Urban Forestry & Urban Greening 13 (2014) 869–877

Contents lists available at ScienceDirect

Urban Forestry & Urban Greening journal homepage: www.elsevier.com/locate/ufug

Fragmentation and recreational use affect tree regeneration in urban forests Susanna Lehvävirta a,b,∗ , Ferenc Vilisics b , Leena Hamberg c , Minna Malmivaara-Lämsä c , D. Johan Kotze b a

Botanic Garden and Herbarium, Finnish Museum of Natural History, University of Helsinki, PO Box 7, FI-00014 Helsinki, Finland Department of Environmental Sciences, University of Helsinki, PO Box 65, FI-00014 Helsinki, Finland c Finnish Forest Research Institute, PO Box 18, FI-01301 Vantaa, Finland b

a r t i c l e

i n f o

Keywords: Boreal forest Edge effects Forest dynamics Recreation ecology Spatial structure Tree saplings

a b s t r a c t The aim of this study was to test whether fragmentation or recreational use affect tree regeneration in urban forests, and to quantify these effects. We sampled tree saplings at different distances from edges in spruce (Picea abies) dominated forests, and at different distances from paths that represented different levels of wear. Generalized linear mixed models were used to test our hypotheses. We found that fragmentation favours the regeneration of deciduous trees in urban spruce dominated forests: distance from the edge had a pronounced effect on regeneration, at least up to 80 m into the forests. Saplings of Betula pendula, Populus tremula, other deciduous species and Pinus sylvestris benefited from edge conditions. Betula pubescens saplings, however, were most abundant in the interior and small Sorbus aucuparia saplings at 25–30 m from the edge. All species suffered from the direct effects of trampling, while varying responses of species to distance from the paths were observed up to 6 m, and possibly further. As trees essentially define the living conditions for other forest species, we suggest that the spatial extent of edge and trampling effects should be studied for different types of forests. This knowledge should then be used in urban forestry and planning to define the threshold value that will allow for at least some “intact” interior. We suggest a diameter larger than 160 m to support indigenous species in boreal spruce dominated forests. © 2014 Elsevier GmbH. All rights reserved.

Introduction It has been argued that natural regeneration dynamics should be allowed in urban forests (Greene et al., 1999; Lehvävirta, 2007), thus not only saving in management costs but also enhancing possibilities for education, research, biodiversity and recreational value, amongst others. In urban environments, however, natural dynamics are altered, which may affect the regeneration of trees (Tonnesen and Ebersole, 1997; Lehvävirta and Rita, 2002; Malmivaara-Lämsä et al., 2008b; Hamberg et al., 2009a). For example, urban forests are usually drier and sunnier due to their small size (DeWalle and McGuire, 1973; Matlack and Litvaitis, 1999), with an increase in temperature and soil nutrient content (Hamberg

∗ Corresponding author at: Botanic Garden and Herbarium, Finnish Museum of Natural History, University of Helsinki, PO Box 7, FI-00014 Helsinki, Finland. Tel.: +358 50 5762952. E-mail addresses: susanna.lehvavirta@helsinki.fi (S. Lehvävirta), [email protected] (F. Vilisics), leena.hamberg@metla.fi (L. Hamberg), [email protected] (M. Malmivaara-Lämsä), johan.kotze@helsinki.fi (D.J. Kotze). http://dx.doi.org/10.1016/j.ufug.2014.10.003 1618-8667/© 2014 Elsevier GmbH. All rights reserved.

et al., 2009b; O’Brien et al., 2012). This may favour drought- and heat tolerant species and those benefiting from high nutrient levels. In addition, urban forests are characterised by continuous and repeated anthropogenic factors. In this study we focus on trampling and fragmentation, as evidence suggests they are good candidates for scientific hypotheses about factors affecting tree regeneration in urban forest. Understanding their effects may essentially help in the successful planning and management of these forests. Fragmentation increases the proportion of edges resulting in smaller areas for forest interior conditions and consequent changes in vegetation (Bannerman, 1998; Hamberg et al., 2008, 2010; Vallet et al., 2010; Ranta et al., 2013). Forest edges experience high variation in daily temperature, decreased moisture and increased air particle levels due to solar radiation and wind (e.g. Chen et al., 1995; Bannerman, 1998; Weathers et al., 2001; Malmivaara-Lämsä et al., 2008a). Trampling in turn disturbs the forest floor and tree regeneration (Lehvävirta, 1999; Hamberg et al., 2008, 2010; Hauru et al., 2012). Residents use urban forests for a variety of activities, such as recreation, exercise, play and shortcuts, thus creating extensive path networks, reducing plant cover, and damaging tree roots

