Data for: Promoting beneficial arthropods in urban agroecosystems: Focus on flowers, maybe not native plants
Data files
Jun 30, 2023 version files 20.20 KB
-
Philpott_etal_2023_Insects.csv
-
README.md
Abstract
Urbanization threatens biodiversity, yet urban native plants support native biodiversity, contributing to conservation and ecosystem services. Within urban agroecosystems, where non-native plants are abundant, native plants may boost the abundance and richness of beneficial arthropods. Nevertheless, current information focuses on pollinators, with little attention being paid to other beneficials, like natural enemies.
We examined how the species richness of native plants, garden management, and landscape composition influence the abundance and species richness of all, native, and non-native bees, ladybeetles, ants, and ground-foraging spiders in urban agroecosystems (i.e., urban community gardens) in California.
We found that native plants (~10% of species, but only ~2.5% of plant cover) had little influence on arthropods, with negative effects only on non-native spider richness, likely due to the low plant cover provided by native plants. Garden size boosted native and non-native bee abundance and richness and non-native spider richness; floral abundance boosted non-native spider abundance and native and non-native spider richness; and mulch cover and tree and shrub abundance boosted non-native spider richness. Natural habitat cover promoted non-native bee and native ant abundance, but fewer native ladybeetle species were observed.
While native plant richness may not strongly influence the abundance and richness of beneficial arthropods, other garden management features could be manipulated to promote the conservation of native organisms or ecosystem services provided by native and non-native organisms within urban agroecosystems.
Methods
Study System
We conducted our research in 19 gardens in the California Central Coast region in Monterey, Santa Clara, and Santa Cruz counties (Figure 1). All gardens are community gardens where vegetables, fruit trees, and ornamental plants are grown. The gardens are managed either collectively or in individual plots or allotments, range in size from 444 m2 to 15,400 m2, and had been cultivated for between 4 and 46 years at the earliest year of the study. The gardens are all managed using organic practices, and none allow chemical pesticides or herbicides to be used. None of the gardens contain large patches of grass, and none experience regular mowing. Each garden was separated from other garden sites by at least 2 km and up to 90 km.
Vegetation Sampling
In each garden, we sampled vegetation and ground cover as metrics of the local habitat management. At the center of each garden, we established a 20 m × 20 m plot within which we measured canopy cover with a concave vertical densiometer (at the center of the plot, and at each edge), counted and identified all trees and shrubs, and counted the number of trees and shrubs in flower. Within four randomly located 1 m × 1 m plots within the 20 m × 20 m plot, we sampled ground cover by visually estimating the percent cover of (a) bare ground, (b) grass, (c) herbaceous plants, (d) rocks/wood panels, (e) leaf litter, (f) mulch, and (g) concrete. In the same plots, we identified and estimated the percent cover of each herbaceous plant species, counted the number of flowers, and measured the height of the tallest herbaceous vegetation. We estimated the percent cover of grass, but did not identify grasses at the species level. Each of these vegetation and ground cover features were sampled six times in 2013 (between 17 May and 2 June, 18 and 24 June, 16 and 22 July, 12 and 21 August, 10 and 16 September, and 11 October and 19 November), six times in 2014 (between 17 and 20 June, 7 and 10 July, 27 and 30 July, 19 and 21 August, 8 and 10 September, and 29 September and 1 October), and six times in 2015 (between 13 and 15 May, 16 and 30 June, and 7 and 11 July, on 2 August, between 1 and 15 September, and between 21 September and 6 October) near to the dates that we sampled arthropods. Our dataset contained several plant morphospecies (<4% of total plant cover) that we were not able to identify at the species level, but these were not included in the analysis.
Native Plant Characterization and Abundance
We classified all herbaceous plant species identified in vegetation surveys according to whether they are native to California. Specifically, we categorized herbaceous plant species as native or not native based on information from four plant databases (e.g., Calflora (https://www.calflora.org/ (accessed on 10 November 2022)), USDA Plants (https://plants.sc.egov.usda.gov/java/ (accessed on 10 November 2022)), Jepson eFlora (https://ucjeps.berkeley.edu/eflora/ (accessed on 10 November 2022)), and the Missouri Botanical Garden (https://www.missouribotanicalgarden.org/plantfinder/plantfindersearch.aspx (accessed on 10 November 2022)). For species of herbaceous plants for which we found conflicting information, we first followed information from Calflora, and then from other databases. For any herbaceous plants for which we could not find information, we assumed them to be non-native to California. To determine the total number of native herbaceous plant species in each site for each year, we used the cumulative number of native herbaceous species encountered across all sample plots and all time periods for that site. To determine the total percent cover of native herbaceous plants, we summed all herbaceous plant cover (minus grass) from all sample points in a site in a year, and calculated the fraction of that cover comprising native herbaceous plant species.
