Long-term warming effects on root morphology, root mass distribution, and microbial activity in two dry tundra plant communities in northern Sweden


Author for correspondence:
Robert G. Björk
Tel: +46 31 7864804
Fax: 46 31 7862560
Email: robert.bjork@dpes.gu.se


  • • Effects of warming on root morphology, root mass distribution and microbial activity were studied in organic and mineral soil layers in two alpine ecosystems over > 10 yr, using open-top chambers, in Swedish Lapland.
  • • Root mass was estimated using soil cores. Washed roots were scanned and sorted into four diameter classes, for which variables including root mass (g dry matter (g DM) m−2), root length density (RLD; cm cm−3 soil), specific root length (SRL; m g DM−1), specific root area (SRA; m2 kg DM−1), and number of root tips m−2 were determined. Nitrification (NEA) and denitrification enzyme activity (DEA) in the top 10 cm of soil were measured.
  • • Soil warming shifted the rooting zone towards the upper soil organic layer in both plant communities. In the dry heath, warming increased SRL and SRA of the finest roots in both soil layers, whereas the dry meadow was unaffected. Neither NEA nor DEA exhibited differences attributable to warming.
  • • Tundra plants may respond to climate change by altering their root morphology and mass while microbial activity may be unaffected. This suggests that carbon may be incorporated in tundra soils partly as a result of increases in the mass of the finer roots if temperatures rise.


The cold climate of tundra ecosystems restricts the availability of nitrogen (N) for microbial and plant growth (Billings & Mooney, 1968). For this reason, partly at least, tundra plants require relatively large root systems (Atkinson, 2000) and tend to have higher root:shoot ratios (6.60, on average) than plants in any other biome (for comparison, the corresponding ratio in boreal ecosystems is 0.32; Jackson et al., 1996). Arctic plants also have shallower root systems than plants in any other biomes; 98% of their total root biomass and 94% of their fine-root biomass (< 2 mm diameter; Jackson et al., 1997) is restricted to the upper 30 cm of the soil because of the permafrost. Thus, roots are extremely important components of arctic ecosystems, where below-ground competition is likely to be greater than above-ground competition (Harper et al., 1991).

However, the arctic climate is warming rapidly and considerable further changes are predicted (air temperatures are expected to rise 4–7°C during the next 100 yr; ACIA, 2005), which are likely to have profound effects on arctic ecosystems. In order to predict the scale and nature of the likely effects, knowledge gained from warming experiments, using open-top chambers (OTCs), in tundra ecosystems is highly valuable. Such experiments have already increased our understanding of above-ground plant growth, the phenological responses of species and ecosystems to simulated climate change, and the effects of temperature increases on carbon (C) balances in arctic ecosystems (Chapin & Shaver, 1985; Chapin et al., 1995; Chapin & Shaver, 1996; Arft et al., 1999; Welker et al., 2000, 2004; Walker et al., 2006). A few studies have investigated the effects of warming on root biomass and its distribution in tundra ecosystems. Notably, Sullivan & Welker (2005) observed a nonsignificant trend towards higher annual root production in the arctic cotton grass Eriophorum vaginatum after 2 yr of OTC warming, in accordance with data from a boreal forest, where warming has been found to result in increased root production and mortality (Majdi & Öhrvik, 2004). Sullivan & Welker (2005) also recorded higher growth rates, earlier maximum growth rates, and significant increases in live biomass within their OTCs during the first two-thirds of the growing season.

However, predicting the effects of climate change on ecosystems, including their root systems, is not straightforward because many variables affect them. For instance, in high arctic heath tundra in north-eastern Greenland, Rinnan et al. (2005) recorded increases in total root density following reductions in solar UV radiation, mainly as a result of increases in fine roots of < 1 mm diameter. Studies in coniferous forests (Ostonen et al., 1999; Majdi & Viebke, 2004) have also shown that additions of calcium, magnesium, phosphorus and potassium (CaMgPK), wood ash, and N can induce increases in root length density (RLD), which are likely to enhance the capture of immobile ions more than that of mobile ions, although when plants are in interspecific competition for a limited N source root proliferation is also important for capturing N (Hodge, 2004). Changes in fine-root (< 2 mm in diameter) biomass as a result of warming and elevated CO2 concentrations in other biomes have also been observed (reviewed by Pendall et al., 2004), but they are not consistent. Furthermore, although a few authors have investigated root morphology and distribution in arctic ecosystems (Callaghan et al., 1991; van Wijk et al., 2003; Sullivan & Welker, 2005), little attention has been paid to date to these variables in a climate change context. Very few investigations have considered the influence of soil temperature on root morphology (cf. Pregitzer et al., 2000), and the potential effects of anticipated climate changes on the interactions of roots with the soil biota have been neglected. Thus, as these interactions profoundly affect nutrient cycling processes (as discussed in the following paragraph) further information on the interactions among climate, root systems and soil biota is required in order to make accurate predictions regarding the potential effects of climatic changes.

For tundra ecosystems, which have very high below-ground C stores, temperature has large effects on rates of organic matter decomposition and nutrient mineralization. Thus, the predicted increases in air temperature during the next 100 yr are expected to have major effects on these variables (ACIA, 2005). The key interfaces for these processes are the fine roots, which are responsible for much of the water and nutrient uptake by plants and soil C inputs, as c. 15–25% of the C allocated to below-ground parts is exuded from the fine roots into the soil (Kuzyakov, 2002; Pendall et al., 2004). These exuded organic substances generally stimulate microbial growth, promote microbial-driven C turnover in the rhizosphere and thereby enhance rates of decomposition of soil organic matter (SOM; Kuzyakov, 2002). It has also been suggested that soil microbes in tundra are N-limited (Schimel & Weintraub, 2003). In a meta-analysis, experimental ecosystem warming was shown to increase the rates of net nitrogen (N) mineralization in the upper organic soil horizon by 46% on average (Rustad et al., 2001). However, the fate of the N is poorly understood, and the microorganisms in tundra soils may utilize N sources other than KCl-extractable N (Björk et al., 2007), which is often used to study net N mineralization.

