Copyright by Megan O'connell 2021

Copyright by Megan O’Connell 2021 The Dissertation Committee for Megan O’Connell certifies that this is the approved version of the following Dissertation:

PLANT-POLLINATOR INTERACTIONS IN THE FACE OF GLOBAL CHANGE

Committee:

Shalene Jha, Supervisor

Stanley Roux

Lawrence Gilbert

Alexander Wild

Thomas Juenger PLANT-POLLINATOR INTERACTIONS IN THE FACE OF GLOBAL CHANGE

Megan O’Connell

Dissertation Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

The University of Texas at Austin May 2021 Dedication

I dedicate this dissertation to anyone who is curious about pursuing the sciences but does not believe they can. To anyone who thinks they are not smart enough or feels they do not have the resources and support to pursue field work, research, and graduate studies. To anyone who does not see their likeness reflected in the images of scientists they see in the media, text books, and names of authors listed on publications. With training, we all can be scientists, we all can earn PhD’s, we all can pursue our curiosities about the world, measure its patterns, and marvel at its wonders. I dedicate my dissertation to anyone who dreams of being a scientist but is too intimidated to pursue their dream. On your behalf, I promise to actively work make my field a more welcoming, diverse, and inclusive community in all my future endeavors.

iv Acknowledgements

This dissertation would not have been possible without the tremendously generous support of so many people, but above all my graduate mentor, Dr. Shalene Jha. Without your open-mindedness and visionary imagination, I would not have felt seen and accepted in the sciences. You helped me carve a niche in the fields of ecology and conservation that feel true to who I am, and I have you to thank for so much of my personal and professional growth. I would also like to thank the many researchers and professionals who helped me get to where I am today: Dr. Andy Jones and Dr. Eric Manzane, you introduced me to one of my greatest loves: researching tropical forests; Dr. Antonio Castilla, you have always believed in me and kindly encouraged me through my self-doubts, I would not have a dissertation if it were not for the massive research efforts you undertook in Panama; Dr. Alex Wild, you gave me a chance to incorporate my inner artist and journalist into my career; Trevor Hance, you graciously continue to give me avenues through which I can bring my research to my community and make real, lasting impacts. I also give immense thanks to all the kind and generous members and friends of the Jha lab: Dr. Kim Ballare, Dr. Sarah Cusser, Dr. Nate Pope, Rebecca Ruppel, Dr. Nathan LeClear, Nick Ivers, Camila Cortina, Laurel Treviño, Dr. Hollis Woodard, Dr. Daniel Katz, Dr. Elinor Lichtenberg, Dr. Gabriella Pardee, Dr. Sean Griffin, Dr. Hannah Gray, Dr. Felicity Muth, Dr. Rodolfo Jaffe, Dr. Jay Banner; you all made me feel safe to ask questions and to be vulnerable with my ideas. I could not have completed my research without the incredibly hard work of my undergraduate and high school student researchers: Pragati Kore, Apoorva Magadi, Leticia Lee, Yadira Rodriguez, DJ Ojeda, Katie Pennartz, Amy Wrobleski, and Jen Schlauch. Immense, heartfelt thanks to my friends and team in Panama who were patient with my rudimentary Spanish and helped me achieve what felt like the impossible: Dr. Angie Estrada, Hilario Espinosa, Nelson Jaen, Maikol Guevara, Valeria Franco, Tyler Macy, Peter

v Marting, Dr. William Wcislo, Leonardo Simmons, Debbie Rivera, and Dr. Alonso Santos- Murgas. Thank you to my committee members who have given me crucial feedback throughout my dissertation: Dr. Stanley Roux, Dr. Alexander Wild, Dr. Lawrence Gilbert, Dr. Thomas Juenger; as well as the UT IB support staff for all the help over the years: Tamra Rogers, Sylvia Moore, Frances Lemear, Sean Schaffer, and Theresa Kelly. Additional thanks to my dear friends and family who have always believed in me through this wild journey: Dr. Jade Florence, Dr. Amanda Vaughn, Dr. Rose Stafford, Olivia Haun, Ash

Dionne, Iffy Roma, David McKay, Gabe Patterson, Clayton Noyes, Ross Woods, Walker Pickens, Gabe Miller, Jerod and Lauri Walz, Denise, Neil, Patrick, Sean and Kacy O’Connell. Lastly, thanks to my grandma and grandpa Fern and Larry O’Connell who are no longer with us: you forged this path into the sciences for me, you showed me what a strong independent woman looks like, you laid the groundwork that allowed me to be the first PhD in our family – we did it grandpa!

PLANT-POLLINATOR INTERACTIONS IN THE FACE OF GLOBAL CHANGE

Megan O’Connell, PhD The University of Texas at Austin, 2021

Supervisor: Shalene Jha

More than 80% of terrestrial plant species are dependent on animal pollinators to facilitate their reproduction and survival via pollen dispersal and pollen-mediated gene flow. With anthropogenic habitat destruction, urbanization, and climate change intensifying, the alteration and loss of pollination services may be one of the greatest threats global biodiversity faces today. Plant-pollinator interactions meet a myriad of synergistic challenges, both spatial and temporal, that impact their frequency and efficacy, ultimately altering the movement of pollen-mediated genetic diversity across landscapes and rendering tangible consequences for plant reproduction. Therefore, the ability for ecosystems to support diverse and robust pollinator communities, that can facilitate sufficient pollination services in quickly changing landscapes, may largely determine the future genetic health and survival of plant communities. The spatial impacts of land-use change and urbanization alter both density- dependent dispersal patterns and pollinator foraging behavior, while climate change may exacerbate these issues by further altering floral resource availability and foraging behavior temporally. To explore these dynamics we conducted extensive field surveys (Chapters 1, 2, 3), molecular analyses (Chapters 1, 2), and pollen analyses (Chapters 2, 3) across two systems: the tropical lowland forests of the Panama Canal region (Chapters 1, 2) and a

vii

network of urban gardens along the central coast of California (Chapter 3). We explored the scales at which pollen dispersal and pollen-mediated gene flow can be influence by deforestation (Chapter 1), finding measurable fine-scale effects in a multipaternal tropical tree species. We then added a temporal aspect to our tropical study system to explore how density-dependence may interact with climate change to impact pollination services after a plant-pollinator network experienced a discrete phenological shift (Chapter 2), finding that the distribution of genetic diversity and the robustness of plant-pollinator networks may play important roles in buffering plants from the negative effects of climatic extremes. We also investigated how the most extreme form of habitat degradation, urbanization, impacts pollinator foraging preferences across a network of urban gardens (Chapter 3), finding clear patterns of how pollinators utilize resource patches within cities as a function of the surrounding urban matrix and the richness of plant communities in these patches. Lastly, I present a portfolio of professional science media products I produced and/or co-produced throughout the course of my dissertation studies (Chapter 4), illustrating the importance of science communication for the fields of ecology and conservation, and the potential ways researchers can participate in the creation of compelling science media products.

viii

Table of Contents

List of Tables………………………………………………………………………………xi

List of Figures…………………………………………………………………………….xii

Chapter 1: Bee movement across heterogeneous tropical forests: multi-paternal analyses reveal the importance of neighborhood composition for pollen dispersal………………………………………………………………………….…...1

Abstract…………………………………………………………………….………...1

Introduction………………………………………………………………………...... 2

Methods………………………………………………………………………………8

Study System and Sample Collection…………………………………….....8

Genetic Analyses…………………………………………………………..11

Statistical Analyses………………………………………………………...13

Results……………………………………………………………………...14

Discussion………………………………………………………………….15

Chapter 2: Landscape genetic diversity and pollinator network specialization buffer plant reproduction and pollen-mediated gene flow from extreme climate events…………………………………………………………………….…29

Abstract……………………………………………………………………………..29

Introduction…………………………………………………………………………30

Results and Discussion………………………………………...……………………33

Methods……………………………………………………………………………..38

Study System and the 2015-2016 El Niño Southern Oscillation…………..38

Study Species, Neighborhood Traits, and Phenology……………………...39

Pollinator Observations and Community Composition, Pollen Load Analyses, and Netowrk Construction………………………………….41

Reproductive Success and Genetic Analyses……………………………...43

Statistical Analyses and Models…………………………………………...44

Chapter 3: Reap what you sow: local plant composition mediates bumblebee foraging patterns within urban garden landscapes………………………………….53

Abstract……………………………………………………………………………..53

Introduction…………………………………………………………………………54

Methods……………………………………………………………………………..59

Study Region and Garden Metrics…………………………………………59

Pollinator Survey…………………………………………………………..61

Bumble bee pollen loads…………………………………………………..62

Reference Collection………………………………………………………63

Impacts of local and landscape features on pollinator diversity…………..64

Within Garden Preference Analysis……………………………………….66

Results………………………………………………………………………………67

Discussion…………………………………………………………………………..70

Conclusions…………………………………………………………………………77

Chapter 4: Science communication, media, and public engagement………………….....90

Abstract……………………………………………………………………………..90

Science communication films: descriptions and links……………………………...91

References………………………………………………………………………………...98

List of Tables

Table 1.1: Chapter 1: Results of Linear Mixed Effects Models……………………...... 22 Table S1.1: Chapter 1: Study region, plant neighborhood, and pollinator community metrics………………………………………………………….26 Table S1.2: Chapter 1: Genetic summary of study regions…………………………..….26 Table 2.1: Chapter 2: Global Regression Models and Results…………………………46 Table S2.1: Chapter 2: Results of Averaged Models………………………………….....50 Table 3.1: Chapter 3: Top Regression Models and Results………………………….…80 Table 3.2: Chapter 3: Results of Pollen Preference Analyses……………………….....81 Table S3.1: Chapter 3: Top models from MuMin model selection..…………………….86 Table S3.2: Chapter 3: Top models from MuMin model selection..…………………….88

List of Figures

Figure 1.1: Chapter 1: Study system imagery and sample pollen dispersal map……….23 Figure 1.2: Chapter 1: Boxplot of pollen dispersal distances per pollinator species ranked from small to large body sizes………………………………………24

Figure 1.3: Chapter 1: Regression plots of the influence of individual and landscape factors on pollen dispersal distance moments ……………………………...25 Figure S1.1: Chapter 1: Regression plots of sire diversity calculation methods………….27 Figure S1.2: Chapter 1: Density distribution of pollen dispersal events……………….…28 Figure S1.3: Chapter 1: Correlograms from spatial autocorrelation analysis………….…28

Figure 2.1: Chapter 2: Regression plots of seed set results……………………………..47

Figure 2.2: Chapter 2: Regression plots of dispersal distance results…………………...48

Figure S2.1: Chapter 2: NMDS Community Dissimilarity plot and SIMPER analyses….49

Figure S2.2: Chapter 2: All effects regression plots of models.………………………….52 Figure 3.1: Chapter 3: Boxplot of proportions of pollen loads collected from in and out of gardens……………………………………………………………….82 Figure 3.1: Chapter 3: Boxplots of plant availability versus usage by bumblebees for crop, ornamental, and weed species…………………………………….83 Figure 3.2: Chapter 3: Regression plots of the inluence of urban cover and garden plant species richness on pollen usage……………………………………...84 Figure 3.3: Chapter 3: Regression plots of the inluence of urban cover and garden plant species richness on usage of crop, ornamental, and weed species……85 Figure S3.1: Chapter 3: Top ten pollen species identified in pollen loads………….……89

xii

Chapter 1: Bee movement across heterogeneous tropical forests: multi- paternal analyses reveal the importance of neighborhood composition for pollen dispersal

ABSTRACT

Animal pollination is critical for maintaining the reproduction and genetic diversity of

many plant species, especially those in tropical ecosystems. Despite the threat to

pollination posed by tropical deforestation, it remains an understudied process. In

particular, little is known about these dynamics in multi-paternal, successional plant

species whose fruits can contain substantial genetic diversity. Given the importance of

successional plants in reforestation, quantifying the factors that impact their reproduction

is essential for understanding plant gene ﬂow in the context of global change. In this study,

we investigated pollen-mediated gene ﬂow at the multi-paternal fruit level to quantify how

tropical pollinators navigate and mediate gene ﬂow in altered forests. Utilizing

microsatellite genotyping and paternity analyses, we revealed that distinct plant

neighborhood and individual factors drive pollen dispersal at the intra-fruit scale. Variance in pollen dispersal distances was greater for neighborhoods with higher conspeciﬁc density, indicating that density dependent reproductive patterns play a role at this scale.

Additionally, both the diversity and evenness of sires mediated by a single

O'Connell, M.C., Castilla, A.R., Lee, L.X., & Jha, S. (2018). Bee movement across heterogeneous tropical forests: multi‐paternal analyses reveal the importance of neighborhood composition for pollen dispersal. Biotropica, 50(6), 908-918. M. O’Connell was responsible for designing research questions, performing research, analyzing data, writing the dissertation, and writing the manuscript.

pollinator were affected by the size of the mother tree, that is, larger mothers received

pollen from a less diverse, less even pool of sires per fruit. Pollinator body size was not

found to be a significant driver of pollen dispersal, indicating that both small- and large- bodied pollinators were equally important pollen dispersers at this scale. By exploring patterns of variation at the intra-fruit level, we show that conspecific density and tree size significantly impact multi-paternal pollen-mediated gene flow, reinforcing the importance of investigating intraspecific, intra-individual variance in plant reproduction.

INTRODUCTION

The tropics are home to unparalleled levels of biodiversity and complex networks of plant–

animal interactions, much of which may be threatened by increasing levels of deforestation

and land alteration. Given that in most studied tropical forest communities, more than 90

percent of tropical plant species are animal-pollinated (Ollerton et al. 2011), the disruption of pollen dispersal services represents one of the greatest threats to tropical biodiversity posed by anthropogenic land conversion (Aizen & Feinsinger 1994, Hadley & Betts 2012,

Hansen et al. 2013). Many animal-pollinated plants are self-incompatible and obligately dependent on their pollen dispersers (Aguilar et al. 2006); therefore, pollen-mediated gene

ﬂow and the subsequent diversity of sired offspring are critically impacted by a pollinator’s ability to effectively disperse pollen between individual plants. As deforestation alters the spatial aggregation and size of plants, these changes to the arrangement of individuals within plant neighborhoods can inﬂuence mutualistic interactions (Kunin 1997, Jones &

Comita 2008); thus, it is possible that neighborhood and individual traits could

influence not only seed production but also pollen-mediated gene flow. At the neighborhood level, positive density dependence is a ubiquitous phenomenon in plant reproduction and is thought to be a strong driver of pollen dispersal and siring patterns (Ghazoul 2005, Dick et al. 2008). At the individual level, traits, such as plant and pollinator size, may also act to influence seed set and pollen dispersal (Lowe et al. 2015, Castilla et al. 2017). Despite the potential importance of such neighborhood- and individual-level traits in mediating plant reproduction across human-altered regions, much remains unknown about the drivers of pollen-mediated gene flow.

Understanding pollen dispersal and siring patterns is particularly critical for tropical pioneer plants, as these species are often dominant in secondary forests, which are becoming increasingly common across the tropics. Such species often have very different life strategies compared to primary forest species (Snow 1965), the latter of which have been the primary focus of many past plant gene ﬂow studies (reviewed in Dick et al. 2008).

Specifically, many pioneer species have multi-seeded fruits that could potentially be fertilized by several sires; thus, the post-fertilization dispersal unit could transfer very different genetic information than that of a single-seeded fruit. Therefore, pioneer species may display unique, but overlooked patterns of pollen movement and siring across altered forests (Rhodes et al. 2017). The role of pollinators may also be different for multi-seeded pioneer plant species, as the spatial scales at which pollinator foraging affects pollen- mediated gene ﬂow include not only the population and individual tree levels (Vranckx et al. 2011), but also the individual fruit, potentially yielding different gene ﬂow patterns depending on the ecological scale at which they are measured (Breed et al. 2015). Despite 3

the high occurrence of multi-paternity in pioneer plant species, and the growing importance

of understanding the regeneration of tropical forests in global conservation efforts, few

studies have examined the variation in siring patterns or pollen dispersal distances within

multi-paternal fruits in these forests (Davies et al. 2015, Ramos et al. 2016) and fewer

have explored its relationship to attributes of (1) the pollinator, (2) the plant, and (3) the

plant neighborhood (Breed et al. 2015, Castilla et al. 2017, Rhodes et al. 2017).

To begin, while we know that pollinator traits can inﬂuence seed production in

multi-paternal plants (Castilla et al. 2015, 2017, Cranmer et al. 2012), our understanding

of how pollinator attributes influence pollen-mediated gene flow in multi-paternal species is limited (Breed et al. 2015, Hasegawa et al. 2015). Past work has suggested that pollinator traits such as body size and foraging behavior can be important drivers of plant reproductive output (Sahli & Conner 2007, Mitchell et al. 2013). Assertions about optimal foraging and pollinator efficacy based on body size have led to the general assumption that larger-bodied pollinators should perform better at various efficacy metrics such as dispersing pollen further distances and depositing more pollen (Lindauer 1957, Stout 2000,

Greenleaf et al. 2007). While some studies support this theory (Hasegawa et al. 2015), many have historically compared pollinators that are functionally and morphologically very different (Breed et al. 2015, Rhodes et al. 2017); or when comparing more functionally similar pollinators, studies have found a more complicated relationship between size, behavior, and effectiveness (Stout 2000). In tropical forests, Castilla et al.

(2017) found that although larger-bodied pollinators set more seeds per visit, smaller- bodied pollinators visited plants more frequently and dispersed pollen similar distances 4

when compared to large-bodied pollinators. Across studies, what remains unknown is how

morphologically and functionally similar tropical pollinators contribute to siring and

pollen dispersal within multi-paternal fruits, particularly in ecosystems where individual

plant traits and neighborhood composition may be variable.

At the scale of the individual plant, ﬂoral display size could be a critical trait in

determining pollen dispersal distance and siring diversity. For example, past work shows

that pollinators may visit larger trees for longer foraging durations within a tree, possibly

increasing the transfer of self-pollen or pollen from nearby, closely related individuals,

and likely promoting greater levels of geitonogamy and inbreeding (Karron & Mitchell

2012, Mitchell et al. 2013). Alternately, trees with larger ﬂoral displays could receive

more occasional floral visitors than smaller trees and thus could support long-distance dispersal events and act to increase genetic admixture (Makino et al. 2007). For tree species that have developed complete self-incompatibility, these influences on within-tree foraging may have significant impacts on the genetic variability of new cohorts. In tropical forests, past research has indicated that larger trees may produce more fruits and seeds per tree but can have a lower proportion of viable seeds per fruit compared to smaller trees (Castilla et al. 2015). These patterns suggest that individual plant traits may contribute to variance in pollinator foraging behavior, differentially impacting seed composition at the fruit, individual, or regional level.

In addition to individual plant traits, plant neighborhood features may also inﬂuence pollen dispersal and siring in plants. Plant conspeciﬁc density has been found to play a critical role in the probability of successful outcrossed mating events, and the 5

survival of future, genetically diverse plant cohorts (Ghazoul 2005, Ismail et al. 2012,

Comita et al. 2014). Spatially isolated trees may engage in fewer mating opportunities

and are often assumed to be more genetically distinct; thus, the ability of pollen dispersers

to facilitate long-distance dispersal events from less dense patches may be particularly

important in the movement of novel genetic diversity to and from spatially isolated plant

neighborhoods (Nei 1972, Hutchison and Templeton 1999, Vekemans and Hardy 2004).

Mechanistically, spatial isolation of plants may reduce pollinator efﬁcacy and plant ﬁtness

through increased transference of self-pollen and biparental inbreeding, which can lead to

increases in seed abortion rates or seed cohorts with reduced ﬁtness (Hufford and Hamrick

2003, Breed et al. 2012, 2014, Rhodes et al. 2017). Interestingly, this may lead to

neighborhoods of high local kinship, a critical trait where neighboring individuals exhibit

high levels of relatedness (Loiselle et al. 1995). In combination with conspeciﬁc density,

two tropical studies have found that high local kinship in dense neighborhoods can

counteract the beneﬁts of proximity to reproductive output, sometimes increasing the transfer of pollen that is too genetically similar, consequently increasing the number of aborted seeds and/or fruits in those neighborhoods (Jones & Comita 2008, Castilla et al.

2015). Overall, the combination of these spatial and genetic patterns may be particularly relevant in fragmented tropical forests where conspeciﬁc density and kinship can be highly variable (Jha & Dick 2010) and natural population densities tend to be low

(Duminil et al. 2016).

Finally, while a number of ﬁelds now emphasize the importance of analyzing ecological variance and measures of intra-individual and intraspeciﬁc variation in 6

ecological systems (Violle et al. 2012), many studies still focus on maximal, mean, or

single- seed measures and sire diversity indices to describe pollen dispersal processes

(Lowe 2005, Lowe et al. 2015) . Such analyses may not be able to capture the true breadth

of genetic information transferred to each new generation via pollen dispersal, particularly

in the case of plants with reproductive strategies that involve multi-seeded, multi-paternal

fruits (Ghazoul 2005, Lowe et al. 2015). Regarding pollinators, mean and maximal

foraging measures may not provide an accurate depiction of a pollinator’s typical foraging

behavior (Roubik 1989, van Nieuwstadt & Iraheta 1996), nor the variation that exists in an individual pollinator’s response to landscape alteration (Wenner et al. 1991). Given these factors, it is likely that changes in pollen dispersal patterns due to land-use and global change described at the plant species or regional scale may exhibit a very different distribution from those described at the individual or intra-individual scale. Thus, to capture patterns of genetic diversity and the reproductive strategies of functionally different plant species more completely, it will be necessary to explicitly analyze variance, diversity, and evenness, especially at ﬁner ecological scales such as the intra-individual, and even intra-reproductive unit level (Bolnick et al. 2011, Breed et al. 2012, Breed et al.

