Introduction

A considerable proportion of Central European forests consists of planted, even-aged, and pure stands with coniferous tree species such as Norway spruce (Picea abies (L.) Karst.) or Scots pine (Pinus sylvestris L.) (Spiecker 2003; Messier et al. 2022) due to land-use and forestry history, respectively (Zerbe 2023). Over the past decades, an ongoing decrease in tree vitality and an increase in tree mortality has been observed, particularly in pure and structurally uniform Norway spruce stands (Marini et al. 2017; Bałazy et al. 2019; Bolte et al. 2021). These phenomena are often associated with the interacting effects of abiotic and biotic stressors (Gardiner et al. 2010; Groot et al. 2019) that are expected to further increase in the future due to climate change (Seidl and Rammer 2017; Bolte et al. 2021).

Since the paradigm shift towards near-natural silviculture in Central Europe in the 1990s, a major goal of current forestry is the conversion and restoration of pure and even-aged coniferous stands towards more diverse and structurally rich mixed forests (Klimo et al. 2000; Ammer et al. 2008; Barsoum and Henderson 2016). Compared to monocultures, mixed forests are thought to exhibit a higher adaptation potential and greater resilience against the effects of climate change (e.g., drought, storms, weather extremes), pest outbreaks, and other abiotic and biotic disturbances (Bolte et al. 2009; Neuner et al. 2015; Bauhus et al. 2017; Jactel et al. 2021). Furthermore, mixed forests may provide more accompanying ecosystem services beyond timber production (Gamfeldt et al. 2013; Schuldt et al. 2018). Recent research shows that mixed forests can produce comparable or even more biomass (Forrester and Bauhus 2016; Pretzsch et al. 2017; Zeller et al. 2018), however, not inevitably concerning harvested products (Puettmann et al. 2016). Mixed forests also contribute more to multifunctionality and social acceptance (Williams 2014; Messier et al. 2022) and can harbour and conserve greater biological diversity (Ampoorter et al. 2020). The positive effects of tree species diversity depends, however, on the environmental conditions and species mixtures (Ammer 2019b) and varies in space and time (Mina et al. 2018).

The long-term conversion of monospecific coniferous forests in Central Europe is often achieved by selectively felling individual trees that have reached a species-specific target diameter to create small canopy gaps that facilitate artificial or natural regeneration (Zerbe 2002; Spiecker et al. 2004; von Lüpke et al. 2004). Alternative approaches range from clearcutting (e.g., following calamities) and replanting of target species and extend to favouring natural regeneration under existing canopies (Stanturf et al. 2014; Hansen and Spiecker 2015). Forest conversion aims at modifying the homogenous forest structure and tree species composition of even-aged coniferous plantations (von Lüpke et al. 2004; Schall and Ammer 2013; Dieler et al. 2017). These modifications change abiotic site conditions and thus resource availability of e.g., light, water, nutrients, and space for species regeneration (Augusto et al. 2003; Barbier et al. 2008; Fischer et al. 2016; Kremer and Bauhus 2020). Depending on the dimension of canopy cover reduction or gap size increase, specific forest structures and tree species are promoted. For instance, pioneer tree species such as rowan (Sorbus aucuparia L.) or silver birch (Betula pendula Roth.) benefit from higher light availability due to more vigorous canopy reduction (Yamamoto 2000; Huth and Wagner 2006; Dobrowolska 2008). In contrast, shade-tolerant species like European beech (Fagus sylvatica L.) profit from less vigorous canopy reduction (Wagner et al. 2010; Fischer et al. 2016). According to the “heterogeneity-diversity relationship”, a high structural diversity is often accompanied by high levels of biodiversity, particularly regarding vascular plants (Stein et al. 2014; Heidrich et al. 2020; Oettel and Lapin 2021). Compared to a combination of pure stands, mixed stands may increase the landscape-scale diversity of forest understoreys (Simmons and Buckley 1992; Gosselin et al. 2017). However, previous studies found that landscape-scale mixtures of tree species promoted understorey diversity better than stand-scale mixtures (Cavard et al. 2011; Heinrichs et al. 2019) challenging the beforementioned results.

To our knowledge, studies on changes in tree species diversity and composition as well as forest structure in the context of the conversion of pure coniferous towards diverse mixed forests of almost three decades are scarce. To fill this gap, we investigated such changes in the Bavarian Spessart mountains in southwest Germany. We compared our results with a large-scale survey from the 1990s (Zerbe 1995, 1999) and assessed them against the background of natural ecological processes and forest restoration activities. Specifically, we expected that during the last three decades former even-aged and pure coniferous stands increased in i) stand structural heterogeneity, ii) tree species richness and diversity, iii) abundance and frequency of broad-leaved tree species, and iv) heterogeneity of tree species composition. From these results, we assess and discuss the status quo of forest conversion in the Spessart mountains and derive implications for forest restoration management in the lower mountain ranges of Central Europe which are still largely covered by pure and even-aged coniferous stands, particularly of increasingly susceptible Norway spruce.

