1 Five treeline species had low seed germination rates and low survivorship and growth of seedlings when transplanted into Alaskan tundra. Seed germination of all species increased with experimental warming, suggesting that the present treeline may in part result from unsuccessful recruitment under cold conditions.
2 Growth, biomass and survivorship of seedlings of treeline species transplanted into tundra were largely unaffected by experimental warming. However, transplanted seedlings of three species (Betula papyrifera, Picea glauca and Populus tremuloides) grew more when below-ground competition with the extant community was reduced. All three measures of transplant performance were greater in shrub tundra than in the less productive tussock or heath tundra. Establishment of trees in tundra may thus be prevented by low resource availability and competition.
3 Two species (Alnus crispa and Populus balsamifera) had low seed germination and survivorship of germinated seeds; transplants of these species did not respond to the manipulations and lost biomass following transplanting into tundra. Isolated populations of these two species north of the present treeline in arctic Alaska probably became established during mid-Holocene warming rather than in recent times.
4 Of all the species studied here, Picea glauca was the most likely to invade intact upland tundra. Its seeds had the highest germination rates and it was the only species whose seedlings survived subsequently. Furthermore, transplanted seedlings of Picea glauca had relatively high survivorship and positive growth in tundra, especially in treatments that increased air temperature or nutrient availability, two factors likely to increase with climate warming.
Both past and present evidence suggest that temperature is a proximate factor limiting the position of the northern treeline. For example, during the middle Holocene, when summer temperatures were higher than they are today, the treeline extended north of its present position (Ritchie et al. 1983; COHMAP 1988; MacDonald et al. 1993). Recent warming (during the past century) has been associated with increased growth of trees at treeline (Innes 1991; Jacoby & D’Arrigo 1995), increases in tree densities in forest stands at treeline (Payette & Filion 1985; Szeicz & MacDonald 1995) and northward expansion of treeline (F. Suarez, D. Binkley and R. Stottlemyer, unpublished data). There is also a compelling present-day correlation between position of treeline and temperature (Sveinbjornsson 1992).
Numerous hypothesized mechanisms by which temperature limits treeline have been proposed (Stevens & Fox 1991; Sveinbjornsson 1992). Some of these mechanisms involve the direct effects of low temperature on physiology (e.g. photosynthesis, growth, and nutrient and water uptake) whereas others involve the indirect effects of low temperature on plant performance (e.g. negative carbon balance resulting from the short growing season, low nutrient availability resulting from cold soils). However, little experimental evidence exists to evaluate these hypotheses.
Given the past association of changes in treeline with climate warming, future climate warming is also likely to cause northward treeline expansion (Starfield & Chapin 1996). Despite the lack of understanding of current treeline, elucidating the factors that control the rate of this expansion is important for several reasons. First, because of the relatively low albedo of boreal forest, replacement of tundra by boreal forest is likely to exacerbate warming that results from the rising concentrations of greenhouse gases (Bonan et al. 1992; Foley et al. 1994). Secondly, knowing the future areal extent of the boreal forest is necessary for predicting future carbon storage by the biosphere (Prentice & Fung 1990; Smith & Shugart 1993). Third, boreal forests are of large economic interest.
In this study, we examined limitations to tree establishment in Alaskan tundra, where, at present, northern treeline coincides with the south slope of the Brooks Range and the range limit of Picea glauca (Viereck 1979). We hypothesized that, although warmer temperatures would directly enhance seed germination, the survivorship and growth of seedlings would be limited by factors that influence below-ground resource availability, namely competition with tundra species and site fertility. We focused our study on four trees (Betula papyrifera Marsh, Picea glauca (Moench) Voss, Populus balsamifera L., and Populus tremuloides Michx.) and one shrub (Alnus crispa (Ait.) Pursh) that currently exist at treeline in northern Alaska. Two of these species (P. balsamifera and A. crispa) also occur in isolated stands north of the Brooks Range (Viereck 1979; Murray 1980; Edwards & Dunwiddie 1985).
Materials and methods
Our experimental site at Toolik Lake, Alaska (68°38′N, 149°34′W, elevation 760 m a.s.l.), is in the northern foothills of the Brooks Range. Treeline presently occurs 50 km south of Toolik Lake, on the southern slope of the Brooks Range, although isolated populations of both A. crispa and Populus balsamifera occur within 20 km of Toolik Lake. Tree seeds for all experiments were obtained primarily from Fairbanks, Alaska (64°53′N, 147°45′W). The remote location of the Alaskan treeline made it impossible to collect adequate seed there, particularly given the large number of experimental treatments and species compared (see below).
