Population dynamics of small mammals in relation to forest age and structural habitat factors in northern Sweden

Authors


Frauke Ecke, Department of Environmental Engineering, Luleå University of Technology, SE-971 87 Luleå, Sweden (fax + 46 920 492830; e-mail Frauke.Ecke@sb.luth.se).

Summary

  • 1In northern Scandinavia there are indications of a long-term decline in the abundance of the three dominant vole species, Clethrionomys glareolus , Clethrionomys rufocanus and Microtus agrestis , since the 1970s. One explanation proposes that intensified clear-cutting has created even-aged and homogeneous forest stands with poor overall conditions for survival and reproduction of the voles.
  • 2We investigated the relationship between forest age and structural habitat factors and its implications for the species richness and abundance of small mammals. In particular, we assessed the population dynamics of C. glareolus , a forest-dwelling species with rather general habitat requirements .
  • 3Extensive snap-trapping of small mammals was conducted during 1998–2000 on 24 study sites in boreal forests in northern Sweden. Trapping was carried out along transects running from immature forests of six age classes (0–50 years) into adjacent reference sites (> 100 years). At each trapping station we recorded 14 habitat variables that were reduced to three principal components (PCs). The PCs were related to late successional traits, such as forest age and cover of tree layers (PC1), cover of tall vegetation in the field layer (PC2) and structural heterogeneity in the forest floor (PC3).
  • 4The species richness of small mammals, as well as the total abundance of C. glareolus , was positively influenced by tall vegetation (PC2) and structural heterogeneity (PC3) but not by late successional traits (PC1). The youngest forests had higher scores for both PC2 and PC3 compared with older forests.
  • 5The youngest forests also had the highest species richness and total abundance of C. glareolus . This was associated with a generally higher rate of change in numbers of C. glareolus during summer in the youngest forests compared with adjacent reference sites. In contrast, survival during winter was lower in the youngest forests. We found this result to be consistent with a source–sink scenario where young individuals, primarily born in old forest stands in early summer, migrate into younger forests to breed, but where the probabilities for winter survival are poor.
  • 6Our study demonstrates that both the species richness of small mammals and the population dynamics of C. glareolus are influenced to a great extent by structural habitat factors that are altered by common forest management practices in northern Sweden. In order to conserve species richness of small mammals and to minimize population fluctuations of C. glareolus in northern Scandinavia, we outline forest management practices that will provide heterogeneous environments, such as leaving logging residues on site after forest harvesting.

Introduction

Small mammals are of crucial importance to boreal ecosystems because they constitute staple food for many mammalian and avian predators (Krebs & Myers 1974; Hörnfeldt et al. 1990) and they consume plants, lichens, fungi and invertebrates (Ericson 1977; Gebczynska 1983; Hansson 1988). In northern Scandinavia there are indications of a long-term decline since the 1970s in the abundance of the three dominant vole species in boreal landscapes (Clethrionomys glareolus Schreber, Clethrionomys rufocanus Sundevall and Microtus agrestis L.) (Hanski et al. 1993; Hörnfeldt 1995). The decline has been most pronounced in C. rufocanus and may be due to changes in forest management practices and increased fragmentation of important habitats for reproduction and survival (Hörnfeldt 1995). A study by Hansson (1999) indicated that population densities and numerical fluctuations of C. glareolus ceased in northern Sweden between 1971 and 1998, possibly linked to altered land use along with deterioration in the quality of remaining old forests.

During the last century forest management practices in northern Sweden have altered the structure of boreal forests. Today, natural forests are dominated by even-aged stands and immature plantations with a relatively low standing volume (Linder & Östlund 1992). Old-growth forests tend to support higher biological diversity than younger managed forests (Svensson 1996). However, several studies of small mammals have shown that species richness and/or diversity may be higher in younger forests than in older natural forests (Kirkland 1977; Sullivan, Lautenschlager & Wagner 1999; Sullivan & Sullivan 2001). In contrast, a long-term study (31 years) in Hokkaido, Japan, revealed that the diversity of small mammals was generally lower in forest plantations compared with natural forests (Saitoh & Nakatsu 1997). Kirkland (1990) reviewed 21 studies from North America investigating the initial response of small mammal communities to clear-cutting. Species richness and diversity did not show any consistent pattern among the studies, whereas the overall abundance of small mammals increased after clear-cutting (Kirkland 1990). In fact, most studies on small mammal community dynamics indicate that both species composition and abundance may be altered by forest management practices like clear-cutting (Kirkland 1990; Saitoh & Nakatsu 1997; Hansson 1999), although their effects depend on the individual species. For example, forest-dwelling species, like many Clethrionomys spp., are usually disfavoured by clear-cutting (Hansson 1978, 1999; Rosenberg, Swindle & Anthony 1994; Sullivan, Lautenschlager & Wagner 1999; Sullivan & Sullivan 2001) whereas species normally found in open habitats, such as many Microtus spp., are favoured (Hansson 1978; Sullivan, Lautenschlager & Wagner 1999; Sullivan & Sullivan 2001). However, Kirkland's (1990) review revealed that the abundance of C. gapperi Vigors, the North American ecological equivalent to the European Clethrionomys glareolus (Kaneko et al. 1998), increased after clear-cutting.

