Effects of experimental warming on nitrogen concentration and biomass of forage plants for an arctic herbivore



  1. In many arctic herbivores, the growth of young depends upon a synchrony between hatching date and seasonal change in plant nutritive quality. If plants respond more quickly than herbivores to climate warming, this may cause a mismatch between the availability of high-quality food and the hatching of young. This study examines the impact of experimental warming on the main food plants of an arctic herbivore, the greater snow goose (Chen caerulescens atlanticaL.) breeding on Bylot Island, Nunavut, Canada.
  2. During summers 2007–2009, we increased the temperature using small glasshouses (open-top chambers, OTC) in two habitats, wetlands and mesic tundra. Every 10 days, we measured above-ground plant biomass and a proxy of nutritive quality, nitrogen concentration, of graminoid plants in warmed and control plots from snowmelt in June until late July.
  3. Open-top chambers increased mean maximum temperature by up to 2.0 °C in wetlands and 4.6 °C in mesic tundra. Annual warming significantly increased biomass of graminoids by up to 29% in wetlands and 20% in mesic tundra. There was no difference in nitrogen concentration of the four plant species sampled (Dupontia fisheri, Eriophorum scheuchzeri, Arctagrostis latifolia and Luzula spp.) early in the season, but the seasonal decline in nitrogen occurred more rapidly in warmed than in control plots (10% to 14% less nitrogen in warmed plots in July). This effect was consistent across the 3 years of the experiment and independent of annual variation in plant phenology. There was either a weak positive effect or no effect of the warming treatment on the nitrogen biomass of plants depending on species or period of the season.
  4. Synthesis. Our results show that warming speeds up plant phenology and the seasonal decline in nutritive quality for arctic herbivores. Because young herbivores like geese are highly sensitive to the nitrogen concentration of their food, a warmer climate will likely reduce their growth. Climate warming may therefore have a negative impact on the population dynamic of arctic herbivores by reducing the quality of their summer forage.


Mean global temperature has increased by approximately 0.74 °C over the last century (1906–2005), and global climatic models predict that this will continue over the next century and reach an increase of 1.8–4.0 °C before 2099 (IPCC 2007). This warming has already had major impacts on ecosystems by altering not only the distribution of species but also their phenology (Walther et al. 2002; Parmesan & Yohe 2003; Oberbauer et al. 2013). Phenology, defined as the seasonal initiation of essential life cycle activities (Berteaux et al. 2004), is the parameter most often affected by climate change (Parmesan 2006). Changes in phenology can affect trophic interactions through a phenomenon known as trophic mismatch, which occurs when species at different trophic levels exhibit differential change in phenology in response to changes in their environment (Durant et al. 2007; Kerby & Post 2013). Trophic mismatch between predators and their prey can reduce reproductive success and/or recruitment, leading in some cases to population declines (Post & Forchhammer 2008; Both et al. 2010). Moreover, in long-distant migrants, changes in climate occurring on the wintering and breeding grounds may progress at different rates, hindering their adaptation to warming and making them more susceptible to trophic mismatch (Visser, Both & Lambrechts 2004; Møller, Fiedler & Berthold 2010).

Polar regions are warming at a faster rate than the rest of the planet and arctic ecosystems are among the most vulnerable to climate change (Post et al. 2009; Gilg et al. 2012). Reproduction of arctic herbivores is highly seasonal and typically timed to coincide with the period when food quality and availability is at its peak (Lepage, Gauthier & Reed 1998; Post et al. 2003). For these reasons, goose populations that breed in arctic regions should be highly susceptible to trophic mismatch. Migratory geese, which are major herbivores in many tundra regions, have a short time window to reproduce during the summer, leading to synchronized reproduction (Lepage, Gauthier & Menu 2000). Gosling growth is highly sensitive to environmental conditions (Gauthier, Fournier & Larochelle 2006) and is directly related to quality and availability of food plants, primarily protein content, on the brood-rearing areas (Larsson & Forslund 1991; Lindholm, Gauthier & Desrochers 1994; Lepage, Gauthier & Reed 1998).

