Bottom-up versus top-down control of tree regeneration in the Białowieża Primeval Forest, Poland


  • Dries P. J. Kuijper,

    Corresponding author
    1. Mammal Research Institute, Polish Academy of Sciences, ul. Waszkiewicza 1d, 17-230 Białowieża, Poland
    Search for more papers by this author
  • Joris P. G. M. Cromsigt,

    1. Mammal Research Institute, Polish Academy of Sciences, ul. Waszkiewicza 1d, 17-230 Białowieża, Poland
    2. Center for Ecological and Evolutionary Synthesis, University of Oslo, PO Box 1066 Blindern, 0316 Oslo, Norway
    Search for more papers by this author
  • Bogumiła Jędrzejewska,

    1. Mammal Research Institute, Polish Academy of Sciences, ul. Waszkiewicza 1d, 17-230 Białowieża, Poland
    Search for more papers by this author
  • Stanisław Miścicki,

    1. Department of Forest Management, Geomatics and Forest Economics, Forestry Faculty, Warsaw University of Life Sciences, ul. Nowoursynowska 166, 02-787 Warszawa, Poland
    Search for more papers by this author
  • Marcin Churski,

    1. Mammal Research Institute, Polish Academy of Sciences, ul. Waszkiewicza 1d, 17-230 Białowieża, Poland
    Search for more papers by this author
  • Włodzimierz Jędrzejewski,

    1. Mammal Research Institute, Polish Academy of Sciences, ul. Waszkiewicza 1d, 17-230 Białowieża, Poland
    Search for more papers by this author
  • Iwona Kweczlich

    1. Department of Forest Management, Geomatics and Forest Economics, Forestry Faculty, Warsaw University of Life Sciences, ul. Nowoursynowska 166, 02-787 Warszawa, Poland
    Search for more papers by this author

Correspondence author. E-mail:


1. We tested the interactions between biotic and abiotic factors in structuring temperate forest communities by comparing tree recruitment after 7 years inside 30 pairs of exclosure (excluding ungulates: red deer, roe deer, bison, moose, wild boar) and control plots (7 × 7 m each) in one of the most natural forest systems in Europe, the Białowieża Primeval Forest (eastern Poland). The strictly protected part of that forest hosts the complete native variety in trees, ungulates and their carnivores and excludes human intervention.

2. We analysed whether the exclosure effect interacted with abiotic factors, relevant for tree recruitment (canopy cover, ground vegetation cover, soil fertility and soil wetness) at different stages of tree regeneration (seedlings, saplings ≤ 50 and > 50 cm).

3. Contrary to our expectations, a single factor dominated at each stage of tree regeneration. Herbaceous vegetation cover was the main factor determining the number of seedlings with an optimum at 38% of cover. Soil fertility determined the density of saplings ≤ 50 cm, with on average three times higher density on eutrophic than on oligotrophic soils. Herbivory was the main factor determining recruitment rate of trees into > 50 cm size classes only.

4. The density of saplings that grew into the > 50 cm size class was more than three times higher in the exclosures than in the control plots during 7 years. In the absence of ungulates, on average 3.1 species recruited into the > 50 cm size class compared to 1.7 in control plots. Tree species occurred in more equal proportions inside exclosures, whereas species composition was pushed towards strong dominance of a preferred forage species, Carpinus betulus, in the presence of ungulates. This suggests that preference of species by ungulates can coincide with tolerance to browsing.

5.Synthesis. The study showed that abiotic conditions dominated the early stages and ungulate impact the later stages of tree regeneration, indicating the context-dependence of herbivore top-down effects. Heterogeneity in abiotic and biotic conditions may, therefore, have an important influence on the strength of top-down effects and the role that herbivores play in natural ecosystems.


There has been an ongoing debate on the importance of bottom-up versus top-down forces in structuring plant communities (McNaughton et al. 1989; Hunter & Price 1992; Power 1992).

Bottom-up forces are environmental conditions affecting primary production and, as a result, higher trophic levels. In contrast, top-down forces are higher trophic levels impacting lower levels (Fretwell 1987; DeAngelis 1992). There is general agreement that both factors affect plant communities (Polis & Strong 1996), but the question remains what the relative strength of such top-down and bottom-up forces is and if we can find general patterns in their effects (Gripenberg & Roslin 2007). Increasing evidence suggests that the balance of top-down and bottom-up forces varies over environmental or productivity gradients (Oksanen et al. 1981; Osem, Perevolotsky & Kigel 2002; Kuijper & Bakker 2005; Bakker et al. 2006). Much of this evidence comes from grassland systems. Here, we investigate the relative effects of both forces on forest dynamics.

Studies from boreal forest systems illustrate the overriding importance of abiotic bottom-up factors relative to biotic herbivore control (Turkington et al. 2002), whereas others show that abiotic and biotic factors change in importance during different stages of boreal forest primary succession (Kielland & Bryant 1998). Recruitment of trees in mature forest typically depends on the formation of gaps in the tree canopy (e.g. Runkle 1981; Bobiec 2007). Most tree species profit from the enhanced light availability and show higher growth and recruitment inside gaps. However, enhanced light levels may also increase herbaceous vegetation cover leading to increased competition with tree seedlings hampering seedling growth or establishment (Modry, Hubeny & Rejsek 2004; Van den Berghe et al. 2006). When light is not limiting, soil productivity may be an important factor determining sapling growth (Sipe & Bazzaz 1995; Lusk & Matus 2000), potentially changing the competitive balance between regenerating tree species (Latham 1992; Modry, Hubeny & Rejsek 2004). Next to abiotic conditions, foraging by large herbivores can top-down regulate both recruitment rate and species composition of recruiting trees in temperate forests (Ammer 1996; Van Hees, Kuiters & Slim 1996; Kriebitzsch et al. 2000; Scott et al. 2000). In the long-term, this browsing may prevent successful regeneration of some species into the tree canopy and alter the structure and dynamics of forest ecosystems (Mladenoff & Stearns 1993; Long, Pendergast & Carson 2007). Recent studies from temperate North-American forest systems illustrate the dominant role of top-down factors. Reintroduction of large predators in Yellowstone National Park led to cascading effects, modifying plant–herbivore interactions (Ripple & Larsen 2000). This suggests that herbivores are top-down regulating tree regeneration in the absence of predators, but that presence of predators may release trees from this control (Ripple & Beschta 2007).