Habitat selection by small mammals is strongly influenced by factors that provide food and/or shelter directly or indirectly (Hansson 1978, 1982; Henttonen & Hansson 1984; Adler 1985; Adler & Wilson 1987; Batzli 1992; Morris 1995, 1996; Hansson 1997). In boreal forests, such factors may include coarse (CWD) and fine woody debris (FWD), vegetation in the field, shrub and tree layer, and burrows and cavities provided by boulders and rocks (Hansson 1978; Cockburn & Lidicker 1983; Pucek 1983; Chetnicki & Mzurkiewicz 1994; Batzli & Lesieutre 1995; Morris 1997; Johannesen & Mauritzen 1999; Ecke et al. 2001). However, forest management practices alter the availability of the amount of CWD and FWD (McCarthy & Bailey 1994; Carey & Johnson 1995; Sturtevant et al. 1997; McGee, Leopold & Nyland 1999; Sullivan, Sullivan & Lindgren 2001) and the composition and structure of the tree layer (Gilliam, Turrill & Adams 1995; Sullivan, Sullivan & Lindgren 2001).

In this study we investigated a wide spectrum of structural variables generally considered to be important for small mammal population dynamics and that may be altered by forest management practices. Specifically, we analysed the relationship between forest age, changes in structural factors, species richness of small mammals and the population dynamics of C. glareolus. We also tested the hypotheses that species richness as well as the overall abundance of small mammals should be positively influenced by increased structural heterogeneity of the habitats and by clear-cutting. Our study focused upon C. glareolus because this species is generally found in most types of forest habitats in northern Scandinavia and is known to be very flexible in its habitat requirements (Hansson 1969; Henttonen & Hansson 1984). Our results are discussed in relation to the general forest management practices in northern Sweden and their implications for the persistence of the small mammal fauna.

Materials and methods

study area

This study was performed within the rural district of Älvsbyn in northern Sweden (Fig. 1), in part of the middle boreal subzone (Sjörs 1999). We studied the population dynamics of small mammals in forests belonging to six different age classes (0–5, 6–10, 11–20, 21–30, 31–40 and 41–50 years) and compared them with an adjacent mature forest (> 100 years old), referred to as a reference site. We used four replicates of each age class, giving 24 study sites (Fig. 1).

Figure 1.

Location of the 24 study sites in the rural district of Älvsbyn (shaded), Sweden, and illustration of the sampling design on study sites.

A geographic information system (GIS) was used to select suitable study sites. The GIS included attribute data about the age and size of forest stands, land use, roads, and streams and lakes. We selected study sites located in coniferous forests of the fresh heath type, which is a common forest type on moraine soils in northern Sweden. Usually, Scots pine Pinus sylvestris L. is the dominant tree species, but Norway spruce Picea abies (L.) Karst is also common. In the field layer Vaccinium myrtillus L. and Vaccinium vitis-idaea L. predominate. Each study site included two 420-m long parallel transects, running from the immature into the adjacent mature forest stand (Fig. 1). Each transect comprised 15 trapping stations spaced at regular intervals of 30 m, with the centre located at the border between the immature and mature forest. The distance between transects was 100 m. To minimize edge effects from the surrounding areas, each study site had an area of at least 620 × 300 m.

Eighty-three per cent of the rural district of Älvsbyn (1813 km2) is covered by forests. Forty-six per cent of this area has forest stands younger than 50 years and 22% is older than 100 years (Skogsvårdsstyrelsen Norrbotten 2000). Most forests used as reference sites were old-growth forests, but some of them were influenced by widely spaced selective cutting. All forests in the age classes 0–39 years originated from clear-cuts, whereas the forests in the age class 41–50 years had been managed by selective cutting.

monitoring of small mammals

From 1998 to 2000 small mammals were trapped twice per year, in early summer (mid-June) and in autumn (mid-September). Each trapping station (Fig. 1) comprised three snap-traps located within a radius of 2 m. Traps were baited with Polish wicks (Grodzinski, Pucek & Ryzkowski 1966) and dried apples. The traps were checked for 3 consecutive days during each trapping session, once per 24 h. In a more intensive study than ours (five traps per trapping station and 10 m distance between trapping stations), P. Christensen & B. Hörnfeldt (unpublished data) did not find any effect of continuous kill-trapping on small mammal populations. We classified specimens of C. glareolus into four functional categories: those trapped in early summer that had survived winter were referred to as overwintered breeders (OWBR); reproductive individuals trapped in early summer, belonging mainly to the first cohort of the year, were referred to as year-born breeders (YBBR); subadults (SUBA) were individuals trapped in autumn that did not attain maturity during their first year; reproductive and/or post-reproductive individuals trapped in autumn were referred to as adults (ADUL). For details on these definitions, see Myllymäki (1977) and Löfgren (1989).

estimation of habitat variables

At each trapping station we quantified 13 structural habitat variables within an area of 10 × 10 m (Table 1). The cover of trees in the lower tree layer (TR2) was also a measure of the age structure of the forest: even-aged forest stands scored zero for TR2. TOP was a combined measure of the structural complexity and relief of the forest floor, including hollows, hummocks, stumps, logs and boulders.

