Recent declines in biodiversity have given new urgency to questions about the relationship between land-use change, biodiversity and ecosystem processes. Despite the existence of a large body of research on the effects of land use on species richness, it is unclear whether the effects of land use on species richness are principally direct or indirect, mediated by concomitant changes in ecosystem processes. Therefore, we compared the direct effects of land use (fertilization, mowing and grazing) on species richness with indirect ones (mediated via grassland productivity) for grasslands in central Europe.
We measured the richness and above-ground biomass in 150 grassland plots in 3 regions of Germany (the so-called Biodiversity Exploratories). We used univariate and structural equation models to examine direct and indirect land-use effects.
The direct effects of mowing (−0.37, effect size) and grazing (0.04) intensity on species richness were stronger compared with the indirect effects of mowing (−0.04) and grazing (−0.01). However, the strong negative effect of fertilization (−0.23) on species richness was mainly indirect, mediated by increased productivity compared with the weak direct negative effect (−0.07).
Differences between regions in land-use effects showed five times weaker negative effects of mowing (−0.13) in the region with organic soils (Schorfheide-Chorin), strong overall negative effects of grazing (−0.29) for the region with organic soils opposed to a similar strong positive effect (0.30) in the Hainich-Dün region, whereas the Schwäbische Alb region displayed a five times weaker positive effect (0.06) only. Further, fertilization effects on species richness were positive (0.03) for the region with organic soils compared to up to 25 times stronger negative effects in the other two regions.
Synthesis. Our results clearly show the importance of studying both direct and indirect effects of land-use intensity. They demonstrate the indirect nature, via productivity, of the negative effect of fertilization intensity on plant species richness in the real-world context of management-induced gradients of intensity of fertilization, mowing and grazing. Finally, they highlight that careful consideration of regional environments is necessary before attempting to generalize land-use effects on species diversity.
Recent global declines in biodiversity have given new urgency to questions about the relationship between land-use change, biodiversity and ecosystem processes, above all productivity (Foley et al. 2005; Balvanera et al. 2006; Fischer et al. 2008). Clearly, land-use change is a major driver of plant species richness, both at local and regional scales (Sala et al. 2000). In Germany and most of central Europe, grasslands comprise about a quarter of the agricultural area and harbour a high diversity of plant species. Most of the grassland is intensively used for hay making and for cattle herding (Statistisches Bundesamt 2009).
Generally, more intensive land use leads to higher productivity, mostly driven by fertilization. Moreover, experimental (Silvertown 1980; Gough et al. 2000; Crawley et al. 2005) and observational (Grime 1973; Al-Mufti et al. 1977; Thompson et al. 2005) studies usually show that species richness declines with increasing site productivity, at least in managed grassland (Adler et al. 2011). Together, this suggests that land-use effects on species diversity may be largely indirect, via changed productivity.
However, different components of land use, including fertilization, mowing and grazing, may also directly affect species diversity (Zechmeister et al. 2003; Stewart & Pullin 2008). Direct effects of mowing involve early and frequent destruction of above-ground plant organs or a generally low ability to resprout, resulting in a rapid decline of sensitive species and enrichment of more disturbance-tolerant ruderal taxa (Grime 2001). Further effects of early and frequent mowing may include the partial or even complete failure of seed production leading to seed and recruitment limitation of the affected species. On the other hand, mowing may strongly enhance seedling germination by the removal of litter and above-ground biomass and the creation of gaps that act as suitable micro-sites for the establishment of small-seeded species (e.g. Foster & Gross 1998). Further, mowing can remove nutrients with the harvested biomass, affecting nutrient levels and biomass production (Oelmann et al. 2009). The seasonal release hypothesis combines several of the above processes in proposing that seasonal dieback of above-ground material can act to release the community from the competitive suppression of the recruitment process and further act to maintain higher species richness (Huston 1994; Grace 2001).
