Spatiotemporal arthropod community dynamics in an agroecosystem

by

Patrick Burgess

A Thesis presented to The University of Guelph

In partial fulfilment of requirements for the degree of Master of Science in Integrative Biology

Guelph, Ontario, Canada

© Patrick Burgess, September, 2020

ABSTRACT

SPATIOTEMPORAL ARTHROPOD COMMUNITY DYNAMICS IN AN AGROECOSYSTEM

Patrick Burgess Advisor: University of Guelph, 2020 John M. Fryxell

Many arthropod communities are undergoing declines, with agricultural activity being a likely cause. However, we know little about what determines arthropod community composition in seasonal agroecosystems. Here I quantify how variation in climate, agricultural intensity, and local habitat along with spatial and temporal covariates impacts arthropod community composition in Southern Ontario, Canada. Local habitat factors, particularly canopy openness and plant community composition, had the strongest effect. Climatic variables followed closely, driven primarily by seasonal variation in temperature and humidity. Agricultural intensity had the smallest effect. Direct effects of spatial and temporal distances also occurred besides differences in environment. I discuss the possible roles of niche and neutral mechanisms in generating these patterns. My findings show that habitat restoration should be a priority and that seasonality plays a strong but underappreciated role in structuring arthropod communities

iii

ACKNOWLEDGEMENTS

I would first like to extend my thanks and gratitude to my advisor, John Fryxell, for providing me with guidance both throughout this M.Sc. degree as well as through the final year of my undergraduate degree. You have given me the freedom to develop and pursue my own ideas and interests, have always provided support when things seemed to be going terribly wrong, and have shown me what it means to conduct rigorous science. I would also like to thank the members of my advisory committee, Cortland Griswold, Dirk Steinke, and Andrew MacDougall for their valuable input. An additional thanks to Cortland Griswold for all the time he spent showing me the way through the world of theoretical ecology. I am very grateful for the researchers at the Centre for Biodiversity Genomics at the Biodiversity Institute of Ontario, who coordinated fieldwork for collecting arthropod samples as well as processing them and generating the metabarcoding data. I also want to thank all the landowners who allowed us to conduct research on their lands. I sincerely hope that my work will contribute to the understanding and conservation of the biological communities found in these wonderful places.

None of my work would have been possible without the team of dedicated field technicians and volunteers including Ashley Cholewka, Caroline Reisiger, Emma Wright, Jillian McGroarty, and Eleish Miller. Thank you all for making field work fun in even the most unpleasant conditions. I would also like to thank my lab and officemates including Rebecca Viejou, Xueqi Wang, Nicola Shaw, Gustavo Betini, Samantha Shaw-McDonald, Rachel Holub, and Eleish Miller for all the laughs and guidance. You all helped me through difficult points of conducting this research and I am grateful for the friendships we have formed. Lastly, I would like to thank my friends and family, particularly my parents Blair and Lynn, for their support throughout this degree. None of it would be possible without you.

iv

TABLE OF CONTENTS

Abstract ...... ii

Acknowledgements ...... iii

Table of Contents ...... iv

List of Tables ...... v

List of Figures ...... vi

Introduction ...... 1

Methods...... 7

Results ...... 15

Discussion ...... 18

References ...... 29

Tables ...... 36

Figures...... 38

Appendix ...... 47

v

LIST OF TABLES

Table 1: Environmental variables used in path analysis and their groupings into either local habitat, agricultural intensity, or climatic variables ...... 36

Table 2: Significant results from path analysis and associated standardized path coefficients .... 37

Table A1: Temporal alignment of plant and arthropod surveys by trap ...... 47

Table A2: Full list of plant taxa from all surveys by plant type ...... 52

vi

LIST OF FIGURES

Figure 1: Distribution of arthropod BINs for the top 6 orders across all samples ...... 38

Figure 2: Total arthropod BIN richness among traps pooled across all time periods...... 39

Figure 3: Total arthropod BIN richness among sample periods pooled across all traps ...... 40

Figure 4: Non-metric multidimensional scaling (NMDS) ordination of arthropod species composition based on pairwise Sørensen dissimilarities among all samples ...... 41

Figure 5: Path analysis of factors influencing spatiotemporal arthropod Sørensen dissimilarities...... 42

Figure 6: Path analysis of factors influencing spatiotemporal arthropod Sørensen dissimilarities aggregated by variable type...... 43

Figure 7: Spatial distance-decay of arthropod community composition based on pairwise Sørensen dissimilarities among samples after filtering the arthropod data to match the plant data ...... 44

Figure 8: Temporal distance-decay of arthropod community composition based on pairwise Sørensen dissimilarities among all samples after filtering the arthropod data to match the plant data ...... 45

Figure 9: Relationship between pairwise Sørensen dissimilarities of arthropod communities and Bray-Curtis dissimilarities of plant communities for all samples after filtering the arthropod data to match the plant data ...... 46

Figure A1: Map of Southern Ontario, Canada showing the location of traps at study sites and their respective management types ...... 58

Figure A2: Boxplots of (a) the percentage of agricultural area, (b) the percentage urban area, and (c) the percentage semi-natural area within a 2000m radius of each trap for each locality type .. 59

INTRODUCTION Recent studies indicate that many arthropod communities around the world are undergoing strong declines in biomass (Hallmann et al 2017), abundance, and species richness

(Sánchez-Bayo and Wyckhuys 2019; Seibold et al., 2019). The causal links behind these declines are not resolved, but sites in landscapes with a higher proportion of agricultural land have experienced steeper declines (Seibold et al. 2019), and Sánchez-Bayo and Wyckhuys (2019) suggest that habitat loss from conversion to agriculture and urban land and the use of pesticides are likely the main factors behind these declines. However, we largely lack empirical evidence to substantiate these claims and thus it is essential that we better understand the effect of agriculture on arthropod communities.

There is immense pressure to maximize food production in order to feed our current global population of 7.8 billion people. This will only increase as the global population rises to a predicted 10 billion by 2050 (Tilman, Balzer, Hill, and Befort, 2011; Gerland et al., 2014).

Agricultural systems have expanded and intensified in response, resulting in the dedication of vast expanses of land to monoculture crop growth and livestock grazing (Tilman, Balzer, Hill, and Befort, 2011). About 40% of land globally is currently dedicated to food production (Fahrig et al., 2011). Such extensive agricultural production often results in the removal and degradation of habitat, the fragmentation and homogenization of the landscape, and the pollution of local ecosystems through fertilizer and pesticide use (Fahrig et al., 2011; Tilman, Balzer, Hill, and

Befort, 2011).

Arthropods are one of the most taxonomically and functionally diverse groups of animals on earth. They serve as vital links in food webs and play a particularly important role, providing ecosystem services including pollination, nutrient cycling, and pest control (Kremen et al., 1993;

1

Stork, 2018). As such, there is a strong need to identify and implement strategies that can effectively feed a growing human population while mitigating the negative impacts of agriculture on arthropod communities.

Several strategies could be used to attain this goal. Reducing pesticide and fertilizer loads can promote diversity in fields and reduce impacts on adjacent ecosystems (Tilman et al., 2002).

However, reducing the degree of chemical treatment can be difficult to implement due to losses in yield (Tilman et al., 2002). One alternative that both reduces the need for pesticide use and promotes diversity is a habitat management approach (Tilman et al., 2002; Fahrig et al., 2011).

Under this framework, agricultural systems are managed to retain local and landscape-level habitat heterogeneity such that a wider array of organismal needs can be met (Fahrig et al.,

2011). This may entail planting several crop types within or between fields, establishing natural areas adjacent to or within cropland, and combinations thereof (Lichtenberg et al., 2017).

Polyculture systems that contain several crop and non-crop species host greater diversity than monocultures by providing limiting resources and are often superior for maintaining ecosystem function and providing ecosystem services (Thrupp, 2004; Fahrig et al., 2011).

Some success has been found in agri-environment programs that restore native habitat for the provisioning of ecosystem services. Several European studies have shown increased arthropod diversity and abundance due to habitat restoration, often with positive effects on herbivore suppression as well as pollination and predation services (Albrecht et al., 2010;

Haaland, Naisbit, and Louis-Félix, 2011). Similar results have been found on farms that work with Alternative Land Use Services (ALUS) Canada, a program that restores forest, wetland, and prairie habitat on marginal lands (ALUS Canada, 2020). Paterson, Cottenie, and MacDougall

2

(2019) and Dolezal (2019) found increased bee and general arthropod abundance and diversity, respectively, mainly due to restored prairie habitat on marginal lands in both cases.

The management of agricultural landscapes not only affects species richness and abundance, but community composition as well. Increasing the diversity and complexity of local and landscape-level habitat can support taxonomically and functionally distinct communities, increasing the total number of species regionally, the acquisition of ecosystem services, and the spatiotemporal stability of resources required by natural communities (Tscharntke et al., 2012).

The quantity, quality, and the spatial arrangement of habitat can all simultaneously impact the composition of arthropod communities (Matson, Parton, Power, and Swift, 1997; Hendrickx et al., 2009; Fahrig et al., 2011). Specifically, variation in arthropod community composition is often linked to local plant resources such as biomass, structural complexity, and community composition (Stinson and Brown, 1983; Schaffers, Raemakers, Sýkora, and ter Braak, 2008;

Prather and Kaspari, 2019) as well as measures of habitat diversity, land-use intensity, landscape connectivity, and the configurational complexity of the landscape (Hendrickx et al., 2007;

Fahrig et al., 2011).

Most of these processes have been investigated primarily from a spatial perspective, but temporal factors can play a central role as well. Seasonality is particularly important in many systems, stemming from an interplay of species-specific responses to abiotic conditions, such as temperature and precipitation, biotic conditions, such as plant resource availability, and stochastic variation through time (Stinson and Brown, 1983; Wolda, 1988; Grøtan et al., 2012;

Hatosy et al., 2013; Shimadzu, Dornelas, Henderson, and Magurran., 2013). Studies in agroecosystems have found that the effect of habitat composition and the configuration of those habitats in the landscape on terrestrial predatory arthropod communities varies across the

3

growing season (Maisonhaute, Peres-Neto, and Lucas, 2015; Bertrand, Baudry, and Burel, 2016) and that landscape composition modulates phenological diversity of bee communities throughout the growing season (Sydenham, Eldegard, and Totland, 2014). Even in tropical systems, arthropod community composition in both natural and rubber forests shows high seasonal turnover (Beng, Corlett, and Tomlinson 2018). However, very few studies on arthropod communities in agroecosystems incorporate seasonality, suggesting that seasonality could play an under-appreciated role in determining the composition of these communities. Seasonality in agroecosystems can unfold by climatic seasonality and associated management-based seasonality of factors such as plowing, planting, and pesticide application. Neither form of seasonality, nor their interaction, is well understood in terms of impacts on arthropods.

From a theoretical perspective, variation in community composition in time and space has been described by two main streams of thought: niche and neutral theories. Under the niche framework, community composition is the result of species-specific responses to environmental variation and interspecific interactions (Vandermeer, 1972). Coexistence between species is promoted if environmental conditions are favourable and if the strength of intraspecific competition exceeds that of interspecific competition, which can be achieved by partitioning limiting resources (Vandermeer, 1972). Under the neutral framework, individuals of all species within a trophic level are demographically and competitively identical. Community composition results from the interaction between stochastic birth and death events (ecological drift) and limited dispersal between localities, which generates spatial and temporal turnover independently of environmental conditions (Bell 2000, 2001; Hubbell, 2001). Niche theory describes the role that deterministic processes play in structuring communities, whereas neutral theory describes

4

the role that stochastic processes play in structuring communities (Vandermeer, 1972; Hubbell,

2001).

