• abundance;
  • aspen;
  • browsing;
  • conservation;
  • demographics;
  • disturbance;
  • forestry;
  • spatial scale;
  • species;
  • ungulates


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1. Management of trees with high conservation value under altered land use is challenging. This applies to European aspen Populus tremula, a keystone tree species for species conservation in northern forests. Fire suppression in managed forest has reduced niches for sexual regeneration in aspen while levels of browsing have increased with increasing numbers of ungulate herbivores.

2. We combined observational and experimental data from 1953 to 2007 to unravel patterns and causes of changes in aspen abundance in Sweden, the country with the largest forests in the EU. The density of small-sized aspen ramets showed a peak in the early 1970s, followed by a marked decrease. Numbers of moose Alces alces, the most important browser on aspen, showed a similar temporal pattern, but moose numbers peaked 10 years later than aspen. During the same time period, the volume of aspen doubled.

3. The changes in aspen abundance correlate to large-scale changes in forestry including the introduction of clear-cutting practices and extensive clearing of aspen, and cessation of forest livestock grazing and abandonment of marginal farmland.

4. Using exclosures, and controlling for time since disturbance and regeneration status, we monitored aspen demographics for 5 years in an aspen rich landscape. There was an eightfold increase in recruitment rate of established ramets to a height safe from browsing (>3 m) in fenced plots. However, the finite rate of increase, λ, derived from a transition matrix model, was consistently below 1, i.e. the aspen growth rate was negative with or without browsing. This was associated with a decrease in sprouting rate over time.

5.Synthesis and applications. Our results suggest that changes in land use practices are the main cause of changes in aspen abundance at regional and national scales in Sweden during the last 50 years. Restoring regeneration niches, most importantly emulating natural disturbance processes, viz. fire at various spatial scales, and retaining aspen in cleaning and pre-commercial thinnings are the most important management recommendations to secure regeneration of aspen. Protecting established aspen ramets at designated sites from browsing either by fencing or reducing ungulate numbers could be used as complementary management tools.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Populations of woody plants in managed forest landscapes may be unsustainable if regeneration niches are reduced by changes in prevailing disturbance regimes. A negative trend can be especially troublesome if the species concerned are susceptible to browsing and management-induced actions increase the carrying capacity of herbivores. This situation applies to aspen Populus tremula L. in Northern Europe, where forest management has practically eliminated fire as an ecological factor and increased the abundance of food for large herbivores. As a consequence, niches for sexual regeneration of aspen have been much reduced, while levels of ungulate browsing on young aspen have increased (Anonymous 2002).

European aspen is host for a great number of critically endangered species from different taxonomic groups, e.g. saproxylic insects (Tikkanen et al. 2006), and is now designated as a keystone species for biodiversity conservation in northern forests (Martikainen 2001; Watson 2003). Aspen regenerates both sexually by seeds and asexually (clonally) through root suckers (Barnes 1966; Worrel 1995). It regenerate well on burned soil (Kay 1993; de Chantal & Granström 2007), but as a result of effective and long lasting fire suppression in managed forests this substrate is rare, and European aspen now regenerates mostly by sprouting, i.e. clonal growth. There are also signs of a decline in its regeneration, which may partly be due to browsing (Linder, Elving & Zackrisson 1997; Angelstam et al. 2000; Kouki, Arnold & Martikainen 2004). In accordance with this hypothesis, the browsing frequency on aspen by ungulates, especially moose Alces alces (L.), is reportedly high (Näslund 1986; Ericsson, Edenius & Sundström 2001; Edenius & Ericsson 2007; Zakrisson, Ericsson & Edenius 2007). However, aspen seems very resilient to high levels of biomass loss (Eiberle 1975) and the role of browsing as a factor limiting aspen abundance in northern Europe needs to be clarified.

In Sweden, as in other countries in Northern Europe, ungulate populations over the last 50 years have been strongly influenced by more intensive forestry, the cessation of livestock grazing in forests and abandonment of marginal farmland (Ahlén 1975; Bergström & Hjeljord 1987; Gordon, Hester & Festa-Bianchet 2004). In order to attain a more thorough understanding of factors limiting tree abundance, the interactions between disturbance, herbivory and land management history need to be considered (Hessl & Graumlich 2002). However, large-scale, long-term longitudinal analyses of changes in abundance of trees in the context of radically changing land use are rare in Northern Europe, particularly for species prone to browsing and with a high conservation value, such as aspen.

