Impervious surface and local abiotic conditions influence arthropod communities within urban greenspaces

Site selection

Arthropods were sampled from eight urban gardens and eight pocket prairies across metropolitan Grand Rapids, Michigan, USA (see Fig. 1). Urban gardens were predominately compromised of edible plants, but often contained non-edible wildflowers. The urban garden sites were located on non-residential properties, and functioned as community garden spaces for groups of individuals to manage. Pocket prairies were defined as habitats that have undergone a restoration process through the intentional planting of native prairie plant species. Sites were compiled by first establishing a list of known pocket prairies and urban gardens within a 25 km radius of Grand Rapids City Hall as a proxy of the city center. Then, permissions were requested for each site to visit and collect specimens during the summer of 2020. We secured permissions from 16 collection sites that were spatially separated by a minimum distance of 750 m between each site to maintain the spatial independence of each research site and reduce pseudoreplication. Sites were managed by the City of Grand Rapids, City of Hudsonville, Township of Ada, Township of Gaines, Creston Neighborhood Association, Blandford Nature Center, New City Neighbors, Urban Roots, Dominican Sisters of Grand Rapids, Fairway Christian Reformed Church, Rosewood Church, Calvin University, and Cornerstone University.

Sampling methods

Each site was sampled once per month (June, July) on sunny days with temperatures above 70° F (21.1 °C). Sampling occurred in the morning (8 am–10 am) and afternoon (3 pm–5 pm) on the same day. During each sampling event, multiple measurements were taken near the center of each site in a region that contained visible vegetation representative of the habitat (urban garden, pocket prairie). The temperature and relative humidity were recorded using a digital psychrometer (Model # CECOMINID048683). From the site center, a 15 m transect was marked and the total number of individual flowering plants and their colors (white, yellow, purple) were recorded within 1 m of the transect, as others have done (Pardee & Philpott, 2014; Otoshi, Bichier & Philpott, 2015; Burdine & McCluney, 2019b). We considered inflorescences to be a single flower. To sample the arthropods, 12 large 350 mL bowls (4 blue, 4 yellow, 4 white) and 12 small 175 mL bowls (4 blue, 4 yellow, 4 white) were placed along the transect. Both small and large sampling bowls were used to capture arthropods of varying sizes. Each bowl contained a soap (Dawn) and water mixture to prevent specimens from escaping the bowls. Bowls were set out in the morning and retrieved in the afternoon on the same day. Upon collection, samples were placed into plastic collection containers filled with 70% ethanol for later identification. The collection containers were then labeled and stored at Cornerstone University (Grand Rapids, Michigan, USA).

All arthropods sampled were identified to morphospecies (S1). Arthropod abundance was calculated as the total number of arthropods measured at each site, and morphospecies richness was calculated as the total number of morphospecies present at each site. For the diversity metric, we calculated the Shannon diversity index. We combined arthropod samples from June and July together, and environmental variables between both months were averaged. For temperature and humidity values, we average the measurements taken during morning sampling events.

Landscape characteristics

The percent of impervious surface for each site was calculated using the 2016 National Land Cover Dataset (NLCD) Percent Developed Imperviousness layer (Homer et al. 2020) in ArcGIS v 10.7. The impervious surface value at each site was calculated by placing a 500 m radius buffer around each site, and averaging the impervious surface grid cells from the NLCD dataset. The area, perimeter, and perimeter-area ratio of each site were measured using Google Earth Pro v 7.3.3.7786.

Statistical methods

All statistical analyses were completed using the statistical program R version 3.6.2 (R Core Team 2019). Within this program, we used the “vegan” package to conduct a permutation analysis of variance test (PERMANOVA) on arthropod community composition. Non-metric multidimensional scaling (meta MDS) was utilized to display significant associations between arthropod community composition and environmental factors. The “vegan” package was also used to calculate the Shannon diversity index. The correlation function (cor) was used to check collinearity between environmental variables (S2). Variables with a correlation coefficient above r = ±0.7 were considered correlated, and when this occurred one of the variables was removed from the statistical analyses.