2014, Breed et al. 2015).

In this study, we investigate the impacts of pollinator, plant, and plant neighborhood traits on pollen-mediated gene flow for the tropical pioneer tree, Miconia affinis. Specifically, we ask whether pollen dispersal distances and the transfer of genetic diversity per pollinator visit to a multi-seeded, multi-paternal fruit differ across different pollinator body sizes, plant sizes, conspecific tree densities, and local kinship levels. To 7 take advantage of the multi-paternal fruit structure of this species, we utilize a unique approach and measure several statistical moments and multiple diversity indices per pollen dispersal event. We examine three facets of pollen dispersal and three facets of sire diversity per pollinator visit: (a) the mean pollen dispersal distance, maximum pollen dispersal distance, and standard deviation of the pollen dispersal distance and (b) the raw sire counts, the Chao estimated sire diversity, and sire evenness. We hypothesize that (1) variation in pollen dispersal distances exhibited in a single visit is enhanced by plant conspecific density, local kinship, mother size, and pollinator body size. Particularly, we predict that variation in pollen dispersal distance will be greatest in more isolated trees that are visited by larger-bodied pollinators. We also hypothesize that (2) sire diversity exhibited in a single visit is driven by plant local kinship and pollinator body size, but not by plant conspecific density or mother size. Specifically, we predict that sire diversity will be greatest in trees within lower kinship neighborhoods and for visits mediated by larger- bodied pollinators.

METHODS

Study System and Sample Collection — For much of the past two centuries, the moist lowland forests of the Panama Canal watershed have been heavily impacted by anthropogenic development; in 2001, it was estimated that only 54 percent of the original forest remained while more than 43 percent had been converted to pasture or shrubland

(Condit et al. 2001). Our study system includes three study regions that exist across a

~3370-ha area and include 1157 individuals of the tropical pioneer tree M. afﬁnis

(Melastomataceae; 3–6 m in height). This tree species has an extensive distribution ranging

from Mexico to Brazil and inhabits a wide range of habitats, from primary and unaltered

secondary-growth forest to highly fragmented forest and grassland systems (further

system description can be found in Table S1). M. afﬁnis is an ideal study species due

to its well-studied ecology (Luck & Daily 2003, Jha & Dick 2010, Castilla et al. 2015), known colonization history in the area (Castilla et al. 2016), and availability of genetic tools to examine pollen dispersal (Jha & Dick 2009). The species is dependent on buzz pollination conducted by a suite of native bee species, is hermaphroditic, and is self- incompatible (Jha & Dick 2010) (Fig. 1A). Mature trees of

M. afﬁnis display 1–3 ﬂowering events during the Panamanian dry season (January to

June), blooming for ~2 d per ﬂoral event. The globose fruits of M. afﬁnis develop to maturity between May and September and are mainly dispersed by native small-bodied frugivorous birds (Luck & Daily 2003, Jha & Dick 2008, 2010) (Fig. 1A).

Single-visit pollination experiments were conducted in 2013 (as described in

Castilla et al. 2017): Five focal inflorescences on randomly selected mother trees (N = 75 mother trees; N = 375 inflorescences) were bagged until the day of flowering and the bags were removed to allow a single pollinator to visit each inflorescence. Each pollinator could visit several flowers within one focal inflorescence; when the pollinator departed, it was collected and stored on 70 percent ethanol for subsequent identification and to measure its intertegular distance as a proxy for body size (ITD; sensu Cane 1987). After a single pollinator visit, each inflorescence was re-bagged until fertilized fruits matured, at which point all fertilized fruits were collected. Dissections on the collected fruits and 9

seed viability counts were performed to determine the seed set (the proportion of viable

seeds out of the total number of seeds) produced by each observed single pollinator visit.

Leaf tissue was collected from all adult M. afﬁnis trees (N = 1157) within a 2 km radius of each study region’s geographic centroid for use in paternity analyses, as this distance has been shown to capture a large portion of the pollen dispersal kernel for the species

(Jha & Dick 2010).

Two individual traits and two plant neighborhood traits were measured per pollination event. Individual traits investigated included the diameter at breast height of the mother trees (dbhmother) as a proxy of ﬂoral display size (correlation validated for this

species in Castilla et al. 2017, sensu Kettle et al. 2011) and the size of each pollinator as measured by their intertegular distance (ITDpollinator), which is often correlated with some

indices of dispersal ability (Greenleaf et al. 2007). The plant neighborhood traits included nearest neighbor distance (NNDmother; the average spatial distance to the ten nearest

conspeciﬁc trees to the measured mother) as an index of conspeciﬁc density, and local

kinship of the mother tree, deﬁned as the mean Loiselle kinship coefﬁcient (Fij; the

pairwise comparisons between the mother tree and all its neighbors within a 400 m radius

sensu Castilla et al. 2017, Hardy et al. 2006). In previous analyses in our study system

(Castilla et al. 2016), a 400 m radius was determined to be an important threshold within

which mother trees displayed positive ﬁne-scale spatial genetic structure, indicating higher

kinship values within this area than expected (Fig. S3). Past work in this system (Castilla

et al. 2015, 2016) and others (Jones & Comita 2008) indicates that these speciﬁc individual

and neighborhood traits have the potential to inﬂuence pollen movement. While forest 10

cover was evaluated in our study regions (Table S1), it correlated highly with NND;

therefore, we opted to retain NND instead as we know from past work in the region that

NND can inﬂuence seed viability and pollen dispersal, even when forest cover is relatively

homogeneous (Castilla et al. 2016).

Genetic Analyses—In order to quantify variation in sire diversity and pollen dispersal distances at the fruit level, we subsampled fruits from the single-visit experiments, only including visits that resulted in >10 viable seeds per fruit, from which ten seeds were randomly selected for the genetic analysis. We also randomly subsampled the resulting fruits to only include one fruit per tree, to minimize resampling of individual mother trees. For our paternity analyses, we only utilized fruits for which we were able to successfully genotype ten viable seeds (N = 34 single fruits; N = 340 seeds). We set the

threshold at ten viable seeds, as previous work in this system has indicated that M. afﬁnis

exhibits high sire diversities at the inﬂorescence level; therefore, we wanted to ensure that

we had an even and sufﬁcient sample size to estimate sire diversity at the fruit level

(Castilla et al. 2015). Further, given that more than 60 percent of all fruits visited by a

single pollinator had greater than ten viable seeds, setting this minimum threshold did

not preclude us from describing a large portion of fruits in our study system. These visits

were conducted primarily by social bee species within the Meliponini tribe; however, the

14 species varied greatly in size, exhibiting ITD measures that ranged from 0.91–7.72

mm (x = 2.19 mm, SD = 1.09 mm). We did not explore pollinator species-level effects given our small per-species sample size and the inability to detect species-level effects in

larger data sets from the study region (See Fig. S1, Castilla et al. 2017). DNA was

extracted from individual seeds using the DNAzol Genomic DNA Isolation Reagent

extraction protocol, and from leaves sampled from all 1157 adult trees using the CTAB

protocol, and then PCR processed using Qiagen Multi- plex PCR Master Mix (Castilla et

al. 2015). The seeds and leaves were genotyped at 12 highly polymorphic microsatellite

loci (Le Roux & Wieczorek 2008, Jha & Dick 2009) and later narrowed to eight

polymorphic markers (Micaff-5, Micaff-7, Micaff-8, Micaff-14, Micaff-16, Micaff-

19, B102, B109) to remove markers with inconsistent PCR recovery or possible presence of null alleles.

Using the genotypes from all ten seeds per visit, we used HP- RARE (Kalinowski

2005) to measure the allelic richness and private allelic richness of each fruit (Table S2).

We then conducted likelihood-based paternity assignments using CERVUS, where sire assignments were only made to seeds with a conﬁdence criterion of >0.80 (Marshall et al.

1998). Paternities were also assigned at 90 percent conﬁdence, but because this reduced our sample size to less half of the original dataset, we focused our analyses on data from the >80 percent conﬁdence assignments (282 seeds), as per other past plant dispersal studies (e.g., Hardesty et al. 2006, Bitten- court & Sebbenn 2007, Lander et al. 2010).

Given that sampling efforts are ﬁnite and limited to a 2 km radius in this study, unassigned parentage is likely the result of dispersal events outside of the sampled plant neighborhoods, as assumed in past studies (Bittencourt & Sebbenn 2007, Bacles & Ennos

2008, Dick et al. 2008). From these assignments, two groups of measures were generated:

(a) three dispersal distance moments and (b) three sire diversity measures. For the dispersal 12

distance moments, we calculated the linear distances between mother and the assigned

pollen donor from the CERVUS analysis (Fig. 1D), and these were measured at three

statistical moments: the mean, the standard deviation, and the maximum pollen dispersal

distances per fruit. For the sire diversity measures, we measured three values: the raw sire

counts per fruit from the CERVUS output (sensu Pelabon et al. 2015), the Chao diversity estimator (Chao 1984) for the CERVUS output (to account for small sample sizes), and sire evenness, deﬁned as the proportional representation of individual sires within a seed set, to account for the likely uneven representation of sires in our progeny arrays (Mitchell et al. 2013). As a second index of sire counts and diversity, we also calculated the number of full versus half sibships per fruit using COLONY (Jones and Wang 2010) and similarly calculated half sibship Chao diversity; we found the two quantiﬁcations of sire diversity to be very similar and therefore present the COLONY results in the Supplemental

Information (Fig. S1) and focused on the CERVUS-based paternity analysis when discussing our results. Finally, for descriptive purposes, we also used GenAlEx (Peakall &

Smouse 2006) to generate a study region-level genetic summary including: N (the total number of seeds that were successfully assigned fathers per study region), Na (the average

number of alleles per locus), He (the expected heterozygosity), and Ho (the observed

heterozygosity)(Table S2).

Statistical Analyses—We used linear mixed-effects models to analyze the inﬂuence of

individual explanatory variables (dbh and ITD) and plant neighborhood explanatory

variables (NND and local Kinship) on the two groups of response variables: (a) dispersal

distance moments (mean, maximum, standard deviation) and (b) sire diversity measures 13

(raw count, Chao diversity, evenness). We also included study region and mother tree as

random factors, with mother tree nested within study region. All our response variables

and NNDs were log-transformed. We found no interactions between variables in our models (Table S3).

RESULTS

Our CERVUS analysis successfully assigned fathers to 83 percent of the 340 sampled seeds

with a conﬁdence criterion of >0.80 (282 seeds). The subsequent dispersal distances

measured from these assignments ranged from2.8 m to 3600 m(l = 905 m, SD = 953 m)

with 66 percent of the pollen dispersal events coming from fathers that were 1 km or closer

to the mother trees (Fig. S2). Pollinator body size did not affect pollen dispersal distances

in our study system (Fig. 2, Table 1). We found a negative relationship between NND and

the standard deviation in pollen dispersal distances per visit/fruit (Table 1, Fig. 3A), but no

signiﬁcant relationships between any of the explanatory variables and either the mean or

maximum pollen dispersal distance (Table 1). In other words, the multiple pollen dispersal

distances from a single visit were more similar to one another (lower in standard deviation)

in spatially isolated trees than in spatially aggregated trees. The raw sire counts from the

CERVUS paternity analysis yielded anywhere from 3 to 9 fathers per visit/fruit (l = 6.4,

SD = 1.74), the Chao sire diversities estimated from the same output ranged from 4 to 10

fathers per visit/fruit (l = 7.3, SD = 1.77), and sire evenness ranged from 0.55 to 1.0 (l

= 0.92, SD = 0.1). Both the raw sire counts and Chao sire diversities were signiﬁcantly

affected by the size of the mother tree (dbh), with raw sire counts and sire diversity

signiﬁcantly decreasing as the mother dbh increased (Table 1, Fig. 3B). Sire evenness per

dispersal event also showed a negative relationship with dbh (Table 1, Fig. 3C).

DISCUSSION

In this study, by investigating multiple measures of pollen dispersal and siring in a multi- paternal species, we determined key drivers of pollen-mediated gene flow for a ubiquitous successional plant species across heterogeneous study regions. Specifically, we found that variation in pollen dispersal distances per visit was significantly greater for trees that were more spatially aggregated. We also show that measures of sire diversity and evenness are most influenced by the size of the mother tree, with sire diversity and evenness significantly decreasing as mother tree size increases. None of our measures of multi-paternal reproductive dynamics were significantly affected by the body size of the pollinator, indicating that in the pollinator community we measured, both large-bodied and small-bodied pollinators are contributing similarly to pollen-mediated gene flow across fragmented tropical landscapes.

First, we found that for our study species, a tropical pioneer tree, there was tremendous variation in pollen dispersal and siring mediated by a single pollinator visit, irrespective of pollinator size. In fact, many of the smaller-bodied pollinators travelled just as far, if not further, than some of the larger-bodied pollinators, depositing pollen from a wide variety of distances and from multiple fathers in a single ﬂoral visit. For example, we found a single small-bodied Trigona muzoensis (ITD = 1.72 mm) deposited pollen collected from an estimated nine sires that were anywhere from 10 m to 2700 m away from

15 the mother. A much larger bee, Xylocopa fimbriata (ITD = 7.72 mm), deposited pollen from an estimated seven sires that were similarly about 12 to 2000 m away from the visited mother. In terms of traditional perspectives regarding pollinator body size and maximal foraging behaviors, these results contradict the assumption that large-bodied pollinators will always travel further distances than small-bodied pollinators and that, despite forest features, larger-bodied pollinators will consistently travel the longest distances (Greenleaf et al. 2007). These results support our previous findings that pollinator body size is not the primary driver of pollen dispersal distance in our study system (Castilla et al. 2017). Other studies have also begun to present mixed findings regarding pollinator body size, indicating that it may be too simplistic of a portrayal of a pollinator species’ biology to accurately describe their foraging response to landscape change. Stout (2000), for example, found that although larger bees visited more flowers, smaller bees were more effective at triggering floral mechanisms that release pollen in Scotch broom (Cystisus scoparius). Our results resonate with a number of studies that indicate that individual functional traits, such as pollinator body size (ITD), are not the best predictors of pollinator behavior and pollen movement in heterogeneous landscapes (Roubik 1989, Stout 2000,

Makino et al. 2007), and that future studies may beneﬁt from including more detailed data on pollinator foraging behaviors.

Instead, we found that the variance in distances from which pollen was dispersed per visit was significantly predicted by the degree of spatial isolation of the mother tree; specifically, our data reveal that variation in the sire distances per pollinator visit increased in neighborhoods with higher conspecific density (lower NND). In other words, more 16

spatially aggregated mother trees received pollen from sires that were located at more

variable distances from the mother tree. This ﬁnding is in contrast to Duminil et al. (2016),

which found that pollen dispersal distances consistently increased with decreasing

conspeciﬁc density, but corroborates patterns found by Ismail et al. (2012) which indicate

that while plants in more dense patches tend to receive pollen from donors within their

own patch, they receive this pollen from a larger number of donors from a variety of

distances within the patch. From the plant’s perspective, spatially aggregated mothers may

beneﬁt from this variance in pollen dispersal distance, as they are likely receiving pollen

from a higher variety of conspeciﬁc neighborhoods. Past work has indicated that both the

kinship and the density of different conspeciﬁc neighborhoods can impact fruit production

(Jones & Comita 2008); speciﬁcally, they found that high conspeciﬁc density can interact

negatively with high local relatedness by counteracting positive density dependent forces

in fruit set with higher biparental inbreeding and thus higher fruit abortion rates. Previous

research in our study system has also found a negative interaction between local kinship

and conspeciﬁc density, which also acted to reduce reproductive output and increase variance in mean seed viability with increasing conspeciﬁc density and local kinship

(Castilla et al. 2015). While we did not ﬁnd a similar interaction between density and kinship in our analysis of multiple seeds per fruit, we did document high variance in pollen dispersal within a fruit, and we reveal that density dependent patterns also exist for variance in pollen dispersal, with potential impacts on post-fertilization gene ﬂow in other multi-paternal systems.

This density dependent relationship with pollen dispersal variation may also be 17

due to the way pollinators forage in resource-dense versus resource-poor patches in the

short 2-d, mass-ﬂowering period of our study species (Delmas et al. 2015). From the

pollinator’s perspective, spatially aggregated patches may act as epicenters of foraging

activity, attracting pollinators, from both near and far, speciﬁcally those who have noted

them as reward centers in their spatial memory (as seen in ﬂight cages, Burns & Thomson

2005). Also, breakdowns in predicted nearest neighbor foraging patterns have been

observed to depend not only on conspeciﬁc densities in remnant patches, but the spatial

aggregation of pollen sources within these patches (Dick et al. 2008). Interestingly, our

pollen dispersal data, which documented a few long-distance and several short-range

foraging events for almost all the observed pollinators, provide rare individual-level

tracking and support for previous hypotheses that the majority (>75%) of bee foraging

activity occurs within the lower 40 percent of a species’ maximum foraging range (Roubik

1989). Several of these previous theories had either been based on con- trolled,

experimental conditions (van Nieuwstadt & Iraheta 1996, Zurbuchen et al. 2010), or ﬁeld-

based observations that were designed to measure only maximum foraging distances

(Roubik & Aluja 1983, Knight et al. 2005), but no studies were aimed at capturing the

breadth of distances pollinators travel during normal foraging bouts for multiple species,

particularly in heterogeneous forests. Although we were unable to separate the potential

contribution of secondary pollen transfer to the long-distance dispersal events we measured, its impacts may be limited due to the short time window (~2-d flowering events) pollinators must forage at M. affinis. Overall, our measures of variance in dispersal distances bolster previous findings of density dependence, detect patterns that were not 18

apparent via mean and maximum measures, and help provide a more complete depiction

of how pollinators are navigating these heterogeneous landscapes.

Third, we found that the diversity and evenness of sires per pollinator visit were

significantly influenced by the size of the mother tree (dbh): Both sire diversity and sire evenness decreased as the size of the mother tree increased. Given that the number of inflorescences per tree strongly correlates with the size of the tree in this system (Castilla et al. 2017), dbh of M. affinis mothers serves as a proxy for floral display size. Taken together, these results indicate that larger M. affinis individuals receive less diverse pollen

from a less even group of sires compared to smaller trees. Pollinators have been found to

respond to tree ﬂoral display size independent of local conspeciﬁc density (Makino et al.

2006); therefore, potential explanations for this pattern could be that ﬂoral display size independently alters pollinator foraging patterns at the individual tree and fruit level.

Speciﬁcally, larger ﬂoral displays may promote longer foraging bouts at a single tree, increasing the transfer of self or non-legitimate pollen, thereby decreasing female reproductive success (Ghazoul 2005, Jones & Comita 2008, Brys & Jacquemyn 2010).

Previous ﬁndings in other systems have also indicated that increased ﬂoral display sizes may lead to increases in near-neighbor matings and reductions in the number of sires due to changes in pollinator foraging patterns, regardless of pollinator species or size (Stout

2000). More speciﬁcally, Mitchell et al. (2004) and Karron et al. (2003) found that several

species of bees that pollinate Mimulus ringens strongly responded to increased ﬂoral

display size: Trees with larger ﬂoral display sizes attracted more ﬂoral visitors, but these

visitors stayed at individual trees for longer and visited more ﬂowers per plant as ﬂoral 19

display size increased, effectively reducing siring success and increasing self-fertilization.

Our results could similarly indicate that M. affinis pollinators are spending more time foraging at trees with larger floral displays, visiting fewer potential pollen donors, resulting in reduced sire diversity and evenness within fruits at such trees. Therefore, similar to our conspecific density and dispersal distance findings, if larger trees act as resource epicenters for pollinator foraging, we may be detecting a trade-off between attracting sufficient pollinators to an area to promote gene flow and receiving sufficiently diverse pollen per pollinator visit (Mitchell et al. 2013). Additionally, our findings support previous suggestions by Stout (2000) that while floral display size may alter pollinator visitation and pollen dispersal rates, it does so independently of pollinator body size, indicating that bees of all sizes respond similarly to floral display size. Overall, these results highlight the importance of diverse tree sizes to promote variation in pollen dispersal and reproductive processes and to prevent erosion of genetic diversity in fragmented tropical plant communities (Ellstrand 2014).

As secondary and fragmented forest structures are becoming increasingly common across tropical regions, it is critical that we improve our understanding of how forest structure impacts species interactions and the reproductive dynamics of

successional plant communities (Girao et al. 2007, Magrach et al. 2014, Lowe et al.

2015). In this study, we found that two key factors, NND and mother tree size,

significantly influenced pollen dispersal and sire diversity at the within-fruit level and thus should be taken into consideration when assessing pollinator-mediated gene flow.

This ﬁnding is novel as it highlights the relevance of both plant neighborhood (conspeciﬁc 20

density) and individual traits (tree size) as drivers of within-individual variance in pollen dispersal and sire diversity across heterogeneous forests. We specifically note that maintaining an array of conspecific tree sizes, not just large trees, in a forest can support high levels of sire genetic diversity and evenness. We also found that pollinators of all body sizes promoted high multi-paternity levels and long-distance gene flow, indicating that pollinator conservation decisions should not necessarily rank species prioritization based on size alone. Importantly, given that we found different drivers of pollen distance and sire diversity when examining multiple seeds in multi-paternal fruit, it is important that we begin to explore a broader array of plant reproductive systems and life history strategies in our investigation of landscape genetics processes for conservation. (Kremen

2005). Finally, our results also illuminate the need to evaluate dispersal and genetic indices beyond mean values and highlight the importance of future work to develop additional methods that capture variation in ecological function.

Table 1.1 Results of linear mixed effects models run for the two response variable groups: (1) dispersal distance moments (standard deviation, mean, maximum) and (2) diversity metrics (the raw counts and Chao extrapolated sire diversity measures, sire evenness). Independent variables tested include nearest neighbor distance (m) (NND), the local kinship of the mother tree, the size of the mother tree (cm) (dbh) and the 22intertegular distance of the pollinator (mm)(ITD). Results are presented as the estimated strength and direction of the relationship, and the t- and P-values. Df = 34. *Signiﬁcant relationships in are bold.