870

S. Lehvävirta et al. / Urban Forestry & Urban Greening 13 (2014) 869–877

(e.g. Cole, 1988, 1995; Lehvävirta, 1999; Littlemore and Barker, 2001; Malmivaara-Lämsä et al., 2008b). Trampling increases soil compaction and run-off (e.g. Liddle and Moore, 1974; Godefroid and Koedam, 2004), and may cause chemical changes – for example, soil pH and nutrient levels increased on and beside paths in Finnish urban forests (Malmivaara-Lämsä et al., 2008a). In sum, both the edge and trampling affect urban forests substantially. We studied whether these two factors also affect the regeneration of trees. We observed the distribution and abundance of saplings in forests in the cities of Helsinki and Lahti, southern Finland. We hypothesised that saplings of light-demanding broadleaf trees (Alnus incana L., Acer platanoides L., Betula pendula Roth., B. pubescens Ehrh., Populus tremula L., Prunus padus L., Quercus robur L. and Salix caprea L.), and the conifer Pinus sylvestris L. would be more abundant near edges than in forest interiors (e.g. Lähde et al., 1999; Lehvävirta and Rita, 2002; Bakker et al., 2004; Hamberg et al., 2008). Saplings of the shade-tolerant Picea abies (L.) Karst. (e.g. Leemans, 1991) were hypothesised to be more abundant in forest interiors than at edges. We also hypothesised no remarkable effects of forest edges on Sorbus aucuparia L. regeneration, as this species thrives in shade and light (Lehvävirta and Rita, 2002; Hamberg et al., 2008). Furthermore, we hypothesised that the direct effects of trampling will diminish the abundances of all saplings on and right next to paths, except for rowan that we expected to be more abundant along the edges of the paths (due to, e.g. soil fertility; Hamberg et al., 2008, 2009b) than further away, in the seemingly un-trampled vegetation. We used the distance from the edge and paths to determine the magnitude and spatial extent of edge and trampling effects on tree regeneration in urban forests, and found considerable effects of both. We quantified these effects in order to discuss the implications for planning and management. Materials and methods Study area and sampling Medium fertile (Myrtillus type) P. abies dominated forest sites were selected in 2003 in two cities of southern Finland: Helsinki (60◦ 10 15 N, 24◦ 56 15 E; 33 forests) and Lahti (60◦ 59 00 N, 25◦ 39 20 E; 19 forests), with 2792 and 669 inhabitants km−2 , respectively. The Helsinki (and Lahti) populations are 604 000 (103 000), of which 15.5% (15.4%) are 0–15 years old, 69.7% (63.4%) are 16–64 years, and 15.8% (21.2%) at least 65 years old. Helsinki (and Lahti) had 3775 (6270) ha of green areas and 216 (over 100) km of maintained recreational or cycling paths in 2012 (Lahden Viheralueohjelma, 2013–2025; Statistical Yearbook of Helsinki, 2013; Tilastotietoja Helsingistä, 2013; Lahti City Database, 2014). Helsinki is situated in the hemi-boreal and Lahti in the southern boreal vegetation zones, both including remnants of indigenous forests. Dominant tree species in the study forests included P. abies, P. sylvestris and B. pendula, and the main canopy consisted of trees more than 80 years old. We defined the edge as a line along the trunks of the outermost mature trees of a forest fragment. The study sites had a south to west exposure in order to maximize the edge effect (exposure to sun and the prevailing wind direction). In order to study responses in relation to the edge we only chose sites with no urban development for at least 10 years (i.e. the edges were older than that), and bordered asphalt roads (not busy highways) or residential constructions with impermeable surfaces within 20–30 m from the forest edge. Sites next to high buildings that would shade the edge, with recent cuttings, or on slopes, were not selected (see Hamberg et al., 2008, 2010 for a detailed description of the sampling design). Our method was that of stratified random sampling, i.e. we sampled randomly but made sure all conditions would be