Generally, grazing can have similar effects as mowing, but in a much more patchy manner in space and time (Rook et al. 2004; Klimek et al. 2007). Depending on stocking densities and grazer species, more unpalatable taxa might benefit from grazing while others are suppressed. At low intensity, mowing as well as grazing impose moderate degrees of disturbance and nutrient stress that usually strongly enhance species diversity in grasslands, whereas high levels or the complete absence of disturbance usually lead to a drastic decline in species richness (Huston 1994). Thus, the effects predicted by the intermediate disturbance hypothesis (Grime 1973) are not only mediated by the impact of disturbance on the amount of standing crop but also by seed limitation at high-disturbance levels and micro-site limitation at low disturbance levels.
Fertilization affects species richness in grasslands via increased above-ground biomass, height and competition for light (Hautier, Niklaus & Hector 2009). Compared with competition for soil resources, competition for light is more asymmetric and thus more likely to drive inferior species to extinction (Grime 1973). Similarly, seedling recruitment is strongly hampered by raised biomass levels (Tilman 1993; Foster & Gross 1998; Gough et al. 2000). Competition for light and seedling recruitment is probably the most decisive for the decline in species richness under high-nutrient levels induced by land use through fertilization (Leps 1999).
Disentangling direct and indirect effects using structural equation modelling (Shipley 2002) allows exploration of the question of whether direct and indirect effects act in the same direction and whether they are of similar importance. Previous studies concentrated on the relative contribution of productivity and disturbance in natural grasslands (e.g. Grace et al. 2007), but studies separating direct and indirect land-use effects in central European grasslands are still scarce and mostly do not cover real-world gradients of land-use intensity (Grace & Pugesek 1997; Gough & Grace 1999; Grace & Jutila 1999; Leps 1999).
Several meta-analyses (Kleijn et al. 2006; Paillet et al. 2010) have revealed large variation in land-use effects on biodiversity among taxa, habitats and regions. Land use will depend on climate, geography and socioeconomic settings. It is therefore important to study land-use effects in different geographical regions to distinguish general patterns from regionally idiosyncratic responses. We studied plant species diversity and productivity on 150 grasslands in 3 different regions considering the impact of 3 components of land-use intensity – amount of fertilizer, mowing frequency and grazing intensity – to address the following main questions:
How important are direct effects of fertilization, mowing and grazing on plant species richness compared with indirect effects mediated via productivity?
To what extent do these effects differ among regions?
Materials and methods
We studied grasslands in three regions in Germany, the so-called Biodiversity Exploratories (Fischer et al. 2010a). All regions are comprised of a large variety of differently used grasslands and forests in areas of c.1000 km2 (Fischer et al. 2010a,b). The Exploratory Schorfheide-Chorin in Brandenburg (NE Germany) is situated in the area of the Biosphere Reserve Schorfheide-Chorin. The Exploratory Hainich-Dün in Thuringia (Central Germany) is situated in the National Park Hainich and surrounding areas. The Exploratory Schwäbische Alb in Baden-Württemberg (SW Germany) is situated in the Biosphere Reserve Schwäbische Alb. In each of the regions, grasslands are surrounded by arable fields, forests and settlement areas.
In each region, we studied 50 grassland plots, consisting of unfertilized and fertilized meadows, pastures and mown pastures (grasslands both grazed and mown within the same year).
Because soil types differ between plots, within and across the regions, a soil sample was taken from each plot and its soil type classified according to the World Reference Base of Soil Resources (IUSS Working Group WRB 2007), the pH of the soil was also measured (after Schlichting & Blume 1966) in each plot (Fischer et al. 2010a,b). Topographical information on elevation and slope was derived from digital terrain models.
A questionnaire submitted to farmers and land owners provided information on land use in the plots (Fischer et al. 2010a). Farmers were asked about the livestock type (cattle, both horse and cattle, or sheep) on pastures and mown pastures. For meadows and mown pastures, the number of cuts per year (from one to four times) was assessed. Moreover, the questionnaire asked whether a grassland site was fertilized or not and whether it was additionally sown with commercial seeds. Additional quantitative information about livestock density, grazing duration and the amount of nitrogen in the fertilizer used on the grassland was also obtained.
