Changes in species abundance after seven years of elevated atmospheric CO₂ and warming in a Subarctic birch forest understorey, as modified by rodent and moth outbreaks

Study site

The study took place in a low-statured birch forest 10 km east of Abisko Scientific Research Station in the Subarctic region of Sweden (68°21′N, 19°4′E, 341 m a.s.l.). As is typical for the heath forest types of the region (Carlsson, Karlsson & Svensson, 1999), mountain birches (Betula pubescens ssp. czerepanovii (Orlova) Hämet-Ahti) were sparsely distributed and the field layer dominated by the three ericaceous dwarf shrubs Vaccinium myrtillus L., V. vitis-idaea L., and Empetrum nigrum ssp. hermaphroditum (Hagerup) Böcher. In smaller amounts, the dwarf shrub V. uliginosum L., the stoloniferous Linnaea borealis L., the herbaceous pteridophyte Equisetum sylvaticum L., and the grasses Avenella flexuosa (L.) Drejer and Calamagrostis purpurea (Trin.) Trin. were found. Common bryophytes were Hylocomium splendens (Hedw.) B. S. & G., Dicranum spp. (mainly D. scoparium Hedw. and D. fuscescens Turn.), Pleurozium schreberi (Brid.) Mitt. and species of Barbilophozia Loeske, while lichens were rare. The area was chosen as to ensure as little variation as possible within the site. The climate at the site is similar to that recorded at the Abisko Scientific Research Station, where the mean annual temperature is –0.5 °C and mean annual precipitation is 320 mm (Eriksson, 1982; Eriksson, 1983). The field experiment was approved by the County Administrative Board in Norrbotten, Sweden (no. 231-6447-01; 2502 035).

Experimental setup

The experiment was established early in the growing season of 2000 and included a warming treatment and an elevated CO₂ treatment in a blocked two-way factorial design, and was run until September 2006, i.e., for seven seasons. We used six c. 45-m² blocks; within each block five experimental plots were established. Each plot was 1.5  × 0.75 m, and four of the five plots in each block were during the growing season (late May to early September) surrounded by a hexagonal open-top chamber (OTC) with a footprint of c. 1.5 m². The walls of the OTC were made of non-UV-proof transparent acrylic sheets.

Four chambers in each of the six experimental blocks were randomly assigned to one of the following four treatments: (1) warmed chambers in which soil and air temperatures were elevated 5 °C above ambient, (2) CO₂ chambers with doubled internal atmospheric CO₂ concentrations, (3) combined warming and CO₂-enrichment, (4) ambient control chambers receiving no elevated CO₂ or warming. In addition, one plot, (5) “non-chambered control”, was established without the OTC in each of the six blocks, to discriminate treatment effects from experimental artefacts. Wooden boardwalks were erected to enable plots to be accessed without disturbing the vegetation.

The soil was warmed using resistance cables threaded through the organic upper layer of the soil. Heating cables were controlled by data loggers coupled to thermistors—three in warmed chambers, two in ambient chambers. The electricity through the cables was switched on and off automatically on a two-minute cycle to maintain the temperature difference between warmed and ambient chambers. Within the warmed chambers, three infrared lamps, suspended 1.2 m above the soil surface, warmed the plant surfaces with 22 Wm⁻².

In CO₂-enriched chambers, atmospheric concentrations were elevated to 730 ± 25 ppm CO₂. A LiCor 6262 Infrared Gas Analyser monitored CO₂ concentrations in four of the six enriched chambers. A data logger coupled to flow sensors and a mass flow controller regulated the flow of CO₂ from the tanks to the chambers. CO₂-enriched air was blown into chambers from two sides to maintain an even concentration of CO₂ across each chamber. Non-CO₂-treatment chambers had an identical design, but received air without CO₂. This ensured that all chambers experienced similar effects of blowing, e.g., reduced convective heating effects within the chambers. Both warming and CO₂-enrichment were applied during the growing season each year. The OTCs were removed during winter.

