Local management and landscape structure determine the assemblage patterns of spiders in vegetable fields

Both field- and landscape-scale factors can influence the predator communities of agricultural pests, but the relative importance and interactions between these scales are poorly understood. Focusing on spiders, an important taxon for providing biological control, we tested the influence of field- and landscape-scale factors on structuring the spider communities in a highly dynamic brassica agroecosystem. We found that local factors (pesticide-use and crop type) and forested landscape significantly influenced the abundance and species richness of spiders, whilst grassland patches significantly affected the spider species richness. Correlation results demonstrated that assemblage patterns of most spider families positively responded to the interplay between local factors and forest patches in the landscape. The spiders abundance was greatest in cauliflower crops surrounded with forest and grassland patches in landscape. Similarly, ordination analyses revealed that organic fields of cauliflower in forested landscapes had a strong positive association with the abundance and species richness of spiders. In contrast, insecticide and synthetic fertilizer-treated fields of Chinese cabbage in landscapes with little non-crop habitat reduced the abundance and species richness of spiders. Our results highlight the extent of interaction between local- and landscape-scale factors, help explain recently reported inconsistent effects of landscape factors on conservation biological control.


Scientific RepoRtS
| (2020) 10:15130 | https://doi.org/10.1038/s41598-020-71888-w www.nature.com/scientificreports/ on the presence of nearby source habitats that allow recolonization in crops after pesticide use and associated episodic extermination of natural enemies within crops 13,14 . Therefore, there is a need to investigate the relative importance of local factors and wider landscape structures together in shaping the natural enemies' community. Structurally complex landscapes exert positive effects on the species abundance and richness patterns of arthropods from medium to large scales, extending to several kilometers from a focal field 1,15 . However, landscape complexity has several facets and the relative importance of each is not completely understood. At a smaller scale, adjacent crops or non-crop vegetation have been shown to act as a refuge for arthropods during times of adverse conditions or disturbance within a focal crop 16,17 . Thus, the arthropod community structures of natural enemies and crop pests can be affected by the availability and type of adjacent habitats 16,18 . For example, Saqib et al. 17 demonstrated that certain forms of adjacent perennial crop and non-crop vegetation types benefited predatory spiders in brassica fields, most likely by serving as source habitats. On the other hand, Perović et al. 19 showed the importance of the diversity of land use types and the spatial arrangement (including connectivity) of land uses up to 3 km from the focal field affected pest and natural enemies. Globally, a recent analysis of an especially large data set found that the presence of natural vegetation (woodland, grassland and scrubland) in the landscape led to inconsistent responses by pests and their predators and unpredictable yield outcomes 20 . This inconsistency suggests that local management factors, such as agronomic practices and inherent properties of crop types may also be important indirectly driving arthropod assemblages and mediating the effects of the wider landscape. Understanding the interplay between these wider landscape scales and local management practices is becoming important in highly disturbed and ephemeral cropping systems such as Brassica vegetables that have short-duration crop cycles, frequent tillage and often intense agrochemical use.
Spiders are an important group of ecosystem service providers, acting as natural enemies of pests 21 . They attack diverse groups of prey including pests such as aphids 22 and lepidopteran larvae 23 . The effectiveness of the biological control of these generalist predators is strongly influenced by the structure and composition of their communities 24 . Previous studies have reported that landscape and local scale factors influenced the assemblages of diverse natural enemies in stable and less disturbed agroecosystems including grasslands 25 , pastures 26 , olive crop 27 , wheat fields 28 , cotton crop 19 and vineyards 29 . But to the best of our knowledge, the present study represents the first attempt to shed light on how spider assemblages are shaped in a highly disturbed and ephemeral ecosystem, featuring different local field management factors, such as chemical use and crop types, and differing landscape structure. We hypothesized that spider density and diversity would be determined by local factors but mediated by the surrounding landscape composition, including proportion of different land uses.
Effects of field scale variables on spiders. Species accumulation curves showed a generally greater richness in cauliflower than in Chinese cabbage, with only Chinese cabbage reaching the asymptote (Fig. 1a). The spider species accumulation rate was increased more steeply across all organic sites compared to conventional sites, and further sampling is inferred to reveal greater species richness of spiders in organic than conventional fields (Fig. 1b).