Fairbanks has a warmer climate than the treeline, which in turn is warmer than our study site in the northern foothills. Comparative climate data for 1975–76 indicate that the number of thawing degree days were 40–100% and 10–60% greater at Fairbanks and treeline, respectively, than in the Toolik region for these two years (Haugen 1979). Measurements made in 1994 indicate a growing season (12 June–31 July) air temperature at Toolik Lake of 10.7 °C, while for the same period air temperature was 11.7 °C at treeline and 16.2 °C in Fairbanks (D. U. Hooper, unpublished data). Little is known about genetic differentiation among populations from treeline and Fairbanks, and examining genotypic variation in response to our manipulations was beyond the scope of this study.
Establishment of plots
We established plots in three tundra vegetation types: a relatively dry lichen-heath tundra, a more mesic acidic tussock tundra, and a wet acidic shrub tundra. All three sites are within 0.5 km of one another in the vicinity of Toolik Lake, Alaska. The lichen-heath tundra is located on a well-drained slope at the boundary between two adjacent glacial surfaces (Itkillik I and II) (Walker et al. 1995). The tussock and shrub tundra are located on the older of these two surfaces. The tussock tundra is located on a gentle slope, but drainage is impeded by underlying permafrost. The shrub tundra occurs along a water track that primarily drains acidic tussock tundra and flows into Toolik Lake.
Within each tundra type, we established 20 split-plots. One half was left uncovered as a control that experienced ambient climate; the other half was designated for cover with a polyethylene greenhouse that warmed the air. We constructed polyethylene greenhouses by stapling 0.15 mm polyethylene to irrigation-tube frames. Greenhouses were 0.5 m tall; they were placed on warmed plots each spring following snowmelt (mid-June) and removed at the end of each growing season (mid-August).
To examine temperature limitation to seed germination in tundra, we obtained seeds of Betula, Populus balsamifera and P. tremuloides, and Picea near the University of Alaska, Fairbanks, and seeds of Alnus from tundra near Slope Mountain (68°55′N, 148°50′W) on the north slope of the Brooks Range. Seeds were collected in 1990 from multiple individuals in single populations and prechilled over winter by storing them frozen until sowing.
Seeds were sown on 21 June 1991 into a 5 × 2 factorial species × warming experiment in gently sloping upland tussock tundra. Four control and four greenhouse-covered plots were randomly assigned to each of the five species. We estimated seed viability immediately prior to sowing by determining maximum germination rates for each species on moist filter paper in Petri dishes in the laboratory. Within each plot we sowed 500 viable seeds into each of two 400 cm2 subplots located in moss mats (dominated by Hylocomium splendens and Aulacomnium turgidum). Because of low germination rates (and thus lack of sufficient seeds), we sowed only 250 viable seeds into Alnus crispa plots. We watered seeds once, on 24 June 1991, with 150 ml lake water subplot−1. These seed densities generally fall within the range of natural seed rains found in interior Alaska for these species; they are higher than natural seed rains of Picea glauca and lower than those of B. papyrifera (Zasada et al. 1992).
To determine seed germination rate, we counted seedlings on 30 June, 12 July, 31 July and 10 August 1991. Additional germination occurred in the spring of 1992 for all but the Populus species. We counted seedlings on 1 July, 11 July, 22 July and 2 August 1992, distinguishing new from previous year's germinants. We calculated the maximum germination rate as the sum of maximum numbers of seedlings counted during 1991 and 1992 divided by the number of viable seeds sown. Seedlings were counted again on 23 July 1993, 28 June 1994 and 26 July 1995. We determined survivorship in any given year by dividing the number of live seedlings by the maximum germination. We compared arcsine square root-transformed maximum germination rates among species and between treatments using two-way analysis of variance (anova). Low survivorship precluded a statistical comparison of seedling survivorship.