sampled sufficiently. Spontaneous paths representing different levels of wear were selected at a variety of edge distances (0–108.5 m from the edge). In order to obtain a representative sample from the edge to the interior, we stratified the forest patches into 15 m wide zones, from the edge into the forest. Within these zones, we sampled paths of each level of wear (see below) that could be found within the zone. The sample plots were placed on paths (in the middle), next to (at 0.45 m) and further away (>1.2 m) from path edges. The ‘away’ plots were placed in seemingly un-trampled vegetation. The positioning of the plots along the paths was randomized. As the focus of this study was not on the maintained official recreational paths, the potential effect of them was controlled for by keeping a minimum distance of 10 meters to these paths. Four path wear classes were distinguished, following the classification of Lehvävirta (1999): Wear class L (low) represents minimum visible effects of trampling. Wear class M (moderate) represents visible effects of trampling with vegetation damaged and reduced in cover, but not completely worn out. At wear class H (heavy), vegetation is mostly absent, the humus layer is not completely worn out, but rocks and tree roots are sometimes uncovered. At wear class T (totally worn) bare mineral soil or a deeply worn humus layer is exposed, tree roots are often uncovered and no vegetation remains except for scanty individuals sheltered by tree roots and rocks. To summarize, we sampled on, next to and away from paths (replicated in each wear class), at varying distances from the edge. Plots next to and away from paths were 1 m × 2 m in size, while plots on the paths were 0.5 m × 2 m to stay on the paths. For the analyses, the number of saplings observed in the path plots was multiplied by 2. Altogether, we collected data from 321 sample plots in Helsinki and 249 in Lahti. All tree saplings in the sample plots were measured (including Rhamnus frangula L.), counted and classified into four size categories: size class 1 included individuals with a height of 30–79 cm, size class 2 saplings 80–129 cm in height, size class 3 saplings 0.5–2.0 cm in diameter at breast height (dbh), and size class 4 saplings 2.1–5.0 cm in dbh. Both S. aucuparia and P. tremula can spread clonally; their saplings were considered individuals if they were at least 50 cm apart and if no underground joint stem or root could be found. The nomenclature for all the sampled species follows the Field Flora of Finland (Hämet-Ahti et al., 1998). We also visually estimated the canopy cover and measured the dbh of trees (>5 cm by dbh) on concentric plots of 50 m2 around the sapling sample plots, to give an approximate description of the study sites at the sampling points. Hypotheses concerning the responses of individual species to the edge effects and trampling were based on literature (e.g. Ingelög et al., 1977; Hämet-Ahti et al., 1998; Reinikainen et al., 2000; Lehvävirta and Rita, 2002; Malmivaara et al., 2002; Hamberg et al., 2008). Before data analyses, the sampled species were divided into three classes with regard to edge and trampling tolerances: edge sensitive, indifferent, edge tolerant and very low, low, and moderate trampling tolerance (see Table 1). Statistical analyses We used generalized linear mixed models (GLMMs; the glmmPQL procedure of the MASS package in R) to test our a priori hypotheses about the effects of forest edges and trampling on the abundances of saplings (Venables and Ripley, 2002; R Core Team, 2012). The most frequent species were analysed individually (in different size categories when there were enough observations). Those of low frequency were analysed as a group of species with similar expected responses (Table 1) – the only group that was formed was the Edge Tolerant Group (ETG). R. frangula and Juniperus communis L. were present in the forests, but not included in the statistical

S. Lehvävirta et al. / Urban Forestry & Urban Greening 13 (2014) 869–877

871

Table 1 A priori hypothesised relative tolerances of the observed tree saplings to trampling and edge effect. Trampling Sensitive Edge

Sensitive Indifferent Tolerant

Low tolerance

Moderate tolerance

Rhamnus frangula Betula pubescens, Pinus sylvestris, Juniperus communis, Edge tolerant group (Acer platanoides, Alnus incana, Prunus padus, Quercus robur, Salix caprea)

Sorbus aucuparia Populus tremula

Picea abies Betula pendula

analyses. R. frangula, being the only species in its category, was not analysed due to low numbers. J. communis, a conifer hypothesised to be edge tolerant, was also low in number but not included in the ETG that consisted only of deciduous species. If a species or species group was not observed in a particular forest site, the site was excluded from that analysis because a zero occurrence in a particular forest patch could be due to a stand level effect. We were interested in the edge and trampling effects within a forest stand, not in the stand effect per se. The GLMMs were constructed as follows. The response variables were the number of saplings (of a certain species and size category or possibly grouped together as described above) per sample plot. We used a negative binomial distribution (see O’Hara and Kotze, 2010), with a log-link function to estimate the models. We included one random factor (forest site) and four fixed effect variables in the models: (a) linear, squared and cubic distances from the forest edge [(x – average(x)), (x – average(x))2 , (x – average(x))3 , m], (b) city (as a factor, 1 = Helsinki, 2 = Lahti), (c) distance from the path (m), and d) wear class of the path (four levels, see above). Furthermore, we included an interaction term between (c) and (d) as we hypothesised the effect of distance from the path to be different for different wear classes. Distance from the edge included three components to allow for a curvilinear response in the sapling numbers with distance from the edge. For the sake of parsimony, the simplest model for each response was chosen using a stepwise model selection procedure. The squared and cubic edge distance terms and the path distance × wear class interaction were dropped if their p-values were larger than 0.20. The interaction between path distance and wear was considered first and if removed, the cubic term of edge distance was considered, followed by the squared term. We used the full dataset (i.e. all sites including individuals of the focal response) to study the effects of edges and trampling, but the models were also run without sample plots on the paths that we expected to have a disproportionate effect on the responses to ‘distance from the path’. The response curves as a function of edge distance were drawn from the full dataset using value 1 for path distance and wear class, with the level of the factor ‘city’ set to Helsinki. Responses to distance from the paths were drawn without

