Habitat area but not habitat age determines wild bee richness in limestone quarries

Authors

  • Jochen Krauss,

    Corresponding author
    1. Department of Animal Ecology I, Population Ecology, University of Bayreuth, Universitätsstrasse 30, D-95447 Bayreuth, Germany
      *Correspondence author. E-mail: Jochen.Krauss@uni-bayreuth.de
    Search for more papers by this author
  • Thomas Alfert,

    1. Agroecology, University of Göttingen, Waldweg 26, D-37073, Göttingen, Germany
    Search for more papers by this author
  • Ingolf Steffan-Dewenter

    1. Department of Animal Ecology I, Population Ecology, University of Bayreuth, Universitätsstrasse 30, D-95447 Bayreuth, Germany
    Search for more papers by this author

*Correspondence author. E-mail: Jochen.Krauss@uni-bayreuth.de

Summary

  • 1Within highly modified European landscapes, limestone quarries can act as important secondary habitats for a range of endangered wild bee species. However, the relative influence of quarry habitat area, habitat age and within-habitat diversity on the conservation value of these secondary habitats is mainly unknown.
  • 2We assessed species richness and abundance of wild bees by variable transect walks in 24 limestone quarries ranging in size from 0·01 to 21·2 ha. Species traits such as social status, resource specialization and nesting substrate were used to define functional guilds of bees.
  • 3In total, 41% of all wild bee species known from southern Lower Saxony, Germany, were found in the studied quarries. Total species richness increased with habitat area but, in contrast to our expectations, not with habitat age, although we tested an age gradient of over 120 years. Solitary species were more strongly affected by decreasing habitat area than social species but response did not differ with respect to habitat age.
  • 4Hierarchical partitioning analyses revealed that habitat area per se, within-habitat diversity and sampling effort were of similar importance in explaining species richness patterns.
  • 5Synthesis and applications. Even newly created limestone quarries provide an important secondary habitat for wild bees. Therefore, maintenance and management of secondary succession of these sites should be given high priority in species conservation plans. Landscape management schemes involving filling or flooding such quarries should be prevented.

Introduction

In Europe, quarries and other secondary habitats (e.g. sand pits, waste rock heaps, brown field sites) created by human exploitation have been identified as valuable conservation areas for plants, vertebrates and several invertebrate groups (Davis 1979; Wheater & Cullen 1997; Prach, Pysek & Bastl 2001a; Benes, Kepka & Konvicka 2003; Gilcher & Tränkle 2005; Tropek & Konvicka 2008). Although local conservation authorities are often confronted with the evaluation and possible restoration of secondary habitats, their importance has been predominantly neglected in scientific research (Gilcher 1995; Poschlod et al. 1997; Prach et al. 2001b; Gilcher & Tränkle 2005). As a consequence, little is known about basic ecological patterns and processes such as species–area relationships, community structure, and secondary succession for these habitats.

While there is consensus that the size of quarries plays an important role for species conservation (Poschlod et al. 1997; Benes et al. 2003; Gilcher & Tränkle 2005), the influence of age of quarries is still unclear. Conservation value has been assumed to increase with the age of quarries (Davis 1979), and for plants, there is some evidence of increasing conservation value with increasing successional age in abandoned quarries (Novak & Konvicka 2006; but see Gilcher & Tränkle 2005). However, xerophilious butterflies have been shown to prefer young and even working limestone quarries compared to old quarries (Benes et al. 2003). Apart from a few well-replicated studies (e.g. Benes et al. 2003; Gilcher & Tränkle 2005; Novak & Konvicka 2006), most other studies focusing on the conservation value of quarries are rather descriptive (e.g. Davis 1979; several case studies in Poschlod et al. 1997).

Following the theory of secondary succession, species richness should increase and community composition should change with successional age (Odum 1969; Brown & Southwood 1987). Accordingly, early successional stages are expected to be dominated by mobile generalists with high dispersal capacity and rapid reproduction, while later stages are expected to develop a higher floristic and structural complexity. Both total species richness and the proportion of specialized species are expected to increase with age, since a higher predictability of resources and smaller niche breadths are expected to enhance the more competitive specialist species compared to generalist species (Brown & Southwood 1983). Non-linear shifts in species richness are possible and intermediate succession stages might contain the highest species numbers according to the intermediate disturbance hypothesis (Grime 1973; Connell 1978). For example, species richness of wild bees in fallows peaked after 2 years of early secondary succession and remained constant during the following years (Steffan-Dewenter & Tscharntke 2001).

