Biotic and abiotic factors drive multi-trophic interactions among spiders at different spatial scales in urban greenspaces

Abstract

Introduction

Urban ecosystems pose unique biotic and abiotic challenges for flora and fauna that influence the local species pools found in cities (Pickett et al. 2001; Mckinney 2002; Miles et al. 2019). For example, the amount of impervious surface such as buildings and roads as well as artificial light at night can disrupt patterns of arthropod dispersal and foraging (Schoeman 2016; Egerer et al. 2017; Johnson, Borowy, and Swan 2018; Philpott et al. 2019). Although these effects of urbanization have reduced arthropod abundance and diversity in some cities (Deichsel 2006; Horváth, Magura, and Tóthmérész 2012; Magura, Nagy, and Tóthmérész 2013), urban ecosystems can support species-rich communities (Dearborn and Kark 2010; Ives et al. 2016; Normandin et al. 2017). Whether a city can support diverse arthropod communities may depend on the amount, local management, and landscape context of its urban greenspaces (McKinney 2006; Kark et al. 2007; Aronson et al. 2014). Although many cities worldwide are growing, legacy cities have experienced prolonged population decline, which has resulted in an abundance of vacant land following the demolition of abandoned structures (Rieniets 2009; Haase 2013; Mallach and Brachman 2013). Vacant land is increasingly recognized as a valuable conservation habitat that can enhance arthropod biodiversity and desired ecosystem services (Gardiner, Burkman, and Prajzner 2013; Burkman and Gardiner 2014; Philpott et al. 2014; Delgado de la flor et al. 2017, 2020; Sivakoff, Prajzner, and Gardiner 2018; Perry et al. 2020; Riley, Herms, and Gardiner 2018). However, understanding the contexts in which cities support biodiversity and associated ecosystem services is required for urban greening efforts to achieve desired gains in conservation. A focus on drivers of community assembly in urban ecosystems will elucidate the processes that shape species diversity in cities.

Community assembly is determined by a series of abiotic and biotic filters that influence which species from the regional species pool can colonize and establish in habitats to form local communities (HilleRisLambers et al. 2012; Aronson et al. 2016; Perry et al. 2020). These abiotic and biotic filters operate at a range of spatiotemporal scales. Successful colonization is influenced by immigration-emigration dynamics as well as chance events (HilleRisLambers et al. 2012; Perry et al. 2020). For example, dispersal ability of species will determine the subset of the regional species pool able to colonize habitat patches (Fournier, Frey, and Moretti 2020). These dispersal dynamics can be further influenced by the composition, configuration, and connectivity of habitat in the surrounding landscape (Delgado de la flor et al. 2020; Perry et al. 2020). Once a habitat is colonized, successful establishment is influenced by local niche processes related to abiotic environmental characteristics and biotic interactions among species. Within a habitat patch, the environment and species interactions determine long-term coexistence of the community through stabilizing (niche differences) and equalizing (fitness differences) mechanisms, which influence the ability of a species to invade or establish (Chesson 2000; HilleRisLambers et al. 2012). For example, species that are poorly adapted to an environment will be unable to establish under the conditions or excluded by a stronger competitor with higher fitness. Conversely, coexistence among species could occur when: (1) small niche differences exist among species that are similarly adapted to an environment (i.e. have similar fitness); or (2) large niche differences exist among species that also have large fitness differences (Chesson 2000). The cumulative effects of these abiotic and biotic filters determine which species from the regional pool comprise the local species pool and which of those species can survive and reproduce within various habitats.