Table 1.  Description of the 14 habitat variables recorded on each trapping station
VariableDescriptionMeasure
AGEForest age, divided into seven age classes: 1–5; 6–10; 11–20; 21–30; 31–40; 41–50; > 100 yearsGIS with information for each forest stand, covering the entire study area
TR1Cover of trees (> 5 m high)%
TR2Cover of trees (> 5 m high) that differ at least 5 m in height from TR1%
SHRCover of shrubs (1·5–5 m high)%
HERCover of vascular plants (< 1·5 m high)%
MOSCover of mosses in the bottom layer%
LICCover of lichens in the bottom layer%
EPLCover of epixylic lichensFour-graded scale: 0%; > 0 ≤ 5%; > 5 ≤ 50%; > 50% of branches and stems at breast height
CWDCover of coarse woody debris (stems > 10 cm diameter)van der Maarel scale*, nine-graded
FWDCover of fine woody debris (stems < 10 cm)van der Maarel scale*, nine-graded
BOUCover of visible boulders > 10 cm diametervan der Maarel scale*, nine-graded
UMBCover of umbrella vegetation, vegetation > 30 cm but < 1·5 m highvan der Maarel scale*, nine-graded
SNANumber of snags, i.e. standing dead trees 
TOPTopographyFour-graded scale: level; sparsely distributed hollows; many hollows but only locally distributed; many hollows, evenly distributed

data analysis

Trapping stations located at the border between the immature and the mature forests were excluded from the analyses. Within any given forest age class there were seven trapping stations on each transect, and two transects per study site (Fig. 1). Each age class was studied at four sites. Thus, our study originally comprised 56 trapping stations per forest age class. However, we excluded data from 84 trapping stations (including one complete study site) that were altered by forest management practices during the study period. For this reason, there were 42 trapping stations in the age classes 0–5 and 31–40 years and on the reference sites adjacent to the age classes 0–40 years. In total, we analysed data from 588 trapping stations.

As a measure of the seasonal variation in abundance of C. glareolus, we calculated the rate of change in numbers from early summer to autumn and from autumn to early summer in the following year. The rate of change was calculated as log10(1 + Xt+1) – log10(1 + Xt), where X refers to the abundance of C. glareolus at a trapping station and t is the time (Steen, Ims & Sonerud 1996).

We used principal component analysis (Jongman, ter Braak & van Tongeren 1995) to reduce the 14 habitat variables to a few essential components. Only principal components (PCs) that explained more than 10% of the variance among the habitat variables were considered. For the interpretation of the PCs, variable loadings |> 0·5| were used.

To analyse relationships among habitat variables, and between the PCs and small mammal trapping data, we used the Spearman's rank correlation (rs). The significance level (α) of all correlation coefficients was adjusted (α′) for the number of correlations performed (k) with the Bonferroni method (α′ = α/k) (Sokal & Rohlf 1995). The overall relationship between the habitat variables was analysed with Kendall's coefficient of concordance (W). Kruskal–Wallis analysis of variance by ranks (H) was used to test the overall differences between the immature forests and the reference sites. For age-specific comparisons between an age class and its respective reference sites, we applied the Mann–Whitney U-test (U). We used logistic regression to examine the importance of the habitat variables (represented by the PCs) in predicting the occurrence of various functional categories of C. glareolus at trapping stations. For these analyses, the abundance of the functional categories at each trapping session was reduced to a binary response variable (present/absent) and stepwise logistic regressions were performed (Quasi-Newton estimation method with 50 iterations; Sharma 1996). The joint significance of the explanatory PCs was tested with the likelihood ratio χ2-statistic. Estimates of the explanatory variables revealed the importance of each PC. Significance was tested with Wald's χ2-statistic.

Results

small mammal abundance

During 1998–2000 we trapped a total of 1816 C. glareolus, seven C. rufocanus, 35 m. agrestis and 133 Sorex spp. Specimens of C. glareolus were found on all study sites; Sorex spp. occurred on 22, C. rufocanus on three and M. agrestis on 10 of the 23 study sites. There was a marked seasonal variation in the abundance of C. glareolus and Sorex spp. (Fig. 2). All species trapped were considered for the analyses of species richness, whereas only data on C. glareolus were used for the analyses of population dynamics.

Figure 2.

