Impact of land-use practices on phytodiversity of mesic grasslands in a sub-mountainous region (Western Germany) Dissertation zur Erlangung des Doktorgrades (Dr. rer. nat.) der Naturwissenschaftlichen Fachbereiche der Justus-Liebig Universität Gießen durchgeführt am Institut für Landschaftsökologie und Ressourcenmanagement Professur für Landschaftsökologie und Landschaftsplanung vorgelegt von Dipl. Biol. Camilla Wellstein Gießen 2006 Dekan: Prof. Dr. Peter R. Schreiner 1. Gutachter: PD Dr. Rainer Waldhardt 2. Gutachter: Prof. Dr. Gerd Esser The doctoral thesis „Impact of land-use practices on phytodiversity of mesic grasslands in a sub-mountainous region (Western Germany)” is based on the following three papers: I. Wellstein, C., Otte, A. & Waldhardt, R.: Impact of site and management on the diversity of Central European mesic grassland. Agriculture, Ecosystems & Environment – accepted. II. Wellstein, C., Otte, A. & Waldhardt, R.: Seed bank diversity in mesic grasslands and their relation to vegetation, management and site conditions. Journal of Vegetation Science – accepted. III. Wellstein, C., Otte, A. & Waldhardt, R.: The population structure of three perennial grassland species (Pimpinella saxifraga L., Leontodon autumnalis L., Sanguisorba officinalis L.) in relation to management and habitat conditions. Applied Vegetation Science – submitted. In paper I, I had the main responsibility for design, field work, data analysis and writing. The co-authors contributed valuable ideas and suggestions for this study. In paper II and III, I had the main responsibility for design, field work, data analysis and performed the writing, while the co-authors contributed valuable comments. TABLE OF CONTENTS Table of contents 1 General Introduction ....................................................................................................................... 1 1.1 Background ............................................................................................................................... 1 1.2 Objectives .................................................................................................................................. 5 2 Study area ............................................................................................................................................. 8 3 Methods ............................................................................................................................................... 12 3.1 Sampling of aboveground vegetation and other habitat variables .......................... 12 3.2 Soil nutrient analyses and ascertaining of other environmental parameters ....... 12 3.3 Seed bank analyses .............................................................................................................. 13 3.4 Assessment of population biological parameters ........................................................ 13 3.5 Data analysis .......................................................................................................................... 13 4 Impact of site and management on the diversity of Central European mesic grassland .......................................................................................................... 15 4.1. Abstract .................................................................................................................................. 15 4.2. Introduction .......................................................................................................................... 16 4.3. Material and Methods ........................................................................................................ 17 4.4. Results .................................................................................................................................... 21 4.5. Discussion ............................................................................................................................. 25 5 Seed bank diversity in mesic grasslands and their relation to vegetation, management and site conditions ............................................................................................... 28 5.1 Abstract ................................................................................................................................... 28 5.2 Introduction ............................................................................................................................ 30 5.3 Material and Methods ......................................................................................................... 32 5.4 Results ..................................................................................................................................... 36 5.5 Discussion .............................................................................................................................. 42 6 The population structure of three perennial grassland species (Pimpinella saxifraga L., Leontodon autumnalis L., Sanguisorba officinalis L.) in relation to management and habitat conditions ............................................................ 50 6.1 Abstract ................................................................................................................................... 50 6.2 Introduction ............................................................................................................................ 52 6.3 Material and Methods ......................................................................................................... 54 6.4 Results ..................................................................................................................................... 59 6.5 Discussion .............................................................................................................................. 64 7 General Discussion and Synthesis ............................................................................................ 69 8 Summary ............................................................................................................................................ 74 9 Zusammenfassung .......................................................................................................................... 75 10 References .......................................................................................................................................... 77 Acknowledgements ..................................................................................................................................... 87 GENERAL INTRODUCTION 1 1 General Introduction 1.1 Background For centuries, semi-natural grasslands have been created and maintained by human land use in Central Europe (Slicher van Bath, 1963). Due to their plant species richness they are highly relevant for the maintenance of biodiversity at multiple spatial scales. Generally, low productivity and recurrent disturbance by mowing or grazing are a prerequisite of high species diversity, while simultaneously hampering competitive exclusion and allowing for coexistence of many plant species (Huston 1994; Grime 2001). At the habitat scale, it has been proved that increased productivity by fertiliser application as well as abandonment or enhancement of disturbance frequency by mowing or grazing lead to changes in floristic composition and finally to a loss in species-richness (e.g. Burel et al. 1998; Korneck et al. 1998; Mac Donald et al. 2000). In Central Europe grasslands declined strongly in number and size over the five last decades due to abandonment of marginal agricultural areas, melioration and subsequent arable use. The remaining grassland areas often underwent intensification of land-use such as fertilisation by mineral fertiliser or manure, drainage or frequent mowing (e.g. Burel et al. 1998; Korneck et al. 1998; Mac Donald et al. 2000). Consequently, many types of unimproved semi-natural grasslands that were common several decades ago have become extinct or fragmented. Besides the particularly endangered wet meadows and dry calcareous grasslands, the formerly widespread mesic grasslands of the order Arrhenatheretalia (Tüxen 1931) are also currently in decline (Burel et al. 1998; Poschlod & Schumacher 1998; Mac Donald et al. 2000). This development has led to the inclusion of this habitat type in the European Fauna-Flora-Habitat Directive of the European Union (92/43/EEC, European Union 1992; Ssymank et al. 1998). Particularly low-land grasslands were threatened by land use intensifications. In contrast, marginal regions, mainly within mountainous areas, are less agriculturally favourable. Such regions are traditionally associated with low-intensity management, and grassland habitats still predominate the agricultural landscape (e.g. Cousins & Eriksson 2002; Vandvik & Birks 2002; SRU 2004). Furthermore, according to OECD (2004), the common agricultural policy (CAP) reform is expected to ensure the maintenance of grassland areas. Due to these preconditions, such regions offer unparalleled opportunities to study different management regimes amongst other determinants influencing phytodiversity in mesic grasslands. The Lahn-Dill Highlands of central Hesse, Germany, are a typical example of marginal rural GENERAL INTRODUCTION 2 landscapes: these are characterised by relatively unfavourable abiotic conditions for cultivation, such as cool climate and shallow soils (Frede & Bach 1999). Furthermore, unfavourable structural conditions for agriculture comprise small-scale part time farming along with alternative incomes in the vicinity. Since the 1950s, the Lahn-Dill Highlands have been subject to major agricultural land-cover changes, resulting in a decline in arable land and an increase in grassland and fallow land (Waldhardt & Otte 2003; Hietel et al. 2005, in press; Reger et al. accepted). In many places, non-intense grassland use has successively replaced the traditional, extremely small-parcelled crop production and crop/grassland rotation. There are different current grassland management regimes in the study area which provide strong differences in disturbance impact. The management regimes range from hay meadows with only one or two cuts per year to pastures grazed from May to September and silage meadows mown three times a year. Management has a considerable influence on almost all aspects of grassland dynamics (Lennartson & Oostermeijer 2001). Hence, diverse management schemes can differently affect the dynamics and composition of plant communities and the dynamics of individual plant species. Species response to management may not only be manifested as presence/absence, but also in population demography, the alternations of which may anticipate possible floristic changes. Different cutting dates and frequencies may have a diversifying impact on vegetation (e.g. Kirkham & Tallowin 1995; Zechmeister et al. 2003) through differences in species’ regenerative abilities. Grazing animals affect vegetation in several different ways: through direct biomass consumption, selective grazing, trampling, urination, defecation, and by acting as dispersal agents (Olff & Ritchie 1998). Consequently, there is a need to assess the suitability of the alternative management options in maintaining grassland communities (Bühler & Schmid 2001; Hegland et al. 2001; Colling et al. 2002). This is particularly true for mesic grasslands, for which knowledge about the quantitative importance of recent and historical management practices in relation to other determinants of plant species richness is scarce. In this context, the analysis of specific effects of management practices and site conditions on phytodiversity at the habitat scale is a challenge for scientific research. Ecologically, mesic grasslands are characterised by a modest productivity and moderate variability in soil water potential. This leads to a relatively high diversity in species