Urban ecosystems are unique in that these systems have novel anthropogenic filters that shape processes of community assembly in cities in addition to typical abiotic and biotic filters (Aronson et al. 2016). These novel anthropogenic filters are created by the past and present values, preferences, and activities of humans living and working within cities (Swan et al. 2011). Aronson et al. (2016) developed a conceptual model that outlines several hierarchical anthropogenic filters that are hypothesized to influence processes of community assembly in urban ecosystems. The urban species pool consists of species from the regional pool that are able to colonize and establish within the city, but this pool also may contain species introduced intentionally or unintentionally by humans, forming novel communities (Aronson et al. 2016, Fournier, Frey, and Moretti 2020). From the urban species pool, filters associated with urban form and history, socioeconomic and cultural factors, human facilitation, and species interactions are hypothesized to influence species occurrences and distributions across habitats within a city (Aronson et al. 2016). Urban form and history, such as city age and the composition, configuration, and connectivity of natural habitat in the surrounding urban landscape can influence processes of community assembly (Aronson et al. 2014; Parker et al. 2020; Perry et al. 2020). For example, greenspaces supported fewer bird species that were habitat specialists and ground nesters as urbanization increased in the surrounding landscape (Clergeau et al. 2006). Small ballooning spiders were able to colonize more isolated greenspace patches within an urban landscape compared to large ground-dwelling spiders with lower dispersal potential (Braaker et al. 2017, Delgado de la flor et al. 2020). At neighborhood scales, socioeconomic and cultural factors are important filters that drive population density, human preferences, the abundance of greenspace such as vacant land, vegetation diversity and structure, soil quality, and site history (Pickett et al. 2011; Aronson et al. 2016). At finer spatial scales such as an individual habitat patch, management of greenspace can further influence the occurrence and distribution of species within a city by acting as an environmental filter. For example, human management of greenspaces can vary along an intensity gradient from highly manicured lawns with frequent mowing and chemical inputs to more natural, heterogeneous habitats with diverse and structurally complex vegetation (Aronson et al. 2017). Important biotic filters related to species interactions such as competition, predation, and facilitation also occur at the local patch scale and can influence species occurrence and community composition (Weiher et al. 2011; Lessard et al. 2016). Urban ecosystems tend to support more generalist and nonnative species (McKinney and Lockwood 1999), which results in novel interactions with native species and communities. Understanding how anthropogenic filters influence processes of community assembly in urban ecosystems will inform conservation of ecologically valuable species.

Spiders have been extensively studied in urban environments due to their high abundance and ecological role as predators that contribute to pest suppression (Shochat et al. 2004; Magura, Horváth, and Tóthmérész 2010; Lowe et al. 2018). Emerging trends suggest urbanization does not strongly impact patterns of spider diversity, but changes in community structure and composition have been observed due to the replacement of habitat specialists by generalists that thrive in disturbed environments (Samu and Szinetár 2002; Fedoriak et al. 2012; Burkman and Gardiner 2015; Egerer et al. 2017). For example, generalist spiders such as species of Pardosa (Family Lycosidae) can dominate in habitats highly polluted with heavy metals, altering patterns of species composition (Jung et al. 2008). Studies investigating processes of spider community assembly in urban ecosystems have focused on the effects of landscape and local environmental filters (Alaruikka et al. 2002; Moorhead and Philpott 2013; Braaker et al. 2017; Meineke et al. 2017; Argañaraz, Rubio, and Gleiser 2018). For example, spiders responded to landscape composition and greenspace connectivity (Braaker et al. 2017; Delgado de la flor et al. 2020) as well as local factors such as prey availability (Harwood, Sunderland, and Symondson 2001; Lowe, Wilder, and Hochuli 2016) and structural complexity of vegetation (Greenstone 1984; Lowe et al. 2018; Gardiner et al. 2021). Moreover, landscape and local filters may differentially impact spider communities based on the functional traits of species, such as body size and hunting strategy. For example, small web-building spiders were more abundant in urban greenspaces that were managed via monthly mowing, while larger ground-hunting spiders were more abundant in urban meadows and pocket prairies that were less intensively managed via annual mowing (Delgado de la flor et al. 2020). At the local scale, spiders establish in habitats with greater prey availability (Wise 1993; Marshall, Walker, and Rypstra 2000; Harwood, Sunderland, and Symondson 2001; Lowe, Wilder, and Hochuli 2016), suggesting a direct relationship among spiders, their hunting strategy, and food resources. However, indirect relationships are less frequently investigated, but are central to multi-trophic interactions in food webs. Rather than directly responding to prey abundance, spiders may respond to local habitat characteristics (Greenstone 1984; Uetz 1991), as the structural complexity and productivity of vegetation enhance arthropod communities broadly (Langellotto and Denno 2004; Borer et al. 2012; Burkman and Gardiner 2014; Sarthou et al. 2014). Therefore, to understand processes of spider community assembly and inform urban conservation, a multi-trophic approach is required that examines the direct and indirect effects of landscape and local environmental filters simultaneously.

