Local and landscape factors differentially influence predatory arthropods in urban agroecosystems

Abstract

Like other urban green spaces, urban community gardens can act as biodiversity refugees, especially for small organisms like arthropods. In turn, arthropods can provide important ecosystem pest control services to these agroecosystems. Thus, an often-asked question among urban gardeners is how to improve gardens and surrounding areas for natural enemies and associated pest control services. We examine how local vegetation and garden characteristics, as well as the surrounding landscape composition, affect ground-dwelling beetles (Coleoptera: Carabidae and Staphylinidae), spiders (Araneae), opilionids (Opiliones), and ladybird beetles (Coleoptera: Coccinellidae), all of which are important predators. In the summer of 2019, we collected predators, vegetation, ground cover, and garden and landscape characteristic data from 10 community gardens in the city of Seattle, Washington. We found that different groups of natural enemies are associated with different environmental variables and at different scales; probably related to differences in their dispersal capabilities, habits, and diets. Floral variables (number of flowers and number of species in flower) had a negative effect on nonflying natural enemies (spiders, opilionids, and ground-dwelling beetles), but not on flying ones (ladybird beetles). The only taxon that was significantly affected by a landscape-scale variable was Opiliones, the only group examined that exclusively disperses by ground. Our results show contrasting results to similar studies in different regions and highlight the need to expand the taxa and regions of study.

INTRODUCTION

As a growing percentage of the global landscape continues to urbanize, it is increasingly important to understand the ecology of urban green spaces and the ecosystem services they provide. The study of biodiversity conservation in urban landscapes has become a more predominant research field in recent years (Barbosa, 2020; Burkman & Gardiner, 2014; McKinney, 2002, 2008), particularly so for urban agroecosytems (Liere & Egerer, 2020). This is because despite the overall detrimental effects of urban sprawl for biodiversity, urban green spaces, including urban farms and community gardens, can foster microcosms of high biodiversity (Burks & Philpott, 2017; Goddard et al., 2010; Lin et al., 2017), especially for small organisms like insects and other arthropods. In turn, these organisms mediate important ecosystem services and disservices (Ziter, 2016). Insect herbivores, for example, are important components of urban food webs (Faeth et al., 2005) but can convey a disservice if their populations grow to the point where they cause significant damage to cultivated plants, including those grown for food or ornamental purposes in urban agroecosystems. However, herbivore populations can be kept from reaching damaging levels by their predators and parasitoids (Tscharntke et al., 2005, 2012). Thus, natural enemies provide an important ecosystem service to urban gardeners and farmers who, due to restrictions on chemical pesticides, rely on natural pest control methods (Oberholtzer et al., 2014).

Because of the importance of natural enemies in urban agroecosystems, many studies now focus on analyzing the environmental factors that affect these organisms in urban farms and community gardens. Studies based in urban community gardens have found that natural enemies and their associated pest control services are affected by local garden features and gardening practices, such as garden size (Burks & Philpott, 2017), plant diversity and abundance, and ground cover materials (Bennett & Gratton, 2012; Burks & Philpott, 2017; Egerer, Arel, et al., 2017; Otoshi et al., 2015; Sattler et al., 2010). For example, in Wisconsin, Bennett and Gratton (2012) found that flower diversity within urban gardens increased parasitic Hymenoptera abundance. In separate studies in central California and Córdoba city, Argentina, spider activity density and richness decreased with percent of bare ground cover in the gardens (Argañaraz et al., 2018; Otoshi et al., 2015).

Natural enemies are also affected by features of the surrounding landscape, such as the proportion of the area covered with natural/seminatural habitats (Burkman & Gardiner, 2014; Egerer, Bichier, et al., 2017; Gardiner et al., 2014), developed areas (Bennett & Gratton, 2012; Egerer, Arel, et al., 2017; Grez et al., 2019) or agricultural land (Philpott, Lucatero, et al., 2020). Within these, the effect of urban cover, estimated as the percent of impervious cover at the landscape scale, has varying effects on different taxa in different regions. Bennett and Gratton (2012) found that landscape-scale impervious cover decreased parasitoid diversity in Wisconsin, while Burks and Philpott (2017) found positive effects of urban cover on parasitoid abundance but negative effects on superfamily and family richness in the California central coast. Similarly, diversity of ladybird beetles in urban gardens increased with urbanization in California but decreased with urbanization in Michigan (Egerer, Li, et al., 2018).