Kruskal-Wallis test for all alpha diversity measures, (including richness, Shannon and Simpson indices) displayed significant (χ 2 = 10.637, p = 0.001; χ 2 = 8.801, p = 0.003 and χ 2 = 6.760, p = 0.009, respectively) differences  Local and landscape interactions. Abundance and species richness of spider families were affected differently by local management and land use proportions in the landscape (Figs. 2, 3). Conventional fields with a higher proportion of forest in the surrounding landscape showed a positively significant relationship with the abundance and species richness of Pisauridae ( Fig. 2a,b). On the other hand, statistically significant negative relationship of conventional fields with a higher proportion of forest in the surrounding landscape were observed with the abundance of Lycosidae (Fig. 2a,b). Moreover, the higher proportion of orchards around organic fields showed a significantly positive relationship with the richness of Oxyopidae (Fig. 2b). Cauliflower fields surrounded with a higher proportion of forested land showed significantly negative correlation with the abundance and species richness of Lycosidae ( Fig. 3a,b). In contrast, however, Chinese cabbage fields with a higher proportion of forests in the landscape were dominated by Gnaphosidae in terms of abundance only (Fig. 3a).
CCA models illustrated the response of spider families assemblages in terms of both abundance and species richness to different explanatory variables (including local factors and proportion of different landuses in surrounding landscape). In CCA models only the "built-up" landscape element was found to be redundant (have VIF > 10), so it was removed from the final CCA models. The CCA models explained 32% of total variability in the assemblages of spider families abundance and 34% of total variability in the assemblages of species in spider families. The results showed that the first three, out of the total 8, constrained eigenvalues (including management practices, crop type and proportion of forests in the landscape) accumulatively explained 82% of total variability in the assemblages of spider families abundance and 91% in the assemblages of species structure in spiders taxa (Fig. S1). In contrast, only small fractions of variability in assemblage structure of abundance and species richness in spider families were explained by the proportions of other land-uses (such as cultivated, grassland, unused, water and orchards).

Scientific RepoRtS
| (2020) 10:15130 | https://doi.org/10.1038/s41598-020-71888-w www.nature.com/scientificreports/ respectively) in spider families. However, proportion of grassland in the landscape significantly explained the variations in only the assemblage of species in spider families (χ 2 = 0.030, F = 2.568, p = 0.021). The assemblage structure of both abundance and species richness in spider families was not significantly influenced by none of the other land uses in the landscape (including cultivated, unused, water, built-up and orchard) ( Table 1). The CCA ordination results indicated that among local scale factors the organic versus conventional fields and cauliflower versus Chinese cabbage were separated on the first and second CCA axes (Fig. 4a,b). However, the landscape scale factors, the forested land was clustered along the first CCA axis. Whilst all other landscape variables were separated by the second CCA axis (Fig. 4a,b). The ordination showed positive response of most spider families (except Lycosidae, Salticidae, Mysmenidae and Gnaphosidae) in terms of abundance and species richness with organic fields, cauliflower crops and proportion of forest area in landscape (Fig. 4a,b). The abundance of Salticidae, Mysmenidae and Gnaphosidae positively responded with the organic practices and proportion of grassland in the landscape (Fig. 4a). Similarly, the abundance of Lycosidae responded positively CUL" = cultivated, "FOR" = forest, "GRA" = grassland, "ORC" = orchard, "UNU" = unused, "WAT" = water) in the landscape based on correlation at levels of Brassica crop types (Chinese cabbage "CC" and cauliflower "CF"). The abundance and species richness of each spider taxa is correlated with each of the environmental variables. "*" is indicating the significant correlation. Table 1. Permutation test for Constrained Correspondence Analysis (CCA) of spider abundance and richness in Brassica crop types (cauliflower or Chinese cabbage) managed under different management practices (conventional or organic) in sites with varying proportions of different land-uses. The significance of constraint variables were tested by performing 999 permutations, "*", "**" and "***" is indicating the significant constraints. www.nature.com/scientificreports/ to Chinese cabbage crops, and proportions of cultivated land, water and orchards in the landscape (Fig. 4a). The abundance and species richness of Araneidae strongly associated with the conventional fields ( Fig. 4a,b).