Seedling transplant experiment
We compared direct limitation by temperature with limitation by competition and site characteristics to growth and survivorship of seedlings transplanted into tundra. We transplanted 2-year-old seedlings of all five species into control and greenhouse-covered plots in all three different tundra vegetation types in August 1990. Seedlings were obtained from seed collected in Fairbanks and grown for 2 years at the Alaska State Forest Nursery (Palmer, Alaska). At the time of transplanting, randomly selected seedlings were harvested to determine pretransplant biomass (n = 5 for Populus spp., n = 10 for all other species). Because of the difficulty of warming individual seedlings, we set up a somewhat complicated split-split-split plot design. The first split refers to the temperature treatment (see establishment of plots), the second split refers to the species of seedling, and the third split refers to the competition treatment.
Each split-plot was further split by species: two seedlings of each of the five species were transplanted into each half of each split-plot. Because of a shortage of transplants, Populus tremuloides and P. balsamifera were only planted in 12 randomly chosen plots of the total of 20 plots within each vegetation type. Split-plots containing three species were 0.8 × 0.5 m; those containing all five species were 1.3 × 0.5 m.
We randomly chose one member of each pair for a reduced below-ground competition treatment, and cut vertically into the peat with a serrated knife down to mineral soil in an 18-cm circle centred on this seedling. Cutting severed the roots of adjacent individuals in the extant tundra community. To prevent root regrowth, we inserted a sheet of 0.15-mm polyethylene to 10 cm depth in the vertical cut, except in the heath site where sheets were inserted to 5 cm depth because of the rocky soil and shallow organic layer.
We scored transplants for survivorship at the end of each growing season (late July to mid-August) in 1991–93. We also measured height and the length of the current year's growth increment on 10 August 1991 and 31 July 1992. In early August 1993 we harvested all transplants to determine biomass. We dug around individual seedlings to recover as much of the below-ground biomass as possible. Harvested plants were divided into above- and below-ground (root) biomass. Roots were frozen and returned to the University of California, Berkeley, where they were separated from soil by gently washing them in a water bath. Root nodules of A. crispa were separated from root biomass. We separated current from previous year's above-ground biomass. All plant parts were dried (65 °C) and weighed.
We compared survivorship of transplants among species, sites, warming treatment and competitive regimes to 1993 using log-linear analysis. To determine if survivorship was independent of species or the various treatments, we first constructed a complete-independence model (Selvin 1995) that included survival, species, site, greenhouse treatment and competition treatment. The model also included all possible interactions (two-, three- and four-way) between species, site, greenhouse treatment and competition treatment, since the margins were fixed for these factors by the experimental design. For this model, all possible interactions (two-, three-, four- and 5-way) between survival and the various treatments were set equal to zero. If the complete-independence model did not sufficiently fit the data, all interaction terms were added back to the model, and each was then eliminated in turn to determine which interaction terms were necessary to ensure good model fit (assessed using both log-ratio and chi-square analyses).
We compared transplant height and current year's growth (1991 and 1992) and biomass (1993) for each species separately using split-split plot analysis of variance (SSP anova), with site as a main effect, the greenhouse treatment as a split-plot effect, and the competition treatment as a split-split-plot effect. Low survivorship of some species prevented analysis of the data as a complete split-split-split-plot design with species as the split-split-plot effect. In 1991, the main stems of most Populus balsamifera individuals died back and they resprouted from buds at the base of the stem. Since height was almost always equal to current year's growth, only growth was analysed for this species. Low survivorship prevented us from including the heath site in the split-split-plot analysis for P. balsamifera in 1992 and for B. papyrifera in both 1992 and 1993. Low survivorship prevented us from using split-split-plot analysis altogether for P. balsamifera in 1993 and A. crispa in 1992 and 1993. When split-split-plot analysis was impossible, we averaged growth, height or biomass over the two levels of the competition treatment (when both members of a species-pair survived) and compared sites and the greenhouse treatment using split-plot analysis of variance (SP anova), with site as a main effect and the greenhouse treatment as a split-plot effect. All heights, growth and biomass measures were ln-transformed for statistical analyses.
To serve as a control for possible transplant effects on survivorship and biomass, we transplanted seedlings of each species into a site near Fairbanks, Alaska (where these species establish naturally) in mid-July 1990 (n = 10). The site was located on an upland south-facing slope above the Tanana River in the Bonanza Creek Long-Term Ecological Research (LTER) site UP1 A. The site burned in 1983 and is currently dominated by P. balsamifera trees with scattered A. crispa seedlings and an understorey of Calamagrostis canadensis. We scored transplants for survivorship and harvested them for above-ground biomass in mid-August 1993.
