Nonnative plant invasion increases urban vegetation structure and influences arthropod communities

Ecological theory and empirical evidence indicate that greater structural complexity and diversity in plant communities increases arthropod abundance and diversity. Nonnative plants are typically associated with low arthropod abundance and diversity due to lack of evolutionary history. However, nonnative plants increase the structural complexity of forests, as is common in urban forests. Therefore, urban forests are ideal ecosystems to determine whether structural complexity associated with nonnative plants will increase abundance and diversity of arthropods, as predicted by complexity literature, or whether structural complexity associated with nonnative plants will be depauperate of arthropods, as predicted by nonnative plant literature.

Arthropods form a necessary component of the global food web for a suite of wildlife and are often ecologically tied to plant communities.This group is diverse, demonstrating variable responses to changes in the environment (Haddad et al., 2009;Hamilton et al., 2013;Koricheva & Hayes, 2018).Arthropods are experiencing global declines, which can disrupt energy transfer in food webs and affect ecosystem services (Hallmann et al., 2017;Lister & Garcia, 2018;Sánchez-Bayo & Wyckhuys, 2019).Arthropods are a particularly important food resources for birds, providing calcium and amino acids, and arthropods in urban forests are often crucial to bird success in urban environments (Long & Frank, 2020;Nagy & Holmes, 2005;Seress et al., 2018).The phenology of strict herbivores like caterpillars, the larval form of butterflies and moths in the order Lepidoptera, are reliant on particular plant communities (Koricheva & Hayes, 2018).When changes in plant communities occur, herbivore communities can respond with changes in abundance and biomass (Koricheva & Hayes, 2018), or richness and diversity (Haddad et al., 2009).Strict predators like spiders, in the order Araneae, are an ecologically important taxa (New, 1999) and can be indirectly tied to plant communities that sustain large communities of prey (Haddad et al., 2009).One study reported a greater effect of plant diversity on predators than on herbivores, suggesting plants may affect predators independent of herbivore dynamics (Koricheva & Hayes, 2018).Haddad et al. (2009) found a threefold increase in predator abundances in areas of high plant richness, without a similar increase in herbivorous prey, suggesting differing dynamics between plant diversity and herbivore and predator responses.
Together, changes in plant communities can drastically alter arthropod abundance, biomass, richness and diversity.
Additionally, nonnative plant species are geographically separated from their native arthropod community and therefore lack an evolutionary history with arthropods native to the invaded area.This means that adding nonnative plants to a community does not necessarily increase food resources for local herbivores and can reduce arthropod diversity as existing species are outcompeted (Adams et al., 2009;Kadlec et al., 2018;Narango et al., 2018).Arthropods frequently encounter nonnative plants in human-dominated landscapes, both in residential plantings and in disturbed areas, where nonnative plants released from the ornamental plant trade often thrive (Dyderski & Jagodziński, 2019;Lehan et al., 2013;Pickett & Cadenasso, 2009;Reichard & White, 2001).Nonnative plant invasions are common in urban forests which are often adjacent to managed landscapes and frequently disturbed by human activities (Malkinson et al., 2018;Pickett et al., 2001).
We were interested in addressing (1) whether arthropods responded, in abundance, biomass, richness or diversity metrics, to the presence of understory vegetation, or specifically to the presence of native or nonnative vegetation, and (2) whether a difference in response could be detected at the order level (all arthropods) or family level (spiders only).We analysed spiders separately from other arthropods because they are strict predators and an ecologically important taxon (New, 1999), abundant and gave an opportunity to compare trends at a refined taxonomic resolution.We also analysed relationships between caterpillar abundance and biomass, and total, nonnative and native structural complexity, to determine if a strict herbivore group responded differently than arthropods combined.