Material and methods

Study area

The study area Spessart is a low mountain range with elevations up to 586 m a.s.l.. It is located in north-western Bavaria in southwest Germany and stretches in the north into the adjacent federal state of Hesse (Fig. 1). The bedrock consists mainly of red sandstone and is locally influenced by loess (Matthes and Okrusch 1965). Thus, acidic and nutrient-poor soils prevail throughout the mountain range. The climate is sub-oceanic with mean annual precipitation ranging from 700 mm in the lower mountain elevations (< 300 m a.s.l.) to about 1000 mm at higher elevations (reference period: 1971–2000, Bayerisches Landesamt für Umwelt (2021)). The mean annual temperatures follow the same gradient ranging from 8 to 9 °C in the valley of the river Main to around 7 °C in the upper Spessart (Zerbe 1999). The natural forest vegetation is mainly beech forest (Fagus sylvatica L.) on acidic soils (Zerbe 1999; Weichhardt-Kulessa 2011). Spessart forests historically experienced many phases of anthropogenic influence, with a spatially varying impact of forest glassworks, hunting, agriculture as well as litter gathering and subsequent afforestation of degraded forest sites about 200 years ago. Particularly related to the history of hunting, this impact resulted in a distinguished separation of main forest types in the northern and southern Spessart. Consequently, pure and even-aged stands of Norway spruce, Scots pine, and European larch (Larix decidua Mill.) are the dominant forest types in the north, while in the south, near-natural broad-leaved forests with European beech and oak (mainly Quercus petraea (Matt.) Liebl.) prevail. Furthermore, Douglas fir (Pseudotsuga menziesii (Mirbel) Franco) has been introduced and managed in the Spessart for more than 150 years (Mergner 2018; Lange et al. 2022).

Fig. 1
figure 1

Map of the study area Spessart with reference to the geographic location in south-western Germany (upper left corner) and distribution of sampling plots

Full size image

Study design and data collection

We re-identified the geolocation of sampling plots within coniferous stand types from detailed forestry map records (spatial resolution of 1:10,000) of the initial studies by Zerbe (1995, 1999). Therefore, we determined geographic coordinates by comparing the historical map information with digital topographical maps (e.g., by major landmarks, isolines, road network) and forestry geodata (e.g., by forest departments, districts, sections, stand numbers, forest roads, paths, skidding trails) via the geographic information system Quantum GIS (version 3.22; QGIS Development Team 2022). In total, we re-identified the geolocation of 108 sampling plots from which the majority were Norway spruce-dominated stands (n = 62), while Scots pine- (n = 27), European larch- (n = 9), and Douglas fir-dominated stands (n = 10) contributed less to the total number of sampling locations (Fig. 1). Due to the varying silvicultural treatments of previously mature stands since the 1990s (c.f. Zerbe 1999), stand ages differed among the sampling locations at the time of the resurvey. We assessed current stand ages from detailed forest management plans (Bavarian State Forest Research Centre, unpublished) and verified them approximately by expert-based estimation in the field. At the time of our resurvey, the stand ages ranged from 13 to 180 years with a mean stand age of 98 years according to the management plans.

Within each stand, we established a sampling plot with a size of 20 m × 20 m (400 m2), which corresponds to the size of the first survey. The exact location was indicated by a handheld GPS device (Garmin eTrex 30x). We adopted site characteristics (i.e., slope [°], elevation [m a.s.l.], terrain location, aspect) from Zerbe (1999) and double-checked the information in the field to minimize potential relocation errors (Kapfer et al. 2016). We did not resurvey plots if there was an inconsistency between GPS-indicated locations and site characteristics of the original plot. We estimated the possible relocation error to be a few tens of metres and considered this accuracy adequate, given the rather spatial homogeneity of the forest stands and the reasonably large sampling plots. We visually assessed the plot-wise assemblage of woody plants (i.e., species identities and respective cover value classes) as well as forest structure (i.e., occurrence and total cover [estimated in %] of each forest layer (i.e., shrub (1 – 5 m growth height), lower canopy (≥ 5 m – 15 m growth height), upper canopy (> 15 m growth height) according to the vegetation sampling method by Braun-Blanquet (1964). For instance, a given sampling plot with an estimated cover value of 0% indicates the complete absence of living biomass, while a cover value of 100% depicts the maximum coverage of living biomass within a given forest layer. We standardized the nine ordinal species cover value classes to intermediate class decimal values (Tremp 2005). The nomenclature of plant species follows Jäger (2017). We conducted field work in summer 2021 and 2022.