In the context of agroecosystems, niche and distance-based neutral processes can be expected to interact significantly but are rarely quantified. Under a niche framework, differences in environmental conditions should be the primary driver of differences in community composition. However, wide compositional differences may unfold simply by stochastic-based dispersal limitation that restricts movement, an effect exacerbated by patch isolation and especially affecting arthropod groups with limited traits for dispersal. One characteristically neutral phenomenon resulting from spatial turnover by distance is unoccupied suitable habitat, where niche-based environmental responses are muted or suppressed simply because species cannot reach areas where their fitness may be maximized. In this case it is pure distance, not environmental differences, that regulates distribution, abundance, and organismal fitness (Bell

2000, 2001; Hubbell, 2001).

There have been very few studies investigating the relative importance of local habitat factors, agricultural intensity, and climate in structuring arthropod communities in agroecosystems generally and even fewer that explicitly incorporate both spatial and temporal variation. This means that while stories of an “insect apocalypse” dominate the media

(https://www.nytimes.com/2018/11/27/magazine/insect-apocalypse.html), we lack a clear mechanistic understanding of how arthropods are regulated on contemporary landscapes, especially agricultural systems. Such research not only would test fundamental regulating processes but have important implications for understanding how habitat management techniques could be used to promote diversity in seasonal agroecosystems. In this thesis, I use a suite of environmental variables to quantify how spatiotemporal variation in climate, agricultural

5

intensity, and local habitat impact the composition of arthropod communities. I additionally quantify direct and indirect effects of community size, species richness, spatial distance, and temporal distance on arthropod community composition and discuss the roles that niche and neutral mechanisms may play in generating these patterns.

I hypothesize that arthropod community composition is affected by (1) climatic variation due to physiological limitations (i.e. ectothermy and desiccation risk) as well as through influences on dispersal, (2) local habitat due to plant resource limitation and species-specific plant resource requirements, and (3) agricultural intensity due to species-specific plant resource requirements and/or uneven risk of anthropogenic mortality among species. In each case I predict a positive relationship between arthropod community dissimilarity and environmental distances.

That is, larger differences in environmental conditions should generate more dissimilar communities for each variable representing either climate, local habitat, or agricultural intensity.

Following the framework of Jabot et al. (2020), I incorporate spatial and temporal distances into the analysis both as direct effects on arthropod community composition to quantify spatial and temporal variation in community composition explicitly, which tie to neutral processes such as dispersal limitation and demographic stochasticity, as well as indirect effects that act through environmental variation.

Community size can be a determinant of community composition and species richness by limiting population sizes of constituent species and thus increasing the importance of demographic stochasticity (Gilbert and Levine, 2017). Species richness can influence dissimilarities through nestedness, where less species-rich communities may be composed of a subset of species that are in the more species-rich communities (Baselga, 2010; Jabot et al.,

2020). Therefore I also expect to see: a negative relationship between arthropod community

6

dissimilarity and mean community size (here the wet weight of each malaise trap bottle is used as a proxy for community size), a positive relationship between differences in community size and differences in species richness, and a positive relationship between arthropod community dissimilarity and differences in species richness.

METHODS

Arthropod sampling and DNA metabarcoding

In late April/early May 2019, 29 Townes style malaise traps were placed at 15 sites in

Southern Ontario, Canada (Figure A1). Malaise traps are well suited for large-scale monitoring as they are easily standardized, time and cost effective, and sample a wide array of arthropod taxa, though they preferentially trap flying insects (deWaard et al., 2018; D’Souza and Hebert,

2018). The placement of malaise traps on the sites represented four broad habitat types in varying proportions: woodland, grassland/meadow, aquatic edge, and crop edge. Typically, traps were placed on the edge of two of these habitat types. Sites varied widely in agricultural intensity, including conventional farms, conventional farms with a higher proportion of natural land (mid-impact), farms with restored habitat on their marginal lands in cooperation with

Alternative Land Use Services Canada (ALUS), and conservation areas.

Arthropods were collected in 500mL plastic bottles filled with 95% ethanol attached to the trap heads. The bottles were collected and replaced on a biweekly schedule until mid October

2019. With a few exceptions (damaged samples or early trap takedown), a total of 12 two-week samples were collected at each trap. All of the collected samples were accessioned and stored at

Centre for Biodiversity Genomics (http://www.biodiversitygenomics.net/). Every other two-

7

week sample was sent to the Centre for Biodiversity Genomics for metabarcoding at their barcoding facility (http://ccdb.ca/).

The bottles that were sent for metabarcoding first had their ethanol filtered off using a sterile Microfunnel 0.2 uM Supor membrane filter (Pall Laboratory) using a 6-Funnel Manifold

(Pall Laboratory). They were then weighed to measure wet arthropod biomass. DNA extractions followed a membrane-based protocol (Ivanova, deWaard, and Hebert, 2006). Wet biomass was used to determine the amount of arthropod lysis buffer to be added to the bottle. After adding lysis buffer, each bottle was incubated gently on a shaker at 56°C overnight. After DNA lysis, 8 technical replicates (repeated samples of lysate from the same bottle) were taken per bottle from different spatial locations within the bottles in 300 μl aliquots. 50 μl of each aliquot were arranged on 96 well microplates along with 8 negative (no DNA) and 8 positive (known community DNA sample: public dataset at http://dx.doi.org/10.5883/DS-AGAKS) controls per block, each of which was purified using 3.0 μm Pall Supor Membrane glass fibre plates (Pall

Laboratory). 100 μl of binding mix was added to the lysate, which was transferred to a column plate and centrifuged for 5 minutes at 5000g. DNA was then purified with three wash steps as follows: 180 μl of protein wash buffer centrifuged for 2 minutes at 5000g followed by two washes with 600 μl of wash buffer centrifuged for 5 minutes at 5000g. The filter plate was then transferred on to a sterile 96-well microplate and incubated at 56°C for 30 minutes. DNA elution was carried out by adding 60 μl of 10 mM Tris-HCl pH 8.0 followed by centrifugation for 5 minutes at 5000g.

A 462 base-pair amplicon of cytochrome c oxidase subunit I (COI) was amplified from two rounds of PCR amplification using the forward primer AncientLepF3 (Prosser, deWaard,

Miller, and Hebert, 2016) and the reverse primer cocktail C_LEPFoIR (containing LepR1 and

8

HCO2198) (Hebert et al. 2004). The reagents included in each reaction were: 1.25 μl of 10x

Platinum Taq reaction buffer (Invitrogen), 6.25 μl of 10% trehalose (Fluka Analytical), 0.625 μl of 50 mM MgCl2, 0.0625 μl of 10 mM dNTPs (KAPA biosystems), 0.125 μl of each primer (1

μM), 0.06 μl of Platinum Taq (5 U/μl), 2 μl of DNA extract, and 2μl of Hyclone ultra-pure water

(Thermo Scientific). All PCR reactions were carried out with the following thermocycler settings: initial denaturation at 94C for 2 minutes, 20 cycles of denaturation at 94C for 40 seconds, annealing at 51C for 1 minute, extension at 72C for 1 minute, and lastly extension for

5 minutes at 72C. First round PCR products were diluted by 2x before the second round of

PCR. Platform-specific unique molecular identifiers (UMIs) were attached using fusion primers along with sequencing adaptors that are required for S5 libraries. Labelled PCR products were pooled, standardized to 1 ng/μl, and sequence libraries were prepared on the Ion Chef™ system

(Thermo Fisher Scientific) with 530 Chips according to manufacturer instructions.

Read libraries from all replicates were uploaded to the mBRAVE platform

(http://www.mbrave.net/). Sequences were only retained if they had a minimum average quality value (QV) of 20, a minimum length of 350bp, less than 25% of bases with a QV of less than 20, and less than 5% of bases with a QV of less than 10. Reads were trimmed 30 bp at the front with a trim length of 450 bp. Reads were queried against mBRAVE reference libraries for chordates, insects, non-insect arthropods, non-arthropod invertebrates, and bacteria. Any reads that diverged from sequences in these reference libraries by greater than 3% were subsequently clustered with an operational taxonomic unit (OTU) threshold of 1% and a minimum of 5 reads per cluster.

Reads were assigned to a Barcode Index Number (BIN) that serves as a species proxy

(Ratnasingham and Hebert, 2013). The BIN system uses the Refined Single Linkage (RESL) algorithm to designate OTUs. RESL assigns OTUs in three steps: (1) a hidden Markov model of

9

the COI protein is used to align input sequences, (2) single linkage clustering is performed, where RESL performs distance and clustering stages concurrently based on a given sequence threshold, (3) clusters are refined using Markov clustering to verify and further delineate OTU assignment when sequence variation is found within clusters (Ratnasingham and Hebert, 2013).

OTUs are then assigned to a BIN, which is a unique identifier that is associated with specimens and their metadata on the barcode of life database (BOLD) (Ratnasingham and Hebert, 2013).

Each OTU can be either associated with an existing BIN, in which case it will add to the metadata of that BIN, or a new BIN entry will be created (Ratnasingham and Hebert, 2013).

BINs are assigned two identifiers: a URI on BOLD and a DOI. Thus, BIN assignments are dynamic and depend on the continual updating of sequence information. The taxonomy reported in this study is current as of November 2019. During the denoising process, BINs were discarded under the following circumstances: (1) they had less than 5 sequence reads summed across all technical replicates, (2) their read count was less than the mean read count for the run in at least

75% of the technical replicates, or (3) their read count was less than 1% of the maximum read count for the run with less than 10 total reads. BINs that showed up in negative controls would have been removed in this process and if the noise could not be removed through these steps, the run was excluded and/or rerun. Only arthropods and non-arthropod invertebrates were included in the final BIN table, though arthropods constituted 99.8% of these BINs.

Plant surveys

Plant surveys of ground vegetation were conducted monthly on a four-week rotating schedule between May and September at 29 traps on 15 sites (Table A1; Figure A1). Sites were selected to maximize the representation of habitat and management types while working within logistical constraints. Two plant survey techniques were used over the course of the sampling

10

period. For the first 3 weeks, 5 plots measuring 1x1m were randomly placed on each side within

25 m of the malaise trap for a total of 10 plots per trap. From week 4 onward, two 25 m transects were placed perpendicularly to each side of trap, or as close to perpendicular as possible if there were large waterbodies next to the trap, and 1x1m plots were placed every 5m along the transects.

The identity of each plant (at the finest taxonomic resolution possible), its flowering status, percent cover, maximum herbaceous vegetation height, visual obstruction, plant biomass, and overhead canopy openness were measured in each plot. Visual obstruction was measured using a 2m tall visual obstruction pole with 10cm bands of alternating colours viewed from 4 m away at a height of 1 m from the ground (Toledo, Abbott, and Herrick, 2008). Visual obstruction values were averaged from 4 viewing points perpendicular to each side of each plot. Plant biomass was estimated using the point-intercept method (Frank and McNaughton, 1990), where a thin rod (2 m height, 3 mm diameter) was placed vertically at 10 points in each plot and plant- rod contacts were counted on a per-species basis. Overhead canopy openness was measured using a convex spherical densiometer, averaged from 4 points perpendicular to each side of each plot.