composition: The moderate range in productivity allows nutrient-demanding species as well as species depending on nutrient poor habitats. Besides species requiring moderate soil moisture, species that are adapted to alternations in soil moisture as well as those adapted to GENERAL INTRODUCTION 3 relatively dry conditions can be found in these grasslands. Hence, mesic grasslands meet the habitat requirements both of competitive species and of stress strategists. Phytodiversity is defined here as the share of biodiversity that is constituted by plants (cf. Noss 1990). In the thesis at hand the impact of management and site conditions on three different components of phytodiversity is studied at the habitat scale: the aboveground vegetation, the soil seed bank and the population biology of selected plant species’ populations. In the following paragraphs, these three components and their reliance on the respective determinants are introduced. The impact of management and site conditions on aboveground vegetation In traditional European phytosociology, management has been regarded as one of the most important factors differentiating vegetation types in mesic grasslands. Grazed (Cynosurion) and mown (Arrhenatherion) grasslands are usually separated at the high syntaxonomic level of the “alliance” (e.g. Dierschke 1994; Rodwell 1998). In addition to this expectation, the assemblage of plant species in seminatural grasslands is related to abiotic factors such as soil and topography (e.g. Cousins & Eriksson 2002; Sebastiá 2004). In general, site fertility is regarded as a crucial factor for phytodiversity (e.g. Grime 1979). Furthermore, the age, site history, and traditional management practices that may have ceased long ago are also influencing factors (e.g. Pärtel & Zobel 1999; Cousins & Eriksson 2002; Waldhardt & Otte 2003; Sebastiá 2004; Maurer et al. 2006). Several studies identified either environmental conditions (e.g. Vandvik & Birks 2002) or current management practices (e.g. Austrheim et al. 1999) to be more relevant for the explanation of floristic variance in grasslands. Moreover, species richness and composition of grassland vegetation depends on the pool of available species (Pärtel et al. 1996; Zobel et al. 1998; Pärtel & Zobel 1999). However, knowledge is scarce about the quantitative importance of recent and historical management practices in mesic grasslands in relation to other determinants of plant species richness and floristic composition, the two major measures of phytodiversity. Such information is of particular relevance for the development of recommendations for future land use with respect to grassland diversity. Significance of grassland soil seed banks Soil seed banks are a source for re-establishment of species which are lost from the aboveground vegetation. Hence, maintenance and restoration of species-rich grasslands will GENERAL INTRODUCTION 4 also depend on the soil seed bank (Grubb 1977; Fenner & Thompson 2005). However, Thompson et al. (1997) found that the investigation of seed banks has concentrated on productive agricultural habitats such as fertile grasslands. Relatively little is known about the seed bank communities and the respective ecology of species occurring in less productive semi-natural grasslands (Thompson et al. 1997: 21). For some grassland species such as Trifolium repens and Agrostis capillaris, which are common in mesic grasslands, there are clear indications for the presence of a persistent seed bank, but many grassland species lack a persistent seed bank (Rice 1989; Thompson et al. 1997; Bekker et al. 1998b, Bekker et al. 2000). However, it remains unclear to what degree the soil seed bank may contribute to the maintenance and restoration of species-rich mesic grasslands. Since changes in land-use and management practices alter disturbance regimes (e.g. Gibson et al. 2005) they can have very distinct impacts on the seed bank and the established vegetation (Bekker et al. 1997; Smith et al. 2002). Theory predicts a close relationship between the degree of disturbance in a habitat and the percentage of species with long-term persistent seed banks (Thompson et al. 1998 (p.168); Grime 2001; Hölzel & Otte 2004). Despite this, few empirical studies of soil seed banks investigate the impact of different types of current management regimes. Response of plant populations to management and site conditions From a conservation perspective, it is necessary to develop criteria in order to make priorities for choosing appropriate management regimes that allow for high species diversity. The performance of viable (i.e. growing or stable) populations is one important criterion for selecting management regimes, since the dispersal capacity of many grassland species is limited in space and time. Thus, the possibility of species enrichment is restricted in floristically impoverished sites (e.g. Bakker et al. 1996; Donath et al. 2003). In semi-natural grasslands, perennial species are representative for a major part of the plant species in the community, since some 90% of the species are relatively long-lived perennials (Lindborg et al., 2005). Due to the differences in species traits such as disturbance tolerance and nutrient requirements, there may be a species-specific impact of the determinants management regime and site conditions. The evaluation of population stage structure has proved to be a useful method to describe the demographic viability of populations in cultural landscapes in relation to management (Bühler & Schmid 2001; Lennartsson & Oostermeijer 2001), management and site conditions (Oostermeijer et al. 1994; Colling et al. 2002; Bissels et al. 2004) or land GENERAL INTRODUCTION 5 use change (Lindborg et al. 2005). Furthermore, populations have been evaluated in their natural habitats in landscapes dominated by natural disturbance regimes (García et al. 2002). Thus, the objectives of this thesis are: I. To evaluate the relative impact of current and past management and site conditions - such as edaphic parameters and topography - on species richness and species composition of mesic grasslands. II. To assess the relative impact of management on the soil seed bank diversity, and to assess the capability of the seed bank to contribute to the maintenance and restoration of species-rich grasslands. III. To evaluate the population structure of three perennial grassland species (Pimpinella saxifraga L., Leontodon autumnalis L., Sanguisorba officinalis L.) in relation to management and site conditions such as edaphic parameters and light supply. 1.2 Objectives The main objective of this study is to assess the impact of different management regimes on phytodiversity of mesic grasslands in the context of other important determinants. Three components of phytodiversity are investigated: the aboveground plant species richness and floristic composition, the seed bank plant species richness and floristic composition, and the population structure of three model species. At first, the study is based on an evaluation of the vegetation composition and species richness of the established grassland vegetation and on the identification and assessment of their determinants (chapter 4). Due to the heterogeneity of the overall study region, there is a high variability of site conditions related to edaphic parameters, topography and the history of land use. To understand patterns of phytodiversity in the grassland stands, the relative importance of current and past management and site conditions such as edaphic parameters and topography was analysed in a comparative study. If species are lost from the aboveground vegetation, the soil seed bank may offer a source for re-establishment. Knowledge on seed longevity is essential to assess the role of persistent soil seed banks in maintenance and restoration of sites (Bekker et al. 1997; Thompson et al. 1997; Hölzel & Otte 2004). Therefore we studied the species richness and composition of grassland soil seed banks, assessed the longevity of all present plant species and evaluated the impact of the different management regimes (chapter 5). GENERAL INTRODUCTION 6 Knowledge about the effects of different types of grassland management on population viability and persistence of grassland species is of high importance. However, most of the recent studies investigated only single species (e.g. Bakker et al. 1980; Bissels et al. 2004; Gibson et al. 2005). In this study (chapter 6), the population stage structure of the model species Pimpinella saxifraga L., Leontodon autumnalis L. and Sanguisorba officinalis L. was analysed in relation to environmental conditions and under the different main management regimes which exist in the region. In the following, the objectives of the thesis, as listed in the preceding chapter, are presented in detail. In chapter 7, the results of the individual studies (chapters 4-6) are summarised and discussed in a general discussion. Relative impact of site conditions and management on grassland vegetation (Chapter 4) This study deals with objective I, as it evaluates the relative impact of current and past management and site conditions such as edaphic parameters and topography on species richness and species composition of mesic grasslands of the order Arrhenatheretalia (Tüxen 1931) using two approaches. First, we compared vegetation types with respect to floristic composition, species richness, site conditions, grassland age as well as management. Secondly, we quantified the impact of these determinants on floristic composition of grasslands. Seed bank diversity (Chapter 5) The second study treats objective II, as it assesses the relative impact of management on the soil seed bank diversity, and the capability of the seed bank to contribute to the maintenance and restoration of species-rich grasslands. The main objectives were to analyse the floristic composition and size of the seed bank and to relate these to aboveground vegetation, site conditions and management. An additional goal was to test the effects of management regimes on functional aspects of the seed bank, such as seed mass, C-S-R strategy and seed longevity. Population structure of Pimpinella saxifraga, Leontodon autumnalis, Sanguisorba officinalis (Chapter 6) The study addressing objective III evaluates the population structure of three perennial grassland species (Pimpinella saxifraga L., Leontodon autumnalis L., Sanguisorba officinalis GENERAL INTRODUCTION 7 L.) in relation to management, site conditions (nutrient availability, soil moisture, pH and light supply), vegetation structure and species composition. The evaluation of management regimes is of high relevance for a successful maintenance of species-rich grassland communities. In this context the viability of model species populations may serve as a particularly useful indicator. We studied the stage structure in 16 populations of each of the perennial species Pimpinella saxifraga, Leontodon autumnalis, and Sanguisorba officinalis with respect to vegetation, site conditions and management. The main objective was to evaluate management options for the sustainable conservation of populations of these species in particular and species-rich grasslands in general. The results also provide useful information about how different management regimes affect populations of species differing in C-S-R strategy, clonal growth and requirements on edaphic conditions. In chapter 7 the most important results of chapters 4, 5 and 6 are summarised and their significance for land-use practices is discussed. STUDY AREA 8 2 Study area The sub-mountainous Lahn-Dill Highlands (LDH) is located in Central Western Germany in the state of Hesse. The area covers about 900 km² and forms the eastern descent of the Rhenish Uplands. The strong geomorphologic heterogeneity of the region is accounted for by structuring the LDH in eleven geographic subunits (Meynen & Schmithüsen 1957; Klausing 1988). Nearly all