Effects of habitat and landscape features on grassland Orthoptera on fl oodplains in the lower reaches of the Tisza River Basin

The Tisza River Basin is an important area as it is a green corridor in which there are highly endangered habitats and a high level of biodiversity. The patterns in the species richness of invertebrates and the environmental conditions affecting these patterns are poorly studied in the grassy habitats in the lower reaches of the Tisza River Basin. The present study focuses on the effects of fl ooding, habitat and landscape features on the species richness of orthopterans at 24 grassland sites in two different landscapes. The relations between the explanatory variables and the pattern of diversity of orthopterans with different life-history traits were studied, using ordination and Generalized Linear Mixed Models. Although the infl uential factors for the different trait groups differed, we suggest that landscape features are the most important in shaping orthopteran assemblages, whereas habitat characteristics and fl ooding have comparatively little effect. Habitat characteristics affected only the non-xerophilous and Ensifera species and only the species richness of non-xerophilous orthopterans in fl ooded and non-fl ooded sites differed. We emphasize that even in countries where there are still considerable areas of high value natural grasslands, such as Hungary, non-protected meadows, linear grassy habitats (dikes, ditch banks, road verges, etc.) need more attention and should be given higher priority in the conservation of invertebrates.


INTRODUCTION
In their natural state, riverine landscapes are characterized by mosaics of various habitat patches.Due to their high heterogeneity and connectivity (Naiman et al., 2005), they can support a diverse fl ora and fauna (Gregory et al., 1991;Zwick, 1992;Ward et al., 1999).However, many European rivers are restricted to narrow riverbeds bordered by dikes and the majority of fl oodplain habitats have been transformed into agricultural land (Tockner et al., 2009), causing a severe decline in biodiversity (Godreau et al., 1999).
The River Tisza is the largest tributary of the Danube and its catchment includes most of the Carpathian Mountains covering approximately 157,000 km 2 (Sommerwerk et al., 2009).The regulation of the Tisza in the 19th century caused profound changes; a considerable amount of the former fl oodplain has since never been fl ooded.However, on this non-fl ooded part of the former fl oodplain (so-called "historical fl oodplain") there were several high value habitats, i.e. pastures, woody pastures and hay-meadows, in this extensively used mosaic landscape until the 1950s (Deák, 2007;Sendzimir et al., 2008).During the socialist era, intensifi cation of agriculture resulted in a decrease in the area of these grasslands (Deák, 2007).Nowadays, sweep netting was carried out along three, 50 m long, fi xed transects in 2009.To avoid periods of fl ooding, sweep netting was carried out in summer.The fi rst samples (30 May-2 June) were collected before mowing and the second samples (22-24 July) when the vegetation started to regrow after mowing.For the data analyses, the sweep netting samples were pooled separately for transects and periods, resulting in a total of 24 statistical samples (one sample per site).

Assessments of explanatory variables
Habitat and landscape features were assessed at each site (Appendix 1).Habitat characteristics included features of the vegetation and soil water content.The vegetation was sampled in three, 1 × 1 m quadrats along each transect.Mean data for the quadrats were used to defi ne variables at the sites sampled.To characterize the structure of vegetation, the average height of the vegetation, the total cover of vegetation at 10 and 40 cm above the ground and cover of litter were recorded.To characterize the richness of vegetation, the total number of species of plants and of only the dicotyledonous plants were recorded in the quadrats.Soil samples were taken from the top 10 centimetres close by the coenological quadrats.The percentage of gravimetric water in soil samples was measured.
To assess the amount of habitat we measured the percentage of the area covered with grassland in a radius of 100, 250, 500 and 750 m around each site using ArcView 3.11 GIS software.

