Diversity and Distributions, (Diversity Distrib.) (2010) 16, 267-276 •111 Half a century of succession in a temperate oakwood: from species-rich community to mesic forest Radim Hédl1*, Martin Kopecký1'2 and Josef Komárek3 1 Institute of Botany of the Academy of Sciences of the Czech Republic, Poříčí 3b, 603 00 Brno, Czech Republic, 2Department of Botany, Faculty of Science, Charles University in Prague, Benátská 2, CZ-128 01 Prague, Czech Republic, 3Agency for Nature Conservation and Landscape Protection of the Czech Republic, Havlíčkův Brod Regional Office, Husova 2115, 580 01 Havlíčkův Brod, Czech Republic "Correspondence: Radim Hédl, Institute of Botany of the Academy of Sciences of the Czech Republic, Poříčí 3b, 603 00 Brno, Czech Republic. E-mail: radim.hedl@ibot.cas.cz ABSTRACT Aim Lowland woodlands in Europe went through dramatic changes in management in the past century. This article investigates the influence of two key factors, abandonment of coppicing and increased pressure of ungulates, in thermophilous oakwoods. We focused on three interconnected topics: (1) Has the assumed successional trend lead to impoverishment of the vegetation assemblages? (2) Has it resulted in vegetation homogenization? (3) Are the thermophilous oakwoods loosing their original character? Location Czech Republic, Central Europe. Methods The vegetation in 46 semi-permanent plots was recorded three times: firstly, shortly after the abandonment of coppicing (1953) and then, after four to six decades of secondary succession and strong game impact (1992 and 2006). Overall trends and changes in species spectra were analysed. Results There is a marked successional shift towards species-poorer communities growing in cooler, moister and nutrient-richer conditions. The change was significantly different in parts affected and unaffected by high numbers of ungulates yet only for herbs, not the woody species. However, observed change in species composition was not accompanied by significant homogenization process that is the general process reported from elsewhere. A sharp decline in plant species typical for thermophilous woodland communities and in endangered species indicates that the original character of the woodland has been gradually lost. Main conclusions Thermophilous oakwoods have been largely replaced by mesic forests. Lowland oakwoods in continental parts of Europe historically depended on active management, which kept the understorey conditions light and warm. Successional processes in the 20th century caused a critical loss of species diversity at various spatial levels. However, artificially high numbers of ungulates, which otherwise have a negative impact, probably held up succession, so that the changes may still be reversible. Keywords Endangered species, homogenization, long-term changes, loss of diversity, natural woodland, semi-permanent plots. INTRODUCTION In densely inhabited European lowland areas, many forests have been intensively managed for millennia. Coppicing, litter raking, hay-making and wood pasture were routine practices all over Europe (Szabo, 2005; Rackham, 2006; Bergmeier, 2008; Gimmi et al, 2008). After many centuries of varying historical management (e.g. Verheyen et al, 1999; Szabo, 2010), European lowland woodlands have faced new threats in the past few decades (Rackham, 2008). Intensive biomass and nutrient output connected with a fine-scale regime of disturbances were replaced by a massive input of nitrogen (Thimo-nier et al, 1994) and other pollutants, while the management regime has changed in favour of timber production. This has gradually lead to taxonomie impoverishment and homogenization of the woodland vegetation (Keith et al, 2009), which © 2010 Blackwell Publishing Ltd DOI: 10.1111/J.1472-4642.2010.00637.x www.blackwellpublishing.com/ddi 267 R. Hédl ef al. may be part of biotic homogenization in general (McKinney & Lockwood, 1999). The abandonment of former management and the subsequent conversion to high forest, including re-planting by non-native species, have been especially prevalent in Central and Northwestern Europe. However, conservation-motivated neglect helped to preserve the ancient character of some forests, now often declared nature reserves because of their biological values. The history of the management regime has been increasingly used as a causal factor in studies of long-term changes in herb layer composition (Baeten ef al, 2009; Corney ef al, 2008; Van Calster ef al, 2008a), soil seed bank (Van Calster ef al, 2008b) and even in soil properties (Strandberg ef al, 2005). Geographically, most studies focusing on vegetation changes in lowland woodlands come from historically little-forested Northwestern Europe (Belgium, England). Information on more continental parts of Europe is largely