Individual Factors Neighborhood Factors Dispersal Distance dbh IDT NND Mother Kinship Est. 0.195 0.149 0.325 0.025 SD t 1.529 1.167 2.573 0.200 P 0.136 0.251 0.015* 0.842 Est. 0.070 0.070 0.192 0.009 Mean t 0.520 0.517 1.433 0.072 P 0.606 0.608 0.161 0.943 Est. 0.007 0.072 0.041 0.026 Max t 0.057 0.632 0.329 0.232 P 0.955 0.533 0.746 0.818 Sire Diversity dbh IDT NND Mother Kinship Est. 0.131 0.019 0.041 0.052 Raw t 2.613 0.380 0.822 1.098 P 0.013* 0.706 0.417 0.280 Est. 0.095 0.023 0.030 0.047 Chao t 2.192 0.524 0.704 1.139 P 0.035* 0.592 0.486 0.263 Est. 0.059 0.007 0.009 0.031 Evenness t 2.951 0.354 0.446 1.633 P 0.006* 0.726 0.659 0.112

Figure 1.1 (A) A Melipona panamica bee visiting an M. affinis inflorescence (Photograph: Antonio R. Castilla), (B) an M. affinis infructescence, fertilized fruits turn dark purple when mature, (C) mature M. affinis fruits can contain ~30–50 seeds, the number of viable seeds per fruit is highly variable among and within mothers, (D) a map detailing the sires where pollen was collected (black) and the mother where the pollen was deposited (pink) during a single pollination event by a Melipona panamica individual in Gamboa, Panama.

Figure 1.2 Dispersal distances for all observed species of pollinators listed in increasing body size (ITD) order. From smallest to largest species (mm): Tetragonisca angustula (1.28), Halictidae sp. (1.69), Trigona muzoensis (1.72), Paratetrapedia lineata (1.86), Trigona fuscipennis (1.96), Trigona fulviventris (2.26), Pseudochloropsis schrottky (2.38), Trigona buyssoni (3.4), Melipona panamica (3.64), Centris dichootricha (5.28), Xylocopa ﬁmbriata (7.72).

Figure 1.3 (A) The negative relationship between the standard deviation in dispersal distances per pollination event and nearest neighbor distance (m), (B) the negative relationship between Chao sire diversity and mother tree diameter at breast height (DBH) (cm), (C) the negative relationship between sire evenness and the size of the mother tree diameter at breast height (cm).

TABLE S1.1 Environmental metrics of study regions, their Miconia affinis neighborhoods, and their pollinator communities across the Panama Canal watershed. Study region names: Alfagia (AG), Camino de Plantaciones (CP) and Gamboa (GB). System Features AG CP GB Geographic Centroids 9°13'32.02"N, 9° 5'27.67"N, 9° 6'58.82"N, (Lat, Long) 79°45'58.57"W 79°39'19.91"W 79°41'38.18"W Total Trees 397 400 360 Reproductive Density (trees/m2) 0.083 0.075 0.014 Forest Cover (%)(SE) 75.96 (1.93) 81.54 (1.44) 83.23 (2.75) DBH (cm)(SE) 27.38 (1.85) 46.77 (3.25) 40.95 (2.52) NND (m)(SE) 53.92 (8.78) 31.17 (17.15) 34.44 (2.44) Elevation (m.a.s.l.) 120 164 26 Pollinator Species 7 5 6 Mean ITD (SE) 2.814 (0.32) 1.952 (0.29) 2.743 (0.52)

TABLE S1.2 Genetic summary of M. affinis seeds in the three study regions (Gamboa (GB), Camino de Plantaciones (CP, and Alfagia (AG)): N is the total number of seeds that were successfully assigned fathers per study region, Na is the average number of alleles per locus, He is the expected heterozygosity, Ho is the observed heterozygosity as calculated by GenAlex. The mean allelic richness and private allelic richness were determined using rarefaction with the HPRARE program. Genetic Measure AG CP GB N 93 102 97 Na (SE) 7.125 (1.231) 5.500 (0.732) 7.250 (1.373) He (SE) 0.639 (0.075) 0.591 (0.070) 0.660 (0.048) Ho (SE) 0.491 (0.079) 0.460 (0.105) 0.532 (0.152) Mean Kinship (SE) -0.001 (0.002) 0.135 (0.012) 0.011(0.003) Mean Allelic Richness 2.526 2.376 2.642 Mean Private Allelic Richness 0.14 0.13 0.12

( A) (B)

( C ) (D)

FIGURE S1.1 (A) Raw sire count (CERVUS, P=0.013), (B) Chao extrapolated sire diversity (CERVUS, P=0.035) (C) The raw sire count (COLONY, P= 0.046), (D) The Chao extrapolated sire diversity (Colony, P= 0.059).

FIGURE S1.2 Density distribution of all pollen dispersal events (m).

FIGURE S1.3 Correlograms from spatial autocorrelation analysis using the Ritland kinship coefficient and even 100m distance classes in the three M. affinis focal study regions (Gamboa (GB), Camino de Plantaciones (CP, and Alfagia (AG)). Dashed lines represent the 95 % confidence intervals around the null hypothesis of absence of spatial genetic structure. Black lines around average Fij values represent 95% confidence intervals generated by jack-knifing loci.

Chapter 2: Landscape genetic diversity and robust pollinator networks buffer plant reproduction and pollen-mediated gene flow from extreme climate events

ABSTRACT

Tropical forests provide ecosystem services that are inarguably vital to global health, yet these forests are dually threatened by intensifying climate change and the highest rates of deforestation worldwide. More than 90% of tropical plant species are animal-pollinated and are largely dependent on mutualistic interactions for reproduction, thus the disruption of pollination services represents a major threat to tropical biodiversity. In this study, we use a uniquely integrated field, network, and molecular approach to quantify the spatial and temporal effects of climate and deforestation on density-dependent plant reproduction and animal-mediated gene flow processes for a common tropical understory tree. We reveal that landscape-scale genetic diversity and pollinator network generalization are important for buffering plant reproduction in deforested areas against the negative effects of climatic extremes.

O’Connell, M.C., Castilla, A.R., Espinosa, H., Kore, P., Magadi, A., Alonso Santos-Murgas. & Jha, S. (in prep for Science). M. O’Connell was responsible for designing research questions, performing research, analyzing data, writing the dissertation, and writing the manuscript. 29

INTRODUCTION

Tropical forests are home to unrivaled levels of biodiversity that interact to facilitate

essential ecosystem functions, such as carbon sequestration, water cycling, and pollination,

all of which are essential to our planet’s well-being (Shimamoto et al. 2018). Despite supporting more than two-thirds of the earth’s known species (Gardner et al. 2009, Gibson et al. 2011, Lewis et al. 2015) and exerting more significant influences on global climate

and agricultural production than any other terrestrial biome (Brandon 2014), tropical

regions and forests are threatened by the highest rates of landscape alteration worldwide

(Hansen et al. 2013, Hansen and Sato 2016). Simultaneously, tropical species, which

evolved within relatively narrow climatic niches (Ghalambor et al. 2006), are facing novel,

anthropogenically-induced climate regimes, that are making tropical forests

uncharacteristically drier and warmer, potentially altering critical time-sensitive species

interactions (Gilman et al. 2010, Bregman et al. 2016, Aguirre-Gutiérrez et al. 2020). While a number of recent studies have investigated the independent effects of climate or land

alteration on tropical ecosystem function (Grimm et al. 2013, Lewis et al. 2015) how these

two pressures interact and synergistically alter vital species interactions, such as plant-

pollinator mutualisms, is largely unexplored (Schweiger et al. 2010, González-Varo et al.

2013, Keith et al. 2008, Fordham et al. 2012).

Upwards of 90% of tropical plant species are critically dependent on animal-

mediated pollination services (Ollerton et al. 2011), but these species interactions are

vulnerable to the synergistic spatial (Kremen et al. 2007) and temporal (Miller-Rushing et

al. 2010) pressures of land-use and climate change, which disrupt animal pollinators’ 30

capacity to provide effective, targeted pollination services (Kearns et al. 1998) and

facilitate pollen-mediated gene flow (Breed et al. 2015). Trees located in dense conspecific

neighborhoods often experience the reproductive benefits of positive density-dependence,

i.e. increased fruit and seed set (Kunin 1993, Jones and Comita 2008, Nattero et al. 2011),

but as deforestation and forest fragmentation alter the distribution of conspecifics, they can

dismantle traditionally observed dispersal patterns, often impeding pollination services and reproductive success (Aizen and Feinsinger 1994, Aguilar and Galetto 2004).

Additionally, the temporal distribution of floral resources in tropical forests are evidenced

to be shifting in response to climate change (Forrest 2015), resulting in phenological

mismatches, reductions in pollination services, and reproductive failure in plants (Harrison

2000, Memmott et al. 2007, Kudo and Cooper 2019). Alterations to plant phenology and

reproduction may be further exacerbated by extreme climate events (Wright and Calderón

2006), such as the El Niño Southern Oscillation (ENSO), which are becoming more severe

and frequent as global temperatures increase (Timmermann et al. 1999, Cai et al. 2014,

Erfanian et al. 2017); such disruptions of pollination services by both land use and climate

change may pose one of the greatest threats to the future of tropical biodiversity (Aguilar

et al. 2006, Traveset et al. 2017) specifically with respect to population genetic diversity

and essential pollen-mediated gene-flow processes (Lowe et al. 2015).

Genetic diversity is assumed to be essential for population persistence in the face

of novel and extreme environmental variation (Pauls et al. 2013), yet the impacts of this

diversity have rarely been evaluated across spatial scales and with respect to reproductive

success across land-use and climate extremes. In addition to reproductive output, pollen- 31 mediated gene flow also tends to follow positive density-dependent patterns where denser, less fragmented forest landscapes exhibit higher levels of genetic admixture (reviewed in

Dick et al. 2008); however it has been found that that high levels of local kinship (low genetic diversity) in dense plant neighborhoods can counteract positive density-dependent processes (Castilla et al. 2017) leading to high levels of biparental inbreeding (Elam et al.

2007), reductions in the genetic quality of deposited pollen (Soutu et al. 2002), reproductive failures (i.e. higher fruit and seed abortion rates, sensu Jones and Comita

2008), and reductions in offspring fitness (Hirao 2010). Although long-distance pollen dispersal via animal pollinators can mitigate the negative impacts of forest fragmentation and high local kinship (Latouche‐Hallé et al. 2004, Jha and Dick 2010, Ismail 2012), these animal-mediated pollination services may be sensitive to spatial and temporal disturbances

(Kremen et al. 2007, Tscharntke and Brandl 2004, Kudo and Ida 2013), raising concern about whether plant-pollinator networks will be resilient enough to continue to facilitate sufficient gene flow in changing forests (Lowe et al. 2015, Kaiser-Bunbury et al. 2017).

Robust plant-pollinator networks that provide ecological insurance, i.e. ecological redundancy, and abundant alternative mutualist partners, may be more resilient and crucial to buffering plant species against the spatial and temporal consequences of global change

(Blüthgen and Klein 2011, Fontaine et al. 2005, Gilman et al. 2012). Network robustness, often the result of the abundance of generalist versus specialist pollinators in a network

(Brosi 2016), is likely to shift with species losses and changes to pollinator community composition (Burkle and Alarcón 2011, Schleuning et al. 2016, Steffan-Dewenter et al.

2006, Brosi and Briggs 2013), and currently we have only a rudimentary understanding of 32

to what degree tropical plant-pollinator networks can actually ensure sufficient pollination

services in rapidly changing forests.

In this study, we leverage a uniquely fine-scale and spatially explicit analysis of the

common tropical understory tree, Miconia affinis, to evaluate the impacts of extreme

climate events, neighborhood spatial and genetic structure, and plant-pollinator networks on critical plant reproduction and pollen-mediated gene flow processes. We surveyed more than 3370 ha of tropical forest, mapped and genotyped 1157 trees using geospatial and molecular tools to quantify neighborhood density and local kinship (sensu Lowe et al.

2015, Castilla et al. 2017), identified more than 69,000 pollen grains to quantify the specialization of pollen-based pollinator foraging networks (Bosch et al. 2009, Ballentyne et al. 2015), and used parentage analyses to determine the impact of these factors on seed set, pollen dispersal, and sire richness (Marshall et al. 1998, Bittencourt and Sebbenn 2007) across two years, one of which occurred during a historically severe El Niño Southern

Oscillation (ENSO) event.

RESULTS AND DISCUSSION

In this study we show that plants in high kinship neighborhoods (low genetic diversity)

are more vulnerable to the effects of severe climatic extremes. Specifically, in a non-

ENSO year, mother trees located in neighborhoods with higher local kinship values did

not significantly affect seed set (z = -0.132, P = 0.227)(Table 1), but during the ENSO

year, high local kinship conferred noteworthy reproductive consequences, significantly

reducing seed set (z = -0.4718, P <0.0001)(Figure 1A). Additionally, we found that while

pollen dispersal distances were significantly positively influenced by local kinship in our non-ENSO study year (z = 0.2084, P = 0.0267)(Table 1) meaning that in high kinship

neighborhoods, trees received pollen from further away, this positive relationship was significantly reduced during the ENSO year (z = -0.2259, P = 0.0322)(Figure 2A). We did not find local kinship to impact sire richness differently between our two study years

(Table 1). While both temperate and tropical studies indicate that trees in higher kinship neighborhoods often exhibit reduced viable seed production (Bosch and Waser 1999,

Castilla et al. 2017), likely due to higher biparental inbreeding or maternal selective

abortion (Collevatti et al. 2001), we show that ENSO conditions further exacerbate local

kinship’s negative effects on plant reproduction, supporting the theory that selective

forces may be stronger (Snow 1994, Hirao 2010), and that the ability for pollinators to

counteract these effects by promoting genetic admixture via long-distance pollen

dispersal may be compromised during environmental extremes (Delattre et al. 2013,

Cormont et al. 2011, Kremer et al. 2012, Ismail et al. 2012). Importantly, we provide

novel, field-based evidence supporting the literature around the Environment-Dependent

Inbreeding hypothesis, that climatic extremes can contribute to significant changes in

inbreeding depression rates in natural plant populations (Fox and Reed 2011). Although

seemingly short-term in their climatic impacts, ENSO-induced extremes could have

lasting impacts on the genetic structure of plant populations, as evidenced by studies that

have found that in some species, adaptive responses to the local environment can lead to

genetically and phenotypically structured plant populations within 2–3 generations

(Kramer et al. 2009, Cahill et al. 2010), which in our focal species is ~6 years (Castilla et 34

al. 2016), similar to the approximate time scale of the ENSO cycle. Overall, these results provide evidence that extreme climatic events can negatively impact the dispersal patterns, genetic structure, and reproductive output of tropical tree species (Cleary et al.

2006).

We also found that our studied pollinator communities differed in their

compositions across the non-ENSO and ENSO study years (Figure S1) and that the

robustness of M. affinis pollen transfer networks was significantly reduced during the

ENSO year compared to the non-ENSO year (P <0.0001). Climate-related changes to

networks have been documented in other study systems, i.e. changes in the degree of

network specialization, connectance, and robustness in response to extreme weather events

and drought (Rafferty et al. 2015, Hoiss et al. 2015, Ramos–Jiliberto et al. 2018), but this

study is the first to quantify the impacts of neighborhood-level plant-pollinator networks on seed production and pollen dispersal patterns. We found that network robustness was

significantly negatively correlated with seed set during both the non-ENSO (z = -0.1797,

P <0.0001)(Table 1), and that the negative influence of increasing network robustness

reduced seed set more severely during the ENSO year (z = -0.6755, P <0.0001)(Figure

1B). Additionally, we found that while increasing network robustness did not have a significant influence on pollen dispersal distances during a non-ENSO year (z = 0.0015, P

= 0.9847), robustness marginally correlated with longer pollen dispersal distances during the ENSO year (z = 0.2611, P 0.0966)(Table 1)(Figure 2B). We did not find the robustness of pollen transfer networks to be a significant predictor of sire richness in this system

(Table 1). 35

Our results indicate that more robust networks are correlated to reductions in seed

set, but increases in pollen dispersal distance, and that both patterns were augmented during

the ENSO study year. Plant-pollinator network robustness has been previously attributed

to a higher presence of generalist as opposed to specialist pollinator species (Fontaine et

al. 2008), which confers resilience to species loss due to environmental duress (Soares et

al. 2017) and land-use change (Memmott et al. 2004, Ferreira et al. 2020) to the plant

species they pollinate. As more robustness plant-pollinator networks may not provide as

specialized and therefore potentially less effective pollination services (Dáttilo et al. 2015), more robust plant-pollinator networks may confer reductions in reproductive output to plants (Tur et al. 2013), highlighting the important contributions of specialized pollinators to networks and plant reproduction (Brosi 2016). Additionally, pollen dispersal by novel

and/or pollinators with more flexible preferences may be able to counteract some of the negative impacts of deforestation and climate-driven phenological shifts, sometimes conferring unexpected genetic benefits (Dick 2001, Noreika et al. 2019) by increasing the frequency of long-distance dispersal events in altered landscapes (Kitamoto et al. 2006,

Millar et al. 2014). The fact that we found the negative influence of robustness on reproductive output and its positive effect on pollen dispersal both became more substantial during an extreme climatic event, illustrates that complex trade-offs between species in plant-pollinator networks may become further exacerbated as climate change continues to expose mutualists to novel interactions (Burkle and Alcarón 2011, Gilman et al. 2012). As landscape alteration and climatic extremes continue to expose plants to novel, altered pollen transfer networks (Miller-Struttman et al. 2015, Rafferty 2017), we illustrate the 36

need for the incorporation of network analyses in discussions of plant reproduction, gene

flow, and forest resilience (Gonzalez-Varo et al. 2013, Hoiss et al. 2015).

Lastly, we show that the positive density-dependent dispersal and reproductive

patterns classically found in plant species (Reviewed in Ghazoul et al. 2005) are altered by

extreme climate events. Specifically, while nearest neighbor distance (NND; our proxy for

deforestation) did not appear to significantly impact seed in the non-ENSO year (z = -

0.0085, P = 0.9446)(Table 1), NND had a significant positive impact of seed set during the

ENSO year, i.e. as nearest neighbors became more dispersed, their seed sets increased (z =

0.1291, P <0.0001)(Figure 1C). We found the positive effects of high conspecific density to be reversed during the ENSO year, i.e. trees in less dense neighborhoods produced higher seed sets on average, but we did not find nearest neighbor distance to have differential impacts on dispersal distances nor sire richness across our study years (Table

1). Dense conspecific plant neighborhoods often receive higher quality pollination services and their resulting reproductive benefits (Kunin 1997, Elliott and Irwin 2009, Nattero et al.

2011), but little field-based evidence exists to evaluate how density-dependent patterns will

change as plants and their pollinators encounter novel and extreme climatic regimes

(Molano-Flores et al. 1999, Rafferty 2017). Our results bolster previous findings that under

environmental duress, density-dependence can become unpredictable (Bachelot et al. 2020,

Duminil et al. 2016) and that seed production and pollen dispersal will not necessarily follow simple linear relationships with population size (Morgan et al. 1999, Dahlgren et al.

2016). In this study, we provide preliminary evidence that classical positive density-

dependent plant reproduction may differ dramatically as a function of local climatic

conditions and extreme climatic events.

In this study we integrate field, molecular, and network data to reveal that climatic

extremes may further exacerbate the negative impacts of deforestation on pollen dispersal

and plant reproduction. Specifically, our analyses reveal that more genetically diverse plant neighborhoods may be buffered against the negative impacts of climatic extremes on reproductive success and pollen dispersal. We also found that extreme climatic events may interfere with density-dependent reproduction patterns by altering not only the movement of pollen in altered plant neighborhoods, but also the robustness and specialization of the networks that pollinate them. Through our results, we illustrate the complex, interacting impacts of global change on sensitive but critical species interactions, and bring forth new insights gained through the pursuit of synergistic research approaches to questions of global change and the future of biodiversity.

METHODS

Study System and the 2015-2016 El Niño Southern Oscillation

We conducted our study in the tropical lowland forests of the Panama Canal watershed,

which have experienced substantial deforestation over the past two centuries and are

heavily impacted by continued anthropogenic development; as of 2011, it was estimated

that only 55% of the watershed was forested (Ibáñez et al. 2002, Simonit and Perrings

2013). For this study, we collected fine-scale data on plant population genetics, reproductive success, and plant-pollinator interactions during the Panamanian dry season

(January to June) in both 2013 (non-ENSO year) and 2016 (ENSO year). Dry season rainfall has been cited as a critical limiting factor influencing tropical plant phenology and reproduction (Wright and Calderón 2006, Wright et al. 2019), therefore to evaluate whether our study years are representative of historic ENSO and non-ENSO annual dry season averages, we examined long-term rainfall data dating back to 1925 (Estoque et al. 1985,

Physical Monitoring Program of the Smithsonian Tropical Research Institute, NOAA) and confirmed that the dry season rainfall average of 127.67 mm that we detected in the non-

ENSO year of our study (2013) was within one standard deviation of the long-term non-

ENSO average (x = 105.03, SD = 50.68) and that the dry season rainfall average of 52.23 mm that we detected in the ENSO year of our study (2016) was within one standard deviation of the long-term ENSO average (x = 49.98, SD = 20.91).

