5. Discussion
We assessed the effects of local and landscape environmental heterogeneity on ant community structure in semi-natural temperate grasslands. We found that extensively grazed pastures and within-site heterogeneity in soil moisture and topographic exposition at the local scale, as well as diversity of land cover types and forest cover at the landscape scale affected ant community composition and evenness. Our findings additionally confirmed the importance of local mean air temperature and mean soil moisture as environmental filters of ant species establishment in temperate grasslands (Dauber and Wolters 2005, Seifert 2017, Heuss et al. 2019). Furthermore, our trait-based approach showed that the response of overall ant community structure to local and landscape environmental heterogeneity was mediated by a set of species traits including morphological, ecological and life history features.
5.1. Effects of local and landscape heterogeneity on ant species composition
Within-site variability of soil moisture affected ant species composition by promoting nest densities of some species and therefore total nest density per grassland site, while an increase of mean soil moisture between grassland sites explained the decrease of nest densities in our studied system. Although both results on soil moisture may seem contradictory, they should be interpreted as the effect of soil moisture as i ) an environmental heterogeneity measure reflecting variability within grasslands (TWIcv) and ii ) an environmental means measure showing variability between grasslands sites (TWImean;sensu Stark et al. 2017). High levels of soil moisture have been reported to adversely affect ant species richness and density in Central European grasslands, with most ant species having a narrow range of tolerance to water content in soils (Dahms et al. 2005, Dauber et al. 2005, Seifert 2017, Heuss et al. 2019). In this sense, it seems likely that a higher environmental heterogeneity in terms of soil moisture (high TWIcv) exerts a positive effect on ant communities within grassland sites by providing more niche-spaces to be exploited by a wider group of species with different tolerances to wetness. On the other hand, the significant negative effect of mean soil moisture on ants likely reflects the gradient in humidity between grassland sites in the study area, where less humid grasslands (lower TWImean) accommodated more diverse ant communities than more humid grassland sites (higher TWImean). This is in line with a more regional-scale pattern found across Central Europe, where ant diversity decreases from warm-dry to humid-cool grasslands (Dekoninck 2007, Seifert 2017). In contrast to soil moisture, the effect of mean air temperature was more evident at species than at entire community level (total nest density per site) where nest densities of the reportedly thermophilic species L. niger , L. flavus , M. rubra and Formica cunicularia(cf. Figure A4) increased along with Tmean (Seifert 2017, 2018).
In addition to the role of soil moisture, our findings showed that extensively managed pastures promote nest density of most of species compared to extensive meadows, which supports previous results highlighting the benefits of extensive grazing on biodiversity compared to annual mowing (Tälle et al. 2016). Although both extensive management types are expected to be beneficial for biodiversity in semi-natural grasslands, the mechanisms behind their effect on grassland vegetation and fauna may differ (Lepš 2014, Tälle et al. 2016). While low intensity (gradual but continuous) grazing by cattle creates a small-scale mosaic of disturbances in soil and spatial structure of vegetation (vertical and horizontal), annual mowing homogenizes the vegetation structure through consistent and uniform biomass removal in a short period of time (Olff and Ritchie 1998, Adler et al. 2001, Lepš 2014, Tälle et al. 2016). Thus, grazing disturbance not only enables openness throughout the sward but also creates a variety of different micro- sites and micro-climates suitable for a wider group of ground-dwelling insects, including ants (Cole et al. 2010, Hoffmann 2010, Jerrentrup et al. 2014). The differential effect of management types on grassland vegetation structure has been shown to affect ants indirectly by changing micro-habitat and soil moisture conditions (Dahms et al. 2005, Dauber et al. 2005, Pérez-Sánchez et al. 2018, Heuss et al. 2019), which correspond to our data showing higher within-site soil moisture heterogeneity in pastures (TWIcv : mean= 14.7, SD= 5.9) than in meadows (TWIcv : mean=8.9, SD= 4.5). Thus, both environmental heterogeneity measures, management type and soil moisture variability, may address the same synergistic effect of providing favorable micro-habitat and soil conditions for ant communities.
At the landscape scale, a high diversity of land cover types had a general positive effect on species composition, while the influence of the surrounding forest varied among species. The effect of land cover heterogeneity on ant communities can be interpreted in two different ways. First, a higher variability in land cover types in the surroundings of the grassland sites may provide source habitats for new colonizing species, thereby enriching the local species pool (Benton et al. 2003, Öckinger et al. 2012). Second, a higher between-habitat heterogeneity in the surroundings increases the range of abiotic conditions, particularly along the edges, thus causing ecotonal effects that may favor thermophilic species near open habitat edges (i.e. arable, built-up land) or moist-tolerant species near shaded edges (i.e. forest) within grassland communities (Dauber and Wolters 2004). In our study region, the higher levels of landscape diversity are located at intermediate elevations where the land cover matrix transitions from a built up- to forest-dominated landscape and the proportions of neighboring arable lands and grasslands reach their peaks (Figure A1). Supporting the ecotonal effect, the observed species composition in many of these grasslands was characterized by species with different habitat preferences: F. fusca (woodland edges), L. acervorum(light forest), F. cunicularia (open verges), L. niger(urban and disturbed open habitats; Seifert 2018). The species-specific effects of surrounding forests detected in our study further support this statement, as F. fusca density was positively affected by %Forest, while densities of L. flavus and F. cuniculariawere negatively affected (Figure A4).