In recent years, numerous studies have examined the response of plants to experimental warming, most notably by the International Tundra Experiment (ITEX) group in polar and alpine regions. These studies have demonstrated that plants respond to warming by an accelerated phenology, increased growth and a greater reproductive effort in the form of increased flower production (Henry & Molau 1997; Arft et al. 1999; Aerts et al. 2004; Aerts, Cornelissen & Dorrepaal 2006; Elmendorf et al. 2012; Oberbauer et al. 2013). While these studies show how warming affects plant phenology, it is unclear how temperature increases could affect the nutrient content of plants for herbivores. Plants at high latitudes and high elevation generally have high nutrient concentrations (Kudo, Molau & Wada 2001), possibly as an adaptation to low air temperature (Weih & Karlsson 2001), and thus we would expect warming to have a negative impact on nitrogen content. However, previous warming experiments show decreases, increases or no effect on the nitrogen content of plants (Lenart et al. 2002; Jónsdóttir, Khitun & Stenström 2005; Welker et al. 2005; Aerts et al. 2009; Natali, Schuur & Rubin 2012; see Appendix S1 in Supporting Information for a review). These discrepancies may be due to differences in species sampled, climatic and edaphic conditions, type of warming experiment or timing of sampling (Jónsdóttir, Khitun & Stenström 2005). It is therefore not possible to make generalizations on how future increases in temperature should affect forage quality of arctic herbivores.

Proteins, generally measured by the nitrogen concentration of plants, are known to be the most limiting nutrient for the growth of arctic herbivores like geese (Sedinger & Raveling 1986; Lepage, Gauthier & Reed 1998). Nitrogen concentration of arctic plants, which is typically highest shortly after the onset of plant growth, decreases slowly throughout the summer, and the hatching of goslings is generally timed to coincide with the period of peak food quality (Manseau & Gauthier 1993; Lepage, Gauthier & Reed 1998; Larter & Nagy 2001; Cadieux, Gauthier & Hughes 2005). If warming was to change this pattern by advancing the date of peak nitrogen or accelerating the seasonal decline in plant quality, this may lead to a mismatch between the hatching of goslings and the availability of high-quality food, with potential negative impacts on gosling growth and survival.

The aim of this study is to examine the effect of short-term experimental warming on the availability and quality of forage plants for an arctic herbivore. To accomplish this, we used small polycarbonate glasshouses to increase air temperature in two different habitat types used by geese during three summers (2007–2009) in order to evaluate the effects of warming on (i) total above-ground biomass and (ii) the timing of the seasonal decline in plant nutritive quality, an important phenological event for growing herbivores during the summer. In this study, nitrogen concentration was the proxy used for nutritive quality of forage for goslings. We predicted that (i) warming would increase the total forage plant biomass but (ii) that this increase would be accompanied by a more rapid decline in plant nitrogen concentration in warmed plots.

Materials and methods

Study Area

This study was conducted on the south plain of Bylot Island (73°08′N, 80°00′W), north of Baffin Island, Nunavut, Canada. This region is one of the most important breeding areas for the greater snow goose (Reed, Hughes & Boyd 2002). Habitats used by geese can be divided in two broad categories (Gauthier, Rochefort & Reed 1996). The first habitat type consists of wetlands characterized by wet polygon fens (Tarnocai & Zoltai 1988). Vascular plant communities in this habitat are dominated by graminoids and sedges, such as Dupontia fisheri R. Br., Eriophorum scheuchzeri Hoppe and Carex aquatilis Wahlenb., growing through a thick carpet of moss. These three species account for > 90% of the vascular plant biomass in this habitat (Manseau & Gauthier 1993). The second habitat type consists of upland mesic tundra, characterized by rolling hills and better-drained soil. Vascular plant communities are more diversified and dominated, in decreasing order of abundance, by prostrate shrubs (Salix arctica Pall. and Cassiope tetragona D. Don), graminoids (Arctagrostis latifolia (R. Br.) Griseb., Luzula nivalis (Laest.) Beurling, L. confusa Lindeberg, Poa arctica R. Br.) and other small herbaceous species (e.g. Stellaria longipes Goldie, Oxytropis maydelliana Trautv., Polygonum viviparum L.) (Duclos 2002; Audet, Gauthier & Lévesque 2007). Though less preferred by geese, mesic tundra vegetation covers approximately 90% of the landscape and is used at low densities throughout the nesting and brood-rearing periods (Hughes, Gauthier & Reed 1994).