sample plots on the paths, using mean values for distance from the edge, with Helsinki representing the level for ‘city’. To allow for easy interpretation by end-users (e.g. foresters and city officials), we present our results in terms of saplings per hectare. With frequent zero counts, we could not estimate GLMMs for separate size categories, and the size categories per tree species were pooled until data became sufficient. For B. pendula, B. pubescens, P. sylvestris, P. tremula, and the ETG, all size classes were combined. Size classes 1 and 2, and size classes 3 and 4 were pooled for P. abies, while all size classes were analysed separately for S. aucuparia. Results Altogether we observed 1501 tree saplings on the 570 plots that we sampled. The study sites in Lahti were more open than those in Helsinki – the canopy cover in the sample plots was 50% while being 74% in Helsinki. The dominant tree species was P. abies, with P. sylvestris, S. aucuparia, B. pendula and pubescens, P. tremula, and Salix species as other components, in decreasing order of summed dbh. S. aucuparia and B. pendula were more frequent in Helsinki than in Lahti (the average summed dbh per sample plot being 9.6 cm for S. aucuparia and 9.4 cm for B. pendula in Helsinki while being 1.9 cm and 1.7 cm in Lahti, respectively). A somewhat different set of sapling species was observed in the two cities, for example, P. tremula and S. aucuparia were significantly more abundant in Helsinki while J. communis was only recorded from forests in Lahti, and P. padus only from Helsinki (Tables 2 and 3). In terms of sapling responses, the forest edge effect seemed to penetrate at least up to 80 m into the forest interior while the effects of paths were observed up to 6 m from the path edge, into the un-trampled vegetation (Tables 3 and 4, Figs. 1 and 2). Edge effects Apart from B. pubescens and S. aucuparia, all saplings responded as hypothesised to distance from the edge (Fig. 1). Unexpectedly, B. pubescens saplings were scarce near the edge but increased from

Table 2 Numbers of tree saplings per hectare in urban forests in Helsinki and Lahti. Species

Acer platanoides Alnus incana Betula pendula Betula pubescens Juniperus communis Picea abies Pinus sylvestris Populus tremula Prunus padus Quercus robur Rhamnus frangula Salix caprea Sorbus aucuparia

Helsinki

Lahti

Mean

SD

Min

Max

Mean

SD

Min

Max

1056 93 1553 1211 0 1708 497 1615 342 124 652 373 18,292

6519 962 6418 7659 0 5736 3404 7354 3089 1109 4452 1897 29,648

0 0 0 0 0 0 0 0 0 0 0 0 0

80,000 10,000 70,000 100,000 0 40,000 40,000 70,000 50,000 10,000 50,000 10,000 220,000

161 964 1566 602 241 5622 964 643 0 80 442 924 7751

1260 6403 6687 3244 1993 15,017 4000 3646 0 894 3941 7643 15,285

0 0 0 0 0 0 0 0 0 0 0 0 0

10,000 60,000 70,000 30,000 20,000 130,000 30,000 40,000 0 10,000 50,000 110,000 90,000

872

S. Lehvävirta et al. / Urban Forestry & Urban Greening 13 (2014) 869–877

b) Edge tolerant species 2500

8000 10000

a) Edge tolerant species

1500 1000

6000 4000

0

500

2000 0 0

20

40

60

80

100

0

20

40

60

80

100

d) Edge indifferent species 4000

3000

c) Edge tolerant species

Sorbus aucuparia SIZE 1 Sorbus aucuparia SIZE 2 Sorbus aucuparia SIZE 3 Sorbus aucuparia SIZE 4