of these were considered in the field studies by randomly selecting grassland sites spread over the whole study area (Fig.1). The ridges in the west-southeast region of the study area reach an altitude of 600 m a.s.l. (Schelde Forest, Hörre). Basins with low mounds at 200 up to 400 m a.s.l. are predominant in the northern and eastern part of the area (e.g. Niederweidbach Basin, Damshausen Mounds, Salzboede Valley). Almost in the centre of the area the planar Bottenhorn Uplands rise up to 500 m a.s.l. The subunits are mainly structured by geomorphology, but studies have shown that they also differ with regard to recent land cover and land-use history (Reger et al. accepted). Predominant soil types are moderately acidic Cambisols and Luvisols with possible gleysation in the valley floors; Regosols are limited to hilltop positions (Harrach 1998; Szibalski 2000). The subatlantic climate of the LDH is characterised by a relatively cool and wet climate typical of sub-mountainous regions. Mean annual temperature and mean annual precipitation show a gradient from 6°C and 1100 mm in the raised western part to 7.5-8°C and 700mm in the eastern basin (Knoch 1950; HELELL 1981). However, the climate is largely modified by the local topography. In conjunction with the edaphic conditions, the wet climate results in a high variability of the soil-water potential. Since the 1950s, the Lahn-Dill Highlands have been subject to major agricultural land-cover changes, resulting mainly in a decline in arable land and an increase in grassland and fallow land (Kohl 1978; Hietel et al. 2004, 2005, in press; Reger et al. accepted). In many places, larger-scaled low-intensity grassland management schemes have replaced the traditional, extremely small-parcelled crop production and crop/grassland rotation. The agricultural land of the LDH, which currently covers about 31% of the entire area, is dominated by grasslands; these cover more than half of the agricultural land area (Reger et al. accepted). Generally, the study region is characterised by a predominance of low-intensity farming, which has its origins in disadvantageous natural site conditions and the political and social history of the region (Nowak 1988, 1992). Iron mining had been the main industry from the Middle Ages to the early 20th Century, whereas agriculture in this region was traditionally based on small- scale farming and predominantly provided only sideline incomes (Hietel et al. 2005). Hence, STUDY AREA 9 the study area is a typical example of marginal landscapes. Since 1976 the entire area has been included in the support scheme for less-favoured areas (EC Regulation No75/268). Typically, a large part of the landscape is managed by part-time farmers, who adhere to traditional agricultural practices (Hietel et al. 2005). The grasslands are partly managed according to EU-based agro-environmental schemes offered by the state of Hesse, focussing on grassland extensification (HMULF 2002). The adoption of agro-environmental schemes ensures a late first mowing not before mid-June, a low input of fertiliser (< 30 kg N ha-1 year-1) or even a ban on fertilising, and extensive grazing with not more then 1.5 life weight units ha-1 (LWU). The grasslands of the study area are mainly grazed with cattle, but also with horses and on a few sites by sheep. As in many other low mountainous regions of central Europe (cf. Gigon 1999) these grasslands are embedded in a small-patch mosaic of arable fields, grassland fields, and old fields with shrub succession (Simmering et al. 2006). The latter are mainly occupied by Cytisus scoparius (Simmering et al. 2001). Fields vary in size from less than 0.5 to 5 ha (Reger et al. accepted). STUDY AREA 10 � � � � ������ ���� ���� ���� ������� ���������� �� ��� ������� �� ����� �� �� �� �� �� �� �� � � �� � � � �� � �� �� �� � � � �� � � � � � ��� ���� �� ���� �� �� ����� �� � � � �� � � � � �� � � � � � �� ���� �� �� � � � �� �� �� � � � ���� � � ��� ���� �� � � � ���� �� ���� �� �� �� �� �� �� �� �� �� �� �� � ��� �� �� ��� � � �� � � ������� �� � � � � �� ���� �� �� �� � ��� � � � �� �� �� �� �� �� �� �� �� �� � � � � � ������� ������� ������� ������ ��� ���������� � � � � � � � � ���� ���� ������� � �������������� !��"�������#"$���������������� %��&����� � '��(�����)����� ����� �� *�������� +�������� �������� ��, � -��. ���������� /���� �� �������"����� 0��.����������� 1��2������������� �������������������� � �3������) � ������������) �� �����)#������� �� ��������4 �����5 � ��������4 �����5 ( 6 ! +6 +! "��������� Fig. 1. Geographic situation of the study region Lahn-Dill Highlands and schematic distribution of the sampling sites, including information on management types, within geographic subunits. Since our main objective was to evaluate the impact of current management regimes on phytodiversity, we selected only grasslands that are managed within the local farming system. We excluded grasslands that are not included in local farming systems, i.e. grasslands of nature reserves receiving nature protection measures for maintenance. Due to the heavy influence on phytodiversity we did not aim to compare grasslands over a large gradient in soil moisture. Therefore wet grasslands were excluded and only mesic grasslands with species compositions referring to the order Arrhenatheretalia (Tüxen 1931) were investigated with STUDY AREA 11 respect to management regimes and site conditions. To account for the topographic and edaphic heterogeneity of the region, 56 grassland fields (size 0.3 – 3 ha) were randomly selected within nine geographic units of the entire Lahn-Dill Highlands (Fig.1). Across all geographic units, the fields were categorised into management types (i-v; see below). Management types were almost evenly spread within the entire Lahn-Dill Highlands (Fig. 1). The five main types are: (i) silage meadows with early and frequent mowing (three cuts per year) and little fertiliser input (up to 30 kg N ha-1 year-1), (ii) hay meadows with late mowing in mid-June and one or two cuts per year, (iii) meadow-pastures with late mowing and subsequent cattle grazing, (iv) pastures grazed by cattle from May to September and (v) pastures grazed by horses between June and September. For the study on aboveground vegetation (chapter 4), all 56 sampling sites covering the five management types were sampled (Fig.1). For the study on the soil seed bank (chapter 5) and the study on the population structure of selected species (chapter 6) a subset of these sites spread over the whole study region and covering the four main important management types (hay meadows, silage meadows, meadow pastures and pastures grazed by cattle) was studied. More detailed information with focus on the different research topics can be found in the material and methods as well as the study area sections of the chapters 4, 5 and 6. METHODS 12 3 Methods This chapter summarises the sampling and data analysis methods used in the thesis. 3.1 Sampling of aboveground vegetation and other habitat variables Different methods for sampling of vegetation and other habitat variables were applied. To describe species composition in the evaluation of aboveground floristic composition (chapters 4 and 5), the abundance of all vascular plant species was estimated on 5 x 5 m plots using a modified Braun-Blanquet scale (van der Maarel 1979). For the investigation of the population structure of three selected grassland species (chapter 6), several indicators of vegetation structure were estimated on 1 x 1 m plots: Total vegetation cover, the coverage of mosses, plant litter, the percentage of bare soil surface and the mean vegetation height. Moreover, light measurements were undertaken in each plot in this study (chapter 6). Using a Line Quantum Sensor of one meter length (LI-COR: LI-191SA) light intensity penetrating to the ground was recorded. Photosynthetically active radiation (PAR, 400-700 nm) was measured simultaneously at ground level and in full light above the canopy. Light penetration was expressed as a percentage of the latter value. 3.2 Soil nutrient analyses and ascertaining of other environmental parameters For the evaluation of soil chemical parameters in chapters 4, 5 and 6 soil samples were collected and soil nutrient analyses were conducted. Plant-available phosphorus and potassium were determined using the Calcium-Acetate-Lactate extraction method (CAL). Total nitrogen and total carbon were assessed using a CN analyser (FlashEA 1112, Thermoquest). The pH values of the fine soil were determined in CaCl2. As there were no limestone formations in the sampling areas, the total carbon of the soil represents the organic carbon and was used to calculate the organic matter in the soil samples. All analyses were done according to Steubing and Fangmeier (1992). The topographic position of the plots within the landscape was categorised in four classes from the valley floor to the upper slope. Furthermore, slope inclination (°), elevation (m a.s.l) and aspect were recorded. Aspect, i.e. the compass-direction of a slope, was characterized by Northness (cosine of aspect) and Eastness (sine of aspect). Additionally, the site conditions were characterised by calculating METHODS 13 cover-weighted averages of the vascular plant indicator values for moisture and nutrients for each plot (Ellenberg et al. 1992). 3.3 Seed bank analyses For the investigation of the soil seed banks of grasslands (see chapter 5), soil seed samples were collected at each site by random collection of 20 cores of 10 cm depth. We analysed seed banks using the seedling emergence method over 4½ months (Roberts 1981; Thompson et al. 1997). After removing vegetative plant material the soil samples were transferred to Styrofoam trays and exposed under warm greenhouse conditions (day 25°C / night 15°C). The trays were watered regularly. Germinating seedlings were identified and removed once every week. Unidentifiable seedlings were transplanted into pots and grown until identification was possible. 3.4 Assessment of population biological parameters In chapter 6 the population stage structure of the perennial species Pimpinella saxifraga, Leontodon autumnalis and Sanguisorba officinalis was investigated. For this purpose 16 grassland sites were chosen for each species where the respective model species was present. In each of the 16 populations two 1 m² plots were randomly selected and the total number of individuals in each plot was counted. To classify the life stage classes for each species the following growth parameters were measured for each individual: existence of cotyledons (all species); leaf morphology of the leaf blades on the ground rosette: length and width (L. autumnalis); number of pinnules (P. saxifraga and S. officinalis) (see chapter 6, Table 2); existence of flowering stalks (all species). 