We investigated the relative importance and strength of landscape and local filters for influencing the structure of hunting and web-building spider communities within an urban ecosystem. Using a path analysis modeling approach (Fig. 1), we tested the direct and indirect effects of landscape composition and connectivity, soil texture and quality, prey abundance and breadth, and habitat management designs on spider communities in Cleveland, Ohio, a legacy city with large holdings of vacant land. We hypothesized that (Fig. 1): (1) greenspace patch isolation will negatively affect active hunting spiders which disperse by walking and web-building spiders which disperse by ballooning; (2) landscape diversity will favor hunting spiders that can actively disperse to exploit surrounding habitat patches; (3) web-building spiders and active hunting spiders will be positively associated with prey abundance and breadth; (4) higher plant biomass will directly positively influence spider communities by providing structural habitat complexity that has a higher diversity of refugia, more sites for web attachment, and protection from desiccation and predation; and (5) soil texture, soil contamination, and plant biomass will indirectly influence spiders via changes in prey availability.

Figure 1:

Hypothesized path analysis model describing the predicted effects of landscape and local environmental filters on the abundance of web-building and hunting spiders in an urban ecosystem. Solid arrows show predicted positive correlations and dashed arrows indicate predicted negative correlations.

Open in new tabDownload slide

Methods

Experimental design

This study was conducted in the city of Cleveland, Ohio, USA. Since the 1950s, Cleveland has experienced significant economic decline and prolonged population loss which has led to the abandonment and demolition of residential properties and creation of over 27 000 vacant lots throughout the city (Western Reserve Land Conservancy 2015). Current vacant lot management strategies employed by the City of Cleveland Land Bank include monthly mowing during the growing season. In 2014, the Cleveland Pocket Prairie Project, a city-wide manipulative field experiment, was established and included two habitat management treatments implemented within 16 vacant lots (former residential properties) replicated across eight residential neighborhoods. Each neighborhood contained one replicate of the two habitat management treatments. Treatments were (1) vacant lots seeded with a turf grass mixture and dominated by spontaneous vegetation mowed monthly from April to September to represent the current management strategy employed by the City of Cleveland (Fig. 2A), and (2) pocket prairies seeded with three native grass and 22 native forb species mowed annually in October to represent urban conservation efforts (Fig. 2B). Pocket prairies were seeded by Ohio Prairie Nursery (Hiram, Ohio, USA) on 3–12 November 2014 following removal of existing vegetation via two applications (28–30 May and 23–25 June 2014) of non-selective glyphosate-based herbicide. Prior to establishment of the treatments, all vacant lots were managed by the city via monthly mowing. All data collection occurred with a 7 × 15 m plot of 105 subplots (1 m² each) placed in the center of each site.

Figure 2:

Habitat management treatments established in 2014 across 16 vacant lot sites in Cleveland, Ohio: (A) Vacant Lots seeded with fescue grass mixture, mowed monthly to reflect existing management practices within the city, and (B) Pocket Prairies seeded with a mixture of three native grasses and 22 flowering plants, mowed annually in October. Photos were taken in 2017.

Open in new tabDownload slide

Arthropod sampling and identification

Landscape variables

The City of Cleveland Planning Commission provided 1 m resolution landcover data from their 2011 Urban Tree Canopy Cover Assessment. Landcover classes were grass/shrubs, bare soil, water, buildings, roads/railroads, other paved surfaces, tree canopy over vegetation/bare ground, tree canopy over buildings, tree canopy over roads/railroads, and tree canopy over other paved surfaces. Landscape diversity was calculated using Shannon’s Diversity Index using all landcover classes in FRAGSTATS v4.2 (McGarigal, Cushman, and Ene 2012). The landcover classes grass/shrubs and tree canopy over vegetation/bare ground were combined into Class Greenspace, and the class-level metric Euclidean Nearest Neighbor Distance (ENN; a measure of patch isolation) was calculated in FRAGSTATS v4.2 (McGarigal, Cushman, and Ene 2012). All landscape variables were calculated at 200 m radii surrounding each site.