missing. In terms of historical management practices, parallels between the Atlantic, continental and southern parts of Europe are apparent (Grove & Rackham, 2003; Loidi, 2005; Szabó, 2005). Nevertheless, the biological and ecological properties of these large biogeo-graphical regions differ in various aspects; therefore, transferring knowledge directly from west to east may be misleading. Only a few studies deal with long-term changes in the lowland woodlands of continental Europe (Chytrý & Danihelka, 1993; Kwiatkowska, 1994; Jakubowska-Gabara, 1996). The latter author attributed the decline of thermophilous oakwood communities to declining human activities, namely pasturing. The decline and abandonment of coppicing, one of the formerly common forms of woodland management in Europe (Rackham, 2006), has until relatively recently not been regarded as a plausible explanation for vegetation changes in the woods of the continental parts of Europe. Recent studies by entomologists have clearly shown a causality between a frustrating decline in the diversity of butterflies (Beneš ef al, 2006; Freese ef al, 2006; Konvička ef al, 2008), xylophagous beetles (Vodka ef al, 2009) or ground invertebrates (Spitzer ef al, 2008) and forest succession after the abandonment of coppicing. Unfortunately, we have scarce evidence for other groups of organisms and communities. In this article, we deal with a thermophilous wood intensively managed for at least six centuries as coppice-with-standards (the first management survey dates back to 1386, Bretholz, 1930; Szabó, 2010), which has undergone a major change since the mid-20th century, including establishment of a large game preserve. Milovice Wood constitutes a core locality of the subcontinental loess thermophilous oak-woods at the north-western fringe of the Pannonian Basin. It is one of the largest remaining oakwoods in this part of Europe. We used a set of re-surveyed semi-permanent vegetation plots first recorded in the early 1950s (Horák, 1972) reflecting the composition of coppice-with-standards oakwood, now converted to oak high forests. Our aims are (1) to test for the effects of abandonment of the traditional management on compositional changes in vegetation, (2) to test for the effects of the game preserve on the assumed successional trend, (3) to detect changes in vegetation heterogeneity, assuming a trend towards vegetation homogenization because of succession and (4) to assess the current conservation status of thermophilous woodland communities based on the occurrence of endangered species and of plant species typical for these communities. Finally, we will discuss the naturalness of Central European thermophilous oakwoods in the light of their dependency on management. METHODS Study site With its 25-km2 area, Milovice Wood is one of the largest continuous complexes of subcontinental oak and Pannonian hornbeam woodland communities in Central Europe and is protected as a Site of Community Interest CZ0624100. The Wood is located in SE Czech Republic, 16°41'38" E, 48°49'26" N, in the north-western edge of the Pannonian biogeographic province according to the Natura 2000 division. The site is a gently undulating loess plateau with altitudes ranging between 180 and 324 m a.s.l. Climate is relatively warm and dry with 9.0°C of average annual temperature and 550 mm of precipitation (Tolasz ef al, 2007). The most frequent soils are luvisols. Three woodland communities prevail in the forest vegetation (cf. Chytrý ef al, 2001; Commission of the European Communities, 2003): Pannonian oak-hornbeam forests, Primulo veris-Carpinetum (part of habitat type 91 GO); Pannonian thermophilous oak woods on loess Quercetum pubescenti-wboris (part of habitat type 9110); and Peri-Alpidic basiphilous thermophilous oak forests, Como-Quercetum (part of habitat type 91H0). Milovice Wood has most probably not changed in extent for at least the past seven centuries. Management consisted of short-rotation coppice-with-standards (cf. Szabó, 2010). The ownership of nearly the entire Milovice Wood was remarkably stable with only two major landlords from c. 1300 to the 1940s. The cutting period of the coppice was only 7 years in the Middle Ages (data from 14th to 15th centuries, Bretholz, 1930) and gradually increased to 12 years in the 17th century. Coppice compartments were several tens of hectares. The demand for underwood (coppice shoots) was high even in the pre-WW II period (Szabó, 2010). The post-WW II confiscation of private estates by the State could have been the principal reason leading to the abandonment of coppicing around the mid 20th century. In 1965 and 1966, two game preserves were established, taking up most of the Wood's area. In 1991, the larger Bulhary preserve (1250 ha) was populated mainly by red deer (Cervus elaphus) and fallow deer (Dama dama). The smaller Klentnice preserve (500 ha) has been burdened by two to three times higher densities of mainly fallow deer and moufion (Ovis musimon). Densities of animals per hectare were 0.34 in Bulhary and 1.13 in Klentnice in 1991, and 0.39 in Bulhary and 0.55 in Klentnice in 2006. (Komárek, 2008). 