• Emissions of isoprene, a reactive hydrocarbon, from Subarctic vegetation are not well documented. However, the Arctic is likely to experience the most pronounced effects of climatic warming, which may increase temperature-dependent isoprene emission.
• Here, we assessed isoprene emission from a Subarctic heath subjected to a 3–4°C increase in air temperature and mountain birch (Betula pubescens ssp. czerepanovii) litter addition for 7–8 yr, simulating climatic warming and the subsequent expansion of deciduous shrub species and migration of the treeline. The measurements were performed using the dynamic chamber method on a wet heath with a mixture of shrubs, herbs and graminoids.
• Isoprene emissions averaged across the treatments were 36 ± 5 µg m−2 h−1 in 2006 and 58 ± 7 µg m−2 h−1 in 2007. The experimental warming increased the emissions by 83% in 2007 (P = 0.021) and by 56% in 2006 (P = 0.056), while litter addition had no significant effects. The net ecosystem CO2 exchange was significantly decreased by warming in 2007.
• These results show that isoprene emissions from Subarctic heaths are comparable to those from Subarctic peatlands. Climatic warming will increase the emissions, and the amount of carbon lost as isoprene, from Subarctic heath ecosystems.
Major scenarios for global warming predict a 4–5°C increase in annual temperatures in the Arctic by 2080 (ACIA, 2004). Warming with a subsequent decrease in snow cover and increasing precipitation in the Arctic have been evident over the 20th century (Serreze et al., 2000). Reasons for the especially pronounced climatic warming in the Arctic include decreasing albedo as a result of melting of reflective snow cover and sea ice, shallower atmosphere than in lower latitudes, and alterations in atmospheric and oceanic circulation (ACIA, 2004). Consequences of the warming climate have already been detected in Arctic vegetation communities through satellite measurements and aerial photography as migration of the treeline northward and upslope and as increasing abundance of deciduous woody shrubs, phenomena that have also been documented in the observations of indigenous inhabitants (Chapin et al., 1996, 2005; Sturm et al., 2001; Callaghan et al., 2004; Tape et al., 2006). These vegetation changes lead to increasing amounts of leaf litter on the ground (Cornelissen et al., 2007). Models predict the expansion of shrubs to further increase during the 21th century (Epstein et al., 2000). An increasing amount of leaf litter not only brings extra nutrients to the soil (Rinnan et al., 2007, 2008) but also decreases the biomass of lichens and mosses by shading close to ground (Cornelissen et al., 2001; Hobbie et al., 2005).
The aim of the present study was to obtain the first estimates of isoprene emissions from a Subarctic dwarf shrub heath ecosystem, and to assess the effects of long-term climatic warming and increasing litterfall on the emissions. Dwarf shrub heaths (also called tundra) are a vast ecosystem covering 1.282 × 106 km2 (23% of the total ice-free land area) in the Arctic (Bliss & Matveyeva, 1992). We hypothesized, firstly, that the experimental warming would increase isoprene emissions as these emissions are strongly dependent on temperature (Guenther et al., 1993; Kesselmeier & Staudt, 1999). Secondly, we hypothesized that increasing litterfall would also increase the emissions. This hypothesis is based on observations of increased isoprene emission in velvet bean (Mucuna sp.) (Harley et al., 1994) and in English oak (Quercus robur) (Possell et al., 2004) under nutrient fertilization. We measured net CO2 assimilation to relate the isoprene emissions to photosynthesis. We also analysed the relationships among environmental factors, the abundance of different vascular plant species, and isoprene emission.
Materials and Methods
The study was conducted in the growing seasons of 2006 and 2007 on an open Subarctic wet heath near Abisko Scientific Research Station in the Lake Torneträsk region in Northern Sweden (68°21′N, 18°49′E, 385 m above sea level). The mean annual temperature (1970–2000) is −0.5°C and the annual precipitation is 315 mm. The mean daytime air temperature during the present study is shown in Figs 1(b) and 2(b). The vascular vegetation of the heath mainly consisted of evergreen and deciduous shrubs, graminoids, herbs and horsetails. The ground layer was dominated by the mosses Sphagnum sp., Tomentypnum nitens (Hedw.) Loeske, the liverworth Ptilidium sp. and lichens. The cover percentage of vascular plant species was analysed by the point intercept method described by Jonasson (1988) in June 2006, and in June, July and August 2007. Mosses and lichens were also included in the calculation but not identified at species level. The vegetation is described in detail in Table 1. The characteristics of the soil are described in Rinnan et al. (2007).