The findings described above illustrate the complexity of the interactions between climate and microbially driven nutrient cycling processes in tundra ecosystems, the key roles that roots play in these processes, and the need to acquire a more profound understanding of the effects of potential climate changes on roots and their interactions with the microbial populations of these systems.

The study presented here is part of the International Tundra Experiment (ITEX) programme, in which OTCs are being used to assess the long-term effects of increased summer air temperatures on tundra vegetation and ecosystem processes. We report the effects of increased temperature on root morphology characteristics, root mass distribution (after 11 yr) and microbial activity (after 10 yr) in two dry tundra plant communities. Two main hypotheses were tested: firstly, that increased summer warming will increase the root growth of tundra vegetation, as it has been shown to enhance above-ground plant production in arctic and alpine tundra (Arft et al., 1999; Walker et al., 2006) and, secondly, that a temperature-mediated increase in N mineralization (Rustad et al., 2001) will affect the populations of nitrifying and denitrifying microbes.

Materials and Methods

Study site

The study was conducted at Latnjajaure Field Station (68°21′N, 18°30′E), located at 980 m above sea level (a.s.l.) in the Latnjavagge valley, a U-shaped glacial valley in the mid-alpine region, c. 15 km west of Abisko, in northern Sweden. The bedrock in the valley is composed of garnet mica schist with inclusions of dolomite on the west-facing slopes. The soil is a shallow Pergelic Cryorthent (Soil Survey Staff, 1994; Marion et al., 1997a) with a loamy-sand texture; additional soil properties are presented in Table 1. Since April 1992, a standardized ITEX climate station (see Molau & Mølgaard, 1996 for more details) has been collecting hourly mean air temperature data at Latnjajaure Field Station. A simple regression model for these data for the years 1993–2006 is shown in Fig. 1. The local climate is of typically low-arctic character, with a mean annual temperature of –2.0°C (1993–2006). The coldest month, February, has a mean temperature of –9.3°C, and the warmest month, July, has a mean temperature of +8.5°C. The mean annual total precipitation is 848 mm (1990–2006), of which about 210 mm falls during the growing season (approx. June–August).

Table 1.  Characteristics of the organic soils in the treatment plots in two dry tundra ecosystems at Latnjajaure, northern Sweden
 Dry heathDry meadow
  1. The standard error of the mean is given in parentheses (dry heath, n = 4; dry meadow, n = 5).

Depth of organic soil (cm)2.1 (0.5)4.5 (1.3)10.2 (2.1)10.6 (2.1)
pH (KCl)3.7 (0.1)3.6 (0.1)4.7 (0.1)4.8 (0.1)
Soil organic matter (%)14.3 (2.2)18.0 (5.3)54.2 (5.2)49.2 (4.9)
Carbon:nitrogen ratio18.0 (0.9)18.4 (1.7)19.8 (1.1)20.6 (2.3)
Figure 1.

Annual average air temperature (regression; y = 0.1245x –250.8) and thawing degree days (TDD) > 0°C (bar plot) from May to September at Latnjajaure, northern Sweden, during 1993–2006.

The vegetation resembles that of the Low Arctic, but is unusually species-rich and highly diverse (Molau et al., 2003; Lindblad, 2007). The two plant communities studied, dry heath and dry meadow, both have a restricted water regime and are often snow-free during the winter as a result of the wind blowing snow into topographic depressions, thus forming snowbeds. The dry heath community is found on acidic glacial moraine ridges and flats, occupying about a quarter of the valley floor (described in detail by Molau & Alatalo, 1998; Lindblad, 2007). Usually there are some active frost processes, such as small mud-boils and local solifluction, but most of these processes have ceased in recent decades. The vegetation cover is sparse and patchy and is dominated by cryptogams. Among the mosses several species of Dicranaceae are dominant and the lichens are dominated by Stereocaulon alpinum and Ochrolechia frigida. The vascular plant cover is characterized by Diapensia lapponica, Cassiope tetragona, Loiseleuria procumbens, Vaccinium vitis-idaea, Salix herbacea, Empetrum hermaphroditum, Calamagrostis lapponica and scattered, extremely low-growing clones of Betula nana. The dry meadow community – also known as Dryas meadow – is associated with warm calcareous sites and has a denser plant cover (described in detail by Lindblad, 2007). The vegetation is more diverse than in the dry heath and is dominated by Dryas octopetala, Vaccinium vitis-idaea, Festuca ovina, Bistorta vivipara, Carex vaginata, Thalictrum alpinum, Carex rupestris, Astragalus alpinus, and Salix reticulata. Cryptogams are not as noticeable as in the dry heath, but among the most common bryophytes are Hylocomium splendens and Tomentypnum nites. Few lichens are adapted to conditions in the dry meadow, although Flavocetraria cuculata and Flavocetraria nivalis are relatively common.