Study Species, Neighborhood Traits, and Phenology

Our study system covers a ~3370-ha area of tropical lowland forest and includes 1157 individuals of the tropical pioneer tree Miconia affinis (Melastomataceae; 3–6 m in height) distributed across three study regions separated by an average of 12.9 kms (described in

Castilla et al. 2017). M. affinis is an ideal study species due to its well-studied ecology and vast range (Luck and Daily 2003, Jha and Dick 2010), the availability of genetic tools for this species (Jha and Dick 2009), its capacity to thrive in a variety of forest contexts

(Castilla et al. 2015), and the fact that it is diploid, hermaphroditic, self-incompatible, and depends on a small community of native bee species to buzz-pollinate, or release pollen from, the plant’s poricidal anthers (Jha and Dick 2010). Adult M. affinis trees display 1–3 flowering events lasting ~2 days each during the Panamanian dry season during which it is 39

pollinated by a relatively small community of native bees, after which the species’ fruits

develop to maturity between May and September and are dispersed largely by small-bodied frugivorous birds (Luck and Daily 2003, Jha and Dick 2008, 2010). To evaluate whether the phenology of M. affinis was altered during the 2016 ENSO event, we examined long- term species-specific phenological data from the same study region dating back to 1987

(Wright and Calderón 2006). We confirmed that the average first day of flowering during

our non-ENSO study year of our study (2013: x = day 91, SD = 42 days) was within one

standard deviation of the average first day of flowering recorded during non-ENSO years

since 1987 (historic: x = day 81, SD = 46 days) and that the average first day of flowering

we observed during our ENSO study year (2016: x = day 104, SD = 29 days) was within

one standard deviation of the average first day of flowering recorded during ENSO years

since 1987 (x = 90, SD = 20.91).

Between the years of 2012 and 2016, we calculated two main neighbor indices per

tree, nearest neighbor distance (NND) and local kinship (Fij). NND was calculated as the average spatial distance to the ten nearest conspecific trees, is commonly used index of conspecific density and proxy of deforestation (Frohn and Hao 2006, Songer et al. 2009), and is highly correlated with forest cover in this study region (rp = −0.40, P = 0.007 as per

Castilla et al. 2017). The second trait, local kinship, is a measure of the nonrandom spatial

distribution of genotypes using the Loiselle kinship coefficient (Fij), which is calculated

via pairwise comparisons between the focal tree and all its neighbors within a 400 m radius;

a radius that was previously determined to be an important threshold within which

neotropical tree species display fine-scale spatial genetic structure (Castilla et al. 2017, 40

Hardy et al. 2006). To calculate local kinship, we collected leaf tissue from every adult M. affinis tree (N = 1157) within a 2 km radius of each study region’s geographic centroid, as this 2 km buffer is a well-established distance to characterize bee foraging and response to landscape-level resources (Steffan-Dewenter et al. 2002, Cusser et al. 2019, Egerer et al.

2017, Quitsberg et al. 2016) and has been shown to capture a large portion of the pollen dispersal kernel for M. affinis (Jha and Dick 2010). We obtained the genotypes from each tree using eight highly polymorphic microsatellite loci, then use this information to calculate the kinship coefficients using SPAGeDi (Loiselle et al. 1995; Hardy and

Vekemans 2002).

Pollinator Community Observations, Pollen Load Analyses, and Network Construction

In both our ENSO and non-ENSO study years we conducted pollinator observations and collected floral visitors at 32 reproductive focal M. affinis mother trees. Pollinators were observed at five focal inflorescences per tree between the hours of 6am and 2pm for one

30-minute session per tree in our non-ENSO study years and one 15-minute session per tree in our ENSO study year (due to delayed flowering in 2016, all study regions flowered simultaneously, limiting the time we could spend at each study region). We considered pollinators legitimate floral visitors if they sonicated the flowers of M. affinis and we allowed them to visit several flowers within the same inflorescence, then collected them as they departed with insect nets and stored them in 70% ethanol upon their departure from their focal inflorescence (sensu Castilla et al. 2017). Pollinator identifications were performed by author A.S.M. and specimens were deposited to the University of Panama’s entomological collection. 41

To understand how the pollinator communities that visited M. affinis during our

ENSO and non-ENSO study years differed, we calculated several indices regarding their species diversity, pollen fidelity, community composition, as well as their degrees of pollen transfer network specialization and robustness. To analyze pollen loads, we assembled a pollen reference collection from the anthers of plant species that we observed to be flowering contiguously with M. affinis. We made pollen slides for both the pollen reference collection and insect pollen loads following protocol using fuchsine dye as described in

Kearns and Inouye (1993). We analyzed the slides using microscope protocol as per Ritchie

et al. (2016), totaling in 97 pollen species for reference. Due to the high quantities of M.

affinis pollen found in each insect pollen load (some pollen loads reached upwards of

>15,000 M. affinis pollen grains), we subsampled each slide by randomly selecting 10

squares from 20x20 grid in which we characterized the species compositions for the first

300 grains, then simply counted the number of grains beyond 300 to quantify total pollen

load size (as per Ritchie et al. 2016).

We calculated pollen load size and fidelity to M. affinis as the proportion of a given

pollen load that was quantified as conspecific versus heterospecific pollen (as per Brosi

and Briggs 2013, Huang et al. 2015). As per past studies focused pollen transfer network

construction around a focal plant species (e.g. Cusser et al. 2019), we also constructed

pollen-load network matrices (as per Alarcón 2010), and quantified pollen transfer network

metrics Robustness and Blüthgen’s network specialization index (H2) per mother tree in

the ‘bipartite’ package in R (Blüthgen et al. 2006, Dormann 2008), where a robustness

value of 0 = the least robust and a value of 1 = the most robust network, and an H2 value 42

of 0 = a completely generalized and a value of 1 = a completely specialized network. We

chose to focus on network robustness because of the metrics cited in the literature, there are an increasing number of studies indicating that robustness is sensitive to climate,

phenology, and temporal variation in the neotropics (Ramos–Jiliberto et al. 2018, Rabeling

et al. 2019). Our measured network robustness values were negatively correlated with

pollen fidelity (defined as the proportion conspecific pollen per pollen load; sensu Lindsey

1984, Kearns and Innouye 1993)(P <0.0001) and conspecific pollen abundance per pollen

load (P <0.0001). Additionally, to quantify and visualize the compositional difference

between the pollinator communities that visited M. affinis in our two study years, we ran

Non-metric Multi-dimensional Scaling (NMDS) analyses (Faith et al. 1987). For our

NMDS ordination, we used the Bray–Curtis dissimilarity index (Beals 1984) and compared

this to a permutational multivariate analysis of variance (PERMANOVA) analysis in the

vegan package in R (Oksanen et al. 2013) along with a similarity percentage analysis

(SIMPER)(Clarke 1993) to find the average contribution of each collected pollinator

species to overall Bray-Curtis dissimilarity across our study years. (Figure S2).

Reproductive Success and Genetic Analyses

We collected 10 mature fruits from each of the 32 focal M. affinis mother trees and

quantified seed set (the proportion of viable seeds out of the total number of seeds) for a

total of 640 seed set measurements across both years. Genomic DNA was isolated each

seed per fruit using the DNAzol® Genomic DNA Isolation Reagent (ThermoFisher

Scientific, CA, USA), and amplified at eight highly polymorphic microsatellites using

Qiagen Multiplex PCR Master Mix Kits (as per Castilla et al. 2015, O’Connell et al. 43

2018). Using the GeneMarker software (SoftGenetics, LLC) we successfully genotyped a total of 320 seeds from 2013 and 317 seeds from 2016 with at least 75% loci coverage

(Castilla et al. 2016, Dick 2001). To determine pollen dispersal distances, we ran likelihood-based paternity assignments using CERVUS software (Marshall et al. 1998), only accepting assignments with a confidence criterion of >80%, as per other past plant dispersal studies (e.g., Hardesty et al. 2006, Bittencourt and Sebbenn 2007, Lander et al.

2010), resulting in a sample size of N=191 assignments from 2013 and N=191 assignments from 2016 from which we measured the Euclidean distances between mother trees and the assigned fathers. Given that our sampling of potential fathers is limited to a 2 km radius per study region, unassigned paternities are likely the result of pollen coming from outside of our samples areas, as assumed in past studies (Bittencourt and Sebbenn 2007, Bacles and Ennos 2008, Dick et al. 2008). Lastly, we calculated the number of sires that successfully fertilized viable seeds via animal-mediated pollen deposition at mother trees

(N=32) for the genotyped offspring at each tree (10 seeds per mother) using the COLONY software which determined the number of half- versus full-sibship seeds (Jones and Wang

2009).

Statistical Analyses and Models

We ran three separate linear mixed-effects models testing whether (1) the seed set, (2) the pollen dispersal distance, and (3) the richness of sires, were impacted by the predictor variables (a) Nearest Neighbor Distance (NND), (b) Local Kinship (Fij) and (c) and pollen transfer network robustness. These three predictor variables were run as interactions with the categorical variable ENSO year versus non-ENSO year, to capture the suite of rainfall 44 and phenological differences between the observed years. In each model, NND, local kinship, and network robustness were scaled to a mean of zero and variance of 1 to allow for comparison across all interaction terms in the model, and we ensured that all predictor variables had VIF scores of <3 in our models using the ‘car’ package (Fox and Weisberg

2019). All models were run using the in the ‘lme4’ package (Bates et al. 2015) with study region and mother as nested random effects: for seed set we fit a binomial distribution and ran a GLMM, for pollen dispersal distance data fit a normal distribution and we used a

LMM, and sire richness fit a gaussian distribution for which we also used a GLMM (as per

Castilla et al. 2017, O’Connell et al. 2018). In addition to running the global models (Table

1) we confirmed the same results using the ‘MuMin’ package (Barton 2009) to perform model selection and averaging for all models within <2 AICc of the top model (Table S2).

Table 2.1 Results of our global linear mixed effects models. The three response variables modeled include (1) Seed Set, (2) Pollen Dispersal Distances, and (3) Sire Richness. Predictor variables included Nearest Neighbor Distance (NND), Local Kinship (Fij), and pollen transfer network robustness, all of which were all ran as interactions with the ENSO versus non-ENSO year as a categorical variable. Significant effects are bolded.

Model 1: Seed Set Model 2: Pollen Dispersal Distance Model 3: Sire Richness Predictor Term Est. Std.Err z value Pr(>|z|) Est. Std.Err z value Pr(>|z|) Est. Std.Err z value Pr(>|z|) Non-ENSO -0.0085 0.1229 -0.070 0.9446 0.0407 0.0852 0.477 0.6351 -0.0100 0.0588 -0.170 0.865 NND ENSO 0.1206 0.1225 0.984 0.3252 -0.0048 0.0873 -0.055 0.9565 -0.0199 0.0640 -0.311 0.756 ENSO*NND 0.1291 0.0324 3.984 <0.0001 -0.0455 0.1016 -0.448 0.6548 -0.0099 0.0870 -0.114 0.909 Non-ENSO -0.0428 0.1119 -0.383 0.7020 0.2084 0.0923 2.258 0.0267 -0.0076 0.0664 -0.114 0.909 Kinship ENSO -0.5146 0.1112 -4.627 <0.0001 -0.0175 0.0859 -0.204 0.8391 -0.0345 0.0600 -0.574 0.566 ENSO*Kinship -0.4718 0.0327 -14.443 <0.0001 -0.2259 0.1051 -2.150 0.0322 -0.0269 0.0896 -0.301 0.764 Non-ENSO -0.1797 0.0271 -6.637 <0.0001 0.0015 0.0774 0.019 0.9847 -0.0126 0.0594 -0.212 0.832 Robustness ENSO -0.8551 0.0562 -15.225 <0.0001 0.2626 0.1346 1.951 0.0532 0.0342 0.0849 0.403 0.687 ENSO*Robustness -0.6755 0.0657 -10.286 <0.0001 0.2611 0.1561 1.672 0.0966 0.0468 0.1037 0.451 0.652

Figure 2.1 Regression plots displaying the influence of (A) local kinship, (B) network robustness, and (C) nearest neighbor distance (m) on reproductive output (seed set: proportion viable to non-viable seeds per fruit) across our non-ENSO (green) and ENSO (orange) study years. Regression lines are visual representations of interaction terms of the raw data, independent of other interactions in the model. *Asterisks in the corner of each plot indicate a significant difference between study years.

Figure 2.2 Regression plots displaying the influence of (A) local kinship, (B) network robustness on pollen dispersal distance (km) across our non-ENSO (green) and ENSO (orange) study years. Regression lines are visual representations of interaction terms of the raw data, independent of other interactions in the model. *Asterisks in the corner of each plot indicate a significant difference between study years.

Figure S2.1 (A) Non-metric multidimensional scaling (NMDS) ordination Figure S2.1 (B) Similarity percentage analysis (SIMPER, Clarke plot created using the Bray-Curtis dissimilarity index illustrating the 1993) displaying pairwise comparisons of collected pollinator species dissimilarity between the pollinator communities that visited Miconia affinis to find the average contribution of each species to overall Bray-Curtis during the non-ENSO (2013) and ENSO years (2016). PERMANOVA dissimilarity across our study years. results: R2 = 0.4644, Pr(>F) = 0.0563. non-ENSO ENSO p-value Augochloropsis sp. 0.0000 0.4621 0.4 Centris dichootricha 0.2310 0.0000 0.1 Frieseomelitta paupera 0.0000 0.9027 0.4 Halictidae sp1 0.6486 0.0000 0.1 Halictidae sp2 0.2310 0.2310 1.0 Melipona comprissepes 0.0000 0.2310 1.0 Melipona fuliganosa 0.0000 0.3662 1.0 Melipona panamica 1.9938 0.9027 0.4 Melissodes sp. 0.0000 0.2310 1.0 Paratetrapedia lineata 0.5973 0.2310 0.6 Paratetrapedia schottky 0.2310 0.0000 0.1 Paratetrapedia sp. 0.0000 1.2614 0.1 Partamona sp. 0.0000 0.8283 0.4 Tetragonisca angustula 2.3890 0.0000 0.1 TRIGONA AMALTHEA 0.2310 0.0000 0.1 Trigona fuliganosa 0.5365 0.7993 1.0 Trigona fulviventris 1.9844 0.4621 0.1 Trigona fuscipennis 2.2445 0.2310 0.1 Trigona muzoensis 2.6754 2.0781 1.0 Trigonisca buyssoni 0.9027 0.0000 0.1 Xylocopa fimbriata 0.2310 0.2310 1.0 Xylocopa sp. 0.0000 0.3662 1.0

Table S2.1 Top Models for each response variable as determined by MuMin Model selection. Specifically, we provided the degrees of freedom (Df), log likelihood (logLik), AIC values (AICc), delta AIC (Delta), and Akaike weights (Weight) for each model. Models within 2 AIC of the best model are listed along with the results of the full model average. Independent variables included in models are as follows: (1) Nearest Neighbor Distance (NND), (2) Local Kinship (Fij), (3) Pollen transfer network robustness, (4) the interaction of NND and Year, (5) the interaction of Kinship and Year, and (6) the interaction of Robustness and Year. Significant effects are bolded.

Response Model Selection: Full Model Averages within 2 AICc of Top Model Variable Independent Df logLik AICc Delta Weight Variable 1234567 10 -4334.36 8689.07 0.00 1 Independent Estimate Std. Error z value Pr(>|z|) Variable Seed Set NND -0.0085 0.1227 0.069 0.9452 Kinship -0.0429 0.1119 0.382 0.7023 Robustness -0.1796 0.0271 6.610 < 0.0001 ENSO*NND 0.1287 0.0332 3.867 0.0001 ENSO*Kinship -0.4718 0.0327 14.404 < 0.0001 ENSO*Robustness -0.6759 0.0660 10.216 < 0.0001 Independent Df logLik AICc Delta Weight Variable 145 7 -541.63 1097.57 0.00 0.24 13457 9 -539.81 1098.11 0.54 0.18 1345 8 -541.30 1098.98 1.41 0.12 1245 8 -541.50 1099.39 1.82 0.10 Pollen Independent Estimate Std. Error z value Pr(>|z|) Dispersal Variable Distance NND 0.0060 0.0307 0.195 0.8451 Kinship 0.2098 0.0886 2.361 0.0182 Robustness 0.0112 0.0554 0.202 0.8400 ENSO*NND NA NA NA NA ENSO*Kinship -0.2285 0.1026 2.220 0.0264 ENSO*Robustness 0.0816 0.1519 0.537 0.5915

Table S2.1, continued Independent Df logLik AICc Delta Weight Variable (Null) 3 -134.95 276.30 0.00 0.29 1 4 -134.74 278.15 1.85 0.12 Independent Estimate Std. Error z value Pr(>|z|) Sire Variable Richness NND -0.0100 0.0588 -0.170 0.865 Kinship -0.0076 0.0664 -0.114 0.909 Robustness -0.0126 0.0594 -0.212 0.832 ENSO*NND -0.0099 0.0870 -0.114 0.909 ENSO*Kinship -0.0269 0.0895 -0.301 0.764 ENSO*Robustness 0.0468 0.1037 0.451 0.652

Figure S2.2 All-Effects regression plots displaying the influence of (A) pollen transfer network robustness, (B) local kinship, and (C) nearest neighbor distance (NND)(m) on seed set (proportion viable to non-viable seeds per fruit) across our non-ENSO (2013) and ENSO (2016) study years. Regression lines are visual representations of interaction terms taking other interactions in the model into consideration. *Asterisks in the corner of each plot indicate a significant difference between study years.

Chapter 3: Reap what you sow: local plant composition mediates bumblebee foraging patterns within urban garden landscapes

ABSTRACT

Although urban gardens are celebrated for supporting bee abundance and diversity within

cities, little is known about how garden management and urbanization levels influence bee

foraging behavior and ability to utilize resources within these landscapes. Specifically, the

preferences and diet breadth of bees may depend largely on local and landscape conditions

in human-altered, urban environments. To understand how foraging patterns and pollen

preferences are influenced by urban landscape composition, we first examined if bees visit

plants grown within urban gardens and then assessed the relationships between local floral

resources, urban land cover, and pollen collection patterns within 20 community gardens

across 125 km of the California central coast. We targeted a well-studied, essential native

pollinator in this ecoregion, Bombus vosnesenskii, and analyzed pollen on the bodies of

individuals collected in our study gardens to compare their contents to local and landscape

garden composition factors. We found that both greater landscape-level urban cover and greater plant species richness in the gardens drove higher within-garden pollen collection.

We also found that B. vosnesenskii preferred ornamental plant species over highly available crop species in the gardens. Our study indicates that landscapes that support plant diversity,

O’Connell, M., Jordan, Z., McGilvray, E., Cohen, H., Liere, H., Lin, B. B., ... & Jha, S. (2020). Reap what you sow: local plant composition mediates bumblebee foraging patterns within urban garden landscapes. Urban Ecosystems, 1-14. M. O’Connell was responsible for designing research questions, analyzing data, writing the dissertation, and writing the manuscript.

including both ornamental plants and sustenance-oriented food crops, promote greater within-garden pollen collection patterns, with likely benefits for urban garden food production.

INTRODUCTION

Urban gardens are increasingly recognized as beneficial, multifunctional green

infrastructure that contribute to a city’s resilience for human inhabitants and biodiversity

alike (Goddard et al. 2010; Barthel and Isendahl 2013; Shwartz et al. 2014). In highly

urbanized matrices, gardens that are largely cultivated to provide human sustenance may

serve a dual purpose, both as a source of food security for gardeners and as a critical

resource-dense patch that helps foster the movement and survival of wildlife across

otherwise inhospitable, altered landscapes (Rudd et al. 2002; Sushinsky et al. 2013). These

green spaces may be especially important for critical ecosystem service providers such as

pollinators, natural enemies of pests, and seed dispersers, but the capacity for urban gardens to support such taxa largely depends on the animals’ access to urban gardens and more importantly, their ability to collect and utilize resources once they are in an urban garden

(Paker et al. 2014; Philpott and Bichier 2017; Johnson et al. 2018). Despite the high

abundance and richness of wildlife recorded in urban gardens (McFrederick and LeBuhn

2006; Baldock et al. 2015; Lin et al. 2015), it is often unclear if garden resources are being

utilized by these animals, and which local- and landscape-scale management practices

might influence resource collection patterns in these spaces (Lowenstein et al. 2019).

Elucidating these patterns could further our understanding of how animals utilize resources

54 in heterogeneous landscapes and would inform urban decision-makers on how to manage green space to support urban wildlife and associated ecosystem functions.

At the local scale, the composition and species richness of plants grown in urban spaces can have large impacts on the presence and abundance of wildlife (Smith et al. 2006;

Kowarik 2011); this is particularly relevant for highly mobile animals found within urban gardens, such as bees (Hulsmann et al. 2015), butterflies (Blair and Launer 1997), and birds

(Paker et al. 2014). There is compelling evidence that plant species composition and vegetative complexity (Lin et al. 2015), as well as access to bare soil and woody plants, correlates with the diversity and richness of beneficial insects that visit urban gardens

(Egerer et al. 2017 ) such as pollinators (Pawelek et al. 2009; Quistberg et al. 2016) and natural enemies of pests (Philpott and Bichier 2017). However, vegetation management within urban gardens is often driven by the needs of the gardeners rather than the needs of wildlife, therefore it is possible that resources provided within urban gardens are not intended or sufficient to support urban-dwelling animals (Loram et al. 2011; Bigirimana et al. 2012; Clarke and Jenerette 2015). For example, urban gardens are often dominated by species that either provide human sustenance (crop species) or aesthetic appeal (exotic ornamental species)(Lowenstein and Minor 2016), even though these same species may not provide optimal resources to urban wildlife with respect to nutrition (Vaudo et al.

2015), habitat requirements (Shapiro et al. 2002), or phenology (Harrison and Winfree

2015). Furthermore, the composition of vegetation within urban gardens is highly variable from garden to garden and may be mediated by extraneous factors, such as the socioeconomic standing of the gardeners (Bigirimana et al. 2012; Clarke and Jenerette 55

2015), which may further drive plant species compositions in urban gardens to be less

similar to naturally occurring plant communities to which native wildlife are accustomed

(Thompson et al. 2003; Turo and Gardiner 2014).