Given some uncertainty about field identification of closely related forbs and grasses that were not in flower, all plant data were analysed at the genus level or higher. Only 361 of the total

12,894 individual observations in the plant dataset could not be identified to any taxonomic level, representing less than 1%. The greatest level of uncertainty was with the identification of

C3 pasture grasses (such as species in the genera Poa or Festuca), and so some of these grasses were classified into the classical tribe Festuceae, although updated nomenclature would place

11

these into the tribe Poeae. (Smith, 2017). All plants were additionally classified into the groups: forb, C4 grass, C3 grass, sedges and rushes, woody plant, crop, or pteridophyte.

Looking at general patterns of Sørensen beta diversity among time periods and traps, the full dataset was used (29 traps and 6 time periods). Since the temporal extent of the arthropod and plant data differed slightly (6 months and 5 months, respectively), the arthropod data was filtered to the samples that were closest in time to the plant data for all analyses that involve environmental effects. This meant that either the first arthropod sample or the last arthropod sample was omitted, depending on the plant survey schedule (Table A1).

Climatic data

Weather data were sourced from the Government of Canada Historical Weather Database

(https://climate.weather.gc.ca/) from four weather stations based on the nearest distance from the sampling sites as well as from temperature loggers that were attached to each malaise trap. The loggers recorded temperature (℃) hourly throughout the entire sampling period for each trap.

Five loggers malfunctioned. In these cases, temperature data were taken from the other trap on the same farm (4 traps) or from the nearest site (1 trap). Hourly relative humidity (%), hourly wind direction (10s deg), and hourly wind speed (km/h) were obtained from all weather stations.

Total daily precipitation (mm) was only available for three stations. All variables were averaged, and the coefficient of variation was calculated for the temperature data to match the two-week sampling periods of the nearest traps throughout the season.

Landcover data

Landcover data were obtained from the 2018 Annual Crop Inventory, which classifies landcover types from satellite images at a 30m spatial resolution using decision tree algorithms

(Agriculture and Agri-Food Canada, 2018). All landcover types were reclassified into cropland, 12

semi-natural, and urban categories prior to analysis. Since the percentages of seminatural and agricultural land were highly correlated, and since I was primarily interested in the effects of agriculture, only percentage of the landscape that is agricultural within a 2000 radius was used in the analysis. This scale better represents landscape-level processes and strong effects of landscape factors on arthropods at 1000-2000 m scales have been observed in the literature

(Watson, Wolf, and Ascher 2011; Gámez-Virués et al., 2015; Siebold et al., 2019). These metrics were calculated using the landscapemetrics R package (Hesselbarth, Sciaini, With, Wiegand, and

Nowosad, 2019).

Statistical methods

To explore general patterns of beta diversity, a dissimilarity approach based on Sørensen dissimilarity was taken where arthropods were grouped by trap across all time periods, by time period across all traps, and for all samples. In the first case, there were 29 units to compare with each other (trap), in the second there were 6 (time periods) and in the third there were 144

(samples; each trap by time combination). In addition, non-metric multidimensional scaling

(NMDS) based on Sørensen dissimilarities among all samples was used to look at the main gradients of variation in arthropod community composition, which was conducted using the R package vegan (Oskanen et al., 2019). Sørensen dissimilarity is a metric that uses presence- absence data to quantify how different two samples are in species composition and is calculated

2푎 as: 훽 = 1 − (Sørensen, 1948; Koleff at al., 2003) where a represents the number of 푠ø푟 2푎+푏+푐 species shared between two samples and b and c represent the number of species unique to each sample. A value of one indicates that no species are shared, while a value of zero indicates that all species are shared.

13

To test my hypotheses, I conducted a distance-based path analysis according to the framework of Jabot et al. (2020), which uses the relative effects of environmental distances, temporal distance, spatial distance, differences in community size (arthropod biomass was used here as a proxy), and differences in richness on Sørensen pairwise dissimilarities between individual samples (trap by time combinations) to infer dominant assembly mechanisms. Highly correlated environmental distances were removed prior to analysis (r > 0.7). The structure of the path model was as follows: arthropod Sørensen dissimilarities between samples were linked to

19 environmental distance variables (including dummy variables). Ten represented local habitat, three represented agricultural intensity, and six represented climatic variation (see Table 1 for a full list). Arthropod Sørensen dissimilarities were directly linked to spatial and temporal distances, and each environmental distance was also linked with spatial and/or temporal distance depending on whether they showed spatial variation, temporal variation, or both (Table 1).

Differences in species richness between samples were linked to arthropod Sørensen dissimilarities, which in turn was linked to differences in community size. Lastly, arthropod

Sørensen dissimilarities were linked to average community size between samples. Environmental distances were calculated as Euclidian distances except for differences in plant communities, which were calculated as Bray-Curtis dissimilarities (an abundance-based extension of the

Sørensen index) of the plant cover data. Spatial distances were calculated as the distance in meters between individual traps using the geodist R package (Padgham and Sumner, 2020), and temporal distances were calculated as the Euclidian distance between sampling periods.

The importance of each significant path in the model was assessed based on standardized path coefficients (SPC) and model fit was based on Standardized Root Mean Square Residual values (SRMR). SRMR is a measure of the absolute difference between observed and predicted

14

covariance matrices and as such values closer to zero indicate better fit (Cangur and Ercan,

2015). Typically, this measure is interpreted with values <0.10 indicating acceptable fit and values <0.05 indicating good fit (Cangur and Ercan, 2015). The significance of parameters was determined using the permutation method of Fourtune et al. (2018) to account for non- independence between dissimilarity values and the Benjamini-Hochberg procedure was used to correct P-values for multiple comparisons (Benjamini and Hochberg, 1995; Jabot et al. 2020).

Values of paths to and from groups of environmental distances were calculated by summing the absolute values of significant standardized path coefficients of individual environmental variables (Jabot et al. 2020). Significant paths from temporal and spatial distances to environmental distances were only visualized for environmental variables that also had significant effect on arthropod community dissimilarity. Thus, direct paths from temporal and spatial distances to arthropod community dissimilarity represent spatial and temporal effects that are not associated with the measured environmental variables while paths from temporal and spatial distances that connect to environmental distances give an indication of the spatiotemporal structure of the important environmental variables. All aspects of model fitting were conducted by modifying scripts provided in the supplementary material of Jabot et al. (2020). The R package lavaan (Rosseel, 2012) was used to fit the model, and the R package MASS (Venables and Ripley, 2002) was used in the permutation procedure. All analyses were carried out using R statistical software version 3.6.2 (R Core Team, 2020) at a significance level of α = 0.05.

RESULTS Prior to temporally matching the arthropod and plant samples, a total of 172 Malaise trap samples were available for analysis (two traps were missing one sample). In total, 10793 BINs were identified, representing 36 orders and 438 families. 53% of BINs belonged to the order

15

diptera, 17% to hymenoptera, 10% to lepidoptera, 6% to coleoptera, 6% to hemiptera, and 2% to aranae (Figure 1). As expected with malaise sampling, these orders except aranae are generally mobile flyers. The number of BINs varied strongly among samples (mean = 346.88, sd=150.79).

There also was substantial variation in the number of BINs identified among traps and time periods. The greatest number of BINs identified in a single trap was 1851, the lowest was 983, and the average was 1335 (Figure 2). The most diverse sampling period was July, with a total of

5059 BINs, and the least diverse sampling period was May, with a total of 1761 BINs. The average number of BINs per sampling period was 3748 (Figure 3). 61% of BINs represented singleton occurrences. After temporally matching the arthropod and plant samples, a total of 144 samples were available for analysis. 10359 BINs were identified representing 34 orders and 428 families. The taxonomic distribution for the top 6 orders remained the same as in the full dataset.

In total, 229 plant genera were identified, though this is a conservative estimate of true richness as some plants were categorized at higher taxonomic levels. There were 48 plants identified at each trap on average, though this tended to be quite variable (sd = 12.12). Pooled across all traps, an average of 182 plants were identified with relatively small variation between time periods (sd = 12.31). Of the plants that were identified, most were forbs (61%), followed by woody plants (20%), C3 grasses (7%), pteridophytes (4%), C4 grasses (3%), crops (3%), and sedges and rushes (2%). Using mean percent cover as an abundance measure, Bray-Curtis dissimilarities (where a value of 1 would indicate that two samples share no species) of plant communities among traps were very high (mean = 0.81, sd = 0.11). Far less variation was observed among sampling periods (mean = 0.24, sd = 0.1). Dissimilarity tended to be highest among individual samples (mean = 0.84, sd = 0.13).

16

Sørensen dissimilarity in arthropod community composition among traps across all time periods was very high (mean = 0.73, sd = 0.09), as was dissimilarity among time periods across all traps (mean = 0.61, sd = 0.12). The highest pairwise dissimilarity among time periods was

0.80, comparing May and August, and the lowest was 0.44, comparing July and August. Overall, dissimilarity tended to highest when comparing individual samples (mean = 0.87, sd = 0.07).

Two general patterns were discernable from the NMDS ordination (Figure 4). First and most notably, samples collected in May clustered together away from all other samples regardless of the trap location. Second, for collection months beyond May, samples clustered into two groups based on the traps they were collected from, likely reflecting local habitat conditions. This shows that species composition was relatively homogenous across the landscape in May, but communities differentiated from each other as the growing season progressed.

The square-root mean residual value (SRMR) of the path model was 0.086, indicating an acceptable absolute model fit (Cangur and Ercan, 2015). The total R2 value for the effect of all variables on arthropod dissimilarity was 0.55. There were significant effects of environmental distances, temporal distance (SPC = 0.23), and spatial distance (SPC = 0.07) on arthropod community dissimilarity (Table 2, Figures 5 – 8), though the effect of spatial distance was relatively weak. Traps were separated by 48,551 metres on average (range = 71.2 – 142,343, sd =

30,533) and traps on the same farm were separated by 371 metres on average.

There was a significant positive effect of differences in arthropod species richness (SPC

= 0.21), and a significant negative effect of mean community size on arthropod community dissimilarity (SPC = -0.24). Differences in arthropod richness were also positively related to differences in arthropod community biomass (SPC = 0.23), indicating that effects of richness on arthropod composition may have been mediated by community size. Spatial and temporal

17

distances were both significantly related to environmental distances, demonstrating that environmental variables were both spatially and temporally structured.

Of the 19 environmental variables considered, 10 had a significant effect on arthropod community composition after Benjamini-Hochberg correction (Figure 5, Table 2; see Figure 9 for example). In order of importance these were: a positive effect of changes in tree canopy openness (SPC = 0.32), a positive effect of plant community dissimilarity (SPC = 0.25), a positive effect of changes in the coefficient of variation in temperature (SPC = 0.20), a positive effect of changes in temperature (SPC = 0.17), a positive effect of conservation areas relative to conventional farms (SPC = 0.14), a positive effect of changes in relative humidity (SPC = 0.08), a negative effect of forest habitat relative to other habitats (SPC = -0.07), a negative effect of crop habitat relative to other habitats (SPC = -0.04), a negative effect of ALUS farms relative to conventional farms (SPC = -0.05), and a positive effect of water habitat relative to other habitats

(SPC = 0.02). Breaking the environmental effects into variable groups of climate, local habitat, and agricultural intensity (Figure 6), local habitat variables had the strongest effect (Σ|SPC| =

0.71), followed by climatic variables (Σ|SPC| = 0.45) and agricultural intensity (Σ|SPC| = 0.19).

