Contrasting E ff ects of Tillage and Landscape Structure on Spiders and Springtails in Vineyards

Interactions between predatory species and their potential prey are little studied in vineyards, especially considering the surrounding landscape structure. We examined the effects of soil tillage intensities in vineyard inter-rows on the activity density and diversity of spiders (Araneae) and springtails (Collembola), their potential preys, and assessed whether these effects are altered by non-crop elements in the surrounding landscape. We collected data in 16 vineyards in Austria; eight were periodically mechanically disturbed (PMD), eight had permanent green cover (PGC). The study vineyards were embedded in landscapes ranging from structurally simple to complex. Both, spiders and springtails were collected with pitfall traps. Data analyses using generalized linear mixed models (GLMM) showed different effects of soil tillage intensities on spiders and springtails and an interaction with semi-natural elements (SNEs) in the surrounding landscape. Activities of springtails were higher under PMD than under PGC while spider activity density remained unaffected. Spider family Shannon diversity was lower under PMD than under PGC, while springtail species Shannon diversity was unaffected by tillage. Under PMD, spider activity and family diversity decreased with increasing SNEs in the surroundings indicating spider emigration away from vineyards. Under PGC, spider activity density increased with increasing SNE proportions in the surroundings when springtail activity density was high. Our findings suggest that recommendations on sustainable vineyard management should include both site and landscape factors.


Introduction
Viticulture is a very old sector of agriculture, dating back to about 6000 BC [1].Modern conventional vineyards are usually intensively managed with frequent use of pesticides and tillage during the vegetation period [2].Inter-row soil management practices are imposed to control weeds and can include intensive tillage (bare soil), alternate tillage where every second inter-row is tilled, or no tillage with permanent green cover, or by the use of herbicides to control weeds [3].Inter-row soil management in vineyards has been shown to affect aboveground [4] and belowground [5,6] biodiversity and associated ecosystem services [7].However very little is known about influences on organisms comprising different trophic levels that might affect predator-prey interactions.
Spiders (Araneae) are all predators comprising the largest order of Arachnids with around 46,800 species worldwide [8].Their habitat requirements are defined by several abiotic (temperature, humidity, wind, light intensity) and biotic (type of vegetation, food supply, intra-and interspecific competition, predators) factors [9,10].Spiders live in different vegetation layers and are often divided in two functional groups: Sedentary web spiders and non-web builders.Spider abundances can vary considerably in different ecosystems with mean abundances in temperate climate in grassland with 92 ind.m −2 , forests with 53 ind.m −2 , and arable fields of 21 ind.m −2 [11].Spiders are important generalist predators in many ecosystems [9,12,13], however their abundance and diversity is commonly decreased in managed agroecosystems [10,14,15].While spider communities, their feeding behaviors and possible implications for biological control have frequently been investigated in arable agroecosystems [16,17], far less is known from vineyards [13, [18][19][20][21].
Springtails (Collembola) are among the most widespread terrestrial microarthropods with about 8800 species worldwide representing an important component of soil fauna in many terrestrial ecosystems [22].Springtail species diversity is usually affected by land use [15,23].Springtails can be grouped into soil dwelling (eu-and hemiedaphics) and surface dwelling species (epedaphics) [24].As members of the decomposer community they are feeding on fungi, bacteria and dead organic matter thereby mobilizing nutrients [25].Abundance of springtails can vary considerably between different ecosystems [11,26], ranging from abundances in grassland of 29,545 ind.m −2 , forests of 1826 ind.m −2 to arable fields of 6056 ind.m −2 [11].Additionally, land management intensity including tillage, residue management or fertilization [27,28], pesticide use [29].Springtails are also an important source of prey to spiders and comprise up to 72% of their diets [16,[30][31][32].Furthermore, springtails provide essential nutrients and amino acids in the diet of spiders [33,34].
Few studies have investigated the effect of management practices in vineyards on spiders [18,20,35,36] or springtails [5].Tillage of vineyard inter-rows commonly has negative effects on the abundance of soil invertebrates, although rapid recovery of specific invertebrate groups (e.g., earthworms) has been shown [6].Studies conducted in arable and viticultural landscapes that assessed combined effects of management and the surrounding landscape on spiders [37][38][39][40] or springtails [5] found that complex landscapes contain a higher biodiversity than simpler structured landscapes [41].However, spiders appeared to respond differently to landscape structure: Ambush and habitat specific spiders are more abundant in diverse landscapes whereas web builders are more abundant in simple landscapes [38].Species richness of spiders was significantly enhanced in wheat fields with more non-crop habitats in the surroundings [17,42,43], although opposite responses have also been reported [44].Also, springtail populations showed a response to land-use intensity and the surrounding landscape structures [5,45,46].However, to the best of our knowledge no study investigated both spiders and springtails in vineyards in response to the surrounding landscape.
The aims of the current study were (i) to assess the response of spider and springtail activity and diversity in vineyards to periodically mechanically-disturbed (PMD) versus permanently green covered (PGC) inter-rows, (ii) to analyze whether the proportion of semi-natural habitats in the surrounding landscape and the presence of springtails alter responses of spiders and (iii) to assess spider family assemblages and their relation with local management and potential springtail prey.We hypothesized that permanent green cover will benefit both spiders and springtails because of less disturbance.A higher proportion of non-crop habitats in the surroundings is expected to increase the diversity and activity of spiders and springtails in vineyards due to migration into vineyards.

