Soil freezing and N deposition: transient vs multi-year effects on plant productivity and relative species abundance

Authors


Summary

  • Plant responses to increased atmospheric nitrogen (N) deposition must be considered in the context of a rapidly changing climate. Reductions in snow cover with climate warming can increase the exposure of herbaceous plants to freezing, but it is unclear how freezing damage may interact with increased N availability, and to what extent freezing effects may extend over multiple years.
  • We explored potential interactions between freezing damage and N availability in the context of plant productivity and relative species abundance in a temperate old field using both snow removal and mesocosm experiments, and assessed the legacy effects of the freezing damage over 3 yr.
  • As expected, N addition increased productivity and freezing damage decreased productivity, but these factors were nonadditive; N addition increased productivity disproportionately in the snow removal plots, whereas extreme freezing diminished N addition responses in the mesocosm experiment. Freezing altered relative species abundances, although only the most severe freezing treatments exhibited legacy effects on total productivity over multiple growing seasons.

  • Our results emphasize that while both increased N deposition and freezing damage can have multi-year effects on herbaceous communities, the interactions between these global change factors are contingent on the intensities of the treatments.

Introduction

Increased atmospheric nitrogen (N) deposition has altered the productivity and plant species composition of numerous terrestrial ecosystems over the last century (Vitousek et al., 1997; Clark & Tilman, 2008), and rates of atmospheric N deposition are projected to increase further in the coming decades (Galloway et al., 2004). A critical question that must be answered to predict future plant responses to a changing environment is how increased N availability may interact with climate change over this time (Hungate et al., 2003). While much of the research on this topic has focused on the importance of increases in mean annual temperature and growing season length, the case has also been made for the potential importance of extreme climate events in affecting plant communities and ecosystem processes (Jentsch & Beierkuhnlein, 2008). The relative importance of extreme, episodic climate events vs chronic environmental change depends not only on the frequency and intensity of extreme events (Jentsch et al., 2007), but also on the resilience and lifespan of the dominant community members. Therefore, while for forest communities the effects of extreme events can be long lasting, for herbaceous plant communities it is less clear to what extent extreme events may have long-lasting effects (Kreyling et al., 2008a,b, 2010).

Change in winter temperatures has been identified as a critical factor that can affect plant performance in temperate regions, but research on this topic has been greatly underrepresented in the literature compared to that of warming effects during the growing season (Kreyling, 2010). Paradoxically, climate warming can increase soil freezing by reducing snow cover and increasing the exposure of soil to cold air temperatures (Groffman et al., 2001). Warming has already decreased the annual extent of snow cover by 10% in the Northern Hemisphere (ACIA, 2005), and increasing air temperatures are expected to further decrease snow fall in many areas of North America (Kapnick & Delworth, 2013). In many temperate regions, climate warming is therefore expected to increase the freeze–thaw exposure of plants that overwinter beneath the snow (Groffman et al., 2001; Hardy et al., 2001; Henry, 2008). Sub-lethal freezing damage can have substantial effects on plant productivity during the subsequent growing season, but with a few exceptions (Kreyling et al., 2010, 2011), most field studies of freezing effects on herbaceous plants have not extended beyond a single growing season.

Numerous mechanisms have been proposed whereby increased soil freezing can potentially interact with N availability to affect plant growth, including freezing-induced changes in microbial N mineralization (Elliott & Henry, 2009), root N uptake (Malyshev & Henry, 2012a) root damage (Tierney et al., 2001), and soil N leaching and trace gas losses over the winter (Muller et al., 2003; Goldberg et al., 2010; Vankoughnett & Henry, 2013). Grass responses to the combined effects of winter warming and N addition have been explored previously in the field using overhead heaters that supply a constant input of warming (Turner & Henry, 2009; Hutchison & Henry, 2010). While warming in the latter studies increased the frequency of freeze–thaw cycles at the soil surface, the intensity of these cycles was mild (a minimum temperature of −2°C was reached at 2 cm soil depth), and not sufficiently cold to damage the dominant plant species (Malyshev & Henry, 2012a). Therefore, it remains unclear how freezing damage to plants, combined with increased N availability, might affect plant productivity and relative species abundance in the field.

