Inter-row Vegetation Increases Abundances of Ground- and Canopy-dwelling Arthropods in Mediterranean Vineyards

The decline of arthropod populations observed in many parts of the world is a major component of the sixth mass extinction with intensive agriculture being one of its main drivers. Biodiversity-friendly farming practices are taking centre stage in the recovery process. In vineyards, vegetation cover is commonly used for production purposes, to reduce soil compaction by machinery use and soil erosion. Here we examined the effects of vegetation cover and soil management on the abundance of ground- (spiders, beetles, Hemiptera and harvestmen) and canopy-dwelling (wild bees, green lacewings, beetles and Hemiptera) arthropods in three categories of vineyards: (i) vineyards with no vegetation, (ii) partially vegetated (every second inter-row is vegetated) and (iii) all inter-rows are vegetated. We recorded a general positive effect of a decrease in soil perturbation intensity and corresponding higher vegetation cover on arthropod abundance. Plant species richness was the most important vegetation parameter, with a positive effect on spiders, harvestmen, hemipterans and beetles (ground and canopy) abundances. Using a path analysis, we also highlighted the central role of inter-row vegetation management in trophic and non-trophic relationships between vegetation and arthropods, and between arthropod groups. Our results demonstrate the benets of a softer soil management preserving a diverse vegetation cover for the conservation of arthropods in Mediterranean vineyards.


Introduction
The decline of arthropod populations observed in many parts of the world over the past decades is a major component of the sixth mass extinction currently observed 1,2 . Although multi-causal, arthropod decline is strongly associated with the intensi cation of agricultural practices since the Green Revolution 3 . This is due to the homogenisation of landscapes leading to the loss of semi-natural habitats, the toxicity of pesticides and the mechanisation of farming practices, reducing diversity and abundance of plants on which arthropods directly depend 4,5 .
Given their role as decomposers 6,7 , herbivores 6 , predators 8 , pollinators 9,10 and prey of many vertebrate taxa, the general decline of arthropods threatens the functioning of ecosystems in agricultural but also in natural environmental contexts. Moreover, arthropods provide ecosystem services that directly affect production (pollination, pest regulation, recycling of organic matter, bioturbation) 11,12 . Their loss is therefore expected to have serious economic consequences in addition to ecological consequences 13 .
In order to stop the loss of arthropods biomass and diversity, and their related services, various agricultural practices have been developed to increase sustainability and conservation of biodiversity and its functions within agricultural ecosystems. The restoration of semi-natural habitats in agricultural areas is a key aspect of agro-ecology 14,15 . Sowing plants to cover the soil is increasingly promoted to improve predation 16 , pest regulation 17 and the refuge function for insect biodiversity 18,19 in addition to soil protection against erosion 20 . The presence of plants other than the cultivated species is therefore expected to provide multiple bene ts for perennial crops, such as the grapevine 21 .
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Diversifying the plant community in crop production by integrating annual and perennial plant species allows arthropods to access more diverse resources and habitats [22][23][24] . Several studies have demonstrated the positive effect of inter-row vegetation on grapevine arthropod communities 18, 25,26 . The establishment of vegetation cover is often accompanied by a general trend towards an increase in the abundance and diversity of arthropods. However, there is a high variability in the response to vegetation when looking at arthropod guilds separately. For example, Sáenz-Romo et al. (2019) reported a higher abundance of epigeic arthropods in grassy inter-rows but a lack of response of those collected in the grapevine canopy. Conversely, Eckert et al. (2019) found that epigeic arthropods were less numerous in densely vegetated inter-rows but more abundant in the adjacent canopy. These contrasting results demonstrate the need to improve the understanding of vegetation effects on arthropods, and to investigate more nely the relationships between the different components in response to the management of vegetation cover in vineyards.
In Mediterranean ecosystems known for their high level of biodiversity 27 , the integration in crop production of more sustainable agricultural practices may represent an important lever for biodiversity conservation. The objective of our study was to test the restoration of inter-row vegetation in Mediterranean vineyards as a measure for arthropod conservation. Using taxonomic and functional approaches, we studied the response of arthropods and the organisation of trophic relationships to three modes of inter-row management: mechanical soil management to remove vegetation, partially vegetated (one out of two inter-rows vegetated) and all inter-rows vegetated. We focused on two ecosystem components: the epigeic fauna (beetles, spiders, harvestmen, Hemiptera) and the ying fauna occurring in the grapevine canopy (beetles, wild bees, green lacewings and Hemiptera). We hypothesised that a dense and diversi ed plant cover increases the abundance of arthropods through trophic relationships between vegetation, phytophagous arthropods and predators both in the canopy and on the ground.