In addition to local-scale resource availability, the presence, abundance, and

foraging patterns of urban-dwelling animals are also a function of larger-scale landscape

composition and the degree to which this landscape supports animal foraging and habitat

needs (McKinney 2002). Within the urban landscape, native vegetation and green spaces

are often replaced by impervious substrates such as pavement and buildings, through which

mobile wildlife must navigate to find resources and suitable habitat (Blair and Launer 1997;

Jha and Kremen 2013). While highly mobile taxa such as birds (McKinney 2002) and

flying insects (McFrederick and LeBuhn 2006) can successfully colonize and survive in

highly urbanized landscapes, the persistence of these animals is often driven by the distance

between resource-rich patches and the landscape composition between these patches.

Indeed, the presence of bird and mobile invertebrate species in urban gardens is sometimes

more influenced by the surrounding habitat than by garden features (Chamberlain et al.

2004; Smith et al. 2006). Past studies have suggested that flying insects may indeed be

sensitive to landscape composition surrounding urban gardens; for instance, natural habitat

cover within 3 km of gardens supports higher insect abundance in gardens, while urban

habitat cover in this area supports lower abundance (Winfree et al. 2008; Ahrné et al. 2009).

This is likely because these animals often utilize ‘partial landscapes’, i.e. forage or nesting space within a habitat type (sensu Westrich 1996), and thus may travel between several

gardens or resource patches to meet their multiple biotic needs (Walther‐Hellwig and 56

Frankl 2000; Rudd et al. 2002). Insect use of multiple habitats for nesting and food

resources may explain why the impacts of urban land cover on this group seems to be

context- and species-specific; for example, past research within urban gardens in coastal

California found that overall bee abundance (both ground and cavity-nesting) was

negatively impacted by urban land cover, presumably due to lower nest-site availability

(Plascencia and Philpott 2017), while research within urban gardens in New York City

found that cavity-nesting bees were positively affected by urbanization, assumedly due to

their opportunistic use of artificial cavities in built structures (Matteson et al. 2008). While

such contrasting responses may be due to differences in animal resource availability and

utilization in urbanized landscapes, such as altered floral or nesting resources (as

hypothesized in Cane 2005), few studies have explicitly investigated this utilization within

urban landscapes.

Beyond the local and landscape dynamics driving the usage of urban gardens by

animals, foraging patterns are often dictated by the degree to which each animal displays

their dietary preference, where preference is defined as the use of a particular resource in

relation to its availability in a landscape (sensu Beyer et al. 2010). For example, many pollinator species prefer foraging at native and ‘pollinator-friendly’ plant species despite their often relatively low abundance in human-altered landscapes (Frankie et al. 2005;

Saifuddin and Jha 2014). Simultaneous reductions in native plant species richness and

habitat function are well documented in urbanized areas (Kowarik 2011). As a result, urban

wildlife may have very few options when their preferred forage species are no longer

available; they can either shift their preferences, find new habitats, or face local extirpation. 57

Thus, in human-dominated landscapes, novel plant communities may drive changes in urban-dwelling wildlife foraging behaviors. For example, pollinator species may shift their foraging behaviors to visit non-preferred, ornamental, and exotic species when native species are locally unavailable; however, these behavioral shifts depend on both plant composition within the landscape and pollinator foraging plasticity (Cane 2005; Williams et al. 2011). More specifically, when found in urban landscapes, some butterflies, bees, hover flies, and hummingbirds display preferences for highly rewarding ornamentals planted by gardeners, even when these are non-native novel resources (Garbuzov and

Ratnieks 2014; Harrison and Winfree 2015). This may explain why pollinator community composition shifts from a broad mix of specialized and generalist consumers to broadly generalist species in urban areas relative to nearby semi-natural areas (Cane 2005; Baldock et al. 2015), indicating that urban plant communities may not support the wide breadth of dietary needs displayed by healthy native wildlife networks (Inouye 1978; Kleijn and

Raemakers 2008). Despite the relevance of diet breadth to biodiversity conservation in urban matrices, little is known about the preferences that mobile ecosystem service providers, like pollinators, exhibit in human-altered, resource dense patches such as urban gardens.

Thus, it is of critical importance that we begin to understand more about the foraging patterns and food preferences of pollinators within rapidly urbanizing landscapes.

Animal pollinators facilitate the reproduction of ~78-94% of wild plant species and over

75% of global crop species (Klein et al. 2006), worth more than $235-$577 million USD globally (IPBES 2016). In the case of urban gardens, a very high percentage of crop plants 58

depend on animal-mediated pollination services (Matteson and Langellotto 2009);

therefore, understanding these dynamics in the urban agricultural context will be crucial in

determining the future of urban agroecosystems and urban food security. In this study, we

determine if an important native pollinator, Bombus vosnesenskii, collects floral resources

in urban gardens and how local and landscape-scale factors mediate its foraging patterns.

We also quantify and compare this pollinator’s forage preference across plant species

cultivated our focal urban gardens. Based on previous findings, we propose three

fundamental hypotheses H1) bees collect pollen resources from within urban gardens, H2)

Local floral richness and urban land cover positively impact within-garden pollen

collection patterns, and H3) Pollinators exhibit a consistent preference for non-crop plant

species, even when their availability is low in urban gardens.

METHODS

Study region and garden metrics—We conducted our study in 20 urban gardens across

three counties in the California central coast (Monterey, Santa Clara, and Santa Cruz) from

June to August of 2016 (Fig. 1). Each site consisted of an urban community garden (0.10 to 3.84 ac) separated from other gardens by >2 km. All gardens are organically managed and have produced food for between 2-50 years. The region hosts ~1 million people and although the study region is in a single geographic area, the region is heterogeneous and gardens vary in temperature, precipitation, management, landscape conditions, and gardener demographics (Egerer et al. 2019).

At the local scale, we measured habitat characteristics (e.g. vegetation and ground

cover) within a 20 x 20 m plot placed at the center of each garden. We measured canopy

cover with a convex spherical densiometer at the center of the plot, and 10 m to the North,

South, East, and West. We counted and identified all trees and shrubs in the plot and noted

the number of individuals in flower. In each plot, we randomly selected eight 1x1 m

quadrats within which we identified all herbaceous plants (except grasses) to

morphospecies, measured height of the tallest non-woody vegetation, noted which species

were in flower and counted their flowers, and assessed percent ground cover of bare soil,

grass, herbaceous plants, leaf litter, rocks, mulch, and straw. Due to high levels of

collinearity between many of these metrics, we only included the following in our

regression model development: the number of trees and shrubs in the 20 x 20 m plot at

each garden, the percent mulch cover in the 1 x 1 m quadrats in each gardens, the average

number of flowers in the 1 x 1 m quadrats in each garden, the size of the garden, and the

estimated herbaceous plant species richness in the 1 x 1 m quadrat in each garden (model

composition similar to past studies in this region, Plascencia and Philpott 2017).

Herbaceous and woody plant species were identified using the USDA PLANTS database

and were binned into the following categories according to human usage: crop species,

ornamental species, and weed species (as designated by the USDA PLANTS database).

Because these vegetation surveys were conducted on one day in each garden within three

time periods (June 6th-9th, July 1st-5th, Aug 1st-3rd) surrounding the pollinator survey

(June 27th - July 11th), we included all three vegetation surveys in our analyses to comprehensively describe the plant community. Further, of the 93 plant species in our 60

vegetation surveys, 11 were not observed flowering during the survey dates but are known

to flower in the summer; for these species we confirmed with resources at the Missouri

Botanical Gardens that flowering typically occurs between May and July (Kemper Center

for Home Gardening 2020).

At the landscape scale, we calculated the proportional cover of four main land-use

categories within 2 km buffers surrounding each garden with data from the 2011 National

Land Cover Database (NLCD, 30 m resolution)(Homer et al. 2015). Specifically, we

selected 2 km buffer zones as this is the largest landscape scale that has been shown to be

predictive of resource usage for bees (Steffan-Dewenter et al. 2002) and is utilized by many other urban bee studies (Cusser et al. 2019, Egerer et al. 2017, Quitsberg et al. 2016). We created four land-use categories and calculated the proportion of area represented for each in the 2 km buffer: 1) semi-natural (deciduous, evergreen, and mixed forests, dwarf scrub, shrub/scrub, and grassland/herbaceous), 2) open (lawn grass, parks, and golf courses), 3) urban (low, medium, and high intensity developed land), and 4) agricultural habitat

(pasture/hay and cultivated crop). Other land cover types covered <5% of the total area and were not included. Due to collinearity between these four main land-use categories, for our initial model development we only included semi-natural habitat and urban habitat cover.

Additionally, we estimated total garden size by ground-truthing GPS points around each garden. We assessed both garden size and land cover with spatial statistics tools in ArcGIS v.10.1.

Pollinator survey—We collected Bombus vosnesenskii between the 27th of June and 11th of July 2016. B. vosnesenskii is an annual, primitively eusocial, ground-nesting, central 61 place foraging bee native to the Pacific Coast states of the United States and is a proficient crop and greenhouse pollinator (Dogterom et al. 1998; McFrederick and LeBuhn 2006).

The species is also often found in urban greenspaces, where it displays relatively generalist foraging behaviors (McFrederick and LeBuhn 2006). We captured 10 B. vosnesenskii individuals in each garden with a net and then transferred them to a ‘kill jar’ with ethyl acetate. To examine pollen that is representative of floral visitation (as in Alarcón 2010), we first removed corbicula pollen from the specimen (less than 15% of individuals were carrying corbicula pollen loads) and then placed individuals into 5 mL test tubes and submerged them with 95% ethanol solution. We rinsed the forceps with the 95% ethanol solution between the processing of each individual to avoid pollen contamination. A few of the samples experienced ethanol leakage and degradation, leaving a total of 189 bees for subsequent analysis (mean = 9.45, SE = 0.08, per garden).

Bumble bee pollen loads—Bees were vortexed in their original 5 mL test tube for 30 seconds, then after removing the bees, the pollen was centrifuged for 2 minutes at 1800 rpm. We then pipetted 15 µL of ethanol plus the suspended pollen pellet into a new tube, added 40 µL of Fuchsine dye, vortexed the mixture for 10 seconds (Kearns and Inouye

1993). We added 50 µL of the vortexed solution to the slide and allowed this to sit for 10 seconds before the cover slip was placed on top and allowed to set for at least 24 hours.

We examined each microscope slide at 20x magnification by beginning at the top left corner and moving down to the bottom edge, then moving over just enough to ensure any grains from the first column would not be in view. Then, we moved back up the slide and repeated this pattern to cover the entire area without coming across the same grain twice. 62

For up to the first 300 grains encountered, we examined and identified the species of each

pollen grain using the reference collection as a comparison. We chose to maintain this 300- grain cutoff to ensure our sampling efforts were consistent across pollen loads (as per

Williams and Kremen 2007, Ritchie et al. 2016). If the original slide had fewer than 300 grains, we counted and identified grains on an additional slide made with the same dye protocol as the original slide until we reached 300 grains. If fewer than 250 grains were found on the first two slides, no more slides were analyzed as we observed diminishing returns in the number of grains on each new slide made from the original sample solution.

We were able to collect pollen and makes pollen slides for all 189 of our study bees. When we found a pollen grain that could not be identified as any of the species in the reference collection, we counted that grain as a morphological species, assigned it a number, and followed the same protocol described above to add it to the reference collection. We were able to successfully identify the majority of the pollen grains collected in the pollen loads to species: less than 0.1% of the collected grains remained unidentified. For all grains identified to species, grains were binned into three categories based on the USDA PLANTS database: crop species (species that were planted intentionally for the production of fruits and vegetables for human sustenance), ornamental species (species that were planted intentionally, but do not produce fruits that are consumed by humans), and weedy species

(species that occurred in the gardens but were not planted intentionally)(Green 2009).

Reference Collection—We created a pollen reference collection by collecting pollen samples from the 160 most abundant floral resources (based on inflorescence counts within the 20x20 m plots) in the urban gardens. Specifically, anthers from these plants were 63

collected from the gardens and placed in ethanol, and later centrifuged to separate pollen

for 1 minute at 1800 rpm. We combined 15 µL of this pollen solution with 40 µL of

Fuchsine dye, vortexed this for 10 seconds, and pipetted 50 µL onto a slide. After 10

seconds, we placed a slide cover on top and allowed 24 hours to set. We then viewed these

slides using a Leica light microscope to find and photograph pollen grains for the reference

collection using LAS v4 software. We took one picture of each pollen species at 20x

magnification, and two pictures at 63x magnification at different focal planes. We made

some exceptions for very large grains, instead taking pictures at 40x magnification for

clarity in viewing significant features. The 20x photographs included a scale bar in

micrometers and were kept to scale for a comparative size reference. We set the saturation

of the photographs to 101 and the Gamma setting to 0.55. We adjusted the brightness as

needed to see the details of the pollen grains but tried to keep the brightness near 80%. This

reference collection then served as a comparison to identify pollen grains found on bumble

bees.

Impacts of local and landscape features on pollinator diversity—We used binomial

generalized linear mixed effects models to examine relationships between local and

landscape predictor variables on three B. vosnesenskii pollen load response variables: 1) the proportions of pollen from within and outside of the gardens, 2) the proportions of crop, ornamental, and weed pollen, and 3) the Shannon-Weiner Diversity of each pollen load.

To ensure an appropriate characterization of proportions and preference, we only retained pollen loads for which we were able to identify a total of 300 pollen grains to species for

our statistical analyses (as per Harmon‐Threatt et al. 2017, Saifuddin and Jha 2014). We 64

calculated Shannon-Weiner Diversity using the ‘vegan’ package in R (Oksanen et al. 2018) to calculate diversity within each pollen load. The predictor variables we originally included in our model development were the following: at the landscape scale, the percent of urban land cover and semi-natural habitat within 2 km of the garden, and at the local scale, the number of trees and shrubs in the 20 x 20 m plot at each garden, the percent mulch cover in the 1 x 1 m quadrats mulch in each gardens, the average number of flowers in the 1 x 1 m quadrats in each garden, the size of the garden, and the estimated herbaceous plant species richness in the 1 x 1 m quadrat in each garden. Given that bumble bees make multiple foraging bouts in a single day and heavily groom themselves between bouts

(Holmquist et al. 2012) and that individual bumble bees in urban gardens tend to forage within the same gardens where they were originally collected (Matteson and Langellotto

2009), we assume that a plant species present in a garden and found on a pollinator was

likely collected from that garden (as per Matteson and Langellotto 2009). After developing

this set of initial predictor variables, we ran tests to identify collinearity of predictors in all

our models by calculating a variance inflation factor (VIF) for each model set using the car

package in R (Fox and Weisberg 2018). We used a VIF cutoff score of 2 and removed the

variables with the highest value in a stepwise fashion until all variables received a score

below the cutoff. Using this process, our three full models contained all the predictor

variables except the number of flowers in the garden (which positively correlates with

herbaceous plant species richness, retained in the model) and the amount of semi-natural

habitat within 2 km of the garden (which negatively correlates with percent urban land

cover within 2 km, retained in the model). We considered each of the 189 pollen loads 65 collected from an individual bee as a replicate response variable in each garden and we included the garden as random effect in our models. We ran GLMs in R using the ‘glmer’ function for response variables 1 and 2 and the ‘lmer’ function for response variable 3 in the ‘lme4’ package (Bates et al. 2011) and model selection using the MuMin package

(Barton 2018) and selected the top model based on the AICc values and model averaging for models that were within 2 AICc of the top model.

Within-Garden Preference Analysis—We calculated plant species preferences for 189 individuals of B. vosnesenskii to determine which plant species the bee is more likely to visit, relative to its availability in the gardens. Specifically, we assessed the species-level composition of pollen loads according to a framework developed for habitat or resource use analysis (Johnson 1980) that is also frequently used for pollinators (e.g., Davis et al.

2012; Jha et al. 2013). These analyses rank resource collection (i.e. proportion of each pollen species) on the pollinator relative to its availability (i.e. percent cover of each plant species) in the garden; pollen species that are collected significantly more than expected respective to the plant species’ floral availability are ranked the highest. We ran these preference analyses using the adehabitatHS package in R (Aebischer et al. 1993; Calenge

2006) by comparing a matrix of floral species availability per garden with a matrix of pollen species use per bee using Wilks’ lambda. From this we constructed a preference matrix, which was evaluated using a randomization test (500 repetitions) to determine significant preference for one plant species over each other species (Aebischer et al. 1993).

For herbaceous plants, we estimated the floral availability in each garden by averaging the percent cover of each plant species that occurred within our eight 1 x 1 m quadrats and 66 averaged these across the three sampling periods. For woody plants, we estimated floral availability in each garden by using shrub count data from our 20 x 20 m survey plot. For these species, we calculated each species’ average circular area available (as per average plant spread measures provided by the Kemper Center for Home Gardening, Missouri

Botanical Garden 2020), summed these values per plot, then divided these totals by the area of the 20 x 20 m plot and averaged this across the three sampling dates. Total percent cover of each plant type (crop, ornamental, weed) was calculated by summing the plant species percent cover within each category. Preference was only analyzed for plant species documented both in the garden and in the pollen load of each B. vosnesenskii, therefore, of the 93 plant species identified in the pollen loads, 25 plant species were excluded from the preference analysis portion of the study because they were not present in the gardens where the bees were collected. Of the remaining 68 species included in the preference analyses, three were removed due to their infrequency (<2 individual occurrences) within the garden or pollen load data (as per Ritchie et al. 2016); this minimum occurrence threshold was set due to the constraints in calculating variance for compositional analysis (Calenge 2006).

We also conducted this preference analysis excluding the most common pollen species, strawberry, to evaluate whether remaining preference rankings remained the same and found similar overall results (Table S1).

RESULTS

Across our 20 study gardens, we observed a variety of landscape contexts, garden compositions, and B. vosnesenskii pollen collection patterns. Our study gardens were

67 located across a gradient of urbanization ranging from 7.7% to 97.3% urban cover within

2 km and ranged in size from 0.04 to 1.55 ha. Our garden plots had anywhere from 7 to 24 herbaceous plant species and on average hosted 60.7% crop cover (μ = 0.6039, SE =

0.1248), 21.7% ornamental cover (μ = 0.2175, SE = 0.1160), and 17.6% weed cover (μ =

0.1761, SE = 0.0088). B. vosnesenskii individuals across these gardens collected pollen from 3 to 35 plant species. On average, 39.2% (μ = 0.1761, SE = 0.0088) of the pollen loads were composed of pollen from plant species identified as being in the study gardens and 60.8% (μ = 0.1761, SE = 0.0088) of pollen from plants that were designated as being outside of the gardens, with the average pollen load being composed of 49.9% crop species

(μ = 0.4721, SE = 0.1313), 45.7% ornamental species (μ = 0.4727, SE = 0.1393), and 4.4% weed species (μ = 0.4334, SE = 0.0463)(Fig. 2).

We found a positive relationship between the percent of urban land cover within 2 km and the proportion of pollen collected from within the gardens (X = 0.3259, P =

0.0195)(Fig. 3A, Table 1). Additionally, we found a positive relationship between the number of herbaceous plant species in the garden and the proportion of pollen collected from within the garden (X = 0.4722, P = 0.0004)(Fig. 3B, Table 1). An increase in herbaceous plant species richness negatively correlated with the collection of pollen from crop plants (X = -0.4800, P = 0.0002), but positively influenced the collection of pollen from ornamental plants (X = 0.5700, P = 0.0004)(Fig. 4A, Table 1). An increase in the percent urban land cover within 2 km positively correlated with crop plant pollen collection

(X = 0.3002, P = 0.0282), but had no effect on the collection of pollen from ornamental plants (Fig. 4B, Table 1). To this end, we verified that no correlations (Pearson correlation 68

coefficients) existed between our measure of urbanization and the percent cover of crop (r

= -0.1874) versus ornamental (r = 0.1066) versus weed (-0.3382) species in our study

gardens. Lastly, an increase in garden size negatively impacted the diversity of pollen

collected as measured via the Shannon-Wiener species diversity index (X = -1.0471, P =

0.0196)(Table 1).

From our preference analysis, we found that B. vosnesenskii highly preferred

strawberry (Fragaria x ananassa) and California poppy (Eschscholzia californica) above

the other plant species in our study gardens (λ = 0.0174, P = 0.0020)(Table 2). Beyond this

common crop species, over half of the species preferred were ornamental and weed species

that were present in low availabilities relative to many of the crop species in the gardens.

Regarding ornamentals, B. vosnesenskii preferred California poppy (Eschscholzia

californica), snapdragon (Antirrhinum majus), rose bush (Rosa L.), and California

buckwheat (Eriogonum fasciculatum)(Table 2, Table S2). B. vosnesenskii also displayed a preference for two weed species: wild mustard (Brassica L.) and wild radish (Raphanus

raphanistrum) which are widely distributed introduced species in the region. Lastly, B.

vosnesenskii also displayed a weak preference for crop plants such as American red

raspberries (Rubus idaeus), arugula (Eruca vesicaria), and eggplant (Solanum melongena)

despite their lower percent coverage compared to crops such as tomatoes and peppers.

Overall, the dominance of ornamentals and weeds in the top ten preferred plants relative to

their lower overall availability in the gardens, shows that B. vosnesenskii exhibits strong

foraging preference for non-crop plant species (Fig. 2).

DISCUSSION

We conducted one of the first analyses of urban bee pollen collection and preference to

reveal strong predictors of urban garden floral usage at both the landscape and local scale,

as well as strong foraging preferences for ornamental plants available within these gardens.