We conducted an experiment in a grass-dominated old field to investigate the combined effects of enhanced freezing (via a single winter of snow removal) and N addition on plant productivity and relative species abundance over three growing seasons. In addition, we exposed plant–soil mesocosms treated with three levels of N addition to a range of controlled freezing treatments in a growth chamber, and then monitored the plant responses over 3 yr in the field. Based on the assumption that increased soil freezing severity would damage roots, we predicted that snow removal would decrease both plant productivity and the ability of plant productivity to respond to N addition. We also predicted that significant treatment effects of freezing on plant productivity would only extend to the second and third growing seasons if there were changes in plant relative species abundance.

Materials and Methods

Site description

The experiments were carried out in a former agricultural field at the Agriculture and Agri-Food Canada research station in London, Ontario, Canada (43°01′46″N, 81°12′52″W). The site was dominated by the perennial grasses Poa pratensis L. and Bromus inermis Leyss., and the forbs Cirsium arvense L. and Lotus corniculatus L. were also common, but patchy. The forbs Asclepias syriaca L., Aster ericoides L. and Solidago altissima L. were also present, but at low densities. The soil was classified as a well to imperfect drained silt loam glacial till (Hagerty & Kingston, 1992), with a mean pH of 7.6 (Bell et al., 2010), and the site had not been ploughed, fertilized or mowed for over 28 yr. According to local climate records (Canadian Climate Normals 1981–2002 or 2006, Environment Canada, National Climate Data and Information Archive) the mean annual air temperature was 7.9°C, with a low monthly mean of −5.5°C (January) and a high monthly mean of 20.8°C (July), and the mean annual precipitation for this site was 1012 mm. Snowfall at the site typically begins in early to mid-December, with snow cover occurring from late December to mid-March. The current atmospheric N deposition rates are 1–2 g m−2 yr−1 in this region (National Atmospheric Deposition Program; Environment Canada).

Snow removal experiment

The snow removal experiment consisted of six spatially distinct blocks each consisting of four (1 m × 1 m) plots (= 24), with each plot placed at least 1 m away from the adjacent plots within each block. Snow was removed from half of the plots in each block, and half of the snow removal and ambient snow plots in each block were fertilized with 6 g N m−2 yr−1of aqueous NH4NO3 for a total of four treatment combinations with six replicate plots each. The N additions were divided equally among three dates in 2010 (8 June, 21 August and 19 October), 2011 (13 April, 10 June and 5 October), 2012 (15 March, 6 June and 1 October), and 2013 (20 March, 6 June and 1 October). The N addition rate approximated projected increases in atmospheric deposition in the study region by the year 2050 (Galloway et al., 2004). Soil temperatures at 5 cm soil depth were recorded in the centers of the plots using Ibutton DS1922L-F5 loggers (Maxim: San Jose, CA, USA; = 3 for each treatment). Before snowfall, white plastic mesh (2 cm × 2 cm; Winter Wrap, Quest Plastic Ltd, Brampton, ON, Canada) was placed on top of the vegetation of both the ambient snow and snow removal plots to minimize litter and soil disturbance during snow removal in the latter. Snow was removed to a depth of c. 2 cm following each snowfall event from 9 December 2010 to 16 February 2011, and periodically when wind-blown snow accumulated in the plots. After 16 February 2011, snow was allowed to accumulate in the plots to avoid differences in soil moisture between the ambient snow and snow removal plots at snow melt. Gravimetric soil moisture content was measured both immediately after snow melt and 4 wk after snow melt by taking 2 cm diameter × 10 cm deep soil cores and drying them for 3 d at 65°C. Snow was not removed over the subsequent winters, and as a result, any snow removal effects in subsequent years were interpreted as legacy effects of snow removal in the first winter.

Mesocosm experiment

Intact plant–soil mesocosms (10 cm diameter × 15 cm deep) were collected by inserting sections of PVC pipe into the ground on 25 May 2010. The mesocosms were collected from six spatially distinct blocks in the field and incubated in the holes from which they had been collected. The mesocosms were randomly assigned to one of three 1-wk freezing treatments (0, −5 and −10°C) and one of three N addition treatments (0, 2 and 6 g N m−2 yr−1 of aqueous NH4NO3). To eliminate any possible carry-over effects of the destructive sampling on the following growing season's productivity, there were three mesocosms per treatment combination in each block, with one assigned to each of the three sampling years, resulting in a total of 162 mesocosms. Nitrogen additions were administered on three dates in 2010 (1 June, 29 July and 11 October), 2011 (13 April, 10 June and 5 October), 2012 (15 March, 6 June and 1 October), and 2013 (20 March, 6 June and 1 October). For the freezing treatments, on 12 March 2011 all mesocosms were removed from the field and placed in incubation chambers at 2°C for 3 h, then brought down at a rate of 0.5°C h−1 to 0, −5 or −10°C for 6 d. Subsequently, the mesocosms were returned to their holes in the field. Although minimum winter air temperatures typically reach from −20 to −25°C at our site, the freezing treatment temperatures of −5 and −10°C covered the range of the most extreme cold temperatures experienced in the soil by the overwintering grasses (Henry, 2008; Malyshev & Henry, 2012a).