Vegetation and biomass samples
Between mid-May and mid-June 2009, we collected vegetation data and above-ground biomass from the plots. Vegetation data were taken in an area of 4 × 4 m situated next to the soil core. In addition, we recorded the percentage cover of each vascular plant species and the total plant cover, and calculated the total cover of three functional groups (herbs, legumes and grasses). All species with very low cover, or those consisting of only single individuals, were assigned a cover of 0.5%. We used these data to obtain the summed cover of all vascular plant species (cover) and the total species number of vascular plants (species richness).
Biomass was clipped in 8 50 × 50 cm plots 2─3 cm above-ground adjacent plots in which cover was recorded. In meadows, our biomass sampling was carried out at the same time as the first cut by the farmer. We also harvested in subareas of pastures and mown pastures, which had been temporarily fenced to ensure that the vegetation had not been grazed before our sampling. The biomass samples were dried for 48 h at 80 °C and weighed immediately after drying. For statistical analyses, the average of the 8 biomass measurements was used.
Linear model (anova)
Firstly, we did univariate tests on the effects of land use on species richness (mean number of vascular plant species on 4 × 4 m), productivity (biomass in g m−2) and the abundance of different functional groups (the cover of herbs, legumes and grasses). In the model, we included region, slope, soil type and soil pH as additional explanatory variables, which were followed by the variables describing the land use. These consisted of the amount of nitrogen fertilizer applied to the field, the number of cuts, day of first cut, number of livestock units per hectare, grazing days (number of days per year animals were grazing on the land), livestock type (sheep, cattle, both cattle and horse) and additional sowing (yes or no), and the interactions of region with the land-use variables. We simplified the model in a stepwise algorithm based on the Akaike Information Criterion (AIC), as implemented in R version 2.10.0 (R Development Core Team 2009), to select the model with the lowest AIC.
Structural equation modelling
As a second – multivariate – approach, to examine the relationship of land use with productivity and species richness more comprehensively and to distinguish between direct and indirect effects, we used structural equation modelling (SEM) for each region separately and for all three combined (Shipley 2002). Three continuous land-use variables were chosen to describe the intensities of mowing, grazing and fertilization: the number of cuts per year, grazing intensity (=number of livestock units × grazing days per year) and the amount of nitrogen fertilizer added to the plot per year. The structural equation model reflected the fact that all three land-use variables affect both productivity and species richness and that productivity affects species richness (indicated by one-way arrows in the path diagrams in Fig. 1). Thus, the structural equation models considered the direct effects of fertilization, mowing and grazing on species richness, along with the indirect effects of these same factors on productivity and, subsequently, species richness. Moreover, we assumed that the number of cuts and the amount of fertilizer correlated with each other (indicated by a two-way arrow in the path diagrams). We calculated the structural equation models in R (version 2.10.0) using the package SEM (R Development Core Team 2009).
Nitrogen fertilization (0–333 kg N−1 ha−1; mean 35 kgN−1 ha−1, min 27 kg N−1 ha−1 on non-fertilized plots) caused a mean drop in species richness of 19% (P <0.001) and a 14% higher cover of grasses (P =0.007). Biomass increased by 13% (P =0.034) in response to nitrogen fertilization (Table 1, univariate models).
Table 1. Summary of analyses of variance of land-use effects on species richness, productivity and cover of vascular plants, herbs, legumes and grasses for 150 grasslands in three regions in Germany
Cover vascular plants
*P <0.05, **P <0.01, ***P <0.001.
Amount of nitrogen (N)
Number of cuts (NC)
Day first cut (DFC)
Live stock unit (LSU)
Grazing days (GD)
Live stock type (LST)
Additional sowing (AS)
E × N
E × NC
E × DFC
E × LSU
E × GD
E × LST
E × AS
Residual mean squares
With increasing cutting frequency (from one to four cuts) species richness decreased by 36% (34.5 to 22.3, independent of region, P <0.001). The earlier the day of the first cut, the lower was species richness (P =0.006) and the higher the cover of grasses (P =0.002).
Higher numbers of livestock units grazing on the grassland slightly decreased the total cover of vascular plants (P =0.046). Longer grazing duration increased the cover of herbs (P =0.0003) and the total cover of vascular plants (P =0.008). Productivity was highest on plots grazed by cattle and horses, and lowest on sheep pastures (P =0.01).