268 Diversity and Distributions, 16, 267-276, © 2010 Blackwell Publishing Ltd Half a century of succession in a temperate oakwood Vegetation datasets Complete lists of plant species with visual estimates of cover-abundances using the Braun-Blanquet scale (Van der Maarel, 2004) were recorded in square plots sized 500 m2. Vertical layers of trees, shrubs and herbs were distinguished. Woody saplings were considered as part of the shrub layer (Chytrý & Danihelka, 1992). The first sampling was performed in 1953-1954 by Horák (1972) before establishing the game preserves, reflecting 'the relatively homogeneous structure of a coppice' (pers. comm. cited by Chytrý & Danihelka, 1993). Horak's original dataset consisted of about 80 vegetation plots marked in topographical map scaled 1:25,000. In 1992, after three decades of the existence of game preserves, 46 of Horak's original plots were re-sampled using the same approach by Chytrý & Danihelka (1992). Horak's map and the information on the slope and aspect of the sites were used for defining the positions of the original plots. Only sites still covered with continuous forest vegetation were included, so that the results reflect mainly the succession process and not effects of cutting or re-planting (Chytrý & Danihelka, 1992). A total of 22 plots were in the Bulhary preserve, 15 in the Klentnice preserve, and nine plots remained outside both preserves. In summer 2006, Komárek (2008) sampled the 46 plots again. In total, 138 vegetation records were used in the analyses. Data analysis All analyses were carried out separately for herb and woody species (tree, shrub and woody saplings layers merged together), numbering 251 and 35, respectively. In all multivariate analyses, we used square-root transformation of percentage mid-points of respective cover-abundance values as species data. As a measure of compositional dissimilarity between plots, we used abundance-based Morisita-Horn index (Horn, 1966), which is insensible to differences in species richness (Wolda, 1981; Chao et al, 2006). In our case, to use a dissimilarity index independent from species richness is necessary, because subsequent vegetation records largely differ in species richness. Using some of other widely used dissimilarity indexes (e.g. Bray-Curtis) would lead to misleading conclusions because of their dependence on species richness (Chao et al, 2006; Jost, 2007). 1. To reveal the overall pattern in the vegetation, we performed non-metric multidimensional scaling (NMDS) in two dimensions and with a maximum of 400 random starts to achieve convergence. To facilitate visualization, the final configuration was rotated by principal components analysis and ordination axis scaled in half-change units. Moreover, we calculated arithmetic means of Ellenberg indicator values (Ellenberg et al, 1992) of herb species present within the plot. We also counted the number of species present in each plot (herb and woody species taken together). To help the interpretation of the pattern visualized by NMDS, we projected the vectors of these variables into an NMDS diagram, but only if they significantly correlated with some of the ordination axes (P < 0.05, based on a permutation test with 1000 permutations). 2. We tested the significance of the vegetation change by permutation-based MANOVA (Anderson, 2001). Time was coded as a continuous variable, i.e. 1953, 1992 and 2006. Because plots were semi-permanent (exact position not known), we used their IDs as covariables to remove the effect of spatial variability at the plot level. We assessed the significance of the vegetation change by a permutation test with 1000 permutations restricted by plot IDs. 3. We used permutation-based MANOVA also for testing the impact of game. The difference of vegetation change between plots inside and outside the game preserves was analysed. As explanatory variable, we used the interaction between time (year of recording) and the position of plots inside or outside the game preserves (binary coding). Time and plot IDs were used as covariables. Significance was tested by nonparametric test with 1000 permutations. We used a permutation scheme for split-plot design - unrestricted permutation of the whole plot (i.e. all three replicates from the same plot) and time series permutation within the whole plot (i.e. among replicates from the same plot). 