Discussion
Local scale management practices such as crop identity and chemical use, had direct effects on the assemblage structure of spider communities in highly disturbed and ephemeral vegetable growing systems, both in terms of abundance and species composition. Likewise, at a landscape scale, we found significant effects of the proportions of non-crop habitats (including forests and orchards) in the adjacent landscape. Other studies of less disturbed and persistent agroecosystems 15,28 has also reported that assemblages of spider communities were significantly affected by both the farming systems and high percentage of non-crop habitats in the landscape. Studies have confirmed the persistent effects of local management practices on natural enemy diversity and their biological control potential 5 , and over wider landscape scale managing spatial distribution of both crop and non-crop habitats can be used as mean to reduce the pest population 12 . Although, purely based on landscape variables, the meta-analysis by Karp et al. 20 found inconsistent responses of pests and predators to the surrounding landscape composition. The authors, however, suggested that local management factors should be incorporated in landscape models to further develop the accuracy of pest control systems. The results of the present study provide an evidence for the value of that suggestion, by highlighting the importance of crop species and management practices, as local factors, as well as the presence of non-crop habitats in the surrounding landscape.
Our results showed the influence of different crop types interact with surrounding landscape structure to differently drive the assemblage patterns of each spider families. This may be due to the differences in the functioning of crops, how efficiently they provide resources (such as food, shelter and prey items), for every spider family at a local field scale 30,31 . Moreover, previous studies of woodlands 32,33 have reported that the density and diversity of spider assemblages are closely related to the structural composition and stability of the local habitat; because the enhancement in the structural complexity and stability of the habitats increased the density and diversity of predators by increasing their required resources (mainly food and shelter) 31,34 . The results of this study confirmed these findings; in a highly disturbed and ephemeral agroecosystem, the assemblage structure and composition of most spider families were positively responded with the cauliflower and presence of seminatural habitats in the surrounding landscape. This is likely to be due to the increased availability of heterogeneous micro-sites within the crop canopy of a complex plant structure, relatively more stable habitat because of long crop duration of cauliflower than Chinese cabbage for colonization/recolonization 35 . In addition to the heterogeneous micro-sites and stable habitat, the nearby seminatural habitat patches may provide more resources for spiders with diverse foraging behavior; such as web-builders (e.g., Pisauridae, Linyphiidae and Theridiidae), ambushers (Thomisidae), stalkers (Oxyopidae) and foliage runners (Clubionidae). In another study, Gangurde 36 , have reported the highest diversity and richness of predator species during the later growth stages of rice than the early growth stages. On the other hand, Chinese cabbage crop and low proportion of non-crop habitats in the landscape were correlated with the colonization of running spiders (Lycosidae). In this case, simple plant structure, small leaf surface area and less stable habitat because of the short Chinese cabbage crop duration for colonization/recolonization and www.nature.com/scientificreports/ that the low proportion of non-crop habitat patches in the surrounding landscape make the environment more suitable for the active hunters and likely to be unsuitable for the assemblages of diverse spider communities. The strong negative influence of Chinese crop growing systems on spider assemblages may be due to the lack of places for hiding and attaching their nets 33 , fewer prey items and high frequency of disturbance. The eminent dominance of Lycosids and lower number of web spiders in this highly dynamic brassica growing system may be because Lycosids have a high potential as early colonizers 35 with high dispersal ability, whilst frequent disturbances (due to harvesting and intensive farming practices) does not favor the web spiders. In another study by Öberg and Ekbom 37 sowing event was a detrimental factor for occurring of most of the spider families, but Lycosids as a group were not affected by the sowing events and were uniformly distributed in the field which may because of their high resilience to the environmental changes. In our study we reported the positive association of Lycosids with lesser non-crop habitat than those in with higher non-crop area. Similar patterns were also reported by Öberg 38 where Pardosa spider was more dominant in the structurally simpler landscape than in structurally complex landscape. Altogether, these results highlighted the importance of crop species, crop duration and availability of non-crop habitats in the surrounding landscape to support the diverse spider assemblages at local field scale, which, in turn, can mediate the ecosystem service of pest suppression 24 .