Local environmental variables

Using the 105 subplots (1 m² each) placed in the center of each site, vegetation biomass (g/m²) was sampled during 17–19 July 2017 using the comparative yield method (Haydock and Shaw 1975) and the dry-weight-rank method (Mannetje and Haydock 1963). First, five 0.5 m² subplots were visually ranked from 1 to 5 reflecting the lowest and highest biomass in each site to establish the five-point standard yield scale. Then, 20 subplots, referred to as comparative yields, were compared to the five standard yields and visually ranked accordingly to estimate the total biomass per site (details in Delgado de la flor et al. 2020). The five standard yields were harvested and transported to the laboratory where they were dried at 75°C for 48 h and weighted (g). Lastly, the weights of the standard yields were used to create a linear equation for each site, and this equation was used to estimate biomass for each of the 20 comparative yield rankings. This resulted in 20 estimated biomass calculations per vacant lot site.

Three composited soil cores (2.5 cm diameter × 15 cm deep) were collected at each site to measure soil quality. Soil texture was analyzed using the micropipette method to measure particle size distribution (Miller and Miller 1987), and percentage sand, silt, and clay were determined. Concentrations of heavy metals measured were aluminum, antimony, arsenic, barium, cadmium, chromium, cobalt, copper, iron, lead, manganese, nickel, vanadium, and zinc. Site-level heavy metal contamination was determined by calculating the Contamination Factor Index (Hakanson 1980; Weissmannová and Pavlovský 2017) for each heavy metal at each site by comparing observed concentrations to background levels determined for the eastern US (US EPA 2007). Next, the pollution load index (Tomlinson et al. 1980; Weissmannová and Pavlovský 2017) was calculated for each site using the site-level Contamination Factors for each heavy metal.

Statistical analysis

All analyses were performed in R v3.6.2 programming language (RStudio Team 2016, R Core Team 2019). Path analysis was used to examine causal pathways among the abiotic and biotic variables outlined in the conceptual model (Fig. 1) to identify factors mediating spider composition and abundance in urban landscapes (Grace 2006). Variables included in the initial model tested the direct and indirect effects of landscape composition and connectivity (landscape diversity and greenspace patch isolation; each at 200 m radii), soil texture and quality (percentage of clay, sand, and silt, and pollution load index), species interactions (abundance and breadth of prey), and habitat management designs (plant biomass). All variables were log-transformed to meet assumptions of normality and linearity (Ullman 2007), and Pearson correlation coefficients were examined to ensure no multicollinearity (r² < 0.7) among variables (Grewal, Cote, and Baumgartner 2004). Percentage silt was highly positively correlated with percentage clay and sand, and therefore, removed from the analysis. The path analysis model was developed using the ‘lavaan’ package in R (Rosseel 2012).

To obtain the final model, we sequentially removed non-significant paths (P > 0.10) until an adequate model fit was obtained, and maximum likelihood methods were used for parameter estimation (Fan et al. 2016). Model performance was evaluated using the chi-square-based goodness-of-fit test, with P > 0.05 indicating a model structure consistent with the data (Grace 2006). We also used the comparative fit index (CFI) and root mean square error of approximation (RMSEA) to assess model fit, with values of >0.97 and <0.05 indicating acceptable model fit, respectively (Schermelleh-Engel, Moosbrugger, and Müller 2003). Lastly, we computed the coefficient estimates of indirect effects using 1000 bootstrap replicates.

Results

In total, 1625 spiders representing 17 families were collected, of which 838 spiders (52%) belonged to the family Lycosidae (hunting spiders) and 508 (31%) belonged to the family Linyphiidae (web-building spiders) (Table 1). Nine families totaling 1018 specimens were classified as hunting spiders, and eight families totaling 607 specimens were classified as web-building spiders. Path analysis indicated that web-building spiders responded differently to abiotic and biotic factors compared to hunting spiders (Fig. 3, Table 3). The final path analysis model was consistent with the data (χ² test, P = 0.78, df = 10), and CFI and RMSEA values were 1.00 and 0.00, respectively, indicating good model fit.