Total number of small mammals trapped per trapping session, in early summer and autumn 1998–2000. Shading denotes winter. C. glareolus (closed circles); C. rufocanus (open circles); M. agrestis (open triangles); Sorex spp. (closed triangles).

general characteristics of the forests

Many of the 14 habitat variables recorded were associated with each other (W = 0·63, n= 588, P < 0·001). The two measurements of tree canopy cover (TR1 and TR2) were positively correlated with forest age (rs = 0·65 and rs = 0·56, respectively, n= 588, P′ < 0·004) whereas the cover of shrubs, FWD, boulders and umbrella vegetation were negatively correlated with forest age (rs = −0·27, rs = −0·21, rs = −0·20 and rs = −0·21, respectively, n= 588, P′ < 0·004). The cover of epixylic lichens and mosses, and the number of snags, were also positively related to forest age (rs = 0·60, rs = 0·56 and rs = 0·39, respectively, n= 588, P′ < 0·004). Forest age was not correlated with the cover of herbs, lichens, CWD or topography. Site-specific comparisons also revealed that several of the habitat variables recorded differed significantly between the immature and adjacent mature forests (Table 2). The 0–5-year-old forests differed in all variables from the mature forests except in the cover of CWD. Forests 0–20 years old had a generally higher cover of umbrella vegetation and FWD than their respective reference sites. In contrast, the 31–40- and 41–50-year-old forests had less cover of both FWD and CWD than their respective reference sites. The cover of herbs (HER) was highest in the 0–5-year-old forests. The 0–10- and 31–40-year-old forests had a more complex forest floor (TOP) than their reference sites. The cover of shrubs was negatively correlated with forest age in general, but was less in the 0–5-year-old forests compared with adjacent mature forests. In contrast, the forests that were 6–40 years old had a generally higher cover of shrubs than their respective reference sites. Generally, the cover of the two tree layers, mosses and epixylic lichens was higher in the mature forests than in the adjacent managed forests that were 0–40 years old. These differences were less pronounced between the 41–50-year-old forests and their respective reference sites. Trapping stations located in the age classes 0–5 and 31–40 years had a higher cover of visible boulders than those in the adjacent mature forests. The number of snags, i.e. standing dead trees 1 m high, was consistently lower in all immature forest age classes than on their respective reference sites (Table 2).

Table 2.  Mean estimates of the structural habitat variables and of the principal component scores in the forests of different age (Im) and in the adjacent mature forests (Ma). Bold figures indicate higher values in the immature forest compared with the mature forest according to the Mann–Whitney U -test ( * P < 0·05, ** P < 0·01, *** P < 0·001 ). PC1, late successional traits; PC2, cover of tall vegetation in the field layer; PC3, structural heterogeneity of the forests floor. See Table 1 for abbreviations and see the Materials and methods for n
VariableAge classes (years)
0–5 vs. > 1006–10 vs. > 10011–20 vs. > 10021–30 vs. > 100 31–40 vs. > 10041–50 vs. > 100
ImMaImMaImMaImMaImMaImMa
TR1 1·5225·26***1·3813·17*** 3·0914·68*** 7·1611·71***12·5712·6715·5220·36*
TR2 0·21 6·19***0·04 3·55*** 0·09 3·73*** 0·18 3·69*** 0·00 3·52*** 2·39 3·95
SHR 1·50 4·07***14·32 2·52***12·98 3·91***8·09 6·21*6·90 2·95*** 6·12 4·62
HER49·0539·64*30·5233·6932·3036·1143·9641·6737·5533·6949·4643·84
MOS15·5768·33***38·8962·62***31·4165·80***46·7574·59***48·4571·43***73·6660·80**
LIC 0·07 0·55***13·07 5·457·52 2·61*** 5·18 1·985·57 2·31***1·93 1·21*
EPL 0·57 2·76***0·50 2·31*** 2·23 3·32*** 2·36 3·29*** 2·64 3·36*** 3·63 3·73
FWD 4·81 3·55***4·18 3·60**4·13 3·63** 3·05 3·10 3·40 4·05** 2·38 3·70***
CWD 2·88 2·692·05 3·10** 3·34 3·48 2·30 2·52 2·05 3·10** 1·29 1·95*
BOU 3·93 2·45***3·93 3·62 4·79 4·57 3·36 2·83 4·90 2·98*** 2·18 2·13
UMB 4·81 3·71**4·93 3·02*** 4·77 3·11*** 4·82 4·57 4·31 3·40 5·20 5·18
TOP 3·36 2·45***2·82 2·57* 2·77 3·04 2·73 2·83 3·05 2·60** 3·11 3·11
SNA 0·12 0·67***0·00 0·59*** 0·09 0·87*** 0·05 0·62*** 0·02 0·43*** 0·16 0·41*
PC1−1·06 1·07***−1·53 0·56***−0·95 0·96***−0·67 0·68***−0·45 0·79*** 0·16 0·86***
PC2 0·41−0·14**−0·22−0·72−0·13−0·36 0·15 0·13−0·09−0·39 0·63 0·51
PC3 1·15−0·20***0·07−0·08 0·40 0·37−0·34−0·34−0·08−0·06−0·81 0·01***

The 14 habitat variables were reduced to three PCs by principal component analysis. The first PC explained 34·7%, the second 26·4% and the third 14·0% of all variance among the habitat variables. The first PC was positively correlated with forest age, the number of snags and the cover of the tree layers, mosses and epixylic lichens (Fig. 3). Therefore, we interpreted the first PC as a factor related to late successional traits. The cover of vascular plants and umbrella vegetation was positively correlated with the second PC, whereas the cover of lichens and boulders was negatively related to this component. Thus, we interpreted the second PC as a factor involving cover of tall vegetation in the field layer. There was a positive correlation of CWD, FWD and topography with the third PC and we interpreted this component as a factor involving general structural heterogeneity of the forest floor (Fig. 3).