Specifically, our results show that the composition of the pollen loads of B. vosnesenskii

in urban gardens were largely influenced by two factors: the percent urban cover within 2

km of the gardens (landscape) and the number of herbaceous plant species planted within

the gardens (local). Additionally, we found that this key native pollinator exhibits a strong preference for ornamental and weedy plants, indicating that urban gardeners should consider cultivating a suite of ornamental plants and allowing some flowering weeds to persist, in addition to their food crops, to attract pollinator visitation and services in urban

gardens.

The proportion of pollen that B. vosnesenskii collected from within the study gardens was strongly positively correlated with the percent of urban land cover within 2 km of the garden. In other words, we found that in landscapes with more urban land cover, a greater the proportion of the bee’s pollen load came from plant species cultivated within the focal urban garden. This trend seems intuitive, considering that these gardens may be acting as highly rewarding resource islands in otherwise forage-poor landscapes

(McFrederick and LeBuhn 2006). Because the degree of urbanization often correlates with the homogenization of floral communities (McKinney 2002; Shwartz et al. 2014), urban gardens in highly urbanized landscapes may contain a higher plant richness relative to their surrounding landscapes (Hulsmann et al. 2015). Increased pollinator visitation to high- 70 reward floral resource patches within resource-poor landscapes has been documented in classic patch dynamics (Goulson 2000; Blaauw and Isaacs 2014) and spatial memory literature (Cartar 2004; Burns 2005; Ohashi et al. 2007); however, its relevance for pollinators within urban landscapes had not been previously determined. Bees are central place foragers (Michener 2000) and although the literature claims that larger bee species can forage at greater distances from their nests (Greenleaf 2007) even large-bodied species, including several Bombus species, predominantly exhibit shorter-distance foraging, with

78% of foraging bouts made within 500 m of nests (Walther‐Hellwig and Frankl 2000). In a study of B. impatiens in urban gardens in New York City, Matteson and Langellotto

(2009) found that 45% of marked individuals were later collected in the gardens where they had initially been documented, indicating that bumble bees in highly urbanized areas may largely forage within a single garden. This is perhaps partially due to the capacity for bumble bees to display ‘area-restricted foraging’, repeatedly traveling along the same foraging routes in a landscape until they optimize and integrate resource location in their spatial memories (Ohashi et al. 2007). Our findings similarly suggest that bumble bees foraging within urban landscapes respond to landscape-scale resource distribution when selecting high-resource patches, as seen in rural systems (Jha and Kremen 2013; Pope and

Jha 2018) and/or additionally experience strong nest site fidelity, as suggested in past landscape genetic studies (e.g., Schenau and Jha 2017).

Interestingly, greater urban cover around our study gardens also correlated with greater proportions of pollen collected from crop plants in the garden; this is particularly interesting given that crop cover was not higher in gardens within more urbanized 71 landscapes. This trend could be a result of lower or more homogenized resources in more urbanized landscapes combined with the greater relative floral availability of crops in our urban gardens. This pattern also may be due to greater levels of generalism or novel-plant visitation by bees in more urbanized landscapes. For certain species of bumble bees, pollen load composition can shift with land-use change over time (Cane 2005). Sustained exposure to novel, but more readily available food sources (such as crops and invasive plants) can lead to these shifts (Kleijn and Raemakers 2008; Williams et al. 2011). In suburban and rural Northern California, B. vosnesenskii, the same species studied here, did not show a preference for native over non-native pollen, suggesting that the species might be more resilient to shifts in vegetation composition (Jha et al. 2013; Saifuddin and Jha

2014). In this study, we could not compare native and non-native pollen collection since more than 80% of the species in our study were non-native, but we were able to document substantial variability in pollen collection in response to urban landscape cover. Overall, this finding provides additional evidence that urban gardens are likely acting as important foraging centers and may be utilized to different degrees depending on landscape composition.

We also found that the number of herbaceous plant species in the gardens impacted several aspects of B. vosnesenskii pollen foraging. As we hypothesized, herbaceous plant species richness (a measurement including crops, ornamentals, and weedy species) in the urban gardens was positively correlated with the proportion of pollen collected from within the garden, rather than from outside the garden. Past work has shown that bee abundance and diversity is higher in urban gardens with higher floral diversity (Ballare et al. 2019) 72 especially when particularly attractive plant species are incorporated (Pawelek et al. 2009).

In fact, one study of common bumble bees in Lüneburg, Germany found that bee persistence in urban landscapes was driven less by the degree of urbanization and more by plant species richness, diversity, and composition across urban greenspaces and gardens

(Hulsmann et al. 2015). Other past studies also support our findings, showing that floral visitation rates (Ebeling et al. 2008) and pollinator persistence (Kleijn and Raemakers

2008) in a patch are correlated with patch plant diversity. This pattern may be particularly clear in urban ecosystems relative to nature preserves and farmland, given that pollinators may exhibit more generalist behaviors, foraging from a greater number of plant species in urban areas (Baldock et al. 2015). Interestingly, in our study system, the number of herbaceous plant species did not drive the diversity of pollen found in the bees’ pollen loads; instead, pollen load diversity was negatively related to garden size. Given that garden size did not correlate with herbaceous plant species in our study, and that previous work in this system has found that garden size does not predict bee composition (Plascencia and Philpott 2017), we hypothesize that larger garden size may simply represent larger patch size and but not greater floral resource availability. In particular, the pattern we found could be reflective of optimal foraging theory predictions where pollinators forage at fewer flowers in larger patches because they are attempting to maximize their rate of resource acquisition (Goulson 2000).

Interestingly, we found that the proportion of pollen collected from ornamental plants was positively correlated with herbaceous plant species richness, while the proportion of pollen collected from crop plants was negatively correlated with herbaceous 73

plant species richness. Specifically, we found that B. vosnesenskii can forage primarily on crop species in some gardens, but if given more plant species options, they will shift to

collecting pollen from more ornamental plant species. Lowenstein et al. (2019) similarly

found that a number of ornamental plants were visited more often than other plant types,

and Hennig and Ghazoul (2011) found that higher floral abundance and diversity can

concentrate pollinators towards more attractive plant species. Overall, these patterns

suggest that small differences in the diversity of plant species within a site can critically

impact pollinator foraging. However, we note that these past studies and our own do not

measure the impacts of such a shift on plant reproduction and therefore it remains a debate

as to whether reductions in a single pollinator’s plant fidelity may be outweighed by

frequent pollinator visitation in terms of influencing plant reproduction (Brosi and Briggs

2013; Geslin et al. 2013). Plant reproductive success relies on the robustness of pollinator

networks (Brosi 2016), but in urban contexts these dynamics likely change as pollinator

networks may be in constant flux in response to land-use change and resource

inconsistency (Geslin et al. 2013). Indeed, Carvalheiro et al. (2012) and Blaauw and Isaacs

(2014) found that small patches of non-crop flowers within large farms increase pollinator-

dependent crop output, suggesting that increases in visitation due to greater flowering

diversity may outweigh the negative reproductive impacts of heterospecific pollen

deposition, depending on how these species are arranged spatially (Werrell et al. 2009).

Other work conducted within our study system also indicates that midday bee visitation

levels increase with greater floral density and diversity; however, floral density and

diversity (not bee visitation) were the best predictors of high crop plant reproductive 74

success, perhaps because floral composition and availability can capture longer-term

pollinator visitation patterns (Cohen et al, in review). Overall, these studies suggest that

small decreases in crop-pollen collection within the more plant species-rich urban gardens

may be outweighed by greater overall pollinator visitation at these sites, but further study

is required to connect urban bee foraging patterns with urban plant reproductive success.

Finally, we ran a preference analysis which revealed that B. vosnesenskii displayed significant foraging preferences overall, where 6 of the top 10 most preferred species were non-crop species even though the majority of plant species in the gardens were crops.

Specifically, 4 of the top 10 preferred plants were ornamentals, likely planted for aesthetics or biodiversity provision since they do not satisfy human sustenance needs. This preference pattern, which is analyzed on a per-bee basis, is consistent with general pollen load composition patterns across bees (45.7% ornamental species, 49.9% crop species, and

4.4% weed species) considering that gardens are heavily skewed toward crop plant cover

(21.7% ornamental cover, 60.7% crop cover, 17.6% weed cover). Although most of the ornamental species in our study gardens were not native to the region, and non-native flowers are believed to be less attractive forage for wild bees in the study region (Morandin and Kremen 2013), B. vosnesenskii does not display a strong preference for only wild

native species (Jha et al. 2013; Saifuddin and Jha 2014; Harmon‐Threatt et al. 2017).

Indeed, one crop species B. vosnesenskii preferred relative to its availability in our study

gardens was strawberry (Fragaria x ananassa). It is possible that this preference came out

strongest, not necessarily because of the plant species’ presence in our observed pollen

loads relative to its availability in our gardens, but because of the presence of large-scale 75

strawberry cultivation in the region, which could increase animal-mediated strawberry pollen carryover into the gardens. Beyond this species, B. vosnesenskii displayed a strong

preference for Eschscholzia californica, or California poppy, and to a lesser extent

perennial shrub Eriogonum fasciculatum, California buckwheat, which are both native to the region and have been cited as preferred foraging resource for bumble bees previously

(Morandin and Kremen 2013; Harmon‐Threatt et al. 2017). Both species are among several

California natives that are being targeted by many local bee conservation planting programs, as their reward offerings appear to be more attractive to native bumble bees than

honeybees (Kremen et al. 2002). Other ornamentals we found that B. vosnesenskii preferred include snapdragons and roses, exotics introduced to the area that are well-known for being

‘bee-friendly’ and particularly suited to bumble bee pollination (Odell et al. 1999; Frankie

et al. 2005). Additionally, B. vosnesenskii preferred two Brassicaceae species, wild

mustard and wild radish, which are both prevalent and naturalized agricultural weeds that

are nectar and pollen sources for a variety of pollinators (Sahli and Conner 2007). Overall,

our preference results align with other studies that have found that urban bee communities

may increase their diet breadth when placed under biotic pressure to forage on exotic and

introduced species within urban landscapes (Cane 2005). Outside of urban systems, bumble

bee species that can survive land use changes are often those able to switch their

preferences toward floral resources in the altered landscape, where their new diets reflect

changes in plant community composition, such as the increasing presence of non-native

plants in altered landscapes (Kleijn and Raemakers 2008; Burkle et al. 2013; MacIvor et

al. 2014). Based on these studies, B. vosnesenskii appears to be a species capable of such

foraging shifts in the face of increasing land-use change.

CONCLUSIONS

We show that gardens across even the most urbanized landscapes function as crucial

resource centers for B. vosnesenskii and that resource collection depends critically on both

local garden vegetation management and landscape composition. Therefore, our work

resonates with other studies that have indicated that effectively supporting healthy

pollinator communities across urban landscapes requires thoughtfully designed and

abundant greenspaces, especially in the mostly densely urban portions of cities (Braker et

al. 2014). Because urban gardens may be one of the few habitats where urban-dwelling

bees can forage, adding elements to these gardens that better support their biology will be

critical; this includes supplementing crop plantings with a higher proportion and

particularly a higher diversity of ornamental plant species and allowing for some flowering

weeds to persist. Gardeners may also see secondary benefits from such planting schemes,

as many studies have found that supplemental non-crop, native wildflower plantings can

attract a higher diversity of insect visitors at higher visitation rates, which can help increase

plant seed set and crop reproductive output (Carvalheiro et al. 2012; Blaauw and Isaacs

2014). Additionally, because a high number of popular fruit and vegetable crops have been

found to be obligately outcrossing, and thus more dependent on animal-mediated pollination to set fruit at all (Klein et al. 2006; Matteson and Langellotto 2009), gardeners may see increases in the number of successfully fruiting crop species if they can

77 successfully attract a variety and abundance of pollinators to their plots. For example, by cultivating a wide variety of ornamentals and native flowers to supplement their crop gardens, it is likely that gardeners will increase the temporal window for which floral forage is available in their gardens, as many of these plant species display varying, but predictable phenologies to which native bees are often accustomed (Hernandez et al. 2009).

Additionally, weeds are known to be important sustenance for pollinators in altered and agricultural landscapes (e.g., Requier et al. 2015); given that some weeds may pose a management challenge for gardeners, gardeners could evaluate this tradeoff and choose to forego weed removal specifically when there is low general floral resource availability.

Establishing an urban garden as a consistent and early resource epicenter may help pollinators develop spatial memories of these resource locations (Cartar 2004; Burns 2005) and can possibly ensure that urban gardeners will receive their pollination services throughout the growing season (Pawelek et al. 2009).

By expanding the way urban gardens are perceived, from isolated patches primarily intended to meet human sustenance needs, to permeable refugia that dually support of both food production and urban biodiversity, these gardens can address the food resource needs of both humans and foraging animals. Regarding B. vosnesenskii, this species is a highly effective pollinator that is a critical actor in the maintenance of genetic connectivity for native and ornamental plant populations, many of which are also obligately outcrossing and need pollinator-assisted pollen distribution to set seed and produce healthy offspring

(Cane 2005). Additionally, because urbanized landscapes may act as a filter for bee communities (Banaszak-Cibicka and Żmihorski 2012), the diets of bee species that persist 78 in these environments, like B. vosnesenskii, need to be considered for the maintenance of healthy bee populations in these resource-limited environments (Ahrné et al. 2009; Vaudo et al. 2015). Further, gardeners and conservationists who cultivate plants according to B. vosnesenskii’s preferences are likely to capture the preferences of many other wild pollinator species found in cities, as they also often display preferences for highly attractive ornamentals (Lowenstein et al. 2019). In growing cities, there is a movement to increase the cultivation of native plant species for pollinator conservation (Southon et al. 2017); we posit that efforts to commit space and resources for non-crop species may be further supported if such implementations are promoted by city or state incentives or if the proposed enhancements are oriented along the garden edge or pathway, limiting competition with human food production goals. Overall, we echo Turo and Gardiner (2019) in proposing that urban gardens be considered critical conservation spaces and the people who plant them as key stakeholders in the discussion around urban planning and pollinator conservation, now and into the future.

Table 3.1 Top models after MuMin model selection for each of our five response variables related to pollen load composition. Models 1-4 were run as general linear mixed effects models with a binomial distribution and model 5 was run as a linear mixed effects model with a gaussian distribution. Garden identity was included in all models as a random effect. *Significant p-values are bolded.

Predictors Est. Std.Err Z value Pr(>|z|)

Response Variables Model 1: Garden Size + Urban 2km + Herbaceous Plant Richness

GardenSize -0.2569 0.1437 -1.787 0.073880 Model 1: Proportion Pollen Urban2km 0.3259 0.1395 2.336 0.019503 from Within Garden HerbPlantRich 0.4722 0.1325 3.565 0.000364

Model 2: Number Trees/Shrubs + Urban 2km + Herbaceous Plant Richness

Model 2: NumTreesShrubs 0.23535 0.12077 1.949 0.051324 Proportion Crop Pollen Urban2km 0.30020 0.13677 2.195 0.028165

HerbPlantRich -0.48001 0.12841 -3.738 0.000185

Model 3: Garden Size + Herbaceous Plant Richness Model 3: GardenSize 0.3187 0.1769 1.802 0.071601 Proportion Ornamental Pollen HerbPlantSp 0.5700 0.1590 3.584 0.000338

Model 4: Model 4: Herbaceous Plant Richness Proportion Weed Pollen HerbPlantSp -0.4504 0.2967 -1.518 0.129

Model 5: GardenSize + Herbaceous Plant Richness Model 5: Shannon’s Pollen GardenSize -1.0471 0.4011 -2.610 0.0196 Species Diversity HerbPlantSp -0.6725 0.3828 -1.757 0.0956

Table 3.2 Assessed pollen preference for Bombus vosnesenskii. The overall preference selection test (p-value and Lambda, left column) and the top ten preferred floral species are listed, with rank denoting the level of preference for that species with “A” being the most preferred species and each letter indicating a significantly different level of preference. P = 0.002000, λ = 0.0173803. Crop/Ornamental/ Preferred Plant Species Plant Common Name Rank Weed Frangaria x ananassa Strawberry Crop A Eschscholzia californica California poppy Ornamental A Brassica sp. Mustard Weed B Raphanus raphanistrum Wild radish Weed C Antirrhinum majus Snapdragon Ornamental C Rosa L. Rose bush Ornamental C Rubus idaeus American red raspberry Crop C Eruca vesicaria Arugula Crop C Solanum melongena Eggplant Crop C Eriogonum fasciculatum California buckwheat Ornamental D

Figure 3.1 The proportion of pollen grains in the pollen loads collected from B. vosnesenskii that were composed of pollen identified as being from plants within the garden versus plants that were outside of the garden plots. Pink represents pollen collected from species located within the study gardens and blue represents pollen collected from species that were not identified as growing in the study gardens.

Figure 3.2 The availability versus usage of crop, ornamental, and weed plants in our study gardens by B. vosnesenskii. These groupings were determined using the USDA PLANTS database and floral resource availability was determined by summing the total percent cover of each classification as a proportion of the garden. Usage was calculated as the proportion of the pollen load that composed of each type. Pink represents plant availability and blue represents plant usage.

(A)

(B)

Figure 3.3 (A) The influence of % urban land cover within 2 km of the study garden plots and (B) plant species richness in the garden plots on the proportion of pollen collected from within (pink) and outside (blue) of the garden plots by B. vosnesenskii.

(A)

(B)

Figure 3.4 The effects of landscape (A: % urban land cover within 2 km of garden plots) and local (B: plant species richness within garden plots) features on the proportion of crop (pink), ornamental (green), and weed (blue) pollen in B. vosnesenskii pollen loads.

Table S3.1 Top Models for each response variable as determined by MuMin Model selection. Specifically, we provided the degrees of freedom (Df), log likelihood (logLik), AIC values (AICc), delta AIC (Delta), and Akaike weights (Weight) for each model. Models within 2 AIC of the best model are listed along with the results of the full model average. Independent variables included in models are as follows: (1) garden size, (2) number of herbaceous plant species in the garden, (3) urban density within 2 km of the garden, (4) number of trees and shrubs in the garden, and (5) mulch cover in the garden. Significant effects are bolded. Response Model Selection Variable Independent Df logLik AICc Delta Weight Variable 123 5 -2115.18 4240.96 0.00 0.25 23 4 -2116.66 4241.72 0.76 0.17 Proportion 1243 6 -2114.77 4242.39 1.42 0.12 1253 6 -2114.93 4242.71 1.74 0.11 Pollen Load Independent Est. Std. Error Z value Pr(>|z|) from Within Variable Garden GardenSize -0.1689 0.1738 0.966 0.3339 HerbPlantSp 0.4936 0.1415 3.448 0.0006 Urban2km 0.2669 0.1778 1.491 0.1358 NumTreeShrub -0.0345 0.0892 0.384 0.7010 Mulch1m -0.0167 0.0845 0.195 0.8451 Independent Df logLik AICc Delta Weight Variable 243 5 -2314.07 4638.74 0.00 0.21 1243 6 -2313.25 4639.35 0.61 0.15 23 4 -2315.80 4640.00 1.26 0.11 2 3 -2317.10 4640.43 1.69 0.09 Proportion 2543 6 -2313.83 4640.51 1.77 0.09 Crop Pollen 253 5 -2314.97 4640.55 1.80 0.09 Independent in Load Est. Std. Error Z value Pr(>|z|) Variable GardenSize -0.0611 0.1193 0.509 0.6107 HerbPlantSp -0.4895 0.1386 3.492 0.0005 Urban2km 0.2030 0.1733 1.165 0.2440 NumTreeShrub 0.1259 0.1436 0.873 0.3829 Mulch1m 0.0279 0.0932 0.298 0.7658 Independent Df logLik AICc Delta Weight Proportion Variable Ornamental 12 4 -2359.27 4726.94 0.00 0.25 Pollen in 2 3 -2360.77 4727.78 0.83 0.17 Load 124 5 -2359.16 4728.92 1.97 0.09

Table S3.1, Continued Independent Est. Std. Error Z value Pr(>|z|) Variable Proportion GardenSize 0.1966 0.2052 0.952 0.3409 Ornamental HerbPlantSp 0.5709 0.1662 3.396 0.0007 Pollen in Urban2km -0.0601 0.1299 0.460 0.6458 Load NumTreeShrub -0.0137 0.0729 0.186 0.8521 Mulch1m -0.0252 0.1052 0.238 0.8121 Independent Df logLik AICc Delta Weight Variable 12 5 -257.57 525.75 0.00 0.18 13 5 -257.86 526.33 0.58 0.14 125 6 -256.82 526.49 0.74 0.13 Shannon 1 4 -259.06 526.52 0.77 0.13 123 6 -256.85 526.54 0.80 0.12 Pollen 1235 7 -256.25 527.64 1.90 0.07 Species Independent Est. Std. Error Z value Pr(>|z|) Diversity Variable GardenSize -1.0244 0.4064 2.491 0.0128 HerbPlantSp -0.3810 0.4419 0.858 0.3911 Urban2km -0.2447 0.3851 0.632 0.5275 NumTreeShrub 0.0208 0.1408 0.146 0.8836 Mulch1m 0.1341 0.3156 0.422 0.6731

Table S3.2 Results of second pollen preference analyses for Bombus vosnesenskii. The overall preference selection test (p-value and Lambda) and the top ten preferred floral species are listed, with rank denoting the level of preference for that species with “A” being the most preferred species and each letter indicating a significantly different level of preference. Analysis 1 include strawberry pollen, this second analysis does not. The species and their rankings remained the same in this second analysis. P = 0.00200000, λ = 0.01577525.