We set up one 0.75 × 0.75-m permanent subplot in each of the five treatments—warming, CO₂, warming plus CO₂, chambered, and non-chambered controls—within each block for vegetation and demography analyses. Homogeneity of vegetation was emphasized in the selection of the subplots to ensure as little variation as possible within each subplot.

Point-frequency analysis

The occurrence of all vascular plants, bryophytes and lichens, as well as litter and bare soil, was recorded using a 10-mm-thick quadratic sheet of transparent acrylic glass. The sheet had 97 holes evenly drilled at 5-cm intervals and was standing on three adjustable metal legs enabling us to position the frame horizontally and in the same place every year, with the help of small metal markers on the soil surface and a water-level. The holes had been precision-drilled using a milling cutter to enable a 5-mm-diameter brass rod to be inserted with a minimum of angular error. All hits on a 20-mm-long zone on the sharpened tip of the rod were recorded as it was moved downward through the vegetation canopy. Whenever possible the final record would be that of litter (one or two hits) or a bryophyte. In the case of bryophytes, most of them might have more than one hit while Barbilophozia sp. never had more than one.

Statistical analyses

Total number of point hits (a proxy for species abundance or dominance in 3D-space) for vegetation collectively (excluding litter), and for V. myrtillus, V. vitis-idaea, E. nigrum ssp. hermaphroditum, A. flexuosa, bryophytes and forbs, were analysed with Mixed effects ANOVAs with the random factor Block, and the fixed factors Year, Warming and CO₂, the last two with two levels, and their interactions (Zar, 1999). To eliminate differences due to initial conditions the analyses were made using number of hits in 2000 as covariables. All data was log-transformed (x′ = ln(x + 1) before analyses to meet the assumptions of the statistical tests.

There were no significant differences in number of hits between non-chambered controls and treatment controls (ambient) for any of the species studied, except V. myrtillus (Two-way ANOVA, Chamber: F_1,5 = 4.8, P = 0.033; Year: F_1,5 = 40.7, P < 0.001; Interaction Chamber × Year: F_1,5 = 0.878, P = 0.502) and A. flexuosa (Two-way ANOVA, Chamber: F_1,5 = 12.2, P = 0.001; Year: F_1,5 = 3.13, P = 0.015; Interaction Chamber × Year: F_1,5 = 0.832, P = 0.533). For both species there were initially more hits in the chambered control plot (V. myrtillus: 15%, A. flexuosa: 25%) (Fig. 1), but as the initial abundances of the species were included as co-variables in the analyses the results suggest that the chamber indeed had a positive influence, albeit small. However, since no other species showed any increased (or decreased) growth in the chambered control compared to the non-chambered control we henceforth did all analyses excluding the non-chambered controls to keep a balanced design.

As a proxy for canopy cover or dominance in 2D-space we analysed number of first hits per species in the same manner as for total number of hits.

We also tested whether there were any preferences by the voles and moths for any particular treatment. We did this by noting the degree of defoliation and the frequency of dead shoots on initially five marked V. myrtillus in each plot 2001–2006. In all cases we used a six-level scale, where 0 means no herbivory effects were detected. As above, a Mixed effects ANOVA with the random factor Block, and the fixed factors Year, Warming and CO₂, the last two with two levels, and their interactions were used to test for differences between treatments. We also checked whether there were any differences in herbivory between chambered and non-chambered controls using a Two-way ANOVA with the fixed factors Year and Chamber and their interaction.

All analyses were made using IBM SPSS Statistics 24.0.0.0 (SPSS, Inc., Chicago, IL, USA).

Plant community responses to herbivory

Irrespective of treatment, there were marked differences between years, particularly regarding the abundance of V. myrtillus (Fig. 1A), with a small decrease in 2001, a recovery in 2002 followed by striking decreases in both 2003 and 2004. The decrease in vegetation abundance in the early phase of the experiment was caused by grazing by the Grey red-backed vole (Clethrionomys rufocanus) during the rodent peak in 2001. Data from separate demographic analyses (B Svensson, pers. obs., 2001)—where we followed the fates of individual shoots—show that 71 out of 183 (39%) V. myrtillus shoots were partly or completely grazed by rodents in 2001. The year after, 2002, we noted that shoots of V. myrtillus were still dying as a result of the herbivore attacks. In 2003, a heavy outbreak of the Autumnal moth (Epirrita autumnata) began in the area, including also our study site. The impact was even more severe in 2004 (Fig. 1A). For example, we observed that in most plots the shoots of V. myrtillus were completely defoliated.