Figure 3:

Final path analysis model. Solid arrows represent positive correlations and dashed arrows indicate negative correlations. The width of the arrows indicates the strength of the relationship, with wider arrows representing stronger relationships among variables. Statistical results are provided in Table 3.

Open in new tabDownload slide

Web-building spiders were negatively associated with plant biomass (Std. Est. = −0.47, P = 0.02, Figs 3 and 4A), but positively associated with the soil pollution index (Std. Est. = 0.49, P = 0.02) (Figs 3 and 4B, Table 3). A weak negative association was observed between web-building spiders and landscape diversity (Std. Est. = −0.37, P = 0.07) (Figs 3 and 4C). Hunting spiders were positively associated with landscape diversity (Fig 4F; Std. Est. = 0.40, P = 0.04), and to a lesser degree, the soil pollution index (Fig 4E; Std. Est. = 0.36, P = 0.06) and prey breadth (Fig 4D; Std. Est. = 0.33 P = 0.08) (Fig 3 and Table 3). Our final model also revealed that the percentage of clay in the soil had a positive influence on vegetation biomass (Std. Est. = 0.67, P < 0.01; Table 3). No significant indirect associations among spiders and abiotic and biotic variables were detected (Table 3).

Figure 4:

Partial residual plots of web spiders (A–C) and hunting spiders (D–F) from our final path analysis model. Grey area represents 95% confidence intervals. Statistical results are provided in Table 3.

Open in new tabDownload slide

Discussion

We investigated the relative importance and strength of landscape and local filters in structuring spider communities across greenspaces in an urban ecosystem. Our findings indicated that spider community assembly was influenced by multiple abiotic and biotic drivers across spatial scales related to landscape composition, soil quality, prey breadth, and habitat management designs. The path analysis model revealed that: (1) hunting spider abundance increased with available prey breadth, (2) hunting and web-building spider abundance increased with soil heavy metal contamination, (3) web-building spiders declined with vegetation biomass, and (4) landscape diversity was associated with the abundance of web-building and hunting spiders. Importantly, web-building spiders and active hunting spiders responded differently to landscape and local drivers, highlighting the need to incorporate a functional perspective based on species’ ecology to understand processes of community assembly.

Landscape diversity, but not greenspace patch isolation, was identified as a moderately strong driver influencing spider communities among greenspaces in the city. Differential responses to landscape diversity were observed among spiders based on their hunting strategy and web-dependency, suggesting that there may be trade-offs in traits related to mobility and resource capture in heterogeneous urban landscapes. Active ground-hunting spiders were positively associated with landscape diversity, which aligns with our second hypothesis. Wolf spiders (Lycosidae) were the dominant taxon of ground-hunting spiders collected during the study, and these spiders are generalist predators that thrive in disturbed environments (Samu and Szinetár 2002; Burkman and Gardiner 2015). Species composition of spiders differ among urban greenspace types (Lowe et al. 2018), suggesting higher landscape diversity in cities may support increased species abundance and diversity. Because active ground-hunting spiders rely on patch connectivity for dispersal corridors among greenspaces (Bonte et al. 2003; Braaker et al. 2017), increased landscape diversity surrounding vacant lots may have resulted in a concentration effect leading to higher abundances within these habitats (Tscharntke et al. 2012). The opposite response was observed for web-building spiders such as species in the families Linyphiidae and Tetragnathidae, which were negatively associated with landscape diversity. Web-building spiders are smaller in size, but considered strong dispersers due to their capacity to balloon throughout their life cycle (Bonte et al. 2004; Blandenier 2009). This method of dispersal should facilitate movement of small web-building spiders among greenspaces in heterogeneous urban landscapes compared to larger ground-hunting spiders (Delgado de la flor et al. 2020), allowing them to exploit diverse habitat patches. However, compared to active hunting spiders, web-building spiders likely have less control over the direction and distance of dispersal via ballooning, which could negatively impact their survival in diverse landscapes. Contrary to our first hypothesis, greenspace patch isolation was not an important driver of spider communities, regardless of their hunting strategy and web-dependency. The overabundance of greenspace represented as vacant land within this urban ecosystem may have reduced the strength of this landscape-level anthropogenic filter.