Study sites and design
This study was carried out in South-eastern France, between the southern slopes of the Luberon mountain and the Durance river (Appendix 1). The landscape is dominated by medium-sized vineyards.
This area is subject to a Mediterranean climate, with hot and dry summers followed by mild winters and rainfall mostly occurring in autumn and spring. Annual temperatures and total annual rainfall average 14.5 ± 0.1°C and 607 mm respectively (Pertuis meteorological station, 1991-2020). In this region, most vineyards are mechanically managed between rows to keep bare soil from mid-spring to the autumn rain seasons. Permanently vegetated vineyards are mowed or laid down before summer.
In 2018, we selected 27 vineyards where volunteer farmers (N = 9) applied on of the three following types of inter-row management: (i) periodic mechanical soil management to remove vegetation (0/2; N = 8), (ii) partially vegetated (every second inter-row is vegetated for a minimum of four years) (1/2; N = 10) and (iii) all inter-rows are permanently vegetated (2/2; N = 9). The selected vineyards had an average size of Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js 7,008 ± 746 m² representative of the study region and were dispersed over an area of 20 km by 6 km (Appendix 1). According to the French guidelines for organic viticulture no chemical was applied to 23 vineyards for at least three years. In the other four vineyards, farmers use chemical fertilisers and fungicides. As none of the 27 vineyards had suffered from pest attacks, no insecticides were applied. The only insect considered as a pest in the region is the Cicadellidae Scaphoideus titanus, which as a limited distribution. The inter-rows of all these vineyards were sown at least once in the last ve years with commercial grass-legume mixtures. However, with the exception of two vineyards, they were all dominated by spontaneous species. All vineyards were planted with either Syrah or Grenache grapevines for more than 15 years, the most common grape varieties in this region.

Characterisation of vegetation cover and arthropod communities
A pair of inter-rows separated by two inter-rows was selected in the centre of each vineyard (Fig. 1). The community of plants was studied in three 2×2 m quadrats in each selected inter-row from May 15th to June 2nd, resulting in a total of six quadrats per vineyard. The distance between quadrats within rows was 10 m. We identi ed all vascular plants species and estimated their respective ground cover (%) then combined to estimate the cover of different groups such as Fabaceae and Poaceae. Plant cover was estimated as the vertical projection of all above-ground organs such that the total vegetation cover may exceed 100%. The cover of owering nectariferous plants and total vegetation cover were thus estimated separately and do not correspond to the sum of individual species cover.
We assessed the abundance of ground-and canopy-dwelling arthropods using pitfall and sticky traps. One pitfall trap was placed in the middle of each vegetation plot in each vineyard (six pitfall traps per vineyard) (Fig. 1). The traps were 11 cm deep and 8 cm in diameter. They were buried up to the rim and lled with propylene glycol to a quarter of the depth. The epigeic arthropods were collected over two sampling periods (mid-May and mid-June), resulting in a total of 324 traps. Sticky traps consisted of yellow PVC sheets (15x20 cm) adhesive on both sides. Five sticky traps were hung to the top wire running alongside the rows of grapevines in each central row between the two inter-rows used for analysis of vegetation (N = 135). The lower edge of each trap was 60 cm above the ground and the distance between two traps was 10m. Sticky traps were set during the second pitfall traps sampling period, i.e. mid-June.
Arthropods were sorted in the laboratory and counted by taxonomic order. We also grouped arthropods according to their diet into: omnivorous, predatory, phytophagous, and detritivorous (Appendix 2). As beetles show a great diversity in their trophic regimes, we identi ed them down to the species level on the basis of taxonomic keys and our personal knowledge.