The annual average air temperature on Bylot Island over the period 1994 and 2007 was −14.5 °C (4.5 °C in the summer and −32.8 °C in the winter; Cadieux et al. 2008). Over the past 35 years, temperature in the area has warmed on average by 2.8 °C in spring and summer and by 4.3 °C in fall (Gauthier et al. 2011). The annual cumulative number of thawing degree-days has also increased from 381 in 1989 to 521 in 2011 (Gauthier et al. 2013).

Experimental Design and Sampling

Air temperature was increased using hexagonal, conical open-top chambers (OTC) made of polycarbonate according to the specifications of the ITEX protocol (Molau & Mølgaard 1996). They measured approximately 1.6 m in diameter at the base, 83 cm at the top and 24 cm high. Design was modified slightly to increase rigidity by bending the top 1 cm of each panel. This allowed the use of thinner polycarbonate (2 mm thick) and reduced their weight and the impact of the chambers on light quality. These warming devices have been used in numerous studies to increase temperature on a small scale (Marion et al. 1997; Hollister & Webber 2000; Hollister, Webber & Tweedie 2005; Hollister et al. 2006; Elmendorf et al. 2012). OTCs allow for normal precipitations, but they alter light conditions and wind patterns around the plants and can advance snowmelt by up to 14 days (Marion et al. 1997). Most of the heating by OTCs occurs when radiation peaks and thus the strongest impact of warming relates to daily maximum temperatures. In previous studies, daily maximums were increased by ~3.5 °C in OTCs, and the change in mean soil temperature (at −3 cm) ranged from −0.2 °C to 1.3 °C (see Marion et al. 1997 for further details concerning OTC performance). Each OTC was paired with an adjacent unwarmed control plot (2 m2) protected from goose grazing by a 30-cm high, 2.5-cm mesh chicken wire (exclosure). This experiment was conducted over three summers (2007–2009) in the two habitat types described above (wetlands and mesic tundra).

Open-top chambers and exclosures were set up at the end of the summer preceding plant sampling at 6 different sites per habitat type and spaced from each other by > 200 m. The same general sites were reused every year, but OTCs and exclosures were moved 10–20 metres each year due to destructive plant sampling. In summers 2007 and 2008, we used 3 OTCs and 3 control plots per site (each spaced out by a few metres), for a total of 18 OTCs and 18 control plots per habitat type. In 2009, the sampling was downscaled to one OTC and one control plot per site, for a total of six OTCs and six control plots per habitat type. The effect of treatments on the microclimate of growing plants was monitored in a subsample of each treatment (typically three to five plots) in 2007 and 2008 using two types of automated temperature loggers: Onset HOBO® Pendant probes (Onset Computer Corporation, Bourne, MA, USA; ~0.5 °C precision) or Model 107 Campbell Scientific temperature probes (Campbell Scientific, Edmonton, AB, Canada; ~0.2 °C precision). In 2007, most temperature loggers in control plots of mesic tundra habitats were destroyed by foxes over the course of the summer, so data from control plots were obtained from a single logger until the end of July. The loggers were buried approximately 2–3 cm under the surface and recorded the temperature every 20 min for the duration of the experiment.