Open-top chambers experiment

The effect of climatic warming was simulated passively using OTCs, in accordance with the ITEX standard (Molau & Mølgaard, 1996). OTCs are considered to be the most suitable devices for warming experiments in most tundra environments (Marion et al., 1997b) and they have been found to provide reasonable analogues of regional climate change (Hollister & Webber, 2000). The OTCs used at Latnjajaure are hexagonal chambers with a top opening of 0.6 m in diameter, which gives a side-to-side basal diameter of 104 cm, made of polycarbonate with high transmittance of solar radiation. The sides are inwardly inclined by 60° and they increase the internal air temperature by as much as 2–3°, but still allow precipitation to enter (Molau & Mølgaard, 1996; Marion et al., 1997b). The OTCs increase the soil surface temperature by approximately 1.5°C (Marion et al., 1997b). In late May 1993, four OTCs were deployed in the dry heath and five in the dry meadow, according to a randomized stratified design (Welker et al., 1997; Molau & Alatalo, 1998). Control samples were taken at points 2 m away from each of the OTCs to represent ambient conditions (n = 4 for the dry heath and n = 5 for the dry meadow).

Soil microbial activity sampling

Two soil samples, from the top 10 cm, were collected from each OTC, a few centimetres from the sides in order to avoid any edge effects, and two from control sampling points 2 m away from each of them, with a soil sample cylinder (diameter 70 mm), on 31 August 2003. The samples were stored at +4°C for 1 wk before they were sieved to a < 2 mm fraction. Combined soil samples were then generated from the pairs of soil samples from each OTC (n = 4 for the dry heath and n = 5 for the dry meadow) and the corresponding controls. These were stored in plastic bags at +4°C for a maximum of 1 month, awaiting further analysis. All samples were then analysed for nitrification enzyme activity (NEA; Lensi et al., 1985, 1986), denitrification enzyme activity (DEA; Smith & Tiedje, 1979; Tiedje, 1994), total C, total N, water and SOM contents, and pH. NEA and DEA measurements provide measures of potential nitrification and denitrification rates, but correlate better with extractable inorganic N concentrations than, for instance, net nitrification (Björk et al., 2007). Furthermore, these variables, in particular NEA, are much more stable than extractable N concentrations, which is highly advantageous when working in remote environments such as the Latnjajaure catchment, where collecting and processing samples quickly is logistically difficult, if not impossible.

Soil characteristics

Soil pH was measured in suspensions of 5.0 g of soil (wet weight) and 30 ml of 1 m KCl. Soil water content was determined as weight loss as a result of drying at 70°C for 48 h. To determine SOM content, the dried samples were ignited in a muffle furnace at 550°C for 24 h and SOM was calculated as the loss on ignition. Soil samples used for total C and N analysis were dried at 70°C until there was no further change in weight. The samples were then ground to a fine powder and analysed by combustion in an elemental analyser (Model EA 1108 CHNS-O; Fisons Instuments S.p.A., Rodano, Italy).

Nitrification enzyme activity

A two-step incubation technique was used to determine NEA, as described by Lensi et al. (1985, 1986) for analysing low nitrification rates in acid soils. First, 25 ml of a solution containing 0.2 mm KH2PO4, 0.2 mm CaCl2, 0.2 mm MgSO4 and 2.5 mm (NH4)2SO4 described by De Boer et al. (1988) with the addition of 20 mm HEPES buffer (as recommended by Jiang & Bakken, 1999), adjusted to pH 8.0 by adding 5% NaCO3, was mixed with 3 g of fresh soil in a 50-ml test tube. The tube was then sealed, wrapped in aluminium foil and incubated for 24 h, in darkness, at room temperature on a rotary shaker rotating at 20 r.p.m. After 0, 3, 6, 12, 18 and 24 h, 1-ml subsamples of the soil slurry were transferred to 22-ml headspace bottles, which were sealed with Parafilm M (Pechiney Plastic Packaging, Chicago, IL, USA) and stored at –20°C.

Next, inline image was reduced to N2O, following the procedure of Lensi et al. (1985), by adding a modified denitrifying bacterium, Pseudomonas chlororaphis ATCC 43928, which lacks the enzyme to reduce N2O to N2. A C source (succinic acid) was also added to avoid the denitrifiers becoming C limited. The samples were defrosted and sealed with crimp caps and rubber septa before they were evacuated with N2. To each sample, 1 ml of succinic acid and 1 ml of bacterial suspension were added (Bäckman & Kasimir Klemedtsson, 2003). To inhibit the reduction of N2O by naturally occurring denitrifiers in the sample, 2.5 ml of N2 was replaced with acetylene. The samples were incubated in darkness, at room temperature for 24 h and then analysed by gas chromatography, as described by Klemedtsson et al. (1997).

Denitrification enzyme activity

The following anaerobic incubation technique, based on acetylene inhibition of N2O reductase, was employed to determine DEA (Klemedtsson et al., 1977; Smith & Tiedje, 1979). First, 20 g of fresh soil and 30 ml of nutrient solution (1 mm potassium nitrate, 0.5 mm glucose, 0.5 mm sodium acetate and 0.5 mm succinic acid) were placed in a 250-ml serum bottle. The bottle was then sealed with a rubber septum, evacuated and flushed with N2, and then filled with a 90 : 10 (volume/volume) mixture of N2 and acetylene. The sample was shaken continuously and gas samples were withdrawn after 10, 30, 50 and 70 min, which were then analysed by gas chromatography, again as described by Klemedtsson et al. (1997). This procedure provides an estimate of the maximum concentration of functional denitrifying enzymes in the soil (Tiedje, 1994).