Site Description and Study Design
The study sites were located in the two neighboring wine regions Carnuntum (48 The study regions are strongly shaped by a Pannonian, continental-like climate with hot, dry summers and cold winters.The most frequent soil types are Chernozems and Cambisols with medium to high usable field capacity, neutral pH-values and high nutrient contents [48]. Within the regions, 16 vineyards were selected for the current investigations (Figure 1).In these regions vineyards are small (0.4-1 ha) and typically surrounded by landscapes with various degrees of heterogeneity.Landscape heterogeneity was calculated as the proportion of semi-natural habitats (hedges, tree rows, grass stripes, natural grassland, pasture, fallow, heathland, wetland and woodlots) within a radius of 750 m centered in the plot.This area is within the activity ranges of both spiders [17,49] and springtails [46] in agricultural landscapes.In eight of the study vineyards, the inter-rows had a permanent green cover (PGC) whereas in eight vineyards every second inter-row was periodically mechanically disturbed (PMD).Both management regimes were distributed across both regions and practiced for at least five years.The inter-row cover crops were sown with seed mixtures, except for three vineyards with spontaneous vegetation.Inter-row vegetation was cut or mulched several times per season.Grapevines were cultivated in a vertical trellis system for at least 10 years with within-row grapevine distance of 0.75-1.3m and inter-row distance varying between 2.15-3 m.On each site one inter-row sampling area of 2 × 40 m in the middle of the vineyard was selected; this area had a distance of at least 5 m to the vineyard edge.Furthermore, in PMD sites samples were taken in the current undisturbed inter-rows that had been cultivated in the year before and thus was vegetated.

Spider and Springtail Sampling
To collect surface active springtails and spiders, six pitfall traps (diameter 17 mm, depth 60 mm) were installed on 21 April 2015 at each vineyard site.Each trap was filled with ethylene glycol and a drop of odorless detergent.They were inserted along a transect in the middle of inter-rows with a distance of 6 m between each other.The focus was to assess spring communities of spiders and springtails to check the pest control potential for the first generation of grape berry moths (Eupoecilia ambiguella Hübner), an important pest insect in the study region.Springtails can act as alternative prey for generalist predators such as spiders and thus sustaining the population of natural enemies within the fields [50,51].After 13 days the traps were removed.Determination of spiders was only performed at family level because of many juvenile individuals in the traps [50,51].Spiders were classified according to their feeding behavior into non-web or web building spiders.Hunters were summed up under the collective name of non-web building spiders including Lycosidae, Gnaphosidae, Pisauridae and Dysderidae.Web builders were comprised of families Linyphiidae, Theridiidae and Hahniidae [52].Springtails were sorted and determined to species level [53][54][55][56][57][58][59][60][61].The number of trapped springtails and spiders during the sampling is referred to as activity density.
All land owners gave their permission to sample spiders and springtails on their private properties.No protected species/families were collected.The study complied with all relevant regulations.

