Introduction

Bird populations are becoming widely used indicators for assessing the state of forest biodiversity (Gregory and van Strien 2010; Voříšek et al. 2020; Oettel and Lapin 2021). At the assemblage scale, forest birds can reflect alterations of stand structure (LaManna and Martin 2016), critical loss of old-forest habitats in landscapes (Schmiegelow and Mönkkönen 2002; Roberge et al. 2018), and broad-scale habitat turnover related to intensive production forestry (Lamanna and Martin 2016, Roberge et al. 2018). While the general development of bird assemblages after clear-cutting or salvage logging is well understood (Roberge et al. 2018; Georgiev et al. 2020; Thorn et al. 2020), its alternative successional trajectories and their underlying forest management decisions have been seldom explicitly studied.

In Europe, late-successional hemiboreal (boreo-nemoral) mixed forests constitute a hotspot of bird diversity where the species pools of boreal and temperate forests meet (Nilsson 1997; Roberge et al. 2018). Notably on fertile soils, the mixtures of coniferous and deciduous trees provide structurally heterogeneous conditions there (Lõhmus and Kraut 2010; Põldveer et al. 2020), supporting more diverse bird assemblages than pure deciduous or conifer stands (Nilsson 1997; Lõhmus et al. 2016). In old-growth stands, a large dead-wood supply provides additional resources for many species (Nilsson 1979a; Rosenvald et al. 2011; Bergner et al. 2015). Although most European hemiboreal forests have a centuries-long silvicultural history, intensification of timber production continues to raise the issues of ecologically sustainable management regimes (Felton et al. 2020), including a response to regional declines in forest bird abundance (Nellis and Volke 2019). This raises the issue of ecologically sustainable management regimes in these forests, specifically the consequences of further intensification of the even-aged systems (Felton et al. 2020).

In the even-aged systems, a major approach is to plant and manage clearcuts as economically beneficial monocultures; in Northern Europe mostly as planted Scots pine (Pinus sylvestris L.) and Norway spruce (P. abies) stands (Spiecker 2003; Roberge et al. 2018). Compared with natural forests, spruce plantations have a particularly homogenous and simplified stand structure (Nilsson 1979b; Swanson et al. 2011; Bergner et al. 2015), causing local reductions in bird species richness (Gjerde and Saetersdal 1997, Felton et al. 2011) and a loss of some specialist species across landscapes (MacKay et al. 2014; Lindbladh et al. 2019). The bird diversity in spruce plantations can be increased by promoting the abundance of deciduous trees (Lindbladh et al. 2017; Felton et al. 2021; Kebrle et al. 2021; Vélová et al. 2021), dead wood (Lindbladh et al. 2019), and large trees (Kebrle et al. 2021). However, the plantation impacts may be context dependent. For example, a small share of spruce plantations may add diversity in homogeneous Scots pine (P. sylvestris) dominated landscapes (Gjerde and Saetersdal 1997), while adjacent deciduous stands can reduce bird richness in the plantations (Vélová et al. 2021).

The current study investigates post-clearcut bird assemblages in the most productive hemiboreal sites, where natural succession would involve a mixture of deciduous tree species, a later development of the shade-tolerant Norway spruce and, possibly, another late-successional shift to shade-tolerant deciduous stands (Liira and Sepp 2009; Lõhmus and Kraut 2010). Such compositional changes are eliminated in spruce plantations, together with an impoverished layer structure and dead wood abundance. Despite such an obvious divergence in forest successional pathways, there is a lack of comparative studies on the successional development of bird assemblages depending on management scenarios.