Here we combine observational data on changes in aspen abundance in Sweden from 1953 to 2007 with experimental data on factors affecting its regeneration. The objectives of the study were: first, to quantify changes in aspen abundance, land use and moose numbers over the past 50 years in Sweden; secondly to evaluate the interactions between these variables at different scales; and thirdly to evaluate the impacts of disturbance and browsing on aspen demographics; and finally, to recommend management strategies to conserve aspen. To achieve the first two objectives, we analysed patterns of abundance of small-sized aspen stems and standing volume of aspen in relation to changes in land use and moose numbers, in order to quantify their co-variation at national and regional levels. To meet the third objective, we compared aspen regeneration within and outside exclosures to disentangle the effects of disturbance and browsing on aspen demographics.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Land use and moose data

We used National Forest Inventory (NFI) data from 1953 (the first year for which digital data were compiled by the Swedish NFI) to analyse the abundance of aspen at both national and regional scales. We collated data on aspen volume and stem numbers in different size classes on forest land, but excluding protected areas. For this analysis, we used density of stems ha−1 in stem diameter class 1–99 mm (diameters measured at 1·3 m height) as an index of aspen regeneration. From 1973, it is possible to distinguish stems of 1–19 mm and 20–99 mm diameters in the data and these two size classes strongly co-vary from 1973 onwards. The total annual number of NFI-plots in northern and southern Sweden was 3400–7200 and 2400–6100, respectively. Aspen were found in 130–770 (northern Sweden) and 190–1100 plots (southern Sweden). Estimated over a 3-year period, the coefficient of variation of estimated number per hectare of aspen stems 1–99 mm diameter was 6–9% (northern Sweden) and 5–8% (southern Sweden).

We also used NFI data to extract information on the area of annually cut forest, area of mechanical clearing and the volumes of cut aspen. Official Statistics of Sweden (Anonymous 2005) were used to collect data on changes in the area of pasture land used as an index of livestock forest grazing. Official hunting bag data from the Swedish Association for Hunting and Wildlife Management were used to index changes in moose numbers. Roe deer Capreolus capreolus L. and hare (Lepus timidus L. and L. europaeus Pallas) also use aspen as food, but moose are the most important browsers on aspen in Sweden (Näslund 1986; Edenius & Ericsson 2007). Therefore, we concentrate on moose numbers in this paper. Reviews in Sweden and elsewhere (e.g. Solberg et al. 1999; Lavsund, Nygrén & Solberg 2003; Seiler 2005) have shown that harvest data index the population development correctly (Ball, Ericsson & Wallin 1999). However, there is a harvest lag in the population development of around 2 years (e.g. Solberg et al. 1999).

Exclosure study

The exclosure study was performed in the hemi-boreal zone (sensuAhti, Hämet-Ahti & Jalas 1968) north of Uppsala (60°00′N, 18°20′E), one of the most aspen-rich regions in Sweden. This forest-dominated area (75% forest, 15% farmland and 10% lakes and mires) is intensively managed for timber and pulp production, by means of clear cutting with rotation periods of about 80–100 years. About 15–20% of the forest land is covered by young (sapling) forests, and stands >100 years old are very uncommon (see Edenius & Ericsson 2007 for further details on vegetation, etc.).

Moose density was estimated at 1·2–1·5 moose km−2, which is high compared with the regional average for the same period (S. Holm, Swedish Association for Hunting and Wildlife Management, pers. comm.).

Selection of aspen clones

Eight 1 × 1 km cells located 5 km apart were selected from the National Grid over Sweden and each cell was screened for occurrence of aspen clones in the field. We defined spatially distinct units of ramets located at least 100 m away from another unit as a clone (see also Edenius & Ericsson 2007). We avoided clones in farmland fringes, in order to remove confounding effects of alternate land use practices and homogenize environmental variation. This left us with aspen clones located in interior forest land. The soils here are relatively uniform with respect to fertility and compared with farmed areas their productivity is lower. Two clones were selected per 1 × 1 km cell.

A total of 16 aspen clones were selected based on two criteria: (i) forest age class, i.e. whether they were located in cutover areas and sapling stands (<2 m in height) or mature/old forest and (ii) regeneration status, defined by the density of small aspen ramets (cf. Shepperd, Bartos & Mata 2001). The density of small ramets (<0·5 m) varied between 1 and 350 ramets 100 m−2. Using 45 ramets 100 m−2 as the cut-off point, 6 and 10 clones were designated as high and low density clones, respectively. Due to difficulties in finding high density clones, the final selection of clones became somewhat unbalanced, with three clones each of high density and five of low density type per forest age class. Eight of the aspen clones were located in coniferous forest (>70% conifer basal area), six in mixed forest (<70% each of deciduous and conifer basal area), and two in deciduous forest (>70% deciduous basal area). In total, 1730 ramets were tallied during the course of the study, excluding ramets in the immediate vicinity of the fences.

Herbivore exclusion and field measurements

Three 5 × 5 m plots located in the centre of each clone were allotted to control, exclusion of moose, roe deer and hare, and exclusion of moose and roe deer only treatments. Fences with 10 × 10 cm mesh size and a height of 3 m were erected in April 2002 around the herbivore exclusion plots. Hares were excluded by covering the lowest 1 m of the exclosure with chicken wire of 2·5 × 2·5 cm mesh size (see Zakrisson, Ericsson & Edenius 2007 for additional details).