Local habitat and agricultural intensity both showed spatial structure (Σ|SPC| = 0.67 and 0.52, respectively), local habitat showed weak temporal structure (Σ|SPC| = 0.04), and climatic variables only showed temporal structure (Σ|SPC| = 1.05).

DISCUSSION

I found simultaneous support for the hypotheses that arthropod community composition is determined by local habitat, climate, and, to a lesser degree, agricultural intensity, with the relationships in path analysis generally supporting my predictions.

18

Local habitat factors played the largest role in determining arthropod community composition, which showed much stronger spatial than temporal structure, although both had detectable importance. This relationship was driven largely by features of local plant communities, particularly tree canopy openness and plant community composition. These effects are striking given that the landscapes of Southern Ontario that I worked on are almost entirely composed of agricultural, urban, or marginal land, totalling ~92% crop cover (Dolezal, 2019). It thus could have been expected that such local variation in habitat would be overwhelmed by the regional dominance of habitat that may be unsuitable for many species. However, my findings agree with Scheper et al. (2013) who found that agri-environment measures such as organic farming, the planting of wildflower strips, or the natural restoration of field margins were most effective for increasing pollinator richness and abundance in structurally simple landscapes.

More generally, my results confirm a relationship that that is often found between arthropod and plant communities, centering on the fact that the former is often strongly specialized on the latter. Arthropod communities have been shown to respond strongly to measures of plant structure (Stison and Brown, 1983; Brose, 2003), plant diversity (Borer, Seabloom, and Tilman,

2012; Ebeling et al., 2018), and plant community composition (Schaffers, Raemakers, Sýkora, and ter Braak, 2008). These effects could be due to species-specific preferences for food (either directly on plants or other organisms that depend on those plants), nesting, shelter, and mating resources, or because plants also act as an effective measure for other environmental gradients such as light availability or soil type (Schaffers, Raemakers, Sýkora, and ter Braak, 2008). From a management perspective, this result suggests that maintaining several habitat types with diverse and compositionally distinct plant communities on local scales is likely to be effective for supporting arthropod diversity in agroecosystems.

19

Variation in climatic conditions played the second most important role in determining arthropod community composition, which path analysis showed was very strongly temporally structured. This could be explained by several mechanisms. The first is that climate has a direct effect on arthropod survival and reproduction. Arthropods are generally constrained to narrow optimum ranges of temperature and humidity and taxa differ widely in their tolerances for climatic conditions, with some species being specialized for early emergence (Høye and

Forchhammer, 2008). Such differences in the phenology of emergence due to climatic conditions could result in compositional turnover throughout the season. It could also partially explain why strong differences were observed with forest canopy, as compositional turnover of insects between shaded cool forest and warmer and often drier herbaceous plant communities tends to be high (Dolezal, 2019). A related explanation for the effects of climatic variation could be resource limitation. Many arthropods depend on specific feeding and nesting resources, and many of those resources are not available early in the season due to plant phenology in the case of herbivores and the phenology of prey in the case of predators (Høye and Forchhammer, 2008). The combined effects of colder temperatures and limited resources likely explains the strong discordance between May arthropod assemblages versus the rest of the summer. However, my results showed weak temporal structure in the local habitat variables that had an effect on arthropod communities, where a significant effect was only observed for plant community dissimilarity. In any case, arthropod communities in this system showed strong affinities with to seasonal changes in climatic conditions as is expected but rarely demonstrated empirically.

I found mixed results regarding the effect of agricultural intensity on arthropod community composition. An increased percentage of agricultural land can result in the fragmentation and loss of other habitat types in the landscape. Since some species fare better in

20

agricultural settings than others, it would be expected that a higher proportion of agricultural land would shift community composition towards species that thrive in highly-fragmented agricultural landscapes, especially herbivorous taxa specialized on crops (Gámez-Virués et al.,

2015; Gossner et al., 2016). In addition, fragmentation could lead to smaller, more disconnected patches of habitat which could filter for species with traits that allow them to survive in small patches of habitat as well as sufficient dispersal capacity to remain functionally connected to other patches of favorable habitat. Indeed, Simons, Wiesser, and Gossner (2016) found that increasing land-use intensity shifts trait distributions to favour smaller organisms that are capable of dispersing longer distances. I did not find evidence that changes in the proportion of agriculture in the landscape affects arthropod community composition. Although there was variation in the proportion of agriculture between traps in the 2000 m radii, ranging from 9% to

80%, most of the landscape of this region is nonetheless dominated by cropland. The relative uniformity of the landscape at larger spatial scales as well as strong effects of local habitat could explain why no effect was found at this scale of analysis.

I did, however, find a significant effect of site management style (conventional,

ALUS/mid-impact farms, and conservation areas). The reasons why are not entirely clear, but some possibilities could be differences in the representation of habitat types and plant communities, differences in pesticide or fertilizer use, and differences in sampling intensity

(ALUS/mid-impact farms were represented more heavily in our dataset than either conservation areas or conventional farms). ALUS/mid-impact farms often contain habitats that are not present or are smaller on conventional farms, but they have similar proportions of agriculture in the landscape (Figure A2a) and their fields are generally a soy-corn rotation that likely uses the same pesticide regime as conventional farms. However, they do tend to have a lower proportion of

21

urban area (Figure A2b) and a higher proportion of semi-natural land (Figure A2c), which could in part explain this effect.

These results corroborate studies that have shown local habitat factors to be more important than landscape-level factors for regulating invertebrate community composition, species richness, and abundance (Mazerolle and Villard, 1999; Stoner and Joern, 2004;

Schaffers, Raemakers, Sýkora, and ter Braak, 2008; Dolezal, 2019), though the literature is mixed on this. For example, Schweiger et al. (2005) found much stronger effects of land-use intensity than local habitat factors for regulating arthropod community composition, and many studies have shown that landscape-level factors can play a strong role in regulating arthropod communities (e.g. Hendrickx et al., 2007; Attwood, Maron, House and Zammit, 2008). It is likely that the importance of these processes is dependent on the study system as well as specific traits of the organisms under investigation, such as their feeding group (Dolezal, 2019).

Landscape and local factors may also have interactive effects, as found by Stoner and Joern

(2004). A balanced representation of management types, the quantification of other landscape metrics such as patch size, connectivity, configurational heterogeneity, and habitat diversity

(Fahrig et al., 2011), and a detailed account of management practices for each farm should be used to examine the effects of agricultural intensity relative to local habitat in greater detail.

When looking at the effects of individual environmental variables, most had a positive relationship with arthropod community dissimilarity as predicted. However, some showed the opposite. For example, a negative relationship was found for forest habitat distance (a binary variable), indicating that traps without forest (a distance of zero) tended to be more dissimilar to each other than the comparison of forests with other habitats (a distance of one). These effects are perhaps not surprising given how much local variation in habitat factors there were between

22

sites as well as the fact that habitat types were not independent in this design. Even in the case of site management, for which the categories were independent, a negative relationship was found comparing ALUS farms to the reference category of conventional farms. As such, these relationships likely reflect heterogenous local habitat conditions within categories that overwhelm the effect of these classifications.

Much of the spatiotemporal variation in arthropod community composition could be explained by environmental factors, indicating a strong role for niche structuring in both space and time in these communities. The direct effect of spatial distance on community composition was significant but surprisingly weak, especially given how far some traps were from each other.

The direct effect of temporal distance was much stronger, indicating changes in arthropod community composition throughout the growing season independent of the environmental variables measured here. The effects of spatial distances as well as temporal distances after controlling for environmental variation combined with effects of community size have been interpreted as reflecting the roles of dispersal and ecological drift (Hatosy et al., 2013; Matsuoka,

Kawaguchi, and Osono, 2016; Jabot et al. 2020).

My results suggest that dispersal is not highly limiting when considering the pure spatial distance between traps even though dispersal limitation might be expected to pervade these communities due to the regional dominance of agriculture and patchiness of natural habitat. That being said, many flying arthropods such as some species of bee are able to forage in ranges upwards of 5km (Greenleaf, Williams, Winfree, and Kremen, 2007) and are liable to drift long distances in windy conditions (Matthews and Matthews, 1971; Pasek, 1988), resulting in many transient species being caught in the traps and high dispersal potential. This could explain the high incidence of singleton occurrences that were observed here as well as the weak effect of

23

spatial distance. The effects of temporal distance and community size on arthropod community composition could have been generated by ecological drift, resulting in the differentiation of small communities as well as shifts in community composition through time that are not associated with environmental variation. However, my dataset is not well-suited to unambiguously address the relative importance of these processes. It is entirely possible that the direct effects of temporal and spatial distances are due to missing environmental variables, dispersal occurring between sampling periods or from outside of the sampling region, stochastic effects of sampling, and so on. Multi-year datasets along with estimations of dispersal rates and dispersal abilities among taxa combined with environmental data would be needed to precisely address the relative roles of environmental filtering, ecological drift, and dispersal.

An important research direction would be to quantify the relative importance of niche and neutral processes in more detail. For instance, several recent studies have shown that ecological drift may play a stronger role in determining the composition of local communities than is commonly thought (Gilbert and Levine, 2017; Sydenham et al., 2017; Jabot et al., 2020), but very few empirical studies have investigated this. This may be because long-term time series data are often needed to infer its importance relative to other mechanisms (Jabot et al., 2020).

Attaining a better understanding of these mechanisms has implications for the management of agricultural landscapes. If niche processes primarily govern the composition of arthropod communities then the focus of conservation efforts might be placed on ensuring that a diverse set of habitat types and local resources are represented in the landscape (Economo, 2011). If neutral processes primarily govern the composition of arthropod communities then more focus might be placed on the size and spatial arrangement of habitat patches due to the roles of ecological drift and dispersal limitation, respectively (Economo, 2011; Gilbert and Levine, 2017).

24

Despite attaining an acceptable model fit in path analysis, I likely did not account for some important site-specific variables. As such, more work on identifying features that are unique to each site would be valuable, especially other variables that are commonly cited as determinants of arthropod community composition such as soil properties (e.g. Antvogel and

Bonn, 2001). Soil is extremely important in agricultural systems, where the success of the crop depends in large part on properties of the soil such as the nutrient composition, biological activity, and physical structure (Parr, Papendick, Hornick, and Meyer, 1992). It is likewise important for many arthropods, particularly species that use soil as a nesting resource such as ground-nesting bees. These species tend to be at high risk of negative impacts due to factors such as the accumulation of pesticides in the soil or the modification of soil structure associated with tilling (Ullmann, Meisner, and Williams, 2016; Chan, Prosser, Rodríguez, and Raine, 2019).

Other important variables to quantify would be pesticide and fertilizer applications, especially if and when they are applied relative to the seasonal variation in climate and the phenology of plants and arthropods on farm landscapes. These relationships could not be quantified here.