Derived from the island biogeography theory, the positive relationship between habitat area and species richness is one of the few basic laws in ecology (e.g. Rosenzweig 1995). However, the factors influencing species–area relationships are still not well understood and three alternative explanations have been developed: (i) the area per se hypothesis, which assumes that species richness increases with area independent of habitat diversity; (ii) the habitat heterogeneity hypothesis, which assumes that the increasing number of micro-habitats (within-habitat diversity or habitat heterogeneity) within a defined larger habitat cause the increase in species richness; (iii) the sampling hypothesis, which assumes that more intensive sampling of larger habitats compared to smaller habitats causes the increase in species richness (Rosenzweig 1995; Ricklefs & Lovette 1999). Hierarchical partitioning is a recently developed statistical method that might solve this dispute by calculating the relative importance of each explanatory variable within multi-factor models. This approach is useful when the explanatory variables are correlated with each other, because independent contribution and joint contribution to the explained variance of the model are calculated for each independent variable separately (Heikkinen et al. 2005).

Apart from the general pattern of increasing species richness with increasing habitat area, it is also recognized that species traits modulate responses to habitat and landscape characteristics (Ewers & Didham 2006; Hambäck et al. 2007). For example, it has been shown that habitat-specialist butterfly species were more strongly affected by decreasing habitat area than generalist species (Krauss, Steffan-Dewenter & Tscharntke 2003). Wild bee species differ greatly in their life histories and several functional guilds can be distinguished based on social status, resource specialization and preferred nesting substrate. As bees are essential pollinators for crops and wild flowering plants, wild bees perform a key function in human-dominated and natural or semi-natural ecosystems (Aguilar et al. 2006; Albrecht et al. 2007; Klein et al. 2007; Winfree et al. 2007). Bee species are also of conservation concern and have been recorded frequently in limestone quarries (Westrich 1989, 1990).

In this study, we analysed the relative influence of habitat area, habitat age, and within-habitat diversity for wild bee species richness, abundance and community composition in limestone quarries. We developed the following predictions, according to the theories of island biogeography and secondary succession:

  • 1Wild bee species richness and abundance increase with habitat area and habitat age.
  • 2Specialist wild bees, like solitary, parasitic, and oligolectic (utilizing a narrow range of pollen sources) species, are negatively affected by habitat loss to a greater extent than generalist bees.
  • 3Specialist and above-ground nesting wild bee species colonize newly created quarries later than generalist and soil-nesting bees.
  • 4Positive species–area relationships can be explained by combined effects of habitat area per se, habitat heterogeneity and sampling effort.

Materials and methods

study sites

A total of 24 limestone quarries in the vicinity of the town of Göttingen in Lower Saxony (Germany) were selected as study sites. The criteria for the selection of the 24 study sites out of 64 quarries in the study region were: (i) the quarries exhibited typical quarry floor and slope characteristics; (ii) the full available gradients across habitat area and habitat age was covered; (iii) habitat area and habitat age of the selected sites were not correlated (see Table 1); (iv) the quarry slopes were south-east to south-west facing; (v) access was permitted by the owners. The age of the selected quarries ranged between 1 and 121 years and was determined from information from quarry owners and the evaluation of historical maps (Kurhessische Landesaufnahme 1878, Preussische Landesaufnahmen 1910–1919, Preussische und aktuelle Messtischblätter 1919–1991). The total area of each of the quarries ranged between 0·01 and 21·17 ha. To define within-habitat diversity (habitat heterogeneity), the area of seven vegetation types (open, sparsely herbaceous, grassy, herbaceous, sparsely bushy, bushy, forest) was estimated separately for floor, slope and edge region of the quarries. Most measurements were taken using a Differential GPS (GEOmeter 12 L, GEOstat GmbH, Wuppertal, Germany) but very small areas were measured by hand. Within-habitat diversity was calculated using the Shannon–Wiener diversity index (Magurran 1988) up to a maximum of 21 different micro-habitat types. As species richness of plants in flower might be seen as a surrogate for within-habitat diversity for pollinators, flowering plant species richness was recorded each time when bee transects were conducted. The plant records covered the whole period of wild bee activity in the year of study. It is recognized that flower occurrence and availability in previous years might also have affected the observed species richness and abundance of wild bees, but there was no data to allow us to consider this in this study.