Thus, the effect of environmental variables on natural enemies can be region-dependent. Moreover, because natural enemies vary greatly in their dispersal capabilities, different taxa are affected by different environmental variables and at different scales (Egerer, Arel, et al., 2017; Grez et al., 2019; Liere et al., 2019). Consequently, in order to comprehensively understand how to enhance natural enemies in urban gardens, it is important to continue expanding the array of studied areas and taxa. Our study investigated local and landscape factors as they relate to abundance and diversity of predatory arthropods with different dispersal capabilities in the Pacific Northwestern US city of Seattle, WA. As past studies investigated natural enemies either in rural systems or in urban settings of other regions, our study fills a regional knowledge gap on the role of local and landscape factors specific to the Pacific Northwest and city of Seattle.

As in other agroecosystems, successful natural pest control in urban community gardens likely benefits from increased complementary effects of different groups of natural enemies with a variety of needs, habits, diets, and dispersal capabilities. Opilionids, for example, mostly feed upon small and soft-skinned prey, such as springtails and aphids (Papura et al., 2020). Spiders range greatly in size, habits, and hunting modes, and thus, their prey are as diverse as their group (from small aphids to large caterpillars, from ground-dwelling to vegetation-dwelling) (Michalko et al., 2019). Ground-dwelling beetles are effective in pest control of ground-dwelling herbivores, such as aphids (mostly fallen of the host plant), fly eggs, beetle larvae, caterpillars, and slugs (Kromp, 1999; Reich et al., 2020). Ladybird beetles mostly feed on vegetation-dwelling aphids and coccids (Hodek et al., 2012) and are very mobile in urban environments (Liere et al., 2019).

Our goal was to understand the relationship between environmental variables and the abundance and richness of ground-dwelling and flying natural enemies in community gardens in Seattle. We investigated the effect of local garden-scale variables including vegetation and garden management factors (age, size, and ground cover type). We also examined the effect of landscape-scale variables, including land-use cover type as characterized in the National Land Cover Database (NLCD), as well as the amount of public urban parks cover in the surrounding landscape. These local and landscape variables were evaluated on their effect on the abundance of ground-dwelling predatory beetles (Coleoptera: Carabidae and Staphylinidae) and opilionids (Opiliones), the abundance and family richness of spiders (Araneae), and the abundance and species richness of ladybird beetles (Coleoptera: Coccinellidae).

METHODS Study site

We conducted the study in the city of Seattle, Washington, located in the US Pacific Northwest (47.6062° N, 122.3321° W). Seattle's population in 2020 was estimated to be 737,015 in an area of 215 km² (Office of Planning and Community Development, 2023). While Seattle is among the fastest growing cities in the United States, the city is committed to protecting urban biodiversity in its various green spaces (City of Seattle, 2018) and has an increasing demand for urban agriculture. The Community Garden program alone oversees 89 community gardens throughout the city. These gardens occupy about 10 ha where food is grown for gardeners and for the general public (City of Seattle, 2023).

Our study took place in 10 of these urban community gardens (Figure 1). The gardens are managed in an allotment style where households rent and cultivate individual plots within the garden. The chosen gardens range in size from 240 to 16,187 m², housing 21 to 259 individual plots, have been in operation from 5 to 46 years, and are >2 km from each other. All selected gardens are administered by Seattle Department of Neighborhoods' P-Patch Program which requires use of organic gardening inputs and methods (Seattle Department of Neighborhoods, 2020). Thus, no synthetic chemicals including pesticides, insecticides, herbicides, weed killers, and fertilizers are allowed anywhere in the gardens.

FIGURE 1. Map of urban community gardens used as study sites in Seattle, WA, USA (47.6062° N, 122.3321° W), and the National Land Cover categories in 500-m buffers around each garden.

To standardize the sampling area of our study sites, we established a 20 × 20 m plot in the center of each garden. Our samplings and observations were limited to these areas for the duration of our study.

Landscape-scale variables

We used land-cover data from the 2011 NLCD, 30-m resolution (Homer et al., 2015), and calculated the percentage of land-cover types in 500-m buffers from the center of each garden. The 500-m buffer has been used to study landscape effects of many taxa (Batáry et al., 2012; Concepción et al., 2008; Otoshi et al., 2015; Schmidt et al., 2008). We used five land-cover categories established by the (NLCD): developed open, developed low, and developed medium/high (we combined the NLCD categories of “developed, medium intensity” and “developed, high intensity into one category),” and natural/seminatural (which included deciduous forest, evergreen forest, mixed forest, shrub/scrub, herbaceous, and hay/pasture), and agricultural (listed in the NLCD as “cultivated crops”) (Multi-Resolution Land Characteristics, 2023). In addition, we calculated the proportion of urban parks in the 500-m buffers using the City of Seattle parks map available through the King County GIS website (https://kingcounty.gov/services/gis.aspx). These parks are managed by the city and have a variety of uses and characteristics (Appendix S1: Table S1).