4. We compared vegetation heterogeneity in three sampling periods through the analysis of multivariate homogeneity of group dispersions (Anderson et al, 2006). To asses whether the dispersions of groups are different, we performed a nonparametric permutation test with 1000 permutations (Anderson, 2006). We performed all multivariate analyses with vegan package (Oksanen et al, 2009) for R software (R Development Core Team, 2009). 5. To assess the status of the thermophilous oakwoods and the degree of assumed 'mesophication' (cf. Nowacki & Abrams, 2008), we analysed plant species typical for thermophilous oakwood and mesic forest communities in the three subsequent datasets. Typical species were statistically defined diagnostic species for higher vegetation units of the Czech vegetation (Chytrý & Tichý, 2003). Vegetation units were alliances Aceri tatarici-Quercion (with six diagnostic species), Quercion pubescenti-petraeae (N = 18) and Quercion petraeae (N = 19) representing thermophilous oakwoods, and alliance Carpinion betuli (N = 55) and order Fagetalia sylvaticae excluding C. betuli (N = 47) representing mesic forests of the Czech Republic. In each of the three datasets (1953, 1992 and 2006), we calculated the sum of presences of typical species for each vegetation unit and divided it by the sum of the presences within a unit in all three datasets. 6. Changes in conservation values were assessed using the occurrence of endangered species. We used the three highest-ranked categories of endangerment from the red list of vascular plant species of the Czech Republic (Holub & Procházka, 2000), i.e. Cl (critically endangered), C2 (strongly endangered) and C3 (endangered). The presence of endangered species in plots was summed within years and plotted. The analysed endangered species hardly overlap with the sets of typical species (see Appendix SI in Supporting Information). Diversity and Distributions, 16, 267-276, © 2010 Blackwell Publishing Ltd 269 R. Hédl ef a/. RESULTS Species richness, vegetation composition and heterogeneity Total species richness in 46 plots decreased between 1953 and 2006 from 181 to 159 herb species and from 33 to 24 woody species (for individual species see Appendix SI). In 1953, about one-third (58 herb species) were rare, present in 1-3 plots, and were subsequently not re-encountered in 1992 and 2006. The same was observed for six woody species. Only 15 rare herb species (occurring in 1-3 plots) were unique for the 1992 dataset and 25 for the 2006 dataset. No woody species were unique for either recent datasets. The vegetation composition of Milovice Wood has markedly changed during the last 53 years. Regarding herb species (NMDS diagram, Fig. la), the 1953 plots are clearly separated from more recent plots (1992, 2006). They were also more species rich and contained more species demanding higher temperature. By contrast, species with a higher demand for soil nutrients and moisture are more common in recent plots. Regarding woody species, the change in species composition is much less pronounced (Fig. lb). The composition of woody species changed much less than the composition of herb species, reflecting rather the changing proportions of present species than species turnover. Overall temporal change tested by permutation-based MA-NOVA explained 9.5% of the variation in herb species composition, and its effect was highly significant (F = 14.6, P < 0.001). For woody species, the change was also apparent; time explained 2.0% of the variance in species data, and its effect was statistically significant (F = 3.0, P < 0.001). Concerning the vegetation development inside and outside the game preserves, our results showed a significant difference, but only for herb species. The interaction between time and game preserve explained 1.2% of the variation in herb species composition and was statistically significant (F = 2.0, P < 0.001). However, the effect was not significant for woody species (F = 0.44, P = 0.27). Vegetation heterogeneity has not significantly changed over time. This holds true for herb species (global test of homogeneity of multivariate dispersion: F = 0.6, P = 0.26) as well as woody species (F = 0.5, P = 0.19). The assumed vegetation homogenization was not confirmed in this respect. Occurrence of typical and endangered species A marked decline was detected in species typical for the three units of thermophilous oakwoods. The most severe was the decline in the alliance Aceri tatarici-Quercion, that is, subcontinental thermophilous oakwoods (Fig. 2). In the other two alliances of thermophilous oakwoods, Quercion pubescenti-petraeae and Quercion petraeae and in the mesic oak-hornbeam woods (alliance C. betuli), there is a nearly identical trend of moderate decline from 1953 to 1992 followed by a milder decrease onwards to 2006 (Fig. 2). Species typical for mesic (a) Herb species