We collected above-ground plant biomass of two common species in each habitat type four times per season (approximately every 10–14 days) from shortly after snowmelt in mid-June until the end of July. Plant species were Arctagrostis latifolia and Luzula spp. (L. nivalis or L. confusa) in mesic tundra and Eriophorum scheuchzeri and Dupontia fisheri in wetlands. These species were chosen for their abundance and their importance in the diet of goslings (Manseau & Gauthier 1993; Duclos 2002; Audet, Gauthier & Lévesque 2007). In wetlands, sampling was done by removing a 20 × 20 cm piece of turf at random within OTCs or exclosures, avoiding sampling directly adjacent to previous sampling. We cut all vegetation present on the pieces of turf at the base of the white basal stem buried in the moss. In mesic tundra, we marked a random 25 × 50 cm area in each plot and removed each shoot of Arctagrostis and Luzula present in this area by pulling it from the root. The size of the sampling area differed between the two habitat types to ensure that we collected sufficient plant material to conduct the nitrogen analyses (stem density was higher in wetland than in mesic tundra). Markers were left in the plots after each sampling to avoid subsequent sampling over the same area. Samples were sorted to separate vegetative green parts (i.e. leaves), flowers (including stems) and dead matter, dried at 45–50 °C for 24–36 h and weighed to the nearest 0.001 g. Green parts (i.e. leaves) of Eriophorum, Dupontia, Arctagrostis and Luzula were ground to a fine powder and analysed for nitrogen concentration using a QuickChem Lachat nutrient auto-analyser (Zellweger Analytic, Milwaukee, WI, USA, QuickChem 4000 Series).

Data Analysis

For temperatures, we averaged values across loggers at each sampling time and extracted daily maximum temperatures. We calculated mean daily maximum temperature over four time periods corresponding to our plant sampling periods.

Linear mixed models were used to determine the impact of treatment (warming vs. control) on above-ground green biomass and on nitrogen concentration (%) and biomass (g m−2). Because the same experimental units were repeatedly sampled over a period of time within years (but differed between years), our model consists of a split-block anova with repeated measures. The year effects appeared in the main part of the model with the random factor site as the blocking factor, while the treatment effect appeared in the subpart of the model. Measures were repeated over sampling periods in experimental units within each site × year × treatment combination. The correlation structure that best fitted the data, based on the Akaike Information Criteria, was the first-order autoregressive structure (Crowder & Hand 1990). Distributions of continuous variables were tested for normality, and, if appropriate, log or square root transformations of skewed variables were used in the analyses. For plant biomass, we summed both plant species within each sampling unit, but each species was treated separately for the nitrogen analyses. Nitrogen biomass was calculated as the product of green biomass (i.e. leaves) by nitrogen concentration in each individual sample. Samples with fewer than 20 shoots were excluded from the biomass analyses. Following significant effects in the anova table, we used the protected LSD multiple comparison method to identify differences among treatments. Statistical analyses were performed with sas (procedure PROC MIXED of sas 9.3; SAS Institute, 2010, Cary, NC, USA) and r (R Development Core Team 2006). Conditional R2 values were calculated following the procedure outlined by Nakagawa and Schielzeth (2013).



Temperature in warmed plots was higher than in control plots in both habitat types, but the difference was most pronounced in mesic tundra (Fig. 1). In wetlands, OTCs increased maximum daily soil temperatures by 0.2 ± 1.0 °C (95% CI) for 10–18 June, 1.4 ± 0.6 °C for 26 June–2 July, 2.0 ± 2.3 °C for 8–14 July and 1.1 ± 1.1 °C for 22–28 July. In mesic tundra, OTCs increased maximum daily soil temperatures by 3.9 ± 0.9 °C for 15–24 June, 4.6 ± 0.9 °C for 27 June–5 July, 1.9 ± 0.7 °C for 13–19 July and 1.4 ± 1.2 °C for 26–29 July.

Figure 1.

Mean maximum daily soil temperature (°C) in (a, b) wetlands and (c, d) mesic tundra in control and warmed plots during summers 2007 and 2008 on Bylot Island, Nunavut. Temperatures were recorded 2–3 cm under the surface. Data points were obtained by averaging temperatures across loggers and calculating the daily maximum from these averaged values.