Data analysis
Generalized linear mixed-effects models (GLMMs) were computed to explore the effects of management type and vegetation characteristics on different response variables (arthropod abundance in pitfall and sticky traps) (package glmmTMB). Vineyards (and sampling period only for pitfall traps) were included Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js as random factors. When modelling the most parcimonious using the dredge function among models exploring the response to vegetation variables. Data analyses were performed using RStudio, version 1.2.5033.
We modelled separately the effects of inter-row management and vegetation. These two approaches are complementary and straightforward: models testing the effect of the inter-row management provide information on the effect of the soil management at the vineyard level whereas models testing the effect of vegetation variables provide information on biotic interactions between arthropods and vegetation at a local scale within vineyards. The following variables describing the vegetation structure and community we tted to the model: (i) vegetation cover, (ii) species richness, (iii) cover of nectariferous owering plants, (iv) cover of Poaceae species, (v) cover of Fabaceae, (vi) perennial to annual species ratio, and (vii) within vineyard beta diversity calculated using the Bray-Curtis dissimilarity index for all pairs of quadrats within vineyards. All vegetation variables (except vii) were recorded at the quadrat level.
We modelled the abundance of each arthropod group (beetles, spiders, hemipteran, harvestmen, bees, lacewings) in pitfall and sticky traps according to inter-row management and vegetation variables separately using GLMMs with a negative binomial distribution. We then speci cally modelled the abundance of beetles per diet and trap according to inter-row management and vegetation variables.
Inter-row management was further tested using the Tukey's post-hoc test, using the lsmeans package.
We also performed a path analyse (PA) to evaluate the effect of soil management on vegetation parameters and biotic relationships in both ground and canopy components (R package lavaan, Rosseel 2012). Path analysis is a speci c structural equation modelling (SEM) tool to disentangle relationships between several observed variables 29 . Based on an a priori knowledge, we developed a conceptual model based upon expected ecological interactions. We assessed the relationships between soil management, the seven vegetation variables (see above), canopy-dwelling arthropods (Hemiptera, wild bees, lacewings, predatory and phytophagous beetles) and ground-dwelling arthropods (spiders, harvestmen, Hemiptera, and plant feeders, detritivorous, omnivorous and predatory beetles). Soil management was coded from 1 for low soil disturbance (fully vegetated vineyards) to 3 for vineyards with periodic mechanical soil management. The full model was simpli ed by stepwise exclusion of non-signi cant variables until a minimum adequate model was reached (Appendix 3). The robustness of the nal model was determined by checking whether predicted and observed covariance matrices did not differ (χ2-squared tests, P > 0.05), whether root mean squared error of approximation index was low (RMSEA < 0.1) and whether comparative t index (CFI > 0.9) and Tucker-Lewis Index (TLI > 0.9) were high. The normal distribution of the residuals of each component model of the full path model was statistically tested. When residuals were not normally distributed, the dependent variables were log-transformed (i.e. abundances of Hemiptera and total predator in the canopy, and abundances of omnivorous and predatory grounddwelling beetles).
Arthropods abundance according to inter-row management and vegetation Overall abundance of arthropods was positively related to plant species richness although the different groups responded differently. In pitfall traps, beetles responded only to the cover of owering plants, with decreasing abundance as ower cover increased (Table 1). Spiders were signi cantly more abundant in the 1/2 and 2/2 vineyards than in the 0/2 ( Table 1; Fig. 2A). Spider abundance also increased with plant species richness. Similarly, Hemiptera were more abundant in 1/2 and 2/2 vineyards than in 0/2 ( Table 1; Fig. 2B). They also responded positively to plant cover ( Table 1). The abundance of harvestmen responded positively to plant species richness and negatively to the increase in both owering plant cover and Bray-Curtis dissimilarity index.
The arthropods sampled by the sticky traps in the canopy were more abundant in fully vegetated vineyards ( Table 2). This response was mostly driven by beetles and hemipterans ( Table 2; Fig. 3B). Both taxa responded also positively to plant species richness ( Table 2). Wild bee abundance responded positively to increasing owering plant cover and perennials/annuals ratio (Table 2).