Between April 2002 and April 2008, old and new ramets <3 m tall within plots were annually tagged, their height was measured and they were checked for signs of browsing. Due to a shortage of taller stems, ramets of unequal width were categorized into three height classes: small, <0·5 m; medium, 0·5–1 m; and large, 1–3 m. Stems >3 m tall were deemed beyond the reach of moose (see Edenius & Ericsson 2007) and were excluded from further analysis.

Statistical analysis

We analysed yearly growth rate and survival of individual ramets for data pooled over clones. When not stated otherwise, the results are presented as annual means for years 2003/2004–2007/2008.

To condense the demographic data into a single population growth estimate, the finite rate of increase, λ, reproduction and survival rates (see Table 1 for definitions) were entered into a deterministic stage-structured population model (Caswell 1989). A 3 × 3 matrix accommodated transitions from small to medium to large height classes. The diagonal elements contained the proportion of ramets retained within each class, for each time step. The reproduction rate, i.e. proportion of new ramets (sprouts), was added to the proportion of small ramets in the first matrix row element. Finite rate of increase measurements are unreliable if the initial size structure deviates significantly from the stable size distribution and the sample size is small (Caswell 1989). Thus, to obtain a reliable sample size, yearly transition matrices were calculated for data pooled over clones.

Table 1.   Demographic measures and variables of aspen used in the statistical analysis in the exclosure experiment
Demographic measureDependent variableIndependent variable
Sprouting rateNo. of sprouts current yearNo. of ramets <3 m preceding year
Transition rate, small to medium sizeNo. of ramets 0·5–1 m entered from <0·5 m height class current yearNo. of ramets <0·5 m preceding year
Transition rate, medium to large sizeNo. of ramets 1–3 m entered from 0·5 to 1 m height class current yearNo. of ramets 0·5–1m preceding year
Transition rate, large to beyond browsing heightNo. of ramets >3 m entered from 1–3 m height class current yearNo. of ramets 1–3 m preceding year
Mortality rate, small rametsNo. of dead ramets in <0·5 m height class current yearNo. of ramets <0·5 m preceding year
Mortality rate, medium to large rametsNo. of dead ramets in 0·5–3 m height class current yearNo. of ramets 0·5–3 m preceding year

Mixed effect modelling was employed for analysing the demographic data. Mixed effect models handle random and fixed effects within a coherent analytical framework and are particularly useful in situations with spatial pseudo-replication in the data (Pinheiro & Bates 2000). Year and clone id were entered as random effects in the models. Fixed effects included time since disturbance (forest age class: cutover and sapling stands vs. mature and old forest), regeneration status (density class) and treatment (exclusions vs. control). Treatment was nested under clone, to account for the experimental design with exclosures and control plots being located within the same clone.

Regression tree analysis was adopted as a means to unravel the hierarchical structure of the exclosure study data. In regression trees, models are fitted using binary recursive partitioning, whereby the data are successively split along the coordinate axes of the explanatory variables, such that maximally distinguishable splits of the response variable are achieved (Crawley 2002). Regression tree graphics are useful because they facilitate visual interpretation of relationships and interactions among multiple explanatory variables (factors) (Breiman et al. 1984).

Variables were log-transformed to meet assumptions of homoscedasticity in the analyses, which were carried out in systat 12.0 statistical software (SYSTAT Software, Inc., Richmond, CA, USA). RAMAS EcoLab software (Applied Biomathematics, Setauket, NY, USA) was used for the population matrix analysis.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Aspen dynamics and land use changes at national scale

The density of aspen in the 1–99 mm size class increased from about 50 stems ha−1 in southern Sweden in 1953 to a peak of about 120 stems ha−1 in the early 1970s, then steadily declined back to the 1953 level by 2007 (Fig. 1). There was a steady increase in standing volume of aspen from c. 1 m3 ha−1 in 1953 to 2 m3 ha−1 in 2007. From 1989 to 2007, the annual cutting of aspen decreased by almost 70%, i.e. there is a close to reciprocal match between growth and cutting volumes of aspen, over the last 20 years.


Figure 1.  Density of aspen stems in diameter class 1–99 mm (3-year moving averages) and number of shot moose in Sweden between 1953 and 2007. Aspen was included in the group of “other deciduous trees” between 1983 and 1987, so these years were omitted.

Download figure to PowerPoint

The annual hunting harvest of moose increased from around 30,000 animals in 1953 to a peak of 175000 in the early 1980s (Fig. 1), after which the numbers of moose shot declined.

The proportion of young forest (all species) (0–30 years old) increased from 15 to 35% of the forest land in northern Sweden between 1955 and 1980. The area of mechanical pre-commercial thinning increased in parallel with the area of young forest, up to the mid-1980s, after which there was a marked decrease. As a consequence, the area of sapling forest in need of pre-commercial thinning increased from the mid-1990s, illustrated by the increasing amounts of birch (Betula pendula Roth and B. pubescens Ehrh.) (Fig. 2). It has been estimated that 30000 ha of young forest was sprayed with broad-spectrum herbicides on an annual basis between 1968 and 1986 (Ekelund & Hamilton 2001).