To characterize the three different sites, we measured percentage cover of the dominant plant species, soil moisture, soil pH, depth to mineral soil and accumulation of nutrients on ion-exchange resins. To characterize the effects of the greenhouse treatment, we measured air and soil temperature, depth of thaw to permafrost, and accumulation of nutrients on ion-exchange resins in greenhouse and control plots in all three sites.
We visually estimated percentage cover of all plant species in two 400-cm2 quadrats randomly located within each control plot. Values for the two quadrats were averaged before averaging all plots for a site. We measured soil moisture gravimetrically (65 °C, 5–10 cm below the moss or soil surface) in mid-July 1993 at six points in each of the three sites. At the same time, we also measured depth of the organic mat (from the moss or soil surface). pH was measured in soil collected from the bottom of the live vegetation to 5 cm depth in five randomly positioned cores in each site in late August. Soil (2 g oven-dry equivalent mass) was shaken in 20 ml 0.01 m CaCl2 for 30 min before measuring with a pH meter (Orion Instruments). Soil moisture, organic horizon depth and pH were compared among sites using one-way anova.
From 1 to 8 August 1993, we measured air temperature 15 cm above the canopy in heath tundra (n = 4) and shrub and tussock tundra (n = 2 for controls, n = 3 for greenhouses). Concurrently, we measured soil temperature 10 cm below the moss surface in heath (n = 3 for controls, n = 4 for greenhouses), tussock (n = 3) and shrub tundra (n = 2 for controls, n = 3 for greenhouses). Temperatures were measured in randomly chosen control or greenhouse plots using copper-constantan thermocouples attached to dataloggers (Campbell CR10 or 21X; Campbell Scientific, Logan, UT) that recorded hourly and daily means of 1-min measurements; replication was limited by the number of channels available on the datalogger. Daily mean air and soil temperatures were compared among sites and between treatments using repeated-measures anova, with day as a repeated measure.
We measured depth of thaw to permafrost twice in each plot in late July 1993 using a stainless steel probe. The two measurements were averaged before statistical analysis. We compared sites and the greenhouse treatment using split-plot anova with site as a main effect and the greenhouse treatment as a split-plot effect. Because of the rocky soils in the heath site, we were unable to measure thaw depth in this site, and we have excluded it from the analysis.
We assessed nutrient availability in the plots using ion-exchange resins (Giblin et al. 1994). Seven millilitres of acid-washed (10% HCl) anion or cation exchange resins were placed in 25-cm2 acid-washed nylon bags (Dowex 50 W-X8 20–50 mesh H+ or 1-X8 20–50 mesh Cl−, respectively). We placed anion and cation resin bags at 5–10 cm depth in randomly chosen control and greenhouse plots in each site in mid-June 1992 (n = 7 in heath and shrub tundra, n = 6 in tussock tundra). At the end of July 1992, we replaced those resin bags with fresh resin bags that we harvested the following spring (mid-June 1993). After each harvest, resins were frozen until extraction with 100 ml of 2 N NaCl (in 0.1 N HCl) through a 30-ml column. Extracts were analysed for NH+4, NO−3 and PO−4 colorimetrically on the Lachat QuickChem Autoanalyser. We compared sites and the greenhouse treatment for each ion separately using split-plot anova with site as a main effect and the greenhouse treatment as a split-plot effect. Values were ln-transformed for analyses.
Statistical analyses were done using systat (systat 1992) and JMP Start Statistics (Sall et al. 1996).
The three sites differed in both vegetation and soil parameters (Tables 1 and 2). The heath site was dominated by the ericaceous shrub, Vaccinium uliginosum, lichens, and the mosses Hylocomium splendens and Aulacomnium turgidum. The shrub site was dominated by two deciduous shrubs, Betula nana and Salix pulchra, and the mosses Hylocomium splendens and Sphagnum spp. The tussock site was dominated by species in several different growth forms, including mosses (Sphagnum spp. and Hylocomium splendens), deciduous shrubs (Betula nana), evergreen shrubs (Vaccinium vitis-idaea) and the tussock-forming sedge, Eriophorum vaginatum. The heath site had drier, slightly less acidic soils and a thinner organic horizon than the shrub and tussock sites (Table 2). The shrub site had greater accumulation of ammonium and nitrate on ion-exchange resins during winter (including spring thaw; Table 3; SP anova, P = 0.01 for ammonium and nitrate).