Table 1. Mean cover percentages (± SE) of vascular plant species, mosses and lichens in control, warming, litter addition and combined treatments in August 2007
The experiment simulating climatic warming and increasing litterfall in the Arctic has been running on the heath since 1999. It consists of four treatments replicated in six blocks making up a total of 24 plots (120 × 120 cm each) within an area of 1000 m2. The treatments are: (1) control, (2) warming, (3) litter addition, and (4) warming combined with litter addition. The treatments within each block have been randomized. The experimental warming is achieved using dome-shaped open-top chambers made of polyethylene (thickness 0.5 mm), which increase the air temperature by 3–4°C and the soil temperature at 0–5 cm depth by 1°C but also reduce the photosynthetically active radiation (PAR) by c. 10% (Havström et al., 1993; Graglia et al., 2001; Rinnan et al., 2008). Plots under litter addition have received 90 g dry weight (DW) m−2 of mountain birch (Betula pubescens ssp. czerepanovii (N. I. Orlova) Hämet-Ahti) litter every autumn since the beginning of the experiment. The added litter contains c. 9.8 mg g−1 DW nitrogen and 1.2 mg g−1 DW phosphorus (Jonasson et al., 2004). The open-top chambers are set up each year in mid-June and taken down in the end of August, and the leaf litter is added in late August.
Isoprene sampling and analyses
Samples for isoprene analysis were taken in six campaigns from 13 June to 29 August 2006 and in eight campaigns from 14 June to 30 August 2007. One campaign in 2007 was carried out while open-top chambers were removed to see whether there were indirect effects of warming. Otherwise, the open-top chambers were in place during sampling. Sampling was performed with a transparent polycarbonate chamber (thickness 1.5 mm, 23 × 23 cm, height 25 cm; Vink Finland, Kerava, Finland). The chamber was placed on an aluminium collar (20 × 20 cm) installed permanently in each plot. The collar grooves were filled with water to ensure air enclosure. The sample was pulled with a small battery-operated sampling pump (12 V; Rietschle Thomas, Puchheim, Germany) into a stainless steel tube (ATD sample tubes; Perkin Elmer, Boston, MA, USA) containing Tenax TA and Carbopack B adsorbents (Tenax TA, 100 mg and Carbopack B, 100 mg, mesh 60/80; Supelco, Bellefonte, PA, USA).
The air flow in the sampling pumps was set to 205 ml min−1 with a flow meter (Agilent Flow Tracker 1000; Agilent Technologies Inc., Wilmington, DE, USA). This corresponds to a flow of 200 ml min−1 through the sample tube. The chamber was maintained slightly over-pressured by pumping in charcoal-filtered air through a MnO2 scrubber with another pump at a rate of 215 ml min−1 in order to avoid gas leakage from outside into the chamber. During the 30-min sampling period the air was circulated inside the chamber with a small fan, and the temperature and air humidity inside the chamber were monitored (Tinyview Plus; Gemini Data Loggers Ltd, Chichester, UK). PAR was monitored on the chamber with a quantum sensor (Li-Cor, Lincoln, NE, USA). The sample tubes were immediately sealed with Teflon-coated brass caps, stored refrigerated and analysed at the University of Kuopio within 3 wk.