Cumulative production estimates

For the snow removal experiment, we estimated the cumulative aboveground plant biomass production in early May, mid-June, early July and early October in each of the three growing seasons using a nondestructive leaf length:mass allometry method adapted from Hutchison & Henry (2010). Specifically, a 10 cm × 10 cm sampling ring was placed randomly in each plot, then all P. pratensis leaves rooted in the sampling area were counted and 9–15 of these leaves were selected at uniform locations for height measurements. The number and heights of inflorescences were recorded separately. Thirty to fifty leaves of varying height were collected from the field outside the plots on the same day, their heights were recorded, and these leaves were individually dried and weighed. For the latter, the logarithm of height was then plotted against the logarithm of mass, and a regression line was fitted to describe the allometric relationship; the latter was used to estimate the masses of leaves in the plots based on their heights. For P. pratensis, r2 ranged from 0.72 to 0.89 over the three sampling years. A similar approach was taken for B. inermis, except that the allometric equations were calculated on a per tiller (i.e. main stem) basis, using the height of the tallest leaf to describe tiller height. For B. inermis, r2 ranged from 0.79 to 0.94 over the three sampling years. The B. inermis tillers senesced in mid-summer following seed set, and then a new round of growth occurred in the late summer. Therefore, for the purpose of estimating cumulative production across the growing season, the production estimates for the first round of growth were added to the late summer/fall production estimates to obtain a cumulative measure of production. P. pratensis did not experience a phase of abrupt leaf senescence in mid-summer, so it was not possible to account for any leaf turnover in the cumulative production estimates for this species. The biomasses of the dominant forbs (C. arvense, A. ericoides and L. corniculatus) were estimated using the same method as for B. inermis, but because these species were less common and patchily distributed, measurements were made for all individuals in each 1 m × 1 m plot for each sampling date (r2 ranged from 0.72 to 0.91 for C. arvense, 0.71 to 0.89 for L. corniculatus, and from 0.81 to 0.84 for A. ericoides, over the 3 yr). The biomass estimates were used to assess relative species abundance. For the mesocosms, the same methods were used.

Root biomass was estimated at snowmelt and on the same sampling dates as the aboveground biomass by coring the soil in each field plot four to five times using a 2-cm diameter corer to a depth of 10 cm, then wet sieving the roots through a 35 mesh sieve (0.5-mm openings) to separate them from the bulk soil. Although the maximum rooting depth of these grasses is at least 30 cm, we chose a sampling depth of 10 cm because it corresponds with the depth where a large proportion of the roots (> 95%) are present (Hutchison & Henry, 2010). A similar approach was used for the mesocosms, except that roots were sampled only in October. In addition, we dried soil sub-samples (c. 10 g) at c. 65°C for at least 3 d to determine the gravimetric moisture content for each root sampling date.

Data analyses

We tested for significant effects of snow removal, N addition and their interaction, on total and species-specific cumulative aboveground production and root biomass using repeated measures two-way analyses of variance (ANOVAs) (JMP 4.0; SAS Institute, Cary, NC, USA). Similar statistical analyses were conducted for the plant–soil mesocosms, but tested for significant differences for temperature, N addition, and their interaction on total aboveground production (Poa + Bromus) and root biomass. Tukey's post-hoc tests were used to assess treatment combination effects on root biomass. All data were checked for normality and log transformed as necessary. For the C. arvense data, which could not be normalized, we used the Scheirer–Ray–Hare nonparametric two-way test in R-3.0.1 (R Development Core Team, 2013), and there was not sufficient representation of the other forbs in the plots to test their responses in all years.