Life-history traits
Dispersal ability, niche breadth and reproductive potential were the traits considered, because they are hypothesized to be key determinants of species persistence (Kotiaho et al., 2005).As a measure of dispersal ability, the mobility index (Reinhardt et al., 2005) was used.However, mobility is not a constant trait for orthopterans; it may differ between and within populations (e.g.Endo, 2006;Poniatowski & Fartmann, 2009).To reduce the effect of this potential variability, broad mobility classes: sedentary, intermediate disperser and mobile species, were identifi ed.Further, intermediate dispersers were excluded from the analyses, as they are often the species whose classifi cation is uncertain (cf.Marini et al., 2009a;2012).The mean number of ovarioles is a rough measure of the reproductive potential of females (Reinhardt et al., 2005) and is generally coded into three categories: low (< 10), intermediate (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25) and high (> 25) (cf.Dziock et al., 2011).However, this trait is proportional to body size and the phylogeny of the species (Dziock et al., 2011), and Ensifera species are usually placed in a higher reproductive category than Caelifera (see also our data in Appendix 1).Further, Ensifera species usually produce larger eggs and lay them individually in plants or under tree bark, behaviour which can increase the chance of hydrochory and thus their passive dispersal ability (Dziock et al., 2011).Therefore, Ensifera and Caelifera were used as examples of the differences in reproductive potential and passive dispersal of the species.
Orthoptera clearly differ in their preference for habitats of different humidity, and this trait is often used to group them in relation to their habitat specialization (cf.Fartmann et al., 2012).As most of the collected species preferred dry habitats and the number of hygrophilous species was rather low, we sorted them into two groups: xerophilous and non-xerophilous (hygrophilous and mezophilous) species.

Data analyses
To evaluate the degree of collinearity, Pearson correlation coeffi cients were computed between habitat variables (soil water content, vegetation height, the cover of vegetation at 10 and 40 cm above the ground, cover of litter, plant species richness) prior ian landscapes, fl ooding and land use pressure have a pronounced infl uence in shaping assemblages of Orthoptera (Dziock et al., 2011).The majority of species of Orthoptera are associated with open grassy habitats; therefore, for these species the amount of grassland in a landscape is important (Marini et al., 2008;Badenhausser & Cordeau, 2012).The "habitat amount hypothesis" (Fahrig, 2013) postulates that patch size and patch isolation effects are both due mainly to the sample area effect, thus patch size and isolation can be replaced with a single variable, the amount of habitat.
In order to determine the main factors affecting orthopterans, we tested the effects of (1) habitat characteristics (soil moisture and vegetation structure and diversity), (2) landscape features (amount of grassland habitat, landscape structure) and (3) fl ooding (fl ooded vs. non-fl ooded sites) on the species composition and richness of assemblages of Orthoptera.The effect of landscape composition on assemblages of Orthoptera is often scale-dependent (e.g.Marini et al., 2009a), thus we also aimed to determine the appropriate spatial scale for assessing the amount of grassland habitat.As species with different life history traits often need different environmental conditions, the effects on species richness of various life-history traits were tested separately.