Landscape Parameters
Field mapping was carried out in July 2015.As a reference we used the official Austrian land utilization map from 2012 (Integriertes Verwaltungs-und Kontrollsystem, INVEKOS).Landscape elements within a 750 m-radius from the center of the vineyards were classified into habitat types according to CORINE Land Cover and EUNIS Habitat Classification [62,63].Of these habitat types the proportion of semi-natural elements (SNE) such as hedges, tree rows, grass strips, natural grassland, pasture, fallow, heathland, wetland and woodlots were calculated because of their importance for spiders and springtails.Mapping and analysis was carried out using the programs ArcGis 10.2.195, QGis 2.8.1, FRAGSTATS 4.296 and CHLOE201297 [64].

Statistical Analysis
To account for significant differences between the two management regimes for springtails and spider activity density and diversity we used Wilcoxon signed-rank tests, a non-parametric statistical hypothesis test used to compare dependent samples from populations with similar distributions.
To account for the effects that soil tillage, landscape heterogeneity as well as alternative prey in the form of the surface-dwelling springtail activity density could have on the spider activity density and family Shannon diversity index, we used Generalized linear mixed models (GLMM) with Poisson error distributions.We built two sets of 18 different models that we tested for the two response variables modelled in this study, spider activity density and family diversity, covering all the possible combinations of the explanatory variables mentioned above with their interactions.Samples of the individual traps were not pooled but treated separately to be able to account for vineyard as a random factor.This way we could include the variability within vineyards and focus the explanatory power on the response variables described above.To account for large eigenvalues, we rescaled the springtail variables by dividing the average of the total springtail activity density by the standard deviation and subtracted it by the total springtail activity density.
Model selection with Akaike information criteria corrected for small sample size (AICs) was used as an alternative to traditional hypothesis testing [65].AIC is an estimator of the relative quality of statistical models for a given set of data.Given a collection of models for the data, AIC estimates the quality of each model, relative to each of the other models.Models with a difference in AICc > 2 indicate that the worse model has virtually no support and can be omitted from further consideration.Furthermore, models with AIC c less than two units above model with the lowest AIC were also considered and were averaged.Selected models were plotted and used for additional interpretation.When relevant, the size of the effect of springtail activity density was illustrated in lines with the values low (5%), moderate (50%) and high (95% of total springtail activity density).Vineyard was used as a random factor in the models.We calculated the R 2 for the best models to account for the variability explained by mixed models with Gaussian or Poisson error distributions [66].Though this approach cannot be applied to models with a negative binomial error distribution, it allows for two components of R 2 to be calculated: (i) A marginal R 2 (R 2 m ) that only includes the variance explained by fixed effects; and (ii) a conditional R 2 (R 2 c ) that represents the variance explained by both fixed and random effects.In case of two or more best fitting models, R 2 m and R 2 c were calculated for the more complex model.For models without random effects we calculated the adjusted R 2 .
In addition, Nonmetric multidimensional scaling (NMDS) was used to investigate the overall pattern of families' composition on the spider assemblages, and group families that could respond similarly to landscape heterogeneity, management system and resource availability.Data were square-root transformed and then submitted to Wisconsin double standardization [67].We used the Bray-Curtis dissimilarity distance to compute the resemblance matrix among sites.To explore associations between particular spider families or assemblages and springtails, we generated smooth surfaces for the springtail activity densities on the NMDS ordination diagram.Smooth surfaces are the result of fitting thin plate splines using general additive models and interpolating the fitted values on the NMDS plot [68].
We found a total of 6,244 surface-dwelling springtails comprising 34 species (Table A2).Across treatments the most abundant surface-dwelling springtail species was Isotoma viridis (33%, 2061 ind.), followed by Heteromurus major (22%, 1356 ind.), Lepidocyrtus cyaneus (11%, 668 ind.), Orchesella cincta (8%, 501 ind.), Entomobrya marginata (5%, 298 ind.) and E. olivacea and (5%, 284 ind.).The mean density was 70.04 ± 6.06 ind.per trap and sampling period.Springtail activity density was significantly higher under PMD than under PGC (p = 0.038, Figure 2A) whereas their Shannon species diversity was unaffected (p = 0.933; Figure 2B).Species richness was similar between the two tillage systems (p = 0.956).Results from the GLMM indicated five best models to explain influence factors on spider activity density which all included effects of the surrounding SNEs, management and springtail activity density (R 2 m = 0.41, Table 1, Figure 3A,B).Under PGC, the overall spider activity density remained relatively unchanged and was not affected by SNE when the springtail density was low or moderate.However, when the activity density of springtails was high, spider activity also increased with increasing SNE.For PMD the spider activity density decreased with an increasing proportion of SNE in the surroundings.For spider family diversity we found four best fitting models containing effects of surrounding SNEs and management, one model also contained the variable springtail activity density (Table 1, Figure 3C,D).Under PMD, the spider diversity decreased with increasing SNEs in the surroundings.Under PGC, the surroundings showed little effect on spider diversity in vineyards.
The relationship between spiders and springtails was analyzed by Non-metric multidimensional scaling (NMDS; Figure 4).It was shown that a higher activity density of springtails led to a higher activity density of Lycosidae spiders (LYC).In contrast, the families Pisauridae, Hahniidae and Salticidae displayed a relationship with lower springtail activity densities.The rest of the spider families were related with medium springtail activity densities.