Clearly, it is difficult to synthesize general effects of both bottom-up and top-down factors on tree dynamics, since various studies result in quite different outcomes. In addition to differences in productivity and ungulate density between systems, one reason for this is that biotic and abiotic factors may strongly interact. Light conditions may affect forage availability but also alter forage chemical quality, both potentially determining forage patch choice of herbivores (Edenius 1993; Molvar, Bowyer & Van Ballenberghe 1993; Hartley et al. 1997). Forest gaps may, therefore, preferentially be visited by ungulates leading to an uneven distribution of ungulate browsing and their effects of browsing in a forest system (Kuijper et al. 2009). In addition, the abundance and effects of herbivores are predicted to change along gradients of productivity (Oksanen et al. 1981; Kuijper & Bakker 2005; Melis et al. 2009). Lastly, browsing or grazing may directly or indirectly change the competitive balance between plant species and hence change the interaction between tree seedlings and herbaceous vegetation (Riginos & Young 2007). These interactions suggest a strong context-dependence of top-down effects of herbivores on tree regeneration and a shifting importance of biotic and abiotic factors during different stages of tree regeneration.

Although several studies have addressed the relative importance of bottom-up versus top-down factors in structuring temperate forest communities, surprisingly little empirical data are available that show how biotic and abiotic conditions may interact and how this can modify the importance of top-down and bottom-up forces (Hunter & Price 1992). Empirical data are especially lacking from systems with limited human intervention and a complete set of natural processes. The lack of a complete guild of ungulate species occurring at natural densities, absence of large predators and rarity of natural tree recruitment without human manipulation can all be factors that conceal the role that top-down effects might play in natural, unmanaged forest ecosystems. In this study, we tested the interactions between herbivore top-down effects and abiotic bottom-up effects during early stages of tree regeneration in one of the last remaining natural, temperate, lowland forest systems in Europe, the Białowieża Primeval Forest in Poland. The strictly protected part of this forest still hosts the complete native variety in both flora and fauna and excludes any human intervention.

Materials and methods

Study site

The study site was the strictly protected zone of the Białowieża National Park (BNP) located in the centre of the Białowieża Primeval Forest (BPF). The BPF, situated in eastern Poland (52°45′ N, 23°50′ E) and western Belarus, is a large continuous forest composed of multispecies tree stands. The entire BPF covers 1450 km2, of which 600 km2 belong to Poland and the remaining 850 km2 to Belarus. In 1921, the best preserved central part of the BPF was proclaimed as the Strict Reserve, which later became BNP. At present, BNP consists of a 47.5-km2 area of strictly protected old-growth forest in which no human intervention has been allowed since 1921 and tourist access is only permitted with a guide. Before 1921, human impact on tree stand structure and composition was small or minimal (Jędrzejewska et al. 1997; Samojlik et al. 2007). In 1996, BNP was enlarged to 105.2 km2 and the strictly protected zone has been surrounded by the partially protected zone since then (see Fig. 2). The area outside BNP is managed for forestry purposes, and hence wood exploitation and regulation of ungulate numbers is taking place. In the strictly protected zone, old forest stands prevail. Young stands with dominate trees of 40 years or less account for 2.4% of the area. Mid-aged stands, dominated by trees aged 41–80 years, cover 16.3% and old stands (81–120 years) cover 41.5%. Very old stands, dominated by trees aged over 120 years, account for 39%. Only 0.8% of the area is lacking tree cover (Michalczuk 2001). The following main forest types can be distinguished (with % share of total area and dominant trees in brackets based on Michalczuk 2001): deciduous forest (54%, Quercus robur, Tilia cordata and Carpinus betulus), mixed deciduous forest (23%, Picea abies, Quercus robur, Tilia cordata and Carpinus betulus), black alder bog forest and streamside alder-ash forest (14%, Alnus glutinosa and Fraxinus excelsior), mixed coniferous forest (13%, Pinus sylvestris, Picea abies and Quercus robur) and coniferous forest (9%, Pinus sylvestris and Picea abies). These forest types are situated along an ecological gradient of soil richness and water availability. Coniferous forest can be found on the driest, nutrient-poor soils. Deciduous forest and streamside alder-ash forest occur at productive soils rich in organic matter but streamside alder-ash forest are inundated (at least part of the year). Mixed coniferous and mixed deciduous forests occupy transitional zones of the soil fertility gradient (Faliński 1986). The mean altitude of the strictly protected part of BNP is 158 m a.s.l. and the total altitudinal range is 23 m. Mean annual air temperature is 6.8 °C with the coldest month in January with on average −4.7 °C and the warmest month is July with 17.8 °C. Mean annual precipitation is 641 mm and mean annual snow cover lasts for 92 days.

Figure 2.

 Location of the study area in Poland, Europe (right panel) and map of paired exclosures–control plots (numbered from 1 to 30) in the strictly protected part (dark grey) of Białowieża National Park (light and dark grey). The surrounding Białowieża Primeval Forest is not indicated on the map.