Study sites and sampling design
Assemblages of Orthoptera were studied at 24 sites in two different landscapes.The landscapes were located on the same side of the river in Csongrád County, Hungary, and were selected based on intensity of land use and landscape structure.The heterogeneous landscape (HET) was situated near the town of Szeged (approx.46°17´34 ̋ N, E 20°12´45 E) and it consisted of a mosaic of various habitats.The percentage of the area covered by intensively managed arable fi elds was high (58.3 ± 3.4%, mean ± SE within a radius of 500 m around the sites).Small patches of meadows with trees and abandoned fi elds were embedded in the matrix of arable fi elds.Numerous trees and bushes also occurred along road verges, but continuous forest occurred only near the river.The percentage of the area covered by forest habitats, including single trees and bushes, was 18.6 ± 5.6%.
The homogeneous landscape (HOM) was situated approximately 30 km north of Szeged (approx.: 46°27´27 N, 20°9´26 E).The intensity of agricultural activities in this area was moderate (the percentage covered by arable fi elds was 16.1 ± 1.9%) and the percentage of the area covered with semi-natural grasslands and forests was high.The percentage covered by forest was higher (30.0 ± 3.3%) than in HET due to the oak and poplar forests that bordered and partly divided the relatively continuous grassland area into two parts.
Among the various grassy habitats on the historical fl oodplain of the Tisza, only those that occurred in both landscapes were selected for this study i.e., sand steppe and alkaline meadows in the non-fl ooded part of the historical fl oodplain, dike-slope meadows (strip-like meadows on the slopes of the dikes) and fl oodplain meadows.Each of these habitats occurred at the three sites sampled, giving a total of 24 sites in the two landscapes.The distance between sites was approximately 500 m, except in the case of fl oodplain meadows.Orthopterans were collected by sweep netting, which is a widely used technique for sampling these insects (e.g.Bauer & Kenyeres, 2007;Torma et al., 2014).At each site, to analyses (Appendix 2).As the variables were highly inter-correlated, a Principal Component Analysis (PCA) was carried out and the scores on the fi rst axis, which explained 85.2% of the total variance and correlated signifi cantly with each of the assessed habitat variables (Appendix 2), were used as a habitat descriptor (cf.Poniatowski & Fartmann, 2011;Münsch et al., 2013).The newly created variable represents a gradient from sites with dry soils and uniform vegetation (low values) to sites with moist soils and a high architectural complexity and diversity of vegetation (high values).
Similar to previous studies (e.g.Steffan-Dewenter et al., 2002;Cozzi et al., 2008), as landscape variables quantifi ed at nested spatial scales were obviously highly correlated, we determined the scale which best explained the variation for orthopterans.To evaluate the spatial scale, we used a Generalized Linear Model (GLM, Poisson errors) to describe the relationships between the response and landscape variables (percentage of surrounding area that is grassland) for each radius separately (cf.Marini et al., 2009b).
To analyse the species composition of assemblages of Orthoptera and its relationship with environmental variables, Non-metric Multidimensional Scaling (NMDS, Bray-Curtis dissimilarity) was used and environmental vectors were fi tted onto the ordination space.Generalized Linear Mixed Models (GLMM, Poisson errors, maximum likelihood fi t) were used to test the effects of explanatory variables and factors on the species richness of the Orthoptera.In the GLMM, the effect of habitats sampled was used as a random effect and the selected explanatory variables (habitat descriptor, amount of grassland) as fi xed effects.To test for signifi cances, the effect of landscape (HOM vs. HET) and of fl ooding (fl ooded vs. non-fl ooded sites) were also subjected to GLMM as fi xed effects.Automated model selection was carried out, and the effects of different explanatory factors and variables were averaged across the best models with delta < 2 (Grueber et al., 2011).
All statistical analyses were carried out in an R Statistical Environment (R Development Core Team, 2013).GLMM was performed using the glmer function in lme4 package (Bates et al., 2013); automated model selection was carried out with the dredge function in MuMIn package (Bartoń, 2013).Ordinations were performed in the Vegan package (Oksanen et al., 2013); environmental vectors and factors were fi tted onto ordination space using the envfi t function.
According to the NMDS, the variation in the composition of the samples was caused mainly by the difference between the two landscapes (Fig. 1, Table 1).

Species richness of orthopteran assemblages
The Poisson regressions indicated that the total species richness and the richness of species of both Ensifera and Caelifera were associated in a scale-dependent manner with the percentage of grassland in the surroundings (Fig. 2).The scale with the largest decrease in residual deviance was 500 m.For sedentary and non-xerophilous species, the scale with the largest decrease in residual deviance was  1. also 500 m; however, the differences between the different scales were not large.For all the species collected, the model selection resulted in an average best model that included the effect of landscape, fl ooding and amount of habitat (Table 2).The species richness decreased signifi cantly with increase in the percentage of grassland in the surroundings (z = 2.73, P = 0.006) (Fig. 3a); landscape (z = 1.33,P = 0.184) and fl ooding (z = 1.28,P = 0.199) were not signifi cant.
The best GLMM explaining the species richness of mobile orthopterans included only the effect of landscape (Table 2), which was signifi cant (z = 2.14, P = 0.032) (Fig. 3e).
For sedentary species, the model selection procedure (Table 2) yielded a best model that only included the signifi cant effect of amount of habitat (z = 2.90, P = 0.004) (Fig. 3c).
The average best model for xerophilous species included the effects of landscape and the habitat descriptor (Table 2), with the difference associated with landscape significant (z = 2.10, P = 0.036) (Fig. 3h) but not that associated with the habitat descriptor (z = 1.42,P = 0.156).
For non-xerophilous orthopterans the model selection (Table 2) resulted in an average best model that included the effects of the amount of habitat, the habitat descriptor, fl ooding and landscape.Only the effect of the landscape was not signifi cant (z = 1.26,P = 0.206), but that associated with the percentage of grassland (z = 2.18, P = 0.029) (Fig. 2d), the habitat descriptor (z = 2.29, P = 0.022) (Fig. 2f) and fl ooding (z = 2.25, P = 0.024) (Fig. 2i) were signifi cant.