Figure 2.  Relative changes in amounts of birch and aspen in northern and southern Sweden between 1983 and 2007. Arrows denote reference year for birch (1984) and aspen (1989).

Download figure to PowerPoint

Effects of exclosures on demographic parameters

The mortality rate for small ramets varied between 9 and 30%; it was lowest in control plots of low density type in cutovers and sapling stands, and highest in both exclosures and controls in high density stand types in mature/old forests (Table 2). There was no significant effect of treatment, time since disturbance or regeneration status on mortality rates in any of the height classes (Table 3). Browsing rate was highest (22%) in high density clones in cutovers/sapling stands. The rate in low density clones in cutovers/sapling stands and all clone types in mature/old forest amounted to 9–10%, i.e. half that in cutovers/sapling stands. A total of 93% of medium-sized ramets browsed for the first time in 2005/2006 (the winter season with most instances of browsing) were alive in April 2007 and 86% in April 2008 (N = 46). Corresponding figures for unbrowsed ramets were 95 and 86%, respectively (N = 154), i.e. almost identical fractions. Of medium-sized ramets browsed for the first time in 2003 (N = 35), 82% were alive in April 2008. A total of 40% of the ramets in this cohort had been browsed two or more times.

Table 2.   Summary statistics of effects of treatment (herbivore exclusion) on aspen, time since disturbance (cutovers/sapling forest vs. mature/old forest) and clonal regeneration status (high vs. low density of small ramets) on demographic variables
Demographic measureMeanRangeHerbivore exclusion plotsControl plots
Cutovers/sapling stageMature/old forestCutovers/sapling stageMature/old forest
High densityLow densityHigh densityLow densityHigh densityLow densityHigh densityLow density
  1. L = lowest, H = highest value. = 16.

Sprouting rate0·080·02–0·14  H   HL
Transition rate, small to medium size0·120·05–0·16H  H  L 
Transition rate, medium to large size0·120·01–0·19 HL     
Transition rate, large to beyond browsing height0·010–0·03H L LLL 
Mortality rate, small ramets0·190·09–0·30  H  LH 
Mortality rate, medium to large ramets0·080·03–0·12L H L H 
Finite rate of increase, λ0·930·89–0·97H  L    
Table 3.   Results from mixed effect model analysis of effects of treatment (herbivore exclusion), time since disturbance (cutover/sapling forest vs. mature/old forest), and aspen clonal regeneration status (high vs. low density of small ramets) on demographic variables in the exclosure study
Demographic variableRandom effects, variance component partitioning, %Fixed effects, level of significance
Clone idYearResidual errorTreatmentTime since disturbanceRegeneration status
  1. N = 16. NS = >> 0·1; *< 0·05, **< 0·01, ***< 0·001.

Sprouting rate291655NSNS**
Transition rate, small to medium size351640·079NSNS
Transition rate, medium to large size34462***NS**
Transition rate, large to beyond browsing height34264*NSNS
Mortality rate, small ramets32761NSNSNS
Mortality rate, medium to large ramets311059NSNSNS
Finite rate of increase, λNANANANSNSNS

The sprouting rate varied from 2% in control plots in mature/old forest of low density type, to 14% in high density clones in mature/old forest (Tables 2 and 3). The regression tree analysis revealed a first split of the data between high and low density clones, with the former having a higher sprouting rate (Fig. 3a). For the high density clones, there was a second split of the data for time since disturbance, with the highest sprouting rate seen in mature/old forest.


Figure 3.  Tree regression graphics displaying effects of treatment (fencing), time since disturbance and aspen clonal regeneration status on: (a) sprouting rates, (b) transition of ramets from medium to large size.

Download figure to PowerPoint

The transition rate from medium to large-size class showed strong deviations, from 1% in exclosures in high density clones in mature/old forest to 19% in exclosures in low density clones in cutovers/sapling stands (Tables 2 and 3). In absolute numbers, these figures correspond to 0·3–3·8 ramets 100 m−2 year−1. There was a strong effect of treatment, and regeneration status, on transition rate (Table 3). In absolute numbers, about eight times more ramets escaped to heights safe from browsing in exclosures than in control plots each year. There was a significant effect of treatment on transition rate, but not of time since disturbance or regeneration status. The regression tree analysis produced a single split of the data between control plots and exclosure plots, for the transition from small to medium size, and from large to beyond browsing reach, respectively. There were higher transitions rates in the exclusion plots. The transition from medium to large size produced a more complex splitting pattern; similar to the transitions in the other size classes, there was a first split of the data between control and exclosure plots, and a second split of the data for exclosure plots for time since disturbance, with the highest transition rate in cutovers/sapling stands (Fig. 3b).

Average residence time within the browsing height interval (0·5–3 m) decreased from 35·7 to 14·3 years following fencing; hence there was a 2·5-fold reduction in the duration of exposure to browsing. No significant effect of treatment (fencing) or time since disturbance on sprouting rate was observed. In contrast, there was a clear effect of regeneration status, with high density clones showing a higher sprouting rate than low density clones (Tables 2 and 3).