Figure 3.

The scores of the 14 habitat variables derived from principal component analysis, plotted against the first and second PC (a) and the second and third PC (b). Scores |> 0·5| (indicated by the dashed lines) were used for the interpretation of the PCs. Variables contributing significantly (circles) and not significantly (triangles) to the PCs, respectively, are shown. The axes were rotated using the varimax normalized rotation technique. n = 588. See Table 1 for abbreviations.

Forest age-specific comparisons revealed that the scores for PC1 (late successional traits) were consistently lower in immature forests compared with adjacent mature forests (Table 2). Only the youngest forest age class (0–5 years) had higher scores for the cover of tall vegetation in the field layer (PC2) and for structural heterogeneity (PC3) than the adjacent reference sites. In fact, with respect to structural heterogeneity, this age class had much higher scores compared with all other age classes (Table 2).

species richness in relation to forest attributes

There was an overall difference in the species richness of small mammals among forest age classes (H6,588 = 16·07, P < 0·05; Fig. 4). Trapping stations in the youngest forests (0–5 years) had the highest species richness, and the trapping stations in the 11–20-year-old forests had lower species richness than those in the adjacent mature forests (Table 3). Species richness was also positively influenced by the cover of tall vegetation in the field layer (PC2) (rs = 0·21, n= 584, P′ < 0·02) and by structural heterogeneity (PC3) (rs = 0·20, n= 584, P′ < 0·02), but not by late successional traits (PC1) as such.

Figure 4.

Species richness per trapping station in relation to forest age. Mean values ± two SE are given. For n see the Materials and methods.

Table 3.  Means of small mammal data in the forests of different age (Im) and the adjacent mature forests (Ma). Bold figures indicate higher values in the immature forest compared with the mature forest according to the Mann–Whitney U -test ( * P < 0·05, ** P < 0·01, *** P < 0·001 ). See the Materials and methods for abbreviations and for n
VariableAge classes (years)
0–5 vs. > 1006–10 vs. > 10011–20 vs. > 10021–30 vs. > 10031–40 vs. > 100 41–50 vs. > 100
ImMaImMaImMaImMaImMaImMa
Species richness1·40 0·90** 1·11 1·00 0·93 1·21** 1·02 1·00 1·05 0·90 0·91 0·80
Abundance of
 C. glareolus4·17 1·81*** 3·63 2·76 3·71 4·98*2·88 1·64* 2·40 2·38 2·79 3·07
 Sorex spp. 0·26 0·26 0·16 0·21 0·16 0·32 0·29 0·240·26 0·05* 0·30 0·18
Abundance of
 OWBR 19980·50 0·17 0·39 0·21 0·23 0·29 0·09 0·10 0·14 0·19 0·13 0·16
 OWBR 19990·05 0·02 0·07 0·02 0·00 0·20** 0·00 0·05 0·02 0·02 0·07 0·16
 OWBR 20000·12 0·17 0·23 0·17 0·14 0·30 0·05 0·00 0·02 0·05 0·07 0·07
 YBBR 19980·50 0·10* 0·23 0·26 0·18 0·21 0·14 0·07 0·21 0·14 0·32 0·20
 YBBR 19990·00 0·00 0·00 0·00 0·00 0·00 0·00 0·02 0·00 0·05 0·02 0·00
 YBBR 20000·10 0·00* 0·05 0·05 0·14 0·18 0·05 0·00 0·07 0·00 0·05 0·02
 ADUL 19980·40 0·24 0·39 0·26 0·34 0·54 0·30 0·33 0·33 0·19 0·25 0·41
 ADUL 19990·07 0·00 0·05 0·05 0·07 0·14 0·04 0·05 0·00 0·00 0·05 0·13
 SUBA 19982·10 1·10** 1·95 1·67 2·34 2·792·05 0·95** 1·43 1·64 1·48 1·41
 SUBA 19990·31 0·02**0·23 0·07* 0·25 0·34 0·13 0·07 0·19 0·10 0·39 0·52
Rate of change during
 Summer 19980·34 0·24 0·36 0·33 0·42 0·510·40 0·28* 0·34 0·32 0·29 0·30
 Summer 19990·09−0·00**0·06 0·02* 0·08 0·07 0·04 0·02 0·05 0·02 0·10 0·10
 Winter 1998–99−0·45−0·28**−0·44−0·38−0·48−0·53−0·43−0·29*−0·37−0·37−0·31−0·30
 Winter 1999–2000−0·07 0·04**−0·02 0·02−0·04−0·03−0·03−0·03−0·05−0·01−0·09−0·13

abundance of c. glareolus in relation to forest attributes

The total number of C. glareolus caught per trapping station over the entire study period differed between the age classes (H6,588 = 12·66, P < 0·05). Age- and site-specific comparisons also revealed that trapping stations located in the 0–5- and 21–30-year-old forests had higher total abundances than trapping stations in the adjacent mature forests. However, the reference sites adjacent to the 21–30-year-old forests appeared to be marginal habitats that were colonized at high abundance only, as in autumn 1998. In contrast, the 11–20-year-old forests had a lower total abundance than the adjacent reference sites (Table 3).