Table 1.  Pearson correlations between five explanatory variables
 Habitat ageSampling effortPlant species richnessWithin-habitat diversity
  1. Significance levels: ****P < 0·0001; ***P < 0·001; *P < 0·05; (*)P < 0·1; NS, not significant.

Habitat area (log10-transformed)−0·23NS0·94****0·74****0·41*
Within-habitat diversity0·06NS0·39(*)0·68*** 
Plant species richness−0·17NS0·72****  
Sampling effort−0·31NS   

data collection

Wild bees and honey bees (Hymenoptera: Apiformes) in each of the 24 study sites were recorded seven times between 19 April to 3 September 1999. The sequence of surveying the study sites was changed each time and arranged to combine small and large, and young and old habitats on any one survey day. The surveys were conducted from 10·00 to 16·00 h in spring and autumn and from 09·30 to 18·00 h in mid-summer. Surveys were only conducted when the temperature was above 16 °C with low wind (< 3 Beaufort) and there was at least 70% sunshine (cloudless sky). Each transect focused on a 2-m corridor to ensure good detection of small bee species, and covered all regions and vegetation types in the quarries. The sampling effort was adjusted to habitat area. Transect time in small quarries (< 0·1 ha) was set at 15 min, increasing to 30 min in 0·1–0·5 ha sites, 45 min in 0·5–1·0 ha, 60 min in 1·0–3·0 ha sites, 75 min in 3·0–10·0 ha sites and 90 min in very large quarries (> 10·0 ha). Transects were conducted at approximately constant pace and the average distance of a 15-min transect length in all study sites was 331 ± 21 m (as measured with a step counter during the last survey). An increase in sampling effort with habitat area has the advantage that the proportion of sampled area covered by transects in large habitats is comparable to the proportion of area sampled in small habitats (Krauss et al. 2003). It also allows a more realistic estimate of total species richness within each site. Within each 15-min interval, the floor, slope and edge region of the quarry were surveyed and as many different vegetation types as possible were visited. The sampling effort within the quarries concentrated on areas where high species richness of wild bees was feasible. This variable transect approach is recommended for wild bee recording (Westphal et al. 2008). All Apiformes were identified to species level and collected if identification was not possible in the field. To protect hibernated young queens, in the first five survey times per study site, Bombus species were only classified into three groups which could be identified in the field (i.e. Bombus terrestris, B. pascuorum and B. lapidarius groupings). In the sixth and seventh survey, all bumble bees were collected to identify them to species level. Bee species identification followed Scheuchel (1995, 1996) for Antophoridae, and for Megachilidae and Melittidae, respectively; Schmidt-Egger & Scheuchel (1997) for Andrenidae; Ebmer (1969–1971) for Halictus and Lasioglossum; Warnecke (1992) for Sphecodes; Dathe (1980) for Hylaeus; Mauss (1986) and Prys-Jones & Corbet (1987) for Bombus and Psithyrus; and Schmiedeknecht (1930) for Colletes. All bee species were allocated according to Westrich (1989) to one of two categories for each of four life-history traits, that is, social vs. solitary (including communal and colonial), parasitic vs. non-parasitic, endogaic (nesting in soil) vs. non-endogaic (epigaic including endo/epigaic and six unknown species) and polylectic (pollen generalists, i.e. use pollen from several plant families for brood rearing) vs. oligolectic species (pollen specialists, i.e. use only pollen from one plant family for brood rearing) (see Supporting Material, Table S1). For the statistical analyses only non-parasitic bees were considered in the social vs. solitary bee comparison to exclude effects of the heterogeneous parasitic bee group.