In line with previous studies 10, 15 of less disturbed and persistent agroecosystems; results of our study of highly disturbed and ephemeral agroecosystem showed that local management practices interact differently with different land use in the landscape to determine the assemblage patterns of each spider family. The different interactions of local management practices with different land use may because of differences of local habitat quality 39 , the availability of prey items 15 and frequency of disturbances between organic and conventional management practices. The results of this study indicated that the presence of non-crop habitats, especially woodlands and orchards, adjacent to the organic fields positively supported the diverse and abundant assemblages of spiders. Also, the management practices applied in organic fields promoting denser and more diverse vegetation 40 are likely to enhance the availability and diversity of resources, which in turn, increase the density and species composition of spiders 41 . A recent study of vineyards by Muneret et al. 29 highlighted that the deployment of organic practices at landscape scale significantly enhances the density and diversity of natural enemies, reduces the pest infestation and increases the productivity of the crop. On the other hand, synthetic pesticides used to causes rapid, detrimental effects on local fauna 5 and our results have shown the impoverished assemblages of both abundance and richness in most spider families under conventional management practices surrounded with other land uses than non-crop habitat in the landscape, which may because the non-crop patches may act as nearby refuge or donor habitat for spiders during the period of disturbances 19 . However, conventional fields with lower proportion of non-crop habitat in the surrounding landscape were mainly dominated by the active hunters (Lycosidae), orb-weavers (Araneidae and Tetragnathidae). As Tetragnathidae and Araneidae spiders are commonly known as leaf curling and subterranean spiders, so their living characteristics may help them to avoid the high level of disturbance events by seeking shelter and nesting in the curled leaves and subterranean habitats. Similarly, Lycosids are active hunters and are also known as cursorial spiders (ground-runners), which are capable to quickly seek shelter during the disturbance regimes and actively forage over a large area in conventionally managed fields. The conventional fields surrounded with high proportion of forests also had significant correlation with another large and robust spider, like the Lycosids, known as roaming hunters (Pisauridae).
The results of this study highlighted a clear and direct influence of local factors on assemblage patterns of spiders both in terms of abundance and species composition. However, we also found support for the equivalent effects of patches of non-crop habitats in the surrounding landscape which has direct effects and substantially interacts with local factors to drive the response of generalist predator in a highly dynamic brassica agroecosystem. Several other studies showed that the occurrence of non-crop habitats drive the assemblages of spiders in an agroecosystem, both individually 14 or by differently interacting with agricultural operation such as fertilizers and pesticides application 42 . This is because the patches of non-crop habitats in the surrounding landscape can supply a variety of alternative resources for a diverse group of predators including overwintering sites, shelter, refuge during the disturbance periods and food resources 31 , and as well as by favoring the dispersal of predators in the landscape matrix 43 .
In landscape studies, choosing the right spatial scale may influence the sensitivity of the method. A number of studies on how landscape elements at different spatial scales influence the spiders dispersal shape their assemblages supported that a relatively small study radius is sufficient to detect the influence of the surrounding landscape 44,45 . In our study, the effects of non-crop habitat on the abundance and species richness of most spider families at this smaller spatial scale were evident in this highly dynamic Brassica growing system. Correspondingly, the previous study of winter oilseed rape growing system by Drapela et al. 45 also indicated that species richness of spiders was positively related to non-crop areas and woody areas at the smallest spatial scale of (250-500 m radius). Moreover, the study by Schmidt et al. 44 in winter wheat indicated the contrasting response of different spider species (Lycosidae, Tetragnathidae and Thomisidae) at the different spatial scale ranging from 95 to 3,000 m, but the species richness of arable spiders were positively affected by the availability of heterogeneous landscape with the higher percentage of non-crop habitat at all spatial scales. Overall, the results of this study and previous studies indicated that different farmland spider species benefited differently from different land-use types in different growing systems. Still, the positive response of most spider families demonstrates the importance of non-crop habitats as a source of immigrants to the brassica fields at smaller spatial scales. In contrast to the previous studies, the sampling sites in this study belong to the smallholder farmers, and their agriculture operations (including high inputs of chemicals and destruction of natural habitat patches) pose severe threats to biodiversity 46 . Therefore, we need an integrated landscape approach with high opportunities for achieving long-term biodiversity conservation in smallholder farmlands at the smaller landscape scale. Thus, the results of this smaller landscape scale and previously studied large landscape scales clearly indicated that Scientific RepoRtS | (2020) 10:15130 | https://doi.org/10.1038/s41598-020-71888-w www.nature.com/scientificreports/ incorporation/conservation of non-crop habitats in the smallholder farms could provide buffer zones, food and shelter to the arable spiders.