Root morphology characteristics

One soil core per treatment (n = 4 for the dry heath and n = 5 for the dry meadow) was collected on 19 August 2004 using a cylindrical soil corer (diameter 72 mm). The soil samples were separated into organic and mineral soil layers and placed in plastic bags. However, some of the samples of the dry meadow were lacking the mineral soil layer (n = 2 for ambient mineral soil and n = 3 for warmed mineral soil). Samples were stored cold (+4°C) for 5 months and then frozen at –18°C awaiting analysis, which commenced in April 2005, when they were thawed, and roots were extracted from the soil particles by hand under a stream of water passing through a sequential sieve block consisting of four sieves with progressively smaller mesh sizes (ranging from 5.6 mm to 300 µm). Hard blocks were soaked overnight to facilitate separation. Roots were further separated from the rhizosphere soil with a small brush and tweezers in water. No distinction between live and dead roots was made. Finally, the roots were sorted into four size classes according to their diameter (< 0.5, 0.5–1, 1–2 and 2–5 mm; no roots exceeded 5 mm in diameter) using a digital calliper preset to the upper boundary and dried for at least 48 h at 70°C, and then the dry weight of each fraction from each sample was measured (cf. Majdi, 2001). The root fractions were then soaked overnight, placed in a water bath and scanned, using a STD1600+ scanner (Regent Instruments, Quebec, Canada). The images obtained were then analysed using WinRHIZO regular (Regent Instruments) to determine the root lengths, root areas and numbers of root tips (RTs) for each root fraction. Root mass (g dry matter (g DM) m−2), root density (g DM dm−3 soil), RLD (cm cm−3 soil), specific root length (SRL; m g DM−1), specific root area (SRA; m2 kg DM−1), number of RTs (NRT; 104 tips m−2), RT density (103 tips dm−3 soil), specific RT density (103 tips g DM−1), and RT per root length (tips cm−1) were calculated from the acquired data. The way in which we processed our root data in dry condition is not the most accurate method and thereby the root lengths and areas and the variables derived from these values are probably underestimated by a factor of 1.3 (R. G. Björk, unpublished data).

Statistical analysis

The root mass and root morphology characteristics as response variables were examined. The experiment was carried out using a partially hierarchical layout, and a corresponding three-factorial linear model was used in the statistical analysis. For each plant community (dry heath and dry meadow) a number of plots were used, eight for dry heath and 10 for dry meadow, and these plots were randomized in a pair-wise manner to the two treatments (ambient and warmed), with an equal number of plots for each treatment; that is, four and five plots, respectively, for each combination of one treatment and one plant community. In each plot, two layers (mineral and organic soil) were studied. This means that in the linear model two factors, plant community and treatment, are crossed and tested, as well as their interaction, against the sum-of-squares attributable to the variation between plots within plant community and treatment, as the error term. The third factor, layers, also with only two levels, and the interactions between this factor and treatment, as well as plant community, are consequently tested against the overall error term. Statistical analyses were carried out using the general linear model (GLM) combined with the least significant difference test (P < 0.05) as implemented in sas version 9.1 (SAS Institute, 2002), which is preferable when the data set is unbalanced (missing data in mineral soil samples in the dry meadow; n = 2 for ambient mineral soil and n = 3 for warmed mineral soil). Because, for all response variables, the three-factor interaction was found to be quite small and nonsignificant, the interaction between treatment and soil layers reveals the most interesting results, which, hence, can be regarded as independent of plant community.

To further determine the effect of simulated climatic warming on microbial activity, a one-way ANOVA was applied to each plant community, with the treatment as a fixed factor (Sokal & Rohlf, 1994). In order to investigate which variable had the greatest effect on microbial activity, a stepwise multiple regression was conducted, with NEA or DEA as the dependent variable, and the water content and C:N ratio as independent variables (spss 14.0, SPSS Inc., Chicago, IL, USA). The independent variables exhibited no collinearity, but correlated best with the omitted variables (Sokal & Rohlf, 1994). Before the GLM, one-way anova, and multiple regression analyses, all data were log-transformed to ensure normality, as indicated by homogeneity of variances; zero skewness was achieved by adding a constant (for further details, see Økland et al., 2001). The NEA in the dry heath could not be log-transformed because of the high number of zeros and was therefore analysed using Mann–Whitney U-tests. To examine the change in annual mean air temperature over the years of the warming experiment, untransformed data were analysed using simple linear regression, with year and annual mean air temperature as independent and dependent variables, respectively. Significant differences refer to P < 0.05 if not otherwise stated.


Root mass distribution under ambient conditions

Under ambient conditions, the total root mass in the dry meadow (533.6 g DM m−2) was nearly twice as high as in the dry heath (290.0 g DM m−2), largely because the mass in the organic layer of roots < 0.5 mm in diameter was nearly fourfold higher in the former (Table 2). However, there were no significant differences in either root fractions or total root mass between the organic soil and the mineral soil in either the dry meadow or the dry heath under ambient conditions (Table 2).

Table 2.  The means and interactions of soil horizons and treatment within each plant community for root mass, root density, and root length density (RLD) of different root diameter sizes in two dry tundra ecosystems at Latnjajaure, northern Sweden
Root fractionDry heathDry meadow
  1. Different superscript letters between soil horizons within each plant community and treatment show a significant difference (LSD; P < 0.05) between LS mean values. A significant interaction (denoted by *) means that the difference between the mineral and organic soils differed significantly between the ambient and warmed treatments. Dry heath, n = 4; dry meadow (organic soil), n = 5; dry meadow (ambient mineral soil), n = 2; dry meadow (warmed mineral soil), n = 3.