Table 1. Areal cover (%) of plant species in the three tundra types into which seedlings were transplanted. Those species whose cover was <5% are indicated by dashes. Values are means (SE)
Table 2. Soil properties in the three tundra types into which seedlings were transplanted. Values are means (SE). Different letters within a row indicate significantly different means (Tukey's HSD, α = 0.05)
Organic horizon depth (cm)
Soil moisture (%)
Table 3. Nutrient accumulation on ion exchange resins (μmol ion bag−1) in three sites and in control and greenhouse treatments during the growing season (summer) and over winter (which includes soil freeze-up and spring thaw). Values are means (SE)
The greenhouse treatment primarily altered the above-ground environment (Table 4). Air temperatures were significantly warmer in the greenhouses by about 1 °C on average (repeated-measures anova, F1,12 = 100.14, P < 0.001), comparable to the difference in mean growing season temperature between Toolik and treeline south of Toolik (D. U. Hooper, unpublished data). Soil temperatures did not change with greenhouse warming (repeated-measures anova, F1,12 = 0.53, P = 0.48). Consistent with the lack of effect on soil temperatures, greenhouses did not alter thaw depth (SP anova, F1,38 = 1.30, P = 0.26) or the amount of nitrogen accumulated on ion-exchange resins as ammonium or nitrate (Tables 3 and 4). Phosphate was significantly higher in greenhouses in winter (SP anova, P = 0.002). Although we did not measure it, the greenhouses probably reduced photosynthetically active radiation by about 20% (Chapin et al. 1995; Hobbie & Chapin 1998).
Table 4. Air and soil temperature and thaw depth in control and greenhouse treatments in the three transplant sites. Air and soil temperatures are 1-week means of daily means measured from 1 August to 8 August. Thaw depth was measured in mid-July. Values are means (SE)
Air temperature (°C)
Soil temperature (°C)
Thaw depth (cm)
Seed germination and survivorship
Air warming approximately doubled seed germination for all five species (Fig. 1; F1,30 = 5.64, P = 0.02). Species also differed significantly in their maximum seed germination rates (F4,30 = 6.97, P < 0.001). Picea glauca and Betula papyrifera had the highest germination rates, between 5% and 20%. All other species had germination rates less than 5%. By 1995, most germinants had died (Fig. 2). Only Picea glauca had significant survivorship of germinants five growing seasons after seeds were sown.
Transplant survivorship and growth
Three growing seasons after transplanting (i.e. in 1993), seedlings of all species had greater than 80% survivorship when transplanted into a site where they would naturally establish (Fairbanks, Alaska), indicating little mortality due to transplanting alone (Fig. 3). However, when transplanted into tundra, survivorship was quite variable. Survivorship differed significantly among sites, and was lowest in the heath site, particularly for Alnus crispa, Betula papyrifera and Populus balsamifera (Fig. 3 and Table 5; significant survivorship × site and survivorship × species × site interactions). Survivorship was also lower in the greenhouse treatment for all species except Populus tremuloides (Table 5; significant survivorship × species × temperature interaction). The competition treatment had no effect on survivorship of any species (Table 5).
Table 5. Log-linear analysis of survivorship for transplanted seedlings to 1993. u1 = survivorship, u2 = species, u3 = site, u4 = temperature treatment, u5 = competition treatment. Y2 = likelihood ratio statistic; χ2 = Pearson chi-square
Term set equal to zero
All of the below
u1 × u2 × u3 × u4 × u5
u1 × u2 × u3 × u4
u1 × u2 × u3 × u5
u1 × u2 × u4 × u5
u1 × u2 × u3
u1 × u2 × u4
u1 × u2 × u5
u1 × u3 × u4
u1 × u3 × u5
u1 × u4 × u5
u1 × u2
u1 × u3
u1 × u4
u1 × u5
All species accumulated biomass when transplanted into the Fairbanks site (Fig. 4). However, only Picea glauca and Populus tremuloides accrued above-ground biomass when transplanted into tundra. Above-ground biomass of the other three species decreased in tundra.
All species except Populus balsamifera responded significantly to warming in terms of growth (length of current year's growth increment) in the first and second years after transplanting (SP anova, P ≤ 0.01, data not shown). Picea glauca was the only species that responded significantly to warming in terms of total biomass by the third year, having greater biomass in the greenhouses (Fig. 5; SSP anova: P = 0.01).