5.2. Effects of local and landscape heterogeneity on ant community evenness
The effect of environmental heterogeneity on diversity has been explained by classical niche theory, where habitat heterogeneity is expected to increase species co-existence and therefore evenness (Stein and Kreft 2015). However, it has been argued that the high diversity of species in seemingly homogeneous habitats cannot be explained exclusively by niche processes, but also by neutral or stochastic processes related to demography and dispersal events (Hubbell 2001, Andersen 2008, Brown et al. 2013). In this regard, evenness has been used as neutrality metric, meaning that habitats with high levels of community evenness are subject to more neutral than deterministic processes (Schowalter 2011, Bar-Massada et al. 2014). In our study system, an important proportion of grassland sites accommodate even communities, which suggests that neutral processes may also contribute to structuring ant communities in addition to environmental heterogeneity. The fact that ant response traits describing colony reproduction and dispersal constraints in semi-natural grasslands (van Noordwijk et al. 2012) were important in our fourth-corner analysis suggest that neutral processes, such as stochastic species dispersal, may contribute to the local assembly of ant communities. Nevertheless, these findings, although remarkably interesting, need to be considered with caution since neutrality levels in species-poor communities are difficult to interpret as the lower bound of evenness decreases with richness (1/0D ), and in our studied system approximately 25% of the grassland sites showed communities with less than three ant species (Jost 2010, Bar-Massada et al. 2014).
In contrast to our expectations, the effect of environmental heterogeneity on ant community evenness was variable at local scale and negative at landscape scale. The effects of TWIcv, TWImean, and SHID on RLE0,2 were exactly opposite to those observed for species composition and density, hence their interpretation is intimately related to our previous discussion. An increase in within-site soil moisture variability and diversity of surrounding land cover types certainly supports total nest density, but also benefits numerical dominant species (i.e. M. scabrinodis or L. niger ) which may lead to uneven or dominated ant communities in some grassland sites. Similarly, high levels of soil moisture may promote community evenness by limiting nest densities of numerical dominant ants in grasslands, which leads to a more even distribution of species densities within the community structure (c.f. sites in cluster 2, Figure 1). Model results additionally showed a positive effect of within-site surface aspect range (ASPrg) on community evenness. Our study system contains grasslands mostly facing from east to west through south azimuths. In grasslands facing east, the soil surface receives radiation earlier in the day when air temperature and evapotranspiration are lower leading to moister habitats (Ashcroft et al. 2008); while on west-facing sites, the soil surface reaches maximum temperatures during the afternoon when the direct radiation is at its maximum creating warmer and drier habitats (Bennie et al. 2008). A higher range of surface exposure within grasslands sites creates highly heterogeneous habitats for ant species, thereby increasing species co-existence and promoting ant diversity and community evenness.
5.3. Community structure response to environmental heterogeneity through species traits
The community structure response to environmental conditions is the result of different sets of species (response) traits, therefore quantifying how traits and environment interact provides a mechanistic understanding of community assembly (Lavorel and Garnier 2002, Zirbel et al. 2017). Our results show that the response of ant community structure to local heterogeneity on topography, soil temperature, and management was strongly related to behavioral dominance and a generalist life history strategy. Behavioral dominance may be an advantageous trait for food acquisition and territory defense in open habitats characterized by patchy vegetation such as pastures (Mtyppasture) with rugged slopes (SLOcv) where the probability of intra- and inter-specific competitive encounters is high (Savolainen and Vepsäläinen 1988, Savolainen et al. 1989, Pérez-Sanchez et al. 2018; but see Stuble et al. 2019). Similarly, species well adapted to deal with low food availability and variable soil temperatures during nest foundation (life history strategy G, according to van Noordwijk et al. 2012, cf. Table 2), may be favored in dynamic and heterogeneous habitats such as extensively managed pastures (Verberk et al. 2010). Surprisingly, we found a negative relationship between strategy G and within-site solar radiation range (SRDrg), expected to be positive according to the environmental heterogeneity hypothesis. We attribute this result to the overpowering negative effect of within-site soil temperature variability (SDRrg) on L. flavus densities, rather than the cumulative effects on all species with life history strategy G (i.e. L. niger and F. fusca ; van Noordwijk et al. 2012).