We included urban parks as one of our landscape variables because from studies in rural agricultural systems, we know that farms embedded in landscapes with a higher proportion of natural habitats (i.e., forests, wetlands, and grasslands) support higher local density and diversity of beneficial arthropods, even in fields with low local vegetation diversity (Bianchi et al., 2006; Chaplin-Kramer et al., 2011; Tscharntke et al., 2005). In cities, especially rapidly expanding ones like Seattle, nearby “natural” or “seminatural” areas consist largely of urban parks and reserves—habitats which may be vital to connect apparently isolated urban green spaces (Langellotto et al., 2018). Much like fragments of forests, grasslands, and wetlands in rural agricultural landscapes (Landis et al., 2000; Schellhorn et al., 2014), urban parks may provide alternative resources, prey, and shelter, thus enhancing natural enemy abundance and diversity in nearby urban agroecosystems.

Garden-scale variables

Vegetation was sampled three times between June and August 2019, approximately a month in between sampling periods. Vegetation was sampled within the same standardized 20 × 20 m plot in each garden. Canopy cover was measured using a concave spherical densitometer at the center of each plot in addition to 10 m to the north, south, east, and west of the center. Inside each of the 20 × 20 m plots, we counted and identified all trees and shrubs (woody vegetation). We also recorded the number of trees and shrubs in flower. Within the 20 × 20 m plot, we then selected eight locations to place 1 × 1 m plots. To randomly select each of the eight locations, we first marked four 5 × 20 m strips within the 20 × 20 m. For each strip, using a random number table from 0 to 20, we chose two random numbers (which represented, in meters, the distance from 0 to 20 m from the beginning to the end of the length of the strip). We then walked along the edge the strip until reaching the randomly chosen distances and then used a second random number table from 0 to 5 (which represented, in meters, the distance from 0 to 5 m from one edge to other of the width of the strip) to choose the location of the plot. We repeated this procedure for the four 5 × 20 m strips for a total of eight randomly chosen plots.

Within each of these plots, we measured the height of the tallest herbaceous vegetation and counted the total number of flowers and total number of crops and ornamentals in flower. We identified each plant species and estimated the percentage of cover of each plant type (crop, grass, ornamental, weed, and herbaceous). Within each of these 1 × 1 m plots, we also estimated the percentage of ground cover make-up of bare soil, mulch/wood chips, straw, and leaf litter.

In addition, we obtained information on garden size (garden area in square meters and number of individual plots) and garden age (in years since establishment) from the city of Seattle community garden information website (City of Seattle, 2023).

Natural enemies

At each garden site, we conducted three rounds of natural enemies sampling. This included sampling ground-dwelling beetles (Carabidae and Staphylinidae), spiders (Araneae) and opilionids (Opiliones), and ladybird beetles (Coleoptera: Coccinellidae). We sampled natural enemies three times between June and August, 2019. The first round of sampling occurred between June 24 and June 26, the second round between July 17 and July 19, and the final round between August 12 and 13. Natural enemies were sampled using a combination of visual and trapping sampling methods (see below). We estimated total abundance across all sampling methods and sampling periods for the focus natural enemies (ground-dwelling beetles, spiders, opilionids, and ladybird beetles) (see Data analysis). We lumped Carabidae and Staphylinidae into one category—ground-dwelling beetles—and estimated abundance for all. Per time limitations, we only were able to further identify spiders (to family) and ladybird beetles (to species). Thus, in addition to abundance, for spiders we also estimated family richness and for ladybird beetles species richness across all sampling methods and periods.

Visual sampling

Using the same randomized methodology described for the vegetative sampling, eight 0.5 × 0.5 m quadrants within each garden's 20 × 20 m plot were selected. In each of these 0.5 × 0.5 m plots, one person visually searched in the vegetation for 10 min for ladybird beetles, spiders, opilionids and ground beetles. All specimens were collected and preserved in vials with alcohol (with the exception of minimal escaped specimens, we were unable to collect; we identified these specimens visually in the field to family for spiders and morphospecies for ground and ladybird beetles). We recorded the number of individuals (for all), family (spiders), and species (ladybird beetles).

Traps

Four random trap locations were selected in each 20 × 20 m plot using the aforementioned randomization methodology. At each location, four 7.62 × 12.7 cm yellow sticky cards (BioQuip Products, Compton, CA, USA) on 20-cm wire stakes were placed in each corner of a 0.5 × 0.5 m quadrant. A pitfall trap was placed in the middle of the quadrant flush with the ground, filled up one-third with water and dish soap. After 24 h, the traps were retrieved and the specimens were identified.