The total cover of vascular plants was highest on plots grazed by cattle and lowest on plots grazed by sheep (P =0.02), and the cover of herbs was lowest on plots grazed by both cattle and horses (P =0.01) and highest on plots grazed by sheep.
In summary, the amount of nitrogen was the only explanatory variable, which affected both species richness and productivity. Further, land use affected the functional type composition of the grassland.
In addition, regional effects were important for all variables. Moreover, several significant interactions between region and land-use variables indicate regional variation in land-use effects (Table 1; for more results on regional effects see below).
Structural equation modelling: general outcome of relationships between land use, productivity and species richness
According to the parameter values (goodness-of-fit index and Bayesian information criterion), the model for the Schwäbische Alb provided the best fit to the data, and the model for the Schorfheide-Chorin provided the least close fit (Table 2). The number of cuts and the amount of nitrogen added to the grassland were highly significantly correlated with each other for all four models. The negative correlation of species richness with productivity was most pronounced in the overall regions model and between regions became weaker from the Hainich-Dün to the Schwäbische Alb to nonsignificant in the Schorfheide-Chorin (Fig. 1).
Table 2. Output of each of the four structural equation models of relationships of the intensities of mowing, grazing and fertilization with the productivity and species richness of 150 grasslands in three regions in Germany (see Fig. 1)
RMSEA, root mean square error of approximation; BIC, Bayesian information criterion.
The effects of management on species richness and productivity in the SEM were very similar to the ones indicated by the linear models. The three exceptions were as follows: the correlation between the amount of nitrogen added and species richness was far stronger in the linear model than in the SEMs, the amount of nitrogen fertilizer added had significant positive effects on species richness in the Schorfheide-Chorin SEM, and the number of cuts had a strong and highly significant negative effect on productivity in the Schorfheide-Chorin SEM.
Structural equation modelling: direct versus indirect land-use effects on species richness
The direct negative effect of higher numbers of cuts on species richness was much stronger than the indirect one via productivity (Table 3, Fig. 1). For the Schorfheide-Chorin this small indirect effect was positive.
Table 3. Direct, indirect and overall effects of the intensities of mowing, grazing and fertilization on species richness of 150 grasslands in three regions in Germany
Number of cuts
The negative direct effect of fertilization on species richness was much weaker than the overall strong indirect one via productivity. Only for Schorfheide-Chorin was this direct effect positive, outweighing the negative indirect effect (Table 3 and Fig. 1). Compared with the effects of mowing and fertilization, the effects of grazing were weak: the positive direct effect of grazing intensity on species richness was weaker than the negative indirect effect via increased productivity. Only for Schorfheide-Chorin this direct effect was negative, adding to the negative indirect effect.
There was significant variation among the three regions in all tested variables, indicating strong regional differences (Tables 1 and 4). Highest biomass was observed in Schorfheide-Chorin (338 g m−2), followed by Hainich-Dün (292 g m−2) and Schwäbische Alb (270 g m−2) (Fig. 2). Species richness was lowest (20 ± 0.6, mean ± SE) in Schorfheide-Chorin, and far higher both in Hainich-Dün and Schwäbische Alb (by 40%, Table 4).
Table 4. Mean values and standard errors (SE) of species richness, productivity and cover of vascular plants, herbs, legumes and grasses for 150 grasslands in three regions in Germany
Productivity (g m−2)
Cover vascular plants
Cattle and Horse
No of cuts
Among the regions, grasslands in the Schorfheide-Chorin had the highest total cover of grasses, while grasslands in the Schwäbische Alb had the highest cover of legumes and herbs (Table 4). Overall, these findings reflect strong regional effects on grassland productivity, species richness and functional group composition.
Effects of environmental covariates (elevation, slope, soil type, pH)
Plots at higher elevation harboured more plant species (P <0.001). Plots on steeper slopes had higher cover of herbs (P =0.001) and grasses (P =0.005; Table 1). The cover of herbs was lower in plots of lower soil pH (P =0.004; Table 1). These significant results confirm that it is important to include these covariates in models when analysing effects of land use.