Prey breadth was weakly positively correlated with active hunting spiders, but web-building spiders were not associated with either prey metric (i.e. breadth or counts). The prey breadth metric incorporates measures of prey abundance and richness such that it represents potential prey availability for spiders within the vacant lot sites. This finding suggests that higher abundance and richness of prey found in a greenspace patch facilitates local establishment of active ground-hunting spiders. Studies have reported that spider populations inhabit greenspace patches with abundant prey (Harwood, Sunderland, and Symondson 2001; Afzal et al. 2013; Lowe, Wilder, and Hochuli 2016), suggesting a direct relationship among spiders and prey resources within local food webs. Contrary to our hypothesis, the abundance of prey alone was not a significant driver, suggesting that the identity of prey is important for active hunting spiders. The most abundant prey taxa collected in these urban vacant lots were Isopoda, Collembola, Acari, and Hymenoptera. When actively searching for prey, ground-hunting spiders can select particular arthropod taxa, and this selective process may be related to factors other than prey abundance such as handling time or nutritional requirements (Schmidt et al. 2012). For example, the taxonomic identity of prey was the most important determinant for predator-prey interactions of the wolf spider Pardosa glacialis (Thorell), as the gut contents of individuals remained similar while the abundance and composition of available prey changed along an environmental gradient (Eitzinger et al. 2019). Because prey availability positively impacts the fitness of active hunting spiders, such as species of Lycosidae (Uetz, Bischoff, and Raver 1992; Schmidt, Harwood, and Rypstra 2013), urban greenspace patches that support a wider prey breadth may accommodate the prey preferences of diverse ground-hunting spider populations. Conversely, web-building spiders primarily use their webs to collect resources. These species may be more constrained by the availability of arthropod prey in a greenspace patch, although there is evidence that some spiders will relocate their webs based on spatiotemporal changes in prey availability (McNett and Rypstra 1997). Webs of linyphiid spiders have been reported to catch small arthropod prey such as Collembola (Harwood, Sunderland, and Symondson 2001). Web-building spiders have evolved the ability to identify and exploit microhabitats that are rich in prey such that the placement of their webs is not random (Harwood, Sunderland, and Symondson 2003). Fine-scale vegetation heterogeneity in urban vacant lots supported greater web capture breadth and reduced prey overlap for web-building spiders, facilitating resource partitioning and supporting more abundant and diverse spider communities (Gardiner et al. 2021). Importantly, the availability and identity of potential prey taxa changes throughout the growing season. However, there is a lack of information regarding species’ prey preferences, the degree of prey selectivity among spiders, and whether exploitative competition is a factor structuring predator communities. Our study took an inclusive approach to defining prey taxa by considering all potential prey for spiders within this system. This coarse approach to assessing predator-prey relationships may have affected the strength of these biotic variables within our model. Moreover, the predator-prey relationships documented in this study represent a short temporal window when prey species are likely abundant but does not capture the interactions that may occur when prey are scarce.