Results

Snow removal experiment

Over the first winter (2010–2011; 1 December to 31 March), the mean air temperature was c. 1°C cooler than and the total snow precipitation was c. 140 mm greater than (almost double) the corresponding mean climate normals (Table 1). In the snow removal plots, soil temperatures decreased on average by c. 1.3°C during the period of snow removal (9 December 2010 to 16 February 2011) and c. 1.1°C throughout the entire winter (1 December to 31 March), with soil temperatures in these plots dropping below 0°C from early to mid-February, and continuing to stay cooler than those of the ambient snow plots until mid-March (Fig. 1). The snow removal plots were below freezing for 43 d, reached a minimum average temperature of −3.1°C at 5 cm depth, and experienced five freeze–thaw cycles, whereas the ambient snow plots experienced a minimum average temperature of 0.3°C, and thus were never below freezing or experienced freeze–thaw cycles over the winter. The gravimetric soil moisture measured immediately after and 4 wk following snow melt during the first growing season was not significantly affected by snow removal (= 0.66 and = 0.20, respectively). The treatment effect on soil moisture was not characterized in the subsequent years, because snow removal was not conducted during the second and third winters.

Table 1. Mean air temperature, total precipitation and water equivalent snow precipitation over the first (2010–2011), second (2011–2012) and third (2012–2013) winters (1 December to 31 March), and mean air temperature and total precipitation during the first (2011), second (2012) and third (2013) growing seasons (1 April to 31 November), relative to the climate normals
 Climate normalsFirst yearSecond yearThird year
  1. Data from Environment Canada, National Climate Data and Information Archive. Mean winter and growing season air temperature, and total winter and growing season precipitation climate normals range from years 1981 to 2002 or 2006.

Mean winter air temperatures (°C)−3.2−4.41.2−2.8
Total winter precipitation (mm)298.7371.8306.6318.3
Total water equivalent snow precipitation (mm)164.9304.0128.9136.0
Mean growing season air temperatures (°C)13.414.614.313.6
Total growing season precipitation (mm)712.9829.0455.5766.7
Figure 1.

Mean air temperature (dotted line) and soil temperature (5 cm depth) in the ambient snow (solid line) and snow removal (dashed line) plots from early December 2010 to early April 2011 (nitrogen treatments had no effect on winter soil temperature and thus were pooled; = 3 for each treatment). Air temperature data were from the Environment Canada National Climate Data and Information Archive.

The mean air temperature during the second winter (2011–2012; 1 December to 31 March) was relatively mild (1.2°C), with total winter precipitation similar to the mean climate normals, but including a lower proportion of precipitation falling as snow (Table 1). Low snow accumulation over the second winter, coupled with low precipitation over the following growing season, resulted in drought conditions over much of the growing season. The third winter (2012–2013; 1 December to 31 March) was a more typical winter, with mean air temperatures only 0.4°C warmer and c. 20 mm greater total winter precipitation than the mean climate normals. Soil temperatures at 5 cm depth stayed above 0°C for both the second and third winters.

During the first growing season, snow removal alone reduced the cumulative aboveground production by 46% and N addition alone increased production by 32%, but there was a significant interaction between snow removal and N addition (i.e. their effects were nonadditive), with N addition increasing productivity by 104% in the snow removal plots (Supporting Information Table S1; Fig 2). The grass responses at the species level largely reflected those of total production. For P. pratensis, snow removal alone decreased productivity by 47%, N addition alone increased productivity by 44%, and there was a significant interaction between the two factors, with N addition increasing productivity by 138% in the snow removal plots (Table S1; Fig. 2). For B. inermis there was a similar response, with snow removal alone decreasing productivity by 49% and N addition alone increasing productivity by 21%; however, there was no interaction between snow removal and N addition for this species (Table S1; Fig. 2). The forb C. arvense contributed < 1% of the total biomass in the plots, and was not significantly affected by the treatments during the first growing season (Fig. 3). For roots, snow removal reduced biomass by 15%, but N addition had no significant effect (Tables 2, S1).

Table 2. Root biomass (all species pooled) for each treatment combination during the first (2011), second (2012) and third (2013) growing seasons in the snow removal experiment
 SnowmeltMay JuneOctober
(g DW m−2)
  1. Corresponding statistical analyses are displayed in Supporting Information Tables S1–S3. Means within columns for each growing season are not significantly different if they share a common lowercase letter (Tukey's test, P < 0.05). Parentheses indicate standard error (= 6).