Herbivores and their population densities

Another unique aspect of BPF is that it is one of the very few European lowland forests that still host the complete native ungulate and carnivore assemblages. Five ungulate species occur throughout the forest system. The most abundant species, both in numbers and crude biomass, is red deer (Cervus elaphus) with a winter density of 3.4 individuals km−2 in 1999 (Jędrzejewski et al. 2002) increasing to 6.0 individuals km−2 according to the last estimate in 2008 (T. Borowik, J. Borkowski & W. Jędrzejewski, unpubl. data). The second-most numerous ungulate is wild boar (Sus scrofa), of which annual density strongly depends on fluctuations in food availability; density varied from 3.4 individuals km−2 in 1999 (Jędrzejewski et al. 2002) to 5.4 individuals km−2 in 2008 (T. Borowik, J. Borkowski & W. Jędrzejewski, unpubl. data). Roe deer, Capreolus capreolus, were present with 2.4 individuals km−2 in the winter of 1999 (Jędrzejewski et al. 2002) and a similar density in 2008. In lower density occur European bison (Bison bonasus) with 0.49 individuals km−2 and moose (Alces alces) with 0.04–0.08 individuals km−2 during 1993–2008 (Jędrzejewska et al. 1997; Mysterud et al. 2007; T. Borowik, J. Borkowski & W. Jędrzejewski, unpubl. data). Since human intervention is prohibited in the strictly protected zone of BNP, the ungulates in that area have not been hunted for over 80 years. However, the regulation of ungulate numbers that occurs outside BNP may to some extent affect the observed numbers inside the protected area. Hunting activities could increase densities in BNP by displacement of individuals to the protected area during the hunting season. However, recent studies suggest that this probably only affects individuals in areas directly bordering the non-hunting areas (Tolon et al. 2009). Hunting outside BNP may also reduce population levels in the protected area if BNP would act as a source area. However, the observed stability of the home ranges of the dominant ungulate species (red deer: Kamler, Jędrzejewski & Jędrzejewska 2008; wild boar: T. Podgórski & W. Jędrzejewski unpubl. data) inside BNP suggests that there is not a large movement of individuals to the areas outside the protected zone. Concluding, we believe that current hunting practices outside BNP will not strongly influence the ungulate densities inside the protected area. Natural predators of these ungulates, wolf (Canis lupus) and lynx (Lynx lynx), are strictly protected since 1989 and are not hunted throughout Poland. In BPF, they occur with average densities around 0.01–0.05 individuals km−2 (wolf) and 0.01–0.03 individuals km−2 (lynx) (Schmidt et al. 2008).

Several medium-sized and small herbivores occur, although reliable figures on their abundance are scarce. Both brown hare (Lepus europaeus) and mountain hare (Lepus timidus) have been recorded but are very rare. The community of herbivorous–granivorous forest rodents is dominated by bank vole (Myodes glareolus) and yellow-necked mouse (Apodemus flavicollis), whereas species such as pine vole (Microtus subterraneus), field vole (Microtus agrestis) and root vole (Microtus oeconomus) occur in forests in lower densities (Jędrzejewska & Jędrzejewski 1998).

Experimental design

To measure the impact of top-down forces on tree regeneration, 30 paired exclosure–control plots (Fig. 1) were erected in BNP in July 2000 (Kweczlich & Miścicki 2004 following Reimoser & Suchant 1992). The location of each exclosure was chosen randomly by placing a grid of 460 cells of 100 × 1000 m over the map of the strictly protected zone, with the shorter side of each cell aligned according to 330° azimuth. Each intersection of the raster was given a unique number and exclosure locations were determined by randomly selecting intersection numbers (Kweczlich & Miścicki 2004). The minimal distance between two exclosures was c. 310 m. The resulting locations (Fig. 2) covered all main soil and forest types, ranging from mesic coniferous forest to wet deciduous forest (see Table 1). The number of exclosures per forest type reflects the occurrence of each forest type inside the strict reserve. Therefore, we see this set of exclosures as a random sample covering the natural variation which exists in this diverse forest.

Figure 1.

 Two examples of exclosures, illustrating the variation in the effect of ungulate exclusion after 7 years of experiment. The upper photo (exclosure 30) shows higher regeneration of maple (Acer platanoides) inside (on the right hand) compared to outside the exclosure, whereas, the photo below (exclosure 23) shows no effect of ungulate exclusion on tree regeneration. For locations of exclosures, see Fig. 2.

Table 1.   Description of habitat types in which exclosure–control pairs were established with the number of exclosure plots (and paired control plots) in each habitat. The main tree canopy species at the location of the exclosures in each habitat at the start of the experiment are indicated (Picea abies, Pinus sylvestris, Quercus robur, Carpinus betulus, Tilia cordata, Acer platanoides, Populus tremula, Fraxinus excelsior, Betula pendula in coniferous and Betula pubescens in alder-ash forest)
Forest typeHydrological conditionsDominating tree canopy speciesNumber of exclosures
ConiferousWetPicea, Pinus2
Mixed coniferousMesic/wetPicea, Pinus, Quercus, Betula2
Mixed deciduousWet/mesicCarpinus, Quercus, Tilia, Picea, Populus3
DeciduousMesicCarpinus, Quercus, Tilia, Picea, Acer13
DeciduousWetCarpinus, Quercus, Tilia, Picea, Acer, Fraxinus8
Streamside alder-ash forestWetAlnus, Betula, Fraxinus1

Each exclosure consisted of a 2-m-high fence surrounding the 7 × 7 m sample plots plus a 50-cm buffer zone along the borders of the sample plot in which no measurements were taken to prevent potential fence effects. Fences were constructed of mesh wire with a mesh size of c. 15 cm (see Fig. 1). This fence effectively excluded all ungulate species (roe deer, red deer, moose, bison and wild boar) from entering, while allowing all small (rodents) and medium-sized herbivores (hares) free access. For each exclosure, we set-up a paired control plot (unfenced) of 7 × 7 m as close as possible to the exclosure with comparable habitat characteristics at the start of the experiment, such as canopy tree species composition, forest type, tree sapling density, herbaceous vegetation cover and canopy openness. Control plots were situated on average 20 m away from their paired exclosure plot with a minimum distance of 5 m. Metal sticks marked the corners of the plots to ensure consecutive measurements were taken at the same location. The surveyed area in each plot (both control and exclosure) was 49 m2. In 2007, one of the exclosures was destroyed by a fallen tree covering the entire plot. Hence, we excluded this plot from the analyses resulting in = 29.