Differential responses of arthropod diet groups
Overall arthropod diet groups In pitfall traps, the abundance of all phytophagous species was higher in the 1/2 and 2/2 vineyards than in the 0/2 (Table 1; Fig. 2A). Phytophagous species responded also positively to the percentage of plant cover. Similarly, on sticky traps, all canopy-dwelling phytophagous species were more abundant in the 2/2 vineyards than in the 0/2 and 1/2 ( Table 2; Fig. 2B). In pitfall traps, predators increased in abundance as plant species richness increased (Table 1).

Beetle diet groups
The abundance of ground-dwelling phytophagous beetles was negatively correlated with the cover of owering plants ( Table 1). The abundance of ground-dwelling predatory beetles increased with increasing plant cover and decreased with increasing Poaceae cover. Ground-dwelling detritivorus beetles were more abundant in vineyards with no vegetated inter-rows and showed a signi cant response to vegetation variables. The abundance of ground-dwelling omnivorous beetles only responded to Poaceae cover, with a negative relation (Table 1). Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js In the grapevine canopy, phytophagous beetles were more abundant in the 2/2 vineyards than in the 0/2 and 1/2 (  Fig. 2B). The abundance of all phytophagous, phytophagous beetles as well as all predatory beetles increased with plant species richness (Table 2).

Relationships between inter-row management, vegetation and arthropods
The tting parameters of the minimal adequate path analysis model (SEM) were very good (P-Value: 0.31; RMSEA: 0.05; CFI: 0.98; TLI: 0.97; Fig. 3). We identi ed a strong link between inter-row management and the vegetation parameters, and between inter-row management and arthropods in the canopy. Plant richness decreased with increasing inter-row management (regression coe cient β = 0.66, P <0.001) as did the percentage of plant cover (β = 0.66, P <0.001) and the cover of owering plants (β = 0.50, P = 0.003) (Fig. 3). Soil management also strongly in uenced phytophagous beetles and Hemiptera caught in the grapevine canopy and in to a lesser extent wildbees (Fig. 3). The increase in plant species richness was directly related to an increase in canopy phytophagous beetles but a decrease in wild bees that positively responded to ower cover.
The link between the intensity of soil management and ground-dwelling arthropods was rather indirect through its effects on vegetation, particularly species richness (Fig. 3). Only one relationship was established between the inter-row management and ground-dwelling arthropods, showing an increase in the abundance of detritivorous beetles with an increase in soil management intensity (β = 0.79, P <0.001). Hemiptera on the ground, detritivorous beetles and spiders were all positively in uenced by plant species richness. The three taxa of ground-dwelling predators (spiders, harvestmen and predatory beetles) were positively related to detritivorous beetles, and for spiders additionally to phytophagous beetles in the canopy (Fig. 3).

Discussion
While the restoration of plant cover in vineyards is increasingly used as a solution to stabilise the soil, prevent erosion and facilitate machinery use, it can also bene t the community of arthropods by offering a wider diversity of micro-habitats. Here we showed that the abundance of arthropods generally increased in vineyards with low intensity of soil management that favoured the development of inter-row vegetation. Plant species richness in particular was found to be positively correlated to the abundance of most groups of arthropods. Moreover, a decrease in inter-row management intensity not only affected each ecosystem component separately (vegetation, soil and canopy fauna) but also interactions between them.