The samples were analysed by gas chromatography–mass spectrometry (Hewlett-Packard 6890, MSD 5973; Hewlett-Packard, Palo Alto, CA, USA) after thermodesorption at 250°C and cryofocusing at −30°C with automatic thermal desorber (ATD400; Perkin Elmer, Wellesley, MA, USA). Isoprene was separated using an HP-5 capillary column (50 m × 0.2 mm; film thickness 0.33 µm). The carrier gas was helium. The oven temperature was held at 40°C for 1 min and then raised to 210°C at a rate of 5°C min−1, and finally raised to 250°C at a rate of 20°C min−1. Isoprene was identified according to the mass spectra in the Wiley data library (Wiley & Sons Ltd., Chichester, UK), and quantified with a pure standard (Fluka, Buchs, Switzerland).
The emission rates were calculated by multiplying the isoprene concentration in the sample by the chamber volume to obtain the absolute amount of isoprene in the chamber. The surface topography inside each collar was taken into account when determining the chamber volume. The absolute amount of isoprene in the chamber was then proportioned to the surface area of the plot and to the sampling time to obtain the emission in µg isoprene m−2 h−1.
The static chamber method was applied to measure the net ecosystem CO2 exchange (NEE) three times in the same experimental plots during the growing season of 2006 and four times in 2007. We used portable infrared gas analysers (EGM-2 in 2006 and EGM-3 in 2007; PP Systems, Hitchin, UK) to measure the CO2 concentration in the transparent chamber headspace. In 2006, a 10-ml air sample was pulled with a syringe from the chamber headspace at the beginning of the sampling period and 2 and 4 min thereafter. The air sample was immediately injected into the EGM-1 set to the static mode and the CO2 concentration was recorded. In 2007, the air was continuously circulated in a closed system from the chamber headspace to the EGM-3 set to the dynamic mode for a 2-min sampling period. The CO2 concentration was recorded every 15 s and the chamber headspace was ventilated. For both years, the PAR and chamber temperature were monitored during the measurements. CO2 exchange rates were calculated from the linear change in CO2 concentration in the chamber headspace. The NEE was proportioned to the surface area of the heath and is presented as mg CO2 m−2 h−1. A positive NEE indicates that the ecosystem acts as a CO2 sink, and a negative value represents a CO2 source.
Statistical analyses were performed using the spss package 14.0 (SPSS Inc., Chicago, IL, USA). The overall effects of warming and litter addition on isoprene emissions and carbon assimilation were tested as repeated measurements with linear mixed models separately for each year. A measurement campaign, carried out after the litter addition in August 2006, and the campaign with removed open-top chambers were omitted from the repeated measurements analysis. In addition, linear mixed models were run separately for each campaign to evaluate treatment effects in single campaigns. The models included warming and litter addition (plus warming × litter interaction) as fixed factors and block as a random factor. Each sample plot was used as a unit of replication (n = 6). P values under 0.05 were considered statistically significant, and P values under 0.1 to express a close to significant tendency.
Principal component analysis (PCA) was performed to analyse the association of block, temperature, PAR, humidity and the abundance of different vascular plant species and isoprene emission separately for the two years. Species that were present in fewer than four plots were omitted from the PCA. After unit variance-scaling of the variables, models with two (for the data of 2007) and three (2006) principal components were extracted. Factors that appeared to be associated with the emissions in the PCA loading plot, and were significant covariates in the mixed models analysis, were included in the mixed model. The PCA was performed using Simca P 11.0 software (Umetrics Inc., Umeå, Sweden).
The effects of warming and litter addition on isoprene emissions
In the growing season of 2006, isoprene emissions averaged over all measurements (± SE) from the Subarctic heath were 33 ± 11, 47 ± 13, 23 ± 7 and 40 ± 10 µg m−2 h−1 under the control, warming, litter addition and combined treatments, respectively (Fig. 1). Experimental warming tended to increase the emissions (P = 0.056, linear mixed models). When the measurement campaigns were analysed separately, the effect of warming was found to be most pronounced on 18 June 2006, when the emissions were generally low, but two times higher from the warmed plots compared with the plots without open-top chambers (P = 0.094). Litter addition had no significant effects on isoprene emissions, not even directly after application of the treatment in August 2006 (P > 0.4; Fig. 1) Isoprene emissions varied from under the detection limit to 324 µg m−2 h−1 (1.3 nmol m−2 s−1). The average emissions remained under 200 µg m−2 h−1 until August despite a period of warm weather in June. The highest emissions were detected in late August.