Like species richness, the total abundance of C. glareolus was positively affected by cover of tall vegetation in the field layer (PC2) (rs = 0·23, n= 584, P′ < 0·02) and by structural heterogeneity (PC3) (rs = 0·28, n= 584, P′ < 0·02) but not by late successional traits (PC1).

However, various functional categories of C. glareolus differed in their relation to the studied forest attributes between both categories and years of trapping (Table 3). With respect to the abundance of overwintered breeders in early summer 1998–2000, there was no consistent pattern in relation to forest age (Fig. 5). However, in 1998, when densities in early summer were extremely high, this category was found more frequently in young forest stands (0–10 years) than in the older ones (H6,588 = 19·85, P < 0·01). Age- and site-specific comparisons also indicated lower abundance of overwintered breeders in forests that were 11–20 years old, than in the adjacent mature forests (Table 3).

Figure 5.

Number of C. glareolus of different functional categories caught per trapping station in 1998–2000 in relation to forest age in 1998 (circles), 1999 (triangles) and 2000 (squares). Mean values ± two SE are given. Overwintered (OWBR) and year-born breeders (YBBR) in early summer, and reproductive/post-reproductive (ADUL) and subadult (SUBA) individuals in autumn, respectively. For n see the Materials and methods.

The pattern of the occurrence of animals born and breeding in the same year (year-born) and overwintered breeders trapped in early summer was similar in the forest age classes. At high abundances, as in early summer 1998, the year-born breeders were most abundant in the youngest forest age class (Fig. 5). Age- and site-specific comparisons also suggested the year-born breeders to be more abundant in the youngest forest age class compared with adjacent mature forests in early summer 2000 (Table 3). In early summer 1999 only a few year-born breeders were trapped and they were all found in the forest age classes of 41–50 and > 100 years old. Among the adults trapped in autumn, i.e. reproductive/post-reproductive individuals, we found no differences in abundance with respect to forest age class (Fig. 5). In contrast, there was a higher abundance of subadults in the young forest age classes compared with reference sites both in autumn 1998 and 1999 (Fig. 5 and Table 3).

Stepwise logistic regressions on the extracted PCs suggested that at high abundance (1998) the occurrence of overwintered individuals was positively influenced by tall vegetation in the field layer (PC2) and by structural heterogeneity (PC3), but negatively influenced by factors related to late successional traits (PC1) (Table 4). At low abundance (1999) PC2, and at intermediate abundance (2000) PC3, positively influenced the occurrence of overwintered breeders.

Table 4.  Stepwise logistic regressions on the occurrence (present/absent) of each functional category of C. glareolus during the trapping session in 1998–2000. Numbers in parentheses represent the number of trapping stations at which a functional category was trapped. n = 588, * P < 0·05, ** P < 0·01, *** P < 0·001 . See Materials and Methods for abbreviations
Functional categoryχ2 ( d.f. = 3)Classification (% correct)FactorEstimateWald's χ2
Early summer
OWBR
1998 (92)31·0***84·2PC1−0·227 3·9*
   PC20·41313·0***
   PC30·43013·5***
1999 (27)10·8*95·4PC20·481 5·4*
2000 (58)19·9***90·0PC30·62118·2***
YBBR
1998 (94)14·3**83·9PC20·38111·0***
1999 (3) 5·1 NS    
2000 (33)11·8**94·3PC20·522 8·2**
Autumn
ADUL
1998 (168)20·3***71·6PC20·246 7·0**
   PC30·32912·2***
1999 (30) 8·6*94·9PC30·427 5·2*
SUBA
1998 (409)22·8***69·5PC20·207 4·7*
   PC30·38015·7***
1999 (104)35·1***82·4PC20·42914·9***
   PC30·48518·1***

Cover of tall vegetation in the field layer (PC2) was in all years the most important factor explaining the occurrence of year-born breeding individuals trapped in early summer.

Structural heterogeneity (PC3) positively influenced the occurrence of adult (reproductive/post-reproductive) C. glareolus individuals trapped in autumn. At high abundances (1998), the cover of tall vegetation in the field layer (PC2) was also important for the occurrence of this category. Among subadult C. glareolus trapped in autumn, both the cover of tall vegetation in the field layer (PC2) and structural heterogeneity (PC3) positively influenced capture probabilities (Table 4).