Preferred Plant Species Plant Common Name Crop/Ornamental/Weed Rank

Brassica sp. Mustard Weed A

Eschscholzia californica California poppy Ornamental A Raphanus raphanistrum Wild radish Weed B

Antirrhinum majus Snapdragon Ornamental B

Rosa L. Rose bush Ornamental B Rubus idaeus American red raspberry Crop B

Eruca vesicaria Arugula Crop B

Solanum melongena Eggplant Crop C Eriogonum fasciculatum California buckwheat Ornamental D

Salvia spathacea Hummingbird sage Ornamental D

Figure S3.1 The top ten plant species collected by B. vosnesenskii (not relative to availability) ranked from greatest to least: tomato (Solanum lycopersicum, SOLY2), rose bush (Rosa L., ROSA5), eggplant (Solanum melongena, SOME), strawberry (Fragaria x ananassa, FRAN), California poppy (Eschscholzia californica, ESCA2), arugula (Eruca vesicaria, ERSA7), sunflower (Helianthus annuus, HEAN3), red raspberry (Rubus idaeus, RUID), wild radish (Raphanus L., RAPHA), lavender (Lavandula L., LAVAN)

Chapter 4: Science communication, media, and public engagement

ABSTRACT

The roles of science communication and public engagement in the fields of ecology and

conservation are critically important for fostering public understanding around issues of

global change and facilitating positive conservation outcomes. As species declines, habitat

loss, and climate change worsen, it is critical that researchers in our fields make more

concerted efforts to engage the public in strategic and accessible conversations around our

relationship with the earth’s biodiversity. As an ecological researcher, I have prioritized

developing my skills as a professional science communicator with the intention of

incorporating cutting edge communications practices into my research and work. To that end

I have participated in a variety of professional public engagement projects throughout my

dissertation to hone my capacity for developing compelling translational communications, to

practice interacting with a broad spectrum of audiences, and to develop expertise in directing

the creation of captivating science media products. Through opportunities to collaborate with

researchers, media producers, science educators, and community organizations I have been

able to lead and participate in the production of several conservation films, videos, and events,

as well as an ongoing community-led urban restoration project. In the following list, I provide

descriptions and links to completed and ongoing professional science communication projects in which I have participated and the role I played in each.

SCIENCE COMMUNCIATION FILMS

BioVida Panamá: La Rana Dorada (2018) Product: short documentary (14 minutes) Role: Director, story development, photography, distribution Contributors: Ash Dionne (production and editing), Dr. Angie Estrada (story development), Leonardo Simmons (photography) Distribution: Screened at 2018 SXSW Films for the Forest, 2018 Association of Tropical Biology Conference Description: Dr. Angie Estrada is passionate about engaging her fellow Panamanians in the conservation of their national amphibian: La Rana Dorada. With direction from Dr. Estrada, we share the stories of Panamanians from a variety of backgrounds who contribute to the incredible story of how this precious species was rescued from local extinction. Link: https://www.meganoconnell.net/#/fikatha/

Scientist in Residence Video Campaign (2017-2018) Product: promotional videos (~3 minutes each) Role: Director, photography, editing, production Contributors: Dr. Jay Banner (content direction) Distribution: Created to promote participation of graduate researchers at UT Austin in the Scientist in Residence program facilitated by the university’s Environmental Science Institute Description: The Scientist in Residence program places graduate researchers in public school classrooms to practice translating their research for public audiences and to help teachers incorporating cutting edge science into their curricula. These videos were used to illustrate the success of the program and to recruit new participants. Link: https://www.meganoconnell.net/#/scires/

Proposed Border Wall will Harm Texas Biodiversity (2018) Product: research video (~3 minutes) Role: Director, photography, distribution Contributors: Olivia Haun (story development, editing), Alejandro Santillana (interviews, photography) Distribution: Produced for UT College of Natural Sciences to accompany the letter “Border Wall: Bad for Biodiversity” by Norma Fowler, Tim Keitt, Olivia Haun, Martin Terry, and Keeper Trout published in Frontiers in Ecology and the Environment. Description: This video accompanied the release of an important article regarding the future of the biodiversity that exists in nature preserves along the Texas-Mexico border and how it would be impacted by the construction of a border wall. It was produced to provide illustrative imagery of some of the sites and wildlife that would be impacted and to provide direct quotes from the authors. Link: https://www.meganoconnell.net/#/biologists-v-border-wall/

Playing with Fire Video Series (2018-2019) Product: research videos (5-10 minutes each) Role: Director, story development, photography, editing Contributors: Dr. Alex Wild (producer, content, and photography direction), Jen Schlauch (photography), Charlie Brahman (story development) Distribution: Produced for NSF-funded research project with the UT Entomology Collection and the Entomology Department at the University of Georgia, shared through the UT Biodiversity Collections blog. Description: These videos were intended to illustrate the current research taking place regarding Red Imported Fire Ants Solenopsis Invicta by the Invasive Species Lab at the UT Brackenridge Field Laboratory and at the University of Georgia. In the first video we follow master’s researcher Charlie Brahman as he pursues a project on a new frontier of the ant invasion: Georgia’s coastal barrier islands. In the second video we learn how researchers collect fire ant colonies in the field, all the steps it takes to go from field collection to genetic data, and how this genetic data informs researchers about the state of the ant invasion. Link: https://www.meganoconnell.net/#/playing-with-fire/

Moving Lines Short (2019) Product: short film (5 minutes) Role: Co-producer, story development, photography, editing Contributors: Ben Mirin (co-producer, story development, photography, editing), Luke Petersen (co-producer, story development, photography, editing), Brett Addis (co-producer, story development, photography, editing) Distribution: Produced for the International Wildlife Film Festival’s Filmmaker’s Labs on behalf of Defenders of Wildlife and the USDA Wildlife Services Department, screened at the 2019 International Wildlife Film Festival. Description: Throughout the US, humans and wildlife share the landscape. Where this is most evident is in places where large predator populations have been restored, but conservation work does not end once scientists have been able to bring species back from the brink of extinction. We created a short film illustrating the collaborative work between two conservation-minded organizations to promote non-lethal management practices to reduce livestock predation by bears and wolves along their territory boundaries in Montana. Link: https://www.meganoconnell.net/#/moving-lines/

Bayou City Short Documentary (2020) Product: short film (15 minutes) Role: Co-producer, story development, photography, editing Contributors: Olivia Haun (Director, story development, photography, editing) Distribution: Produced for Texas Parks and Wildlife as a part of the 2020 Wild Texas Film Tour with support of TPWD and Fin & Fur Films. Description: Houston is the “Bayou City”, but one would not know it by looking at the current state of the city’s waterways. Houston has historically altered its bayou systems to prevent flooding, but in this film we explore whether Houstonian’s have succeeded at overcoming its flooding issues and whether there might be a better way. We share the perspectives of various members of Houston’s nature preservation and conservation community and what Houston’s bayous could look like and do if they were restored to their natural ecosystem function. Link: https://www.meganoconnell.net/#/bayou-city/

Biodiversity at Work: Can bees help us solve crimes? (2020) Product: explainer/teaser (1.5 minutes) Role: Director, story development, photography, editing Distribution: Produced as a teaser for a longer film regarding research done by the Jha Lab at UT Austin, posted on the UT Biodiversity Center blog. Description: Ecological knowledge can be useful in surprising ways and creative cross- disciplinary collaborations can foster widespread support and appreciation for the detailed study of the earth’s flora and fauna. The Jha Lab at UT Austin has entered an unconventional collaboration with the U.S. Federal Bureau of Investigation to determine how ecological field data regarding regional plant and pollinator communities might help them better geolocate items involved in criminal investigations. Link: https://www.meganoconnell.net/#/biodiversity-work/

References

Aebischer, N. J., Robertson, P. A., & Kenward, R. E. (1993). Compositional analysis of habitat use from animal radio‐tracking data. Ecology, 74(5), 1313-1325.

Aguilar, R., L. Ashworth, L. Galetto, and M. Aizen. 2006. Plant reproductive susceptibility to habitat fragmentation: Review and synthesis through a meta-analysis. Ecology Letters, 9: 968–980.

Ahrné, K., Bengtsson, J., & Elmqvist, T. (2009). Bumble bees (Bombus spp) along a gradient of increasing urbanization. PloS one, 4(5), e5574.

Aizen, M. A., and P. Feinsinger. 1994. Forest fragmentation, pollination, and plant reproduction in a Chaco dry forest, Argentina. Ecology, 75: 330-351.

Alarcón, R. (2010). Congruence between visitation and pollen‐transport networks in a California plant–pollinator community. Oikos, 119(1), 35-44.

Bacles, C. F. E., and R. A. Ennos. 2008. Paternity analysis of pollen- mediated gene ﬂow for Fraxinus excelsior L. in a chronically fragmented landscape. Heredity, 101: 368– 380.

Bachelot, B., Alonso‐Rodríguez, A. M., Aldrich‐Wolfe, L., Cavaleri, M. A., Reed, S. C., & Wood, T. E. (2020). Altered climate leads to positive density‐dependent feedbacks in a tropical wet forest. Global Change Biology.

Baldock, K. C., Goddard, M. A., Hicks, D. M., Kunin, W. E., Mitschunas, N., Osgathorpe, L. M., ... & Vaughan, I. P. (2015). Where is the UK's pollinator biodiversity? The importance of urban areas for flower-visiting insects. Proceedings of the Royal Society B: Biological Sciences, 282(1803), 20142849.

Ballare, K. M., Neff, J. L., Ruppel, R., & Jha, S. (2019). Multi‐scalar drivers of biodiversity: local management mediates wild bee community response to regional urbanization. Ecological applications, 29(3), e01869.

Banaszak-Cibicka, W., & Żmihorski, M. (2012). Wild bees along an urban gradient: winners and losers. Journal of Insect Conservation, 16(3), 331-343.

Barthel, S., & Isendahl, C. (2013). Urban gardens, agriculture, and water management: Sources of resilience for long-term food security in cities. Ecological Economics, 86, 224-234.

Barton, K. (2009) Mu-MIn: Multi-model inference. R Package Version 0.12.2/r18. http://R- Forge.R-project.org/projects/mumin/

Bates D, Mächler M, Bolker B, Walker S (2015). “Fitting Linear Mixed-Effects Models Using lme4.” Journal of Statistical Software, 67(1), 1–48.

Barton, K. (2018). MuMin: Multi-Model inference. R package version 1.15. 6. 2016.

Bates, D., Maechler, M., Bolker, B., Walker, S., Christensen, R. H. B., Singmann, H., ... & Grothendieck, G. (2011). Package ‘lme4’. Linear mixed-effects models using S4 classes. R package version, 1-1.

Beals, E. W. (1984). Bray-Curtis ordination: An effective strategy for analysis of multivariate ecological data. Advance Ecology Research, Elsevier, pp. 1–55.

Beyer, H. L., Haydon, D. T., Morales, J. M., Frair, J. L., Hebblewhite, M., Mitchell, M., & Matthiopoulos, J. (2010). The interpretation of habitat preference metrics under use– availability designs. Philosophical Transactions of the Royal Society B: Biological Sciences, 365(1550), 2245-2254.

Bigirimana, J., Bogaert, J., De Cannière, C., Bigendako, M. J., & Parmentier, I. (2012). Domestic garden plant diversity in Bujumbura, Burundi: role of the socio-economical status of the neighborhood and alien species invasion risk. Landscape and Urban Planning, 107(2), 118-126.

Bittencourt, J., and A. Sebbenn. (2007). Patterns of pollen and seed dispersal in a small, fragmented population of the wind-pollinated tree Araucaria angustifolia in southern Brazil. Heredity, 99: 580–591.

Blaauw, B. R., & Isaacs, R. (2014). Flower plantings increase wild bee abundance and the pollination services provided to a pollination‐dependent crop. Journal of Applied Ecology, 51(4), 890-898.

Blair RB, Launer AE. 1997. Butterfly diversity and human land use: Species assemblages along an urban gradient. Biological Conservation, 80: 113–125.

Blüthgen, N., Menzel, F., & Blüthgen, N. (2006). Measuring specialization in species interaction networks. BMC ecology, 6(1), 9.

Blüthgen, N., and Klein, A. M. (2011). Functional complementarity and specialisation: the role of biodiversity in plant–pollinator interactions. Basic and Applied Ecology, 12(4), 282-291.

Bolnick, D. I., P. Amarasekare, M. S. Arau Jo, R. Burger, J. M. Levine, M. Novak, and D. A. Vasseur. 2011. Why intraspeciﬁc trait variation matters in community ecology. Trends in Ecology Evolution, 26: 183–192.

Bosch, M., and Waser, N. M. (1999). Effects of local density on pollination and reproduction in Delphinium nuttallianum and Aconitum columbianum (Ranunculaceae). American Journal of Botany, 86(6), 871-879.

Bosch, J., Martín González, A. M., Rodrigo, A., & Navarro, D. (2009). Plant–pollinator networks: adding the pollinator’s perspective. Ecology letters, 12(5), 409-419.

Braker, S., Ghazoul, J., Obrist, M. K., & Moretti, M. (2014). Habitat connectivity shapes urban arthropod communities: the key role of green roofs. Ecology, 95(4), 1010-1021.

Brandon, K. (2014). Ecosystem services from tropical forests: review of current science. Center for Global Development Working Paper, (380). Breed, M. F., M. J. Christmas, and A. J. Lowe. (2014). Higher levels of multiple paternities increase seedling survival in the long-lived tree Eucalyptus gracilis. PloS One, 9: e90478.

Breed, M. F., M. H. K. Marklund, K. M. Ottewell, M. G. Gardner, J. B. C. Harris, and A. J. Lowe. 2012. Pollen diversity matters: Revealing the neglected effect of pollen diversity on ﬁtness in fragmented landscapes. Mol. Ecol. 21: 5955–5968.

Breed, M. F., Ottewell, K. M., Gardner, M. G., Marklund, M. H., Dormontt, E. E., & Lowe, A. J. (2015). Mating patterns and pollinator mobility are critical traits in forest fragmentation genetics. Heredity, 115(2), 108-114.

Brosi, B. J., & Briggs, H. M. (2013). Single pollinator species losses reduce floral fidelity and plant reproductive function. Proceedings of the National Academy of Sciences, 110(32), 13044-13048.

Brosi, B. J. (2016). Pollinator specialization: from the individual to the community. New Phytologist, 210(4), 1190-1194.

Brys, R., and H. Jacquemyn. 2010. Floral display size and spatial distribution of potential mates affect pollen deposition and female reproductive success in distylous Pulmonaria ofﬁcinalis (Boraginaceae). Plant Biology, 12: 597–603.

Burkle, L. A., and Alarcón, R. (2011). The future of plant–pollinator diversity: understanding interaction networks across time, space, and global change. American journal of botany, 98(3), 528-538.

100

Burkle, L. A., Marlin, J. C., & Knight, T. M. (2013). Plant-pollinator interactions over 120 years: loss of species, co-occurrence, and function. Science, 339(6127), 1611-1615.

Burns, J. G., and J. D. Thomson. 2005. A test of spatial memory and movement patterns of bumblebees at multiple spatial and temporal scales. Behavioral Ecology, 17: 48–55.

Cahill, A. E., Aiello-Lammens, M. E., Fisher-Reid, M. C., Hua, X., Karanewsky, C. J., Yeong Ryu, H., ... & Wiens, J. J. (2013). How does climate change cause extinction?. Proceedings of the Royal Society of London. Series B: Biological Sciences, 280(1750), 20121890.

Cai, W., Borlace, S., Lengaigne, M., Van Rensch, P., Collins, M., Vecchi, G., ... & Jin, F. F. (2014). Increasing frequency of extreme El Niño events due to greenhouse warming. Nature climate change, 4(2), 111-116.

Calenge, C. (2006). The package “adehabitat” for the R software: a tool for the analysis of space and habitat use by animals. Ecological modelling, 197(3-4), 516-519.

Cane, J. H. 1987. Estimation of bee size using intertegular span (Apoidea). Journal of the Kansas Entomological Society, 60: 145–147.

Cane, J. H. (2005). Bees, pollination, and the challenges of sprawl. Nature in fragments: the legacy of sprawl. Columbia University Press, New York, 109-124.

Cartar, R.V. (2004) Resource tracking by bumble bees: Responses to plant-level differences in quality. Ecology, 85:2764–2771.

Carvalheiro, L. G., Seymour, C. L., Nicolson, S. W., & Veldtman, R. (2012). Creating patches of native flowers facilitates crop pollination in large agricultural fields: mango as a case study. Journal of Applied Ecology, 49(6), 1373-1383.

Castilla, A., N. Pope, and S. Jha. (2015). Positive density-dependent reproduction regulated by local kinship and size in an understory tropical tree. Ann. Bot. 117: 319–329.

Castilla, A. R., Pope, N., & Jha, S. (2016). Positive density-dependent reproduction regulated by local kinship and size in an understorey tropical tree. Annals of botany, 117(2), 319- 329.

Castilla, A. R., Pope, N. S., O’Connell, M., Rodriguez, M. F., Treviño, L., Santos, A., & Jha, S. (2017). Adding landscape genetics and individual traits to the ecosystem function paradigm reveals the importance of species functional breadth. Proceedings of the National Academy of Sciences, 114(48), 12761-12766.

101

Chamberlain, D. E., Cannon, A. R., & Toms, M. P. (2004). Associations of garden birds with gradients in garden habitat and local habitat. Ecography, 27(5), 589-600.

Chao, A. (1984). Nonparametric estimation of the number of classes in a population. Scandinavian Journal of Statistics, 11: 265–270.

Clarke, L. W., & Jenerette, G. D. (2015). Biodiversity and direct ecosystem service regulation in the community gardens of Los Angeles, CA. Landscape Ecology, 30(4), 637-653.

Clarke, K. R. 1993. Non‐parametric multivariate analyses of changes in community structure. Australian Journal of Ecology, 18:117–143.

Cleary, D. F., Fauvelot, C., Genner, M. J., Menken, S. B., & Mooers, A. Ø. (2006). Parallel responses of species and genetic diversity to El Niño Southern Oscillation‐induced environmental destruction. Ecology letters, 9(3), 304-310.

Climate Prediction Center. (2020). Cold & Warm Episodes by Season. NOAA/National Weather Service. https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ ensostuff/ ONI_v5.php

Comita, L. S., S. A. Queenborough, S. J. Murphy, J. L. Eck, K. Xu, M. Krishnadas, N. Beckman, Y. Zhu, and L. Gomez-Aparicio. (2014).Testing predictions of the Janzen–Connell hypothesis: a meta-analysis of experimental evidence for distance- and density-dependent seed and seedling survival. Journal of Ecology, 102: 845–856.

Condit, R., W. D. Robinson, R. Ibanez, S. Aguilar, A. Sanjur, R. Mart Inez, And S. J. Wright. (2001). The Status of the Panama Canal Watershed and Its Biodiversity at the Beginning of the 21st Century: Long-term ecological studies reveal a diverse ﬂora and fauna near the Panama Canal, harbored within a corridor of forest stretching from the Caribbean to the Paciﬁc, but deforestation, land degradation, erosion, and overhunting remain threats. BoScience, 51: 389–398.

Cormont, A., Malinowska, A. H., Kostenko, O., Radchuk, V., Hemerik, L., WallisDeVries, M. F., & Verboom, J. (2011). Effect of local weather on butterfly flight behaviour, movement, and colonization: significance for dispersal under climate change. Biodiversity and Conservation, 20(3), 483-503.

Cranmer, L., D. Mccollin, and J. Ollerton. (2012). Landscape structure inﬂuences pollinator movements and directly affects plant reproductive success. Oikos, 121: 562–568.

Cusser, S., Neff, J. L., & Jha, S. (2019). Landscape context differentially drives diet breadth for two key pollinator species. Oecologia, 191(4), 873-886.

102

Cusser, S., Grando, C., Zucchi, M. I., López-Uribe, M. M., Pope, N. S., Ballare, K., ... & Jha, S. (2019). Small but critical: semi-natural habitat fragments promote bee abundance in cotton agroecosystems across both Brazil and the United States. Landscape Ecology, 34(7), 1825-1836.

Dahlgren, J. P., Bengtsson, K., & Ehrlén, J. (2016). The demography of climate‐driven and density‐regulated population dynamics in a perennial plant. Ecology, 97(4), 899-907.

Dáttilo, W., Aguirre, A., Quesada, M., & Dirzo, R. (2015). Tropical forest fragmentation affects floral visitors but not the structure of individual-based palm-pollinator networks. PloS one, 10(3), e0121275.

Davies, S. J., S. Cavers, B. Finegan, A. White, M. F. Breed, and A. J. Lowe. (2015). Pollen ﬂow in fragmented landscapes maintains genetic diversity following stand-replacing disturbance in a Neotropical pioneer tree, Vochysia ferruginea Mart. Heredity, 115: 125129.

Davis, E. S., N. Reid, and R. J. Paxton. (2012). Quantifying forage specialisation in polyphagic insects: The polylectic and rare solitary bee, Colletes ﬂoralis (Hymenoptera: Colletidae). Insect Conservation and Diversity, 5: 289–297.

Delattre, T., Baguette, M., Burel, F., Stevens, V. M., Quénol, H., & Vernon, P. (2013). Interactive effects of landscape and weather on dispersal. Oikos, 122(11), 1576-1585.

Delmas, C. E. L., N. Escaravage, P. O. Cheptou, O. Charrier, S. Ruzafa, P. Winterton, and A. Pornon. (2015). Relative impact of mate versus pollinator availability on pollen limitation and outcrossing rates in a mass-ﬂowering species. Plant Biology, 17: 209– 218.

Dick, C.W. (2001). Genetic rescue of remnant tropical trees by an alien pollinator. Proceedings of the Royal Society of London. Series B: Biological Sciences, 268(1483), 2391-2396.

Dick, C. W., Hardy, O. J., Jones, F. A., & Petit, R. J. (2008). Spatial scales of pollen and seed- mediated gene flow in tropical rain forest trees. Tropical Plant Biology, 1(1), 20-33.