Measurements of tree regeneration

During July–August 2000, 2002, 2004 and 2007, we recorded all trees occurring in the plots. The first inventory was carried out just after the exclosures were built. During each inventory, we counted the number of trees separately for each species. In addition, individuals per species were assigned to eight different size classes: seedlings (younger than 1 year, no woody stem), tree saplings < 10, 11–25, 26–50, 51–75, 76–100, 101–130, 131–200 and > 200 cm tall.

To compare species diversity between treatments, taking the relative abundances of species into account, we calculated the Shannon index (Pielou 1975) for each plot for different size classes according to:


where S is the number of species and p the proportion of species i of all tree individuals per height class. The minimum value of the Shannon index is 0 when there is only one species present, and the index is highest when each species is present in equal proportions. Low values indicate a high dominance of a single species, either due to low number of species or relatively high abundance of one species. We determined Shannon’s diversity for the 2000 data at the start of the experiment and for the 2007 data after 7 years of potential ungulate impact.

Measurements of abiotic factors at each location

To test for possible bottom-up effects on tree regeneration, we measured those abiotic factors that have been illustrated by previous studies to be important for recruitment of trees. Canopy cover determines the amount of available light on the forest floor and is the prime factor shaping tree recruitment (Runkle 1981; Bobiec 2007). Herbaceous vegetation can compete with tree seedlings, retarding their growth and establishment (Modry, Hubeny & Rejsek 2004; Van den Berghe et al. 2006), whereas soil fertility and wetness determine growth of established seedlings and hence recruitment rate in larger size classes (Sipe & Bazzaz 1995; Lusk & Matus 2000).

As a measure of light availability at each plot, we visually estimated tree canopy cover during each inventory (using 5% classes) and averaged the values per plot. Canopy cover in general was relatively high in both controls (mean ± SE: 73.0 ± 4.1%) and exclosures (72.5 ± 4.7%), typical for this old-growth forest, and ranged in both treatments from 5% to 100%. In addition, we estimated the total ground cover of bushes, herbs, grasses and ferns at each plot to the nearest 5%. The values of ground vegetation cover ranged from 1% to 96% inside controls (mean 51.9 ± 4.4%) and exclosures (59.4 ± 3.5%) during the study period. Both measures were recorded at the same time as the tree measurements in July–August, i.e. at time of peak vegetation biomass. These measurements were conducted at each plot (controls as well as exclosures) because both factors could vary considerably at small scale (metres). Moreover, although at the start of the experiment control plots were selected to have similar canopy and ground cover to those in exclosure plots, both factors changed quickly over time and in some cases led to considerable differences between paired plots.

Data on soil fertility were extracted from a map of soil types in the strict reserve of BNP (Prusinkiewicz & Michalczuk 1998). Plots occurred on nine main soil types (with number of exclosures in brackets): cambisols (7), phaeosols (6), luvisols (5), epigleysols (3), podzolsols (3), gleysols (2), cambiarenosols (1), gleypodzolsols (1) and eutrohistosols (1). These soil types were further aggregated in two categories indicating habitat productivity based on Prusinkiewicz & Michalczuk (1998): eutrophic soils (cambisols, phaeosols, luvisols and eutrohistosols) and oligotrophic soils (epigleysols, podzolsols, gleysols, cambiarenosols, gleypodzolsols). As an additional factor influencing soil productivity, we recorded soil wetness according to two categories: mesic (15 exclosures) and wet (14 exclosures). Values for both soil fertility and soil wetness were the same for paired exclosure and control plots due to their proximity.

Statistical analyses

We aggregated tree height classes into three classes: seedlings, saplings ≤ 50 cm and saplings > 50 cm. We choose the cut-off value of 50 cm because red deer have a preferred foraging height around 50–100 cm (Renaud, Verheyden-Tixier & Dumont 2003) and hence size classes above and below 50 cm were expected to show a differential response to ungulate exclusion. As the number of seedlings showed large year-to-year variation, we calculated the cumulative amount of seedlings by adding up the counted seedlings on each plot in 2000, 2002, 2004 and 2007. We chose to analyse the cumulative number of seedlings because we were not interested in showing how fluctuations in seedlings affected regeneration of different species. Instead, we wanted to show how bottom-up versus top-down factors influenced the net seedling input over the whole 7-year study period, as part of the forest regeneration process.

We performed all below-described analyses for each tree size class separately. First, we used paired-sample t-tests to see if species richness and the number of tree seedlings and saplings were similar in exclosure and control plots at the start of the experiment in 2000. We then tested for effects of top-down and bottom-up factors on the regeneration of trees in the different height classes over our 7-year study period. Because all our tree regeneration data represented count data (number of trees per size class), we used generalized linear models with a Poisson error distribution and log-link function. We used a quasipoisson distribution in all models to correct for overdispersion because the residual deviance was much larger than the residual degrees of freedom in all cases (> 10, Crawley 2007). As we were interested in the effect of herbivory and its potential interaction with abiotic conditions, we first tested for an overall effect of exclosures on the number of woody individuals. Then, we tested whether the effect of exclosure depended on the environmental factors (soil fertility and soil wetness) and covariates (canopy cover and herbaceous vegetation cover averaged over the study period). As we did not find any significant interactions between exclosure and environmental treatments, we continued with testing for effects of environmental variables without including the exclosure effect.

Soil fertility and wetness did not vary between paired exclosure and control plots, while canopy and ground cover did because they varied at much finer spatial scale. Therefore, we tested for an effect of factors (soil fertility and soil wetness) and covariates (canopy cover and vegetation cover) in separate models. First, we tested for an effect of the covariates using all our plots as replicates (= 58). Secondly, we tested for an effect of the factors using only the control plots (= 29). For the latter, we could not use all plots because paired control and exclosure plots always occurred on the same soil type, hence representing pseudoreplicates. We then repeated the test with the exclosure plots to check the robustness of the result.