3.5 Data analysis Several methods of both exploratory data analysis (e.g. ordination) and inferential statistics were applied in the thesis. The inferential statistics cover methods that allow statistical hypotheses testing, such as regression analysis and analysis of variance (Jongman et al. 1995). Analysis of complex data sets like plant community data often combines these two approaches (Jongman et al. 1995). The following methods of multivariate analysis of ecological data were conducted: METHODS 14 o Detrended Correspondence Analysis (DCA) was used to explore gradients in the floristic composition of vegetation samples (chapters 4, 5, 6). o Partial Canonical Correspondence Analysis (CCA) performed a decomposition of variance and was used to isolate the effect of management and other important determinants on floristic composition (chapters 4, 5). o Two-Way Indicator Species Analysis (TWINSPAN) (Hill 1979) allowed a classification of vegetation types of the sampled grasslands by examining the main groupings in the data set and the assignment of sampling sites to the respective groups in chapter 4. o Indicator Species Analysis (Dufrêne & Legendre 1997) was applied to identify significant indicators of vegetation types (chapter 4) and of seed banks of differently managed grasslands (chapter 5). Furthermore the following statistical methods were used: o Multiple regression analysis (GRM module in Statistica) was performed to assess the importance of management and environmental variables in grasslands for the population structure of selected species (chapter 6). o Multivariate analysis of variance (MANOVA) and covariance (MANCOVA) were applied to test for overall effects of possible determinants on the dependent variable of interest (chapters 4, 6). o Univariate analysis of variance (ANOVA) and covariance (ANCOVA) were applied to test for differences between two groups of interest in normally distributed variables and variables that could be adequately transformed (e.g. the numbers of seeds in chapter 5). To analyse significant differences between several groups in detail, post hoc tests were used (e.g. Tukey´s Honest-Significance test (HSD) (chapters 4, 5, 6). o In all cases of correlative analysis the non-parametric Spearman Rank Correlation was used (chapters 4, 6). More detailed information with focus on the different research topics can be found in the material and methods sections of the chapters 4, 5, and 6. IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 15 4 Impact of site and management on the diversity of Central European mesic grassland Camilla Wellstein, Annette Otte & Rainer Waldhardt Agriculture, Ecosystems and Environment: accepted 4.1 Abstract The main objective was to quantify the relative impact of current management types on plant species-richness and composition of mesic grasslands with regard to other important determinants such as topography, soil chemical parameters and grassland age. The grasslands were (i) differentiated into management types and vegetation types, (ii) these types were tested for differences in site conditions and species-richness, and (iii) the relative impact of management, site conditions, grassland age and regional scale geomorphology on floristic composition was quantified. TWINSPAN classification of the vegetation separated nutrient- poor from nutrient-rich sites. Results of ANCOVA revealed that vegetation types indicating high nutrient levels showed significantly higher contents of plant available phosphorous and younger grassland age. In partial CCA analyses, the geomorphology accounted for almost one third of explained variance. The current management had a relatively low explanatory value. Soil chemical variables and topography, in contrast, explained together almost twice as much variation in floristic composition. Keywords: Land use, Marginal landscape, Soil fertility, TWINSPAN, Indicator Species Analysis, ANCOVA, Partial CCA IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 16 4.2 Introduction Plant species-richness and floristic composition in grasslands are shaped not only by current site conditions, species pool and management but also by age, site history, and traditional ancient management practices that may have ceased long ago (e.g., Pärtel et al., 1996; Cousins and Eriksson, 2002; Waldhardt and Otte, 2003; Sebastiá, 2004). Phytodiversity over a broad range of environments has been shown to be determined mainly by the overall productivity and the land-use history of the study systems (e.g., Milchunas and Lauenroth, 1993). Studies on the relative importance of management and environmental factors on floristic composition in grasslands have shown either environmental conditions (e.g., Vandvik and Birks, 2002) or current management practices (e.g., Austrheim et al., 1999) to explain relatively more of the floristic variance. The assemblage of plant species in seminatural grasslands is often related to abiotic factors such as soil and topography (e.g., Cousins and Eriksson, 2002; Sebastiá, 2004). Soil fertility has been shown to be an important factor for phytodiversity (Janssens et al., 1998) and increasing amounts of fertilisers in agricultural practice are generally accepted as the main cause of the decline in grassland phytodiversity (e.g. Gough and Marrs, 1990; Smith, 1993; Korneck et al., 1998; Zechmeister et al., 2003). Grazing animals affect vegetation in several different ways, through direct biomass consumption, selective grazing, trampling, urination, defecation, and by acting as dispersal agents (Olff and Ritchie, 1998). Moreover, species-richness and species composition of grassland vegetation depend on the available species pool (Pärtel et al., 1996; Zobel et al., 1998). Many grassland studies focus on the phytodiversity of highly endangered communities. In central Europe, these are unimproved semi-natural grasslands like the particularly endangered wet meadows and dry calcareous grasslands. But due to agricultural intensification and abandonment, the overall area of the formerly widespread mesophilous grassland of low mountainous regions is currently also in decline (Burel et al., 1998; Mac Donald et al., 2000). This has led to the inclusion of this habitat type in the European Fauna-Flora-Habitat Directive of the European Union (92/43/EEC, European Union, 1992; Ssymank et al., 1998). Given this background, the aim of this study was to assess and quantify the impact of former and current land use practices, site conditions, and regional scale geomorphology on the grassland phytodiversity of a marginal region, which is characterised by extremely small- scaled fields and highly diverse management schemes. Grassland management practices in the area provide strong differences in disturbance impact ranging from low-intensity pasturing without fertiliser application to mowing three times a year for fodder production (silage). The IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 17 specific questions addressed in this paper are: 1. How are grassland vegetation types differentiated in terms of floristic composition, species-richness and site conditions? 2. How important is current management in relation to other factors such as abiotic site conditions, grassland age and regional scale geomorphology for the floristic composition of grassland stands? 4.3 Material and Methods Study region and sampling The entire study region (Lahn-Dill-Highlands, Germany) has been included in the support scheme for less-favoured areas since 1976 (EC Regulation No75/268). Typically, a large part of the landscape is managed by part-time farmers, who adhere to traditional agricultural practices (Hietel et al., 2005). Since the 1950s, the Lahn-Dill Highlands have been subject to major agricultural land-cover changes, resulting mainly in a decline in arable land and an increase in grassland and fallow land (Hietel et al., 2005). In many places, extensive grassland use has replaced the traditional, extremely small-parcelled crop production and crop/grassland rotation. In the study region, a large part of the grasslands is managed according to EU-based agri-environmental schemes, focussing on grassland extensification. The adoption of the agri- environmental schemes ensures a late first mowing not before mid of June, a low input of fertiliser (< 30 kg N ha-1 year-1) or even the ban of fertilising, and low-intensity pasturing with < 1.5 life weight units ha-1 (LWU). Pasturing on grasslands of the study area is done mainly with cattle, also with horses and few sites are grazed by sheep. Predominant soil types are moderately acidic Cambisols and Luvisols with possible gleysation in the valley floors; Regosols are limited to hilltop positions. The climatic conditions in the region are relatively unfavourable, indicated by a mean annual temperature of 6 to 8°C and average annual precipitation ranging from 650 to 1100 mm. In conjunction with the edaphic conditions, the wet climate results in a high variability of the soil-water potential. The Lahn-Dill Highlands are divided in different geomorphological subunits (GU) (Klausing, 1988) (see Table 1). The subunits are mainly structured by geomorphology, but they differ also with regard to recent land cover and land-use history (Hietel et al., 2005). The heterogeneity across these subunits may have had an effect on the development of different species pools in grassland vegetation, which may be due to a combined effect of contrasting large scale differences in soil properties, climate and land-use history. IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 18 To account for the topographical and edaphical heterogeneity of the region, 56 grassland fields (size 0.3 – 3 ha) were randomly selected within nine GU of the entire Lahn-Dill Highlands. Across all GUs, the fields were then categorised into management types (i-v; see below). Management types were almost evenly spread within the entire Lahn-Dill-Highlands (Table 1). To account for the sometimes high internal variability of vegetation within stands, three plots (5 m x 5 m) were randomly placed within each of the 56 fields and these were used as basic sampling unit. Geographical coordinates of plots were recorded using a Garmin GPS. To avoid edge effects, a minimum distance of 10m to the border of the fields was kept. The composition of vascular plant species was recorded within plots between May and September in 2003 and 2004. Species abundance was estimated on a modified Braun-Blanquet-scale (with cover degree 2 subdivided into 2a = > 5 - 15% and 2b = > 15 - 25%). The nomenclature of the vascular plant species followed Wisskirchen and Haeupler (1998). Table 1. Distribution of plots with different management types within geomorphological subunits (GU) of the Lahn-Dill Highlands (Klausing, 1988) Management types: H, hay grassland; S, silage grassland; M, meadow; P[c], cattle grazed pasture; P[h], horse grazed pasture. Geomorphological subunit GU ID Management types of plots (n=166) Elevation a.s.l. [m] (GU) H S M P[c] P[h] Bottenhorn Uplands 1 6 9 9 6 3 470-530 Hoerre 2 3 6 6 3 0 330-410 Niederweidbach Basin 3 6 0 0 3 2 280-315 Zollbuche 4 9 3 6 3 3 340-490 Krofdorf-Königsberg Forest 5 0 6 3 3 0 320-390 Schelde Forest 6 3 6 0 3 2 350-420 Damhausen Mounds 7 6 0 3 3 0 280-340 Salzboede Valley 8 3 6 0 3 3 245-290 Upper Dill Valley 9 6 6 9 3 3 270-360 The current grassland management was classified based on information obtained from personal interviews with farmers conducted prior to the study in 2003. Additional background information was gained from the specific provisions of the agri-environmental schemes mentioned above. This combined data provided information on mowing intensity (frequency and time), grazing intensity (duration, number and kind of grazing individuals per area), the amount and type of fertiliser used. Five management practices were differentiated: (i) silage grassland with early and frequent mowing (three cuts per year) and little fertiliser input (up to 30 kg N ha-1 year-1), (ii) hay grassland with late mowing in mid-June and one or two cuts per year, (iii) meadows with late mowing and subsequent cattle grazing, (iv) cattle grazed pasture IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 19 (grazed from May to September) and (v) horse grazed pasture (grazed between June and September). The four last practices were characterised by the lack of fertiliser input. There was no aftermath grazing in case of management practices (i) and (ii). Information on the consistency of management type was also derived from the personal interviews with farmers. All selected fields were managed accordingly for the last 25 years, i.e. since 1979. However, as major land-cover changes occurred during the first decades after World War II, and memories of farmers were incomplete for this period, the duration of grassland use was checked for each field by visual interpretation of black-and-white aerial photographs available from the years 1953, 1962, 1967, 1973, and 1979. The age (A) of grassland fields was quantified according to