Thus, the degree of herbivory on V. myrtillus differed markedly between years (Fig. 2, Table 1). There was, however, no difference in the degree of herbivory between treatments (Mixed ANOVA, Table 1). Also, there was no difference between chambered and non-chambered controls (F_1,60 = 0.527, P = 0.471).

Plant community responses to warming and CO₂ enrichment

A notable change was detected in the warmed plots in 2004 and 2005, where we found a shift in community composition (measured as the total number of hits), from the deciduous V. myrtillus to the evergreen V. vitis-idaea becoming the most common species (Fig. 3). In 2006, the two species were equally common so the shift may have been of a transient nature.

Because of initial differences between chambers we illustrate treatment effects using differences between each year and the starting year (i.e., no. of hits in 2001 minus no. of hits in 2000, no. of hits in 2002 minus no. of hits in 2000, etc. (Fig. 4). Positive effects of warming were found for V. myrtillus, V. vitis-idaea, E. nigrum ssp. hermaphroditum, A. flexuosa, and a near to significant positive effect also on forbs collectively (Fig. 4, Table 2). Although there was a strong general decrease in the number of hits for V. myrtillus (due to herbivory), V. myrtillus in the warmed plots fared relatively better (Fig. 4A, Table 2b). For V. myrtillus we also found a significant three-way interaction effect Warming × CO₂ × Year (Table 2b), probably due to the years 2002 and 2006 where warming alone had a more positive effect than combined with elevated CO₂, as opposed to the other years (Fig. 4A). Not until the last year (2006) were the abundance values back to the same level as before the vole and moth outbreaks, and only so in the warmed plots. One of the blocks had fewer point hits initially and also reacted differently to vole grazing. This was due to the lower abundance of V. myrtillus in this block.

A statistically significant negative effect of added CO₂ was found for the other common dwarf shrub, V. vitis-idaea, (Fig. 4B, Table 2c). This species did not react as negatively to the vole and moth outbreaks; its abundance never went below the starting level (Fig. 4B). In 2005 and 2006, V. vitis-idaea recovered from the herbivore attacks also in the other two treatments. In contrast to V. myrtillus, no grazing was observed on V. vitis-idaea leaves.

The dwarf shrub Empetrum nigrum ssp. hermaphroditum did not seemingly react as negatively to the vole or moth outbreaks as V. myrtillus but still had its lowest abundance values in 2004. Empetrum responded positively to warming as can be seen by the increased difference in number of hits in 2005 and 2006 in the two warmed treatment combinations (Fig. 4C, Table 2d). There was only one graminoid species in the analysed plots, viz. Avenella flexuosa, and this species also increased in the two warmed treatments (Fig. 4D, Table 2e). The other, less common herbaceous vascular plant species (i.e., forbs collectively) were unevenly distributed between the plots and only showed a significant response to the experimental treatments at the P < 0.1 level (Fig. 4E, Table 2f), nor could we see any herbivory effects.

In contrast to the dwarf shrubs, warming decreased the abundance of bryophytes (Fig. 4F, Table 2g), particularly in the last year. Also, bryophytes did not respond strongly to herbivory outbreaks as the notable decrease in the years with herbivore attacks for V. myrtillus is less marked for bryophytes (Fig. 4F). We also found a significant positive effect of added CO₂, and a significant interaction effect between added CO₂ and warming (Table 2g), as seen by the decrease in the number of hits in the combined treatment compared to the slightly positive effect in the CO₂-only plots (Fig. 4F).