| Site selection and vegetation sampling
This research was conducted in Raleigh, North Carolina (35.778928,) and Newark, Delaware (39.683016, -75.753548)USA as part of the FRAME (FoRests Among Managed Ecosystems, https:// sites.udel.edu/frame/)network of forests used to study urban forest ecology.We sampled 12 urban forests in each city for a total of 24 temperate deciduous and mixed forests within the eastern US (Mitchell et al., 2023b, EMS Figure 1).We established a 25 m × 25 m flagged grid within each forest and randomly selected 10 points from each, with the stipulation that they could not be immediately next to each other.We sampled plant communities at these points, hereafter called sampling locations, during the summer growing season in Newark forests in 2015 and in Raleigh forests in 2017, for a total of 239 sampling locations (one small forest only accommodated nine; Mitchell et al., 2023b;Trammell et al., 2020).Plant communities within the FRAME have been extensively described, including overstory and understory trees and understory woody and herbaceous vegetation (Table 1; Landsman et al., 2019;Mitchell et al., 2023b;Trammell et al., 2020).We measured understory vegetation at two strata within a 2.5-m radius (19.6-m 2 area) flagged circle surrounding sampling locations; we estimated percentage cover by ground layer vegetation, and we counted stems in the shrub layer vegetation.We defined ground vegetation as any plant species, including woody and herbaceous species, <0.5 m tall.We estimated percentage cover by nonnative and native ground vegetation, and identified the most abundant nonnative and native species.We defined a nonnative species as one not occurring in the lower 48 states of America prior to 1600 CE (Gleason & Cronquist, 1991;Trammell et al., 2020;Weakley, 2015).We defined shrub layer vegetation as any plant stem, including woody and herbaceous species, reaching 1.0 m or taller and less than or equal to 2.5-cm diameter at breast height (1.4 m; Nowak et al., 2008; Figure 1).We identified and categorized counted stems as native or nonnative, and we combined ground cover and stem density species identifications to determine total understory plant richness.
All species were identified in the field with select identifications confirmed by the North Carolina State University Herbarium Curator.was vacuumed for 3 min (Figure 2); contents were contained in a mesh bag and stored in the field in a large kill jar until returned to the lab and transferred samples to the freezer until processing (Mitchell, 2021).

| Arthropod collecting and processing
All non-arthropod material was carefully removed from vacuum samples in the lab prior to arthropod identification and weighing (Mitchell, 2021).Arthropods were identified to order and placed in separate, based on sampling location, pre-weighed 1.5 mL vials (VWR® Micro Centrifuge Tubes, Cat.No. 10025-724) and topped with 95% ethanol for temporary storage.If present, caterpillars (213 larval Lepidoptera from 200 sampling locations) were placed in a second pre-weighed vial and topped with 95% ethanol.Spiders (order Araneae) were identified further to family (Bradley, 2019;Ubick et al., 2009), and placed in a third pre-weighed vial and topped with 95% ethanol.To quantify biomass, vials were left open under a hood for 24 h to allow excess ethanol to evaporate, then dried in a 50°C drying oven for 48 h (McCluney et al., 2017).Once dry, all vials were closed and weighed in milligrams and biomass was determined by subtracting each vial's empty weight from its full weight (Mitchell, 2021).

| Vegetation structural complexity metrics
To separate the effects of nonnative and native vegetation from the structure of all understory vegetation, we created three similarly constructed vegetation structural complexity metrics.Each metric was created using three vegetation variables collected from the same sampling locations (N = 239): percentage ground cover, stem counts and understory species diversity, which was determined using the Shannon diversity index (H).To represent total structural complexity, our metric incorporated percentage of total ground cover, total stem counts and shrub layer plant diversity.We combined data from both cities to determine the full data range for total ground cover (1%-100%), total stem counts (1-182), and total shrub layer diversity (0.1-1.8); this was to ensure valid comparisons between cities.Since each vegetation variable was measured in a different unit (percent, count, number), we divided each range of raw data values sampled for each variable into five, equally divided bins (equivalent to 0.1-1.0,1.1-2.0,2.1-3.0,3.1-4.0,4.1-5.0)that represented increasing complexity.Each raw data value was associated with its representative bin of increasing complexity (1-5; least to most complex), and once all raw data values were reassigned with a bin number, we summed bin numbers across the three variables to create a metric value representing total understory structural complexity for each sampling location (possible range: 0-15).The nonnative and native complexity metrics were then created using the same bin ranges created for the total complexity metric, to ensure they were at comparable scales for analyses.However, the nonnative structural complexity metric was created using raw data values of percentage nonnative ground cover, nonnative stem counts and nonnative understory diversity, while the native structural complexity metric was created using raw data values of percentage native ground cover, native stem counts and native understory diversity.
TA B L E 1 Summary of urban forest size and dominant plant species found within the sampled area.