In the growing season of 2007, the emissions were significantly increased by warming (P = 0.021, linear mixed models, Fig. 2). Litter addition had no statistically significant effect on the emissions (P > 0.8; Fig. 2). Isoprene emissions averaged over all measurements were 46 ± 10, 84 ± 21, 35 ± 10 and 67 ± 14 µg m−2 h−1 under control, warming, litter addition and combined treatments, respectively. When the campaigns were analysed separately, the emissions were found to be significantly increased by warming on 10–11 July 2007 (P = 0.034). Isoprene emissions varied from under the detection limit to 738 µg m−2 h−1 (3.2 nmol m−2 s−1). The emissions stayed below 1 µg m−2 h−1 in June and rose to close to 200 µg m−2 h−1 during July. A cool weather period in mid-July slightly decreased the emissions. The highest emissions were detected in early August.
One measurement campaign (17 July 2007) was carried out without the experimental warming (the open-top chambers were removed) to test for indirect effects of warming. The emissions were not significantly different among treatments in that campaign, although litter addition and the combined treatment resulted in slightly lower emissions than the control and warming treatments (Fig. 2).
All isoprene emissions presented and discussed in the present study are actual emissions. To allow comparison with earlier isoprene emission studies, emissions standardized to a temperature of 30°C and a PAR of 1000 µmol m−2 s−1 (Guenther et al. (1993) are presented in Table 2. The statistical analyses were also performed for these standardized emissions, and similar results were obtained as for the actual emissions.
Table 2. Mean (± SE, n = 6) standardized (to 30ºC and 1000 µmol m−2 s−1 photosynthetically active radiation; Guenther et al., 1993) isoprene emission in control, warming, litter addition and combined treatments in 2006 and 2007
W + L
After litter addition.
Open-top chambers removed during sampling.
Close to significant effect of warming (2006, P = 0.076; 4 July, P = 0.061, linear mixed model).
Significant effect of warming (2007, P = 0.031; 10–11 July, P = 0.015).
Relationships among plant species, environmental factors and isoprene emission
Principal component analysis was used to assess the relationships among vascular plant species abundances, environmental factors and isoprene emission. Variables located to the same direction in the PCA loading plot were considered associated with each other.
In the data of 2006, the first principal component (PC) explained 17% of the variance in the data and mainly described the division of the plots into wetter (mosses and herbs abundant) and drier (lichens, horsetails and evergreen shrubs abundant) microsites (data not shown). The second and third PCs (14 and 12% of the variance explained, respectively) mostly accounted for the variance in isoprene emission (Fig. 3a). The abundances of Tofieldia pusilla, Carex vaginata, Arctostaphylus alpinus and Salix myrsinites were slightly positively correlated with isoprene emission in the PCA loading plot. However, as the correlation was not strong and S. myrsinites was present in only a few plots, these species were not used as covariates in the statistical analyses for the 2006 data. According to the PCA loading plot, isoprene emission was in general associated to a greater extent with wet microsites with abundant moss cover than with drier microsites with lichens (Fig. 3a).
In the data of 2007, the first and second PCs (17 and 15% of variance explained, respectively) both accounted for the variance in isoprene emission and the difference between wetter and drier microsites (Fig. 3b). Isoprene emission was, as for 2006, associated with the abundance of C. vaginata and to a lesser extent with T. pusilla and Carex rupestris according to the PCA loading plot. Carex vaginata and T. pusilla also appeared to be significant as covariates in the mixed model (T. pusilla, P = 0.020; C. vaginata, P = 0.001) and were therefore included in the analyses. The abundance of lichens and evergreen shrubs did not correlate with isoprene emission according to the PCA loading plot (Fig. 3b).
Isoprene emission was positively correlated with temperature and light intensity in the 2006 and 2007 data sets according to the PCA loading plots (Fig. 3). Relative humidity was negatively associated with isoprene emission in PCA but appeared to be nonsignificant as a covariate in linear mixed models and was therefore excluded from the analysis. The location of the plots at the site (indicated as blocks) did not correlate with isoprene emission in the PCA loading plot.