Generalized linear models (GLMs) were used to compare environmental variables (percent impervious surface, site perimeter-area ratio, temperature, humidity, total flowers, white flowers, purple flowers) against response variables (abundance, diversity, morphospecies richness). For arthropod orders with more than 500 individuals sampled, we used the abundance and morphospecies richness of the order as response variables. We established a list of candidate models that examined each environmental variable separately (S3). From this list of candidate models, the model for each response variable with the lowest AICc value was selected. Models within 2 AICc units were considered equivalent, and these models were combined to determine whether an additive or interactive model was a better fit (S4). When the most parsimonious model for a response metric was within 2 AICc units of the null model, we reported the results of the null. In addition, we used the site type (urban garden, pocket prairie) as an additive variable to evaluate the sites utilized in our study. We chose this process to simplify model selection because additive and interactive models are problematic with model averaging approaches (Cade, 2015; Harrison et al., 2018). We tested for overdispersion in all models, and made adjustments by altering family distributions when overdispersion was an issue. The distribution families used after testing for overdispersion were negative binomial (total arthropod abundance, Hymenoptera abundance, Hemiptera abundance, Diptera abundance), Gaussian (diversity), and Poisson (morphospecies richness, Diptera richness, Hemiptera richness, Hymenoptera richness). Assumptions of normality and equal variance were assessed by residual plots, and data transformation were performed when necessary.

Summary statistics

A total of 5,468 individuals were collected that represented 14 different taxonomic orders of arthropods and 66 morphospecies. The majority of arthropods sampled (∼88%) were classified into three orders: Diptera (2,604 individuals), Hemiptera (1,156 individuals). and Hymenoptera (1,026 individuals). The orders Araneae (128 individuals), Coleoptera (330 individuals), and Orthoptera (145 individuals) were less abundant, but were represented by at least 100 individuals. The remaining 8 orders contained a combined abundance of 79 individuals (∼1%). Looking at the response variables at the site level, we report the ranges in total arthropod abundance (126–586 individuals), diversity (2.10–2.73), and morphospecies richness (23–39 morphospecies).

The environmental variables we measured also displayed a broad range of values that reflect the conditions of each site. We report the ranges of total flower abundance (23–188 flowers), purple flower abundance (0.5–50.5 flowers), white flower abundance (1.5–70 flowers), humidity (37.7%–65.1%), temperature (21.8 °C–32.1 °C), impervious surface (21.9%–62.3%), and perimeter-area ratio (1.1–5.1).

Community composition

The PERMANOVA analysis yielded three environmental variables that were significantly associated with arthropod community composition (Table 1). These variables included impervious surface (F_1,7 = 2.62, p = 0.004, R² = 0.13, Fig. 2), white flower abundance (F_1,7 = 2.17, p = 0.032, R² = 0.107, Fig. 2), and humidity (F_1,7 = 1.93, p = 0.047, R² = 0.095, Fig. 2). Site type (urban garden, pocket prairie) was marginally significant (F_1,7 = 1.798, R² = 0.088, Fig. 2), and the distinction between pocket prairies and urban gardens can be seen in the NMDS plot. Nonmetric dimensional scaling plots indicated that the genus Chironomus (midges) were positively associated with impervious surface, and four morphospecies were negatively associated with urbanization: Geometridae, P. rapae, Coenagrionidae, and Camponotus. Humidity was positively associated with the genus Ammophila (thread-waisted wasp).The presence of white flowers appeared to be positively associated with Lasioglossum, Drosophila, Echenopa, and P. japonica.

Abundance, diversity, and richness

Here we report results on the most parsimonious model for each response metric (Table 2). Arthropod abundance (AICc = 194.04, R² = 0.39, Fig. 3A) and arthropod diversity (AICc = −22.23, R² = 0.47, Fig. 3B) were positively associated with humidity. Arthropod morphospecies richness was negatively associated with the number of purple flowers (AICc = 96.07, R² = 0.39), but this was equivalent to the null model (AICc = 97.95). Hemiptera abundance was best explained with an additive model of impervious surface and humidity (AICc = 161.21, R² = 0.57, Figs. 3A and 3B). Hymenoptera abundance was negatively associated with temperature (AICc = 145.32, R² = 0.29, Fig. 4C), and positively associated with perimeter-area ratio (AICc = 147.25, R² = 0.2, Fig. 4D). Hymenoptera abundance was also explained by an additive model of temperature and perimeter-area ratio (AICc = 147.32, R² = 0.36). The remaining response metrics (Diptera abundance, Hymenoptera richness, Hemiptera richness, Diptera richness) were not statistically different from the null models (Table 2).