Effects of landscape features and spatial scale
Based on the scale-dependent effect of the composition of the landscape recorded in the present study, the percent-Fig.2. Scale-dependent effects of the percentage of grassland in the surroundings on the number of all Orthoptera (diamonds), Ensifera (circles), Caelifera (squares), sedentary (triangles) and nonxerophilous (crosses) species.The Poisson regressions between orthopteran species richness and the percentage of grassland in a radius of 100, 250, 500 and 750 m around the sites sampled indicates a decrease in the residual deviance (percentage).In the case of other trait groups, no signifi cant scale-dependent effect of the percentage of grassland was detected.
Table 2.The GLMM components included in the average best models explaining the species richness of the different trait groups of Orthoptera.Abbreviations of explanatory variables and factors: homogeneous vs. heterogeneous landscape (Landscape), fl ooded vs. non-fl ooded sites (Flooding), percentage of grassland within in a radius of 500 m of the site sampled (Habitat amount), scores of the fi rst PCA axis as habitat descriptor (Habitat).age of grassland in the surroundings explained the majority of the variability in orthopteran species richness within a radius of 500 m; this scale has been shown to be relevant in terms of the spatial effects for several arthropod groups (e.g.Schmidt & Tscharntke, 2005;Öberg et al., 2007;Torma & Császár 2013;Torma et al., 2014).With an increase in the percentage of the area covered by grassland, the species richness of orthopterans decreased, which is contrary to the prediction of the habitat amount hypothesis (Fahrig, 2013).However, Fahrig (2013) emphasized that the amount of habitat can be a good predictor for species richness, but is only part of the effect of the surrounding landscape.The studies (e.g.Marini et al., 2008Marini et al., , 2009aMarini et al., , 2010;;Badenhausser & Cordeau, 2012) that report a negative relationship between the proportion of grassland in the surrounding area and Orthopteran species richness emphasize the importance of ecotones.These papers suggest that contrary to the expectation based on the higher mortality of Orthoptera in large mown grasslands, in a landscape with a relatively low amount of grasslands, the local Orthoptera diversity can benefi t from the presence of ecotonal habitats due to what are referred to as rescue effects.
In a previous study, Torma & Császár (2013) show that different landscapes along the lower reaches of TRB host similar assemblages of Heteroptera.Contrary to this, assemblages of Orthoptera in the present study differed in their species composition in the two landscapes, despite sampling similar habitats.Similar to other studies (e.g.Torma et al., 2014) we suggest that different insect groups responded differently to habitat and landscape features, and orthopterans are more affected by landscape features than habitat characteristics in this region.This suggestion is corroborated by the signifi cant positive association between the heterogeneous landscape and species richness of many trait groups recorded in this study.Presumably, in a more heterogeneous landscape, a grassland patch is likely to be colonized by more species from ecotonal habitats as it provides suitable conditions for foraging and reproduction (Marini et al., 2008).

Effect of habitat characteristics
Surprisingly, habitat characteristics affected only the species richness of Ensifera and non-xerophilous species.Non-xerophilous species obviously preferred sites with moist soil and thus more dense vegetation.Species of Ensifera also responded to the habitat characteristics of the sites as they preferred moist sites with more complex vegetation.The importance of humidity for egg and larval development of orthopterans is emphasized in several studies (e.g.Hodek, 2003;Wünsch et al., 2012), and Ensifera generally need more water for egg development than Acrididae (Ingrisch & Köhler, 1998).Differences between Caelifera and Ensifera with regard to vegetation characteristics are also reported by e.g.Marini et al. (2009b), who suggest that Caelifera, unlike Ensifer, prefer regularly mown, less dense vegetation.