Data analysis

All data are available from Dryad (Liere & Cowal, 2024). For abundances of spiders, opilionids, and ground beetles, we summed the total number of individuals from both the pitfalls and visuals (none were found in sticky cards) and across the three sampling periods. Similarly, for ladybird beetles we summed the total abundance data from both sticky cards and visuals (none were found in pitfalls) and across the three sampling periods. For species/family richness (ladybeetles and spiders, respectively), we counted the total number of species/families sampled across all sampling methods and periods. These total abundances and species/family richness across sampling methods and periods were then used as response variables for the analyses.

Because of the large number of predictor variables, we first ran Pearson's correlations to identify correlated variables and to select variables for subsequent analysis (see Appendix S1: Table S2). We grouped variables into biologically relevant groups (e.g., garden features, vegetation variables, ground cover variables, and landscape variables). We then identified variables that were significantly correlated with one another and selected non-correlated variables per group to include in the analysis. There were 11 final variables included in the analysis: tree/shrub richness, number of flowers, number of species in flower, number of crop species, grass cover, ornamental cover, herbaceous cover, mulch cover, number of weed species, garden size, and proportion of parks in a 500-m radius.

Of the selected variables, those that did not meet conditions of normality were log transformed (trees/shrub richness, number of flowers, grass cover, ornamental cover, garden size, and proportion of parks in 500 m). To assure that we did not have collinearity between some of the remaining variables (Zuur et al., 2009), we checked the variable inflation factor (VIF) with the vif function in the car package version 3.0-2 (Fox & Weisberg, 2011). All variables had VIF scores below 3.

To examine the effect of the predictor variables on ground beetle abundance, opilionids abundance, spider abundance and family richness, and ladybeetle abundance and species richness, we used generalized linear models (GLMs) with the glm function in R (R Development Core Team, 2018).

We first ran a global model including all the 11 predictor variables. We then tested all combinations of the selected predictor variables with the glmulti function (Calcagno & de Mazancourt, 2010) and selected the top model based on the Akaike information criterion corrected for sample size (AIC_c) values. Then, we averaged all models within delta 2 of the best model, with the model.avg function in the MuMIn package (Barton, 2012) and reported conditional averages for significant model factors. We visualized all significant predictors of each dependent variable from either top-averaged model with the visreg package in R (Breheny & Burchett, 2013). For all models we used a Gaussian distribution.

RESULTS

In our three monthly surveys (July–August), we found a total of 1144 arthropods, including 125 individuals of ground-dwelling beetles (69 Carabidae; 56 Staphylinidae), 276 ladybird beetles (Coccinellidae) from the 11 species (the most abundant ladybird beetles being Harmonia axyridis with 30% of all individuals, Stethorus punctum with 23.6%, Coccinella septempunctata with 11.6%, and Scymnus marginicollis with 11.6%). We found 331 spiders from 11 families and 412 opilionids (Table 1).

TABLE 1 Abundance of orders, families, and species found in urban community gardens in Seattle, WA.

Order	Family/species	No. gardens found	Abundance
Araneae
	Lycosidae	9	188
	Theridiidae	10	56
	Araneidae	8	33
	Clubionidae	2	3
	Salticidae	8	18
	Dictynidae	1	2
	Dysderidaea	1	3
	Agelenidae	1	1
	Unknown	8	25
	Linyphiidae	1	1
	Thomisidae	1	1
Opiliones		10	412
Coleoptera	Coccinellidae
	Harmonia axyridis	10	83
	Stethorus punctum	9	65
	Coccinella septempunctata	8	32
	Scymnus marginicollis	9	32
	Psyllobora renifer	2	25
	Unknown spp.	10	18
	Cycloneda polita	5	11
	Adalia bipunctata	4	4
	Cycloneda sanguinea	2	2
	Nephus binaevatus	2	2
	Scymnus coniferatum	1	2
	Carabidae	10	69
	Staphylinidae	10	56

Note: For the analysis, Carabidae and Staphylinidae were grouped into one category (ground-dwelling beetles). For abundances of spiders, opilionids, and ground beetles, we summed the total number of individuals from both the pitfalls and visuals (none were found in sticky cards) and across the three sampling periods (June, July, and August 2019). Similarly, for ladybird beetles we summed the total abundance data from both sticky cards and visuals (none were found in pitfalls) and across the three sampling periods. For species/family richness (ladybeetles and spiders, respectively), we counted the total number of species/families sampled across all sampling methods and periods. These total abundances and species/family richness across sampling methods and periods were then used as response variables for the analyses.

^aOnly family not commonly known to use ballooning as dispersal method.

For ground-dwelling beetle abundance, we found that 17 models were within two AIC_c points of the best model (Table 2). These averaged models included six variables but only three of those had (marginally) significant effects. Ground-dwelling beetle abundance (marginally) significantly decreased with trees/shrub species richness, with the number of flowers, and with the number of species in flower (Table 2 and Figure 2).