In the logistic regression models, the estimates of the PCs were higher in 1999 and 2000 than in 1998. They were also generally higher in early summer than in autumn (Table 4). Thus, habitat preferences appeared to be more pronounced at low/intermediate abundances than at high abundances, and more pronounced in early summer than in autumn.

seasonal variation in the abundance of c. glareolus in relation to forest attributes

The increase in abundance of C. glareolus during summer was higher in 1998 than in 1999 (Fig. 6). In contrast, survival during winter was generally lower in winter 1998–99 than in winter 1999–2000. Thus, the large increase in abundance of C. glareolus during summer 1998 was followed by poor survival during winter, whereas the small increase during summer 1999 was followed by a relatively high survival during the following winter. The rate of change in C. glareolus from autumn 1998 to early summer 1999, i.e. survival during winter, also differed between the forest age classes (H6,588 = 17·03, P < 0·01) (Fig. 6). Survival of C. glareolus during both winter 1998 and winter 1999 was higher in the mature forests compared with the adjacent 0–5-year-old forests. A similar pattern was observed in the 21–30-year-old forests compared with the adjacent mature forests during winter 1998–99 (Table 3). In contrast, the increase in abundance of C. glareolus during summer 1998 was higher in the 21–30-year-old forests compared with the adjacent mature forests. In summer of 1999 it was higher in the 0–10-year-old forests compared with their reference sites (Table 3).

Figure 6.

Rate of change of C. glareolus per trapping station in relation to forest age. During summer (a) 1998 (circles) and 1999 (triangles), and during winter (b) 1998–99 (circles) and 1999–2000 (triangles). Mean values ± two SE are given. For n see the Materials and methods.

The rate of change in C. glareolus was positively correlated with structural heterogeneity (PC3) in summer 1998, and with both the cover of tall vegetation in the field layer (PC2) and PC3 in summer 1999 (Table 5). In contrast, during both winter 1998–99 and 1999–2000 the rate of change was negatively correlated with both the second and third PC (Table 5).

Table 5.  Relationship ( rs ) between the rate of change of C. glareolus in different years and seasons and the scores of the three principal components. For interpretation of principal components see Table 2. n  = 588. * P ′ < 0·02 Bonferroni corrected
Year and seasonPrincipal component
123
Summer
1998−0·034 0·083 0·160*
1999−0·059 0·152* 0·164*
Winter
1998–99 0·087−0·125*−0·222*
1999–2000 0·038−0·161*−0·066

Thus, the rate of increase during summer was positively influenced by the cover of tall vegetation in the field layer and by structural heterogeneity, but trapping stations with high scores for these factors also seemed to have poor survival during winter. Survival during winter was not generally related to late successional traits. However, survival during winter 1998–99 and 1999–2000 was still higher in the mature forests compared with the adjacent 0–5-year-old forests (Table 3).

Discussion

Our study revealed a higher species richness of small mammals in the youngest forest age class (0–5 years) compared with mature forests. Also, in British Columbia, Canada, Sullivan, Lautenschlager & Wagner (1999) and Sullivan & Sullivan (2001) found species richness to be higher on clear-cuts. In our study species richness was lower in forest stands that were 11–20 years old, which is similar to the result of Sullivan, Lautenschlager & Wagner (1999), who found an increase in species richness 9–10 years after clear-cutting followed by a decline in species richness after 11 years. We also found that species richness was positively influenced by habitat variables related to the cover of tall vegetation in the field layer (PC2) and to structural heterogeneity on trapping stations (PC3). Heterogeneous environments are generally held to enhance species richness (Kerr & Packer 1997), and indeed the youngest forest stands (0–5 years) were far more heterogeneous, as implied by PC3, compared with all other forest age classes investigated in this study. Also, the cover of tall vegetation in the field layer was on average much higher in the youngest forest age class. Both these characteristics should provide shelter, and indirectly also food, for small mammals (Hansson 1978; Cockburn & Lidicker 1983; Adler 1985; Harmon et al. 1986; Batzli & Lesieutre 1995; Morris 1997). Even though we only studied four species of small mammals over 2 years, the species richness of small mammals in this type of landscape initially seems to be promoted by clear-cutting, especially when the clear-cutting contributes to increased structural heterogeneity and cover of tall vegetation in the forest floor. However, subsequently, when these forests become 10–20 years old, it is likely that species richness will decline due to reduced vegetation cover.

Clethrionomys glareolus is generally known to be a forest-dwelling species but may also occur in open habitats ( Hansson 1969, 1978 ; Henttonen & Hansson 1984 ). Our study suggests rather general habitat requirements for C. glareolus , as this species was found in all forest age classes. However, the abundance of C. glareolus was highest in the youngest forests (0–5 years). This result is not consistent with studies suggesting that C. glareolus is more common in forests than on clear-cuts and reforestations in northern Sweden ( Hansson 1978, 1999 ). Sullivan, Lautenschlager & Wagner (1999 ) and Sullivan & Sullivan (2001 ) studied C. gapperi in North America and found this species to be less abundant on clear-cuts than in old forests, whereas Kirkland (1990 ) reported the reverse. Hence it appears that the effect of clear-cutting on these Clethrionomys spp. may be quite different, possibly related to factors other than forest age. Our study suggests that the overall abundance of C. glareolus was to a great extent dependent on the habitat characteristics in the field layer. Specifically, we found that cover of tall vegetation in the field layer (PC2) and structural heterogeneity on trapping stations (PC3) positively influenced the total abundance of C. glareolus . Because the scores of these factors were higher in the youngest forest age class compared with most other forests and the reference sites, the youngest forests should indeed have the highest total abundance of C. glareolus .