Above-Ground Biomass

Warming increased live above-ground biomass of the sampled plants in both wetlands (F1,15 = 11.6, = 0.004; R2 of the full model = 0.72) and mesic tundra (F1,15 = 9.32, = 0.008; R2 of the full model = 0.65). In wetlands, a posteriori tests indicated that biomass was 21%, 29% and 17% higher in warmed plots in late June, mid-July and late July, respectively (Fig. 2a). In mesic tundra, biomass was 20% higher in warmed plots in mid-July and 15% higher in late July, but the trend was already apparent in early July (Fig. 2b). Although there were annual variations in live above-ground biomass (wetlands: F2,10 = 5.35, = 0.026; mesic tundra: F2,10 = 34.0, < 0.001), there were no significant interactions between year and the warming treatment (> 0.28), indicating that the effect of warming was consistent across all 3 years of the experiment.

Figure 2.

Dry above-ground biomass of two species of goose forage plants in (a) wetlands and (b) mesic tundra in control and warmed plots on Bylot Island, Nunavut. Each data point is based on 42 replicate plots over three summers (2007–2009). Asterisks indicate significance (< 0.05) based on individual least-square means comparisons. Error bars represent ± 1 SE.

Nitrogen Concentration

In wetlands, plants reached a peak in nitrogen concentration in late June/early July, after which nitrogen declined rapidly (Fig. 3a, b). The same decline was found in mesic tundra, but the peak in nitrogen concentration possibly occurred earlier because the highest nitrogen values were recorded during the first sampling period in mid-June (Fig. 3c, d; note that the first sampling date was slightly later in mesic tundra than in wetlands due to logistical constraints). Overall, warming affected the nitrogen concentration of all four plant species but the effect of the treatment changed over the season as the interaction period × treatment was always significant (Table 1). Warming had no effect on the nitrogen concentration of Dupontia early in the growing season but there was a significant negative impact of warming in mid- and late July (Fig. 3a). In Eriophorum, nitrogen concentration was significantly higher in warmed plots at the first sampling period, but this effect was reversed in mid- and late July (Fig. 3b). Nitrogen concentration of Dupontia was 14% and 10% lower in warmed plots than in the control in mid- and late July, respectively, and 13% and 14% lower in Eriophorum over the same periods. We found similar results in mesic tundra, but the effect of warming on nitrogen concentration occurred earlier during the season (Fig. 3c, d). Nitrogen concentration of warmed Arctagrostis samples was 12%, 14% and 14% lower than control and that of warmed Luzula samples was 11%, 10% and 7% lower at the three sampling dates in July.

Table 1. Effects of sampling period, year and treatment (warming vs. control) on nitrogen concentration (%) of four plant taxa on Bylot Island, Nunavut during three summers (2007–2009)
d.f.Dupontia (= 320)Eriophorum (= 300)Arctagrostis (= 321)Luzula (= 309)
  1. The F-values for the main effects and the interactions (anova for repeated measures) are presented with their level of significance (*< 0.05; **< 0.01; ***< 0.001). Conditional R 2 values were calculated following the procedure outlined by Nakagawa and Schielzeth (2013).

  2. a

    d.f. of error term = 272 (Dupontia), 252 (Eriophorum), 273 (Arctagrostis), 261 (Luzula).

  3. b

    d.f. of error term = 15.

  4. c

    d.f. of error term = 10.

Perioda 3150.5*** 73.8*** 214.0*** 151.0***
Treatmentb 110.4** 7.7* 24.3** 13.3**
Yearc 21.82.617.5** 1.8
Period × Treatmenta 34.4** 8.7*** 2.6* 2.9*
Period × Yeara 67.0*** 4.2** 10.9*** 17.6***
Treatment × Yearb
R 2 of the model 0.730.640.800.76
Figure 3.

Nitrogen concentration of above-ground green parts of (a) Dupontia fisheri, (b) Eriophorum scheuchzeri, (c) Arctagrostis latifolia and (d) Luzula spp. in control and warmed plots on Bylot Island, Nunavut. Each data point is based on 42 replicate plots over three summers (2007–2009). Asterisks indicate significance (< 0.05) based on individual least-square means comparisons. Error bars represent ± 1 SE.