| Statistical analyses
All analyses were conducted in R version 4.0.5 (R Core Team, 2021).
We pooled arthropod data collected at our two time points by sampling location (N = 239) and calculated total abundance, biomass, richness (order level for non-spider arthropods and family level for spiders), and Shannon diversity index (H; order level for non-spider arthropods and family level for spiders) for non-spider arthropods and spiders.We separately analysed relationships between caterpillar abundance and biomass, and total, nonnative and native structural complexity.
To determine how well native and nonnative plants predicted total structural complexity, we used linear regression analyses to elucidate relationships between predictor variables (nonnative understory richness, native understory richness, nonnative structural complexity metric, and native structural complexity metric) and the response variable, total structural complexity metric.Our response variable met the assumptions for normalcy, so we fit these regressions to a gaussian distribution (Ives, 2015;Zuur et al., 2010).
Though we anticipated a Poisson distribution, comparisons of 'best fit' confirmed the use of gaussian models in analyses.To determine how arthropods responded to total structural complexity, we used linear regressions to determine relationships between our predictor variable, total structural complexity, and arthropod metric response variables (non-spider arthropod and spider abundance, biomass, richness and diversity).We log (x + 1) transformed non-spider arthropod and spider abundance and biomass data to improve normalcy but did not transform non-spider arthropod and spider richness or diversity, and fit regressions to a gaussian distribution (Ives, 2015, Zuur et al., 2010).To determine how arthropods responded specifically to nonnative and native structural complexity, we used the lmer function from the 'lme4' package (Bates et al., 2015) to create mixed effect models for each city.We specified nonnative plant structural complexity, native plant structural complexity, and their interaction as our predictor variables and ran models for each of the following response variables: non-spider arthropod abundance, biomass, richness, diversity and spider abundance, biomass, richness and diversity.Forest site (N = 24) was included in all models as a random variable.

| Nonnative and native plant structural complexity and arthropod communities
In Raleigh, non-spider arthropod abundance (r 2 = .330,p = .050; Figure 6a) and biomass (r 2 = .400,p = .010;Figure 6c) increased as nonnative structural complexity increased.The nonnative structural complexity metric did not correlate with any other non-spider or spider arthropod metric (p > .05).There was no significant effect of the native structural complexity metric (p > .05),or the interaction between the nonnative and native structural complexity metrics (p > .05), on non-spider or spider arthropods (Figure 6).
In Newark, non-spider arthropod abundance increased both as nonnative (r 2 = .450,p = .031)and native (r 2 = .450,p = .001; Figure 7a) structural complexity metrics increased.Non-spider arthropod biomass also increased with increased nonnative (r 2 = .500,p = .001)and native (r 2 = .500,p = .001;Figure 7c) structural complexity metrics and there was a significant negative interaction between these metrics (p = .001).We interpret this interaction as the positive relationship between nonnative and native structural complexity metrics and non-spider arthropod biomass being weaker when one of those metrics is high, which makes sense considering the sum of both values does not exceed 100 in our dataset, and both metrics cannot continuously increase (Table S1.3).Spider abundance (r 2 = .180,p = .037;Figure 7b) and biomass (r 2 = .360,p = .018;