We showed that mowing and fertilization have a larger impact on grassland species richness than grazing does. Further, mowing and grazing rather had direct effects on species richness, whereas the negative effect of fertilization on species richness was mainly indirect, mediated by increased productivity. Moreover, regional differences modulated these findings.
Direct effects of land-use intensity on plant species richness, functional groups and on productivity in grasslands
Many previous studies have demonstrated the effects of single, often categorical, land-use parameters including fertilization (Gough et al. 2000; Crawley et al. 2005), grazing (Collins 1998; Klimek et al. 2007) and mowing (Zechmeister et al. 2003) on species richness. Our study design involved many sites with detailed quantitative information on environmental and land-use variables and therefore allowed us to quantify and interpret the relative strengths of these effects.
As we considered not only fertilization per se, but also quantified the amount of fertilizer, we were able to quantify the mean effect of fertilization on species richness as −19%/35 kg N per hectare and year overall. The decrease in species richness of 37% (per 35 kg N per hectare and year) in Schwäbische Alb (the region with the most species-rich grasslands), 13% in Hainich-Dün, and 5% in Schorfheide-Chorin are similar to results in Stevens et al. (2004). They reported a decrease of 38.4% of plant species because of 25 kg N per hectare per year. These results emphasize the importance of considering regional effects when quantifying impacts of fertilization.
Grasslands with higher mowing intensity had lower species richness than those that were mown less frequently or later in the year. This agrees with other reports that more frequent and earlier mowing decreases plant species richness (Hansson & Fogelfors 2000; Köhler et al. 2005; Knop et al. 2006; Gross et al. 2009) and supports consistently negative effects of increased mowing frequency on diversity (Zechmeister et al. 2003).
The positive correlation between the number of cuts and fertilization in our study (Fig. 1) illustrates that intensities of fertilization and mowing are generally related to each other (Collins 1998; Dickson & Foster 2011). However, as we combined quantitative information on both fertilization and mowing in one model, we could show that mowing effects remain significant even after correcting for fertilization in linear models (Table 1). Moreover, the structural equation models (Fig. 1) indicate that mowing intensity itself has an important and consistently negative effect on species richness.
The grazing parameters of livestock density, grazing duration and livestock type all affected the total cover of vascular plants, and these effects were also significant after correcting for fertilization. Overall, more intense grazing and its components (density of livestock units, grazing duration and livestock type) had a slightly positive impact on the total cover of vascular plants and herbs. In our study, grazing intensity, assessed as product of the number of grazing days and of livestock units, had a large impact on herb abundance, whereas livestock type had an overall weak impact on the total cover of vascular plants and herbs.
Overall, the strength of effects on species richness and functional group composition differed between fertilization, grazing and mowing. Both fertilization and mowing had a large negative impact on species richness, whereas grazing had a slight positive effect on the cover of plants and herbs, but not their species richness.
Direct versus indirect effects of land-use intensity on species richness
Despite pronounced regional differences (see discussion below), certain patterns were strong enough in some regions to remain highly significant when combining data from all regions. Thus, overall, the direct effects of mowing and grazing intensity on species richness were larger than the indirect ones, whereas the indirect effect of fertilization on species richness via productivity was larger than the direct effect (Table 3, Fig. 1). Land use by region interactions (Table 1) and separate analyses for the three regions (Table 3, Fig. 1) showed weaker negative effects of mowing and even positive effects of fertilization on species richness for the region with organic soils, the Schorfheide-Chorin.
The strong indirect, and additional weak direct, effect of fertilization on species richness underlines the importance of fertilization for affecting species richness through increases in productivity (Grime 1973; Silvertown 1980; Gough et al. 2000; Loreau 2000; Rajaniemi 2002; Schmid 2002; Stevens et al. 2004; Crawley et al. 2005; Kahmen et al. 2005; Harpole & Tilman 2007; Clark & Tilman 2008) and clearly attributes most of it to the indirect pathway via increased productivity. Although a recent analysis of a comprehensive data set on productivity and species richness of many sites around the world did not find a universal relationship at the within-site scale, comparison across sites revealed that species richness declined at high levels of productivity, especially in anthropogenic grasslands (Adler et al. 2011). From a nature conservation point of view, this implies that the conflict of interest between the agricultural goal of increasing productivity and the conservation goal of increasing species richness cannot simply be alleviated by reducing direct fertilization effects on species richness while maintaining increased productivity.