Plant biomass was negatively correlated with web-building spiders, but unrelated to active hunting spiders, which provides partial support for our hypothesis. Vegetation structural complexity has been associated with spider community composition (Uetz 1991, Langellotto and Denno 2004, Burkman and Gardiner 2014), as well as arthropod populations more broadly including prey (Langellotto and Denno 2004, Borer et al. 2012, Burkman and Gardiner 2014, Sarthou et al. 2014). The web-building spiders collected within this study were generally small (∼3 mm) and such species prefer structurally simple vegetation such as low growing turf grass because their webs are constructed close to the ground (Delgado de la flor et al. 2020). In this study, urban vacant lots were seeded with a turf grass mixture following demolition of residential buildings, but these sites were commonly colonized by weedy plant species, often referred to as urban spontaneous vegetation (Robinson and Lundholm 2012; Riley et al. 2018). The colonization and establishment of urban spontaneous vegetation increased the habitat heterogeneity within vacant lot sites, but not to the extent of the seeded native grasses and forbs in the pocket prairies. Additionally, regular monthly mowing of vacant lots maintained lower plant height and biomass in vacant lots compared to pocket prairies where plants grew throughout the summer and were mown annually in the fall. Linyphiid and tetragnathid (i.e. Glenognatha foxi McCook) web-building spiders have been found in large numbers in urban vacant lots (Burkman and Gardiner 2015, Delgado de la flor et al. 2020; Gardiner et al. 2021), where their small size and the close proximity of their webs to the ground allow them to survive periodic mowing disturbance.

Abundance of web-building spiders and active hunting spiders were positively correlated with the concentration of heavy metals in the soil. Elevated levels of contamination such as heavy metals are common within the surface soils of vacant lots (Jennings et al. 2002; Sharma, Basta, and Grewal 2015). For example, in Cleveland, OH, USA, more than half of the residential vacant lots investigated had soils with lead levels that exceeded the US EPA remediation threshold of 400 mg kg⁻¹ (Perry et al. 2020). In this study, urban vacant lot soils were moderately contaminated with heavy metals, but deposition of pollutants in the soil can be highly heterogeneous. High spatial variability in the contamination levels of vacant lot soils is influenced by legacies related to distance to historic industrial sources, major roadways, and built structures, as well as age of demolished structures and length of vacancy (Schwarz, Pouyat, and Yesilonis 2016). Exposure directly or indirectly to elevated levels of heavy metals including lead can reduce the fitness of arthropods through effects on their development (Scheifler et al. 2002; Lagisz 2008; Cheruiyot et al. 2013), reproduction (Hendrickx et al. 2003; Lagisz and Laskowski 2008; Eraly et al. 2011), immune response (Stone, Jepson, and Laskowski 2002; Migula et al. 2004; Sorvari et al. 2007), and behavior (Eraly, Hendrickx, and Lens 2009; Sorvari and Eeva 2010), including in spiders (Yang et al. 2016). These fitness effects can manifest in arthropod populations, resulting in simplified communities and altered ecosystem services such as pest suppression (Gardiner and Harwood 2017). For example, vacant lots contaminated with heavy metals such as lead, cadmium, zinc, copper, arsenic, and antimony had greater abundances of small, soil-dwelling Acari and Collembola, but lacked larger predator and detritivore species such as spiders (Perry et al. 2021). We hypothesize that the increased abundance of web-building spiders and active hunting spiders in moderately contaminated sites may reflect their exploitation of available niches (Jung et al. 2008; Żmudzki and Laskowski 2012) due to the loss of other predatory arthropod taxa that may have reduced heavy metal tolerance. It is possible that these populations of arthropods living within the city have adapted to higher levels of soil contamination that are common in urban areas (Jacquier et al. 2021). For example, species of Pardosa (Family Lycosidae) can tolerate and resist heavy metal contamination in soils via mechanisms such as accumulation of metals in their bodies and elevated detoxification pathways (Wilczek and Migula 1996; Wilczek and Babczyńska 2000). Pardosa milvina (Hentz) is a dominant, generalist species in urban greenspaces, including the vacant lots sampled in this study (Burkman and Gardiner 2015; Delgado de la flor et al. 2020). Exposure to heavy metal contamination can have fitness trade-offs for species inhabiting urban greenspaces (Chen et al. 2011; Eraly et al. 2011; Gardiner and Harwood 2017) that likely contribute to patterns of species abundance and community composition by favoring disturbance-adapted species. In this study, total concentrations of heavy metals were quantified, which does not necessarily reflect bioavailable levels in the soil, but they are often correlated.