First growing season: 2011
Ambient497a (58)443a (29)612a (95)462a (40)
Ambient +N454a (41)655a (69)538a (132)686a (62)
Snow removal390a (54)442a (20)430a (61)548a (104)
Snow removal + N463a (71)518a (71)398a (45)524a (75)
Second growing season: 2012
Ambient236a (77)312a (53)283a (33)480a (53)
Ambient +N293a (67)330a (54)312a (27)547a (122)
Snow removal221a (47)185a (24)259a (23)508a (105)
Snow removal + N224a (33)216a (42)352a (27)486a (89)
Third growing season: 2013
Ambient474a (127)301a (74)341a (44)600a (104)
Ambient + N711a (161)197a (55)366a (56)708a (60)
Snow removal395a (122)385a (130)351a (91)469a (79)
Snow removal + N424a (111)455a (59)312a (58)629a (71)
Figure 2.

Total, Poa pratensis, and Bromus inermis cumulative net aboveground production (g m−2) for the first (2011), second (2012) and third (2013) growing seasons in the snow removal experiment. Data are presented for ambient snow plots (open circles), and snow removal plots (open triangles) and nitrogen (N) addition (filled symbols). Back-transformed means are displayed (= 6); error bars, ± SE. Significant ANOVA effects are displayed in each panel (SR, snow removal; N, nitrogen; D, date).

Figure 3.

Biomass of Cirsium arvense in the first, (2011), second (2012) and third (2013) growing seasons in the snow removal experiment. Data are presented for ambient nitrogen (N; open bars) and added N (closed bars). Raw means are displayed (= 6); error bars, ± SE. Significant effects (ScheirerRayHare two-way nonparametric tests) are displayed in the panel (SR, snow removal).

During the second growing season, the cumulative aboveground production was reduced by 36% in response to the legacy of snow removal, N addition alone increased production by c. 10%, and these effects were additive (Table S2; Fig. 2). At the species level, the cumulative production of P. pratensis decreased by 36% as a legacy of snow removal, N addition alone increased productivity by 24%, and these effects were additive. There were no significant treatment effects for B. inermis (Table S2; Fig. 2). The biomass of C. arvense almost tripled during the second growing season in response to the legacy of snow removal, although this species still only contributed 2 and 0.5% of the total production in the snow removal and ambient snow plots, respectively (Table S2; Fig. 3). N addition had no effect on C. arvense biomass (Table S2; Fig. 3), and both the legacy of snow removal and N addition had no significant effects on the biomass of other forbs. There was a trend towards decreased root biomass as a legacy of snow removal during the second growing season (= 0.057; Tables 2, S2), but no significant effect of N addition.

During the third growing season, N addition alone increased the cumulative aboveground production by 35%, but unlike the previous year, there was no legacy effect of snow removal (Table S3; Fig. 2). At the species level, there were no significant legacy effects of snow removal for P. pratensis, but N addition alone increased production in this species by 33%. While B. inermis was unaffected by both N addition and the legacy of snow removal (Table S3; Fig. 2), Cirsium arvense biomass remained elevated in response to the legacy of snow removal, but was unaffected by N addition (Table S3; Fig. 3). There were no effects of N addition or the legacy of snow removal on root biomass (Tables 2, S3).

Mesocosm experiment

Similar to the snow removal experiment, the cumulative aboveground grass production in the mesocosms was reduced during the first growing season with increasing freezing severity, and N addition increased production. However, unlike the snow removal experiment, the N addition effect and temperature manipulation effects were additive (Table S4; Fig. 4). Root biomass also decreased with increased freezing severity (Tables 3, S4). Similar responses were observed for aboveground production and root biomass during the second and third growing seasons, but only the legacy of the most severe freezing treatment (−10°C) elicited significant responses (Tables 3, S4; Fig. 4). Likewise, the significant N addition effect was driven primarily by increased production in the 6 g N yr−1 treatment in both of these years (Table S4; Fig. 4).

Figure 4.

Total grass cumulative production in the first (2011), second (2012) and third (2013) growing seasons in the mesocosm experiment. Data are presented for 0°C (white circles and dotted lines), −5°C (white triangles and dashed lines) and −10°C (white squares and solid lines), and 2 g m−2 yr−1 of NH4NO3 (gray symbols), and 6 g m−2 yr−1 of NH4NO3 (black symbols). Back-transformed means are displayed (= 6); error bars, ± SE. Significant ANOVA effects are displayed in each panel (T, temperature; N, nitrogen; D, date).