0

1000

500 1000

2000

2000

3000

Populus tremula ALL SIZES Pinus sylvestris ALL SIZES

0

Predicted number of saplings per ha

Acer−Alnus−Prunus−Quercus−Salix ALL SIZES Betula pendula ALL SIZES

2000

Betula pubescens ALL SIZES

0

20

40

60

80

100

0

20

40

60

80

100

5000

e) Edge sensitive species

0

1000

2000

3000

4000

Picea abies SIZES 1−2 Picea abies SIZES 3−4

0

20

40

60

80

100

Distance from the forest edge (m) Fig. 1. Responses of tree saplings (individual species and the species group) to distance from forest edges (see Table 1 for predictions and species groups and Table 3 for GLMM results). All predicted values are numbers of saplings per hectare.

the edge up to 80 m into the forests (Table 3, Fig. 1). Betula pendula and the ETG were more abundant at the edge than in the forest interior, as expected, but instead of a simple decreasing response peaked again at approximately 40 m from the edge (Fig. 1). The predicted number of B. pendula saplings decreased from ca. 1000 ha−1 to almost zero at ca. 70 m into the forests. P. sylvestris and P. tremula numbers decreased from the edge to the interior. Contrary to our hypothesis, small S. aucuparia (size classes 1 and 2) were not indifferent to distance from the edge, being most abundant at 25–30 m

from the edge. The number of P. abies saplings did not respond statistically significantly, yet the results suggested better sapling survival in the urban forest interiors, as hypothesised (Table 3, Fig. 1). Trampling effects Saplings were sensitive to direct trampling on the paths, and to distance from paths in the seemingly un-trampled vegetation (Tables 3 and 4, Fig. 2). Yet, the responses to distance from the

S. Lehvävirta et al. / Urban Forestry & Urban Greening 13 (2014) 869–877

1

3

5

4000 3000 0

1000

2000 1000 0

0

−1

e) P. sylvestris All sizes

1

3

5

−1

f) P. abies Sizes 1−2

1

3

5

g) P. abies Sizes 3−4

−1

1

3

5

−1

1

3

5

6000 4000 0 −1

i) S. aucuparia Size 2

1

3

5

3

5

−1

1

3

5

3

5

2000

k) S. aucuparia Size 4

500

1000

3000

0

0

1000

1000 0 1

1

1500

5000

5000 3000

5000 3000 0 1000 −1

−1

j) S. aucuparia Size 3

7000

h) S. aucuparia Size 1

2000

1000 500 0

0

500

2000

4000

1500

1500

6000

2000

2500

8000

2500

d) P. tremula All sizes

0

Predicted number of saplings per ha

−1

c) Edge tolerant gr All sizes

2000

3000

80000 60000 40000 20000

Wear class 1 Wear class 2 Wear class 3 Wear class 4

b) B. pendula All sizes 4000

a) B. pubescens All sizes

873

−1

1

3

5

−1

1

3

5

Distance from the path (m) Fig. 2. The responses of tree saplings (individual species and the species group) to trampling on and off spontaneous paths (see Table 4). A priori hypotheses and the species group are shown in Table 1. Short response lines on the left are mean numbers of individuals in each wear class on the path, and the curves to the right represent responses off the path. All predicted values are numbers of saplings per hectare.

paths varied. Some saplings (236 out of 1501, i.e. 15.7%) could be found even on paths when trampling was light, or when sheltered by stones or tree roots sticking out of the ground. Six per cent of all saplings occurred on the paths with low amount of trampling, while only 1% of all saplings occurred on the totally worn paths. Betula pendula, a species hypothesised to be sensitive to trampling, increased in number with increasing distance from the paths (Table 4, Fig. 2), though not statistically significantly. Unexpectedly,

more saplings were predicted to occur off the paths of higher trampling intensity compared to the low and moderate wear classes. The other trampling sensitive species, P. abies, responded variably – the number of saplings next to paths with low wear was higher than further away from these paths, whereas next to totally worn paths the number of saplings was low increasing with increasing distance from the path edge. This pattern was evident in both size classes (Fig. 2), but is difficult to interpret since responses off the heavily

– – – 1.150 1.231 0.727 – – – −0.329 −0.041 – – – 0.528 0.508 −0.034 – – – −0.200 0.498

Pathdist:H

0.963 0.733 −2.228 −2.038 −2.511 −3.109 −0.251 −0.453 0.090 −0.001 −0.680

Pathdist:M T H

−0.182 1.079 −1.901 −0.617 −0.515 −0.673 0.043 −0.409 0.081 0.021 −1.066 0.027 0.024 0.041 0.005 0.008 −0.029 −0.049 −0.004 0.001