Another consideration is that no information on the history of the sites was included nor were species interactions. Historic assembly processes can affect estimates of beta diversity independently of current dispersal and environmental gradients (Dexter, Terborgh, and

Cunningham, 2011). An example of this is priority effects, where colonizers at earlier time periods can determine species composition in later time periods, which could occur both intra- and interannually (Wainwright, Wolkovich, and Cleland, 2011; Fukami, 2015). Recent work has also shown that including residual species-to-species association matrices as an estimate of interspecific interactions can improve the explanatory power of community models in both space and time (Ovaskainen et al., 2017). On a larger temporal scale, it is possible that much of the

25

historic species pool may already have been filtered to favour species that are relatively insensitive to agriculture. This has been shown to be the case in forest plant communities in

Europe, where even fairly ancient land use history can strongly impact contemporary biodiversity (Dupouey, Dambrine, Laffite, and Moares, 2002). As well as this, site-specific temporal patterns of variation would not have been captured comprehensively in my analyses.

Sites likely vary from each other in their patterns of seasonal turnover, and some may be more or less deterministically structured depending on environmental context. For instance, disturbance can alter the relative strength of niche and neutral mechanisms of assembly in a non-monotonic fashion, where intermediate levels of disturbance are associated with the strongest niche assembly (Dong, Lytle, Olden, Schriever, and Muneepeerakul, 2017). If agricultural activity is a source of disturbance to many species, this same pattern could be observed over a gradient of agricultural intensity. The relative strength of niche and neutral mechanisms also varies based on community size and spatial scale, with ecological drift playing a more important role in small communities and single habitat types (Dexter, Terborgh, and Cunningham, 2011; Gilbert and

Levine, 2017). Testing these and other hypotheses of community assembly in agroecosystems should prove a fruitful area of future research.

Lastly, there are technical limitations associated with using malaise traps and metabarcoding. Malaise traps preferentially capture flying arthropods (Matthews and Matthews,

1971), which tend to be highly mobile. Flying arthropods may be able to better track preferred environmental conditions and as such may show weaker spatial effects than less mobile groups and could be better able to buffer negative effects of agricultural intensity (Simons, Wiesser, and

Gossner, 2016). In the case of metabarcoding, my analyses were restricted to presence-absence data since read counts are not a reliable indicator of abundance (Elbrecht and Leese, 2015). This

26

would over-represent certain species and under-represent others, which could affect the interpretation of my results as rare and common species could differ in their responses to environmental, spatial, and temporal variables. There are also primer biases that can result in the preferential amplification of certain sequences (Elbrecht and Leese, 2015).

My findings advance what is currently known about the impact of agriculture on arthropod biodiversity in several ways. It is rare for studies to examine arthropods in agroecosystems at such fine taxonomic resolution for many different orders over the seasonal plant growth cycle and at a relatively large spatial scale. My ability to do so was largely due to the combined use of metabarcoding and malaise traps, which is a highly scalable methodology

(deWaard et al., 2018). Though not without limitations, metabarcoding has many advantages over traditional morphological identification. It provides a standardized method for species assignment even when a species has not been formally described, allows for finer taxonomic resolution, allows for much faster sample processing time, and is very cost-effective (deWaard et al., 2018; Bush et al., 2020). Using barcoding rather than morphological identification can also increase estimates of species richness and beta diversity by revealing cryptic species (Brehm et al., 2016; D’Souza and Hebert, 2018), which allows for a more robust assessment to be made about the factors that drive variation in community composition (Bush et al., 2020).

My work also included a large suite of environmental factors, touching on several themes that are thus far uncommon in this literature. The effects of seasonal variation in climatic factors are not often investigated in agroecosystems, and I found this to be one of the most important factors explaining variation in arthropod community composition. Local habitat factors and agricultural intensity are also not often investigated in the same analyses (but see Schweiger et al., 2005 and Dolezal, 2019), especially on the spatiotemporal scale at which surveys were

27

conducted here. Many studies focus exclusively on landscape structure or site management and only use GIS data, but I found that local habitat factors, particularly tree canopy openness and plant community composition, were much more important determinants of arthropod community composition. However, I still found significant effects of site management on arthropod community composition, indicating that land-use should be considered in tandem with local habitat in restoration plans and studies of arthropod communities in agroecosystems.

To conclude, in this thesis I demonstrated that arthropod communities in agroecosystems are extremely variable among both localities and time periods. This variation was best explained by niche mechanisms, mainly local habitat factors and seasonal variation in climatic conditions, though small effects of site management were detectable. I showed that significant effects of spatial and temporal distance as well as community size remained after accounting for the effects of environmental distances, suggesting that neutral mechanisms may also have been at play. My findings show that habitat restoration on marginal land is likely to prove an effective strategy for increasing arthropod biodiversity in agroecosystems and that studies with a snap-shot design are missing a large amount of information about arthropod community composition. Given the immense challenge that the world faces in feeding a large and growing human population, it is critical that we continue to better understand and implement the strategies that can work alongside agriculture to maintain viable habitat for the benefit of human and non-human communities alike.

28

REFERENCES Agriculture and Agri-Food Canada, 2018. Annual Crop Inventory 2018. Government of Canada. https://open.canada.ca/data/en/dataset/1f2ad87e-6103-4ead-bdd5-147c33fa11e6.

Albrecht, M., B. Schmid, M. K. Obrist, B. Schüpbach, D. Kleijn, and P. Duelli. 2010. Effects of ecological compensation meadows on arthropod diversity in adjacent intensively managed grassland. Biological Conservation 143:642–649.

ALUS Canada. 2020. ALUS Norfolk. https://alus.ca/alus_community/alus-norfolk/. Downloaded on May 19, 2020. Antvogel, H., and Bonn, A. 2001. Environmental parameters and microspatial distribution of insects: a case study of carabids in an alluvial forest. Ecography 24:470–482.

Attwood, S. J., M. Maron, A. P. N. House, and C. Zammit. 2008. Do arthropod assemblages display globally consistent responses to intensified agricultural land use and management? Global Ecology and Biogeography 17:585–599.

Baselga, A. 2010. Partitioning the turnover and nestedness components of beta diversity. Global Ecology and Biogeography 19:134–143.

Bell, G. 2000. The distribution of abundance in neutral communities. American Naturalist 155:606–617.

Bell, G. 2001. Neutral macroecology. Science 293:2413–2418.

Beng, K. C., R. T. Corlett, and K. W. Tomlinson. 2018. Seasonal changes in the diversity and composition of the litter fauna in native forests and rubber plantations. Scientific Reports 8:1–13.

Benjamini, Y., and Y. Hochberg. 1995. Controlling for the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B. 57:279– 300.

Bertrand, C., F. Burel, and J. Baudry. 2016. Spatial and temporal heterogeneity of the crop mosaic influences carabid beetles in agricultural landscapes. Landscape Ecology 31:451– 466.

Borer, E. T., E. W. Seabloom, and D. Tilman. 2012. Plant diversity controls arthropod biomass and temporal stability. Ecology Letters 15:1457–1464.

Brehm, G., P. D. N. Hebert, R. K. Colwell, M. O. Adams, F. Bodner, K. Friedemann, L. Möckel, and K. Fiedler. 2016. Turning up the heat on a hotspot: DNA barcodes reveal 80% more species of geometrid moths along an andean elevational gradient. PLoS ONE 11:1–15.

29

Brose, U. 2003. Bottom-up control of carabid beetle communities in early successional wetlands: Mediated by vegetation composition or plant diversity? Oecologia 135:407–413.

Bush, A., et al. 2020. DNA metabarcoding reveals metacommunity dynamics in a threatened boreal wetland wilderness. Proceedings of the National Academy of Sciences of the United States of America 117:8539–8545.

Byrne, K. M., W. K. Lauenroth, P. B. Adler, and C. M. Byrne. 2011. Estimating aboveground net primary production in grasslands: A comparison of non-destructive methods. Rangeland Ecology and Management 64:498–505.

Cangur, S., and I. Ercan. 2015. Comparison of model fit indices used in structural equation modeling under multivariate normality. Journal of Modern Applied Statistical Methods 14:152–167.

Chan, D. S. W., R. S. Prosser, J. L. Rodríguez-Gil, and N. E. Raine. 2019. Assessment of risk to hoary squash bees (Peponapis pruinosa) and other ground-nesting bees from systemic insecticides in agricultural soil. Scientific Reports 9:1–13.

D’Souza, M. L., and P. D. N. Hebert. 2018. Stable baselines of temporal turnover underlie high beta diversity in tropical arthropod communities. Molecular Ecology 27:2447–2460. deWaard, J. R., et al. 2018. Expedited assessment of terrestrial arthropod diversity by coupling Malaise traps with DNA barcoding. Genome 11:1–11.

Dexter, K. G., J. W. Terborgh, and C. W. Cunningham. 2012. Historical effects on beta diversity and community assembly in Amazonian trees. Proceedings of the National Academy of Sciences 109:7787–7792.

Dolezal, A. 2019. Habitat-based drivers of arthropod abundance and richness in an intensively farmed agricultural landscape. Master's Thesis. The University of Guelph, Guelph, Ontario, Canada.

Dong, X., D. A. Lytle, J. D. Olden, T. A. Schriever, and R. Muneepeerakul. 2017. Importance of neutral processes varies in time and space: Evidence from dryland stream ecosystems. PLoS one 12:1–14.

Dupouey, J. L., E. Dambrine, J. D. Laffite, and C. Moares. 2002. Irreversible impact of past land use on forest soils and biodiversity. Ecology 83:2978–2984.

Ebeling, A., J. Hines, L. R. Hertzog, M. Lange, S. T. Meyer, N. K. Simons, and W. W. Weisser. 2018. Plant diversity effects on arthropods and arthropod-dependent ecosystem functions in a biodiversity experiment. Basic and Applied Ecology 26:50–63.

Economo, E. P. 2011. Biodiversity conservation in metacommunity networks: Linking pattern and persistence. American Naturalist 177:E167–E180

30

Elbrecht, V., and F. Leese. 2015. Can DNA-based ecosystem assessments quantify species abundance? Testing primer bias and biomass-sequence relationships with an innovative metabarcoding protocol. PLoS ONE 10:1–16.

Fahrig, L., J. Baudry, L. Brotons, F. G. Burel, T. O. Crist, R. J. Fuller, C. Sirami, G. M. Siriwardena, and J.-L. Martin. 2011. Functional landscape heterogeneity and animal biodiversity in agricultural landscapes. Ecology Letters 14:101–112.

Fourtune, L., J. G. Prunier, I. Paz-Vinas, G. Loot, C. Veyssière, and S. Blanchet. 2018. Inferring causalities in landscape genetics: An extension of wright’s causal modeling to distance matrices. American Naturalist 191:491–508.

Frank, A. F., and S.J. McNaughton. 1990. Aboveground biomass estimation with the canopy intercept method: a plant growth form caveat. Oikos 57:57–60.

Fukami, T. 2015. Historical contingency in community assembly: Integrating niches, species pools, and priority effects. Annual Review of Ecology, Evolution, and Systematics 46:1–23.

Gámez-Virués, S. et al. 2015. Landscape simplification filters species traits and drives biotic homogenization. Nature Communications 6.

Gerland, P., et al. 2014. World population stabilization unlikely this century. Science. 346:234– 237.

Gilbert, B., and J. M. Levine. 2017. Ecological drift and the distribution of species diversity. Proceedings of the Royal Society B: Biological Sciences 284:20170507.

Gossner, M. M. et al. 2016. Land-use intensification causes multitrophic homogenization of grassland communities. Nature 540:266–269.