Table 3. Mesocosm root biomass (all species pooled) for each treatment combination in October of the first (2011), second (2012), and third (2013) growing season
 First growing season: 2011Second growing season: 2012Third growing season: 2013
(g DW m−2)
  1. Corresponding statistical analyses are displayed in Supporting Information Table S4. Means within columns are not significantly different if they share a common lowercase letter (Tukey's test, P < 0.05). Parentheses indicate standard error (= 6).

0°C241ab (40)137a (28)233ab (81)
0°C, 2 g N257ab (37)126a (59)296a (79)
0°C, 6 g N336a (62)102a (19)188ab (33)
−5°C130ab (25)97a (22)87b (39)
−5°C, 2 g N159ab (62)68a (28)128ab (19)
−5°C, 6 g N212ab (70)67a (28)140ab (23)
−10°C102b (35)30a (10)38b (16)
−10°C, 2 g N95b (25)34a (10)102ab (29)
−10°C, 6 g N98b (33)55a (15)92ab (31)

Discussion

We predicted that the response of plant productivity to N addition would be reduced in the snow removal plots, and although the highest total productivity was observed in the ambient snow plots that received N, the largest increase in productivity in response to N addition occurred in the snow removal plots. Thus, the influence of snow removal on the N response was the opposite of what we predicted. Our prediction had been based on the assumption that freezing damage to roots in the snow removal plots would decrease the capacity for root N uptake, as has been observed in forest systems (Fitzhugh et al., 2001). In our study, snow removal reduced soil temperatures to −3.1°C at 5 cm depth, which is sufficient to damage the roots of Poa pratensis (Malyshev & Henry, 2012a). Consistent with this observation, snow removal decreased aboveground production and root biomass (pooled over all species) over the following growing season. However, the explanation for why plants in the ambient snow plots were less responsive to N addition than the plants in snow removal plots is likely based on growth-limiting factors other than N. Specifically, with N addition, the plants in ambient snow plots exhibited relatively high aboveground biomass, and thus may have experienced increased light limitation (e.g. Wilson & Tilman, 1991) or possibly phosphorus limitation (e.g. Menge & Field, 2007) relative to the plants in the snow removal plots.

The interaction between N addition and snow removal only lasted a single season, possibly as a result of the drought conditions experienced in the second growing season, which limited total productivity across all treatments. Nevertheless, the legacy effects of snow removal on plant productivity carried over to the second growing season. The latter result was inconsistent with the other (albeit a limited number) studies that have examined multi-year responses to frost events in grasses (Kreyling et al., 2010, 2011), although the freezing treatment in our study was likely more severe than those used in these. Even over the third growing season, when the legacy effect of freezing on total productivity was no longer present, increases in the biomass of C. arvense persisted. Circium arvense is an opportunistic species that often increases in biomass in response to reduced competition (Thrasher et al., 1963; Wilson & Kachman, 1999; Edwards et al., 2000). Therefore, the observed increase in C. arvense biomass was likely caused by this species capitalizing on the reduced grass biomass in the snow removal plots, and its expansion may have been further aided by the reductions in grass biomass in response to drought in the second growing season. Episodic events are important catalysts that can change the competitive interactions among plants and alter successional pathways (Jentsch & Beierkuhnlein, 2003; Jentsch et al., 2007). Accordingly, beyond several years, the longer-term effects of freezing damage in herbaceous systems may be most evident when changes in species composition or relative species abundance occur. Cirsium arvense can spread from seed, but germination is often low and erratic (Tiley, 2010); thus it is most likely that the increase in biomass observed in the snow removal plots was likely caused by vegetative growth. Although the increases in C. arvense in the snow removal plots were relatively minor as a proportion of the total plant biomass in these plots, this species can have prolonged negative effects on grass yield as a result of resource competition (O'Sullivan et al., 1982, 1985; Grekul & Bork, 2004).

The mesocosm experiment confirmed that grass recovery over multiple years is contingent on freezing severity, based on the observation that plant productivity only remained depressed in the second and third growing seasons in response to the most severe freezing treatment (−10°C). There are often large differences in freezing severity experienced by grasses in unmanaged vs managed systems; at our field site, the combination of snow and a thick litter layer kept soil temperatures near 0°C, and even when the snow was removed, soil temperatures only reached an overnight minimum of −3.1°C during the winter, with similar temperatures persisting over several days, despite air temperatures periodically reaching −20°C (Fig. 1). Only when litter and snow removal have been conducted simultaneously have soil temperatures approached −10°C at our field site (Malyshev & Henry, 2012a). The latter approximates what is often experienced in highly managed systems, where grasses are mowed or grazed, leaving behind a relatively thin litter layer and exposing overwintering shoot bases and roots to the air.