Root mass (g DM m−2)
< 0.5 mm61.9a62.2a–0.325.6a133.7b–108.1*164.0a247.0a–83.016.7a277.8b–261.1*
0.5–1 mm34.3a44.1a–9.88.6a61.3b–52.7*69.9a99.2a–29.32.7a84.7b–82.0*
1–2 mm25.1a21.2a3.91.5a24.6a–23.169.3a63.8a5.515.0a77.4b–62.4*
2–5 mm5.5a35.7a–30.20a5.4a–5.436.1a36.1a029.5a32.0a–2.5
Root density (g DM dm−3 soil)
< 0.5 mm1.12a2.61a–1.491.13a3.14b–2.01*2.47a2.46a0.010.05a3.22 b–3.18*
0.5–1 mm0.55a1.97b–1.420.37a1.59b–1.22*1.36a0.93a0.430.28a1.08a–0.80
1–2 mm0.38a0.95a–0.570.06a0.47a–0.411.11a0.50a0.610.06a0.89a–0.83
2–5 mm0.08a1.19a–1.110a0.14a–0.140.24a0.24a00.68a0.27a0.41
RLD (cm cm−3 soil)
< 0.5 mm9.5a21.4a–11.914.3a36.8a–22.530.2a31.5a–1.30.3a44.2b–43.9*
0.5–1 mm0.3a 1.4b–1.10.2a0.8a0.60.9a0.5a0.40.2a0.8b–0.6*

In addition, there were no significant differences in the densities of any root fraction, or total root density, between the two plant communities in either soil layer. However, in the dry heath the total root density was significantly higher in the organic soil than in the mineral soil (6.7 and 2.1 g DM dm−3 soil, respectively) primarily as a result of a significant difference between the soil layers in the 0.5–1 mm fraction (Table 2). In the dry meadow, the root density did not significantly differ between the organic soil and the mineral soil under ambient conditions (Table 2).

Warming effects on root mass distribution

Following the OTC treatment, the differences between the plant communities were greater than under ambient conditions. The total root mass was more than twice as high in the dry meadow (580.0 g DM m−2) as in the dry heath (260.7 g DM m−2). Under warmed conditions in the dry heath (where the organic layer tended to be twice as thick as in the ambient controls, although this was not a significant difference; Table 1), there were nonsignificant increases and decreases in total root mass in the organic soil (from 163.2 to 225.1 g DM m−2) and mineral soil (from 126.8 to 35.6 g DM m−2), respectively, compared with the corresponding masses observed in ambient conditions. However, these nonsignificant changes led to a significant change in the distribution of the root mass between the soil layers, total root mass being six times higher in the organic layer than in the mineral soil, mostly as a result of changes in the two finest root fractions, < 0.5 and 0.5–1 mm (Table 2). This treatment effect was almost certainly attributable to the depth of the organic soil being greater in the warmed dry heath plots (Table 1). In the dry meadow the nonsignificant increase in total root mass (mainly in the < 0.5 mm root fraction) was not as pronounced as in the dry heath, but nevertheless it led to the total root mass being significantly (approx. 4.5 times) higher in the organic soil than in the mineral soil (472.0 and 108.0 g DM m−2, respectively). The proportions of the two finest root fractions varied little between treatments and horizons in the dry meadow; ranging from 55 to 61% for < 0.5 mm roots and from 18 to 25% for 0.5–1 mm roots. However, the changes in the finest root fraction in the dry heath increased the proportion of < 0.5 mm roots in the organic soil from 38% of the total mass under ambient conditions to 59% in the OTCs. Similarly, in the mineral soil, there was an increase from 48 to 67%. The 0.5–1 mm root fraction exhibited no significant between-treatment differences in the dry heath and its proportion ranged from 20 to 27%.

The total root density stayed within the same range as under the ambient conditions in both plant communities (Table 2). However, nonsignificant changes in the soil layers in the dry meadow, as observed in mass, resulted in the total root density being significantly (2.5 times) higher in the organic soil than in the mineral soil (5.5 and 2.1 g DM dm−3 soil, respectively), primarily as a result of a significant difference between the soil layers in the finest fraction (< 0.5 mm), suggesting that the whole rooting zone of fine roots had shifted up, towards the surface of the soil into the organic layer (Table 2).

Warming effects on root morphology

None of the root fractions exhibited any significant changes in RLD, NRT, RT per root length (ranging from 0.8 to 2.5 and from 6.7 to 11.0 for the 0.5–1 and < 0.5 mm fractions, respectively), or RT density with respect to the OTC treatment (Tables 2, 3). However, in the dry heath the specific RT density of the finest fraction (< 0.5 mm) was higher in the mineral soil in the OTCs (Table 3), and the SRL and SRA of this fraction were higher in both the organic soil and mineral soil in the OTCs than in the controls (Fig. 2). In contrast to the dry heath, the only variables that significantly differed in the dry meadow between ambient and warmed conditions were the NRT of the < 0.5 mm root fraction (Table 3). In addition to these within-community differences, several warming-related changes in between-community differences in root morphology were detected. Notably, the SRL and SRA of the < 0.5 mm root fraction were 1.5 times higher in the dry meadow, in both the organic and mineral soil, than in the dry heath under ambient conditions, while under warmed conditions there were no significant differences between the plant communities in these respects, except that the SRA in the organic layer was still significantly (1.2 times) higher in the dry meadow than in the dry heath (Fig. 2). In addition, there were > 2.5 times more root tips of the finest root fraction (< 0.5 mm) in the organic layer of the dry meadow than in that of the dry heath under both conditions. Otherwise, there were no visible differences between the plant communities in root characteristics.