TABLE 2 Generalized linear model results (averaged top models [2 AIC of best model]).

Dependent variable and factors in averaged model	No. models in which factor was included	No. models averaged	Factor weight	Estimate	z	p
Ground-dwelling beetle abundance
No. trees/shrub species	12	17	0.74	−2.331	1.841	0.066
No. species in flower	8	17	0.46	−2.645	1.673	0.094
No. flowers	7	17	0.41	−0.991	1.657	0.097
No. weed species	6	17	0.34	0.462	1.408	0.159
Garden size	4	17	0.19	0.906	1.233	0.218
No. crop species	2	17	0.08	−0.240	1.053	0.292
Spider abundance
No. species in flower	6	6	1	−5.844	2.532	0.011
No. trees/shrub species	2	6	0.36	2.791	1.447	0.148
No. weed species	2	6	0.34	−0.706	1.396	0.163
Ornamental cover	1	6	0.13	1.577	1.109	0.268
Proportion of parks in 500 m	1	6	0.09	−7.483	0.787	0.431
Opilione abundance
No. flowers	3	3	1	−2.553	2.345	0.019
Proportion of parks in 500 m	3	3	1	−26.751	2.275	0.023
No. trees/shrub species	3	3	1	6.348	2.704	0.007
No. weed species	1	3	0.25	−0.727	1.153	0.249
No. crop species	1	3	0.23	−0.490	1.1	0.271
Ladybird beetle abundance
Herbaceous cover	6	6	1	0.230	2.272	0.023
Grass cover	4	6	0.69	2.943	1.792	0.073
Mulch cover	4	6	0.53	0.237	1.415	0.157
Proportion of parks in 500 m	1	6	0.14	−13.145	1.39	0.165
No. trees/shrub species	1	6	0.1	−2.333	1.213	0.225
Spider family richness totals
No. trees/shrub species	1	3	0.234	0.351	0.839	0.401
Mulch cover	1	3	0.223	0.028	0.789	0.430
Ladybird beetle species richness
Grass cover	3	3	1	0.643	2.016	0.043
Herbaceous cover	3	3	1	0.038	2.028	0.042
Mulch cover	3	3	1	0.088	2.936	0.003
No. crop species	1	3	0.27	0.077	1.234	0.217
No. weed species	3	3	0.23	0.106	1.134	0.256

Note: The factors that were present in the best model (lowest AIC_c) appear in boldface. In cases in which no factors appear in boldface, the best model was the null model. Abundance and richness values were estimated as the totals across all sampling methods (visuals, pitfalls, and sticky cards) and sampling periods (June, July, and August 2019).

Abbreviation: AIC_c, corrected Akaike information criterion.

FIGURE 2. Generalized linear model results of averaged models of environmental variables on predatory arthropods in urban gardens in Seattle, WA. Significance represented with size of circles (small, p = 0.05–0.1; medium, p = 0.01–0.05; large, p [less than] 0.01). Variables present in averaged best models (but not significant) shown without circles. Directionality represented with + and blue color (positive) and − and red color (negative). Ground beetles, Coleoptera: Carabidae and Staphylinidae; spiders, Araneae, various families; Opiliones, Opiliones, various families; ladybird beetles, Coleoptera, Coccinellidae. prop., proportion. Illustrations by Charlotte Grenier.

For spider abundance, five variables were present in the averaged top models (six models). Of those, number of species in flower was the only one that had a significant and negative effect (Table 2 and Figure 2). The best-fit model for spider family richness was the null model; no variables included in the best averaged models had significant effects (Table 2 and Figure 2). Opilionid abundance was negatively affected by the number of flowers and the proportion of parks in the landscape, while tree and shrub species richness increased opilionid abundance.

Lastly, ladybird beetle abundance increased with herbaceous and grass cover; the remaining three variables present in the averaged six models had no significant effects (Table 2 and Figure 2). Ladybird species richness was also positively associated with herbaceous and grass cover, as well as with mulch cover (Table 2 and Figure 2).

DISCUSSION

We found that, in general, local garden characteristics were more important for the natural enemies we examined than the landscape context in which the gardens were embedded (which, in our case, was the proportion of area covered by urban parks). In addition, we found variable associations between different natural enemies had local variables, which likely reflects natural enemies' habits and dispersal capabilities.