During the last few decades, severe loss of biodiversity 47 has driven agroecological researchers to consider interventions for more sustainable crop production system in which ecosystem services are promoted 7,8,12,20,39 . Our analysis of spider assemblages in a highly dynamic vegetable production systems shows that local factors and the patches of non-crop habitats in the surrounding landscape individually or substantially interact to drive the assemblage patterns of this important taxon of natural enemies. The net result of these changes in assemblage patterns of spiders is likely to affect the strength of top-down pest suppression by these important ecosystem providers. This study suggested that the inconsistencies in responses of predators to the surrounding landscape composition, as highlighted by Karp et al. 20 , were potentially caused by local factors, including chemical use and type of crops. The population structure of generalist predators in this study provides a solid foundation for future studies to determine the relative extent to which local factors and different landscape variables influence the relative strength of their ecosystem function. Efforts in this direction will provide important strategic guidance to farmers and policymakers for how best to allocate effort to interventions at differing spatial scales in the pursuit of sustainable ecological intensification.

Methods
Study region and site selection. All experimental protocols were carried out in strict compliance with the State Key Laboratory of Ecological Pest Control for Fujian and Taiwan Crops and were approved by the ethics committee of Fujian Agriculture and Forestry University. The informed consent was obtained to assure that the farmers know every aspect of their participation. Field sites were located in Fujian Province, southeastern China, and data were collected during the main Brassica growing season (August-November) in 2017. The Fujian province in china is mostly mountainous and it has a subtropical, warm and humid, climate. Typical farming systems here are smallholders of highly dynamic Brassica crops that are globally common in other agricultural systems. The region's average temperature is 7-10 °C in mid-winter, while the average temperature is 23-33 °C at the peak of summer. Fujian has an annual precipitation of 1,400-2,000 mm. Fields were 1,300-2,000 m 2 in size which is typical for the region and were selected to represent different management practices and crop types with various proportions of land uses in the surrounding landscapes. The 23 fields were marked out using GPS (GPSMAP 60CSx-Garmin) and were at least 1 km apart to minimize the influence of spatial autocorrelation (Fig. 5a).
Fields were allocated into two groups, according to management practices: seventeen conventionally managed (i.e., synthetic pesticides or fertilizers applied) and six Chinese government-certified organics (no synthetic chemical inputs). The uneven number of fields in these two groups reflected their relative representation in this region (Table S1). Statistically adequate replicates of cauliflower (Brassica oleracea) and Chinese cabbage (Brassica rapa pekinensis) crop types were represented among both organic and conventional management. Direct sowing and seedling transplanting cropping methods were used to grow Chinese cabbage and cauliflower, respectively. Chinese cabbage is known as small head Brassica cultivar, so Chinese cabbage seeds directly sowed on the soil beds of approximately sized ~ 600-900 mm. However, for cauliflower, known as large head Brassica cultivar, seedlings were grown first, and then two weeks after sowing seedlings were transplanted to the field on both edges of soil beds (sized ~ 600-900 mm) with the plant to plant distance ~ 150 mm.
Sampling and identification of spiders. Spiders were collected from each site following a sampling method 4 given in the Sørensen et al. 48 and Mader et al. 49 due to the suitability and less chances of damaging the leafy Brassica crops. In each brassica field, spiders were collected individually by visually searching the soil surface and plants for one hour by two persons. A total of 35 sampling visits was performed over 23 sampling sites. Spiders were placed individually in vials, transferred to an ice box for transportation to the laboratory and thereafter kept at − 80 °C. All individuals were photographed and were assigned to a morphospecies. Later, molecular identification was used to determine the actual species of visually sorted morphospecies (see below).
Molecular identification of morphospecies. Genomic DNA was extracted from an excised leg from a representative specimen of each visually assigned morphospecies using Qiagen DNeasy blood and tissue kit (Qiagen Inc., USA) following the manufacturer's instructions. Prior to DNA extraction, each specimen was surface sterilized using 100% ethanol and then rinsed with purified water. Polymerase chain reaction (PCR) was performed to amplify the ~ 658 bp region of mitochondrial COI gene using well established universal primer pair for arthropods (e.g., Forward-LCO1490 5′-GGT CAA CAA ATC ATA AAG ATA TTG G-3′ and Reverse-Cheli-cerateReverse2 5′-GGA TGG CCA AAA AAT CAA AAT AAA TG-3′). PCR reaction contained the 1 × Hieff PCR Mas-