the determined land-cover at each of these dates. The probability that fields used as grassland in 1953 had also been traditionally used as grassland in prior times appeared to be very high. Thus, fields with an age of 50 years (grassland use in all photographs) were presumably ancient grasslands. In autumn 2003, composite soil samples from each of the 168 plots were obtained by pooling 20 randomly sampled cores (3 cm diameter, 10 cm depth). Total nitrogen (Nt) and total carbon (Ct) levels were determined using a CN-analyser (FlashEA 1112, Thermoquest). Levels of plant available phosphorus (PCAL) and potassium (KCAL) were estimated by calcium- acetate-lactate (CAL) extraction method. PH-values of the fine soil were determined in CaCl2. As there were no limestone formations in the sampling areas, the total carbon of the soil represented the organic carbon and was used to calculate the organic matter in the soil samples. The relative topographic position of the plots within the landscape was categorised in four classes from the valley floor to the upper slope. Slope inclination (°), elevation (m a.s.l) and aspect were recorded as well. Aspect, i.e. the compass-direction of a slope, was characterised by Northness (cosine of aspect) and Eastness (sine of aspect). Additionally, the site conditions were characterised by calculating cover-weighted averages of the vascular plants indicator values for moisture and nutrients (Ellenberg et al., 1992) for each plot. Data analysis/Statistics Prior to all analyses of this study, two data modifications were performed. First, according to an Outlier Analysis (included in the software package PC-ORD 4) two plots were omitted as multivariate outliers (McCune and Grace, 2002), resulting in a total of 166 plots for further analyses. Second, seasonal (first or second crop) and interannual (year 2003 or 2004) variation of the vegetation samples was tested by the method of Indicator Species Analysis IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 20 with ‘season’ and ‘year’ as grouping factors (Dufrêne and Legendre, 1997; McCune and Grace, 2002). To calculate the indicator value of a species, its mean abundance in one group compared with its mean abundance in all groups was multiplied by its relative frequency in the samples of that group. The obtained values were tested for significance by Monte Carlo statistics with 1000 random permutations (McCune and Mefford, 1999). ‘Year’ showed no significant differences of species occurrences. ‘Season’ resulted in significant differences of abundance and frequency of six species with an early phenology which were omitted from the data set. Correlation between species richness with and without these omitted species across all plots was 0.97 (Spearman’s r). The omitted species were not specific for particular management regimes, thus, exclusion of these species caused no bias in the comparison between management types. Adjusting the data for season-specific species resulted in reduction of species number by two species per plot on average. Classification of vegetation types was achieved by a Two-Way Indicator Species Analysis (TWINSPAN) (Hill, 1979). The TWINSPAN run resulted in an ordered table that allowed examining the main groupings in the data set and the assignment of sampling sites to the respective clusters. Subsequently, Indicator Species Analysis (see above) was used to detect species characterising the groups (clusters) generated by TWINSPAN. Numerical analyses were performed using the software package PC-ORD 4 (McCune and Mefford, 1999). Multivariate Analysis of Covariance (MANCOVA) was carried out (using Wilks Lambda as test statistic) to test whether species-richness and site conditions differed among (i) the four TWINSPAN clusters and (ii) the five management types. Significant differences between groups for particular dependent variables were assessed using subsequent one-way ANCOVAs with either (i) management type or (ii) cluster membership as fixed effect. In case of significant univariate effects, means were compared using Tukey Honest-Significance test (HSD) for unequal sample sizes. To account for the effects of spatial autocorrelation in our data, the geographical coordinates of the plots were used to construct nine spatial variables, i.e. the terms of a cubic trend- surface polynomial (Borcard et al., 1992): f(x, y) = b1x + b2y + b3x² + b4xy + b5y²+ b6x³+ b7x²y+ b8xy² +b9y³ (1) These were included as covariables in MANCOVA to partial out the spatial component of variation in analyses of variance. Multivariate and univariate ANCOVA and associated tests were carried out using STATISTICA 6.0 (Anon., 1998). IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 21 Prior to ordination analyses, species with less than three occurrences across all plots were excluded to reduce their influence on ordination results. Subjecting the species data set (166 relevés, 114 species) to a Detrended Correspondence Analysis (DCA) (McCune and Grace, 2002) revealed a gradient length on the first axis of 2.34 SD, which showed a modest unimodal response and thus the appropriateness of CCA (Ter Braak and Šmilauer, 1998). Environmental variables (see paragraph ‘environmental variables’) for CCA were selected by the CANOCO procedure of forward selection (Palmer, 1993). Except for Eastness, all examined variables yielded significant contributions to data structure and were retained for CCA. To receive ecologically interpretable variance components, the explanatory variables were grouped into the following five sets: Management type (M; including five different management types) (see Table 1); geomorphological subunit (GU; including nine different subunits) (see Table 1); soil chemical parameters (S; including organic carbon, total nitrogen, pH value, and plant available P and K contents); topography (T; including elevation a.s.l., slope inclination, topographic position, and Northness); grassland age (A). Spatial autocorrelation of plots (SP) was controlled for in CCA by including again a covariable matrix containing the terms of the cubic trend surface polynomial (equation 1) following Borcard et al. (1992). To distinguish between the gross and net effects of these sets on floristic composition, gross effects were first quantified by performing a series of CCAs (controlled for SP) for each set of explanatory variables. To obtain the net effect of a given set of variables, additionally a series of partial CCAs were performed controlling for all other sets (Økland and Eilertsen, 1994). For all CCAs, significance was tested by permutation tests (1000 permutations). The ratio of a given canonical eigenvalue to the sum of all eigenvalues (total inertia) was used to estimate the proportion of explained variation. CCAs were performed using the program package CANOCO (Ter Braak and Šmilauer, 1998). 4.4 Results The TWINSPAN classification of vegetation types resulted in four ecologically meaningful clusters: Judging from floristic differences between the clusters (Table 2) the first division in TWINSPAN apparently separated nutrient-rich (clusters I, II) from nutrient-poor (clusters III, IV) sites. IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 22 Table 2. Indicator species (after Dufrêne and Legendre, 1997) of TWINSPAN clusters I-IV. Significance obtained by Monte-Carlo-Permutation test is given at three levels: * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001, Ind. value, indicator value. Species with non significant or low indicator values (< 25) are not shown; (N) notifies species indicating nutrient-rich habitats (Ellenberg nutrient value > 7; Ellenberg et al., 1992. Cluster I (n = 32) Ind. value Cluster II (n = 65) Ind. value Lolium perenne (N) 47.6 *** Trisetum flavescens 51.3 *** Phleum pratense s. str. (N) 46.8 *** Arrhenatherum elatius (N) 46.6 *** Taraxacum officinale agg. (N) 46.2 *** Dactylis glomerata agg. 36.6 *** Leontodon autumnalis 40.8 *** Trifolium pratense 33.5 *** Trifolium repens 30.8 ** Galium album 33.5 *** Plantago major s. l. 29.3 *** Achillea millefolium 30.1 ** Heracleum sphondylium (N) 28.6 ** Leucanthemum vulgare 26.2 ** Cluster III (n = 39) Ind. value Cluster IV (n = 30) Ind. value Festuca rubra 43.1 *** Sanguisorba officinalis 59 *** Deschampsia cespitosa agg. 36.8 *** Lathyrus pratensis 35.1 *** Ranunculus acris 31.8 *** Leontodon hispidus 34.4 *** Potentilla erecta 31.8 *** Agrostis capillaris 31.4 ** Holcus lanatus 29.2 * Centaurea jacea s.l. 28 *** Table 3. Mean values and standard error of site and vegetation parameters for TWINSPAN-derived clusters of grassland samples. Asterisks denote the significance levels revealed by univariate ANCOVA (* P ≤ 0.05; ** P ≤ 0.01; *** P ≤ 0.001) after significant one-way MANCOVA (F 30,435 = 5.667, P < 0.001). Different letters indicate significant differences among clusters, according to a Tukey test (for unequal n). Cluster I Cluster II Cluster III Cluster IV (n = 32) (n = 65) (n = 39) (n = 30) Age of grasslands (years)*** 38±1.4 a 41±1.0 a 49±1.3 b 50±1.5 b Number of species* 19±1.0 a 23±0.7 b 25±0.9 b 27±1.0 b Elevation a.s.l.* 351±6.7 b 376±4.7 c 432±6.1 d 324±6.9 a Total nitrogen (%)*** 0.40±0.019 a 0.40±0.014 a 0.52±0.018 b 0.46±0.020 a b Organic matter (%)* 6.8±0.35 a 7.1±0.25 a 9.3±0.32 b 7.8±0.36 a pH value [CaCl2]** 5.12±0.075 b 5.14±0.052 b 4.54±0.068 a 4.90±0.077 b Phosphorus [mg/100g]*** 7.4±0.63 b 7.8±0.44 b 3.9±0.57 a 3.1±0.65 a Potassium[mg/100g]* 24±1.9 c 16±1.4 b 10±1.8 a b 7±2.0 a Moisture value (MV)* 5.25±0.066 b 5.00±0.047 a 5.51±0.060 c 5.56±0.069 c Nutrient value (NV)*** 6.4±0.11 c 5.8±0.07 b 5.0±0.10 a 5.2±0.11 a IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 23 Figure 1. Percentage distribution of current management types of grassland samples within TWINSPAN clusters, i.e. vegetation types. This was confirmed by the MANCOVA analyses. The comparison of the derived clusters by MANCOVA revealed significant differences in average site conditions and species-richness (F 30,453 = 5.667, P < 0.001). Tukey tests (Table 3) showed that clusters I and II contained the more fertile sites indicated by the higher content of available phosphorus, potassium, and higher Ellenberg nutrient values. Younger grasslands belonged exclusively to these clusters. In contrast, clusters III and IV contained less productive sites indicated by significantly lower contents of available phosphorus and potassium, and lower Ellenberg nutrient values. However, sites of these clusters were characterised by a significantly higher Ellenberg moisture value. Coherences between these explanatory variables became obvious also by significant, moderate correlation (Spearman`s r) between the age of the fields and the available phosphorus (r = -0.536), total nitrogen (r = -0.593) and organic matter (r = -0.528) across all plots. There was also significant but low correlation of species richness and the explanatory variables age (r = 0.371), phosphorus (r = -0.284) and nutrient value (r = -0.394). At the level of the second TWINSPAN division, the current grassland management became apparent. Cluster I predominantly contained sites with rather intense use for silage and pasture (Fig. 1). Floristically, this was indicated by species preferring high nutrient levels and tolerating high levels of disturbance such as Lolium perenne, Phleum pratense s. str. and Plantago major s. l., and ruderals such as Taraxacum officinale agg. (Table 2). Sites in this cluster contained significantly fewer species and were the youngest stands of all grasslands. Cluster II was dominated by hay grassland (Fig. 1) with species indicating nutrient-rich conditions such as Arrhenatherum elatius and Heracleum sphondylium (Table 2). Cluster III contained many pastures and meadows (Fig. 1). The impacts of grazing and relatively nutrient-poor and acidic site conditions were indicated by the dominance of species such as Festuca rubra and Deschampsia caespitosa agg.. This corresponded