Direct relationships among spiders and landscape and local environmental filters were prevalent within the studied urban ecosystem, but we did not detect any indirect relationships. Although less frequently investigated, indirect interactions are hypothesized to be important drivers of multi-trophic interactions in food webs. For example, soil properties and plant biomass can indirectly influence spider communities via changes in prey availability (Wilder et al. 2011; Maathuis and Diatloff 2013; Philpott et al. 2014; Gardiner and Harwood 2017). Moreover, soil properties and quality can enhance the diversity of plant and arthropod communities via nutrient availability (Wilder et al. 2011; Maathuis and Diatloff 2013; Perry et al. 2021), or can prevent the establishment of arthropods via contamination or reductions in prey availability (Jung et al. 2008; Gongalsky, Filimonova, and Zaitsev 2010; Gardiner and Harwood 2017). Warmer temperatures from the urban heat island effect can influence predatory-prey interactions in urban ecosystems (Evans et al. 2013; Dale and Frank 2014; Meineke, Dunn, and Frank 2014) via changes in prey abundance (Raupp, Shrewsbury, and Herms 2010; Meineke et al. 2013) and altered spider community composition (Meineke et al. 2017). Although evidence suggests that indirect relationships are prevalent, our path analysis model did not identify any significant interactions, and therefore, did not support our hypothesis that local environmental factors such as soil properties and plant biomass would indirectly influence spider communities via changes in prey availability. It is possible that the short temporal window investigated as well as the timing and spatial scale of investigation in this study limited our ability to detect indirect relationships among spider communities and the biotic and abiotic drivers.

In this study, we used a multi-trophic approach to examine the simultaneous direct and indirect effects of landscape and local abiotic and biotic filters on mid-summer spider community assembly. We demonstrated that multiple landscape and local filters associated with urban ecosystems influenced spider community structure across greenspaces in a legacy city with large holdings of vacant land. Landscape diversity was a stronger driver of spider community structure among vacant lots than greenspace patch isolation, suggesting that legacy cities with large holdings of vacant land can support diverse arthropod communities and have conservation value. At local scales, arthropod biodiversity and services can be supported in urban greenspaces by diversifying habitat management strategies, as prey breadth and plant biomass influenced the abundance of different spider guilds. Because spiders responded differently to these landscape and local drivers based on their foraging strategies (i.e. web-building vs active hunting), a combined taxonomic and functional approach is recommended in future studies investigating processes of arthropod community assembly. Additionally, our findings indicated that supporting biodiversity and associated ecosystem services in cities requires a multi-scale approach to conservation management that addresses stressors associated with landscape (composition and diversity of patches) and local (soil quality, prey availability, and habitat management) factors. Importantly, this multi-scale approach is needed in addition to urban greening efforts to achieve desired conservation outcomes.

Acknowledgements

We thank the undergraduate research assistants, Jennifer Thompson, Michael Friedman, Ryan Crozier, Ryan Byler, Alison Zahorec, and Zach Hart, for their help with field collections and laboratory work. This work was supported by NSF CAREER DEB Ecosystem Studies Program (CAREER-1253197) to M.M.G., USDA AFRI Agroecosystem Management Program (20166701925146) to M.M.G. and P.L.P., and the NSF Graduate Research Fellowship Program (DGE-1343012) to Y.A.D.

Author contributions

Yvan A. Delgado de la flor (Conceptualization [equal], Data curation [lead], Formal analysis [equal], Investigation [lead], Methodology [lead], Visualization [equal], Writing—original draft [supporting], Writing—review & editing [supporting]), Kayla I. Perry (Data curation [supporting], Visualization [equal], Writing—original draft [lead], Writing—review & editing [lead]), Lyndsie M. Collis (Data curation [supporting], Formal analysis [equal], Methodology [supporting], Validation [lead], Visualization [supporting], Writing—review & editing [supporting]), P. Larry Phelan (Conceptualization [equal], Investigation [supporting], Methodology [supporting], Writing—review & editing [supporting]), and Mary Gardiner (Conceptualization [equal], Funding acquisition [lead], Investigation [equal], Methodology [equal], Project administration [lead], Visualization [supporting], Writing—review & editing [supporting])

Conflict of interest statement

None declared.

Data availability

The data and analysis code that support the findings of this study are openly available on GitHub: https://github.com/kiperry/Spider_Path_Analysis.

References

Author notes