Among the most likely mechanisms for the productivity-mediated indirect fertilization effects on species richness are increased competition (Grime 1973; Wilson & Tilman 1991), reduced light for subdominants (Hautier, Niklaus & Hector 2009) and increasing litter layer (Foster & Gross 1998). Further, the negative effect of soil acidification caused by atmospheric N deposition causes changes in plant species diversity (Bobbink et al. 2010). Direct effects of fertilization on species richness may reflect selection of those species able to deal with and readily use large amounts of available nitrogen.
The land use by region interactions in univariate models (Table 1) and the comparison between the four structural equation models (Fig. 1) clearly showed that the relative contribution of land-use components varied considerably among regions. In particular, the Schorfheide-Chorin region differed from the two other ones, both in direction and strength of the effects of fertilization and mowing. These differences may be associated with differences in soil types, as the grasslands in Schorfheide-Chorin occur mostly on highly organic soils, whereas those of Hainich-Dün and Schwäbische Alb occur on less organic soils. Organic soils are characterized by generally high levels of nutrient concentrations and moisture, both making plants less dependent on additional nutrient supply. This may explain the weak effect of fertilization on species richness in Schorfheide-Chorin. Together with the slightly positive effect of fertilization on species richness in Schorfheide-Chorin, this indicates that general fertilization and productivity effects are difficult to define for all types of grassland (Adler et al. 2011).
Even though the direct and overall effects of mowing intensity on species richness were negative in Schorfheide-Chorin, they were much weaker than in the other two regions. It may be that the combination of mowing and the high soil moisture of organic soils (Schmid, Bolzern & Guyer 2007) favour shallow-rooted grass species (Table 4) and not deeper rooted herb and legume species. Grazing had a strong negative effect on species richness in the Schorfheide-Chorin but weakly positive effects in the two other regions (Table 3). This may be due to different species composition on pastures and mown pastures in the Schorfheide-Chorin and because of differences in soil types between the regions, which may influence the degree to which grazing affects sward heterogeneity. Overall, regional differences in land-use effects on productivity and species richness apparent from interactions between region and land-use intensity (Table 1) and from the different relationships in the separate structural equation models (Table 3, Fig. 1) clearly demonstrate that land-use effects cannot simply be extrapolated across geographical regions.
We showed that mowing and fertilization have a large impact on grassland species richness, whereas grazing had a weaker effect. Further, effects of mowing and grazing on species richness were governed by direct effects, whereas indirect effects mediated by increased productivity governed the effect of fertilization on species richness. Differences in magnitude and sometimes even direction of land-use effects between regions indicate clearly that careful consideration of regional environments is necessary before land-use effects can be extrapolated between regions. Management recommendations that aim to increase plant species richness must therefore take such regional particularities into account.
We thank Eduard K. Linsenmair, François Buscot, Dominik Hessenmöller, Jens Nieschulze, Ingo Schöning, Ernst-Detlef Schulze, Wolfgang W. Weisser and the late Elisabeth K. V. Kalko for their roles in setting up the Biodiversity Exploratories project and Swen Renner, Konstans Wells, Sonja Gockel, Andreas Hemp, Martin Gorke, Simone Pfeiffer and Ilka Mai for maintaining plot and project infrastructure. Further, we thank Ralf Lauterbach, Martin Fellendorf, Jörg Hailer, Uta Schumacher, Ulf Pommer, Claudia Seilwinder, Katrin Wuchter, Nico Straube and several student helpers for their field support and Eric Allan and two anonymous reviewers for helpful comments on the manuscript. For funding we thank the DFG Priority Program 1374 ‘Infrastructure-Biodiversity-Exploratories’ (FI1246/6-1, FI1246/9-1) and the University of Bern. Field work permits were given by the responsible state environmental offices of Baden-Württemberg, Thüringen, and Brandenburg (according to § 72 BbgNatSchG).