As a result of the intense freezing experienced by grasses in managed systems, coupled with the economic losses caused by frost kill (Ouellet, 1976), freezing studies conducted on forage and turfgrasses have mainly focused on lethal temperatures (Gudleifsson et al., 1986; Tcacenco et al., 1989; Hanslin & Höglind, 2009); however, these lethal temperatures (−15 to −30°C based on the previous studies) are typically much lower than the soil temperatures that occur in many temperate regions, and lower than the soil temperatures predicted to occur with climate change in these regions over the next century (Henry, 2008). Therefore, our study adds to the growing body of results that demonstrate the potential importance of sub-lethal freezing damage to the apical meristems and roots of grasses that occur at less severe freezing temperatures (Eagles et al., 1993; Malyshev & Henry, 2012a,b).

In the first year of the mesocosm experiment, freezing damage intensified from 0 to −5 to −10°C. However, the effect of 1 wk at −10°C persisted over all 3 yr and, consistent with our initial prediction (and contrary to the snow removal experiment), the N addition response was reduced in these mesocosms, which may have occurred as a result of the elevated freezing intensity relative to the snow removal experiment. Freezing damage can increase in severity at increasingly severe sub-zero temperatures because the concomitant decrease in water potential exacerbates cellular dehydration (Gusta et al., 1975). However, in addition to minimum temperature, other factors such as freezing duration can increase damage (Malyshev & Henry, 2012a), which is potentially caused by plant cellular dehydration damage increasing in severity the longer that tissue is frozen. Furthermore the timing of freezing can be important, because warm spells can prematurely deacclimate plants and leave them susceptible to freezing damage if snowmelt is followed by the return of severe frost (Bokhorst et al., 2008, 2009, 2011). In the snow removal experiment, plants experienced 43 d of freezing soil temperatures, five freeze–thaw cycles, and 2 d of unseasonably warm weather with air temperatures reaching a high of 9°C at the end of December and the beginning of January (Fig. 1), all of which may explain why these plants exhibited reduced productivity 2 yr after the treatment, while those exposed to −5°C over the short term (1 wk) in the mesocosms only showed a significant response after 1 yr.

Much like the temperature response, N addition level was influential in the mesocosm experiment, with productivity increasing significantly in response to 6 g N m−2 yr−1, but not in response to 2 g N m−2 yr−1. This distinction is important, given that 2 g N m−2 yr−1 represents the lower bound on estimates of N deposition increases in our region, whereas 6 g N m−2 yr−1 is slightly above the upper bound (Galloway et al., 2004). However, the chronic effects of moderate rates of N deposition could potentially affect productivity through changes in species composition over the longer term (Wedin & Tilman, 1996; Isbell et al., 2013). In addition, in the context of the broader ecosystem, herbivores can play an important role in plant responses to climate change and N availability in our system (Moise & Henry, 2012), and the limited spatial scale of our experiment (both the mesocosms and the 1-m2 area plots) was such that the herbivores would not have always been directly exposed to the treatments.

Conclusions

Our results revealed interactions between plant freezing damage and N addition; N addition promoted recovery from freezing damage in the snow removal experiment, whereas the legacy effects of extreme freezing diminished N addition responses in the mesocosm experiment. However, over multiple seasons, the additive effects of N addition and freezing damage were dominant in defining the overall plant responses. While extreme freezing (−10°C for 1 wk) had large effects on plant production over multiple years, in the case of more realistic freezing intensity (snow removal), the only lasting effects were on the increased expansion of C. arvense, a clonal forb. Moderate rates of N addition did not elicit a significant plant response, although unlike extreme frost, N deposition is a chronic effect and N responses may thus accumulate over time. Overall, our results emphasize that while both increased N deposition and freezing damage have the potential for multi-year effects in grass-dominated systems, the potential interactions between these global change factors are highly contingent on the intensities and possibly the duration of the treatments.

Acknowledgements

We thank V. Karas, J. Fraser, J. Joo, J. Lee, B. Mcwhirter, B. Tom and A. Wojcik for their assistance in the field and laboratory. Research funding was provided by an NSERC Discovery Grant awarded to H.A.L.H.

Ancillary