Landscape Composition and Diversity
We used information from the 2011 National Land Cover Database (NLCD, 30 m resolution) to measure land cover composition within 2 km of each garden. We extracted data for all NLCD land cover classes, and combined classes to create four landscape variables for this study: (1) natural habitat (which combined deciduous, evergreen, and mixed forests, dwarf scrub, shrub/scrub, and grassland/herbaceous cover classes); (2) open habitat (which combined lawn grass, parks, and golf courses); (3) urban habitat (which combined low-, medium-, and high-intensity developed land); and (4) agricultural habitat (which combined pasture/hay and cultivated crops). We excluded land cover types that did not comprise more than 5% of the surrounding land cover for any garden (e.g., open water, wetlands; and barren land).
Arthropod Sampling
We sampled four groups of arthropods, namely bees (Apoidea), ladybeetles (Coleoptera: Coccinellidae), ants (Hymentopera: Formicidae), and spiders (Arachnida), and chose sampling methods recommended specifically for each of these groups. We sampled arthropods between May and October in 2013, 2014, and 2015, and visited gardens five or six times each summer. We surveyed bees in 2013 and 2015, ladybeetles in 2014 and 2015, and ants and ground-foraging spiders in 2013.
We sampled bees with elevated pan traps and aerial hand netting designed to specifically collect bees. We constructed pan traps using 400 mL yellow, white, and blue plastic bowls painted with Clear Neon Brand and Clear UV spray paint. We mounted traps on 1.2 m PVC pipes and filled bowls with a water (300 mL) and soap (4 mL) solution. We placed traps (one yellow, one white, one blue) 5 m apart from each other at the center of the 20 m × 20 m plot in each garden. We placed traps between 8 and 9 AM and collected them between 5 and 7 PM on the same day. We emptied traps into containers and transported them to the lab for sorting and pinning. We put out elevated pan traps six times in 2013 (29–31 May, 25–27 June, 23–25 July, 12–15 August, 17–20 September, 9–11 October) and four times in 2015 (6–13 April, 16–19 June, 8–10 July, and 11–14 August). We netted bees from flowers found within the 20 m × 20 m plot and within 20 m of the plot boundaries (for a total of a 60 m × 60 m plot) for 30 min on warm, sunny days. We collected bees six times during 2013 (17–22 May, 18–24 June, 16–22 July, 12–21 August, 10–11, 23 September, 11–15 October) and six times in 2015 (16–19 June, 7–10 July, 31 July–1 August, 11–14 August, 1–3 September, 21–24 September). We identified bees using online resources, image databases, books, and dichotomous keys, and by comparing specimens to bees in the Kenneth S. Norris Center for Natural History on the University of California, Santa Cruz campus. We identified all specimens to the highest taxonomic level possible, with more difficult groups identified to the morphospecies level.
We sampled ladybeetles in the 20 m × 20 m plots at the center of each garden six times in 2014 (17–20 June, 7–10 July, 27–30 July, 19–21 August, 8–10 September, and 29 September–1 October) and five times in 2015 (16–19 June, 7–10 July, 31 July–1 August, 11–14 August, 1–3 September, and 21–24 September). We used visual surveys and sticky traps. We visually surveyed and collected ladybeetles in eight randomly selected 0.5 m × 0.5 m plots within the 20 m × 20 m plots. Second, we placed four 7.62 cm × 12.7 cm yellow sticky strip traps (BioQuip Products Inc., Rancho Dominguez, CA, USA) on wire stakes placed in the ground next to vegetation at four random locations and collected them after 24 h. All lady beetles were identified at species level, or genus level when species identification was impossible (e.g., Scymnus sp. on sticky traps), using online resources and identification guides.
We sampled ground-foraging spider (hereafter spider) and ant activity density with pitfall traps placed for 72 h in each site four times in 2013 (20–23 May, 17–20 June, 15–18 July, 11–14 August, and 9–12 September). We constructed pitfall traps from 350 mL plastic tubs (11.4 cm diameter × 7.6 cm deep). We placed traps in the middle of the 20 m × 20 m plot in each garden arranged in two rows of three traps, with every trap separated from others by 5 m. We buried traps at the level of the soil surface, and filled traps with 200 mL of a saturated saline solution with a drop of unscented detergent to break the surface tension. We placed green plastic plates (7.62 cm diameter) 7–8 cm over each trap to prevent the capture of non-target taxa (e.g., flies) and to limit overflow for traps from overhead irrigation (Figure S1). Upon collection, we rinsed arthropods with water, separated them by order, and then stored insects in vials with 70% ethanol. Adult spiders in common families were identified at species level, and other individuals of other families were identified at the morphospecies level using several guides. Ants were identified at species level using an online guide to the ants of California.