Differential responses of arthropods to inter-row management and vegetation
Although most of the studied groups showed a positive link to vegetation, three main arthropod responses to inter-row management can be identi ed. Some arthropods responded linearly to inter-row management intensity, while other showed either a steep decline in periodic mechanically manage vineyards or a steep increase in fully vegetated vineyards.
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js In our study, the most abundant group was the order of Coleoptera. Interestingly, beetles responded only in the grapevine canopy with higher abundance in fully vegetated vineyards only. In the literature, the response of beetles to vegetation and management is also ambiguous. While some studies revealed a positive effect of inter-row vegetation on beetle abundance 18 , others did not nd such an effect 30 . As beetles are highly diverse regarding their diet and ecology, we classi ed them into four groups. The increased abundance of beetles occurring in the canopy of fully vegetated vineyards was driven by the dominance of families that feed on plants such as Mordellidae (393 individuals) and Buprestidae (176 individuals). Canopy-dwelling predatory beetles, that hunt directly on plants and can use plant resources such as pollen and/or nectar as supplementary food, also bene ted from increased plant cover. Carabidae and Staphylinidae, the most abundant families in our samples, are known to bene t from the vegetation on the ground 31 . These generalist predators are favoured by the diversi cation of the agroecosystem and the complexity of resources provided by the vegetation 22 . By their direct and indirect relationships with the vegetation, these species are therefore more likely to bene t from inter-row vegetation 32 .
On the opposite, detritivorous beetles showed a preference for vineyards with intensive soil management, as observed in previous studies 26 . Species of the Anthicidae family largely dominated this group. Several non-exclusive hypotheses may explain this response. First, such beetles can be locally abundant and may bene t from habitat disturbance, probably by feeding on vegetation litter that is provided by regular soil management in the vineyards. They may be also favoured by a lower level of competition and/or predation regularly observed in disturbed habitats 33,34 . Finally, this increase in Anthicidae abundance might also be the result of a higher probability of capture in pitfall traps. It is known that arthropod abundance data can be in uenced by habitat structure, with trappability increasing as habitats become more open 35 . The reduction in habitat complexity in the absence of vegetation may facilitate their movements on the ground, whereas Anthicidae were commonly observed on plants in vineyards with ground cover (pers. obs.).
Hemipterans, largely dominated by leaf hoppers, responded positively to a decrease in soil management both on the ground and in the canopy. Since leaf hoppers are phytophagous, this result indicated the occurrence of a bottom-up effect with primary consumers favoured by an increase of primary producers 36 as observed in previous studies 25,30,37 . The increase of phytophagous species in vegetated vineyards may be problematic if it includes pest insects. So far, neither in our vineyards nor in vineyards of other studies 38-40 such an increase in insect pest infestation in response to increasing inter-row cover has been detected. Improved biological control of potential pests by an increase in the number of predators such as spiders favoured by vegetation cover may compensate for potentially positive effects on pest insects 17,41 .
We recorded a clear positive response of spider abundances to a decrease in soil management intensity and an increase in plant species richness. This latter relationship was also observed for harvestmen, whereas harvestmen remained largely unaffected by inter-row management as previously observed 40 .
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Similar to Anthicidae, harvestmen use to move over vegetation, which may reduce their trappability as habitats become more complex. Ground-dwelling predators may use rich inter-row vegetation as a source of prey 42 . This hypothesis is supported by the indirect relationships that we found between spiders and plant richness via an increase in both detritivorous beetles and phytophagous canopy beetle abundances, groups known to be prey for spiders 43 . Vegetation cover may also offer better microclimate (temperature and/or humidity) facilitating the development and/or reproduction of predators 44 .
Inter-row vegetation provides resources for pollinators, particularly for wild bee species. Several studies have already shown a positive impact of reduced soil management on pollinators 45 . Contrary to Kratschmer et al. (2019), the abundance of wild bee species did not respond to the increase in total plant cover but rather to the increase in owering plants. A similar pattern has been recently observed between parasitoids and plant cover providing nectar 46 . The relationship con rms the importance of the plant species composition alongside with plant species richness and functional diversity to increase food availability for arthropods. In contrast, lacewings, that have the potential to improve pest regulation, did not show any response to inter-row management or vegetation variables, as already observed in previous publications 47,48 . Lacewings change their dietary habits during their life cycle, from generalist predators as larvae to pollinators as adults. Here, we only captured adults, which limits our understanding of the lacewing response to soil management and inter-row vegetation.