| DISCUSS ION
Based on the consistency in direction and strength in linear response, total structural complexity in urban forest understories was more related to nonnative plant richness and structural complexity than native plant richness and structural complexity (Figure 3).not evenly distributed (Gotelli, 2008), suggesting only certain nonspider orders and spider families increased in abundance as total structural complexity increased.
These results suggest urban forests invaded by nonnative plants can support some arthropod populations which in turn provide food resources for other wildlife, including birds, that depend on calcium and amino acid-rich foods to feed their young (Frank et al., 2019;Long & Frank, 2020;Nagy & Holmes, 2005;Parsons et al., 2020).
Birds rely on many food resources, but caterpillars are a particularly importance source of protein for developing young (Cramp, 1998).
Previous reports suggest native plant species are essential for arthropod communities required by insectivorous and other bird species (Burghardt & Tallamy, 2013, 2015;Tallamy & Shropshire, 2009) and that nonnative plants can indirectly reduce bird population growth (Narango et al., 2018).While our arthropod collection protocol was not designed to quantify urban forest caterpillar communities specifically, we employed a consistent sampling technique and investigated how caterpillar abundance and biomass correlated with understory vegetation structural complexity.We found no relationship between caterpillars and total, nonnative or native structural complexity.

| Plant origin and understory plant structural complexity
Relationships between total structural complexity and nonnative and native understory plant richness differed by city.Previous reports detailing FRAME vegetation communities found Raleigh forests had more nonnative understory vegetation and Newark forests had more native understory vegetation (Mitchell et al., 2023b).The most abundant understory species found in Raleigh were primarily nonnative (Nonnative: Hedera helix, Ligustrum sinense, Lonicera japonica, Microstegium vimineum, Vinca sp.; Native: Smilax rotundifolia, Vitis rotundifolia) with vining and low-growing herbaceous structure.
In Raleigh, increased native species richness did not contribute to increased total structural complexity, while increased native species richness did contribute to increased total structural complexity in Newark.In Raleigh, this suggests total structural complexity is so predominately comprised of nonnative plants that an increase in native species richness does not increase vegetation structure.

| Understory plant structural complexity and arthropod communities
Arthropod metrics generally increased with greater total structural complexity, but specific relationships differed by city.Similarly, other studies have shown variable responses of arthropod communities to vegetation structural complexity.Greater understory structural complexity is associated with greater insect and spider abundance, but reduced sider richness, in rural Maryland forests (Landsman & Bowman, 2017), greater arthropod occupancy in Australian urban green spaces (Threlfall et al., 2017), greater butterfly richness in urban Brazilian forests (Orlandin & Carneiro, 2021), and reduced ant richness in Australia woodlands (Lassau & Hochuli, 2004).These comparisons suggest a similarity in vegetation structure serving as food resources and refugia for arthropods, but enforce the speciesand regionally-specific nature of these relationships.From our results, non-spider arthropod and spider diversity in Newark were the only arthropod response metrics to negatively correlate with increased structural complexity (Figure 4).Dense shrub layer vegetation correlated with reduced carabid beetle diversity in French urban forests (Croci et al., 2008), while Argentinian forests with low vegetation complexity had reduced arthropod diversity (Gardner et al., 1995).Relationships with arthropod diversity are likely more nuanced than trends among abundance and biomass because species have specific life history traits, for example host specificity and behaviour with or without other organisms, that may not always be known and would contribute to unpredictably.Non-spider arthropod diversity in Raleigh was greater than in Newark, yet spider diversity was similar among cities.This may indicate some influence of Raleigh's greater proportion of nonnative plant species on nonspider arthropod diversity, as Landsman and Bowman (2017) reported greater spider diversity in areas of dense nonnative ground cover compared to structurally less complex control areas.
Due to logistics of resources and the global pandemic, we were unable to collect temporally paired vegetation and arthropod community datasets, which may limit the interpretability of our results and conclusions.Additionally, the low taxonomic resolution of our arthropod dataset likely masks the specific responses that could be found if we had identified arthropods to the genus or species level.
However, we argue that our use of a large order level dataset, plus the companion analyses using spiders identified at a more refined taxonomic level, is appropriate for our purpose of understanding broad relationships between plant and arthropod communities across a regional scale.And perhaps, the order and family level results presented here represent a conservative detection of real response.