statistical analyses

The statistical analyses were performed using the software r 2·5·0 for Windows (r Development Core Team 2004) and Statgraphics Centurion XV Version 15·1·03. The response variables ‘bee species richness’ and ‘number of bee individuals’ met the assumptions of normality and homoscedasticity in the statistical models. For the comparison of slopes for species groups in species–area relationships, untransformed species richness and, additionally, log10 + 1-transformed species richness (z-values), were calculated. To increase linearity of relationships, the explanatory variable ‘habitat area’ was always log10-transformed, while all other explanatory variables, habitat age, within-habitat diversity, plant species richness and sampling effort, showed linear relations with the response variables. General linear models were used to test the effects of habitat area and habitat age on bee species richness and number of individuals. Linear mixed effects models were used to test for differences between species groups (categories) in relation to habitat area and habitat age, with the fixed effects entering the model in the following sequence: category, habitat area, habitat age, category × habitat area interaction and category × habitat age interaction; site identity was used as random effect to correct for pseudo-replication (Pinheiro & Bates 2004). Hierarchical partitioning models (hier.part package in r) were used to provide an estimate of the relative importance of the correlated explanatory variables (i) habitat area per se, (ii) within-habitat diversity (and flowering plant species richness), and (iii) sampling effort, to the explained variance of bee species richness and number of bee individuals in multi-factor models (Chevan & Sutherland 1991; MacNally and Walsh 2004; Heikkinen et al. 2005).

Results

A total of 6863 wild bee individuals from 124 species were collected in the 24 limestone quarries (see Supporting Material, Table S1). Another 8860 individuals belonging to the domesticated honey bee Apis mellifera were excluded from all analyses. The 124 wild bee species represent 41% of the wild bee fauna of southern Lower Saxony, Germany (Theunert 2003). Most of the observed wild bee individuals belonged to the genera Bombus (3597 individuals) and Lasioglossum (1182 individuals).

Habitat area of the quarries was a strong predictor for bee species richness (explained variance 77%) and bee abundance (63%), while habitat age did not show any significant relationships (Table 2; Fig. 1). The strong habitat-area effect was significantly different for species with different life-history traits. As predicted, species richness of solitary bees decreased more steeply with decreasing area than did social bees (Tables 3 and 4; Fig. 2a). Solitary species were also more strongly affected by decreasing habitat area than social bee species, when species richness was log-transformed (P < 0·001). In contrast to expectations, the slopes of species–area relationships were steeper for non-parasitic and polylectic compared to parasitic and oligolectic bee species (Tables 3 and 4; Fig. 2c,e). However, parasitic species (including social and solitary species) and especially oligolectic species were species-poor guilds and were completely missing from some small habitats. In addition, slopes of log-transformed species richness did not differ significantly between parasitic vs. non-parasitic (P = 0·122) and oligolectic vs. polylectic (P = 0·504) species richness. Species richness of endogaic bees decreased more steeply with decreasing area than non-endogaic species (Tables 3 and 4; Fig. 2g). Slopes of endogaic vs. non-endogaic log-transformed species richness was not significant (P = 0·616). The habitat age of the quarries showed no significant effect on bee species richness of any species group (Tables 3 and 4; Fig. 2b,d,f,h).

Table 2.  Multiple regression analyses, with two explanatory variables of limestone quarries, habitat area and habitat age on the response variables bee species richness and bee individuals (see Fig. 1). Habitat area is log10-transformed
  d.f.FPSlopeInterceptR2
Bee species richnessHabitat area1,2170·81< 0·000116·05−29·830·77
Habitat age1,21< 0·010·959−0·1035·06< 0·01
Bee individualsHabitat area1,2136·73< 0·0001214·90−517·320·63
Habitat age1,210·280·601−0·88328·31< 0·01
Figure 1.

Effects of habitat area and habitat age of limestone quarries on species richness (a, b) and abundance (c, d) of wild bees. For the statistics, see Table 2.