In this study, our main focus is the parallel development of bird assemblages in planted versus naturally regenerated stands, between the reference conditions in clearcuts and late-successional assemblages. We ask three questions: (I) how does the origin (spruce plantation vs naturally regenerated stand) affect bird assemblages along with the succession? In general, maturing plantations may retain a species-poorer and sparser bird assemblage (Sweeney et al. 2010; MacKay et al. 2014), but another question is whether, and at which rate, they compositionally converge toward natural stand conditions. This could imply cost-effective restoration for biodiversity either through setting aside plantations or harvesting and replacing these with early successional natural regeneration. (II) Which forest structural elements affect the post-clearcut bird assemblage development? It is well known that several key structural characteristics develop along with the stand age, e.g., dead wood, large trees and vertical stratification (Fuller 2003; Roberge et al. 2018; Klein et al. 2020). An advanced question for management is, however, their contribution independent of age, and depending on the regeneration pathway (e.g., given that plantations lack certain tree species). (III) How are the characteristic bird species of old forests represented across management types, and are resident species more likely to belong to this group? Previous studies have indicated that high site productivity accelerates also the development of old-forest species assemblages (e.g., Nilsson 1997; Lõhmus et al. 2020; 2021). We thus expect their presence already prior to the old-growth stage.

Materials and methods

Study area and sampling design

The study was conducted in the mainland of Estonia, between 57°32′–59°12′ and 23°50′–27°5′. The region belongs to the non-oceanic section of the hemiboreal vegetation zone (Ahti et al. 1968); the mean air temperature is 16.5 °C in July and 6.5 °C in February; the average annual precipitation is 650–750 mm. Forests cover about half of the Estonian area; 46% of the forests belong to the State Forest Management Centre; the average stand age is 55 years. (Raudsaar and Valgepea 2020). Clear-cutting is the dominant approach to timber production in the country.

Our study plots (N = 40) were located in a recently established forest reserve network (see Lõhmus et al. 2020). Before protection, the areas were under intensive management, so that only 7% of the network area was covered with stands > 100 years old. The rest of the area comprised open clear-cuts (10%), and regenerating stands of different age. Importantly for this study, 45% of the network area were artificially regenerated stands, mostly of Norway spruce (Lõhmus et al. 2020).

We sampled five ‘stand types’ that were distinguished according to stand age, origin, and management history: (1) old deciduous-dominated stands and (2) old conifer-dominated stands (> 90 years age), with no thinning or sanitary cutting in the last 20 years (6 and 4 stands, respectively); (3) 10 mixed natural stands of age 30 to 70 years; (4) 10 artificially cultivated spruce stands of age 30 to 70 years; and (5) 10 recent clear-cuts either without or < 20 years old tree layer (Pass et al. 2022). Stand selection was based using the state forest survey and cuttings databases (provided by the Estonian Environmental Agency and the Estonian Environmental Board).

The size of each individual plot was 5 ha and if possible, a cluster of plots consisting each type was located in the same reserve (Fig. 1). The distances between nearest plots within a cluster varied between 100 m and 5 km, depending on the landscape structure. The plots were near-equally divided between eutrophic (22 stands; Aegopodium site type) and meso-eutrophic forests (18 stands, mostly Hepatica site type; sensu Lõhmus 1984). The most frequent dominant tree species in the stands was Norway spruce (21 stands), followed by birch Betula sp. (12), Scots pine (4), and European aspen Populus tremula L. (3).

Fig. 1
figure 1

Locations of the 40 study sites in Estonia. The zoom-in exemplifies a site cluster of four 5-ha sites.

Bird counts

In each of the 40 plots, breeding birds were surveyed in May–June 2020 by one of two observers; each observer studied a similar mixed set of stand types. We used a standard two-visit survey: one visit between 10 and 20 May and the second between 25 May and 10 June (Rosenvald et al. 2011). Each visit comprised two mapping censuses: a standard morning census between the sunrise and 10 a.m., supplemented by an evening census between 6 p.m. and the sunset (see Tomiałojć and Lontkowski 1989 for the rationale). In each of the four censuses, the observer mapped all birds and their behaviour within the 5-ha plot and around its borders by walking slowly at approximately 50–70 m distances (ca. 10–30 min spent per hectare, depending on visibility).