Forest management in the past 30 years

To evaluate the data besides natural ecological processes also in the context of silvicultural activities, extensive records from management chronicles of the Spessart forest offices have been sighted, compiled, filtered, and processed concerning wood extraction, tree plantings, soil preparation, fertilization, weeding, and fencing. During the evaluation process it became apparent that the raw data were deficient for further statistical analyses due to several reasons: some foresters wrote down detailed management decisions, some either more or less, some did not. For some forestry districts such records were just not available. Lacking data were probably resulting from different forestry reformations and several consolidations of forest offices in the Spessart during the last decades. Furthermore, the demarcation and sizes of stands changed over time and did not allow for a precise comparison of management data. Thus, a consistent and reliable management history was not possible to reconstruct for a statistically sound analysis. However, qualitative insights from the researched management chronicles are incorporated in the discussion section. Generally, forest conversion management applied in our study area comprised the following options: a) single-tree selection to promote natural or artificial regeneration in small canopy gaps, b) planting seedlings of coniferous and broad-leaved tree species in advance below the canopy of mature trees, and c) sanitary cuttings after extreme events (e.g., storm damage, insect calamities, etc.) followed by natural regeneration and/or planting of target tree species (Möges and Zanker 2008; Treutlein and Achhammer 2018; Management chronicles unpublished). Consequently, since the 1990s, there has been no one-size-fits-all conversion method applied. However, our study results depict a typical situation of adaptive forest management in an extensive montane forest landscape in Central Europe (Bolte et al. 2009; von Lüpke et al. 2004).

Data analysis

Concerning the assessment of species diversity, we followed the Hill number framework, specifically the three estimators proposed by Chao et al. (2014). Thus, we calculated plot-wise species richness (q0), the exponential of Shannon’s entropy index (q1; hereafter: true diversity), and the inverse of Simpson’s concentration index (q2; hereafter: Simpson diversity) for both sampling periods. While species richness is biased towards rare species, very sensitive to sample size, and offers no information about relative abundances, the frequencies of the more common species get more weight with increasing Hill number (Chao et al. 2014). We then calculated the delta (Δ) of Hill numbers and stand structural attributes between both sampling periods to obtain a measure of temporal change per plot.

We tested, using simple linear models (Chambers 1992), changes in Δ mean cover of forest layers, species richness (q0), true diversity (q1), and Simpson diversity (q2) per plot against the null hypothesis of no change over time for i) all plots and ii) stand types separately. This approach enabled us to capture the overall temporal trend and simultaneously identify the specific stand types that drive the overall trend. We considered the upper canopy layer (UCL; > 15 m), lower canopy layer (LCL; > 5–15 m), and shrub layer (SL; 1–5 m) each separately as well as the combined total canopy layer (TCL; 1 – > 15 m). Furthermore, we calculated the cumulated cover of all three forest layers (hereafter: total cover) according to Ewald et al. (2011). This approach prevents an overprediction of the cumulated cover by subtracting the overlap of forest layers from their sum: total cover [%] = (UCL/100 + LCL/100 + SL/100 – UCL/100 × LCL/100 × SL/100) × 100. After fitting simple linear models, followed by ANOVA-testing (and in the case of stand types), Tukey HSD post hoc tests for pairwise comparisons among stand types were performed to avoid multiple testing (Benjamini and Hochberg 1995). Some response variables of the aforementioned models had to be rank-transformed because they did not meet the parametric assumptions on normally distributed residuals or homogeneity of variance.

To test for the significance of tree species compositional change, we computed a repeated measurement permutational analysis of variance (PERMANOVA) with the function adonis2 of package vegan (Oksanen et al. 2022) using 999 permutations and Bray–Curtis dissimilarity. This performs a permutational test that uses distance matrices of the tree species composition of plots to find significant differences among sampling periods (Anderson 2001). We used the R2 values from the PERMANOVA models as a measure of the magnitude of temporal change to compare between stand types. To examine the temporal shift in β-diversity (i.e., the variability in species composition among stand types) we used permutational analysis of multivariate dispersion (PERMDISP) with the function betadisper (Anderson 2006) of the package vegan (Oksanen et al. 2022). This analysis evaluates the homogeneity of within-group dispersions based on Bray–Curtis distances with significance testing using permutation. An increase in multivariate distance between sampling plots and the time-specific centroid is explained as biotic differentiation, while a decrease suggests biotic homogenization.

We visualized the variation in tree species composition (cover values of each species) within and among the four different stand types for both sampling periods by ordination [Non-metric Multidimensional Scaling (NMDS)]. We applied the NMDS with two dimensions using the function metaMDS of the package vegan (Oksanen et al. 2022). Stress values smaller than 0.20 generally result in viable interpretations (Clarke 1993). We applied the function envfit to fit explanatory vectors of stand structural attributes onto the ordination.

To test the affiliation of woody species in coniferous stands either to the 1990s or 2020s and thus identify “winner” and “loser” species, we conducted an Indicator Species Analysis (ISA; Dufrêne and Legendre 1997) on species-specific cover values with the function indval of the package labdsv (Roberts 2022). The ISA calculates an indicator value for each species based on the product of its relative frequency and its relative mean abundance across plots and sampling periods. The statistical significance of indicator values is assessed via permutation tests. The ISA was applied separately to the upper and lower canopy, as well as to the shrub layer.

We performed all statistics using R, version 4.1.1 (R Core Team 2022).