While testing for effects of environmental variables, we started with full models (including interactions) and used likelihood ratio tests to simplify the full models to find the most parsimonious model (Crawley 2007). Although technically quasi-likelihoods (as used in our models) are not real likelihoods, the use of F-test-based likelihood ratio tests is considered appropriate to simplify quasipoisson models (Crawley 2002; Venables & Ripley 2002; Bolker 2008). For the covariates, we also considered models with quadratic terms of the covariates to test, whether possible effects were nonlinear rather than linear. In the Results section, we present statistics for the most parsimonious models only. We used r version 2.7.1 for these analyses (R Development Core Team 2008).

Differences between exclosures and controls in species composition of trees and Shannon-diversity index were tested pair-wise for cumulative seedlings, saplings < 50 cm and saplings > 50 cm separately using nonparametric Wilcoxon signed-rank tests. All percentage data (canopy cover and vegetation cover) were arcsine-transformed prior to analyses (Zar 1984). These statistical analyses were performed using spss 15.0 (SPSS Inc., Chicago, Illinois, USA).


Effects of ungulates and abiotic factors on tree regeneration

The density of trees in three size classes, the number of species and Shannon’s diversity indices were similar in exclosures and control plots at the start of the experiment in 2000. After 7 years, exclosures did not show a higher cumulative number of seedlings than control plots (Table 2). Moreover, we did not find any interaction between exclosure and environmental factors (F1,54 = 0.55, = 0.46 for all factors, Table 3). The most parsimonious model explaining the variation in cumulative number of seedlings among the plots and including the covariates, only showed vegetation cover as explanatory variable with its quadratic term (F1,55 = 20.56, < 0.001). The density of seedlings increased with vegetation cover until an optimum of 7.2 seedlings m−2 at 38% vegetation cover, and decreased with increasing vegetation cover (Fig. 3). Soil fertility and wetness did not affect the density of seedlings in the control plots (soil fertility: F1,27 = 2.38, = 0.13; soil wetness: F1,27 = 2.37, = 0.13) or exclosure plots (soil fertility: F1,27 = 0.61, = 0.44; soil wetness: F1,27 = 0.59, = 0.45).

Table 2.   Density (individuals m−2 on 49-m2 plots), number of species and Shannon index of species diversity (mean ± SE) of trees in different size classes in the control and exclosure plots at the start of the experiment (2000) and after 7 years (2007). Seedlings in 2000 refer to seedlings present in that year, whereas seedlings in 2007 refer to the cumulative number counted during surveys in 2000, 2002, 2004 and 2007. Sample sizes differ because Shannon indices could only be calculated when trees were present on a plot. Differences between exclosures and controls were tested by Wilcoxon signed-rank tests. Significant differences are indicated in bold
Size classSurvey in 2000
 Seedlings0.49 ± 0.200.44 ± 0.18−0.31290.758
 Saplings ≤ 50 cm1.59 ± 0.391.43 ± 0.31−0.38290.707
 Saplings > 50 cm0.07 ± 0.010.12 ± 0.05−0.60290.550
Number of species present
 Seedlings1.65 ± 0.211.76 ± 0.18−0.43290.665
 Saplings ≤ 50 cm3.72 ± 0.293.51 ± 0.31−0.70290.484
 Saplings > 50 cm0.86 ± 0.181.10 ± 0.17−1.01290.312
Shannon index
 Seedlings0.48 ± 0.070.29 ± 0.07−1.85220.064
 Saplings ≤ 50 cm0.79 ± 0.080.76 ± 0.08−0.41280.683
 Saplings > 50 cm0.21 ± 0.090.28 ± 0.08−0.25110.799
 Survey in 2007
 Seedlings3.89 ± 1.083.79 ± 0.77−0.65290.516
 Saplings ≤ 50 cm1.46 ± 0.341.14 ± 0.24−0.42290.673
 Saplings > 50 cm0.21 ± 0.050.69 ± 0.18−3.32290.001
Number of species present
 Seedlings3.90 ± 0.283.93 ± 0.27−0.06290.960
 Saplings ≤ 50 cm3.72 ± 0.344.00 ± 0.28−0.74290.458
 Saplings > 50 cm1.72 ± 0.193.07 ± 0.30−3.37290.001
Shannon index
 Seedlings0.79 ± 0.080.62 ± 0.08−2.04270.041
 Saplings ≤ 50 cm0.77 ± 0.080.89 ± 0.07−1.25280.210
 Saplings > 50 cm0.45 ± 0.060.72 ± 0.08−2.02230.044
Table 3.   Results of quasipoisson regression models that tested for the effect of exclosure or exclosure and one of four environmental factors (soil fertility, soil wetness, canopy cover and herbaceous vegetation cover) on the number of individuals from three tree size classes. Only the interaction effect is shown from the analysis of the full model, i.e. including main effects of the factors in the interaction term. Significant effects are indicated in bold
Size classModelsd.f.Residual d.f.FP-value
SeedlingsExclosure156< 0.010.95
Exclosure × Soil fertility1540.550.46
Exclosure × Wetness1540.510.48
Exclosure × Canopy cover154< 0.010.97
Exclosure × Vegetation cover1540.030.86
Saplings ≤ 50 cmExclosure1560.600.44
Exclosure × Soil fertility1540.130.72
Exclosure × Wetness1540.080.77
Exclosure × Canopy cover1540.010.93
Exclosure × Vegetation cover1540.220.64
Saplings > 50 cmExclosure15610.20< 0.01
Exclosure × Soil fertility1540.140.71
Exclosure × Wetness1540.230.63
Exclosure × Canopy cover1540.090.77
Exclosure × Vegetation cover154< 0.010.97
Figure 3.