Humidity and temperature

Temperature and humidity were both important abiotic factors that influenced arthropod communities. Water availability can influence the behavior (Green, Scharf & Bennett, 2005) and physiology (McCluney, Burdine & Frank, 2017; Burdine & McCluney, 2019a) of arthropods, and can limit arthropod populations (Allen et al., 2014; Khaliq et al., 2014; Lagucki, Burdine & McCluney, 2017). Many arthropods that dwell in soil and plant roots require high moisture levels to prevent desiccation (Bayley & Holmstrup, 1999), and we know that impervious surfaces in cities change soil moisture by altering hydrology (Shuster et al., 2005). Since urban gardens and parks often utilize irrigation systems, these could be important inputs for maintaining abundant and diverse arthropod communities. In addition, repurposing vacant land into raingardens could provide multiple ecosystem services by capturing excess storm water runoff (Turo & Gardiner, 2019), and providing habitat for arthropods. Others have suggested that shade trees in urban environments may be important in providing moisture for arthropods. McCluney & Sabo (2009) found that crickets increased the consumption of moist leaves to meet their water needs when moisture availability was low. Adding sources of shade and moisture may counteract the negative impacts reduced humidity has on arthropods.

In terms of temperature, many studies have documented the impacts urban warming has on Hymenoptera. Thermal tolerance is often used to explain changes in arthropod community structure (Youngsteadt et al., 2017; Hamblin et al., 2017; Miles et al., 2019), and species-specific responses (Diamond et al., 2017; Burdine & McCluney, 2019a) across urbanization gradients. For instance, Hamblin, Youngsteadt & Frank (2018) found sharp declines in bee abundance across an urbanization gradient, and temperature was the dominant variables in explaining these changes. Therefore, it is not surprising that we detected declines in Hymenoptera with temperature. We did identify a positive association between white flower abundance with Lasioglossum (sweat bees), but it is notable that overall floral resources availability had no influence on Hymenoptera. This suggests that simply planting more flowers may not impact Hymenoptera unless interventions addressing temperature are addressed.

Site structure

Impervious surface was a strong driver of arthropod community composition. In particular, we observed declines in Hemiptera (true bugs) with increasing impervious surface. In general, arthropod declines are common in cities (Ahrné et al. 2009, Fenoglio, Rossetti & Videla, 2020), as impervious surfaces replace natural vegetation cover. However, even at our most urban site (62.3% impervious surface) there is still available habitat for arthropods to utilize. Knop (2016) documented the broad homogenization that true bugs have experienced in cities, and the loss of host plants may underlie this homogenization. We did find that Enchenopa (tree hoppers) were positively associated with white flower abundance, but this was the only Hemiptera morphospecies associated with floral resources. In addition, a weak negative relationship between morphospecies richness and purple flower abundance was found. These results suggest that specific types of floral resources are required for some arthropod taxa, and that simply increasing overall floral resource availability may not be enough. Additional research is needed to investigate whether targeted plant additions at high impervious sites can counter the negative impacts of impervious surfaces, and how much space is required to make an impact.

Urban greenspaces are often small, and embedded within an urban matrix that promotes generalist arthropod species (Gaublomme et al., 2008). Small urban sites generally do not contain the core habitat needed to maintain diverse arthropods (Christie, Cassis & Hochuli, 2010), and may be more prone to edge effects. Studies have shown that urban greenspace patch size is a correlate of communities composition for ants (Uno & J. Cotton, 2010) and natural enemies (Burkman & Gardiner, 2014). Our results suggest that Hymenoptera are also influenced by site structure, as Hymenoptera abundance increased with perimeter-area ratios. Strategies to increase the size and connectivity of urban greenspaces may be important in conserving arthropods, particularly low-mobility arthropod species (Braaker, Ghazoul & Moretti, 2014). Sources of new greenspace habitat is available in vacant lots (Gardiner, Burkman & Prajzner, 2013), green roofs (Braaker, Ghazoul & Moretti, 2014), or backyard gardens (Pardee & Philpott, 2014), amongst others. Increasing the quantity and quality of urban greenspaces not only benefits arthropods, but may enhance the quality of arthropod-mediated ecosystem services like pollination and biological control (Sánchez Domínguez et al., 2020). As cities continue to grow and expand into the surrounding peri-urban environment, city planners should consider leaving patches of non-impervious habitat to maintain diverse arthropod assemblages in cities.

Future research should be conducted on the specific ecosystem services provided by individual taxa to better understand how urbanization influences the delivery of these services. There is also a strong need for manipulative studies to investigate whether the mitigation of urban effects (increasing greenspace size, adding shade trees, irrigation) has a positive impact on arthropod communities. Many of the abiotic conditions we documented could also be explained by loss of greenspace habitat, and manipulative studies would provide more precise conclusions. These types of studies would provide critical pieces of information that land managers and urban dwellers could implement on their properties. This study is limited in its ability to provide clear recommendations on the importance of flower species, and future studies to identify floral species in lieu of using floral colors as environmental factors should be necessary. As many arthropods utilized specific host plants, this information would produce a more applicable recommendation to land managers.