There were significant interactions between sampling periods and year for all four species sampled (Table 1), which suggests annual variations in plant phenology (i.e. the seasonal decline in nitrogen concentration occurred earlier in some years and later in others). However, there was no significant interaction between the warming treatment and year for any species, indicating that the decline in nitrogen concentration was consistently steeper in warmed plots than in the control for all 3 years regardless of annual differences in phenology.

Nitrogen Biomass

In wetlands, nitrogen biomass (g m−2) of plants showed a sharp increase at the beginning of the season in late June, continued to increase at a slower rate in the first half of July and tended to decrease at the end of July (Fig. 4a, b). In mesic tundra, nitrogen biomass of Arctagrostis increased in late June but reached a plateau in July with little change thereafter (Fig. 4c). Nitrogen biomass of Luzula was very low compared to the three other species and showed a weak increase over time (Fig. 4d). In wetlands, there was no significant effect of the warming treatment on the nitrogen biomass of Dupontia, but the effect was nearly significant for Eriophorum (= 0.055; Table 2). Generally, warming had a positive effect on nitrogen biomass of wetland plants, but a significant difference was only found for Dupontia in late June when it was 17% higher in warmed plots than in the control (Fig. 4a, b). There was no effect of warming on plant nitrogen biomass in mesic tundra (Table 2 and Fig. 4c, d). Overall, there was no change in the impact (or lack thereof) of warming over the season (interaction period × treatment was not significant for any plant species), and there was no evidence that the effect of warming changed between years (treatment × year interaction was never significant; Table 2). Finally, nitrogen biomass varied annually in all plant species except Dupontia.

Table 2. Effects of sampling period, year and treatment (warming vs. control) on nitrogen biomass (g m−2) of four plant taxa on Bylot Island, Nunavut during three summers (2007–2009)
d.f. Dupontia (= 307) Eriophorum (= 242) Arctagrostis (= 278) Luzula (= 248)
  1. The F-values for the main effects and their interactions are presented, together with their level of significance (+0.05 < < 0.10; *< 0.05; **< 0.01; ***< 0.001). Conditional R2 values were calculated following the procedure outlined by Nakagawa and Schielzeth (2013).

  2. a

    d.f. of error term = 259 (Dupontia), 194 (Eriophorum), 230 (Arctagrostis), 200 (Luzula).

  3. b

    d.f. of error term = 15.

  4. c

    d.f. of error term = 10.

Perioda 370.3*** 14.3*** 7.2** 5.7**
Treatmentb 12.34.4+ 2.50.7
Yearc 20.45.9* 29.5*** 4.6*
Period × Treatmenta 32.5+
Period × Yeara 63.6** 2.2* 1.82.7*
Treatment × Yearb
R 2 of the model 0.650.570.560.19
Figure 4.

Nitrogen biomass (g m−2) of above-ground green parts of (a) Dupontia fisheri, (b) Eriophorum scheuchzeri, (c) Arctagrostis latifolia and (d) Luzula spp. in control and warmed plots over the summer on Bylot Island, Nunavut. Each data point is based on 42 replicate plots over three summers (2007–2009). Asterisks indicate significance (< 0.05) based on individual least-square means comparisons. Error bars represent ± 1 SE.


Effects of Warming on Plant Growth

Differences in temperature between warmed and control plots were similar to those found by previous studies using OTCs in arctic or subarctic systems (Marion et al. 1997; Jónsdóttir et al. 2005; Hollister et al. 2006). As temperature was recorded 2–3 cm in the soil, the temperature increase experienced by plants at the ground level or slightly above was probably higher. This temperature increase was within the lower range of the warming predicted for polar regions over the next century (IPCC 2007). Wetland soils were saturated in water for most of the growing season, and this probably explains why the impact of OTCs on temperatures was more pronounced in mesic tundra than in wetlands, since water has a relatively high thermal inertia.