 The relationship between vegetation cover and cumulative density of seedlings (in 2000–07) in control and exclosure plots. The parabolic line shows the fit of the most parsimonious model (ground cover: F1,56 = 8.15, = 0.006; vegetation cover2: F1,55 = 20.56, < 0.001) with the following parameter estimates: intercept = 0.083 ± 0.73, t = 5.58, < 0.001; effect of vegetation cover = 0.099 ± 0.032, t = 3.09, = 0.0031; effect of vegetation cover2 =−0.0013 ± 0.00034, t = −3.80, < 0.001 (Log(y) = 0.083 + 0.099x − 0.0013x2).

Exclusion of ungulates also did not affect the density of tree saplings shorter than 50 cm, and this effect did not depend on the environmental factors (effect of exclosure: F1,56 = 0.60, = 0.44; see Table 2 for interaction effects). Canopy and vegetation cover did not affect the density of saplings ≤ 50 cm in the plots (ground cover: F1,56 = 0.23, = 0.64; canopy cover: F1,56 = 1.10, = 0.30). While soil wetness did not affect the density of saplings in control plots (F1,27 = 0.29, = 0.60) or exclosures (F1,27 = 0.022, = 0.88), soil fertility did. The density of saplings ≤ 50 cm was higher on eutrophic soils than on oligotrophic soils in control (F1,27 = 3.61, = 0.0068) as well as exclosure plots (F1,27 = 7.93, = 0.0090). Density of saplings was on average three times higher on eutrophic soils (Fig. 4).

Figure 4.

 Density of tree saplings (N m−2) shorter than 50 cm on eutrophic versus oligotrophic soils for the control and the exclosure plots in 2007. Significant differences (< 0.01) are indicated by asterisks.

Herbivory clearly affected the density of tree saplings taller than 50 cm (F1,56 = 10.20, = 0.0023) and this effect did not depend on environmental conditions (see Table 3 for interactions). The density of saplings that grew into the > 50 cm size class was more than three times higher in the exclosures than in the control plots during 7 years (Table 2). Canopy cover (F1,56 = 2.05, = 0.16) or vegetation cover (F1,56 = 0.95, = 0.33) had no influence on the density of saplings taller than 50 cm and neither soil fertility nor soil wetness did influence density in control (fertility: F1,27 = 0.064, = 0.80; wetness: F1,27 = 0.23, = 0.88) or exclosure plots (fertility: F1,27 = 0.14, = 0.71; wetness: F1,27 = 0.84, = 0.37).

Effects of herbivory on species composition of regenerating trees

After 7 years, hornbeam, Carpinus betulus made up the largest proportion of cumulative seedlings but in general the species composition of seedlings was similar in exclosure and control plots (Fig. 5a). Only one species, Tilia cordata, had a significantly higher abundance inside control plots than in exclosures (z = −2.945, = 29, = 0.003). The total number of species encountered in the seedling group did not differ between exclosures and controls. The Shannon index showed a small but significant difference and was higher in controls than in exclosures (Table 2).

Figure 5.

 Species composition of (a) tree seedlings (cumulative over 2000–07), (b) saplings ≤ 50 cm and (c) saplings > 50 cm inside exclosures and control plots in 2007 (mean density ± SE). The order of tree species in the graphs is the same as in the legend.

Unlike the seedling community, the community of tree saplings ≤ 50 cm was not dominated by a single species in 2007 (Fig. 5b). Overall, exclosures did not show consistent effects on species abundances of saplings ≤ 50 cm. Occurrence of only two species significantly differed between exclosures and controls. Quercus robur was found more often inside exclosures (z = −2.454, = 29, = 0.014), whereas Fraxinus excelsior attained higher numbers in controls (z = −2.189, = 29, = 0.029). The mean total number of species of tree saplings ≤ 50 cm and the Shannon-diversity index did not differ between exclosure and control plots (Table 2).

After 7 years, all species showed on average a higher density of tree saplings taller than 50 cm inside exclosures than in controls (Fig. 5c). This difference was significant for Acer platanoides (z = −2.536, = 29, = 0.011), Sorbus aucuparia (z = −2.609, = 29, = 0.009) and marginally significant for Ulmus glabra (z = −1.904, = 29, = 0.057). Pinus sylvestris, although rare, was found only inside exclosures. The effect of herbivory was also reflected in the number of tree species present. While species richness in exclosures and controls was similar at the start of the experiment, a significantly higher number of species had recruited into the > 50 cm size class inside exclosures (3.1 ± 0.3) than in the control plots (1.7 ± 0.2) after 7 years of excluding ungulates (Table 2). Due to the recruitment of more tree species and the more even distribution of the species composition of the recruiting trees in 2007, Shannon indices were significantly higher inside exclosures (0.72 ± 0.08) compared to control plots (0.45 ± 0.06, Table 2). In the presence of ungulates, recruitment of tree saplings into the > 50 cm size class was strongly dominated by one species, Carpinus betulus, which on average made up 68% of trees > 50 cm in control plots, compared to 38% in exclosures.


The present study shows that the relative importance of top-down biotic and bottom-up abiotic factors changed during different stages of tree regeneration in an old-growth forest with a complete native ungulate and carnivore assemblage. Whereas biotic (vegetation cover) and abiotic factors (soil fertility) determined variation in the numbers of seedlings and tree saplings ≤ 50 cm, top-down effects of ungulates controlled recruitment into taller size classes. The strength of these top-down effects differed among tree species and resulted in a shift in species composition of regenerating trees in the presence of ungulates compared to plots, where ungulates were excluded. In general, species diversity was lower in the presence of ungulates, because of a higher dominance of one species, Carpinus betulus.