Results

Forest structure

The occurrence of an upper canopy layer on a given plot decreased from 108 to 94 plots between the 1990s and 2020s. This decrease occurred predominantly in spruce stands (n = 13) and only little in pine stands (n = 1). The occurrence of a lower canopy layer increased from 41 to 106 and the shrub layer occurrence increased from 74 to 105 since the 1990s (Fig. 2).

Fig. 2
figure 2

Occurrence of forest layers in 108 sampling plots between the sampling periods 1990s and 2020s, with upper canopy layer (> 15 m), lower canopy layer (> 5–15 m), and shrub layer (1–5 m)

Full size image

The mean cover of the upper canopy layer per plot decreased significantly by 15.3% ± 2.7 (mean ± SE) between the two sampling periods (t = -5.79, P < 0.001). This overall trend was mainly driven by the reduction of the spruce stands (− 31.7% ± 2.5), while upper canopy cover increased significantly for the pine stands (+ 11.0% ± 3.8; Fig. 3B, Table 1). There was no significant change in upper canopy cover of the Douglas fir (decreasing trend) and larch stands (increasing trend). The mean cover of the lower canopy layer per plot significantly increased by 37.8% ± 2.9 from the 1990s to the 2020s (t = 13.20, P < 0.001). This increase occurred over all four stand types and was strongest in the spruce stands (+ 44.0% ± 3.4), intermediate in the Douglas fir (+ 36.6% ± 9.2) and larch stands (+ 33.6% ± 9.7), and weakest in the pine stands (+ 22.5% ± 5.6) (Fig. 3C, Table 1). The mean cover of the shrub layer per plot significantly increased by 12.3% ± 1.3 between both sampling periods (t = 9.63, P < 0.001). The increase occurred over all four stand types (+ 6.6% ± 1.6 to 14.8% ± 4.3) (Fig. 3D, Table 1). Taken together, the mean total cover by woody plants (upper canopy + lower canopy + shrub layer) per plot significantly increased by 22.2% ± 3.5 and similarly between the stand types (+ 16.4% ± 3.1 to 37.9% ± 8.1) since the first survey (t = 18.08, P < 0.001; Fig. 3A, Table 1).

Fig. 3
figure 3

Change (Δ) in the mean cover [in %] of forest layers (A = total canopy, B = upper canopy, C = lower canopy, D = shrub layer) per plot between the 1990s and 2020s. The left part of each sub-figure (“Plots”) shows the overall change for all sampling plots (n = 108). The right part (“Stand types”) depicts the change for each stand type (Douglas fir [n = 10], spruce [n = 62], pine [n = 27], and larch [n = 9]. The grey horizontal line represents the Null hypothesis of no temporal change. Positive values imply an increase, negative values a decrease in the cover of forest layers. Significant effects are marked with stars: *** P < 0.001, ** P < 0.01, * P < 0.05. Bars with different letters are significantly different (Tukey HSD post hoc comparisons)

Full size image
Table 1 Absolute temporal changes in plot-wise mean cover values [in %] of the different forest layers for i) all sampling plots and ii) stand types
Full size table

Tree species richness and diversity

The initial survey in the 1990s identified 20 woody species across all forest layers in 108 sampling plots. The total species number was eleven in the upper canopy, six in the lower canopy and 18 in the shrub layer. Our resurvey yielded a total of 30 woody species in all forest layers. Across all plots and stand types, the upper canopy had ten, the lower canopy had 20, and the shrub layer had 28 woody species. Between both sampling periods, the total number of woody species remained approximately constant for the upper canopy, however, increased for the lower canopy (+ 14 species), the shrub layer (+ 10 species) as well as the combination of all three forest layers together, i.e., the total canopy (+ 10 species).

The mean species richness of the total canopy layer per plot significantly increased by 2.1 ± 0.2 (mean ± SE) between both sampling periods (t = 9.06, P < 0.001). Spruce stands contributed most to the increase in species richness (+ 2.8 ± 0.3), Douglas fir (+ 1.5 ± 0.7) and pine stands (+ 0.9 ± 0.4) had a less strong influence, while larch stands (+ 1.3 ± 0.8) did not change significantly over time. The total canopy layers' mean true diversity (t = 9.20, P < 0.001) and Simpson diversity (t = 8.45, P < 0.001) per plot increased significantly by 0.9 ± 0.1 and 0.7 ± 0.1 from the 1990s to the 2020s. Spruce (+ 1.1 ± 0.1; + 0.8 ± 0.1) and Douglas fir (+ 1.1 ± 0.3; + 1.0 ± 0.3) stands contributed most to the increase in diversity indices, while the increase was less strong and not significant in pine (+ 0.4 ± 0.2; + 0.3 ± 0.2) and larch (+ 0.5 ± 0.3; + 0.4 ± 0.3) stands (Fig. 4A; Table 2).