Our plots were warmed for only one growing season, and thus, we were unable to measure the cumulative effect of warming on plants. Nonetheless, short-term experiments are relevant to feeding conditions encountered by herbivores, at least during the initial stages of warming, because seasonal changes in nutrients result from physiological processes that are expected to be sensitive to temperature. Moreover, by repeating our experiment over 3 growing seasons, we were able to show that the effects of warming were consistent across all years despite annual variations in weather conditions.

As predicted, above-ground live biomass of plants was higher in warmed plots in both wetlands and mesic tundra, which is in agreement with previous studies using similar manipulations (Wookey et al. 1994, 1995; Arft et al. 1999; Rustad et al. 2001; Hollister, Webber & Tweedie 2005). This increase may be an effect of increased photosynthetic activity at higher temperatures and/or an effect of increased nutrient availability due to higher rates of mineralization in the soil. A recent, multisite analysis of warming experiments (combining 61 tundra warming experiments) showed that the response of tundra vegetation to warming shows strong regional variations, but that linear increases in effect size were common, with little evidence of acceleration or saturation over time (Elmendorf et al. 2012). This meta-analysis also suggests that graminoids are more responsive to warming at colder sites, with a larger response of sedges in wet habitats and of grasses in dry habitats. Nonetheless, other studies reported that vegetative growth was only significantly higher in the early years of experimental warming and that after 4 years of warming, increases were no longer significant at some sites, possibly due to nutrient depletion (Arft et al. 1999). Therefore, under some circumstances, the short-term response of plants to temperature may have long-term costs if increased growth occurs at the expense of below-ground stored reserves. Although an increase in temperature typically leads to increased mineralization of nitrogen, this may not fully compensate for nutrient depletion because a large percentage of the nitrogen released is quickly taken up and stored by the microbial community, especially in nutrient-poor sites (Rustad et al. 2001).

Although the concentration of nitrogen varied among species and sometimes by year, possibly in relation to annual climatic variations in the latter case, the same seasonal pattern was observed in all plant taxa and years. We did not observe an initial increase in nitrogen concentration in mesic tundra, but this is probably because our first sampling period occurred too late to capture it. The seasonal decline in plant nutritive quality is ubiquitous in high-latitude plants and may be caused in part by an increase in structural material (fibre) during the growing season, as previously shown in wetland plants at our study site (Manseau & Gauthier 1993; Piedboeuf & Gauthier 1999). This increase results in a decline in nitrogen concentrations even though total nitrogen biomass may be increasing (Chapin 1980). In our study, total nitrogen biomass generally increased initially, but this increase quickly levelled off and nitrogen biomass tended to decrease after mid-July. This late summer decline is probably due to translocation of nitrogen to the roots, leading to a decrease in total nitrogen content of green parts (Jonasson & Chapin 1991; Jonasson & Shaver 1999).

Nitrogen concentration of plants in warmed plots decreased more rapidly than in control plots, which supports our second prediction. This is possibly because the dilution effect caused by the seasonal increase in carbon-rich tissues (i.e. fibre) occurred more rapidly in warmed plots due to the speeding up of plant growth, a result consistent with other warming experiments conducted in polar regions (Tolvanen & Henry 2001; Day, Ruhland & Xiong 2008). However, the reduction in nitrogen concentration in warmed plots lagged behind the concomitant increase in biomass, which started earlier in the season. A possible explanation is that fibre accumulation is typically more important during mid-summer in arctic graminoids (Manseau & Gauthier 1993), when plants become taller and presumably require more structural support. Alternatively, plants could perhaps partly adjust their nitrogen uptake to initially match the increase in biomass production but only up to a certain limit. When we combine the opposite trends of increased biomass but reduced nitrogen concentration in warmed plots, we found that warming tended to increase total nitrogen biomass. However, this potential positive effect of warming on total plant nitrogen was only present in wetland plants in the first half of the growing season as it weakened or disappeared later on.