Garden vegetation effects

Management of vegetation cover, diversity, and abundance is unique to each garden and varies greatly even between individual plots within a garden (Philpott, Egerer, et al., 2020). Gardeners are making constant decisions on the abundance and diversity of crops, medicinal, and ornamental plants to grow in their plots, as well as how much and how often to remove associated vegetation (“weeds” and other herbaceous plants) from their plots and the gardens' common areas. Consequently, an often-asked question is how to manipulate vegetation characteristics to enhance habitat quality for natural enemies and support better natural pest control services. We did not find any one variable that had an overall effect—positive or negative—on all the studied groups, meaning that there is no single management practice that would render the garden “better” for all natural enemies. However, with one exception, we also did not see conflicting effects of these variables on the different taxa. That is, there appear to be no noteworthy trade-offs in vegetation management that would benefit one group at the expense of others. Unfortunately, though not examined directly here, there may be potential trade-offs between managing for pollinators and ground-dwelling natural enemies. Particularly, we found that the number of flowers and number of species in flower, which are beneficial for pollinators in urban agroecosystems (Cohen et al., 2020; Lowenstein et al., 2019; Matteson & Langellotto, 2010; Tasker et al., 2020), had a negative association with opilionids and ground-dwelling beetles. This trade-off, however, did not include ladybird beetles, which are capable of directed flight and use pollen and nectar as alternative food sources (Hodek et al., 2012).

Why the number of flowers and species in flower had a negative association with opilionids spiders and ground-dwelling beetles is not clear. Higher floral abundance may attract other aggressive predators, such as ants and predatory wasps (which were found in our pitfalls and sticky cards but not quantified due to time limitations), that may interfere and compete with, or even prey upon, spiders, opilionids, and carabids (Karami-jamour et al., 2018; Rosenheim et al., 1995). In particular, generalist, opportunistic, and competitive dominant species like the invasive Argentine ants (Linepithema humile (Mayr)), common in urban habitats (Carpintero & Reyes-López, 2014; Uno et al., 2010), can disrupt the activity of other natural enemies (Daane et al., 2007; Milosavljević et al., 2021).

Further evidence that higher floral abundance may indirectly represent an added stress to some species of natural enemies comes from a study in California where gardens with high flower abundance were associated with ground beetle species with small body size (Philpott et al., 2019), a trait often associated with highly disturbed sites (Magura, 2017). However, contrary to our results, spider density and richness have been found to increase with increasing abundance of flowering plant species (Otoshi et al., 2015) and may depend on the floral species present, demonstrating that these effects may be context-dependent and temporal. Notably, in our study, all ground-dwelling taxa (ground beetles, spiders, and opilionids) showed similarity in their responses to vegetation variables and differed from those of ladybird beetles. Thus, it is likely that different functional groups of natural enemies and their associated pest control services may be respond to different components of local vegetation variables (Philpott & Bichier, 2017; Shackelford et al., 2013).

Ground cover

Within the associations between ground cover variables and natural enemies we examined, the only one of statistical significance was between mulch cover and ladybird beetle species richness; namely, species richness of this group increased with higher percentage of ground covered by mulch. Interestingly, the opposite trend was found in other studies in urban community gardens, albeit the effects of mulch being only significant in certain landscape contexts (Egerer, Bichier, et al., 2017). On the one hand, mulch cover may represent a disturbance and physical barrier to some species (Burkman & Gardiner, 2014; Quistberg et al., 2016) but, on the other hand, and similarly to leaf-litter cover, mulch can provide additional hiding places, alternative prey, and overwintering sites (Liere et al., 2019; Philpott et al., 2019) when other features of the garden like trees, shrubs, hedges, or “weedy” strips are absent or scarce. Thus, ground cover garden features may be as important as vegetation features for managing for beneficial arthropods, even flying ones like ladybird beetles (Liere et al., 2019), but the magnitude and direction of the effects may be context-dependent.

Landscape effects

Because the proportion of the surrounding area covered by parks was positively correlated with the proportion of area covered by natural and seminatural habitats (as classified by the NLCD) and negatively correlated with urbanization cover—both important drivers of arthropod communities in urban green spaces (Burkman & Gardiner, 2014; Burks & Philpott, 2017; Egerer, Arel, et al., 2017; Egerer, Li, et al., 2018; Fetridge et al., 2008; Hernandez et al., 2009; Matteson et al., 2008; Pardee & Philpott, 2014)—we were expecting to see significant effects of this landscape variable on the taxa we studied. In particular, based on previous studies, we were expecting stronger effects of landscape-scale variables on natural enemies that disperse mostly by flying than on those that disperse by walking. Interestingly, the only taxa that was associated with the proportion of parks in the surrounding landscape was Opiliones, which among the taxa we studied is the only group that has no species able to disperse by air and thus colonizes new habitats on the ground (Drapela et al., 2008).