All arthropods are stored at the Philpott Laboratory at the University of California, Santa Cruz. We used species richness in this paper to refer to the respective level of identification for each taxon (e.g., to species or morphospecies). We summed the number of individuals captured with any sampling method to determine the total abundance for bees and ladybeetles and the activity density of spiders. For ants, we used occurrence (presence of a species in a trap) instead of number of individuals as our abundance metric.
Native Arthropods Characterization
To classify sampled arthropods as native or non-native, we used several resources for each taxon. For bees, we used natural history information, distribution maps, and distribution information from three resources. For species that were only identified at the genus and morphospecies levels, we examined information and distribution maps for all species within each genus found in California. If one of the resources described a genus as restricted to or only found in California, we classified all morphospecies in that genus as native. If all species within a genus found in California were only distributed in N. America, we considered the species to be likely native. For genera for which certain species are distributed across the Holarctic region, or are not restricted to N. America, we considered the species to have an unknown distribution. Bee species were considered to be non-native only if this was confirmed in the literature. Ladybeetle species were classified as native or non-native following and sources therein, and two online resources. We classified spiders using three resources. For spiders identified at species level, we determined whether they were native or not native based on information in these resources. For spiders that were identified at family or morphospecies level within a family, we examined the ‘Pacific Coast Fauna’ sections. If all members of a family were described as native, we classified the species as native. If members of the family (and genera within that family) were described as having limited distributions restricted to California or the Western states, we classified the species as likely native. For spider families with large numbers of poorly known genera, or confirmed introduced and native species, we classified these species as unknown. We determined the native status of ants with information from two sources and references therein. (See publication for references)
Data Analysis
We used generalized linear models (GLM) or generalized linear mixed models (GLMM) to examine the relationships between the abundance and richness of bees, ladybeetles, ants, and spiders and the local and landscape characteristics of urban gardens. For the two groups sampled in two years (bees and ladybeetles), we used GLMM, and for the groups sampled in only one year (spiders and ants), we used GLM. For each arthropod taxon, we included six dependent variables: (a) the abundance (or number of occurrences) of all species, (b) abundance of native species, (c) abundance of non-native species, (d) species richness of all species, (e) species richness of native species, and (f) species richness of non-native species. For bees and spiders, where many organisms were only identified to the morphospecies level, we included two additional variables: (e) the abundance of native species plus likely native species, and (f) the richness of native species plus likely native species. Thus, we included both a conservative (only species confirmed to be native) and a more inclusive (species confirmed to be native plus species likely to be native) metric for native species abundance and richness in our sites. For our predictor variables, we chose six local scale variables (garden size, number of native plant species, mulch cover within 1 m, number of flowers, number of trees and shrubs) and one landscape variable (percent of natural land cover within 2 km) that represent the variation in garden ground cover and vegetation characteristics and landscape cover and that have been used in other urban garden studies. We only used vegetation data that corresponded to the same year in which arthropods were sampled (i.e., we used 2013 vegetation data as predictors of 2013 arthropod data). We used a variance inflation factor (VIF) cutoff of 3 to select predictor variables that were not highly correlated or collinear with each other. We used natural-log-transformed variables for garden size, number of native plant species, number of flowers, and number of trees and shrubs. Prior to running the models, we used the ‘DHARMa’ package in R to determine the best error distribution for each model based on visual assessment of QQ and standardized residual plots. We conducted all analyses in R statistical software version 4.2.2.
For GLMM, we used the ‘dredge’ function in the ‘MuMIn’ package version 1.42.1 to run all iterations of predictor variables, and ran model selection with the AIC scores to select the best models. If any models were within 2 AIC scores of the best model, we used the ‘model.avg’ function to average the top models. We used a Gaussian distribution for all models. For generalized linear models (GLMs), we tested all combinations of the six explanatory variables with the ‘glmulti’ function and selected the top model based on the AICc values. For all models where the AICc value was within two points of the next best model, we averaged them with the ‘model.avg’ function in the ‘MuMIn’ package and reported conditional averages for significant model factors. We used a Gaussian distribution for all spider abundance variables, and a Poisson distribution for all spider richness variables. For ants, we used a Poisson distribution for the number of ant occurrences, a negative binomial for the number of native ant occurrences, and a Gaussian distribution for both ant richness variables. We visualized all significant predictors of arthropod abundance and richness for both GLM and GLMM models with ‘visreg’ in R.
All results for the conservative and more inclusive metrics of native bee and spider abundance and species richness were qualitatively similar, and thus we only report the results for the conservative metric here. We also tested all models using observed species richness and estimated species richness with the Chao2 species richness estimator, and the results were qualitatively similar, so we report the output for observed species richness.