The population growth rate, finite rate of increase (λ), varied between 0·89 and 0·97. It was lowest in exclosures in closed-canopy mature and old forests of low density type and highest in exclosures in open-canopied cutovers and sapling stands of high density type (Table 2). There was no significant effect of treatment, forest state or developmental stage on λ (Table 3). The mean λ value, 0·93, was significantly lower than 1, and thus a negative population growth rate, giving an average 50% reduction in aspen population over 10 years.

Among random effects in the mixed effect models, the proportion of variance accounted for by clone id was 32% on average, with little variation among the demographic variables (Table 3). In comparison, the effect of year, i.e. temporal variation, was much smaller, averaging 7% (range 1–16%). The strongest effect of year was seen in sprouting rate, with a steady decline over time (Fig. 4).


Figure 4.  Changes in sprouting rate over time in fenced aspen clones over the period 2003/2004–2007/2008. Spouting rate was close to zero in low density cones in older forest. Mean values ± SE.

Download figure to PowerPoint


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

It is evident that aspen abundance has undergone major changes in Sweden over the last 50 years. The initiation of the sharp increase in abundance during the 1950s and 1960s can be attributed to the broad-scale introduction of clear-cutting practices in forestry, which facilitated the establishment of early successional species, such as aspen (Bergström & Hjeljord 1987; Edenius et al. 2002). Cessation of livestock grazing in forest and abandonment of marginal farmland during this period are also contributory factors, particularly in southern Sweden where the human population density is higher (Ahlén 1975). Aspen would also have experienced a relatively low browsing pressure from wild herbivores during the initial increase phase, as indicated by the moose population data.

The steady increase in density of small aspen stems came to an abrupt end in the early 1970s. The forest industry had been combating the encroachment of young forest by undesirable deciduous trees, including aspen, with mechanical clearing and herbicides for many years. It is therefore reasonable to assume that these efforts had finally been successful. In addition, regenerating aspen now faced higher browsing pressure from increasing numbers of ungulates. However, based on these trends, the role of ungulates as a regulatory factor for aspen abundance is not clear. The moose population continued to increase dramatically for ten years, while aspen stem numbers dropped significantly after the early 1970s peak. Encroachment of forest fringes by aspen following abandonment of marginal farmland and cessation of livestock grazing should be contributory factors. We have no reliable data on the area of forest livestock grazing, but it is estimated that pasture land, in general, decreased from about 1 million ha in the mid-1930s to less than 300,000 ha in the mid-1960s, in southern Sweden (Anonymous 2005). Another plausible cause might be that the increase in density of small aspens stems, mediated by the intensified forestry, was of shorter duration in southern Sweden, due to the faster growth of trees there.

Densities of small aspen stems are now at the same levels as before the broad scale introduction of clear-cutting practices in Swedish forestry. Given the recent increase in the density of small birch stems one might ask why aspen has not responded in the same way. The exclosure study revealed some interesting results with respect to causative factors. Sprouting rate was higher in high density than in low density clones, regardless of time since cutting and whether there was protection from browsing or not. It has been shown that sprouting rate and ramet density is positively correlated and that this relationship may be only weakly related to ecological factors (Zakrisson, Ericsson & Edenius 2007). Clone id accounted for a relatively high share of the variance in sprouting rate (29%). Such a high share of variance attributable to the clone id may reflect a genetically determined variation in self-replacement capability, but also in palatability to herbivores (Jelinski & Fisher 1991). Demonstration of small scale genetic structuring in aspen (Suvanto & Latva-Karjanmaa 2005) supports this. In effect, this suggests that some aspen clones may have the capability to produce an abundance of new shoots, even with high levels of browsing. An important implication of this variability in sprouting rate is that some clones have a disproportionate impact on aspen regeneration at the landscape scale (Zakrisson, Ericsson & Edenius 2007; Edenius & Ericsson 2007).

In the fenced exclosures, aspen growth rate increased more than twofold, mean residence time within browsing height was reduced by a factor of 2·5 and recruitment into height classes safe from browsing was eight times faster than controls. Browsing thus had a significant impact on transition rates, particularly for the transition from medium (0·5–1 m) to large (1–3 m) aspen size class. Transition rate was faster for fenced plots in cutover/sapling forest sites than in mature/old forest, implying that browsing had a much stronger effect on aspen growth in young open forests. We hypothesise that this difference arises from the fact that browsing rate is much higher in young forests. Spatial variation in food resources at the patch and forest stand level is an important determinant of ungulate foraging behaviour (Edenius, Ericsson & Näslund 2002b; Kuijper et al. 2009), with resultant cascading effects on other biota (Gordon, Hester & Festa-Bianchet 2004). Moose preferentially browse in young forest, because of the abundance of food there, particularly during winter (Edenius et al. 2002).