Table 3.  Mean number of root tips (NRT), root tip (RT) density and specific root tip density and their interactions with soil horizons and treatment within each plant community in two dry tundra ecosystems at Latnjajaure, northern Sweden
Root fractionDry heathDry meadow
Mineral soilOrganic soilDifferenceMineral soilOrganic soilDifferenceMineral soilOrganic soilDifferenceMineral soilOrganic soilDifference
  1. Different superscript letters between soil horizons within each plant community and treatment show a significant difference (LSD; P < 0.05) between LS mean values. A significant interaction (denoted by *) means that the difference between the mineral and organic soils differed significantly between the ambient and warmed treatments. Dry heath, n = 4; dry meadow (organic soil), n = 5; dry meadow (ambient mineral soil), n = 2; dry meadow (warmed mineral soil), n = 3.

NRT (104 tips m−2)
< 0.5 mm440.3a403.9a36.4255.0a1011.8b–756.8*1891.1a2504.7a–613.6303.6a2515.1b–2211.5*
0.5–1 mm2.5a2.4a0.10.9a3.4a–2.55.0a7.4a–2.40.7a8.9b–8.2*
RT density (103 tips dm−3 soil)
< 0.5 mm86.0a180.7a–94.7124.9a250.9a–126.0255.9a233.1a22.826.4a308.1b–281.7*
0.5–1 mm0.4a1.3a–0.90.4a0.8a–0.40.8a0.7a0.10.1a1.1b–1.0*
Specific RT density (103 tips g DM−1)
< 0.5 mm74.6 a74.1a0.5117.3a77.0a40.3102.3a91.2a11.1132.5a89.2a43.3
0.5–1 mm1.8a0.7a1.10.8a0.6a0.20.7a1.0a–0.31.5a1.3a0.2
Figure 2.

Mean specific root length (SRL) and specific root area (SRA) for the root diameter classes 0.5–1 and < 0.5 mm in dry heath and dry meadow at Latnjajaure, northern Sweden. Bars indicate ± SE. Different superscript letters between treatment and soil horizons within each plant community show a significant difference (LSD; P < 0.05) between LS mean values. Dry heath, n = 4; dry meadow (organic soil), n = 5; dry meadow (ambient mineral soil), n = 2; dry meadow (warmed mineral soil), n = 3.

Soil microbial activity

Neither NEA nor DEA exhibited any significant differences in the top 10 cm of soil between the ambient and warmed plots (Fig. 3). However, DEA was more than 25 times higher in the dry meadow than in the dry heath, whereas NEA was very similar in the two plant communities. The NEA stepwise multiple regression model (r2 = 0.489, P = 0.001), indicated that NEA was positively associated with soil water content (β = 0.722, P = 0.001) but not significantly related to either RLD or the C:N ratio. In the DEA stepwise multiple regression model (r2 = 0.707, P < 0.001), DEA showed a positive relationship with both soil water content (β = 0.799, P < 0.001) and RLD (β = 0.307, P = 0.034), but the C:N ratio had no effect at all on DEA.

Figure 3.

Nitrification enzyme activity (NEA; ng N g−1 organic matter (OM) h−1) and denitrification enzyme activity (DEA; ng N g−1 OM h−1) in dry heath and dry meadow at Latnjajaure, northern Sweden. Bars indicate ± SE (n = 4−5). Mean values are significantly (P < 0.05) different among treatments if superscript letters differ within a plant community. Black bars, ambient soil; grey bars, warmed soil.


The annual average air temperature data collected at Latnjajaure Field Station during this study period (1993–2006) clearly indicate that this part of the arctic region (at least) is rapidly warming, the regression model showing a significant increase of 0.12°C (± 0.02°C) per year (Fig. 1). In Alaska and Siberia, temperatures have increased by 2–3°C in the past 50 yr, while data for the Arctic (north of 60°N) as a whole show a significant warming trend of 0.09 degrees per decade (ACIA, 2005). In the northernmost part of Sweden the temperature increase has been more rapid in the last 15 yr (1.0°C over the period 1991–2006), and even more rapid during the winter (2.0–2.2°C over the period 1991–2006; SMHI, 2006). Increases in the length of the growing season have also been observed in Fennoscandia (Carter, 1998), which are supported by phenological data from Latnjajaure (Molau et al., 2005). The increase, 1.2°C per decade, seen in our temperature record is more rapid than previously observed in Arctic regions, and higher than predicted for the sub-Arctic region of Sweden (ACIA, 2005).