In the growing season of 2006, the NEE was not affected by warming (P = 0.735, linear mixed models) or by litter addition (P = 0.926) (Fig. 4a). The mean NEEs before litter addition in 2006 were 211 ± 66, 262 ± 64, 279 ± 65 and 181 ± 74 mg CO2 m−2 h−1 under control, warming, litter addition and combined treatments, respectively. In the last measurement, performed immediately after litter addition, the NEE was significantly decreased by litter addition (P = 0.004) as a result of the combined effects of decomposition of the added litter and shading of vegetation.
In 2007, the NEE was significantly decreased by warming (P < 0.001, linear mixed models) and litter addition (P = 0.050) (Fig. 4b). The interaction warming × litter was not significant (P = 0.433). The mean NEEs in 2007 were 466 ± 67, 163 ± 83, 340 ± 76 and −113 ± 183 mg CO2 m−2 h−1 under control, warming, litter addition and combined treatments, respectively. When the measurement campaigns were analysed separately, warming was found to significantly decrease the NEE on 15 June and 12 July (P = 0.025 and P = 0.029). A close to significant tendency was detected on 8 July (P = 0.053). The effect of litter addition in decreasing the NEE was most pronounced on 12 July 2007 (P = 0.093).
Effects of warming and increasing litterfall
The experimental warming increased isoprene emissions from the Subarctic heath, as hypothesized. A close to significant tendency was detected in 2006, and significantly higher emissions were observed under warming in 2007, after eight growing seasons under the exposure. The effect of warming was not long term or indirect (i.e. a result of changed vegetation cover) but rather a direct response to increased temperature, which indicates a rapid physiological effect of current temperature on isoprene production (Lerdau & Gray, 2003). This was demonstrated by the similar isoprene emissions across the treatments in a measurement campaign carried out on plots directly after suspension of the long-term warming.
In contrast to our hypothesis, isoprene emission was not increased by litter addition, although it brought extra nutrients, especially phosphorus, to the soil and increased the greenness of the vegetation (indicated by the normalized differential vegetation index, NDVI) (Rinnan et al., 2007, 2008). This is in contrast with earlier studies on isoprene emissions under nutrient fertilization treatments in the laboratory. Increased isoprene emissions have been detected under enhanced nutrient availability in velvet bean (Mucuna sp.) (Harley et al., 1994), aspen (Populus sp.), white oak (Quercus alba) (Litvak et al., 1996) and English oak (Quercus robur) (Possell et al., 2004). However, the present study was carried out at the plant community level, which limits comparison with earlier studies. Laboratory incubation of the added litter, both dry and moist, showed that it does not release isoprene.
The heath ecosystem was a sink of carbon during the growing seasons 2006 and 2007. However, the net ecosystem CO2 assimilation was significantly reduced by warming and litter addition in 2007. The result is in accordance with earlier studies in the same region as our experiment (Christensen et al., 1997) and at the same experimental site in which ecosystem respiration was significantly increased under warming (A. Michelsen et al., unpublished). Furthermore, the net CO2 assimilation became negative under the combined treatment in the late season of 2007 as ecosystem CO2 fluxes were controlled more by ecosystem respiration than by photosynthesis. Occasions of negative NEE are not unusual in Subarctic areas with large soil carbon stocks which have accumulated during previous colder periods. For instance, at a similar nearby heath ecosystem, NEE was negative on four out of eight occasions during the growing season (Illeris et al., 2004).
The heath as an isoprene source
We showed here that the Subarctic heath is a significant source of isoprene. The average emissions were similar, 36 and 58 µg m−2 h−1, in the two growing seasons. The largest emissions were detected in both years in August, although periods of warm weather were experienced in June and July. This suggests that fully developed plants have the largest capacity for isoprene emission in the heath. A similar seasonal trend was previously observed in Eriophorum russeolum-dominated Subarctic peatland (Tiiva et al., 2007b) as well as in, for eample, Salix phylicifolia (Hakola et al., 1998), Populus tremuloides (Fuentes et al., 1999) and in several other broadleaved species (Lerdau & Gray, 2003).