Community response to local mean air temperature and soil moisture was strongly related to worker and colony size traits and, to a lesser extent, to life history strategies F and T (cf. Table 2). Higher mean temperatures negatively affected worker size, but positively affected colony size. The relationship between body size and temperature has been traditionally related to the heat conservation hypothesis, i.e. a larger body size has adaptive value in cold temperatures due to lower surface-to-volume ratios (Cushman et al. 1993, Blackburn et al.1999). This hypothesis has been questioned for ants, as a larger body size in ectotherms also reduce the rate of heat gain, which is as important as decreasing heat loss (Cushman et al. 1993). However, it may be suitable at the colony level as long as i ) worker body size is positively correlated to colony size, and ii ) colony size is correlated to thermoregulatory capabilities (Cushman et al. 1993, Kaspari and Vargo 1995). In our case, some Formica species meet such conditions and were certainly associated with cooler temperatures in upland sites or grasslands surrounded by forest (i.e. F. fusca, F. pratensis ). Nevertheless, it is more likely that this particular result is influenced by species with small workers but large colonies such asL. flavus and L. niger , which showed high nest densities in grassland sites with high temperature records. On the other hand, the interaction of life history strategies with temperature suggests that grassland sites with high mean air temperatures have a low nutrient content for ants (strategy F), but may shelter species whose nest foundation is highly dependent on soil temperature (strategy T; van Noordwijk et al. 2012).
The strongest interactions between traits and environment were detected at landscape scale, with a functional response to land cover types (SHID) similar to local mean air temperature (Tmean). Although both predictors, Tmean and SHID, did not exceed the collinearity threshold (|r | < 0.7; Dormann et al. 2013), their effect on ant communities seems to be related. In a forest-dominated landscape, a high number and even distribution of open land cover types (i.e. grasslands, arable lands and urban areas) in the surroundings likely increases air and soil temperature of grasslands; while a decrease in grassland temperatures can be expected when surrounding landscape is dominated by forest due to shading (Figure A2; Krämer et al. 2012, Öckinger et al. 2012, Liivamägi et al. 2014). The strong and negative relationship between life history strategy T and %Forest supports this statement, as species depending on high soil temperature for nest foundation showed low densities or were absent in grasslands surrounded completely or almost entirely by forest. In this sense, a high proportion of surrounding forests not only contributes to grassland isolation by imposing a physical barrier for species colonization, but also limits abiotic conditions for the establishment of grassland specialist and thermophilic species (Krämer et al. 2012, Öckinger et al. 2012).
6. Conclusions
Investigating the impacts of environmental heterogeneity on species diversity and functional traits at multiple spatial scales is crucial to better understand community patterns in ecology, and is especially relevant for biodiversity conservation (Costanza et al. 2011, Ovaskainen et al. 2017). This study shows that local environmental heterogeneity within and between grasslands, and the surrounding landscape play a relevant deterministic role in structuring ant communities in temperate semi-natural upland grasslands. Heterogeneity in habitat structure caused by grassland management (i.e. low intensity grazing) and within-site heterogeneity in soil moisture were the most important local environmental measures shaping ant species composition and density. At the landscape scale, a high land cover heterogeneity increased overall species density within grasslands, although the effect of each surrounding habitat type depended on species-specific abiotic requirements. The variation of community evenness against environmental heterogeneity showed an opposite pattern compared to community composition due to the inherent relationship between richness and evenness (Jost 2010). However, this approach revealed that a higher range of surface aspect (exposure of the slope) within grassland sites exerts a positive effect on community structure by promoting species co-existence. Overall, the proportion of even communities was high (54% of the surveyed sites) compared to uneven (35%) or dominated communities (11%), suggesting additional (neutral) processes operating on ant community assembly (Bar-Massada et al. 2014). The response of ant community structure to environmental heterogeneity at both spatial scales was mediated by morphological (worker size), ecological (dominant behavior) and life history (colony size and strategies related with colony reproduction- dispersal-foundation) traits. The interaction of the aforementioned response traits and the environment framed in a fourth-corner scheme proved to be a valuable approach for understanding how community structure is driven by environmental heterogeneity. We conclude that ant communities in semi-natural grasslands are not only driven by environmental heterogeneity but also by complementary neutral processes related to stochastic habitat colonization, as dispersal events between isolated grasslands are limited in a forest-dominated landscape. Both environmental heterogeneity and stochastic explanations are not exclusive and may represent opposite ends of the same continuous gradient denoting the relative contributions of both niche and neutral processes to the assembly of ant communities in temperate semi-natural grasslands (sensu Gravel et al. 2006).