Shifting importance of top-down and bottom-up factors affect tree recruitment

In the present study, we aimed at studying the interacting effects between biotic and abiotic factors. On the one hand, abiotic factors have been shown to be important in determining tree regeneration (Runkle 1981; Sipe & Bazzaz 1995; Modry, Hubeny & Rejsek 2004; Van den Berghe et al. 2006; Bobiec 2007). On the other hand, they influence forage availability and quality for herbivores (Edenius 1993; Molvar, Bowyer & Van Ballenberghe 1993; Hartley et al. 1997), which is likely to result in interacting effects. In contrast to our expectation, we did not find interacting effects of abiotic and biotic (herbivory) factors. Instead, a single dominant factor determined recruitment at different stages of tree regeneration. Abundance of seedlings was reduced at high herbaceous vegetation cover, suggesting competitive effects of vegetation on seedling establishment, which is in line with other studies (Modry, Hubeny & Rejsek 2004; Van den Berghe et al. 2006). Interestingly, the number of seedlings showed an optimum at intermediate herbaceous vegetation cover. Lower seedling abundance at low vegetation cover could be related to low light or nutrient availability that limit growth of herbaceous vegetation and establishment of seedlings. However, we did not find any effect of canopy cover and soil fertility on seedling density, suggesting that light and nutrient availability per se are not the most important factors explaining the low abundance of seedlings at low vegetation cover. We suggest that, alternatively, this pattern could be explained by the process of associative resistance, where tree seedlings are protected from browsing because they grow among unpalatable herbaceous cover (Smit, Ouden den & Müller-Schärer 2006). Hence, a minimum amount of herbaceous vegetation cover would be needed to provide safe sites for tree seedling establishment, which could explain the low seedling abundance at low vegetation cover in our study. Whereas several studies have illustrated the importance of gap formation in the regeneration process (e.g. Runkle 1981; Bobiec 2007), we did not find that canopy cover significantly affected the density of seedlings or saplings. In contrast, the effect of canopy cover was overruled by other factors at all stages of the tree regeneration process that we studied. In the case of seedlings, the most influential factor (vegetation cover) is only indirectly related to canopy cover, and the most influential factor for saplings ≤ 50 cm (soil fertility) is not related to canopy cover in our study system. The absence of a relation between canopy cover and recruitment into the larger size classes (saplings > 50 cm) can for a large part be explained by the relatively closed canopy observed in all our study plots, which is typical for old-growth forest. Additionally, selective foraging of ungulates inside forest gaps compared to closed forest (Kuijper et al. 2009) may repress enhanced tree regeneration inside forest gaps.

The effects of ungulates on total tree recruitment depended on the size class of trees. Neither did they affect the cumulative number of seedlings throughout the study period nor the number of trees ≤ 50 cm, indicating that the input of seedlings and saplings in the tree regeneration process is not different in the presence or absence of ungulates. The lack of effects of ungulates on seedlings and trees in the smallest size class is in contrast to results from studies in other temperate forest systems which illustrated that ungulate browsing can strongly reduce seedling density (Kumar & Shibata 2007; Long, Pendergast & Carson 2007; Olesen & Madsen 2008). This may first of all be related to the large differences in ungulate density between these areas and our study area. The densities of the dominant browser in these systems were 2- to 13-fold higher than those in the present study system (17.5–39.5 sika deer km−2: Kumar & Shibata 2007; 12 white-tailed deer km−2: Long, Pendergast & Carson 2007; 24 roe deer km−2: Olesen & Madsen 2008). These large differences may reflect the dissimilarity in hunting management, presence of predators or productivity between their and our study systems. A second important factor explaining the observed differences in effects of ungulates may be related to structural diversity of the forest in different study areas. Naturally developed forests, as in the present study, are characterized by rich undergrowth and a multi-layered tree structure, in contrast to managed forests (Jędrzejewska et al. 1994). Hence, food for browsing ungulates is abundant at different height classes in natural forests, whereas it may comprise only certain height classes in more even-aged, managed forest systems. As the preferred foraging height of red deer, the most abundant ungulate species in our system, is 50–150 cm (Renaud, Verheyden-Tixier & Dumont 2003; S. Miścicki, unpubl. data), a higher food availability at this preferred forage height may have prevented intense foraging on trees at lower, less preferred heights.

In the present study, ungulates played a dominant role in affecting tree recruitment only into size classes > 50 cm. Within this size class, more than three times higher numbers of trees occurred inside exclosures compared to controls after 7 years. Therefore, the presence of ungulates results in a ‘herbivore trap’ at about 50 cm from which many trees cannot escape. We suggest that this ‘herbivore trap’ might be functionally similar to the ‘fire trap’ that was shown in studies from African savanna systems (Skowno et al. 1999). The regular occurrence of fires can prevent trees from regenerating into taller size classes in which they would be resistant to fire (Skowno et al. 1999). Constant browsing by ungulates may act in a similar way, in which small but relatively old trees are being trapped at a certain height. These suppressed individuals, called ‘gullivers’ by Bond & van Wilgen (1996), are only able to escape from this herbivore trap once they are released from herbivore control caused by a population crash of browsing ungulates (Prins & Van der Jeugd 1993). This illustrates the important top-down effects that ungulates can have in natural temperate forest systems in determining the recruitment of trees and ultimately may affect species composition and age structure of the tree stand.

Ungulate impacts on tree species composition

In the present study, besides influencing total recruitment, ungulates affected species composition of trees recruiting into the > 50 cm size classes, whereas they did not affect species composition in either the seedling or sapling (≤ 50 cm) size class. Species composition of trees taller than 50 cm was affected by ungulates in two different ways. On the one hand, the number of tree species increased inside exclosures. This difference could not be explained by higher numbers of species which exclusively occurred inside exclosures, as Pinus sylvestris was only found inside exclosures. All other species were found both in exclosures and controls, albeit in lower numbers per species and in a lower number of plots in the controls. Hence, the difference in the number of species is mainly the result of simultaneous occurrence of several species inside exclosures versus recruitment of a single species in controls. On the other hand, tree species occurred in more equal proportions inside exclosures compared to the controls, as indicated by the higher Shannon index. Whereas Carpinus betulus strongly dominated the control plots, comprising more than 68% of all present trees, its dominance was reduced to 38% inside exclosures due to higher recruitment of other species. The impact of ungulates on tree species composition shows that apart from an overall reduction of recruitment, the ‘herbivore trap’ acts as a filter, which only some species are able to pass so that they can successfully regenerate. This results in a higher dominance of only few or a single species in the presence of ungulates, which has also been observed in other studies carried out in temperate forest systems (Ammer 1996; Kriebitzsch et al. 2000; Find’o & Petráš 2007). Such a shift in species composition of regenerating trees may affect forest communities and alter species composition (Kwiatkowska & Wyszomirski 1990).