Table 3.  Mixed effects models for the two explanatory variables habitat area and habitat age and the species groups (category) and their interactions on the response variable bee species richness. Habitat area is log 10-transformed
  d.f.FP
Sociality
social, solitaryCategory1,21152·06< 0·0001
Habitat area1,2170·10< 0·0001
Habitat age1,21< 0·010·943
Category: habitat area1,2168·88< 0·0001
Category: habitat age1,210·230·642
Parasitism
parasitic, non-parasiticCategory1,21120·35< 0·0001
Habitat area1,2170·10< 0·0001
Habitat age1,21< 0·010·943
Category: habitat area1,2115·260·0008
Category: habitat age1,210·370·550
Food specialization
polylectic, oligolecticCategory1,21321·57< 0·0001
Habitat area1,2170·10< 0·0001
Habitat age1,21< 0·010·943
Category: habitat area1,2154·77< 0·0001
Category: habitat age1,21< 0·010·979
Nesting habitat
endogaic, non-endogaicCategory1,2152·99< 0·0001
Habitat area1,2160·97< 0·0001
Habitat age1,210·050·819
Category: habitat area1,218·360·009
Category: habitat age1,211·270·272
Table 4.  Simple regression for species richness of each species group (category) separately for habitat area and habitat age (see Fig. 2). Habitat area is log 10-transformed
 SlopeInterceptR2PSlopeInterceptR2P
Socialitysocial   solitary   
Habitat area2·62−1·490·54< 0·000113·42−28·280·78< 0·0001
Habitat age−0·018·61< 0·010·759−0·0926·390·050·305
Parasitismparasitic   non-parasitic   
Habitat area4·69−10·350·59< 0·000111·35−19·420·70< 0·0001
Habitat age−0·017·86< 0·010·700−0·0927·140·050·289
Food specializationpolylectic   oligolectic   
Habitat area14·52−25·850·75< 0·00011·51−3·910·56< 0·0001
Habitat age−0·0932·900·040·356−0·012·100·020·553
Nesting habitatendogaic   non-endogaic   
Habitat area10·64−18·510·62< 0·00014·89−9·270·60< 0·0001
Habitat age−0·0925·760·060·2480·018·74< 0·010·889
Figure 2.

Effects of habitat area and habitat age on species richness of different functional guilds: social versus solitary species (a, b), parasitic vs. non-parasitic species (c, d), polylectic vs. oligolectic species (e, f) and endogaic vs. non-endogaic species (g, h). For the statistics, see Tables 3 and 4.

Habitat area alone explained 77% of variance of bee species richness. As habitat area was correlated with sampling effort, within-habitat diversity and plant species richness (Table 1), we used hierarchical partitioning to reveal the independent effects of these correlated explanatory variables. A general linear model containing these three variables explained 81% of variance of bee species richness. The hierarchical partitioning model indicated that habitat area per se explained 46% of variance, within-habitat diversity 13%, and sampling effort explained 41% of variance (Fig. 3). For the number of bee individuals, sampling effort explained a larger proportion of the variance than habitat area but the effect of within-habitat diversity was still small. When we replaced within-habitat diversity by plant species richness as an indicator of within-habitat diversity, the amount of variance explained was split in roughly similar proportions between the three variables, highlighting the greater importance of plant species richness compared to habitat diversity (Fig. 3). Considering also the combined effects of each variable did not substantially change the relative importance of the variables. On average, the combined effects of each variable explained a similar proportion of variance as the independent effect of each variable, resulting in a very similar total amount of variance explained by each variable (results not shown).

Figure 3.

Results of hierarchical partitioning of independent effects of habitat area per se, within-habitat diversity or plant species richness and sampling effort on bee species richness and bee individuals. Plant species richness explains a greater amount of variance compared to habitat diversity.

Discussion

Our results show that limestone quarries are important secondary habitats for wild bees. Wild bee species richness and community structure depend strongly on habitat area but habitat age did not appear to affect wild bees. Solitary bee species, which can be considered as specialists, were affected more strongly by lower habitat area than social bee species.

The negative effects of habitat fragmentation, and especially the reduction of habitat area, have been shown before for wild bees (Aizen & Feinsinger 1994; Steffan-Dewenter 2003; Aguilar et al. 2006), while other studies question the importance of habitat fragmentation for pollinator diversity (Cane 2001; Cane et al. 2006). It is still unclear which species are the first to go extinct when habitat area declines (Holt et al. 1999). We show that solitary bees are more sensitive than social bee species to limitations in habitat area. This is in accordance with another study of wild bees in calcareous grasslands (Meyer B, Krauss J, Steffan-Dewenter I, unpublished data) and is also significant if z-values (log-log regression slopes) are compared (Steffan-Dewenter et al. 2006).