We found mortality of ramets to be little affected by browsing. Strand et al. (2009), in a simulation study, identified a drop in aspen regeneration at a threshold of 26% browsing, i.e. at a higher level of browsing than in this study. Other studies have also reported a small impact of mammalian browsing on ramet mortality in aspen (P. tremuloides Michx., Yellowstone National Park, Romme et al. 2005; coastal northern Sweden, Zakrisson, Ericsson & Edenius 2007; Eastern Finland, Latva-Karjanmaa, Penttilä & Siitone 2007 and den Herder, Kouki & Ruusila 2009). Up to 30% of the ramets died on an annual basis in the exclosure study, which seemed high. However, self-thinning is a pronounced trait in aspen (Peterson & Jones 1997). Besides mammalian herbivory, ramets succumb to drought, fungi and insect attacks, factors that may eliminate whole cohorts of ramets (Bärring 1988). Therefore, a high turnover of ramets is expected, even without browsing and, as a consequence, only a small fraction of sprouts reach maturity.

An interesting question emerging from our analyses is whether regeneration today is sufficient to sustain long term aspen abundance. In terms of frequency, extent and intensity, the disturbance regime is very different compared with the pre-industrial era, when conditions for sexual regeneration of early successional deciduous trees such as aspen were much better (Zackrisson 1977; Niklasson & Granström 2000). The data at hand indicate that aspen regeneration in Sweden was caused by changes in land use practices after World War II, and that conditions for regeneration thereafter deteriorated when the large-scale transformation of the forest landscape was completed. This is in line with observations that periods of high regeneration in aspen are interwoven with long periods of limited regeneration (Kay 1997). Regeneration may be slow over extended periods of time, without necessarily threatening long term persistence. Aspen has the capacity to respond rapidly when optimum conditions for establishment and survival of ramets coincide. It has been speculated that an increase in spring early-summer drought events induced by a warmer climate, have reduced regeneration of aspen in the western USA (Worral et al. 2008). We cannot rule out climate as a cause for variation in aspen regeneration in Sweden, but drought should be a less import factor in the moister and cooler boreal forests compared to temperate forest. It has also been speculated that, as a consequence of a more conifer-dominated landscape, a gradual build up of a conifer seed source will make it more difficult for aspen to re-establish itself (Strand et al. 2007).

Methodological considerations

We used densities of small-sized aspen ramets as a proxy of regeneration instead of age. One obvious reason was that no age data on small aspen ramets were available in NFI. Size might also be more relevant than age for analysing browsing impacts, because browsing rate, growth and mortality in trees are strongly size-dependent (Edenius & Ericsson 2007 and references therein). The stage-structured approach is common in analysis of tree demographics (e.g. Huenneke & Marks 1987; Rooney et al. 2000; Kwit, Horwitz & Platt 2004; Edenius & Ericsson 2007).

The clonal growth habit of aspen poses a problem with regards to applying an experimental protocol with treatment units nested within clones. As an integrated unit, deficits in terms of biomass loss due to browsing in control plots may be compensated for by nutrient translocation from the fenced plots within the clone. It should be noted though, that viewing the clone as an integrated “super-unit” could be overly simplistic as the root system gradually begins to disintegrate as the clones matures, i.e. clusters of more or less independent and self-sustaining ramets arise over time (Barnes 1966). We decided on this design based on the pronounced clone-to-clone variability between size structure and ungulate use in an earlier study of unfenced clones in the same area (Edenius & Ericsson 2007). We also opted for to minimize environmental heterogeneity by selecting aspen clones in interior forest land only. The findings in this study confirm the great between-clone difference in growth traits.

Implications for conservation and management

Aspen ramets are relatively short-lived, but the clone (genet) as an entity may be very long-lived. Therefore, conservation has to be conducted within a relevant temporal framework. Our findings illustrate the complexity of managing tree species such as aspen that are preferentially browsed by ungulates, particularly when the priority is to protect their conservation values under rapidly changing land use. This is especially true for Northern Europe, where forest systems have been transformed, in terms of carrying capacity for browsers (Cederlund & Markgren 1987; Gordon, Hester & Festa-Bianchet 2004; Milner et al. 2006).

Today, the policy is to increase the abundance of deciduous trees in managed forests (FSC 2009). The reduced cutting of aspen during the last 20 years suggests that changes in nature conservation policy have contributed to the recent increase in aspen volume. In conclusion, our findings attest to the pivotal role of land use changes as a driver of change in aspen abundance and suggest a limited role of browsing. However, browsing reduces recruitment rate of large aspen trees, which may negatively affect organisms dependent on old growth stages. This would be most pronounced in areas where aspen is less abundant. Here fencing may help to increase transition rate beyond browsing height and speed up recruitment. Fencing should preferably target clones with a high capacity for regeneration as these clones have a disproportionate impact on regeneration (Edenius & Ericsson 2007).