Three species responded positively in terms of growth (current year's biomass) to decreased below-ground competition and site variation in productivity. Betula papyrifera and Picea glauca had significantly greater growth in the reduced competition treatment (Fig. 6; SSP anova, P < 0.001). Betula, Picea and Populus tremuloides all had significantly greater growth in the shrub site (Fig. 6; SSP anova, P < 0.01); for Betula, the site differences were greater in the greenhouses (SSP anova, site × greenhouse interaction, P = 0.009). Betula, Picea and Alnus crispa had greater total biomass in the shrub site (Fig. 5; SSP anova, P < 0.001 for Betula and Picea; SP anova, P = 0.01 for Alnus). For Populus tremuloides the total biomass response was complicated by a significant site × temperature × competition interaction (SSP anova, P = 0.006). Populus balsamifera responded little, if at all, to any of the treatments in terms of growth or biomass (Figs 5 and 6). None of the species responded to treatments with change in relative allocation to roots and shoots, except Betula papyrifera, for which warming significantly increased the root: shoot ratio, particularly in the tussock site (Fig. 7; SSP anova, site × greenhouse interaction, P = 0.005; site × greenhouse × competition interaction, P = 0.01).
This study suggests that treeline in part results from the inability of trees to establish in tundra, even if dispersal is adequate. The species examined showed limited potential for recruitment into upland tundra, the most widespread tundra community type in Alaska. All had generally poor seed germination and little or no survivorship of seedlings after germination. Furthermore, seedlings transplanted into tundra had poor growth relative to those transplanted into a site where these species establish naturally. Our experiments suggest that recruitment of trees in tundra is currently prevented both directly and indirectly by cold temperatures. Seed germination appears to be directly temperature-limited, while seedling growth and survivorship of at least some treeline species is limited by the availability of below-ground resources, as indicated by the response of the transplants to site differences and reduced competition.
The positive response of seed germination to warming is consistent with previous studies showing that seed germination of boreal species is generally temperature-limited (Black & Bliss 1980; Zasada et al. 1992). However, the fairly low germination rates and survivorship of germinants even in the greenhouse treatment suggest that either higher temperatures than those imposed by our manipulation or other factors besides warm temperatures are also required for successful recruitment of these species in tundra. This seems particularly true of Alnus crispa and Populus spp., whose germination rates even in the greenhouses were less than 5% of viable seed.
An additional factor that may limit seed germination of trees in tundra is the availability of suitable substrate. Many studies have shown that germination rates of boreal species are often lower on organic than on mineral substrates (Putnam & Zasada 1986; Walker et al. 1986; Zasada et al. 1992). The moss substrate of tussock tundra may have been suboptimal for seed germination, and fires, which are important in creating suitable seed beds in boreal forest (Walker et al. 1986), are rare in tundra (Wein 1976). Successful recruitment may be restricted to relatively rare areas of exposed mineral soil, such as stabilized frost boils, within upland tundra (Gartner et al. 1986).
In contrast to seed germination, the growth and survivorship of tree seedlings in tundra were not directly limited by cold air temperatures. In fact, survivorship was reduced by the warming treatment. The only positive response to temperature manipulation was a growth response in the first 2 years after transplanting. That the response disappeared within 3 years suggests that tree seedlings could only respond to warming while they still had access to reserves accumulated prior to transplanting. In contrast to the other species, Picea glauca showed an increase in biomass with warming after 3 years, suggesting that cold temperature directly limits its growth at treeline.
For three of the species studied, Betula papyrifera, Picea glauca and Populus tremuloides, seedling growth and survivorship appeared limited by the availability of below-ground resources. Growth of these three species was greater when below-ground competition with the extant community was reduced, suggesting that soil resources were limiting their growth. Although we did not distinguish experimentally between competition for water or nutrients, water is unlikely to limit growth in wet tundra soils (Oberbauer & Dawson 1992). In contrast, nutrients often limit plant growth in tundra (Chapin & Shaver 1985; Shaver et al. 1986; Chapin & Shaver 1996), so competition for nutrients is likely.