The field maps were later interpreted by the author E.P. for nesting territories (‘pairs’) using conventional criteria and a common procedure (e.g., Wechsler 2018). The interpretation first distinguished the most obvious territories based on nests or behaviour directly related to nesting (e.g., nest building), or singing or alarmed adults or pairs at the same spot in subsequent censuses. If two or more males of the same species were singing simultaneously on the same visit, these were counted as separate territories. The following observations were counted as 0.5 territories: (i) territorial birds moving across plot borders; (ii) species with large home range if the observations indicated a nesting territory but no nest was found (e.g., woodpeckers and raptors); (iii) multiple observations of a species that only forms unstable pairs (e.g., Scolopax rusticola L.).

The tests of these field and interpretation methods in Estonian forests have shown that, under bird-rich conditions (such as in this study), the method provides slightly (ca. 20%) lower abundance estimates than a standard multiple-visit mapping (Lõhmus and Rosenvald 2005; Lõhmus 2020). However, since the surveys were performed in the same year and by the same observers, the relative assemblage characteristics should be well comparable among sites.

Stand structure measurements

Each stand was described for the structure of the woody vegetation and dead wood, following the procedures described by Lõhmus and Kraut (2010). In a central 2-ha area in each plot, we pre-planned, in a geographical information system, four straight 50-m spaced-out transect lines. Along each 50-m transect, all living trees (diameter at breast height, DBH ≥ 10 cm), snags (DBH ≥ 1.0 m; including whole standing dead trees) and windthrow mounds (≥ 30 cm in height) were described on strip transects at both sides of the line; coarse downed dead wood items (diameter ≥ 10 cm) were registered at crossings with the transect lines; and fine woody items (> 0.3 cm in diameter) were censused in 1 × 1 m subplots. Tree species, diameter, and condition parameters (such as decay stage of dead wood) of each item were described.

For this paper, we used the plot-scale estimates (the four transects pooled) of the volumes of live trees and of coarse downed dead wood, and the densities of snags, windthrow mounds, and saplings per hectare. Forest stand age (mean age of overstorey trees), overstorey tree species per plot, and proportion of overstorey deciduous tree species (hereafter: deciduous component) were calculated using the data from the national forest survey database of the Estonian Environment Agency.

Linear modelling of the assemblage characteristics

We used general linear models (GLM) to analyze three bird assemblage characteristics as dependent variables: density (no. of nesting territories ha−1); resident bird density (a subset of the former, including partial short-distance migrants that are abundant year-long (see suppl. data) and species richness (total number of species having nesting territories in the plot). Prior to a multi-factor approach, we explored one-way ANOVA models for the stand type effects (Tukey HSD post hoc comparison for pairwise differences). All the models were run with STATISTICA 7.

To explain the relationship between these assemblage characteristics and forest structure (study question II), we used type III sums of squares GLM analyses. Due to fundamental structural differences, clear-cuts were not included in these models. The assumptions of normal distribution of the errors were checked, as well as correlations between the independent factors (forest structure). The continuous factor variables included were: (i) stand age (years); (ii) downed dead wood (m3 ha−1); (iii) tree stock (volume m3 of live trees ha−1); (iv) deciduous component (% deciduous trees in the overstorey); (v) number of tree species, and (vi) snag density (no. ha−1).

For each bird assemblage characteristic, we performed two analyses. First, as a basic model, we analyzed the set of independent variables, except “stand age” and “tree stock” that were closely (p < 0.001) correlated with each other (r = 0.61) and with the downed dead wood (stand age: r = 0.68; tree stock: r = 0.58). Then, we explored alternative models where the stand age or tree stock were added to the basic model. We report the alternative models only if the addition was either significant or exposed an effect of another, previously non-significant variable.

Species composition analyses

We explored differences in the bird species composition among stand types and their specific indicator species using PC-ORD 6.0 program (McKinney and Lockwood 2011). Separate analyses were carried out with and without clear-cuts, which had distinctly lower bird densities and species richness than stocked stands. The data matrix included the number of territories of each species by plot, excluding 13 species that were observed in only one plot.