The high overall abundance of C. glareolus in the youngest forests in the present study was mainly due to high numbers of subadults in autumn. High abundance of subadults was also closely linked to a high rate of change during summer. In contrast, and compared with reference sites, survival during winter in young forests was rather poor and the abundance of overwintered breeders in the following summer was no higher than in adjacent mature forests. Despite poor survival during winter, year-born breeders were at both high (1998) and intermediate (2000) abundances in young forests. Reproductive females of C. glareolus are territorial and they need exclusive space to breed (Bujalska 1985; Löfgren 1995). Hence, the observed pattern in our study may be explained with a source–sink scenario where young individuals, primarily born in old forest stands in early summer, immigrate into younger forests to breed, but where the probability of winter survival is low. A similar mechanism was suggested by Hansson (1999), who proposed that reforested areas may act as sinks for surplus individuals of C. glareolus produced in old forests. With respect to the occurrence of individuals that had survived winter , the cover of tall vegetation in the field layer (PC2) and structural heterogeneity (PC3) appeared to be important during both high and low abundances. Most probably this result simply reflects the fact that there was a very high abundance of subadults in this type of habitat in the preceding autumn. The cover of tall vegetation in the field layer (PC2) was the only factor that positively influenced the occurrence of year-born breeders, whereas both this factor and structural heterogeneity (PC3) were important explanations for the occurrence of adults and subadults in autumn. In none of these categories was the abundance found to be positively influenced by late successional traits. Therefore, it seems likely that forest age as such is not important for the overall distribution of this species. However, we found that winter survival on our reference sites was higher than in the adjacent young forests (0–5 years old), which suggests that old forests are likely to be important refuges for survival during winter.

Despite the differences in several of the 14 habitat variables between the forests aged 31–50 years old and their reference sites, there were no pronounced differences with respect to the cover of tall vegetation in the field layer and structural heterogeneity in the forest floor. This similarity in habitat characteristics may explain why there were neither differences in species richness nor in the abundance and rate of change in C. glareolus in the 31–50-year-old forests compared with the adjacent mature forests. Hence, it seems plausible that the habitat conditions for C. glareolus, and for small mammals in general, were similar in these age classes as on the reference sites, for example in quality of CWD and FWD. During the 31–50 years since these clear-cuts were created, the woody debris left behind should have at least partly decayed (Harmon et al. 1986), providing a similar amount of cover and food as the woody debris in mature forests. The abundance of C. glareolus has been shown previously to be positively correlated to the cover of FWD, CWD and umbrella vegetation in subalpine forests (Ecke et al. 2001).

This study has demonstrated that species richness of small mammals and population dynamics of a forest-dwelling species like C. glareolus are influenced by structural habitat factors that are altered by common forest management practices in northern Sweden. Based on our results, we suggest that where clear-cutting is used in forest management, logging debris should be left on site. This should increase the overall habitat heterogeneity and the cover of tall vegetation in the field layer, thereby increasing availability of food and shelter and promoting reproduction and/or survival of several species of small mammals during summer. However, these factors appear to be less important during winter for C. glareolus in this study, as survival was poor on young clear-cuts compared with old forests. As a consequence, the yearly fluctuations in numbers of C. glareolus were higher in young reforestations compared with middle-aged and old forests. This result is in line with the hypothesis of Van Horne (1983), which suggests that low-quality habitats may support high densities but are mainly occupied by immigrants, whereas high-quality habitats may have lower but less fluctuating densities of small mammals. If so, population fluctuations of a habitat generalist like C. glareolus should be greater in a landscape where the proportion of young forests is high compared with a landscape where middle-aged and old forests dominate. Large population fluctuations in small mammals are known to be accompanied by vole damage to forestry seedlings during peak years (Hörnfeldt, Löfgren & Carlsson 1986; Hansson 1988, 1999). In order to reduce population fluctuations of small mammals it is important to consider both the size of the logging areas and the surrounding mosaic of reforestations and mature forests. We suggest that future studies should focus not only on alternative techniques for logging and harvesting forests, but also on the effects of forest management practices at different spatial scales. The overall habitat quality in existing mosaics of forest stands should be evaluated for possible influences on the long-term persistence and population stability of specific species of small mammals.

Acknowledgements

We very much appreciate the help during fieldwork by Bertil Marklund, Lars Berggren, Fredrik Burström, Anders Drugge, Lars-Göran Ek, Gunnar and Rebecca Fors, Anita Heikkilä, Krister Higberg, Jens Iversson, Britt-Inger Jonsson, Håkan Karlsson, Mikhail Kalinkin, Robert Lindström, Jimmy Lundgren, Satu Mustonen, Ann Rask, Katja Ruth, Krister Salmela and Jörgen Wåhlström. The assistance of Bertil Marklund in the laboratory is also greatly appreciated. Dr Takashi Saitoh gave valuable comments on the sampling design. We are also grateful to Peter Söderberg at the National Board of Forestry who provided us with data on the forest stand characteristics in the rural district of Älvsbyn. Financial support was given by grants to Ola Löfgren from ‘Olle och Signhild Engkvist Stiftelser’ and from the Faculty of Engineering, Luleå University of Technology.

Ancillary