We observed several morphological responses of the finest roots in the dry heath after 11 yr of treatment with OTCs. However, because of methodological inaccuracies in our estimates of root length and diameter, the variables derived from these estimates (i.e. SRL and RLD) are underestimated. These estimates are underestimated by a factor of 1.3 (R. G. Björk, unpublished data) and should be considered with caution as root length and diameter were measured in a dry–rewetted state. In the dry heath, the SRL and SRA of roots thinner than 0.5 mm had increased in both the organic and mineral soils, whereas the specific RT density had only increased in the mineral soil. In previous studies of tundra ecosystems, most attempts to sort roots into size classes have used the traditional fine root (< 1 mm) and coarse root (> 1 mm) categories, although Pregitzer et al. (2002) have shown that fine roots with different branching orders have different functions, that is, turnover and respiration. It has also been shown that the position of a root within the branched hierarchy of the root system, the root order, has a major impact on the SRL and N status of the root (Pregitzer et al., 1998; Wells & Eissenstat, 2001; Pregitzer et al., 2002; Guo et al., 2004). The closer a root is located to the distal end of the root system, the higher its SRL and the lower its C:N ratio (Pregitzer et al., 2002). However, the root architecture of the tundra plants in the communities we studied is different from that of tree species reported in the literature (Pregitzer et al., 1998; Wells & Eissenstat, 2001; Pregitzer et al., 2002; Guo et al., 2004) as their branching density is low. In a review, Pregitzer et al. (2000) found considerable evidence to suggest that the rate of root length extension in tree fine roots is positively related to soil temperature. Moreover, a fertilization experiment in tussock tundra found increases in both biomass and root lengths in fertilized plots (van Wijk et al., 2003), although no distinction between root size classes was made. In a study of the four most abundant species in tussock tundra, Chapin & Shaver (1996) found no temperature effects in below-ground biomass, but following a combined temperature and fertilization treatment the below-ground biomass was found to decrease in Eriophorum vaginatum. However, Majdi & Viebke (2004) recorded increases in RLD, SRL and the number of mycorrhizal root tips for coniferous tree fine roots in response to additions of CaMgPK and wood ash, which were more pronounced in the 0–1 mm root diameter class than in the larger classes. The SRL reflects the potential for plants to exploit the soil, and by altering their SRL (i.e. altering their within-root system allocation strategy; Atkinson, 2000) they can change their potential for nutrient and water uptake relative to the cost of resources used for the construction and maintenance of root tissue (Eissenstat, 1992; Atkinson, 2000). Changes in SRA have also been suggested to provide a way for plants to respond to changes in environmental conditions (Lõhmus et al., 1989) and to be connected with the physiological activity of fine roots (Ostonen et al., 1999). Although we did not determine the root order, our results are consistent with the conclusions of Pregitzer et al. (2000, 2002) as we found a morphological response – increases in SRL and SRA – in the finest fraction (< 0.5 mm) as a result of the OTC treatment. However, there was no general response in root morphology, as the dry meadow did not exhibit any change as a result of the OTC treatment. The optimal temperature for root growth seems to vary widely among different taxa (Pregitzer et al., 2000), which may explain (partly at least) the observed between-community differences in root responses. Another possible explanation, or contributory factor, is that the plants in the dry meadow already had high SRL and SRA at the start of the experiment (see ambient data), and as suggested by Hill et al. (2006) species with very fine root systems may have less capacity to adjust their SRL. As our results demonstrate, responses to global warming are likely to differ among the different tundra plant communities, highlighting the need for multicommunity approaches when simulating climate change in arctic ecosystems.

Overall plant productivity and above-ground biomass appear to increase in long-term warming experiments (Chapin & Shaver, 1985, 1996; Chapin et al., 1995; Walker et al., 2006). Furthermore, species richness decreases, as a result of loss of less abundant species, while the initially most abundant species often remains the most abundant species after long-term warming treatments (Chapin et al., 1995; Walker et al., 2006). In general, warming increases the cover and height of deciduous shrubs and graminoids, and decreases both species diversity and evenness, thus causing shifts in plant community composition (Walker et al., 2006). In the dry heath at Latnjajaure there was an increase in vascular plant cover even after 2 yr of increased temperature, although the diversity was unaffected (Molau & Alatalo, 1998). After 5 yr of warming Betula nana was the only species that increased in biomass at Latnjajure, although the plant community composition was still unaffected (Jägerbrand, 2005). This implies that the shift in SRL observed in our study does not reflect a change in species composition, but rather a change in root morphology per se.

Overall, root mass, SRL and SRA are higher in the dry meadow than in the dry heath. However, the root:shoot ratio in the dry meadow (0.84; above-ground biomass data from Lindblad, 2007) is lower than that in the dry heath (1.47; above-ground biomass data from U. Molau et al., unpublished). These root:shoot ratios are much lower than those suggested by Jackson et al. (1996) for the tundra biome (6.60). However, it has been suggested that slow-growing species in nutrient-poor habitats often require a large root system, and the root systems of fast-growing species in nutrient-rich habitats are considered more able to respond to changes in nutrient concentrations (Grime, 1979; Chapin, 1980). Our data support this hypothesis, as we found the proportion of biomass allocated to the root system to be lower in the dry meadow, and the meadows at Latnjajaure have been shown to have higher N turnover (Björk et al., 2007) and mineral nutrient contents (Lindblad, 2007) than the heaths. Thus, the dry meadow has twice as high a root mass as the dry heath, although there are no differences in root density between the two plant communities.

Our results further show a shift in mass distribution as the whole rooting zone has moved towards the organic layer as a consequence of warming. However, specific features of that response differ between the two plant communities. In the dry heath, the proportion of < 0.5 mm roots increased from c. 40% of total root weight in ambient plots to c. 60% in the warmed plots (Table 2). In addition, there was a tendency for the organic layer to double in depth as a consequence of the OTC treatment, while the total root mass in the organic layer increased from being 1.3 times higher than the total root mass in the mineral layer (a nonsignificant difference) in ambient conditions to being more than 6 times higher in warmed conditions. In contrast, no such change was observed in the dry meadow, although the root density changed significantly; densities were very similar in the organic and mineral layers in ambient conditions, but 2.5 times higher in the organic layer in warmed conditions. The reasons for these changes in mass distribution may be related to changes in nutrient supply, either in nutrient availability or in rates of diffusion of nutrients to roots. Rates of N mineralization are expected to increase as soils warm (Rustad et al., 2001), but nutrient uptake rates per unit root length also increase with increasing temperatures (BassiriRad, 2000). Nevertheless, as the availability of N in tundra ecosystems is limited, ammonium (inline image) is probably the critical N pool associated with plant–microbe competition (Jackson et al., 1989) and both heterotrophs and plants out-compete nitrifiers for inline image (Johnson & Edwards, 1979; Zak et al., 1990). This may explain the lack of response in NEA to the warming observed in our study, as an increase in the rate of N mineralization (not measured in our study) may not lead to increased inline image availability for nitrifiers, contrary to our hypothesis. However, this does not necessarily exclude the possibility that N mineralization may increase, because both heterotrophic microbes and plants out-compete nitrifiers for inline image. At Latnjajaure there is no longer any permafrost (Beylich et al., 2003), which could restrict root growth, so roots can grow well below 30 cm, the depth generally accepted to be the lower limit for root penetration in the tundra (Jackson et al., 1997). Nevertheless, the root distribution in the dry heath mineral soil suggests that not only may root growth in tundra soil be restricted by permafrost, it may also be affected by the depth of the organic layer, which may also increase in a future warmer climate.