with low pH values and higher elevations of the stands (Table 3). Total nitrogen IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 24 and organic matter contents were also higher in the soils of these permanent grasslands. In contrast to cluster III, cluster IV predominantly contained stands with different mowing regimes and very few pastures (Fig. 1). Species characteristics of this cluster comprised indicators of base-rich and relatively nutrient-poor sites with an alternating moisture regime (Table 2). In contrast to the differentiating results obtained by the comparison of vegetation types, grassland plots revealed only slight differences in site conditions and species-richness when compared with respect to their management type alone (MANCOVA F 40,559 = 2.7, P < 0.001). Hay grasslands differed significantly from cattle grazed pastures in age and potassium content, and from meadows in their nitrogen and organic matter contents (Table 4). Species- richness differed significantly only between silage grasslands and meadows. Table 4. Mean values and standard error of site and vegetation parameters for grassland samples of five different management types: H, hay grassland; S, silage grassland; M, meadow; P[c], cattle grazed pasture; P[h], horse grazed pasture. Asterisks denote the significance levels revealed by univariate ANCOVA (* P ≤ 0.05; ** P ≤ 0.01; *** P ≤ 0.001) after significant one-way MANCOVA (F 40,559 = 2.700, P < 0.001). Different letters indicate significant differences among management types, according to a Tukey test (for unequal n). H S M P[c] P[h] (n = 42) (n = 42) (n = 36) (n = 30) (n = 16) Age of grasslands (years)* 41 ±1.3a 41 ±1.3 a 46±1.5 a b 48±1.5 b 46±2.3 a b Number of species** 25±0.9 a b 22±0.9 a 26 ±1.0 b 23±1.1 a b 22±1.4 a b Elevation a.s.l. ** 355±5.7 a 371±5.7 a b 398±6.1 c 383±6.7 b c 367±9.2 a b c Total nitrogen (%)* 0.40±0.017 a 0.45±0.017 a b 0.48±0.018 b 0.44±0.020 a b 0.40±0.027 a b Organic matter (%)** 7.0±0.30 a 7.8±0.30 a b 8.6±0.33 b 7.8±0.36 a b 7.1±0.49 a b pH value [CaCl2] n.s. 4.97±0.068 5.02±0.068 4.87±0.072 5.04±0.080 4.79±0.110 Phosphorus [mg/100g] n.s. 5.6±0.58 6.5±0.58 6.1±0.63 6.8±0.68 3.7±0.94 Potassium [mg/100g]** 10±1.7 a 15±1.7 a b 16±1.9 a b 20±2.1 b 12±2.8 a b Moisture value (MV)* 5.1±0.06 a 5.3±0.06 a b 5.3±0.07 a b 5.3±0.07 a b 5.5±0.10 b Nutrient value (NV) n.s. 5.6±0.11 5.9±0.11 5.5±0.12 5.6±0.13 5.3±0.17 Classification results for vegetation and management types confirmed the importance of site conditions for floristic composition and revealed a moderate impact of management. Taking all sets of variables into account as constraining variables in CCA and controlling for spatial autocorrelation, they explained 27.9% of total variation in floristic composition (Table 5). After controlling for the other sets, each set studied here still yielded a significant net effect on floristic composition in partial CCA. Comparing the net effects, the geomorphological IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 25 subunit (GU) explained almost one third (8.5%) of explained variance. The current management practice (M) had an explanatory value of 4.2%, which indicates its moderate role for the floristic variance. The net effects of site conditions, i.e. soil (S) and topography (T), in contrast, explained together almost twice as much variation in floristic composition (10.2%). Grassland age (A) explained only 1.8% of the variance. Table 5. Results of selected CCA analyses adjusted for the spatial component (SP), isolating the effect of the age of sampled grasslands (A), geomorphological subunit (GU), management type (M), soil chemical parameters (S), and topography (T) as explanatory variables on the vegetation (n = 166). Expl. Var. = Explanatory variables; Covar. = Covariables; Eigenv. = Sum of all canonical eigenvalues – measure for explanatory power of the explanatory variables (total inertia = 2.333); % = percentage of explained variance; F = F-ratio for the test of significance of all canonical axes (test on the trace), P = corresponding probability value obtained by the Monte-Carlo-permutation test (1000 permutations). Expl. Var. Covar. Eigenv. % F P M,S,T,A,GU SP 0.651 27.9 2.576 0.001 SP - 0.088 3.8 3.205 0.001 Net effects GU S,T,A,M,SP 0.198 8.5 2.352 0.001 T S,A,M,GU,SP 0.140 6.0 2.182 0.001 S T,A,M,GU,SP 0.099 4.2 1.856 0.001 M S,T,A,GU,SP 0.097 4.2 2.257 0.001 A S,T,M,GU,SP 0.043 1.8 4.017 0.001 Gross effects GU SP 0.235 10.1 2.416 0.001 T SP 0.193 8.3 2.622 0.001 S SP 0.144 6.2 2.297 0.001 M SP 0.121 5.2 2.396 0.001 A SP 0.056 2.4 4.408 0.001 4.5 Discussion With respect to the impact of site conditions, differences between vegetation types of relatively nutrient-rich (clusters I and II) and nutrient-poor sites (clusters III and IV) were floristically well characterised by indicator species (Table 2). The results of Indicator Species Analysis were supported by ANCOVA results, which confirmed that plant-available phosphorus and potassium as well as the mean Ellenberg nutrient values were significantly higher in younger grassland sites (clusters I and II), in contrast to the species-rich long-term grasslands (clusters III and IV, Table 3). Correlation between organic matter and total nitrogen with grassland age were in line with other studies indicating self-mulching in long- term grasslands (e.g., Gough and Marrs; 1990). IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 26 Since most grassland species lack a long-term persistent seed bank (Thompson et al., 1997), their soil seed banks are rapidly depleted. Additionally, ploughing of grassland allotments will also deplete the soil seed bank of species with long-term persistent seeds (e.g., Bakker et al., 1991). Therefore, species of grassland communities may only re-establish in abandoned arable land by dispersal from elsewhere. Dispersal limitation is a constraint in intensively managed regions where grassland habitats are often fragmented and source populations may be far away from a site (Bischoff, 2002). Even under favourable conditions, with viable remnant populations of species in the vicinity of such sites as in the study area, dispersal is an uncertain and time-demanding process (e.g., Bischoff, 2002). It remains an open question as to what extent the lower species-richness of the nutrient-richer and younger sites can be attributed to enhanced competitive exclusion (resulting from the higher soil fertility levels), compared to factors that limit dispersal. Relative impact of management types and site conditions The degree of floristic variance that was explained by the current grassland management type in CCA was relatively low in the present study (4.2%; Table 5). In contrast, the geomorphological subunit (GU) of the sampled grasslands explained the highest amount of floristic variance. Even after adjustment for all other determinants (soil chemical parameters, topography, grassland age, management type, spatial component), GU retained remarkable explanatory power (8.5%), which can be interpreted as a strong influence of local species pools (Pärtel et al., 1996) on floristic composition of grassland relevés. On the other hand, when accounting for the effect of GU, spatial component and all other explanatory sets (management type, soil chemical parameters, topography, grassland age), each set dropped some percent in the amount of explained variance, but remained significant. The fact that only 27.9% of the total variance in floristic composition could be explained by these sets is not surprising (Lepš and Šmilauer, 2003), given the complexity of the natural communities studied. Another study in the treeline-ecotone of the Swiss Alps reported differences in species-richness between traditionally mown stands compared to sites grazed for up to 50 years by cattle (Fischer and Wipf, 2002). Aftermath and winter grazing has been shown to be important to maintaining the characteristic species composition of upland meadows in northern England (Smith & Rushton, 1994). But why did the current management type prove to be only moderately important in our study? One possible explanation may be the low overall level of land-use intensity compared to other studies and the lack of any steep gradient in the intensity of current grassland IMPACT OF SITE AND MANAGEMENT ON THE DIVERSITY OF CENTRAL EUROPEAN MESIC GRASSLAND 27 management. In contrast, Zechmeister et al. (2003) demonstrated the impact of land-use practices on species-richness along a strong gradient in land-use intensity with a range of fertilisation levels between 0 and 168 kg N ha-1 year-1. The total species-richness of these grasslands differed significantly in response to the level of fertiliser application (below or above 90 kg N ha-1 year-1) and in response to mowing frequency (range from two to four cuttings per year). Studies of modern grassland management often investigate the influence of the cutting regime together with the level of fertilisation because these are commonly connected in practice. Conclusions for management practices In accordance with other studies our analyses show that there is a wide variety of low- intensity management options to contribute to the maintenance of mesic grassland diversity in Europe (e.g., Cousins and Eriksson, 2002; Vandvik and Birks, 2002; Zechmeister et al., 2003; Pykälä, 2005). Even though they have been applied for the least 25 years, currently applied management types did not appear to cause severe constraints on current floristic composition and species-richness of the studied grassland sites. In particular, meadows with an intermediate disturbance impact on the plant biomass, the sward, and the topsoil may support higher species-richness. Differences in phytodiversity were only partly related to the variation of low-intensity management types in this study. Edaphic parameters, topography and regional scale geomorphology had a relatively higher impact on species composition. Our results also suggested that historical land use with fertiliser application on ex-arable fields caused a differentiation in the productivity of the stands that has lasted to the present. Acknowledgements We are grateful to Norbert Hölzel for critical comments and suggestions in statistical analysis and data interpretation. Josef Scholz-vom Hofe provided assistance in field work and the soil chemical analyses, Oliver Ginzler helped in collecting the relevés. We thank Dietmar Simmering for critical revision of this paper and Lutz Eckstein and two anonymous referees for helpful comments and linguistic improvement. We are very grateful to the landowners who allowed us to use their fields. This study was carried out as a part of the Deutsche Forschungsgemeinschaft (DFG) project ‘Land-use Options for Peripheral Regions (SFB 299)’. We would like to thank the DFG for financial assistance. SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 28 5 Seed bank diversity in mesic grasslands and their relation to vegetation, management and site conditions Camilla Wellstein, Annette Otte & Rainer Waldhardt Journal of Vegetation Science: accepted 5.1 Abstract Question: (i) Is there an impact of different management types (i.e., hay meadow, silage meadow, meadow-pasture, pasture) on the size and composition of the seed bank of mesic grassland (Arrhenatheretalia)? (ii) How strong is the effect of management on the seed bank in relation to aboveground vegetation, edaphic factors and land-use history? (iii) Are there differences in C-S-R plant strategy types and seed longevity between managements? Location: Lahn-Dill Highlands in central-western Germany. Methods: Aboveground vegetation and the soil seed bank of 63 plots (at 21 sites) in mesic grasslands were studied. Differences between management types in quantitative seed bank traits and functional characteristics were tested by ANOVA. The impact of management, aboveground vegetation, site conditions and land-use history on seed bank