Fig. 4
figure 4

Change (Δ) in mean Hill numbers (species richness [q0], true diversity [q1], Simpson diversity [q2]) per plot for A = total canopy, B = upper canopy, C = lower canopy and D = shrub layer between the 1990s and 2020s. The left part of each sub-figure (“Plots”) shows the overall change for all sampling plots (n = 108). The right part (“Stand types”) depicts the change for each stand type (Douglas fir [n = 10], spruce [n = 62], pine [n = 27], and larch [n = 9]. The grey horizontal line represents no change. Positive values imply an increase, negative values a decrease in Hill numbers over time. Significant effects are marked with stars: *** P < 0.001, ** P < 0.01, * P < 0.05. Boxplots with different letters are significantly different (Tukey HSD post hoc comparisons)

Full size image
Table 2 Absolute temporal changes in plot-wise species richness (q0), true diversity (q1), and Simpson diversity (q2) for all sampling plots and different stand types by forest layer
Full size table

In the upper canopy layer, we found no significant differences in mean species richness (t = 1.60, P = 0.12) and Simpson diversity per plot (t = 1.33, P = 0.19) among sampling periods. However, the mean true diversity per plot increased significantly by 0.3 ± 0.1 (t = 4.98, P < 0.001). The pine stands were the only stand type that increased significantly in all three Hill numbers since the 1990s (Fig. 4B; Table 2).

In the lower canopy layer, mean species richness (t = 12.79, P < 0.001), true diversity (t = 6.45, P < 0.001), and Simpson diversity (t = 9.65, P < 0.001) per plot increased significantly by 2.7 ± 0.2, 0.5 ± 0.1, or 1.0 ± 0.2 between the 1990s and 2020s. All three variables increased in almost each single stand type over time, even though with different magnitude. The increase was strongest in the spruce stands, intermediate in the Douglas fir and larch stands, and lowest in the pine stands. True diversity and Simpson diversity of larch stands did not show significant differences between sampling periods (Fig. 4C; Table 2).

In the shrub layer, mean species richness (t = 8.16, P < 0.001), true diversity (t = 3.99, P < 0.001), and Simpson diversity (t = 18.10, P < 0.001) per plot increased significantly by 2.2 ± 0.3, 0.5 ± 0.1, or 0.6 ± 0.1 among both sampling periods. This positive trend was driven by the spruce and Douglas fir stands, while pine and larch stands did not show significant differences in any Hill number over time (Fig. 4D; Table 2).

Tree species composition

The only partial overlap of sampling plots in the NMDS graph between the 1990s and 2020s indicates changes in tree species composition (Fig. 5A). The composition of woody species in all forest layers differed significantly between the two sampling periods (PERMANOVA: F = 34.41; R2 = 0.14; P < 0.001). The ordination shows that the species composition was more heterogeneous in the 1990s than in the 2020s and thus, suggests an overall homogenization of the composition of woody species across all sampling locations. The test results verify this visual impression (Table 3; PERMDISP). The overall homogenization can be attributed to a shift from distinct conifer-dominated stands (i.e., Douglas fir, spruce, pine, larch) of the 1990s towards mixed stands with broad-leaved (e.g., beech, oak, birch) and coniferous (e.g., silver fir, spruce, Douglas fir) trees in the 2020s (Fig. 5A).

Fig. 5
figure 5

Non-metric multidimensional scaling (NMDS) ordination based on species abundance data of woody species (upper canopy + lower canopy + shrub layer) for A = each sampling plot (n = 108) and B = each stand type (n = 4) for both sampling periods (using Bray–Curtis dissimilarity, final stress for two-dimensional solution: 0.19). Sub-figure A shows additionally all significant woody indicator species (see Table 4). The ellipses show the standard deviation of plot coordinates (dark grey: 1990s, light grey: 2020s). The ellipses in sub-figure B show the standard deviation of plot coordinates per stand type and sampling period. Black arrows indicate significant correlations of forest layer coverage (ucl = upper canopy layer, lcl = lower canopy layer) with axes values (P < 0.05)

Full size image
Table 3 Test results for temporal shifts in β-diversity (PERMDISP) and species composition (PERMANOVA) of tree communities between sampling periods of the 1990s and 2020s
Full size table

The missing or only minor overlap of sampling plots of the same stand type between both sampling periods suggests a shift in the composition of woody species over time (Fig. 5B). Tree species composition within stand types changed significantly between the 1990s and 2020s for all four stand types (PERMANOVA, Table 3). The NMDS revealed a shift in species composition towards the centre of ordination space (Fig. 5B). Contrary to the overall homogenization of the composition of woody species of all stand layers across all sampling plots, spruce and Douglas fir stands experienced a compositional differentiation over time, while trends were not significant for the pine and larch stands (PERMDISP; Table 3).