We also tested whether species composition of the top-most, first hit, differed between treatments, which would suggest a change in canopy cover between species. The significant positive effect of warming remained for the vegetation as a whole and for V. vitis-idaea and E. nigrum ssp. hermaphroditum but disappeared for V. myrtillus and A. flexuosa. The pattern that in 2004 and 2005 V. myrtillus became less common than V. vitis-idaea in the warmed plots remained, however. There was still a strong negative effect of warming on the abundance of bryophytes.

Accelerated growth in the warmed plots

The evergreen dwarf shrub V. vitis-idaea and the deciduous V. myrtillus were the most responsive species to the warming treatments (warmed and combined warming + CO₂). The third ericaceous species, the evergreen E. nigrum ssp. hermaphroditum reacted similarly and positively to warming, but not as strongly. An increase in shrub abundance is as expected (Myers-Smith et al., 2011) and also in accordance with the meta-analysis results of Arft et al. (1999).

As hypothesized, V. myrtillus was the species responding the strongest to warming, as was also found by Van Wijk et al. (2003) in their meta-analysis of over 40 studies. There was also a small positive chamber effect on the growth of this species. However, the difference between treatments was not particularly large, probably because the response was obscured by the herbivore attacks. Vaccinium myrtillus was particularly liked by the Grey red-backed vole during its population peak in 2001 (pers. obs.), which explains the decrease in abundance of V. myrtillus in this year compared to the start of the experiment (2000). Judging from the strong negative response in V. myrtillus in the years 2003 and 2004, where the average number of hits per plot was reduced by 47 and 91% (Fig. 1), respectively, it seems very likely that this species was particularly palatable also to E. autumnata larvae.

In addition, herbivores may affect soil nutrient availability as a direct effect of faeces deposition. Since the herbivores may eat substantial amounts of the green biomass, an increase in available soil nutrients is expected from dying roots and from faeces, particularly from invertebrate faeces (frass), as these are readily decomposed (Kagata & Ohgusju, 2012) and can increase nutrient availability (Hollinger, 1986; Lightfoot & Whitford, 1990; Frost & Hunter, 2004; Frost & Hunter, 2007). For example, it has been suggested that frass excreted by E. autumnata larvae may fertilize Betula pubescens spp. czerepanovii (Haukioja, Suomela & Neuvonen, 1985), the overstorey species at our experimental site. This is not surprising as frass may contain nearly 90% of the leaf N that was originally consumed (Hollinger, 1986; Lightfoot & Whitford, 1990; Frost & Hunter, 2007). Avenella flexuosa was the only grass species present in the analysed plots and, as hypothesized, responded positively to warming. This fits well with the above as an increase in grasses is expected when nutrient availability is increased, which has been demonstrated in a number of studies (e.g., Karlsen et al., 2013), and in relation to global warming in a meta-analysis study by Van Wijk et al. (2003). We also detected an increase in A. flexuosa flowering frequencies in 2005 and 2006, which may be due to the increased light levels after the herbivore attacks (Scurfield, 1954). Due to the positive feedback loops emanating from the released nutrients from herbivores (carcasses and frass) and dead roots (Kaarlejärvi, Hoset & Olofsson, 2015) the vegetation will recover quickly, even in the absence of extra CO₂ or warming. An additional nutrient source may be the litter from E. nigrum ssp. hermaphroditum. We observed that the cuticulas on young shoots of this species were gnawed by the moths, and litter from this species increased at the same time as living shoots increased. Interestingly, due to the high phenolic content in E. nigrum ssp. hermaphroditum leaves, its litter decomposes slowly and an organic top soil layer is formed. Nutrients are thus mainly available for plants with ericoid mycorrhiza (Tybirk et al., 2000 and references therein) and this may benefit the two Vaccinium species in our study.

We did not find the expected higher levels of herbivory in the warmed plots. If the effect of warming induces higher resilience from stochastic disturbances in subarctic systems by boosting plant growth after herbivore attacks, the effect of the herbivores may be obscured. Or is it the herbivores that contribute to the resilience of the system by intermittently removing biomass, returning the vegetation to its original lower level, as suggested by Kaarlejärvi, Hoset & Olofsson (2015)? Such top-down controls have been shown to be strengthened by warming in cold regions (Marino, Romero & Farjalla, 2018).