The multi-component effect of inter-row management
The path analysis clearly illustrated the key role of inter-row management intensity in vineyard ecosystems and its major effects on their functioning. Inter-row management and vegetation cover do not only have direct effects on ground-dwelling primary consumers such as detritivorous beetles and phytophagous hemipterans but also on ground-dwelling predators, such as spiders, and on almost all studied canopy-dwelling arthropods. Among the analysed vegetation parameters, plant species richness was by far the most important one, vegetation cover being secondary. Plant richness affected groundand canopy-dwelling arthropods, from primary consumers to predators, while ower cover only in uenced wild bee species, and plant cover only phytophagous beetles. This result suggests that plant species richness, more than vegetation cover, increases the diversity of resources for primary consumers and by makes the habitat more complex, thus offering more ecological niches for arthropods 44 .
This analysis also illustrated the complexity of these agro-ecosystems, with numerous links between different arthropod groups of the grapevine canopy and in the inter-row. Increases in both primary consumers on the ground and in the canopy in uenced spider populations in the inter-row. Similarly, predatory beetles in the canopy responded positively to the increase in ground-dwelling hemipterans. These relationships con rm the movement of individuals from the grapevine canopy to the inter-row and vice versa, an important spill-over effect of inter-row vegetation on the functioning of grapevine agroecosystems.

Ecosystem restoration in Mediterranean vineyards
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js Currently, winegrowers are reluctant to maintain or plant vegetation between rows because of competition for water, a particularly important stress factor in water-limited regions such as the Mediterranean 49,50 . However, several studies have demonstrated that inter-row vegetation also provides important ecosystem services 21,51 . In a recent study, we showed that inter-row vegetation improves predation by arthropods 16 . Here, we further highlighted a strong and positive effect of inter-row vegetation on arthropod communities. This positive effect on the abundance of several groups was not limited to ground-dwelling arthropods but was also detected in the grapevine canopy and in trophic relationships between soil surface and canopy. In particular, plant species richness (spiders, harvestmen and canopy beetles) and ower cover (pollinators) had a positive effect on arthropods indicating that highly diverse, ower-rich mixtures are most bene cial to arthropods whereas grass-dominated inter-row vegetation had predominantly negative effects. In a context of growing interest in sustainable and environmentally sound crop management, we showed that the use of plant cover is a key tool to restore arthropod communities in vineyards.

Declaration of competing interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper. function for vegetation variables and Tukey posthoc test for inter-row management. 0/2periodic mechanical soil management to remove vegetation (N = 8); 1/2 partially vegetated (every second interrow is vegetated for a minimum of four years) (N = 10); and 2/2 all inter-rows are permanently vegetated (N = 9). Values in bold indicate signi cant effects, codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1. Table 2. Effect of inter-row management and vegetation parameters on arthropod and on diet group abundance on sticky traps. Estimates (± SE) of the nal models resulting from selection by the "dredge" function for vegetation variables and Tukey posthoc tests for inter-row management. 0/2 periodic mechanical soil management to remove vegetation (N = 8); 1/2 partially vegetated (every second interrow is vegetated for a minimum of four years) (N = 10); and 2/2 all inter-rows are permanently vegetated (N = 9). Values in bold indicate signi cant effects, codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1. As the variable "predatory arthropods" was only represented by beetles, it was not included in the table.    signi cant co-variation between variables. Arrow widths are proportional to standardised path coe cients that are shown next to the arrows and its signi cance is denoted as ***P < 0.001; **P < 0.01; Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js