Discussion
We found contrasting responses of spider activity density and family diversity to interrow tillage: Spider activity density was unaffected by tillage, but family diversity lower under PMD.Springtail activity density was higher under PMD while springtail diversity remained unaffected by tillage.This can be explained by more surface roughness and better shelter for both taxa.Likewise, PGC vineyards tended to have a more compacted topsoil providing less favorable habitats for springtails [5].Contrarily, others have shown that periodic soil tillage can lead to prey diversification in soil which could also enhance spider diversity [72,73].
In line with our expectations spider family diversity was higher when less soil disturbance occurred under PGC.We attribute this result to the smoother and more compacted soil surface enabling a higher spider family diversity to establish.Indeed, the spider families that benefited under PGC where non-web builders of the families Lycosidae, Gnaphosidae, Pisauridae and Dysderidae known to be sedentary waiting for passing prey [74].Whether diversity of prey other than springtails was also higher under PGC, we did not investigate in the current study.In contrast to our findings, others have shown that a higher disturbance level in vineyards (mulching vs. mowing) actually favored spiders [20].In contrast, in salt grasslands no effect on spider diversity was seed between grazed, i.e., disturbed, compared to ungrazed grassland [75].Hence, it is likely that more factors than only disturbance influence the diversity of spider communities.
The prevalence of Lycosid spiders in our study is in contrast to others that found Linyphiids dominating in vineyards [76,77].This might also be due to the sampling method we used.While pitfall traps are commonly used to study spiders in ecosystems they underestimate foliage-dwelling spiders [37,78,79].Another reason for the small number of Linyphiid spider findings may also be the sampling time.We wanted to assess spiders as potential predators for early instars of the Grape berry moths, therefore, our sampling took place in spring.At this time juvenile Linyphiid spiders start to recolonize surrounding landscapes by frequently ballooning after using vineyard inter-rows as overwintering sites or refuge habitats [17,80].In our traps, most Lycosid spiders were adults or too large to use ballooning because this ability is restricted to small-sized instars which leads to the assumption, that the closer surrounding is supposed to be of greater importance for them [17].
Overall, our data suggest that spiders in vineyards were less closely linked to SNE than spiders in arable fields where a major impact of SNEs on spider activity density, diversity and species richness was found [37,43].Perhaps, spider activity densities in perennial systems like vineyards are less dependent on SNE because vineyards provide food and shelter and vegetated inter-rows act as quasi non-crop refuges [81].However, our findings also indicate that spiders are affected by an interaction between vineyard inter-row management, springtails and the surrounding SNEs.Under PMD, both spider diversity and activity decreased with increased SNE, while under PGC, spider activity increased with increased SNE especially when a high activity density of springtails was found within the vineyards.Although, we cannot prove whether an increasing spider activity density responded to an increasing springtail activity density, others have shown that springtails are indeed an important source of prey for spiders [30,82,83].NMDS also indicated that mainly spiders of the families Lycosidae, Gnaposidae and Tomisidae benefited from higher springtail activity densities.Linyphiid spiders have also been shown to locate their webs in areas with high springtail densities which let us conclude that an increase in the number of spiders can be associated with an increase of surface-dwelling springtail density [84].However, it is important to note that also different habitat properties differentially affect activity density of spiders and Collembola [85], and more detailed studies are needed to better understand predator-prey relationships between spiders and springtails.
The decrease of spider activity density and diversity with increasing SNE under PMD suggest that disturbances make spiders leave the vineyard to find shelter in SNEs in the surrounding landscape.Additionally, vineyards are disturbed by various other management activities such as pruning, foliage work or pesticide applications and therefore refuge habitats in the landscape are of great importance for spiders and other arthropods [4].