We tested the compositional differences between stand types using multi-response permutation procedures (MRPP) with Sørensen (Bray-Curtis) distance measure. The differences were illustrated using non-metric multi-dimensional scaling (NMS). The number of axes was chosen with medium autopilot mode. Three sets of NMSs with real (and randomized) data with 250 runs each were performed manually. Linear correlations between the ordination scores and forest structural variables were calculated; the variables considered included in the GLM analyses (Section 5.4), as well as shrub cover (%), canopy cover (%), sapling density (no. ha−1), and windthrow density (no. ha−1).

To distinguish stand type-specific (characteristic) species, we calculated the indicator values, which combines the abundance and frequency of each species by pre-defined classification of sites (Dufrêne and Legendre 1997). Their p-levels in PC-ORD are estimated for the maximum observed value by comparing these with the mean and variance of 1000 Monte Carlo permutations. The values calculated are not originally corrected for multiple tests, and should be interpreted together with the MRPP tests to reduce the type I error probabilities.

Results

In the 40 5-ha plots, a total of 1386 nesting territories of 64 species were distinguished (6.9 territories ha−1). These numbers included 460.5 territories (33%) of 26 species of resident birds. Red crossbill (Loxia curvirostra L.) was also a possible breeder (< 5 pairs), but since its fledglings were already dispersed by the survey time, we omitted this species from the analysis.

The stand types differed significantly in terms of the overall breeding density (ANOVA: F4, 35 = 25.2; p < 0.001), species richness (F4, 35 = 26; p < 0.001), and resident bird density (F4, 35 = 39.6; p < 0.001) (Fig. 2). Clear-cuts had significantly smaller values of all these plot-scale characteristics (Tukey HSD post hoc tests p < 0.05); for example, the mean total densities between clear-cuts and late-successional forests differed on average by 7.3 nesting territories ha−1. In contrast, no significant contrasts were detected between planted spruce stands and successional naturally regenerated stands. Late-successional conifer stands were represented with the smallest sample and the confidence intervals of the estimates were wide; nevertheless, their differences from planted spruce stands also appeared relatively small (no significant post hoc contrasts).

Fig. 2
figure 2

Mean (± 95% CI) density (a) and species richness (b) of breeding birds in 5-ha plots by management type: CC, clear-cut; SP, planted spruce stands; NR, successional naturally regenerated stands; LSC, late-successional conifer; LSD, late-successional deciduous

The basic models explaining those differences in 30 forest plots (clear-cuts omitted) showed that all the three bird assemblage characteristics increased along with the downed dead wood volume; the species richness also responded to the “deciduous component” (Table 1). Adding to the basic model “stand age” and, in the case of resident bird density, also “tree stock”, further explained the variation, but (i) the downed dead wood variable lost its significance, and (ii) previously hidden positive influences were exposed in the breeding density model (“deciduous component”) and in the species richness model (“tree species/ha”) (Table 1). Notably, “stand age” was not significant for the resident bird density, even if “tree stock” was removed from the model.

Table 1 General linear models of the effects of stand characteristics on total and resident bird density (no. of pairs ha−1) and species richness at the 5-ha plot scale

The MRPP test confirmed that bird species compositions differed among the stand types (p < 0.01), except between the pairs of naturally regenerated successional vs. late-successional deciduous stands (p = 0.59), and planted spruce vs. late-successional conifer stands (p = 0.32). In terms of the ordination axes, the bird species composition varied along with the stand age, tree stock, downed dead wood, and the deciduous component (Fig. 3).

Fig. 3
figure 3

NMS ordination graphs of bird communities in all 40 sites (a), and for the subset of 30 sites with tree cover (b). Each symbol refers to one 5-ha site. Two most informative axes (% variance explained indicated) of three-dimensional solutions are shown. Environmental gradients are depicted as arrows proportional in length to their co-variation with the axes (those with r2 > 0.1 shown). DDW, volume of downed dead wood

Out of 52 bird species included in the indicator species analysis, 23 species showed significant stand-type associations. Fifteen species were primarily associated with late-successional deciduous stands, three species with late-successional conifer stands, and five species with clear-cuts. Of the 18 species associated with late-successional stands, ten species were resident birds, while no resident species was associated with clearcuts (Table 2). No significant indicator values were detected for either planted spruce or naturally regenerated stands. However, both types of successional stands pooled, common chiffchaff (Phylloscopus collybita Vieill.), and icterine warbler (Hippolais icterina Vieill.) reached significant indicator values, and all 2.5 pairs of the hazel grouse (Tetrastes bonasia L.) were found there.