Effect of fl ooding
Flooding is considered to be a major disturbance for invertebrate assemblages in riparian habitats (Foeckler et al., 2006).Although the dispersal ability of species is important for the structuring of invertebrate assemblages in fl ooded habitats (Rothenbücher& Schaefer, 2006;Lambeets et al., 2009) before and after fl ooding (Rothenbücher & Schaefer, 2006), we did not fi nd any signifi cant effect of fl ooding on the different mobility trait groups.Instead of active dispersal, Dziock et al. (2011) emphasize the importance of passive dispersal ability of orthopterans in fl oodplains.Certain orthopterans lay eggs in plants or under tree bark and their eggs can be transported together with their substrates by water.This potentially enables these species to colonise even remote areas; however a larger number of offspring (eggs) is needed, because a large proportion of them are likely to end up in unfavourable habitats.Compared to Caelifera, Ensifera species are usually more dependent on passive dispersal and have more ovarioles, but we did not fi nd any signifi cant effect of fl ooding on Ensifera species richness.
However, our results are hard to generalize as the structure and composition of animal assemblages can change rapidly in riparian areas that are frequently fl ooded (Lambeets et al., 2009).Presumably, the results of a study following severe fl ooding would reveal a stronger and more consistent effect of fl ooding.

Implication for conservation
As their diversity is currently declining in many temperate regions, assemblages of Orthoptera are the focus of numerous conservation studies.In fact, more than half of the orthopteran species are endangered in Europe (Ingrisch & Köhler, 1998;Reinhardt et al., 2005).The present study was carried out in habitats of low natural value compared to Natura 2000 and other protected areas in Hungary, despite the occurrence of G. glabra and R. nitidula, which are endangered or critically endangered in surrounding countries (Berg & Zuna-Kratky, 1997;Maas et al., 2002;Krištin et al., 2007;Liana, 2007) and P. vittata, which is close to extinction along the edge of the Pannonian Region (Holusa et al., 2012).These facts confi rm that the habitats in the TRB can serve to maintain high orthopteran (and presumably other invertebrate) diversity in the Pannonian Region.We agree with Hernández-Manrique et al. ( 2012), who conclude that existing conservation strategies, which are based mainly on the protection of certain areas, often selected based on the presence there of particular plant and vertebrate species, may be insuffi cient for ensuring the conservation of invertebrate species.Therefore, we emphasize that even in countries where considerable areas of high natural value grassland still exist, non-protected meadows, strip-like grassy habitats such as dikes, ditch banks, road verges should receive more attention and should be given a major role in the conservation of invertebrate diversity.30.02 20.76 13.33 13.43 13.03 10.4 42.11 60.12 60.32 61.54 81.56 80 33.33 70.01 46.67 50.43 90 66.67 31.01 11.67 21.67 36.67 43.33 41.67 species richness of dicots  17.806 15.686 9.447 6.917 6.444 14.245 24.268 23.191 23.385 19.269 17.162 18.567 19.314 16.964 15.041 23.191 24.269 20.528 10.582 12.438 8.289 19.694 18.390 10.262 Proportion of grasslands in a radius of 100 m

Fig. 3 .
Fig. 3.The signifi cant effects of fl ooding, habitat characteristics and landscape variables on the species richness of different life-history trait groups of Orthoptera delineated using model selection of GLMM.The effects of the explanatory factors and variables were averaged across the best models with delta < 2. Scatter plots show the relations between the percentage of grassland in the surroundings and the total species richness (a), the species richness of Ensifera (b), sedentary (c) and non-xerophilous (d) orthopterans; the relationship between the habitat descriptor and the species richness of Ensifera (e) and non-xerophilous (f) orthopterans.Bar charts represent the differences in the species richness of Caelifera (g) and mobile orthopterans (h) in the two landscapes; the differences in the species richness of non-xerophilous species (i) at fl ooded and non-fl ooded sites.For signifi cances see the results of the model averaging cited in the text.

Appendix 1 .
Collected species of Orthoptera and assessed habitat and landscape features.meadow Dike slope meadow Floodplain meadow Dike slope meadow Alkaline meadow Sand steppe meadow site 1 site 2 site 3 site 4 site 5 site 6 site 7 site 8 site 9 site 10 site 11 site 12 site 13 site 14 site 15 site 16 site 17 site 18 site 19 site 20 site 21 site 22 site 23

Table 1 .
The signifi cances of fi tted environmental variables and factors on the NMDS ordination.P values based on 999 permutations.