Regarding the comparison of parasitic vs. non-parasitic bee species and polylectic vs. oligolectic species, our results seem to contradict our predictions, as in theory, higher trophic level and specialist species are expected to be more strongly affected by habitat fragmentation than species at low trophic ranks and more generalist species (Holt et al. 1999; Ewers & Didham 2006). However, the comparison of slopes for functional groups represented by very different numbers of species might give misleading results. The oligolectic species group in particular contained only 16 species and 257 individuals and, hence, was possibly too small to show consistent results. Furthermore, the analyses of log-log regression slopes between parasitic versus non-parasitic and polylectic versus oligolectic species did not show significant differences. The differences between parasitic and non-parasitic bee species in their dependence on habitat reduction might be also explained by the heterogeneous parasitic species group including social bees of the genus Psithyrus and solitary bee species (e.g. from the genus Nomada). Without further evidence, any conclusions for these species groups must remain speculative.

Soil-nesting (endogaic) species were more strongly affected by lower habitat area than above-ground nesting (epigaic) bee species. It could be speculated that larger quarries contain more heterogeneous habitats for the species-rich endogaic species compared to epigaic species. However, differences were not significant, when comparing z-values of the two groups (Steffan-Dewenter et al. 2006).

In contrast to habitat area, we found no significant effects of habitat age of the limestone quarries on wild bee community structure. This is surprising, as our study sites represented an age gradient of 120 years and secondary succession theory predicts an increase of species richness. We assumed that particularly old quarries might harbour relict populations of wild bee species with a formerly more widespread distribution. Earlier studies and observations showed increasing plant species richness with time since abandonment of quarries (Davis 1979; Novak & Konvicka 2006; but see Gilcher & Tränkle 2005), and bee species richness is known to increase with increasing flowering plant diversity (e.g. Potts et al. 2003; Albrecht et al. 2007; Holzschuh et al. 2007; Holzschuh, Steffan-Dewenter & Tscharntke 2008). Our data also support the close relationships between flowering plant species richness and bee species richness and abundance. Nevertheless, as plant species richness was not related to habitat age in our study, habitat age was independent from this factor and had no significant effect on bee species richness or abundance. Moreover, there is no evidence that young limestone quarries contain different species communities than older quarries, as was assumed by Gilcher (1995). According to our data, habitat age of limestone quarries is only of minor importance for wild bees, and this might suggest that wild bees have the ability to colonize new quarries within a short time period. However, age of quarries might be important for other species guilds with low dispersal capacity like reptiles or for those with high dependence on structural complexity like birds (Poschlod et al. 1997). Any conservation efforts aiming to protect high species richness of wild bees should therefore primarily focus on large quarries.

As there is disagreement whether large habitats contain more species due to their larger size, their larger within-habitat diversity (habitat heterogeneity) or due to increasing sampling effort (Rosenzweig 1995; Ricklefs & Lovette 1999), we conducted a statistical procedure to estimate the relative importance of the assumed causal explanatory variables. We showed that habitat area per se (25–46%) and sampling effort (28–55%) both contributed substantially in multi-factor models, accounting for 81–89% and 77–86% of the amount of explained variance in bee species richness and bee abundance, respectively. Habitat diversity, measured as Shannon–Wiener index of different vegetation types in three different limestone regions, accounted for only a minor proportion of the explained variance (10–13%), while flowering plant species richness accounted for substantially more (34–40%).

In conclusion, it is best for the conservation of wild bee species to protect large limestone quarries with high vascular plant species richness, independent of the age of the quarry. A high proportion of wild bee species can be protected by conserving these quarries, and it is possible that this may even help compensate for the loss of natural and semi-natural habitats (e.g. calcareous grasslands). Management schemes which aim to fill or flood such quarries should be avoided. Our recommendations are similar to the findings from previous studies on other groups, such as birds, butterflies and vascular plants (Benes et al. 2003; Gilcher & Tränkle 2005; Novak & Konvicka 2006), and hence, it is feasible that multiple conservation benefits could be achieved through an appropriate selection and management of limestone quarries.

Acknowledgements

We thank Andreas Erhardt, Martin Konvicka and one anonymous reviewer for helpful comments, Davy McCracken and Carmela Herrmann for editorial comments, Rainer Theunert for taxonomic support and the owners of the quarries for logistic help. This study was financially supported by the Sixth European Union Framework programme (EU-Project COCONUT, contract no. 2006-044346 to I.S.D. and J.K.).

Ancillary