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) and SLU’s Thematic Research Program Wildlife and Forestry and the research program Adaptive Management of Fish and Wildlife provided funding for the study. Jean-Michel Roberge and Nils Bunnefeld and three anonymous referees gave valuable comments on an earlier draft of the manuscript. We also thank land-owners and field workers for their dedicated assistance.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Ahti, T., Hämet-Ahti, L. & Jalas, J. (1968) Vegetation zones and their sections in northwesten Europe. Annales Botanici Fennici, 5, 169211.
  • Ahlén, I. (1975) Winter habitats of moose and deer in relation to land use in Scandinavia. Viltrevy, 9, 45192.
  • Angelstam, P., Wikberg, P.-E., Danilov, P., Faber, W.E. & Nygrén, K. (2000) Effects of moose density on timber quality and biodiversity restoration in Sweden, Finland, and Russian Karelia. Alces, 36, 133145.
  • Anonymous (2002) Skogsdata 2002, Tema ungskogar. Official data about Swedish’ forests from the National Forest inventory. Swedish University of Agricultural Sciences, Umeå.
  • Anonymous (2005) Svenskt jordbruk i siffror 1800–2004. Statistik från Jordbruksverket, Rapport 2005:6. (25.10.2009).
  • Barnes, B.V. (1966) The clonal growth habit of American aspens. Ecology, 47, 439447.
  • Bärring, U. (1988) On the reproduction of aspen (Populus tremula L.) with emphasis on its suckering ability. Scandinavian Journal of Forest Research, 3, 229240.
  • Bergström, R. & Hjeljord, O. (1987) Moose and vegetation interactions in Northwestern Europe and Poland. Swedish Wildlife Research, Supplement, 1, 213228.
  • Breiman, L., Friedman, J.H., Olshen, R.A. & Stone, C.I. (1984) Classification and Regression Trees. Chapman & Hall, Belmont, California, USA.
  • Ball, J. P., Ericsson, G. & Wallin, K. (1999) Climate change, moose and their human predators. Ecological Bulletins, 47, 178187.
  • Caswell, H., (1989) Matrix Population Models. Sinauer Associates, Sunderland, Massachusetts, USA.
  • Cederlund, G. & Markgren, G. (1987) The development of the Swedish moose population, 1970–1983. Swedish Wildlife Research, Supplement 1, 1, 5561.
  • Crawley, M. J. (2002) Statistical Computing. Wiley, Chichester.
  • de Chantal, M. & Granström, A. (2007) Aggregations of dead wood after wildfire act as browsing refugia for seedlings of Populus tremula and Salix caprea. Forest Ecology and Management, 250, 38.
  • Edenius, L. & Ericsson, G. (2007) Aspen demographics in relation to spatial context and ungulate browsing: implications for conservation and forest management. Biological Conservation, 135, 293301.
  • Edenius, L., Bergman, M., Ericsson, G. & Danell, K. (2002) The role of moose as a disturbance factor in managed boreal forests. Silva Fennica, 36, 5767.
  • Edenius, L., Ericsson, G. & Näslund, P. (2002) Selectivity by moose vs the spatial distribution of aspen: a natural experiment. Ecography, 25, 289294.
  • Eiberle, K. (1975) Results of a simulation of game damage by cutting the shoots. Schweizerische Zeitschrift für Forstwesen, 126, 821839.
  • Ekelund, H. & Hamilton, G. (2001) Skogspolitisk historia. Skogsstyrelsen, Rapport nr 8A. (In Swedish)
  • Ericsson, G., Edenius, L. & Sundström, D. (2001) Factors affecting browsing by moose (Alces alces L.) on European aspen (Populus tremula L.) in a managed boreal landscape. Ecoscience, 8, 344349.
  • FSC (2009) Forestry Stewardship Council. (25.10.2009).
  • Gordon, I. J., Hester, A. J. & Festa-Bianchet, M. (2004) The management of wild large herbivores to meet economic, conservation and environmental objectives. Journal of Applied Ecology, 41, 10211031.
  • den Herder, M., Kouki, J. & Ruusila, V. (2009) The effects of timber harvest, forest fire, and herbivores on regeneration of deciduous trees in boreal pine-dominated forests. Canadian Journal of Forest Research, 39, 712722.
  • Hessl, A.E. & Graumlich, L.J. (2002) Interactive effects of human activities, herbivory and fire on quaking aspen (Populus tremuloides) age structure in western Wyoming. Journal of Biogeography, 29, 889902.
  • Huenneke, L. & Marks, P. L. (1987) Stem dynamics of the shrub Alnus incana ssp. rugosa: transition matrix models. Ecology, 68, 12341242.
  • Jelinski, D. E. & Fisher, L. J. (1991) Spatial variability in the nutrient composition of Populus tremuloides: clone-to-clone differences and implications for cervids. Oecologia, 88, 116124.
  • Kay, C. E. (1997) Is aspen doomed? Journal of Forestry, 95, 411.
  • Kay, C. E. (1993) Aspen seedlings in recently burned areas of Grand Teton and Yellowstone National Parks. Northwest Science, 67, 94104.