Conclusions
Taken collectively, our findings suggested that both spiders and springtails in vineyards might be altered by SNEs in the surrounding landscape.Obviously, our results also showed that vineyards are important habitats for spiders and springtails.Both activities of spiders and springtails seem to benefit from alternating tillage activities, however in order to increase spider diversity vineyards with permanent vegetation seem advisable.To what extent a higher spider activity density and diversity contributes to biological control of viticultural pest species would need to be investigated in more detailed experiments.

Appendix A
• 4 N, 16 • 47 E) and Neusiedlersee-Hügelland (47 • 54 N, 16 • 41 E) about 40 km south east of Vienna, Austria.The annual total precipitation for the region Bruckneudorf (which is located in the center of the study regions) is around 619 mm at an altitude of 167 m and an annual average temperature of 10.2 • C [47].

Figure 1 .
Figure 1.Location of study sites (circles with radius 750 m) in Eastern Austria in the viticultural areas Carnuntum (upper eight circles) and Neusiedlersee-Hügelland (bottom eight circles).Black circles denote vineyards with permanent green cover (PGC), black triangles those with periodically mechanical disturbance (PMD).

Figure 2 .
Figure 2. Springtail activity density (A) and Shannon species diversity (B) in vineyard inter-rows with permanent green cover (PGC; green) or periodically mechanical disturbance (PMD; red).Asterisk denotes significant difference between treatments.

Figure 3 .
Figure 3. Spider activity density (A,B) and spider Shannon family diversity (C,D) in vineyards in response to surrounding semi-natural elements and two different tillage systems (PMD = periodically mechanically disturbed; PGC = permanent green cover) with effect size of springtail activity density level (low = 5%, moderate = 50%, high = 95% of total springtail activity density).

Table 1 .
Listing of best fitting models (using AICc) for spider activity density and family diversity in vineyards.The best models (lowest AICc) with which the average will be formed, are indicated in bold.R 2 m (R 2 marginal), R 2 c (R 2 conditional) refer to the most complex model of the best models.Mgmt = Management intensity, Coll = Collembola activity density, SNE = Semi-natural element.

Table A1 .
Spider families collected with pitfall traps in vineyards with periodical mechanical disturbance (PMD) or permanent green cover (PGC).Means ± SD.

Table A2 .
Springtail species collected with pitfall traps in vineyards with periodical mechanical disturbance (PMD) or permanent green cover (PGC).Means ± SD.