composition were analyzed by partial CCAs. Results: Management had no significant impact on species richness and density of the seed bank but significantly influenced their floristic composition and functional characteristics. CCA revealed that even after adjustment for soil chemical parameters and aboveground vegetation management still had significant impact on seed bank composition. ANOVA revealed that silage meadows contained higher proportions of R-strategy compared to hay meadows. In contrast, in hay meadows and meadow-pastures proportions of S-strategy were higher than in silage meadows. Conclusions: The type of grassland management has relatively little impact on quantitative seed bank traits. Management types with a high degree of disturbance lead to an increase of species following a ruderal strategy in the seed bank. Irrespective of management type only a limited proportion of characteristic grassland species is likely to re-establish from the seed bank after disappearance from aboveground vegetation. SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 29 Keywords: grazing; mowing; semi-natural grassland; land-use; marginal region; C-S-R strategy; seed accumulation index (SAI). Nomenclature: Wisskirchen & Haeupler (1998) Abbreviations: ANOVA = Analysis of variance; C-S-R = Competition-Stress-Ruderality; CCA = Canonical Correspondence Analysis; SAI = Seed Accumulation Index. SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 30 5.2 Introduction Despite a sharp decline in mesic grasslands all over Europe due to intensification and abandonment (e.g., Mac Donald et al. 2000), these habitats are still a typical feature of marginal regions, mainly within mountainous areas associated with traditional low intensity management. They have been a topic of several studies on the aboveground flora focussing on the maintenance and restoration of species rich plant communities, for the ecological evaluation of land-use and for modelling purposes. Soil seed banks are a source for re- establishment of species which are lost from the aboveground vegetation. Hence, maintenance and restoration of species rich grasslands will also depend on the soil seed bank (Grubb 1977; Fenner & Thompson 2005). In contrast, the investigation of seed banks is concentrated on productive agricultural habitats such as fertile grasslands. Therefore relatively little is known about the seed bank communities and the respective ecology of species occurring in less productive semi-natural grasslands (Thompson et al. 1997: 21). Basically, the composition of a seed bank depends on the contribution of present and former aboveground plant communities (Rice 1989), seed rain from adjacent areas (Hutchings & Booth 1996) and on seed longevity (Rice 1989). Especially the historical composition of the aboveground vegetation has often been identified as one key factor that determines the subsequent composition of the seed bank (Bekker et al. 1997). However, few studies on soil seed banks investigated the impact of different types of current management. Changes in land-use and management practices can have very distinct impacts on the seed bank and the established vegetation (Bekker et al. 1997; Smith et al. 2002) since they alter disturbance regimes (e.g., Gibson et al. 2005), whereas the impact of site conditions on soil seed bank is considered to be mainly indirect. For example, the soil nutrient level is an important factor influencing composition of aboveground vegetation rather than having a direct species specific influence on seeds. Disturbance of grassland swards through grazing creates gaps for seed germination (Burke & Grime 1996), while at the same time limiting the rate of recolonization (e.g., Bakker et al. 1996, Osem et al. 2006). Also different cutting dates and frequencies can have large impact on aboveground vegetation composition (e.g., Zechmeister et al. 2003) through differences in species regenerative abilities. Germination of small seeds is supposed to be favoured through open vegetation and topsoil disturbances (Grime 2001; Fenner & Thompson 2005) which are typical of grazed sites. Reactions to disturbance differ since species are adapted in distinct ways to these impacts. Species groups can be differentiated through C-S-R strategies. The triangular model of SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 31 ecological primary strategies (Grime 1988) discriminates strategies of competitiveness, stress tolerance and ruderality using resource availability and disturbance as two orthogonal dimensions for plant classification. Moreover, differences of seed attributes such as seed mass are important in determining seed bank behaviour. Seed mass and seed longevity were found to be negatively correlated since smaller seeds are more likely to become buried (Bekker et al. 1998a; Hölzel & Otte 2004). Due to its disturbance tolerance and the capacity of fast colonization, the ruderal (pioneer) strategy, along with a higher proportion of seeds with lower seed mass, would be expected mainly in management types with higher disturbance impact, e.g. pastures and silage meadows. Theory predicts a close relation between the degree of disturbance in a habitat and the percentage of species with long-term persistent seed banks (Thompson et al. 1998 (p.168); Grime 2001; Hölzel & Otte 2004). To evaluate the regeneration and maintenance potential of species rich plant communities of mesic grasslands from the seed bank, the persistence of seeds can be assessed by calculating the seed accumulation index (SAI) which has high correlation with the seed longevity index (Hölzel & Otte 2004). The SAI expresses the tendency of a species to accumulate seeds in the soil relative to its cover in the established vegetation using the frequency and abundance of seeds in the soil seed bank relative to the frequency and abundance of a species in the aboveground vegetation. Much research, however, suggests that seed longevity of grassland plant species is low (Rice 1989; Thompson et al. 1997; Bekker et al. 1998b, Bekker et al. 2000). Given this background, the aim of this study was to assess and quantify the seed bank diversity of mesic grasslands and its relation to aboveground vegetation, seed longevity of species, current management and site conditions. Grasslands of the Lahn-Dill Highlands (Hesse, Germany) have been subject to traditional low intensity management ranging from grazing and mowing without fertilization to silage meadows with low fertilization and early mowing until today. These management practices provide strong differences in disturbance impact ranging from hay meadows with only one or two cuts per year to pastures grazed from May to September and silage meadows mown three times a year. The main objective of this study was to analyse the effects of management on seed bank diversity. Specifically we hypothesised that (1.) current management regime has a significant impact on seed bank composition even after adjustment for other important factors such as overlying vegetation type and edaphic conditions, and (2.) management regimes with greater SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 32 levels of disturbance (silage meadows, pastures) select for more ruderal plant functional types, which is reflected in the composition of the seed bank. 5.3 Material and Methods Study region The Lahn-Dill-Highlands cover about 900 km2 of Hesse, Germany. These highlands are a typical example of marginal rural landscapes, which are characterised by relatively unfavourable abiotic conditions for cultivation (Frede & Bach 1999) such as cool climate and shallow soils. Since the 1950s, the Lahn-Dill Highlands have been subject to major agricultural land-cover changes, consisting mainly of a decline in arable land and an increase in grassland and fallow land (Waldhardt & Otte 2003; Hietel et al. 2005). In many places, non-intense grassland use has successively replaced the traditional, extremely small-parcelled crop production and crop/grassland rotation. In the study region a large part of the grassland is managed according to EU-based agri-environmental schemes offered by the state of Hesse, focussing on low intensity grassland use. Current grassland management and land use history The current grassland management was classified based on information obtained from personal interviews with farmers conducted prior to the study in 2003. Four management practices were differentiated: (i) hay meadows with late mowing in mid-June and one or two cuts per year, (ii) silage meadows with early and frequent mowing (three cuts per year) and little fertiliser input, (iii) meadow-pastures with late mowing and subsequent cattle grazing, and (iv) pastures grazed by cattle from May to September. Except type (ii) that obtains up to 30 kg N ha-1 year-1 all other managements are characterized by a present lack of fertiliser input. All selected sites were managed accordingly for the last 25 years, i.e. since 1979. We checked the former duration of grassland use for each site by visual interpretation of available black-and-white aerial photographs from the years 1953, 1962, 1967, 1973, and 1979. We quantified the age (A) of grassland sites according to the determined land-cover (land use) at each of these dates. The studied grasslands are either long-term-grasslands or were under arable cultivation before 1979. Thus, the factor age is a measure of land use history. Sampling design A total of 21 sites (five to six sites per management type) spread over the whole study region were chosen for seed bank and vegetation sampling. On each site, three randomly situated SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 33 permanent plots of 25 m² were analysed. To avoid edge effects, for each permanent plot a minimum distance of 10 m to the border of the site was kept. Sampling of vegetation and seed bank Species composition of vascular plants in aboveground vegetation was sampled in 2003 and 2004 on the 25 m² permanent plots. Species cover-abundance was visually estimated on a modified Braun-Blanquet-scale (with cover degree 2 subdivided into 2a and 2b). Sampling of seed banks was carried out in February 2004 when cold stratification had taken place naturally during the winter period. Within each 25 m² permanent plot, 20 cores of 10 cm depth were taken at random locations using a soil corer of 3 cm in diameter. After removing the litter layer soil cores were divided into 0-5 cm and 5-10 cm sections. Thus, the data represent the soil seed bank in the strict sense without the superficial diaspore litter deposited during the preceding vegetation period. The soil samples represent 141 cm² of the soil surface and 1410 cm³ of the soil volume in each plot and were thus well above the minimum requirements for studies of seeds in grasslands (Oomes & Ham 1983). Using this sampling setup the minimal detectable density with a 95% confidence level was ca. 214 seeds m-² based on a Poisson distribution of seeds (Thompson et al. 1997). Following the seedling emergence method (Roberts 1981; Thompson et al. 1997) the soil samples were concentrated by eliminating the coarse stones and vegetative plant material and transferred in a 2 cm thin layer to 18 cm x 28 cm styrofoam trays. The trays were exposed under warm greenhouse conditions and watered regularly. Seedlings were identified and removed once every week and later once every few weeks. Unidentifiable seedlings were transplanted into pots and grown until identification was possible. In case of identification to the genus level only, seedlings were pooled (Betula ssp. and Carex ssp). After 4 ½ months <1% germination and no additional species were recorded and the experiment was terminated. Site characteristics/Soil nutrients In autumn 2003, soil samples were collected by randomly taking 20 cores of 10 cm depth and 3 cm diameter within each of the permanent plots. We determined total nitrogen (Nt) and total carbon (Ct) levels using a CN-analyser (FlashEA 1112, Thermoquest), amounts of plant available phosphorus (PCAL) and potassium (KCAL) using the calcium-acetate-lactate (CAL) extraction method and pH (in CaCl2) of the fine soil. As there were no limestone formations in the sampling areas the total carbon of the soil represents the organic carbon and was used to calculate the organic matter in the soil samples. SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 34 Data analysis Cover-weighted averages of the Ellenberg indicator values for soil nutrients (NV) and soil moisture (MV) given by Ellenberg et al. (1992) were calculated for aboveground vegetation of each plot. Similarity between established vegetation and seed bank of each plot was determined using the Euclidian distance after standardizing the data (Z-transformation) (McCune & Grace 2002). To evaluate the regeneration and maintenance potential of target communities of mesic grasslands, the persistence of seeds was assessed by calculating the seed accumulation index (SAI, Hölzel & Otte 2004). The SAI is a continuous estimator of seed persistence which combines two indices to express the relationship between the presence of a certain species in aboveground vegetation and in the soil seed bank. The first index relates the plot frequency of a certain species in the soil seed bank (SBf) to its frequency in aboveground vegetation (AVf) and seed bank: AV/SBfreq index = (SBf / (SBf + AVf)) * 100 (1) The second index relates the total number of seeds recorded in the seed bank over all plots (SB q) to the cumulative cover of a certain species over all plots (AVq) plus the total number of seeds: AV/SBquant index = (SBq / (SBq + AVq)) * 100 (2) Both indices range between 0 (only present in aboveground vegetation) and 100 (only present in the soil seed bank). To integrate quantitative aspects of species occurrence in aboveground vegetation and seed bank, Hölzel & Otte (2004) merged the two indices into a single one, the SAI, by the addition of both indices and division by two. SAI was calculated across all sampled plots (n = 63) for all 207 species recorded in the study (Appendix 1). For seed bank species, mean abundance-weighted (i.e. weighted for seed density) SAI value was calculated for each plot. Abundance-weighted (i.e. weighted for seed density) calibrated C-S-R strategy types by Grime et al. (1988) were calculated for each plot based on the species found in the soil seed bank. The calibration of unbalanced C-S-R radii for species was performed according to Ejrnæs & Bruun (2000). We allocated a total of 60 points to each species divided on the three strategies C, S and R. A species recorded as R/CSR was consequently assigned to ten points for competitive ability, ten points for stress tolerance and 40 points for ruderal adaptation. Only species categorized by Grime et al. (1988) were included in the analysis when considering C-S-R as dependent variable; they SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 35 comprised about 80% of the entire species pool and all of the frequent and abundant species. Uncategorized species always had low abundances, i.e. less than five seeds per plot. Data on seed mass of common species were derived from Korsmo (1930), Grime et al. (1988), Hölzel & Otte (2004) and Otte et al. (2006). Mean abundance-weighted seed weight of seed bank species was calculated for each plot. To test for differences between the four management types in (i) site characteristics, (ii) similarity between established vegetation and seed bank and (iii) differences in functional traits of seed bank species, we calculated one-factorial analyses of variance (Table 1, Fig. 2, Fig. 3). Prior to analyses, for each variable tested mean values for each site were obtained by averaging the values of the respective three plots and the numbers of seeds were log- transformed to fulfill the assumption of normally distributed data required for ANOVA. In case of significance, ANOVA was followed by the Tukey HSD test for unequal sample sizes. To keep the type I error at 5% despite multiple testing, the significance level α was adjusted by the sequential Bonferroni procedure (Holm 1979). ANOVA and associated tests were carried out using STATISTICA 6.0 (Anon. 1998). Two-way analysis of variance was performed to test the effects of the four management types and the two soil depths on (a) species richness (α-diversity) of seed bank species (calculated as mean species number per 25 m²) and (b) seed density (calculated as mean number of seeds m-²). In a second step, ruderal species, i.e. typical species of disturbed habitats according to Ellenberg et al. (1992) and Oberdorfer (1994) which occurred only irregularly were excluded from the analysis. For this reduced data set of the seed bank with grassland species only, which made up 70% of the species in the seed bank, tests (a) and (b) were repeated. The second analysis was considered to be more meaningful for assessing the potential of the seed bank for contributing to the maintenance of mesic grassland vegetation. Untransformed mean values of species richness and seed density for each management type and soil depth are given in Table 2. The relative importance of current management, aboveground vegetation and other important factors on floristic composition of the seed bank was quantified by a series of Canonical Correspondence Analyses (CCA). In order to assess the appropriateness of CCA, we first subjected the seed bank species data set to a Detrended Correspondence Analysis (DCA) (McCune and Grace 2002). DCA revealed a gradient length of 3.23 SD on the first axis, which indicates the appropriateness of a unimodal response model (Ter Braak & Smilauer SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 36 1998). We also performed a DCA of the aboveground vegetation species matrix resulting in three ordination axes representing 23.52% of the total variance in species data. The sample scores of the three DCA-axes were subsequently used as constraining (‘environmental’) variables in canonical correspondence analyses (CCA) of the seed bank matrix. Prior to CCAs, all examined variables were submitted to a forward selection procedure (Lepš & Šmilauer 2003). They all yielded significant contributions (p < 0.05) to data structure and were therefore retained for CCA. To receive ecologically interpretable variance components, the explanatory variables were grouped into the following four sets: Management type (M; including four different management types); soil chemical parameters (S; including organic matter, total nitrogen, pH value, and plant available P and K contents); Aboveground Vegetation (AV; DCA axes scores of vegetation relevès); grassland age (A). In order to isolate the effect of these sets of explanatory variables on seed bank composition we performed a decomposition of variance by running a series of partial CCAs as proposed by Ter Braak & Šmilauer (1998, p.258). All ordinations were done with CANOCO 4 software (Ter Braak & Šmilauer 1998). Significant indicators of management types were detected by the method of Indicator Species Analysis (Dufrêne & Legendre 1997, McCune & Grace 2002). To calculate the indicator value of a species, its mean abundance in one group compared with its mean abundance in all groups is multiplied by its relative frequency in the samples of that group. The obtained values were tested for significance with a Monte Carlo permutation test (1000 random permutations). Indicator Species Analysis was done with PC-ORD (McCune & Mefford 1999). 5.4 Results The aboveground vegetation contained 151 taxa, 56% (84 species) of these were represented also in the soil seed bank. A total of 9209 seedlings emerged from the soil samples which could be assigned to 140 different taxa, 128 in 0-5 cm and 107 in 5-10 cm depth. The observed seed densities ranged from 5 225 to 24 421 with a mean of 10 367 seeds m-² over all sites. The soil seed bank was dominated by a few species which occurred with high densities (see Appendix 1): Trifolium repens (1274 seeds m-²), Agrostis capillaris (1029 seeds m-²), Plantago lanceolata (952 seeds m-²), Juncus bufonius (883 seeds m-²), Leontodon autumnalis (778 seeds m-²), Poa trivialis (545 seeds m-²), and Cerastium holosteoides (482 seeds m-²). SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 37 Fig. 1. Absolute frequencies of species of disturbed habitats, grassland habitats and other habitats in different seed accumulation index (SAI) classes (n = 100 species; only species with a frequency of at least three plots of 21 were included). These seven species contributed 50% of all seeds found, whereas the majority of species had lower seed densities (see Appendix 1). The application of the SAI to our data set resulted in a continuous and differentiated ranking of species (Appendix 1). Species with high accumulation of seeds in the soil, i.e. high SAI, which are likely to form a long time persistent seed bank were mostly scarce or absent in aboveground vegetation. Exceptional species which were frequent in the seed bank but also in the aboveground vegetation were Anthoxanthum odoratum, Cerastium holosteoides, Holcus lanatus, Leontodon autumnalis, Luzula campestris, Plantago lanceolata, Poa trivialis and Veronica chamaedrys. A comparison of absolute frequencies of frequent species (n = 100, frequency at least three from 21 sites; Appendix 1, Fig. 1) of different habitat types over five classes of SAI revealed that most species of disturbed habitats (86%, n = 22) and a remarkable proportion of species from grassland habitats (51%, n = 61), such as Anthoxanthum odoratum and Poa trivialis, showed high SAI values (SAI ≥ 50). The species with preferences for ‘other habitats’ with high SAI values (71%, n = 17) consisted mostly of plants related to disturbed situations e.g. Solidago virgaurea, Epilobium angustifolium, and Aethusa cynapium or anemochorous pioneers like Betula ssp. and Salix caprea as well as undetermined groupings of Carex ssp. Species with low SAI values (SAI < 50) were mostly grassland species. Differences between management types Tests of several environmental factors for differences between management types (Table 1) revealed that all variables had non-significant effects and showed no ecologically meaningful differences among managements. As revealed in one-way ANOVA (Table 1), similarity between species composition of established vegetation with those of the respective seed bank D = species of disturbed habitats (n = 22) G = species of grassland habitats (n = 61) others (n = 17) SEED BANK DIVERSITY IN MESIC GRASSLANDS AND THEIR RELATION TO VEGETATION, MANAGEMENT AND SITE CONDITIONS 38 (Euclidian distance) was not affected by management type, and also mean SAI of seed bank species did not show significant differences between management types. Two-way ANOVA revealed no significant differences in the quantitative traits species richness (α-Diversity) and seed density between management types, but soil depth was a significant factor for these quantitative traits (p ≤ 0.0005, adjusted α = 0.0125). There was no interaction between management and soil depth. The result was identical when only grassland species were considered. Comparing the two depth fractions, the upper soil layer contained higher species richness (seed bank; seed bank of grassland species only) (29 vs. 21; 25 vs. 16) and seed density (7700 vs. 2667; 7088 vs. 1915) (Table 2). Table 1. Abiotic site conditions, similarity between aboveground vegetation and seed bank and the seed accumulation index (SAI) of four management types. Mean values ± standard deviation for sites of each management type are given. For the variable ‘age of grasslands’ the range is given in brackets. Differences between management types were not significant (one-way ANOVA). Management type Hay meadow (n = 6) Silage meadow (n = 5) Meadow-pasture (n = 5) Pasture (cattle) (n = 5) Environmental factors Ag