Several stand structural attributes fit significantly into ordination space: envfit results showed a decrease in cover of the upper canopy layer (r2 = 0.07, P < 0.001) approximately orthogonal to an increase in cover of the shrub (r2 = 0.06, P < 0.001) and lower canopy (r2 = 0.07, P < 0.001) layer as well as the cover of the total canopy (r2 = 0.11, P < 0.001). These changes are directed roughly along the direction of compositional change. While the cover of the upper canopy layer can be mainly associated with spruce stands of the 1990s, the tree species composition of the 2020s is associated with the cover of the shrub and lower canopy layer as well as the total canopy cover (Fig. 5B).

“Winner” and “loser” tree species

In total, ten tree species significantly changed in relative frequency or relative mean abundance in the different forest layers among both sampling periods. In the upper canopy layer, we found two tree species that changed significantly over time. P. abies showed a declining trend and F. sylvatica increased. In the lower canopy layer, ten tree species (e.g., F. sylvatica, P. abies, Pinus strobus, etc.) revealed an increasing trend, while none showed a significant decline. In the shrub layer, eight tree species (e.g., F. sylvatica, P. abies, P. menziesii, etc.) indicated an increasing trend, whereas no woody species declined from the 1990s to the 2020s. Thus, almost all significant indicator tree species (except for spruce in the upper canopy) can be considered as “winner” species. Especially beech can be assessed as the overall winner because it achieved the highest indicator scores for each forest layer. Additionally, spruce may be highlighted due to its extensively increasing regeneration in the shrub- and lower canopy layer compared to the initial survey (Table 4).

Table 4 Indicator values (0 – 1) of all tree species that showed a significant (P < 0.05) preference either for the 1990s or 2020s, separated by forest layer
Full size table

Discussion

Forest structure

Our first expectation (i.e., stand structural heterogeneity of coniferous stands increased during the past 30 years) can be confirmed since the occurrence and cover of the upper canopy layer decreased, while it increased in the lower canopy, the shrub layer and across all forest layers. Especially in the spruce stands, the significant reduction in overstorey density presumably increased resource availability (e.g., light, space, water, nutrients) for the underlying forest layers and thus, stimulated artificial or natural regeneration of trees and shrubs (Augusto et al. 2003; Kremer and Bauhus 2020). The resulting increase in understorey coverage was probably further promoted due to the extensive regeneration of broad-leaved tree species. Especially the tree architecture of highly promoted European beech is known for its enormous shading potential and competitiveness against other tree species (Hagemeier 2002; Heiri et al. 2009; Mölder et al. 2014). The amplified emergence of multi-layered stands emphasise an increase in stand structural heterogeneity (Pretzsch et al. 2016; Oettel and Lapin 2021). Mixed stands with species that feature different ecological traits have been shown to enhance stand structure and heterogeneity compared to monospecific stands (Bauhus 2009; Pretzsch and Schütze 2016; Juchheim et al. 2020). The detected changes in stand structure may be seen in the context of forest management because canopy cover is an important management indicator, especially in coniferous forests (Coote et al. 2013). Also, extreme events (e.g., wind-throw, bark beetle calamity) may play a vital role as some stands in the Spessart showed clear signs of natural disturbances concerning their current structure and species composition. Based on data from the German national forest inventory, Storch et al. (2019) found that low to moderate harvesting intensities, i.e., the removal of ≤ 70% of standing timer volume, positively influenced most aspects of structural diversity. It is well known that irregular canopy openings stimulate tree regeneration locally (Wagner et al. 2011) and result in horizontally and, later, vertically structured stands. Thus, forest management can intentionally emulate small-scale canopy disturbances to promote structural diversity (Doyon et al. 2008; Moretti et al. 2010; Puettmann et al. 2016). The structural diversity of such stands increases over time if some parts are kept dense for at least one or two decades (von Lüpke et al. 2004). If, as in classical shelterwood cuttings, the entire stand was opened up, the regeneration layer would develop to a rather uniform stratum resulting in a less structured future forest stand (Meyer and Ammer 2022). Moreover, the careful density reduction in the overstorey combined with advanced regeneration in the understorey also offers the option to never fall below a certain amount of carbon stored in the living biomass (Ammer 2019a). Besides the cost-efficient use of natural regeneration, underplanting can contribute to the structural complexity of managed forests (Löf et al. 2019). In this context, the spatial and temporal heterogeneity of forest structure can promote biodiversity at the landscape scale (Ammer et al. 2017).

Generally, our results provide scientific support for the guidelines of the Bavarian State Forest Administration (BaySF) that planned to convert pure and even-aged conifer-dominated stands into more diverse and structurally rich mixed forests (Möges and Zanker 2008; Treutlein and Achhammer 2018). However, the current stand structure still seems to be in a transitional phase because the tree regeneration has in most cases not reached the upper canopy yet. This result seems not surprising because average rotation periods of Central European commercial forests may take several decades up to few centuries, depending e.g. on the specific tree species and site conditions (Pach et al. 2018). For instance, the rotation length of spruce stands often exceed more than one century in many European regions (Lindner et al. 2000). Thus, extensive time is needed for forest landscape restoration (Stanturf et al. 2019). Nevertheless, according to Bolte et al. (2021), the efforts to convert coniferous plantations need to be multiplied to reduce the damages that are expected to occur with ongoing climate change.