Greenleaf, S. S., N. M. Williams, R. Winfree, and C. Kremen. 2007. Bee foraging ranges and their relationship to body size. Oecologia 153:589–596.

Grøtan, V., R. Lande, S. Engen, B. E. Sæther, and P. J. Devries. 2012. Seasonal cycles of species diversity and similarity in a tropical butterfly community. Journal of Animal Ecology 81:714–723.

Haaland, C., R. E. Naisbit, and L. F. Bersier. 2011. Sown wildflower strips for insect conservation: A review. Insect Conservation and Diversity 4:60–80.

Hallmann, C. A., et al. 2017. More than 75 percent decline over 27 years in total flying insect biomass in protected areas. PLoS ONE 12: e0185809.

Hatosy, S. M., J. B. H. Martiny, R. Sachdeva, J. Steele, J. A. Fuhrman, and A. C. Martiny. 2013. Beta diversity of marine bacteria depends on temporal scale. Ecology 94:1898–1904.

31

Hebert, P. D. N., E. H. Penton, J. M. Burns, D. H. Janzen, and W. Hallwachs. 2004. Ten species in one: DNA barcoding reveals cryptic species in the neotropical skipper butterfly Astraptes fulgerator. Proceedings of the National Academy of Sciences of the United States of America 101:14812–14817.

Hendrickx, F., et al. 2007. How landscape composition, land-use intensity and habitat diversity affect components of total arthropod diversity in agricultural landscapes. Journal of Applied Ecology 44:340–351.

Hendrickx, F., et al. 2009. Pervasive effects of dispersal limitation on within- and among- community species richness in agricultural landscapes. Global Ecology and Biogeography 18:607–616.

Hesselbarth, M. H. K., M. Sciaini, K. A. With, K. Wiegand, and J. Nowosad. 2019. landscapemetrics: an open-source R tool to calculate landscape metrics. Ecography 42:1648–1657.

Høye, T. T., and M. C. Forchhammer. 2008. Phenology of high-arctic arthropods: Effects of climate on spatial, seasonal, and inter-annual variation. Advances in Ecological Research 40:299–324.

Hubbell, S. P. 2001. The unified neutral theory of biodiversity and biogeography. Princeton University Press, Princeton, New Jersey, USA.

Ivanova, N. V., J. R. deWaard, and P. D. N. Hebert. 2006. An inexpensive, automation-friendly protocol for recovering high-quality DNA. Molecular Ecology Notes 6:998–1002.

Jabot, F., F. Laroche, F. Massol, F. Arthaud, J. Crabot, M. Dubart, S. Blanchet, F. Munoz, P. David, and T. Datry. 2020. Assessing metacommunity processes through signatures in spatiotemporal turnover of community composition. Ecology Letters doi: 10.1111/ele.13523.

Koleff, P., K. J. Gaston, and J. J. Lennon. 2003. Measuring beta diversity for presence-absence data. Journal of Animal Ecology 72:367–382.

Kremen, C., R. K. Colwell, T. L. Erwin, D. D. Murphy, R. F. Noss, and M. A. Sanjayan. 1993. Arthropod assemblages: their use in conservation planning. Conservation Biology 7:796– 808.

Lichtenberg, E. M. et al. 2017. A global synthesis of the effects of diversified farming systems on arthropod diversity within fields and across agricultural landscapes. Global Change Biology 23:4946–4957.

Maisonhaute, J. É., P. Peres-Neto, and E. Lucas. 2015. Seasonal modulation of landscape effects on predatory beetle assemblage. Bulletin of Insectology 68:181–191.

32

Matson, P. A., W. J. Parton, A. G. Power, and M. J. Swift. 1997. Agricultural intensification and ecosystem properties. Science 277:504–509.

Matsuoka, S., E. Kawaguchi, and T. Osono. 2016. Temporal distance decay of similarity of ectomycorrhizal fungal community composition in a subtropical evergreen forest in Japan. FEMS Microbiology Ecology 92:fiw061.

Matthews, R. W., and J. R. Matthews. 1971. The Malaise trap: Its utility and potential for sampling insect population. Michigan Entomologist 4:117–122.

Mazerolle, M. J., and M. A. Villard. 1999. Patch characteristics and landscape context as predictors of species presence and abundance: A review. Ecoscience 6:117–124.

Oskanen, J. et al. 2019. Vegan: Community ecology package. R package version 2.5-6. https://CRAN.R-project.org/package=vegan.

Ovaskainen, O., G. Tikhonov, A. Norberg, F. Guillaume Blanchet, L. Duan, D. Dunson, T. Roslin, and N. Abrego. 2017. How to make more out of community data? A conceptual framework and its implementation as models and software. Ecology Letters 20:561–576.

Padgham, M. and D. Sumner. 2020. Geodist: fast, dependency-free geodesic distance calculations. R package version 0.0.4. https://CRAN.R-project.org/package=geodist.

Pasek, J. E. 1988. Influence of wind and windbreaks on local dispersal of insects. Agriculture, Ecosystems and Environment 22–23:539–554.

Parr F, J., R. Papendick I, S. Hornick B, and R. Meyer E. 1992. Soil Quality: Attributes and relationship to alternative and sustainable agriculture. American Journal of Alternative Agriculture 7:5–11.

Paterson, C., K. Cottenie, and A. S. MacDougall. 2019. Restored native prairie supports abundant and species-rich native bee communities on conventional farms. Restoration Ecology 27:1291–1299.

Prather, R. M., and M. Kaspari. 2019. Plants regulate grassland arthropod communities through biomass, quality, and habitat heterogeneity. Ecosphere 10.

Prosser, S. W. J., J. R. Dewaard, S. E. Miller, and P. D. N. Hebert. 2016. DNA barcodes from century-old type specimens using next-generation sequencing. Molecular Ecology Resources 16:487–497.

R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/

Ratnasingham, S., and P. D. N. Hebert. 2013. A DNA-based registry for all animal species: The Barcode Index Number (BIN) system. PLoS ONE 8.

33

Rosseel, Y. 2012. Lavaan: An R package for structural equation modeling. Journal of Statistical Software 48.

Sánchez-Bayo, F., and K. A. G. Wyckhuys. 2019. Worldwide decline of the entomofauna: A review of its drivers. Biological Conservation 232:8–27.

Schaffers, A. P., I. P. Raemakers, K. V Sýkora, and C. J. F. ter Braak. 2008. Arthropod assemblages are best predicted by plant species composition. Ecology 89:782–794.

Scheper, J., A. Holzschuh, M. Kuussaari, S. G. Potts, M. Rundlöf, H. G. Smith, and D. Kleijn. 2013. Environmental factors driving the effectiveness of European agri-environmental measures in mitigating pollinator loss - a meta-analysis. Ecology Letters 16:912–920.

Schweiger, O. et al. 2005. Quantifying the impact of environmental factors on arthropod communities in agricultural landscapes across organizational levels and spatial scales. Journal of Applied Ecology 42:1129–1139.

Seibold, S., et al. 2019. Arthropod decline in grasslands and forests is associated with landscape- level drivers. Nature 574:671–674.

Shimadzu, H., M. Dornelas, P. A. Henderson, and A. E. Magurran. 2013. Diversity is maintained by seasonal variation in species abundance. BMC Biology 11 doi:10.1186/1741-7007-11- 98.

Simons, N. K., W. W. Weisser, and M. M. Gossner. 2016. Multi-taxa approach shows consistent shifts in arthropod functional traits along grassland land-use intensity gradient. Ecology 97:754–764.

Smith, J. P. 2017. The subfamilies & tribes of the grasses of the United States: Evolving concepts. Botanical studies. Sørensen, T.A. 1948. A method of establishing groups of equal amplitude in plant sociology based on similarity of species content, and its application to analyses of the vegetation on Danish commons. Kongelige Danske Videnskabernes Selskabs Biologiske Skrifter 5: 1–34.

Stinson, C. S. A, and V. K. Brown. 1983. Seasonal changes in the architecture of natural plant communities and its relevance to insect herbivores. Oecologia 56:67–69.

Stoner, K. J. L., and A. Joern. 2004. Landscape vs. local habitat scale influences to insect communities from tallgrass prairie remnants. Ecological Applications 14:1306–1320.

Stork, N. E. 2018. How many species of insects and other terrestrial arthropods are there on earth? Annual Review of Entomology 63:31–45.

Sydenham, M. A. K., K. Eldegard, and Ø. Totland. 2014. Spatio-temporal variation in species assemblages in field edges: Seasonally distinct responses of solitary bees to local habitat characteristics and landscape conditions. Biodiversity and Conservation 23:2393–2414.

34

Sydenham, M. A. K., S. R. Moe, M. Kuhlmann, S. G. Potts, S. P. M. Roberts, Ø. Totland, and K. Eldegard. 2017. Disentangling the contributions of dispersal limitation, ecological drift, and ecological filtering to wild bee community assembly. Ecosphere 8:e01650.

Thrupp, L. A. 2004. The importance of biodiversity in agroecosystems. Journal of Crop Improvement 12:315–337.

Tilman, D., C. Balzer, J. Hill, and B. L. Befort. 2011. Global food demand and the sustainable intensification of agriculture. Proceedings of the National Academy of Sciences 108:20260– 20264.

Tilman, D., K. G. Cassman, P. A. Matson, R. Naylor, and S. Polasky. 2002. Agricultural sustainability and intensive production practices. Nature 418:671–677.

Toledo, D., L. B. Abbott, and J. E. Herrick. 2008. Cover pole design for easy transport, assembly, and field use. Journal of Wildlife Management 72:564–567.

Tscharntke, T. et al. 2012. Landscape moderation of biodiversity patterns and processes - eight hypotheses. Biological Reviews 87:661–685.

Ullmann, K. S., M. H. Meisner, and N. M. Williams. 2016. Impact of tillage on the crop pollinating, ground-nesting bee, Peponapis pruinosa in California. Agriculture, Ecosystems and Environment 232:240–246.

Vandermeer, J. H. 1972. Niche Theory. Annual Review of Ecology and Systematics 3:107–132.

Venables, W.N. and B.D. Ripley. 2002. Modern applied statistics with S. Fourth edition. Springer, New York, USA.

Wainwright, C. E., E. M. Wolkovich, and E. E. Cleland. 2012. Seasonal priority effects: Implications for invasion and restoration in a semi-arid system. Journal of Applied Ecology 49:234–241.

Watson, J. C., A. T. Wolf, and J. S. Ascher. 2011. Forested landscapes promote richness and abundance of native bees (Hymenoptera: Apoidea: Anthophila) in Wisconsin apple orchards. Environmental Entomology 40:621–632.

Wolda, H. 1988. Insect seasonality: Why? Annual Review of Ecology and Systematics 19:1–18.

35

TABLES

Table 1. Environmental variables used in path analysis and their groupings into either local

habitat, agricultural intensity, or climatic variables. Type of variation refers to whether the

explanatory variable exhibits spatial (S) temporal (T) or spatiotemporal (S+T) variation.

*Limited spatial variation based on closest weather stations.