  • Kouki, J., Arnold, K. & Martikainen, P. (2004) Long-term persistence of aspen - a key host for many threatened species – is endangered in old-growth conservation areas in Finland. Journal for Nature Conservation, 12, 4152.
  • Kuijper, D.P.J., Cromsigt, J.P.G.M., Churski, M., Adam, B., Jedrzejewska, B. & Jedrzejewska, W. (2009) Do ungulates preferentially feed in forest gaps in European temperate forest? Forest Ecology and Management, 258, 15281535.
  • Kwit, C., Horwitz, C.C. & Platt, W.J. (2004) Conserving slow-growing, long-lived tree species; input from the demography of a rare understory conifer, Taxus floridana. Conservation Biology, 18, 432443.
  • Latva-Karjanmaa, T., Penttilä, R. & Siitonen, J. (2007) The demographic structure of European aspen (Populus tremula) populations in managed and old-growth forests in eastern Finland. Canadian Journal of Forest Research, 37, 10701081.
  • Lavsund, S., Nygrén, T. & Solberg, E. J. (2003) Status of moose populations and challenges to moose management in Fennoscandia. Alces, 39, 109130.
  • Linder, P., Elving, B. & Zackrisson, O. (1997) Stand structure and successional trends in virgin boreal reserves in Sweden. Forest Ecology and Management, 98, 1733.
  • Martikainen, P. (2001) Conservation of threatened saproxylic beetles: significance of retained aspen Populus tremula on clearcut areas. Ecological Bulletins, 49, 205218.
  • Milner, J. M., Bonenfant, C., Mysterud, A., Gaillard, J. -M., Csányi, S. & Stenseth, N. C. (2006) Temporal and spatial development of red deer harvesting in Europe: biological and cultural factors. Journal of Applied Ecology, 43, 721734.
  • Näslund, B.-Å. (1986) Simulation of Damage and Mortality in Young Stands and Associated Stand Development Effects. Department of Silviculture, Swedish University of Agricultural Sciences, Umeå, 147 pp.
  • Niklasson, M. & Granström, A. (2000) Numbers and sizes of fires: long-term spatially explicit fire history in a Swedish boreal landscape. Ecology, 81, 14841499.
  • Peterson, C.J. & Jones, R.H. (1997) Clonality in woody plants: a review and comparison with clonal herbs. The Ecology and Evolution of Clonal Plants (eds Kroon & J.van Groenendael), pp. 263289. Backhuys Publishers, Leiden.
  • Pinheiro, J.C. & Bates, D.M. (2000) Mixed-effect Models in S and S-Plus. Springer, New York, USA.
  • Romme, W.H., Turner, M.G., Tuskan, G.A. & Reed, R.A. (2005) Establishment, persistence, and growth of aspen (Populus tremuloides) seedlings in Yellowstone National Park. Ecology, 86, 404418.
  • Rooney, T. P., McCormick, R. J., Solheim, S. L. & Waller, D. M. (2000) Regional variation in recruitment of hemlock seedlings and saplings in the Upper Great Lakes. Ecological Applications, 10, 11191132.
  • Seiler, A. (2005) Predicting locations of moose–vehicle collisions in Sweden. Journal of Applied Ecology, 42, 371382.
  • Shepperd, W.D., Bartos, D.L. & Mata, A. (2001) Above- and below-ground effects of aspen clonal regeneration and succession to conifers. Canadian Journal of Forest Research, 31, 739745.
  • Solberg, E. J, Sæther, B.-E., Strand, O. & Loison, A. (1999) Dynamics of a harvested moose population in a variable environment. Journal of Animal Ecology, 68, 186204.
  • Strand, E. K., Vierling, L. A., Bunting, S. C. & Gessler, P. E. (2009) Quantifying successional rates in western aspen woodlands: current conditions, future predictions. Forest Ecology and Management, 257, 17051715.
  • Suvanto, L. I. & Latva-Karjanmaa, T. B. (2005) Clone identification and clonal structure of the European aspen (Populus tremula). Molecular Ecology, 14, 28512860.
  • Tikkanen, O.-P., Martikainen, P., Hyvärinen, E., Junninen, K. & Kouki, J. (2006) Red-listed boreal forest species of Finland: associations with forest structure, tree species, and decaying wood. Annales Zoologici Fennici, 43, 373383.
  • Watson, A. (2003) Aspen – boreal symbol. Taiga News, 42, 5.
  • Worral, J. J., Egeland, L., Eager, T., Mask, R. A., Johnson, E. W., Kemp, P. A. & Shepperd, W. D. (2008) Rapid mortality of Populus tremuloides in southwestern Colorado, USA. Forest Ecology and Management, 255, 686696.
  • Worrel, R. (1995) European aspen (Populus tremula L.): a review with particular reference to Scotland I: Distribution, ecology and genetic variation. Forestry (Oxf.), 68, 93105.
  • Zackrisson, O. (1977) Influence of forest fires on north Swedish boreal forests. Oikos, 29, 2232.
  • Zakrisson, C., Ericsson, G. & Edenius, L. (2007) Effects of browsing on recruitment and mortality of European aspen (Populus tremula L.). Scandinavian Journal of Forest Research, 22, 324332.