the study ponds (Fig. 6, parts a–c), likely due to evaporative concentration and phytoplankton growth. Increases in these variables can also be due to increased allochthonous inputs that occurred in the spring, but there was a low amount of winter snowfall observed (Fig. 3). Individual catchment characteristics may result in these processes differing among ponds (Wolfe et al., 2011) and, thus, may account for the wide range of values observed for concentrations of nutrients, DOC, and chlorophyll a during July in the LDCF ponds. After mid-July, the ponds experienced a decrease in water isotope compositions due to a period of heavy rainfall in August (>double the climate normal; Figs. 3, 4) and, correspondingly, concentrations of nutrients and chlorophyll a were diluted in all ponds (Fig. 6). Similar seasonal changes in nutrient concentrations in response to hydroclimatic conditions have also been reported for other shallow ponds in the Churchill area (Macrae et al., 2004; White et al., 2014).

In contrast to TKN, DOC, and chlorophyll a, seasonal variation of TP concentration differed between WAP 20 and the LDCF ponds (Fig. 6, part d). TP concentrations rose between June and July at the LDCF ponds, but declined at WAP 20 despite modest increase in water-column chlorophyll a concentration. A rise in chlorophyll a concentration in the water column is typically associated with a coincident increase in TP concentration, because phosphorus is contained within the phytoplankton (Wetzel, 2001). Consequently, the decline in TP concentration at WAP 20 between June and July, when values increased in the LDCF ponds, suggests P demand became greater at WAP 20. As described above, evidence from C isotope measurements and pond water pH indicates that the substantially greater C demand at WAP 20 in July was likely driven by higher rates of photosynthesis than occurred at the LDCF ponds. The inferred increase in P demand also is likely associated with the elevated rates of photosynthesis that produced the C demand. However, phytoplankton productivity does not appear to explain the greater demand for C and P at WAP 20, because concentrations of chlorophyll a in the water column were comparable at WAP 20 and the LCDF ponds throughout the ice-free season. Instead, we suggest that stimulation of photosynthesis by benthic algae at WAP 20 is a main cause of the increased C and P demand. WAP 20 is very shallow (<0.5 m), has sparse macrophytes, and so the luxuriant benthic mat along the pond bottom has potential to contribute a large fraction of the pond's total primary production. We note that lower-thanexpected phytoplankton chlorophyll a concentrations have been found in other Arctic lakes and ponds and attributed to benthic productivity (e.g., Van Geest et al., 2007; Côté et al., 2010). Moreover, a recent nutrient enrichment experiment that used microcosms containing water from a nearby coastal fen pond, with and without surficial pond sediments and associated benthic algal mat, demonstrated that these ponds rapidly assimilate additions of dissolved inorganic N (nitrate, ammonia) and P (phosphate) when surficial sediments and associated benthic biofilm are present (Eichel et al., 2014). Phosphate was taken up more rapidly from the water than inorganic N, and benthic algal biomass (as chlorophyll a concentration) became elevated when inorganic N and P were co-added to the microcosms, but not when N was added alone. Collectively, results of our study and the experiment by Eichel et al. (2014) suggest that nutrients supplied by waterfowl disturbances (feces, soil erosion) are rapidly taken up by benthic algae, which leads to increased demand for C and P. Measurements of benthic chlorophyll a and field-intensive nutrient studies would be needed to test this hypothesis. Stimulation of zooplankton grazing is another process that may explain the low chlorophyll a concentrations. In temperate locations, persistent N limitation as a consequence of waterfowl disturbance has often been reported and has been attributed to the low N:P ratio of bird feces (Manny et al., 1994; Post et al., 1998; Olson et al., 2005). However, in our study the comparable seasonal changes in TKN concentration between WAP 20 and the LDCF ponds suggest that demand for N is lower than demand for P. Other studies have also found significant alteration of phosphorus dynamics by waterfowl activities in both Arctic (e.g., Keatley et al., 2009; Côté et al., 2010) and temperate locations (e.g., Manny et al., 1994; Post et al., 1998). Consequently, seasonal measurements of TP concentration appear to provide a more effective measure of the effects of the LSG disturbance on coastal fen ponds compared to measurements of TKN or chlorophyll a concentrations. However, based on our case study, C-isotope measurements appear to provide the clearest distinction between WAP 20 and the LDCF ponds.