Table 2 Bird species associated with certain stand types (resident species in bold). “Observed IndVal” refers to the association strength with the preferred stand type, i.e., the maximum indicator value observed (Dufrêne and Legendre 1997)

Discussion

The structure of forest bird assemblages

A major result was that the forest bird densities, species richness and resident bird densities were roughly ordered at the scale of naturalness (Fig. 2) and stand age (Fig. 2; Tables 1 and 2). The average values indicated that planted spruce stands had the most depauperate bird assemblages among stocked stands, and the values grew with the stand age (study question I). Surprisingly, the MRPP test did not support species composition differences between the planted and late-successional conifer stands, which can be partly due to our small sample of the latter (i.e., low test power). Nevertheless, all conifer stands obviously provide some distinct conditions, such as conifer seeds, year-round shade and, in older stands, bark beetle infestation sites (Hughes and Hudson 1997; Williams et al. 2017; Vélová et al. 2021). Considering also the influences of stand structural factors, we conclude that bird assemblages in ageing, structurally rich spruce plantations can develop toward native late-successional conifer stands (Steverding and Leuschner 2002; Yamaura et al. 2019).

Thus, spruce planting after clear-cutting transforms bird assemblages fundamentally for at least one tree generation. This is distinct from a site-type specific development in naturally regenerated stands where the general bird composition resembled (impoverished) late-successional deciduous stands. There can be species-level exceptions, however. Notably, in line with Swenson and Angelstam (1993), we found that the hazel grouse (Tetrastes bonasia) inhabited successional stands of both regeneration types.

For interpreting the absolute densities recorded, one must acknowledge possible census bias. Our estimates seem consistent with the previous research in Estonia, although they can still be best comparable with the studies using the same methodology. Of these, Rosenvald et al. (2011) found slightly higher bird densities in boreo-nemoral old-growth stands (12 pairs ha−1) than we had in our late-successional stands (10.2 pairs ha−1). This may indicate that the densities still increase between 100 and 200 years of stand age (e.g., Zawadzka et al. 2018). The fact that siskin (Spinus spinus L.) and goldcrest (Regulus regulus L.) numbers were remarkably higher in our study may refer to a favourable year for those species. In various clear-cuts of the same site types in Estonia, Rosenvald and Lõhmus (2007) documented 1.6–3.7 pairs ha−1, while our estimate was 3.2 pairs ha−1. Moreover, using the multiple-visit mapping method, Lõhmus (2020) found similar densities in mid-aged spruce plantations (5.5 pairs vs our 6 pairs ha−1).

Also, the effects of forest structure were broadly consistent on different bird assemblage characteristics (question II; Fig. 3). An enrichment was provided by old-forest elements (downed dead wood, large and old trees; Nilsson 1979b; Mag and Ódor 2015) combined with multiple tree species, notably the deciduous component. Those structures are likely to improve various conditions, notably food resources, shelter options and nest sites for birds (Nilsson 1997; Rosenvald et al. 2011; Vélová et al. 2021). Interestingly, the deciduous component had an independent effect on bird species richness while accounting for tree species richness. This may indicate a role of certain deciduous species, e.g., for cavity nesters or arboreal foragers. Similarly, living tree stock had a positive influence on resident bird density, even when accounting for the stand age that has been previously stressed as important for resident bird communities (Imbeau et al. 2001; Zawadzka, et al. 2018). In contrast, stand age or living tree stock fully explained the effect of downed dead wood, supporting the idea that the downed dead wood rather reflects a complex of successional changes (Lõhmus et al. 2021). Contrary to numerous studies (e.g., Gutzat and Dormann 2018; Lewandowski et al. 2021), snag abundance was not related with bird assemblage characteristics.