Further evidence for limitation by nutrient availability comes from the patterns of biomass, growth and survivorship of these species among the different vegetation types. Growth, biomass and survivorship were generally highest in the shrub site and lowest in the heath site. The shrub tundra site differed from the other sites primarily in its higher N availability, although its soils were also wetter than those of the other two sites. Other studies have shown that shrub tundra has higher nutrient availability (Kielland 1990) and is more productive (Shaver & Chapin 1991) than tussock or heath tundra. In addition, the shrub tundra site studied here occurred along a water track; water tracks have higher productivity than the surrounding tundra, and this higher productivity has been attributed to greater nutrient mineralization and bulk flow of nutrients within water tracks (Chapin et al. 1988). Fertilizer studies have demonstrated that productivity of heath tundra is extremely nutrient-limited: heath tundra responds the most and shrub tundra the least to nutrient addition (Chapin et al. 1992). Low nutrient availability has also been shown to limit growth of adult trees at treeline in Sweden (Sveinbjornsson et al. 1992).
Two of the species studied, Populus balsamifera and Alnus crispa, responded little, if at all, to the manipulations and had generally poor growth and survivorship. This suggests that some other factor besides air temperature or below-ground resource availability was limiting their growth and survivorship. One possible limiting factor that was not manipulated here is cold soil temperature. Root growth and nutrient and water uptake in boreal forest trees are all inhibited at low temperatures (Tryon & Chapin 1983; Goldstein et al. 1985; Chapin et al. 1986). Moreover, isolated tundra populations of Populus balsamifera are generally restricted to warm springs or steep, south-facing slopes where soil temperatures are relatively warm (Murray 1980).
It is unclear whether our choice of seed provenances (from single populations in Fairbanks) has biased our results in any way. Studies of isozyme variation in several of the species studied reveal little genetic differentiation among populations across broad boreal and subboreal regions, perhaps because of high gene flow (promoted by wind pollination and seed dispersal) combined with the relatively recent arrival of these species to boreal regions (Alden & Loopstra 1987; Farmer et al. 1988; Lund et al. 1992). On the other hand, isozyme variation may be a poor indicator of genetic variation in response to environmental factors. Some research has shown differences in growth potential of Picea glauca that correlate with provenance latitude (Li et al. 1997). However, we are unaware of studies specifically examining genotype by environment interactions for traits related to migration in boreal species.
The current distribution of isolated tree populations in the Alaskan Arctic shows little relationship to the species differences in establishment success that we observed. Alnus crispa and Populus tremuloides, the two treeline species that have been reported repeatedly in arctic Alaska (see the Introduction), were the two species least successful in establishing in our experiments. Thus these isolated populations are probably relics, maintained by clonal growth, of populations that established during mid-Holocene warming.
Given adequate seed dispersal, future warming will likely increase the probability that seeds of the species studied here will germinate in tundra. However, the probability of those seedlings establishing will probably increase only if warming increases nutrient availability, or possibly growing season length (not studied here). This is particularly true for Betula papyrifera, Picea glauca and Populus tremuloides. Because of low availability of suitable sites for germination, recruitment may be generally low, except along riparian zones or unless fire frequencies increase.
If establishment of seedlings proves to be an important bottleneck to treeline expansion, Picea glauca will probably invade tundra more readily than the other species studied here. Picea glauca is more likely than other species to establish from seed in intact upland tundra vegetation since it had relatively high germination rates and was the only species whose seedlings survived during the study once seeds had germinated. Furthermore, P. glauca transplants had relatively high survivorship and positive growth in tundra and responded both to increased temperature and to treatments that probably increased nutrient availability, a situation likely to occur with climate warming. Root growth and nutrient uptake are less sensitive to temperature in P. glauca than in the other species we studied (Tryon & Chapin 1983; Chapin et al. 1986). Our suggestion that P. glauca will probably invade tundra with warming is consistent with other studies showing establishment of Picea glauca north of treeline in northern Alaska during recent warm decades (Cooper 1986; F. Suarez, D. Binkley and R. Stottlemyer, unpublished data). Because P. glauca is one of only two evergreen species occurring at North American treeline, its expansion could significantly decrease regional albedo with implications for the global climate.
We thank Jennifer Dekker and Eric Stendell for help processing samples, Chris Lund and Anna Shevtsova for help in the field, and the Toolik Lake LTER program for use of their dataloggers. Funding was provided by an NSF predoctoral fellowship, a NASA Global Change Fellowship, and an NSF Doctoral Dissertation Improvement Grant (BSR 91-22791) to SEH and an NSF Grant (BSR 87-05323) to FSC.
Received 24 April 1997revision accepted 17 November 1997