| Nonnative and native plant structural complexity and arthropod communities
Significant relationships between arthropod metrics and nonnative and native structural complexity were less numerous than those between arthropod metrics and total structural complexity.Responses differed by city, but in general, structurally complex nonnative and native vegetation was associated with increased arthropod abundance and biomass and did not correlate with richness and diversity.
We suggest this is because increased ground cover and stem count is more linearly related to increased physical structure than species diversity.In an uninvaded, predominately native understory, it could be reasonable to assume that as native plant richness increases, so does the volume and therefore structure of the understory.However, nonnative species, by definition, have limited evolutionary histories in their areas of invasion and tend to become invasive by their ability to rapidly propagate and spread (Aronson et al., 2007).Therefore, plant diversity could be lower in an urban forest heavily invaded by a few dominant species, compared to a less-invaded urban forest dominated by multiple species historically evolved for the area and consequently less dense.This then disrupts a theoretical linear correlation between understory density and diversity, as the relationship between density and diversity diverge once total vegetation is separated into nonnative and native species with unique life histories.
Previous reports documented that nonnative plant richness correlated with reduced arthropod abundance and biomass (Landsman & Bowman, 2017), nonnative plant density was associated with increased arthropod abundance and spider diversity (Landsman et al., 2020), and native plant richness and density correlated with increased arthropod abundance and biomass (Adams et al., 2020;Landsman & Bowman, 2017;Threlfall et al., 2017).In studies comparing arthropod communities between nonnative and native congeners, native plants had greater arthropod abundance, biomass, and richness (Ballard et al., 2013;Southwood et al., 2004).In Raleigh, insect predator abundances were similar on nonnative and native congeners of urban trees, but predator community composition on nonnative congeners differed from those on native species (Frank et al., 2019).Together, these reports indicate that nonnative and native plant richness and structural complexity influence arthropod metrics and alter arthropod community compositions.

| Conclusions and management implications
We documented support from two cities for the hypothesis that greater vegetation complexity will increase arthropod abundance and biomass, and against the hypothesis that nonnative vegetation will decrease arthropod abundance and biomass.We found evidence in urban forests that greater nonnative and native structural complexity was associated with increased arthropod abundance and biomass, but not diversity, as predicted by plant complexity literature.Native, not nonnative, structural complexity was associated with increased spider abundance and biomass in Newark, but nonnative plants were not depauperate of arthropods, as predicted by nonnative plant literature.This suggests urban forests invaded by nonnative plants may still provide adequate food biomass to transfer energy to the next trophic level but may not provide adequate resources to support the full suite of species found in urban forests.
This could indicate that heavily invaded urban forests may fail to provide ecological services and functions offered by diverse species integral to forest ecosystems, like forest specialists (Magura & Lövei, 2021;Martinson & Raupp, 2013;Mitchell et al., 2023b).
Urban land managers should consider plant species origin when managing urban green space and prioritize the removal of nonnative structure and replacement with native species when allocating vegetation maintenance resources.While total vegetation complexity correlated with increased arthropod abundance and biomass, we only support the planting of native species to promote arthropod abundance and biomass.Nonnative plant invasion can reduce plant diversity (Trammell et al., 2012), which can deplete available resources for arthropods and reduce diversity (Narango et al., 2018;Tallamy, 2004;Tallamy & Shropshire, 2009).But nonnative vegetation removal alone is not sufficient; once removed of nonnatives, areas can be left barren and susceptible to further invasion if not revegetated with native species (Moore et al., 2023).
Specific species depend on geographic location, but our results suggest fast growing and persistent nonnative species, particularly vining and evergreen species, would be primary targets for removal.Following removal, cleared areas should be planted with native ground cover and shrub layer plant species that produce berries, nuts or other resources in addition to foliage to support native species from multiple trophic levels.This report details relationships between urban forest arthropods and understory vegetation structural complexity, and demonstrates that nonnative and native structural complexity exert different influences on arthropod abundance and biomass.
We collected arthropods within each forest from the same locations established for plant community sampling (N = 239) but expanded the sampling circle to 5 m radius (78.5 m 2 ).In 2019, vacuum samples were collected twice in Raleigh between June 14-19 and July 17-19 and in Newark between June 22-24 and July 22-25.Understory vegetation F I G U R E 1 Schematic diagram (left) showing the flagged grid across the sampled forest area (white dots) and 10 sampling locations (dark green dots) where plant and arthropod sampling occurred; a depiction (right) of how understory plant stems were counted (adapted from Mitchell et al., 2023b, figure 1).
Dominance was determined for overstory vegetation (above 2.5-cm diameter at breast height; 1.4 m) by incorporating species abundance and volume.Dominance was determined for understory vegetation (<2.5-cm diameter at breast height; 1.4 m) by abundance.F I G U R E 2 Differences between (a) high structural complexity and (b) low structural complexity, as found within one urban forest in Newark.