We found a negative relationship between opilionid abundance and the proportion of the buffer areas covered by parks. It is possible that because of the lack of alternative habitats, opilionids may aggregate in community gardens in areas with no park cover, similar to what ladybird beetles do in highly urbanized landscapes (Egerer, Liere, et al., 2018). In contrast, in areas with a large proportion of park cover, opilionids may tend to aggregate in parks instead. Because their dispersal is limited to the ground, it is likely that once they arrive at an adequate habitat, opilionids tend to stay there. Contrastingly, however, the other nonflying natural enemies we studied, ground-dwelling beetles and spiders, did not show the same response to this landscape variable. It is also possible that predators of opilionids congregate in higher abundances in areas with a larger proportion of park cover. For example, Passerine birds, the most frequent predators of opilionids (Powell et al., 2021), may be more abundant in more forested areas, thus exhibiting higher predation on opilionids nearby to parks.

Though ground beetles mostly disperse by walking (Zaller et al., 2008), and these beetles' migration between isolated patches in cities is rare due to highways (Koivula & Vermeulen, 2005), there are some species that are able to fly (Do et al., 2014; Kromp, 1999). So, even though abundance of ground-dwelling beetles in our study gardens was not associated with the proportion of parks in the area, it is possible that flying and nonflying species may have been differently affected. We also did not find significant effects of landscape variables on spider abundance and family richness. Accordingly, spiders have often been found not to be greatly affected by urbanization gradients or distance to wooded habitats (Alaruikka et al., 2002; Argañaraz et al., 2018; Melliger et al., 2018; Schmidt et al., 2005). This is because even though they are incapable of directed flight, most of them are able to disperse great distances by ballooning and thus may be less prone to fragmentation effects in urban areas than other ground-dwelling predators (Vergnes et al., 2012).

In regard to flying natural enemies, we found that ladybird beetle abundance and richness were associated with local-level factors but not with landscape-level ones. These results contrast with similar studies which have found that landscape variables are important drivers for ladybird beetle communities in urban green spaces. For example, in California, the amount of natural habitat cover in the landscape had a significant influence on ladybird beetle abundance and richness (Egerer, Bichier, et al., 2017). In fact, Grez et al. (2019) found that local habitat variables were not important predictors of coccinellid richness and abundance in green spaces across an urbanization gradient extending from the city of Santiago, Chile; instead, they found that both native and non-native species were negatively affected by urbanization.

In our study, among the most common ladybird beetle species were H. axyridis (30% of all individuals) and C. septempunctata (11.6%), both of which are non-native to the area. When separating out native and non-native ladybird beetles (Appendix S1: Table S3), we found that although only 3 of the 11 ladybird beetle species recorded are non-native, they appear to be driving the overall ladybird beetle responses to local and landscape factors (showing a positive and statistically significant association with % grass cover and % herbaceous plant cover). The analysis for the native ladybird species did not yield any significant effects, probably due to their low abundances. It is worth noting that even though non-native ladybird beetles are likely contributing to pest regulation, the biological pest control benefits of H. axyridis and C. septempunctata may be coupled with negative impacts on native coccinellid species and, in some cases, may disrupt pest suppression (Koch & Galvan, 2008). Future studies should aim to determine if this is the case in urban agroecosystems, where sensitive species have already been filtered out by high levels of disturbance.

Context-dependent results

These contrasting results underline the importance of context, both in regard to region and taxa of study. Egerer, Li, et al. (2018) propose that the urbanization history of the landscape as well as local weather conditions can explain the divergent patterns found in different studies. Seattle is characterized by a temperate marine climate with frequent precipitation (166 days in 2014), cool winters, and mild, short, dry summers (Zhao et al., 2019). In contrast, studies in California (Egerer, Arel, et al., 2017; Egerer, Bichier, et al., 2017; Egerer, Liere, et al., 2018) and Phoenix feature little rain year-round, and both Baltimore and Ohio experience frost. Thus, differences in the effect of floral resources on natural enemies in our study compared with others may be related to the unique climatic characteristics of the Pacific Northwest. For example, because of shorter and milder summer conditions, floral resources in Seattle may not be so limited to irrigated environments as they are in regions with longer, hotter, and dryer summers (e.g., California, Arizona), even if studies take place during the same time period (June–September).