The mean isoprene emissions from the Subarctic heath were comparable to emissions from a Subarctic peatland (Tiiva et al., 2007b). However, the emissions from the peatland increased 10-fold during an exceptionally warm period (Tiiva et al., 2007b) while less radical fluctuations were observed for the heath in the present study.
Our results show that Subarctic heaths are more important sources of isoprene than Subarctic forests. A low isoprene flux, 14 µg m−2 h−1, was observed by Rinne et al. (2000) from a Betula pubescens ssp. czerepanovii- and Picea abies ssp. obovata-dominated Subarctic forest with the gradient technique in mid-July 1996. Tarvainen et al. (2007) have estimated isoprene emission from northern boreal forests at 45 kg km−2 per growing season. According to our results, the total growing season isoprene emission from Subarctic heaths is substantially larger, 96 kg km−2 (with a growing season of 111 d and 2041 h of daylight; Pearsall & Newbould, 1957). Emissions from the total heath area in the Lake Torneträsk catchment area in Northern Sweden (1858 km2; 47% of the catchment area; Christensen et al., 2007) are 178 tonnes of isoprene during a growing season.
The mean flux of CO2 into the heath ecosystem was 234 and 220 mg m−2 h−1 in 2006 and 2007, respectively. Therefore, the loss of carbon as isoprene to the atmosphere was in general < 0.1% of the net ecosystem carbon assimilation in the heath. During periods of the highest isoprene emissions, however, the carbon loss reached 1% of the mean net ecosystem carbon assimilation rate. This amount of carbon is at the same level as previously detected from single species under nonstressed conditions (e.g. Monson & Fall, 1989; Sharkey et al., 1991; Pegoraro et al., 2004). However, direct comparison with single-species studies is impossible as carbon exchange in the heath also includes soil respiration, and not all plant species in the heath are isoprene emitters.
Isoprene emissions from the heath were associated with the abundance of C. vaginata and T. pusilla. Significant emissions of isoprene have previously been observed from peatlands with abundant Carex sp. (Janson & De Serves, 1998; Haapanala et al., 2006; Hellén et al., 2006) whereas there are no earlier observations of isoprene emissions from ecosystems with Tofieldia sp. or other species in the family Liliaceae.
Several moss species have been reported to emit isoprene in laboratory trials (Hanson et al., 1999; Tiiva et al., 2007b). In the present study the abundance of mosses was not distinctly associated with isoprene emission, although the emission in general correlated more with mosses than with lichens and other species in drier microsites. Some of the moss species present in the plots are strong isoprene emitters, such as Sphagnum sp. (Hanson et al., 1999), which would explain the positive correlation with the emission. However, the positive correlation between the emission and the vascular species C. vaginata and T. pusilla in both years indicated that these species contribute more than mosses to isoprene emission in the heath.
To conclude, we have shown that heaths are significant sources of isoprene in the Subarctic. The emissions are on the scale of those from Subarctic peatlands but smaller than those from forests in the same region. Under long-term field manipulation, the emissions were substantially increased by warming. Thus, the predicted climatic warming in the Arctic will increase isoprene emissions from the heath ecosystem. As Arctic ecosystems form a substantial store of carbon (Post et al., 1982), the increasing carbon loss as isoprene may have implications for terrestrial ecosystem carbon accumulation if periods of high emissions become more frequent as a result of the warming climate.
We thank Timo Oksanen for constructing the equipment, Juhani Tarhanen for assistance with gas chromatography–mass spectrometry analyses and Annika Kristoffersson for weather data. Facilities for writing were provided by the Measurement and Sensor Laboratory of the University of Oulu. The study was financially supported by the Emil Aaltonen Foundation, an EU ATANS grant, the European Commission (MC-RTN-CT-2003-504720, ‘ISONET’), the Danish Research Council for Nature and Universe, and Abisko Scientific Research Station.