Figure 7d )
Figure 7d) increased as native structural complexity increased.There were no significant responses of non-spider arthropod or spider richness (p > .05)or diversity (p > .05) to nonnative or native structural complexity metrics (see Table S1.3 in Appendix S1).When considering caterpillars alone, neither caterpillar abundance or biomass had a significant response to nonnative (Figure 5b,e) or native (Figure 5c,f) structural complexity metrics, in either city.
This indicates understory structural complexity in the urban forests we sampled consisted more of nonnative plant structure than native plant structure.Increased understory vegetation complexity increased arthropod abundance and biomass (Figure 4), despite the associated increase in nonnative vegetation.Total structural complexity generally increased arthropod abundance and biomass but had no effect on arthropod richness, and reduced diversity in Newark.This indicates the increase in arthropod abundance was F I G U R E 3 Influence of (a) nonnative understory plant richness, (b) native understory plant richness, (c) nonnative structural complexity, and (d) native structural complexity on total understory structural complexity (collected from sampling points, N = 239).Raleigh sites are shown in orange and Newark sites are shown in purple, significant relationships shown with 95% confidence intervals.Data points were jittered to reduce overlap.F I G U R E 4 Influence of total structural complexity on (a) non-spider arthropod and (b) spider abundance, (c, d) biomass, (e, f) richness, and (g, h) diversity (collected from sampling points, N = 239).Raleigh sites are shown in orange and Newark sites are shown in purple, significant relationships shown with 95% confidence intervals.Note the y axes are fitted to the displayed data and not consistent across all graphs; data points were jittered to reduce overlap.

F
Influence of total structural complexity, nonnative structural complexity, and native structural complexity on (a-c) caterpillar abundance and (d-f) biomass (from sampling points where immature lepidoptera were collected, N = 200).Raleigh sites are shown in orange and Newark sites are shown in purple, no significant relationships were detected.Data points were jittered to reduce overlap.F I G U R E 6 Influence of nonnative (orange) and native (pale orange) structural complexity metrics on (a) non-spider arthropod and (b) spider abundance, (c, d) biomass, (e, f) richness, and (g, h) diversity in Raleigh (collected from sampling points, N = 119).Significant relationships shown with 95% confidence intervals.Data points were jittered to reduce overlap.| 1273MITCHELL et al.

F
Influence of nonnative (purple) and native (pale purple) structural complexity metrics on (a) non-spider arthropods and (b) spider abundance, (c, d) biomass, (e, f) richness, and (g, h) diversity in Newark (collected from sampling points, N = 120).Significant relationships shown with 95% confidence intervals.Yellow star indicates a significant interaction; data points were jittered to reduce overlap.