Landscape composition varies greatly between different cities and so does the range of cover of different land-use types. For example, the cover of natural and seminatural land in 500-m buffers around our study sites in Seattle ranged from 0% to 13% (mean of 4%), while in Egerer, Liere, et al. (2018)'s study in California, natural and seminatural land cover ranged from 0% to 52.63% (mean of 10%), and in Grez et al. (2019)'s study in Santiago, Chile, the range was 6%–67% cover (mean of 28%; thought their buffers had a 1000-m radius). If the effects of the different land-use types on mobile organisms is not linear across the whole range, then these widely different ranges (and maximum values) may explain why the different studies found such contrasting results. That is, our sites had a relatively low natural/seminatural habitat cover, and we found that ladybird beetles were not associated with this variable (or to the significantly correlated proportion of park cover). In California, the range of natural/seminatural habitat cover was relatively larger, and this variable had a significant and negative effect on ladybird beetles (Egerer, Bichier, et al., 2017); in Chile, with an even larger range, this land use had the opposite effect on ladybird beetles (Grez et al., 2019). Further, it should be noted that our sample size of 10 gardens was smaller than aforementioned comparable studies which included 12, 19, 24, or 82 gardens, which may explain why some of the cited studies found patterns that we did not.

In addition, some urban landscapes may lack land uses that are common in others, or be separated by physical barriers. For example, while several studies have shown the importance of agricultural cover on different natural enemies in urban gardens (Grez et al., 2019; Otoshi et al., 2015; Philpott, Lucatero, et al., 2020), we did not evaluate the effects of this land use because only two of our sites had any agriculture cover in their surrounded areas. Furthermore, Seattle Metro is a coastal city with the Puget Sound to the west and, in addition, Lake Washington to the east, a barrier that may hinder westward spillover of natural enemies from nearby agricultural or wooded areas. These unique urban landscape features and climatic characteristics may explain why some studies find significant effects of landscape variables while others, like ours, do not.

LIMITATIONS AND FUTURE DIRECTIONS

Given the variability in garden management and our sample size of only 10 garden field sites, we are limited in our ability to find significant patterns; however, trends still emerge which indicate generally how different taxa are associated with garden management practices in Seattle. Further, for the visual counts, some specimens managed to escape and thus were only identified visually in the field rather than in the lab; even though these were a small portion of our samples, it could be a potential source of error in our analysis. However, since the visual counts were done by the same people, using the same methods across all gardens, this source of error was not likely a confounding variable in our study. Lastly, due to the observational nature of our study, we cannot conclude causal relationships between the variables we examined. However, similarly to the many studies examining local and landscape patterns on agroecosystem communities, our study contributes to identifying important links between environmental variables and beneficial arthropods in urban green spaces.

All taxa examined are important predators and have different dispersal, diet, and habitat requirements. Consequently, it is not surprising to find that different taxa showed different associations with environmental filters. Even though our study suggests that there is no single management strategy that would likely enhance the habitat for all beneficial insects, increasing vegetation complexity seems to have an overall positive effect. While floral variables had a negative effect on ground-dwelling predators, probably because of indirect intraguild interactions, flowers are ephemeral, and their negative effects may be temporal.

Our results suggest that using a variety of garden practices to enhance the habitat quality for the different taxa could promote successful natural pest control in urban community gardens by increasing the complementary effects of each group of natural enemies and their varying needs, habits, diets, and dispersal capabilities. Positive and negative intraguild interactions as well as potential trade-offs with other ecosystem service providers are worth further investigation. Future studies should examine the services and disservices of non-native predatory insects on pest regulation and their competition with native species, and the overall effect on urban biodiversity. Lastly, the majority of herbivorous insects present in urban agroecosystems do not cause significant damage to cultivated plants and are an important component of biodiversity. Because varying levels of predation pressure can either facilitate or hinder prey coexistence and diversity, understanding how management practices impact predatory insects is an important step toward understanding the top-down control forces that shape biodiversity within urban ecosystems.

AUTHOR CONTRIBUTIONS

Both authors contributed to the study conception and design. Material preparation and data collection were performed by Heidi Liere and Sanya Cowal. Data analysis was performed by Heidi Liere. The first draft of the manuscript was written by Heidi Liere and both authors commented on previous versions of the manuscript. Both authors read and approved the final manuscript.

ACKNOWLEDGMENTS

Funding for this project came from the Murdock College Research Program for Natural Sciences (grant no. G01821-NP-AAS), Seattle University's Summer Faculty Fellowship Program, College of Science and Engineering Undergraduate Student/Faculty Research Award, and the Environmental Studies Department. Caroline Grandia, Brandon McWilliams, and Emily Nguyen provided invaluable field assistance. We thank King County P-patch coordinators and gardeners at Burke-Gillman, Magnuson Park, Picardo's P-Patch, Phinney Ridge, Jackson Park, Queen Ann, Hillman City, New Holly, Up Garden, and Beacon Food Forest for access to the gardens and permission to collect the arthropod specimens. We thank Charlotte Grenier for the illustrations.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Data (Liere & Cowal, 2024) are available from Dryad: https://doi.org/10.5061/dryad.3tx95x6mx.