As expected, late-successional stands hosted the largest numbers of characteristic species (question III; Table 2), but new information was that these were largely resident species. Characteristic migrants nevertheless included some previously identified focal species for conservation, such as the red-breasted flycatcher (Ficedula parva Bechst.) (Lõhmus et al. 2020) and greenish warbler (Phylloscopus trochiloides Sund.) (Väli and Vaan 2020; Lõhmus et al. 2021). However, several species identified as old-growth associated by Rosenvald et al. (2011) were rarely encountered in our study, e.g., three-toed woodpecker (Picoides tridactylus L.) and long-tailed tit (Aegithalos caudatus L.). A reason may be that most of our late-successional sites were small fragments retained in production forest landscapes. Thus, a high density and species richness in our late-successional stands might be due to the surrounding managed landscape (Wesolowski 2007) and edge effects (Gimenes and Dos Anjos 2003; Batáry et al. 2014).

Implications for sustainable forest management

Our main implications are related to the overall bird abundance along forest succession, and to the specific factors shaping habitat quality in spruce plantations. First, we found that by clear-cutting a late-successional stand, on average 6–8 pairs are lost per hectare in the productive sites studied. The post-clearcut bird densities recovers rapidly up to ca 30-year-old stands when the densities are, based on our study, roughly 65% of late-successional levels and the compositions remain impoverished. After that, the recovery rate is only about one pair ha−1 per 20 years up to at least 100 years of stand age, when the forests still lack most old-growth specialist species (Rosenvald et al. 2011). Such a pattern appears despite a structurally rapid forest recovery on fertile soils (Nilsson 1997; Lõhmus et al. 2020; 2021). We thus concur with other studies (Schmiegelow and Mönkkönen 2002; Fraixedas et al. 2015; Zawadzka et al. 2018) on the importance of old-growth reserves, but we separately highlight the issue of overall losses of bird abundance on production forest landscapes.

Our silvicultural implications are most relevant for even-aged systems. It is well known and confirmed in our study (see also Hansson 1994; Lõhmus 2004), that clear-cuts have distinct bird assemblages of mostly common, but also some declining open habitat or shrubland species (such as Carpodacus erythrinus Pall.). A fundamental difference emerges in the thicket phase, depending on whether the stand is naturally regenerated or planted with spruce. Integrating planted stands into landscapes and retaining a share of natural regeneration would constitute landscape-scale precautionary approaches against biotic homogenization.

Also, the habitat value of spruce plantations can still be facilitated through later management. The ultimate range of techniques depends on the economic options available for the managers, but our study helps prioritize the actions by their ecological effectiveness. A priority for birds would be to establish deciduous gaps in homogenous conifer plantations (Lindbladh et al. 2017; Horák et al. 2019; Vélová et al. 2021), with a long-term aim to increase tree species richness and abundance of deciduous trees. Another major effect (notably on resident birds) is provided by the presence of large trees, which can be achieved through long-term planning of the retention forestry techniques (Rosenvald and Lõhmus 2007; Mag and Ódor 2015; Rosenvald et al. 2019; Gustafsson et al. 2020). Thus, managing even-aged successional forests for birds starts from the clear-cutting stage. The thinning operations in older stands can be crucial for the development of dead wood pools and retaining cavity trees, but their potential has already been shaped by earlier silvicultural decisions.

Conclusions

1. Bird assemblages in planted spruce forests were compositionally distinct compared with naturally regenerated mixed stands and impoverished compared with late-successional forests.

2. Due to relatively rapid stand development on productive soils, avian assemblages can host some old-forest species within even-aged systems, although full old-growth assemblages probably further develop in stands well over 100 years old and some species cannot inhabit the production landscape mosaics.

3. Both in planted and naturally regenerated stands, the main silvicultural targets of bird habitat quality are similar (tree species mixtures; large-tree retention; dead wood).