The higher dominance of Carpinus in the presence of ungulates is interesting as Carpinus constitutes an important food plant in the diet of the three most common ungulates (red deer, roe deer and bison) in the system (Gębczyńska 1980; Gębczyńska, Gębczyński & Martynowicz 1991). In addition, of all trees present in BPF Carpinus was shown to have the highest proportion of individuals in the size class up to 130 cm with its leader shoot browsed (Miścicki 1996; B. Brzeziecki, unpubl. data). This indicates the strong selective foraging of ungulates on this species compared to other species. Hence, the dominance of Carpinus in the recruitment in control plots is not the result of avoidance by ungulates. Other studies from temperate forest systems showed that by selective foraging ungulates may differentially affect the recruitment of tree species, either leading to a reduction (Horsley, Stout & DeCalesta 2003; Modry, Hubeny & Rejsek 2004; Long, Pendergast & Carson 2007) or an increase (Tilghman 1989; Van Hees, Kuiters & Slim 1996) in relative abundance of the preferred forage species. A possible mechanism for increasing abundance of preferred species in the presence of herbivores is that preference coincides with a high browsing-tolerance (Augustine & McNaughton 1998). The observed patterns suggest that Carpinus is one of the most browsing-tolerant species in the studied system and can regenerate despite high browsing pressure. This might be a result of ‘apparent competition’, where a high tolerance to browsing together with a high preference by ungulates favours Carpinus as long as other potentially competing species are also browsed but show lower tolerance (see also Kriebitzsch et al. 2000).

How the observed effects of herbivory on the recruitment and species composition of relatively small trees translate into later stages and finally into the structure of mature stands may not be straightforward (Mladenoff & Stearns 1993; Woodward et al. 1994). Population dynamics of trees are shaped by inter- and intraspecific competition after the sapling stage (Hulme 1996). However, the above-mentioned patterns in species composition are supported by the observed long-term changes (1936–2002) in tree recruitment rates in BPF in relation to changes in ungulate density (D. P. J. Kuijper et al., unpubl. data) and the observed changes in tree canopy composition in the same period (Bernadzki et al. 1998). Periods of several years with low ungulate density resulted in high recruitment of virtually all tree species. However, Carpinus was the only species that showed high regeneration under levels of high ungulate density. This suggests that the effects of ungulates during early life stages carry over on the mature tree stands, and they do play an important role in tree canopy composition.

Biotic and abiotic interactions shape ungulate top-down effects

Most studies on the impact of ungulates in temperate forest systems have been carried out in single-ungulate-species-dominated systems (see for example Ammer 1996; Van Hees, Kuiters & Slim 1996; Kriebitzsch et al. 2000; Scott et al. 2000; Long, Pendergast & Carson 2007) mainly because most systems have a greatly impoverished faunal community. Interactions between multiple ungulate species that co-occur in an area can alter their net effects on the plant community (Gordon 1988; Ritchie & Olff 1997; Latham et al. 1999; Young, Palmer & Gadd 2005). In addition, the presence of carnivores can influence behaviour, habitat choice and spatial distribution of ungulate species (Creel et al. 2005; Fortin et al. 2005; Frair et al. 2005). Recent studies suggest that these indirect (non-lethal) effects of carnivores may be as important (Schmitz, Beckerman & O’Brien 1997) or even more important in influencing herbivore–plant interactions as the release form herbivore top-down control by population reduction per se (Oksanen et al. 1981; Creel & Christianson 2008). The present study illustrates the context-dependence of herbivore top-down effects. It thus shows that besides biotic factors, abiotic bottom-up factors can shape herbivore top-down effects. Heterogeneity in abiotic and biotic conditions may, therefore, have an important influence on the strength of top-down effects and the role that herbivores play in natural ecosystems (see also Gilliam 2006). In most managed forest systems, many natural components, such as diverse tree species composition, undisturbed forest regeneration, heterogeneous soil conditions, multiple large ungulate species or presence of predators are missing. In such more homogeneous environments, one dominant factor is expected to prevail, leading to a directional selection pressure favouring only a few tree species during the stages of the regeneration process. The relative lack of heterogeneity and stochasticity in managed forest systems may prevent the alternating effects of biotic and abiotic factors influencing tree recruitment observed in the present study. These alternating effects may warrant that most tree species do show successful recruitment, albeit at a lower rate, in the presence of ungulates.

There has been a recent paradigm shift in forest management towards a more ecosystem-oriented approach (Puettmann & Ammer 2007). However, knowledge on forest ecosystem functioning is scarce and often based on highly managed forest systems. Our study area comprises the least disturbed and most complete temperate forest system in lowland Europe (Jędrzejewska et al. 1994). Hence, it offers insight into the role played by ungulates in an intact old-growth forest and can serve as a reference for other forest systems in the temperate region.


This study was financed by the grant 5P06H 03418 from the Ministry of Science and Higher Education, Poland. The work of D.P.J.K. and J.P.G.M.C. has been supported by a Marie Curie Transfer of Knowledge Fellowship BIORESC of the European Community’s Sixth Framework Programme under contract number MTKD-CT-2005-029957. We are very thankful to several people who helped with data collection: Małgorzata Piotrowska, Renata Czarnecka, Edyta Nowicka, Dariusz Mróz, Matt Hayward, Anne Velenturf, Basia Bańka. We thank Roman Kozak for his technical help in maintaining the exclosures.