Our results also show that long-term temperature increases have minor effects on NEA and DEA in dry tundra plant communities, contrary to our hypothesis; other variables such as soil water content, SOM, and pH seem to be more important. However, DEA also appears to be positively affected by the presence of roots. Denitrifiers, in contrast to nitrifiers, are heterotrophs and denitrification is never limited as a result of there being insufficient denitrifiers present, because they usually constitute a reasonably large fraction of the bacterial population in soils ( Myrold, 1999). Root tissue quality may play a key role in the responses of microbial communities to a changing environment (Zak et al., 2000), and as the C:N ratio decreases with increasing SRL (Pregitzer et al., 2002) the availability of rhizodeposits is likely to increase. Thus, the positive relationship between root variables and DEA suggests that the denitrifiers may use C derived from rhizodeposition.

Another interesting finding from our study is the tendency for the depth of the organic soil horizon in warmed plots in the dry heath to increase. These plots also exhibit increased SRL and SRA. In contrast to predictions by models indicating that tundra ecosystems will predominantly be C sources if temperatures increase (ACIA, 2005), this may suggest that incorporation of C into this ecosystem is occurring, hence the positive relationship between roots and DEA. In a warming experiment in a boreal forest, Majdi & Öhrvik (2004) found that warming led to increased root production and mortality. Moreover, Reich et al. (1998) found root respiration and N uptake to be highly correlated with SRL among nine boreal tree species, and Joslin et al. (2006) showed that fine roots (< 0.5 mm) turned over more rapidly than 0.5–2 mm roots. Because SRL increased in the dry heath in our study as a consequence of warming, it is likely that the root turnover also increased. This may have contributed to the increased depth of the organic horizon. Furthermore, Kuzyakov (2002) identified three scenarios in which rhizosphere priming by root exudates and/or nutrient uptake by roots may adversely affect SOM decomposition. Firstly, roots may take up the exudates in sufficient quantities to reduce the amounts of available C and thus microbial activity, leading to reduced rates of SOM decomposition. Secondly, when two substrates with very different availabilities (e.g. exudates and SOM) are present at a single location, micro-organisms generally use the readily available source (the exudates) first, so the rate of decomposition of the less readily available one (SOM) may decline, at least initially. Thirdly, as microbial growth in the rhizosphere is limited by N availability, root uptake of N may increase the competition for nutrients and decrease microbial growth and metabolism, thereby depressing SOM decomposition. The third scenario, in combination with increased root turnover, may explain our results, as it has been suggested that certain microbes living in tundra soils are N limited (Schimel & Weintraub, 2003) and that plants out-compete them for N as SRL increases. Consequently, microbial growth and metabolism are inhibited, thereby depressing SOM decomposition and as a result the organic layer increases in depth.

This study demonstrates that there were considerable differences in responses to OTC treatment between the two investigated dry tundra plant communities. The proportion of the finest roots (< 0.5 mm) as well as their SRL and SRA increased in the dry heath, whereas in the dry meadow the whole rooting zone only shifted towards the surface soil into the organic layer as a result of warming. Furthermore, our study suggests that warming may lead to the incorporation of C in tundra soils, in contrast to climate change predictions. However, our results give indications concerning the warming effect on root standing crops at a single sampling occasion, and do not show changes in root dynamics, such as changes in production and mortality during the growing season. To fully understand the effects of global warming on below-ground production in tundra ecosystems, it is therefore also important to perform warming experiments on root dynamics using minirhizotrons.


The authors thank Anja Ödman, Josefine Norman and Olga Sandberg for their assistance in the field and laboratory. We also acknowledge the Abisko Scientific Research Station and its staff for their help and hospitality. Furthermore, we are grateful to two anonymous reviewers for valuable referee comments and to Alastair H. Fitter for valuable editorial comments. This research was supported by the Adlerbertska Research Foundation, the Foundation of Helge Ax:son Johnson, the Wilhelm & Martina Lundgrens Science Foundation Fund, the Royal Society of Arts and Sciences in Göteborg, the P.-A. Larssons Fund, the Royal Swedish Academy of Sciences (grant to RGB), the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (grant nos 22.0/2004-0449 and 21.0/2004-0518; to LK), the Swedish Research Council (grant no. 621-2003-2730 to LK), and the Kempe Foundation (grant to UM). The contributions of HM and LK were made within the ‘Land use strategies for reducing net greenhouse gas emissions’ project, supported by the Foundation for Strategic Environmental Research (MISTRA). All sources of funding are gratefully acknowledged.