GROUP VARIABLE TYPE OF VARIATION LOCAL HABITAT CANOPY OPENNESS S ELEVATION S VISUAL OBSTRUCTION S+T PLANT GENUS RICHNESS S+T FLOWER RICHNESS S+T BRAY-CURTIS PLANT COMMUNITY S+T COMPOSITION PRESENCE OF FOREST HABITAT S PRESENCE OF CROP HABITAT S PRESENCE OF GRASSLAND/MEADOW S HABITAT PRESENCE OF WATER BODY S AGRICULTURAL % AGRICULTURE IN 2KM RADIUS S INTENSITY ALUS FARM S CONVENTIONAL FARM S CONSERVATION AREA S CLIMATIC AVERAGE TEMPERATURE S+T AVERAGE RELATIVE HUMIDITY S+T* CV TEMPERATURE S+T AVERAGE WIND SPEED S+T* AVERAGE WIND DIRECTION S+T* AVERAGE PRECIPITATION S+T*

36

Table 2. Significant results (P < 0.05 with Benjamini-Hochberg correction) from path analysis and associated standardized path coefficients.

GROUP PATH STANDARDIZED PATH COEFFICIENT Δ ALPHA ~ Δ COMMUNITY BIOMASS 0.22626 BETA_SØR ~ Δ ALPHA 0.20966 BETA_SØR ~ Δ TIME 0.232 BETA_SØR ~ Δ SPACE 0.07362 BETA_SØR ~ MEAN COMMUNITY BIOMASS -0.2398 LOCAL BETA_SØR ~ Δ CANOPY OPENNESS 0.32158 HABITAT BETA_SØR ~ Δ PLANT COMMUNITY 0.25305 BETA_SØR ~ Δ FOREST HABITAT -0.0732 BETA_SØR ~ Δ CROP HABITAT -0.04 BETA_SØR ~ Δ WATER HABITAT 0.02335 Δ CANOPY OPENNESS ~ Δ SPACE 0.06048 Δ PLANT COMMUNITY ~ Δ SPACE 0.19782 Δ PLANT COMMUNITY ~ Δ TIME 0.04091 Δ FOREST HABITAT ~ Δ SPACE 0.18204 Δ CROP HABITAT ~ Δ SPACE 0.18765 Δ WATER HABITAT ~ Δ SPACE 0.05106 Δ GRASSLAND/MEADOW HABITAT ~ Δ SPACE 0.13196 Δ VISUAL OBSTRUCTION ~ Δ SPACE 0.09807 Δ VISUAL OBSTRUCTION ~ Δ TIME 0.08618 Δ PLANT RICHNESS ~ Δ SPACE 0.07044 Δ FLOWER RICHNESS ~ Δ TIME 0.0591 Δ ELEVATION ~ Δ SPACE 0.43038 CLIMATIC BETA_SØR ~ Δ CV TEMPERATURE 0.20414 BETA_SØR ~ Δ TEMPERATURE 0.16602 BETA_SØR ~ Δ HUMIDITY 0.0837 Δ CV TEMPERATURE ~ Δ TIME 0.21181 Δ TEMPERATURE ~ Δ TIME 0.29265 Δ HUMIDITY ~ Δ TIME 0.54891 Δ WIND SPEED ~ Δ SPACE 0.28304 Δ WIND SPEED ~ Δ TIME 0.16162 Δ WIND DIRECTION ~ Δ TIME 0.16174 Δ PRECIPITATION ~ Δ TIME 0.12811 Δ PRECIPITATION ~ Δ SPACE 0.15744 AGRICULTURAL BETA_SØR ~ Δ CONSERVATION AREA 0.14276 INTENSITY BETA_SØR ~ Δ ALUS/MID-IMPACT FARM -0.0514 Δ CONSERVATION AREA ~ Δ SPACE 0.3539 Δ ALUS/MID-IMPACT FARM ~ Δ SPACE 0.17165 Δ % AGRICULTURE ~ Δ SPACE 0.2538

37

FIGURES

Figure 1. Distribution of arthropod BINs (Barcode Index Number; a species proxy) for the top 6 orders across all samples. Represents 94% of total filtered BINs.

38

Figure 2. Total arthropod BIN richness (Barcode Index Number; a species proxy) among traps pooled across all time periods.

39

Figure 3. Total arthropod BIN richness (Barcode Index Number; a species proxy) among sample periods pooled across all traps.

40

Figure 4. Non-metric multidimensional scaling (NMDS) ordination of arthropod species composition based on pairwise Sørensen dissimilarities among all samples. Samples represented by month of collection.

41

Figure 5. Path analysis of factors influencing spatiotemporal Sørensen dissimilarity of arthropod community composition. < J > is mean community size (arthropod biomass used as a proxy), S is species richness, βsor is arthropod pairwise Sørensen dissimilarity.

Values correspond to standardized path coefficients. Only significant paths are shown, and paths from spatial and temporal distances to environmental distances are only visualized for variables that also had a significant effect on arthropod dissimilarity. The standardized root mean square residual (SRMR) value of the model was 0.086.

42

Figure 6. Path analysis of factors influencing spatiotemporal Sørensen dissimilarity of arthropod community composition. < J > is mean community size (biomass used as a proxy), S is species richness, βsor is arthropod pairwise Sørensen dissimilarity. Only significant paths are shown.

Values correspond to standardized path coefficients, and values towards or away from variable groups (e.g. climate) represent the sum of the absolute values of standard path coefficients for each variable within those groups. Paths from spatial and temporal distances to environmental distances are only visualized for variables that also had a significant effect on arthropod dissimilarity. The standardized root mean square residual (SRMR) value of the model was 0.086.

43

Figure 7. Spatial distance-decay of arthropod community composition based on pairwise

Sørensen dissimilarities among samples after filtering the arthropod data to match the plant data.

Shading around the regression line indicates a 95% confidence interval.

44

Figure 8. Temporal distance-decay of arthropod community composition based on pairwise

Sørensen dissimilarities among all samples after filtering the arthropod data to match the plant data. Shading around the regression line indicates a 95% confidence interval.

45

Figure 9. Relationship between pairwise Sørensen dissimilarities of arthropod communities and

Bray-Curtis dissimilarities of plant communities for all samples after filtering the arthropod data to match the plant data. Shading around the regression line indicates a 95% confidence interval.

46

APPENDIX

Table A1. Temporal alignment of plant and arthropod surveys by trap. Start and end date refer to a collection period for a given malaise trap bottle. Samples were aligned to based on the closest dates, meaning that either the first or last arthropod sample was omitted depending on the plant survey schedule.

Trap Cycle Start Date End Date Plant survey number Plant survey date

1 A 2019-05-01 2019-05-15 1 C 2019-05-29 2019-06-12 1 2019-05-30 1 E 2019-06-26 2019-07-10 2 2019-06-27 1 G 2019-07-24 2019-08-07 3 2019-07-25 1 I 2019-08-21 2019-09-04 4 2019-08-22 1 K 2019-09-18 2019-10-02 5 2019-09-19

2 A 2019-05-01 2019-05-15 2 C 2019-05-29 2019-06-12 1 2019-05-30 2 E 2019-06-26 2019-07-10 2 2019-06-27 2 G 2019-07-24 2019-08-07 3 2019-07-25 2 I 2019-08-21 2019-09-04 4 2019-08-22 2 K 2019-09-18 2019-10-02 5 2019-09-19 3 A 2019-05-07 2019-05-21 1 2019-05-21 3 C 2019-06-04 2019-06-18 2 2019-06-18 3 E 2019-07-02 2019-07-16 3 2019-07-16 3 G 2019-07-30 2019-08-13 4 2019-08-13 3 I 2019-08-27 2019-09-10 5 2019-09-10

3 K 2019-09-24 2019-10-08 4 A 2019-05-07 2019-05-21 1 2019-05-21 4 C 2019-06-04 2019-06-18 2 2019-06-18 4 E 2019-07-02 2019-07-16 3 2019-07-16 4 I 2019-08-27 2019-09-10 5 2019-09-10

4 K 2019-09-24 2019-10-08

5 A 2019-04-30 2019-05-14 5 C 2019-05-28 2019-06-11 1 2019-05-28 5 E 2019-06-25 2019-07-09 2 2019-06-25 5 G 2019-07-23 2019-08-06 3 2019-07-23 5 I 2019-08-20 2019-09-03 4 2019-08-20 5 K 2019-09-17 2019-10-01 5 2019-09-17

6 A 2019-04-30 2019-05-14

47

6 C 2019-05-28 2019-06-11 1 2019-05-28 6 E 2019-06-25 2019-07-09 2 2019-06-25 6 G 2019-07-23 2019-08-06 3 2019-07-23 6 I 2019-08-20 2019-09-03 4 2019-08-20 6 K 2019-09-17 2019-10-01 5 2019-09-17 7 A 2019-05-08 2019-05-22 1 2019-05-22 7 C 2019-06-05 2019-06-19 2 2019-06-19 7 E 2019-07-03 2019-07-17 3 2019-07-17 7 G 2019-07-31 2019-08-14 4 2019-08-14 7 I 2019-08-28 2019-09-11 5 2019-09-11

7 K 2019-09-25 2019-10-09 8 A 2019-05-08 2019-05-22 1 2019-05-22 8 C 2019-06-05 2019-06-19 2 2019-06-19 8 E 2019-07-03 2019-07-17 3 2019-07-17 8 G 2019-07-31 2019-08-14 4 2019-08-14 8 I 2019-08-28 2019-09-11 5 2019-09-11

8 K 2019-09-25 2019-10-09 9 A 2019-05-01 2019-05-15 1 2019-05-15 9 C 2019-05-29 2019-06-12 2 2019-06-12 9 E 2019-06-26 2019-07-10 3 2019-07-10 9 G 2019-07-24 2019-08-07 4 2019-08-07 9 I 2019-08-21 2019-09-04 5 2019-09-04

9 K 2019-09-18 2019-10-02 10 A 2019-05-01 2019-05-15 1 2019-05-15 10 C 2019-05-29 2019-06-12 2 2019-06-12 10 E 2019-06-26 2019-07-10 3 2019-07-10 10 G 2019-07-24 2019-08-07 4 2019-08-07 10 I 2019-08-21 2019-09-04 5 2019-09-04

10 K 2019-09-18 2019-10-02 11 A 2019-04-30 2019-05-14 1 2019-05-14 11 C 2019-05-28 2019-06-11 2 2019-06-11 11 E 2019-06-25 2019-07-09 3 2019-07-09 11 G 2019-07-23 2019-08-06 4 2019-08-06 11 I 2019-08-20 2019-09-03 5 2019-09-03

11 K 2019-09-17 2019-10-01 12 A 2019-04-30 2019-05-14 1 2019-05-14 12 C 2019-05-28 2019-06-11 2 2019-06-11 12 E 2019-06-25 2019-07-09 3 2019-07-09 12 G 2019-07-23 2019-08-06 4 2019-08-06 12 I 2019-08-20 2019-09-03 5 2019-09-03

12 K 2019-09-17 2019-10-01

48

13 A 2019-05-07 2019-05-21 1 2019-05-24 13 C 2019-06-04 2019-06-18 2 2019-06-20 13 E 2019-07-02 2019-07-16 3 2019-07-18 13 G 2019-07-30 2019-08-13 4 2019-08-15 13 I 2019-08-27 2019-09-10 5 2019-09-12