A shift in community composition?

Vaccinium vitis-idaea strongly benefitted from the reduced presence of V. myrtillus, as it increased in all plots, particularly in the warmed plots. In particular, the abundance of taller shoots of V. myrtillus was reduced (as seen by the non-significant effect of warming on the number of first hits). This resulted in a community shift. Likewise, Rinnan, Stark & Tolvanen (2009) simulated V. myrtillus herbivory and found decreased abundance, which at the same time benefitted evergreen shrubs (Ylänne, Stark & Tolvanen, 2015). However, as the two species were again equally common in the last year of the experiment, the shift may be of a transient nature, perhaps because of the particularly positive effect of warming on V. myrtillus. The fact that V. myrtillus reacted positively (albeit weakly) on the presence of the chamber itself supports this.

Even though V. vitis-idaea is only occasionally used as a food source for most vole species (Kalela, 1957; Batzli & Lesieutre, 1991), an explanation for the positive response in the warmed plots could be an increased defence from secondary substances in the leaves of V. vitis-idaea. Concentrations of tannins have been suggested to increase after warming and nutrient addition (Graglia et al., 2001; Hansen et al., 2006). However, this does not seem likely in our case as the degree of herbivory did not differ between treatments. In addition, others have found little or no effects on tannin concentrations with warming (Kaarlejärvi et al., 2012). Interestingly, during our demographic analyses (B Svensson, pers. obs., 2003; 2004) we observed that shoot apices, rather than leaves, of V. vitis-idaea were consumed by the Autumnal moth, which in the following year was followed by a substantial increase in the number of side shoots. So, irrespective of a possible leaf tannin concentration increase, V. vitis-idaea increased in abundance due to released apical dominance resulting in increased branching (Svensson & Callaghan, 1988) as a response to herbivory.

Bryophytes

Bryophytes reacted negatively to warming, which has been shown before (Healey, Oberbauer & Hollister, 2016). However, a low number of hits for bryophytes, particularly in combination with many hits for vascular plants, does not necessarily mean that bryophytes have decreased in frequency when the abundance of vascular plants increased. Rather, as the vegetation at this site is dense it may be an effect of bryophytes being less visible under the vascular plant layer and an artefact of the point-frequency analysis method. This is indicated by the increased number of bryophytes hits in the year 2004, when herbivory intensity was high and vascular plant abundance reduced. This increase could also, however, be a response to increased light levels this year.

Minor effects of added CO₂ alone

We did not expect any response to elevated CO₂ alone, as tundra plants generally are nutrient limited. However, in an earlier study using the same experiment (Olsrud et al., 2010) where we evaluated the effects on global climate change on ericoid mycorrhizal function after two years’ exposure, we found an increased soil organic N availability in the CO₂-only plots. This could probably be a response to mycorrhizal activity in CO₂-plots (Emmerton et al., 2001; Read & Perez-Moreno, 2003) as a simultaneous increased mycorrhizal colonization was found under elevated CO₂ (Olsrud et al., 2010). This might explain the transient positive response to CO₂ found for V. myrtillus during the first two years of the experiment. The significant negative effect of added CO₂ on V. vitis-idaea contradicts this. However, see Feng et al. (2015) for possible mechanisms whereby increased CO₂ might constrain nitrogen acquisition and growth.

In many global climate change experiments, we find transient above-ground positive growth responses to enhanced CO₂ (e.g., Oechel et al., 1994; Arft et al., 1999). For example, an increased growth of forest trees was seen initially after which it declined (Fig. 4 in Körner, 2006). This is because the plants enter a new steady-state regarding, e.g., leaf area and root turnover (Körner, 2006). However, herbivory may interact with how the vegetation is responding to these changes in CO₂ concentration. A strong herbivore attack will offset this steady-state as a result both from released above-ground competition (as the canopy is more or less destroyed, depending on the size of the herbivore attack) as well as an increase in soil nutrients after the decomposition of dead roots and the direct fertilizing effect of herbivores in the form of frass. Vegetation responses to enhanced CO₂ will start anew after the herbivore attack.