Consequences of Warming on Herbivore Food Quality

For herbivores like geese, nitrogen concentration is probably more important than total nitrogen biomass in determining plant nutritive quality, especially in growing goslings. Unlike most herbivores, geese have a rapid passage time of food in their gut, are unable to digest most of the cellulose, and have low nitrogen retention efficiency (Mattocks 1971; Sedinger 1997). They must therefore select food with a high nitrogen concentration to meet their nutrient requirements, a constraint that is exacerbated in goslings by the synthesis of new tissues during growth (Sedinger 1997). Goslings' guts already operates at their maximum capacity, and any decrease in nitrogen concentration of their food plants cannot be compensated by an increase in total food intake and thus results in a reduction in their growth, as shown by experiments conducted in captivity and in the wild (Lindholm, Gauthier & Desrochers 1994; McWilliams & Leafloor 2005; Gauthier, Fournier & Larochelle 2006). The natural, seasonal decline in plant nitrogen can have major consequences on growth rate of goslings, especially in those hatched late in the season, which suffer from an increased mismatch between the timing of hatch and the period of highest food quality (Sedinger & Raveling 1986; Lindholm, Gauthier & Desrochers 1994; Lepage, Gauthier & Reed 1998). Indeed, Lepage, Gauthier and Reed (1998) have shown that nitrogen availability when goslings are 11–25 days old was the best predictor of structural size and body mass of greater snow goose goslings near fledging at our study site. This period of gosling growth coincides with the 3rd and 4th sampling periods in our study, where the negative impact of warming on plant quality was most pronounced. Therefore, an acceleration of the seasonal decline in plant quality due to climate warming, as documented here, can have serious consequences for gosling growth, regardless of concurrent increases in plant biomass. Additionally, as the negative impacts of warming on plant nitrogen occurred relatively early in the season (July), geese are apparently not able to adjust for a steeper decline in plant nutritive quality by advancing their breeding phenology (Gauthier et al. 2013).

Body size and body condition of goslings at the end of summer are strongly linked to their first-year survival and future reproductive success (Owen & Black 1989; Schmutz 1993; Sedinger, Flint & Lindberg 1995; Menu, Gauthier & Reed 2005). Such effects are not unique to geese, as annual productivity of other arctic herbivores such as caribous (Rangifer tarandus) is also affected by their body condition and quality of their summer forage (Post & Klein 1999; Post & Forchhammer 2008). Thus, a reduction in growth of young arctic herbivores due to a climate-induced mismatch with the phenology of their food plants could have important consequences on their population dynamics (Dickey, Gauthier & Cadieux 2008; Kerby & Post 2013). While arctic herbivores may benefit from climate change in other ways (i.e. latitudinal and altitudinal expansion of breeding grounds, longer summers, higher forage biomass; Jensen et al. 2008; Gilg et al. 2012), our results show that these benefits may be counterbalanced by a loss in forage quality, at least in the short term. To get a more complete understanding of how climate change may affect arctic herbivore populations, further studies are needed to better evaluate the demographic consequences of trophic mismatch and the long-term impact of warming on plant phenology and community composition.


We thank many field assistants who contributed to collecting plant data on Bylot Island over the years, particularly Heidi Kristenson, Mélanie Veilleux-Nolin, Christine Demers, Audrey Jobin-Piché, Marie-Claude Martin and Leslie Qanguq. We are grateful to Gaétan Daigle for his assistance with the statistical analyses. This project was funded by grants from the Natural Sciences and Engineering Research Council of Canada, the Fonds Québécois de Recherche sur la Nature et les Technologies (FQRNT), the International Polar Year program of the Government of Canada, ArcticNet and the Northern Scientific Training Program of the Department of Northern and Indian Affairs Canada. MD was partly funded by a FQRNT scholarship during this project. Logistic support was provided by the Polar Continental Shelf Program (Natural Resources Canada) and Parks Canada. We also thank the Joint Park Management Committee for their assistance.

Data accessibility

Impacts of experimental warming on graminoid plants in wet polygon fens and mesic tundra: Polar Data Catalogue (www.polardata.ca), CCIN Reference Number: 9894.