13 K 2019-09-24 2019-10-08

14 A 2019-05-07 2019-05-21 14 C 2019-06-04 2019-06-18 1 2019-06-04 14 E 2019-07-02 2019-07-16 2 2019-07-02 14 G 2019-07-30 2019-08-13 3 2019-07-30 14 I 2019-08-27 2019-09-10 4 2019-08-27 14 K 2019-09-24 2019-10-08 5 2019-09-24

15 A 2019-05-07 2019-05-21 15 C 2019-06-04 2019-06-18 1 2019-06-04 15 E 2019-07-02 2019-07-16 2 2019-07-02 15 G 2019-07-30 2019-08-13 3 2019-07-30 15 I 2019-08-27 2019-09-10 4 2019-08-27 15 K 2019-09-24 2019-10-08 5 2019-09-24

16 A 2019-05-08 2019-05-22 16 C 2019-06-05 2019-06-19 1 2019-06-06 16 E 2019-07-03 2019-07-17 2 2019-07-04 16 G 2019-07-31 2019-08-14 3 2019-08-01 16 I 2019-08-28 2019-09-11 4 2019-08-29 16 K 2019-09-25 2019-10-09 5 2019-09-26

17 A 2019-05-08 2019-05-22 17 C 2019-06-05 2019-06-19 1 2019-06-06 17 E 2019-07-03 2019-07-17 2 2019-07-04 17 G 2019-07-31 2019-08-14 3 2019-08-01 17 I 2019-08-28 2019-09-11 4 2019-08-29 17 K 2019-09-25 2019-10-09 5 2019-09-26 18 A 2019-05-02 2019-05-16 1 2019-05-16 18 C 2019-05-30 2019-06-13 2 2019-06-13 18 E 2019-06-27 2019-07-11 3 2019-07-11 18 G 2019-07-25 2019-08-08 4 2019-08-08 18 I 2019-08-22 2019-09-05 5 2019-09-05

18 K 2019-09-19 2019-10-03 19 A 2019-05-02 2019-05-16 1 2019-05-16 19 C 2019-05-30 2019-06-13 2 2019-06-13 19 E 2019-06-27 2019-07-11 3 2019-07-11 19 G 2019-07-25 2019-08-08 4 2019-08-08 19 I 2019-08-22 2019-09-05 5 2019-09-05

49

19 K 2019-09-19 2019-10-03 20 A 2019-05-02 2019-05-16 1 2019-05-16 20 C 2019-05-30 2019-06-13 2 2019-06-13 20 E 2019-06-27 2019-07-11 3 2019-07-11 20 G 2019-07-25 2019-08-08 4 2019-08-08 20 I 2019-08-22 2019-09-05 5 2019-09-05

20 K 2019-09-19 2019-10-03 21 A 2019-05-02 2019-05-16 1 2019-05-16 21 C 2019-05-30 2019-06-13 2 2019-06-13 21 E 2019-06-27 2019-07-11 3 2019-07-11 21 G 2019-07-25 2019-08-08 4 2019-08-08 21 I 2019-08-22 2019-09-05 5 2019-09-05

21 K 2019-09-19 2019-10-03 22 A 2019-05-07 2019-05-21 1 2019-05-24 22 C 2019-06-04 2019-06-18 2 2019-06-20 22 E 2019-07-02 2019-07-16 3 2019-07-18 22 G 2019-07-30 2019-08-13 4 2019-08-05 22 I 2019-08-27 2019-09-10 5 2019-09-12

22 K 2019-09-24 2019-10-08 23 A 2019-05-07 2019-05-21 1 2019-05-24 23 C 2019-06-04 2019-06-18 2 2019-06-20 23 E 2019-07-02 2019-07-16 3 2019-07-18 23 G 2019-07-30 2019-08-13 4 2019-08-05 23 I 2019-08-27 2019-09-10 5 2019-09-12

23 K 2019-09-24 2019-10-08

24 A 2019-05-02 2019-05-16 24 C 2019-05-30 2019-06-13 1 2019-05-30 24 E 2019-06-27 2019-07-11 2 2019-06-27 24 G 2019-07-25 2019-08-08 3 2019-07-25 24 I 2019-08-22 2019-09-05 4 2019-08-22 24 K 2019-09-19 2019-10-03 5 2019-09-19

25 A 2019-05-02 2019-05-16 25 C 2019-05-30 2019-06-13 1 2019-05-30 25 E 2019-06-27 2019-07-11 2 2019-06-27 25 G 2019-07-25 2019-08-08 3 2019-07-25 25 I 2019-08-22 2019-09-05 4 2019-08-22 25 K 2019-09-19 2019-10-03 5 2019-09-19

26 A 2019-05-01 2019-05-15 26 C 2019-05-29 2019-06-12 1 2019-05-29 26 E 2019-06-26 2019-07-10 2 2019-06-26 26 G 2019-07-24 2019-08-07 3 2019-07-24

50

26 I 2019-08-21 2019-09-04 4 2019-08-21 26 K 2019-09-18 2019-10-02 5 2019-09-18

27 A 2019-05-01 2019-05-15 27 C 2019-05-29 2019-06-12 1 2019-05-29 27 E 2019-06-26 2019-07-10 2 2019-06-26 27 G 2019-07-24 2019-08-07 3 2019-07-24 27 I 2019-08-21 2019-09-04 4 2019-08-21 27 K 2019-09-18 2019-10-02 5 2019-09-18

28 A 2019-05-08 2019-05-22 28 C 2019-06-05 2019-06-19 1 2019-06-05 28 E 2019-07-03 2019-07-17 2 2019-07-03 28 G 2019-07-31 2019-08-14 3 2019-07-31 28 I 2019-08-28 2019-09-11 4 2019-08-28 28 K 2019-09-25 2019-10-09 5 2019-09-25

29 A 2019-05-08 2019-05-22 29 C 2019-06-05 2019-06-19 1 2019-06-05 29 E 2019-07-03 2019-07-17 2 2019-07-03 29 G 2019-07-31 2019-08-14 3 2019-07-31 29 I 2019-08-28 2019-09-11 4 2019-08-28 29 K 2019-09-25 2019-10-09 5 2019-09-25

51

Table A2. Full list of the plant taxa from all surveys by plant type. Genus (or higher) Plant type Agrostis C3 grass Bromus C3 grass Dactylis C3 grass Elymus C3 grass Festuca C3 grass Festuceae C3 grass Glyceria C3 grass Hordeum C3 grass Leersia C3 grass Lolium C3 grass Patis C3 grass Phalaris C3 grass Phleum C3 grass Phragmites C3 grass Poa C3 grass Poaceae C3 grass Typha C3 grass Andropogon C4 grass Cynodon C4 grass Digitaria C4 grass Echinochloa C4 grass Panicum C4 grass Setaria C4 grass Sorghastrum C4 grass Glycine Crop Pisum Crop Raphanus Crop Sorghum Crop Triticum Crop Zea Crop Abutilon Forb Acalypha Forb Achillea Forb Ageratina Forb Agrimonia Forb Alisma Forb Alliaria Forb Alyssum Forb Amaranthus Forb

52

Ambrosia Forb Amphicarpaea Forb Anagallis Forb Anemone Forb Apocynum Forb Aralia Forb Arctium Forb Arisaema Forb Asarum Forb Asclepias Forb Berteroa Forb Bidens Forb Boehmeria Forb Brassica Forb Campanula Forb Capsella Forb Cardamine Forb Cerastium Forb Chamaesyce Forb Chelidonium Forb Chelone Forb Chenopodium Forb Cichorium Forb Circaea Forb Cirsium Forb Claytonia Forb Clematis Forb Clinopodium Forb Convolvus Forb Coptis Forb Crepis Forb Cryptotaenia Forb Cynanchum Forb Cypripedium Forb Daucus Forb Desmodium Forb Dianthus Forb Dipsacus Forb Draba Forb Echinocystis Forb Epilobium Forb

53

Epipactis Forb Erigeron Forb Erythronium Forb Eupatorium Forb Eurybia Forb Euthamia Forb Fallopia Forb Fragaria Forb Galium Forb Gentiana Forb Geranium Forb Geum Forb Glechoma Forb Helianthus Forb Heliopsis Forb Hesperis Forb Hippuris Forb Hydrophyllum Forb Hypericum Forb Impatiens Forb Lactuca Forb Laportea Forb Lemnoideae Forb Lepidium Forb Leucanthemum Forb Lilium Forb Lobelia Forb Lotus Forb Lycopus Forb Lysimachia Forb Lythrum Forb Maianthemum Forb Medeola Forb Medicago Forb Melilotus Forb Mentha Forb Monarda Forb Monotropa Forb Myostis Forb Myrrhis Forb Nabalus Forb

54

Nepeta Forb Oenothera Forb Osmorhiza Forb Oxalis Forb Penstemon Forb Penthorum Forb Persicaria Forb Phlox Forb Pilea Forb Pilosella Forb Plantago Forb Podophyllum Forb Polygonatum Forb Polygonum Forb Potentilla Forb Prunella Forb Pycnanthemum Forb Ranunculus Forb Ratibida Forb Rudbeckia Forb Rumex Forb Sagittaria Forb Sanguinaria Forb Saponaria Forb Scutellaria Forb Silene Forb Sinapsis Forb Smilax Forb Solanum Forb Solidago Forb Sonchus Forb Stellaria Forb Symphyotrichum Forb Symplocarpus Forb Tanacetum Forb Taraxacum Forb Thalictrum Forb Thlapsi Forb Tragopogon Forb Trientalis Forb Trifolium Forb

55

Trillium Forb Tussilago Forb Verbascum Forb Verbena Forb Veronica Forb Vicia Forb Vinca Forb Viola Forb Athyrium Pteridophyta Dryopteris Pteridophyta Equisetum Pteridophyta Gymnocarpium Pteridophyta Matteuccia Pteridophyta Onoclea Pteridophyta Osmunda Pteridophyta Polystichum Pteridophyta Pteridum Pteridophyta Carex Sedges and rushes Eleocharis Sedges and rushes Juncus Sedges and rushes Scirpus Sedges and rushes Acer Woody plants Actea Woody plants Alnus Woody plants Amelanchier Woody plants Berberis Woody plants Betula Woody plants Carya Woody plants Caulophyllum Woody plants Ceanothus Woody plants Cornus Woody plants Corylus Woody plants Crataegus Woody plants Euonymous Woody plants Fagus Woody plants Fraxinus Woody plants Gaultheria Woody plants Hamamelis Woody plants Juglans Woody plants Lindera Woody plants Liriodendron Woody plants

56

Lonicera Woody plants Malus Woody plants Mitchella Woody plants Ostrya Woody plants Parthenocissus Woody plants Patanus Woody plants Physocarpus Woody plants Picea Woody plants Pinus Woody plants Populus Woody plants Prunus Woody plants Quercus Woody plants Rhamnus Woody plants Rhus Woody plants Ribes Woody plants Robinia Woody plants Rosa Woody plants Rubus Woody plants Salix Woody plants Smilax Woody plants Tilia Woody plants Toxicodendron Woody plants Tsuga Woody plants Ulmus Woody plants Vaccinium Woody plants Viburnum Woody plants Vitis Woody plants

57

Figure A1. Map of Southern Ontario, Canada showing the location of traps at study sites and their respective management types.

58

(a)

(b)

(c)

Figure A2. Boxplots of (a) the percentage of agricultural area, (b) the percentage urban area, and (c) the percentage semi-natural area within a 2000m radius of each trap for each locality type.

59