Tree species richness, diversity, and composition

Our second expectation (i.e., tree species richness and diversity of coniferous stands increased during the past 30 years) can be partly confirmed and particularly depends on the forest layer and measure of diversity in focus. Despite our expectations, we did not find a significant increase in species richness (q0) of the upper canopy, although the total number of woody species increased by 50% across all sampling plots since the 1990s. Apparently, forest conversion towards mature mixed forests is after about 30 years still in an early phase and re-growing trees did not reach the upper canopy yet or got outcompeted by the dominant trees. However, true diversity (q1) increased. The related increase in diversity can be explained by the combined effect of species richness and evenness (Pielou 1966) that gives more weight to the abundance of typically occurring species compared to solely species richness that weights each species equally regardless of their abundance (Chao et al. 2014). Thus, the current individual numbers seem to be distributed more balanced between the occurring species compared to the 1990s.

In contrast, species richness (q0) and diversity (q1, q2) increased in the lower canopy and shrub layer as well as across all layers and sampling plots. At this stage of forest development, the tree species in the regeneration include pioneer (e.g., birch, rowan), broad-leaved (e.g., beech, oak) and coniferous (e.g., Douglas fir, silver fir) tree species. Our findings are in line with many other studies that found canopy disturbance accompanied with increasing light availability to promote tree regeneration abundance or diversity (Canham et al. 1990; Yamamoto 2000; Jonášová and Matějková 2007; Heinrichs and Schmidt 2009; Seliger et al. 2021). Due to the increasing heterogeneity of environmental conditions in the understorey induced by canopy gaps, a diverse tree regeneration was established which supports the heterogeneity-diversity hypothesis (Helbach et al. 2022). The variability in temporal changes of Hill numbers within stand types and the overall facilitation of light-demanding as well as shade-tolerant tree species that we found highlight the significance of different stand development stages on diversity measures (Ujházy et al. 2017). The increase in all investigated measures of sub-canopy diversity suggests a development towards more diverse forests and an increase in evenness of the present species in the community.

Our third expectation (i.e., abundance and frequency of broad-leaved tree species in coniferous stands increased during the past 30 years) can be confirmed. The tree regeneration was strongly promoted as many broad-leaved and coniferous species experienced an increase in abundance or frequency since the 1990s. Only spruce in the upper canopy showed a significantly declining trend. Both results provide scientific support to the regional guidelines of forest restoration (Möges and Zanker 2008; Treutlein and Achhammer 2018). Further, the promotion of broad-leaved tree species, especially European beech, fit tendencies that have been reported for other European regions (Sterba and Eckmuellner 2008; Kudernatsch et al. 2021; BMEL 2021). The significant increase in F. sylvatica in the over- and understorey seems to be the result of both, silvicultural facilitation (e.g., by planting as documented for several stands in the management chronicles) and natural regeneration due to its strong competitiveness over most parts of Central Europe (Ellenberg and Leuschner 2010). In this regard, Axer et al. (2022) found natural regeneration from admixed European beech groups within spruce stands in the Ore mountains up to distances of 69 m and recommend groupwise mixtures of beech with a distance of 40–50 m as a restoration approach. Considering its silvicultural promotion, general competitiveness, potential dispersal distance and the fact that mature beech individuals were present in sight distance in the great majority from our sampling plots, its extensive spread appears comprehensible.

Although, mature spruce was evidently reduced in the upper canopy, the tree species regenerated in parts extensively. In this context, is has been reported that spruce has a competitive advantage against beech concerning its growth if the relative diffuse radiation is over 20% (Unkrig 1997; von Lüpke and Spellmann 1997; Kühne and Bartsch 2003). Other studies found this advantage of spruce only at a light availability that exceeds 40% (Ammer 2005). Thus, it can be inferred that a more vigorous reduction of canopy cover with increasing light availability in the understorey may lead to a promotion of spruce regeneration. Dobrovolny (2016) found that already spruce group-fellings of about 0.04 hectare, which interestingly corresponds to the plot size of our study, is enough to facilitate spruce against beech. Furthermore, in the context of extreme events (e.g., wind-throws, forest fires) that are prognosed to increase in frequency and magnitude under climate change (Seidl et al. 2017), the process of “direct re-growth” of pre-disturbance species has been reported (Romme et al. 2011). Given this, the dominant pre-disturbance species may gain a re-dominance after one or two decades without silvicultural interventions (Priewasser 2013; Kramer et al. 2014), which illustrates an unfavourable scenario in terms of current restoration objectives. In this context, regeneration of pioneer tree species may play an increasingly important role to compensate the ecological deficits resulting from the long-term cultivation of pure coniferous stands and can facilitate the development towards diverse mixed forests (Raspé et al. 2000; Zerbe and Meiwes 2000; Hynynen et al. 2010).

If the current re-growing diverse tree community is to be retained to stand maturity it probably needs active management, especially due to the strong competitive ability of beech and in parts dense regeneration of spruce. Otherwise, there may be a major risk of favouring the next generation of monospecific stand types over an extensive area. In this context, European beech admixtures of at least 30% in the future stands are recommended to have a positive influence on the understorey diversity (Walentowski 1998; Kudernatsch et al. 2021). However, very high proportions of beech may have negative effects on phytodiversity and can lead to a reduction of plant diversity (Zerbe 1993; Budde et al. 2011).

Our fourth expectation (i.e., heterogeneity of tree species composition in coniferous stands increased during the past 30 years) can be partly confirmed. Our results question the assumed increase in β-diversity as a result of an increase in tree species richness (Hilmers et al. 2018; Schall et al. 2018) for coniferous stands that are currently under conversion. Instead, we found an overall shift and the convergence of tree species composition across the forest landscape since the 1990s. The former clearly definable tree communities of pure coniferous stand types (i.e., Norway spruce, Scots pine, Douglas fir, and European larch) experienced a floristic homogenization on the landscape scale. This increasing homogeneity across the forest environment has also been reported from other parts of Europe (Kjučukov et al. 2022) and particularly from forest understoreys (Keith et al. 2009; Reinecke et al. 2014; Prach and Kopecký 2018). However, an overall floristic homogenisation of understorey vegetation was not conclusively detected in our study area but trends became apparent (Seliger et al. 2023). Here, the shift in tree species composition can mainly be attributed to an extensive increase in broad-leaved tree species, especially European beech. The former exclusively coniferous stands developed in a similar manner towards mixed forests with a higher share of broad-leaved tree species. This development corresponds largely to the objectives of forest conversion and management plans (Möges and Zanker 2008; Borrass et al. 2017; Treutlein and Achhammer 2018). Furthermore, it may be assessed beneficial from a perspective of naturalness because the native prevalent forest community in the Spessart mountains is beech forest on acidic soils (Luzulo-Fagetum community; Zerbe 1999; Weichhardt-Kulessa 2011).

In contrast to the decline of overall forest landscape heterogeneity, we found a differentiation of tree communities within specific stand types since the 1990s, especially in spruce stands. Probably due to the active (silvicultural measures) and passive (natural disturbances) transformation towards mixed forests, the relatively homogenous species composition of pure coniferous stands (Zerbe 1999; Hunter 2004) got more heterogeneous on the stand level. The resulting mosaic of remaining old stands and more or less intensively converted stands possibly led to a more heterogeneous species composition within single coniferous stand types of the Spessart. In this context, Heinrichs et al. (2019) found that tree mixtures on the landscape scale are more effective in promoting the diversity of vascular plants, bryophytes, and lichens compared to tree mixtures on the stand scale. The heterogeneity of species composition is seen as one of the key properties that make forests a complex and adaptive system (Bauhus et al. 2013; Puettmann et al. 2013).

Aiming to increase the spatial heterogeneity and diversity of tree communities, it may be neither beneficial to convert pure coniferous stands with European beech uniformly over the whole landscape nor with beech solely (Heinrichs et al. 2020; Kudernatsch et al. 2021). Instead, promoting beech groups can increase species-turnover on the stand scale and simultaneously increase landscape heterogeneity (Kudernatsch et al. 2021). According to these studies it seems beneficial to favour different stand types and development stages if the aim is to achieve a diversification on the landscape scale. This may include also to retain some old coniferous stands, given the environmental suitability of the specific tree species under expected climate change scenarios.

Conclusions

We found that the even-aged and pure coniferous stands of the Spessart mountains overall experienced a diversification in forest structure and tree communities since the 1990s. This is especially true for spruce stands, which showed the greatest change concerning stratification, tree species richness, diversity, and composition. Forest conversion seems to be in full play and observed changes trend currently towards restoration objectives, i.e., structurally heterogeneous and species-rich mixed forests. However, the studied stands are still in a transitional phase, and tree regeneration has mostly not reached the upper canopy yet. Furthermore, regeneration of spruce increased in some stands extensively and there possibly bears the risk of a direct re-growth (Romme et al. 2011) and the potential development of spruce-dominated stands if no intervention takes place. On the other hand the extensive promotion of European beech may lead to homogenous forest properties and relatively low understorey species diversity in the long term (Hobi et al. 2015). In this context, the recently observed crown dieback and mortality of beech in many parts of Central Europe after extensive drought (Schuldt et al. 2020; Obladen et al. 2021) question its future dominance. These findings underline the advantage to diversify commercial forests as a measure of risk management under climate change. Concerning the conversion of pure and even-aged coniferous stands, we propose the cost-efficient use of natural regeneration of autochthonous broad-leaved tree species, wherever possible, and supporting active silvicultural interventions in commercial stands to promote diverse and structurally rich mixed forests. In accordance with other authors (e.g., Stanturf et al. 2019; Bolte et al. 2021), we suggest that the effort to convert coniferous plantations need to be multiplied to reduce the damages that are expected to occur with ongoing climate change.