Dogterom, M. H., Matteoni, J. A., & Plowright, R. C. (1998). Pollination of greenhouse tomatoes by the North American Bombus vosnesenskii (Hymenoptera: Apidae). Journal of Economic Entomology, 91(1), 71-75.

Dormann CF, Gruber B, Fründ J. (2008). Introducing the bipartite package: analysing ecological networks. Rnews, 8: 8-11

103

Dormann, C. F., Fründ, J., Blüthgen, N., & Gruber, B. (2009). Indices, graphs, and null models: analyzing bipartite ecological networks. The Open Ecology Journal, 2(1).

Dormann CF. (2011). How to be a specialist? Quantifying specialisation in pollination networks. Network Biology, 1: 1 – 20.

Duminil, J., Daïnou, K., Kaviriri, D. K., Gillet, P., Loo, J., Doucet, J. L., & Hardy, O. J. (2016). Relationships between population density, fine-scale genetic structure, mating system and pollen dispersal in a timber tree from African rainforests. Heredity, 116(3), 295-303.

Ebeling, A., Klein, A. M., Schumacher, J., Weisser, W. W., & Tscharntke, T. (2008). How does plant richness affect pollinator richness and temporal stability of flower visits?. Oikos, 117(12), 1808-1815.

Egerer, M. H., Arel, C., Otoshi, M. D., Quistberg, R. D., Bichier, P., & Philpott, S. M. (2017). Urban arthropods respond variably to changes in landscape context and spatial scale. Journal of Urban Ecology, 3(1), jux001.

Egerer, M., Ordóñez, C., Lin, B. B., & Kendal, D. (2019). Multicultural gardeners and park users benefit from and attach diverse values to urban nature spaces. Urban Forestry & Urban Greening, 46: 126445.

Elam, D. R., Ridley, C. E., Goodell, K., & Ellstrand, N. C. (2007). Population size and relatedness affect fitness of a self-incompatible invasive plant. Proceedings of the National Academy of Sciences, 104(2), 549-552.

Elliott, S. E., and Irwin, R. E. (2009). Effects of flowering plant density on pollinator visitation, pollen receipt, and seed production in Delphinium barbeyi (Ranunculaceae). American Journal of Botany, 96(5), 912-919.

Ellstrand, N. C. (2014). Is gene ﬂow the most important evolutionary force in plants? American Journal of Botany, 101: 737–753.

Erfanian, A., Wang, G., & Fomenko, L. (2017). Unprecedented drought over tropical South America in 2016: significantly under-predicted by tropical SST. Scientific reports, 7(1), 1-11.

Estoque, M. A., Luque, J., Chandeck Monteza, M., & Garcia, J. (1985). Effects of El Nino on Panama rainfall. Geofisica Internacional, 24(a3).

Faith, D. P., P. R. Minchin, and L. Belbin. (1987). Compositional dissimilarity as a robust measure of ecological distance. Vegetatio, 69: 57–68.

104

Ferreira, P. A., Boscolo, D., Lopes, L. E., Carvalheiro, L. G., Biesmeijer, J. C., da Rocha, P. L. B., & Viana, B. F. (2020). Forest and connectivity loss simplify tropical pollination networks. Oecologia, 192(2), 577-590.

Fontaine, C., Dajoz, I., Meriguet, J., & Loreau, M. (2005). Functional diversity of plant– pollinator interaction webs enhances the persistence of plant communities. PLoS Biology, 4(1), e1.

Fontaine, C., Collin, C. L., & Dajoz, I. (2008). Generalist foraging of pollinators: diet expansion at high density. Journal of Ecology, 96(5), 1002-1010.

Fordham, D. A., Resit Akçakaya, H., Araújo, M. B., Elith, J., Keith, D. A., Pearson, R., ... & Tozer, M. (2012). Plant extinction risk under climate change: are forecast range shifts alone a good indicator of species vulnerability to global warming?. Global Change Biology, 18(4), 1357-1371.

Forrest, J. R. (2015). Plant–pollinator interactions and phenological change: what can we learn about climate impacts from experiments and observations?. Oikos, 124(1), 4-13.

Fox, C. W., & Reed, D. H. (2011). Inbreeding depression increases with environmental stress: an experimental study and meta‐analysis. Evolution: International Journal of Organic Evolution, 65(1), 246-258.

Fox, J., & Weisberg, S. (2018). An R companion to applied regression. Sage Publications.

Frankie, G. W., Thorp, R. W., Schindler, M., Hernandez, J., Ertter, B., & Rizzardi, M. (2005). Ecological patterns of bees and their host ornamental flowers in two northern California cities. Journal of the Kansas Entomological Society, 78(3), 227-247.

Frankie, G.W., Vinson, S.B., Rizzardi, M.A., Griswold, T.L., Coville, R.E., Grayum, M.H., Martinez, L.E.S., Foltz-Sweat, J. and Pawelek, J.C. (2013). Relationships of bees to host ornamental and weedy flowers in urban Northwest Guanacaste Province, Costa Rica. Journal of the Kansas Entomological Society, 86(4), 325-351.

Frohn, R. C., and Hao, Y. (2006). Landscape metric performance in analyzing two decades of deforestation in the Amazon Basin of Rondonia, Brazil. Remote Sensing of Environment, 100(2), 237-251.

Garbuzov, M., & Ratnieks, F. L. (2014). Quantifying variation among garden plants in attractiveness to bees and other flower‐visiting insects. Functional Ecology, 28(2), 364- 374. Gardner, T. A., Barlow, J., Chazdon, R., Ewers, R. M., Harvey, C. A., Peres, C. A., & Sodhi, N. S. (2009). Prospects for tropical forest biodiversity in a human‐modified world. Ecology letters, 12(6), 561-582. 105

Geslin, B., Gauzens, B., Thebault, E., & Dajoz, I. (2013). Plant pollinator networks along a gradient of urbanisation. PloS one, 8(5).

Ghalambor, C. K., Huey, R. B., Martin, P. R., Tewksbury, J. J., & Wang, G. (2006). Are mountain passes higher in the tropics? Janzen's hypothesis revisited. Integrative and comparative biology, 46(1), 5-17.

Ghazoul, J. (2005). Pollen and seed dispersal among dispersed plants. Biological Reviews, 80(3), 413-443.

Gibson, L., Lee, T. M., Koh, L. P., Brook, B. W., Gardner, T. A., Barlow, J., ... & Sodhi, N. S. (2011). Primary forests are irreplaceable for sustaining tropical biodiversity. Nature, 478(7369), 378-381.

Gilman, S. E., Urban, M. C., Tewksbury, J., Gilchrist, G. W., & Holt, R. D. (2010). A framework for community interactions under climate change. Trends in ecology & evolution, 25(6), 325-331.

Gilman, R. T., Fabina, N. S., Abbott, K. C., & Rafferty, N. E. (2012). Evolution of plant– pollinator mutualisms in response to climate change. Evolutionary Applications, 5(1), 2-16.

Girao, L. C., A. V. Lopes, M. Tabarelli, and E. M. Bruna. 2007. Changes in tree reproductive traits reduce functional diversity in a fragmented Atlantic forest landscape. PloS One, 2: e908.

Goddard, M. A., Dougill, A. J., & Benton, T. G. (2010). Scaling up from gardens: biodiversity conservation in urban environments. Trends in ecology & evolution, 25(2), 90-98.

González-Varo, J. P., Biesmeijer, J. C., Bommarco, R., Potts, S. G., Schweiger, O., Smith, H. G., ... & Vilà, M. (2013). Combined effects of global change pressures on animal- mediated pollination. Trends in ecology & evolution, 28(9), 524-530.

Goulson, D. (2000). Why do pollinators visit proportionally fewer flowers in large patches?. Oikos, 91(3), 485-492.

Green, W. (2009). USDA PLANTS compilation, version 1. NRCS: The PLANTS Database. Retrieved January 14, 2020 from https://plants.sc.egov.usda.gov/java/.

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

106

Grimm, N. B., Chapin III, F. S., Bierwagen, B., Gonzalez, P., Groffman, P. M., Luo, Y., ... & Williamson, C. E. (2013). The impacts of climate change on ecosystem structure and function. Frontiers in Ecology and the Environment, 11(9), 474-482.

Hadley, A. S., and M. G. Betts. (2012). The effects of landscape fragmentation on pollination dynamics: Absence of evidence not evidence of absence. Biological Reviews, 87: 526– 544.

Hansen, M. C., P. Potapov, R. Moore, M. Hancher, S. Turubanova, A. Tyukavina, D. Thau, S. Stehman, S. Goetz, T. Loveland, A. Kom-Mareddy, A. Egorov, L. Chini, C. O. Justice, and J. R. Townshend. (2013). High-resolution global maps of 21st-century forest cover change. Science, 342: 850–853.

Hansen, J., and Sato, M. (2016). Regional climate change and national responsibilities. Environmental Research Letters, 11(3), 034009.

Hardesty, B. D., S. P. Hubbell, and E. Bermingham. (2006). Genetic evidence of frequent long-distance recruitment in a vertebrate-dispersed tree. Ecological Letters, 9: 516–525.

Hardy, O. J., L. Maggia, E. Bandou, P. Breyne, H. Caron, M. Chevallier, A. Doligez, C. Dutech, A. Kremer, C. Latouche-Halle , V. Trois- Poux, V. Veron, and B. Degen. (2006). Fine-scale genetic structure and gene dispersal inferences in 10 neotropical tree species. Molecular Ecology, 15: 559–571.

Harmon‐Threatt, A. N., de Valpine, P., & Kremen, C. (2017). Estimating resource preferences of a native bumblebee: the effects of availability and use–availability models on preference estimates. Oikos, 126(5), 633-641.

Harrison, T., & Winfree, R. (2015). Urban drivers of plant‐pollinator interactions. Functional Ecology, 29(7), 879-888.

Hasegawa, Y., Y. Suyama, and K. Seiwa. (2015). Variation in pollen-donor composition among pollinators in an entomophilous tree species, Castanea crenata, revealed by single-pollen genotyping. PloS One, 10: e0120393.

Harrison R.D. (2000). Repercussions of El Niño: drought causes extinction and the breakdown of mutualism in Borneo. Proceedings of the Royal Society of London Series B: Biological Sciences, 267, 911– 915.

Hennig, E. I., & Ghazoul, J. (2011). Plant–pollinator interactions within the urban environment. Perspectives in Plant Ecology, Evolution and Systematics, 13(2), 137- 150.

107

Hernandez, J. L., Frankie, G. W., & Thorp, R. W. (2009). Ecology of urban bees: a review of current knowledge and directions for future study. Cities and the Environment (CATE), 2(1), 3.

Hirao, A. S. (2010). Kinship between parents reduces offspring fitness in a natural population of Rhododendron brachycarpum. Annals of Botany, 105(4), 637-646.

Hoiss, B., Krauss, J., & Steffan‐Dewenter, I. (2015). Interactive effects of elevation, species richness and extreme climatic events on plant–pollinator networks. Global Change Biology, 21(11), 4086-4097.

Holmquist, K. G., Mitchell, R. J., & Karron, J. D. (2012). Influence of pollinator grooming on pollen‐mediated gene dispersal in Mimulus ringens (Phrymaceae). Plant Species Biology, 27(1), 77-85.

Homer, C., Dewitz, J., Yang, L., Jin, S., Danielson, P., Xian, G., ... & Megown, K. (2015). Completion of the 2011 National Land Cover Database for the conterminous United States–representing a decade of land cover change information. Photogrammetric Engineering & Remote Sensing, 81(5), 345-354.

Huang, Z. H., Liu, H. L., & Huang, S. Q. (2015). Interspecific pollen transfer between two co-flowering species was minimized by bumblebee fidelity and differential pollen placement on the bumblebee body. Journal of Plant Ecology, 8(2), 109-115.

Hülsmann, M., von Wehrden, H., Klein, A. M., & Leonhardt, S. D. (2015). Plant diversity and composition compensate for negative effects of urbanization on foraging bumble bees. Apidologie, 46(6), 760-770.

Hufford, K., and J. L. Hamrick. (2003). Viability selection at three early life stages of the tropical tree, Platypodium elegans (Fabaceae, Papilionoideae). Evolution, 57: 518–26.

Hutchison, D., and A. Templeton. (1999). Correlation of Pairwise Genetic and Geographic Distance Measures: Inferring the Relative Inﬂuences of Gene Flow and Drift on the Distribution of Genetic Variability. Evolution, 53: 1898–1914.

Ibáñez, R., Condit, R., Angehr, G., Aguilar, S., Garcia, T., Martinez, R., ... & Heckadon, S. (2002). An ecosystem report on the Panama Canal: monitoring the status of the forest communities and the watershed. Environmental Monitoring and Assessment, 80(1), 65- 95.

Ismail, S. A., Ghazoul, J., Ravikanth, G., Uma Shaanker, R., Kushalappa, C. G., & Kettle, C. J. (2012). Does long‐distance pollen dispersal preclude inbreeding in tropical trees? Fragmentation genetics of Dysoxylum malabaricum in an agro‐forest landscape. Molecular Ecology, 21(22), 5484-5496. 108

Inouye, D.W. (1978). Resource partitioning in bumble bees: experimental studies of foraging behavior. Ecology, 59, 672–678.

IPBES. (2016). The assessment report of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services on pollinators, pollination, and food production. IPBES, Bonn, Germany.

Jha, S., and Dick, C. W. (2010). Native bees mediate long-distance pollen dispersal in a shade coffee landscape mosaic. Proceedings of the National Academy of Sciences, 107(31), 13760-13764.

Jones, F. A., and Comita, L. S. (2008). Neighbourhood density and genetic relatedness interact to determine fruit set and abortion rates in a continuous tropical tree population. Proceedings of the Royal Society B: Biological Sciences, 275(1652), 2759- 2767.

Jha, S., and C. W. Dick. (2008). Shade coffee farms promote genetic diversity of native trees. Current Biology, 18: 1126–1128.

Jha, S., and C. W. Dick. (2009). Isolation and characterization of nine microsatellite loci for the tropical understory tree Miconia afﬁnis Wurdack (Melastomataceae). Molecular Ecology Resources, 9: 344–345.

Jha, S., and C. W. Dick. (2010). Native bees mediate long-distance pollen dispersal in a shade coffee landscape mosaic. Proceedings of the National Academy of Sciences, 107: 13760–13764.

Jha, S., & Kremen, C. (2013). Resource diversity and landscape-level homogeneity drive native bee foraging. Proceedings of the National Academy of Sciences, 110(2), 555-558.

Jha, S., Stefanovich, L. E. V., & Kremen, C. (2013). Bumble bee pollen use and preference across spatial scales in human‐altered landscapes. Ecological entomology, 38(6), 570- 579.

Johnson, A. L., Borowy, D., & Swan, C. M. (2018). Land use history and seed dispersal drive divergent plant community assembly patterns in urban vacant lots. Journal of applied ecology, 55(1), 451-460.

Johnson, D. H. (1980). The comparison of usage and availability measurements for evaluating resource preference. Ecology, 61(1), 65-71.

Jones, O. R., and J. Wang. (2010). COLONY: a program for parentage and sibship inference from multilocus genotype data. Molecular Ecology Resources, 10: 551–555.

109

Jones, F. A., and L. S. Comita. (2008). Neighbourhood density and genetic relatedness interact to determine fruit set and abortion rates in a continuous tropical tree population. Proceedings of the Royal Society B: Biological Sciences, 275: 2759– 2767.

Kaiser-Bunbury, C. N., Mougal, J., Whittington, A. E., Valentin, T., Gabriel, R., Olesen, J. M., & Blüthgen, N. (2017). Ecosystem restoration strengthens pollination network resilience and function. Nature, 542(7640), 223-227.

Kalinowski, S. T. (2005). HP-RARE 1.0: A computer program for performing rarefaction on measures of allelic richness. Molecular Ecology Resources, 5: 187– 189.

Karron, J. D., and R. J. Mitchell. (2012). Effects of ﬂoral display size on male and female reproductive success in Mimulus ringens. Annals of Botany, 109: 563–570.

Karron, J. D., R. J. Mitchell, K. G. Holmquist, J. M. Bell, and B. Funk. (2003). The influence of floral display size on selfing rates in Mimulus ringens. Heredity, 92: 242–248.

Kearns, C. A., & Inouye, D. W. (1993). Techniques for pollination biologists. University press of Colorado.

Kearns, C. A., Inouye, D. W., & Waser, N. M. (1998). Endangered mutualisms: the conservation of plant-pollinator interactions. Annual review of ecology and systematics, 29(1), 83-112.

Keith, D. A., Akçakaya, H. R., Thuiller, W., Midgley, G. F., Pearson, R. G., Phillips, S. J., ... & Rebelo, T. G. (2008). Predicting extinction risks under climate change: coupling stochastic population models with dynamic bioclimatic habitat models. Biology letters, 4(5), 560-563.

Kemper Center for Home Gardening. "Plant finder." Missouri Botanical Garden. Retrieved January 14, 2020 from https://www.missouribotanicalgarden.org/plantfinder/ plantfindersearch.aspx.

Kettle, C. J., C. R. Maycock, J. Ghazoul, P. M. Hollingsworth, Khoo, E., Sukri, R. S. and D. F. Burslem. (2011). Ecological implications of a ﬂower size/number trade-off in tropical forest trees. PloS One, 6: e16111.

Kitamoto N, Ueno S, Takenaka A, Tsumura Y, Washitani I, Ohsawa R (2006). Effect of flowering phenology on pollen flow distance and the consequences for spatial genetic structure within a population of Primula sieboldii (Primulaceae). American Journal of Botany, 93: 226–233.

110

Klein, A. M., Vaissiere, B. E., Cane, J. H., Steffan-Dewenter, I., Cunningham, S. A., Kremen, C., & Tscharntke, T. (2006). Importance of pollinators in changing landscapes for world crops. Proceedings of the royal society B: biological sciences, 274(1608), 303-313.

Kleijn, D., & Raemakers, I. (2008). A retrospective analysis of pollen host plant use by stable and declining bumble bee species. Ecology, 89(7), 1811-1823.

Knight, M. E., A. P. Martin, S. Bishop, J. L. Osborne, R. J. Hale, A. Sanderson, and D. Goulson. (2005). An interspeciﬁc comparison of foraging range and nest density of four bumblebee (Bombus) species. Molecular Ecology, 14: 1811–1820.

Kowarik, I. (2011). Novel urban ecosystems, biodiversity, and conservation. Environmental pollution, 159(8-9), 1974-1983.

Kramer, K., Degen, B., Buschbom, J., Hickler, T., Thuiller, W., Sykes, M. T., & de Winter, W. (2010). Modelling exploration of the future of European beech (Fagus sylvatica L.) under climate change—range, abundance, genetic diversity, and adaptive response. Forest Ecology and Management, 259(11), 2213-2222.

Kremen, C., Bugg, R. L., Nicola, N., Smith, S. A., Thorp, R. W., & Williams, N. M. (2002). Native bees, native plants, and crop pollination in California. Fremontia, 30(3-4), 41- 49.

Kremen, C. (2005). Managing ecosystem services: What do we need to know about their ecology? Ecological Letters, 8: 468–479.

Kremen, C., Williams, N. M., Aizen, M. A., Gemmill‐Herren, B., LeBuhn, G., Minckley, R., ... & Vázquez, D. P. (2007). Pollination and other ecosystem services produced by mobile organisms: a conceptual framework for the effects of land‐use change. Ecology letters, 10(4), 299-314.

Kremer, A., Ronce, O., Robledo‐Arnuncio, J. J., Guillaume, F., Bohrer, G., Nathan, R., ... & Schueler, S. (2012). Long‐distance gene flow and adaptation of forest trees to rapid climate change. Ecology letters, 15(4), 378-392.

Kudo, G., and Ida, T. Y. (2013). Early onset of spring increases the phenological mismatch between plants and pollinators. Ecology, 94(10), 2311-2320.

Kudo, G., and Cooper, E. J. (2019). When spring ephemerals fail to meet pollinators: mechanism of phenological mismatch and its impact on plant reproduction. Proceedings of the Royal Society of London. Series B: Biological Sciences, 286(1904), 20190573.

Kunin, W. E. (1993). Sex and the single mustard: population density and pollinator behavior effects on seed‐set. Ecology, 74(7), 2145-2160. 111

Kunin, W. E. (1997). Population size and density effects in pollination: pollinator foraging and plant reproductive success in experimental arrays of Brassica kaber. Journal of Ecology, 225-234.

Lander, T. A., D. H. Boshier, and S. A. Harris. (2010). Fragmented but not isolated: Contribution of single trees, small patches, and long-distance pollen ﬂow to genetic connectivity for Gomortega keule, an endangered Chilean tree. Biological Conservation, 143: 2583–2590.

Latouche‐Hallé, C., Ramboer, A., Bandou, E., Caron, H., & Kremer, A. (2004). Long‐distance pollen flow and tolerance to selfing in a neotropical tree species. Molecular Ecology, 13(5), 1055-1064.

Le Roux, J. J., and A. M. Wieczorek. (2008). Isolation and characterization of polymorphic microsatellite markers from the velvet tree, Miconia calvescens DC. (Melastomataceae). Molecular Ecology Resources, 8:961–964.

Lewis, S. L., Edwards, D. P., & Galbraith, D. (2015). Increasing human dominance of tropical forests. Science, 349(6250), 827-832.

Lin, B. B., Philpott, S. M., & Jha, S. (2015). The future of urban agriculture and biodiversity- ecosystem services: Challenges and next steps. Basic and applied ecology, 16(3), 189- 201.

Lindauer, M. (1957). Communication among the honeybees and stingless bees of India. Bee World, 38: 3–14.

Lindsey, A. H. (1984). Reproductive biology of Apiaceae. I. Floral visitors to Thaspium and Zizia and their importance in pollination. American Journal of Botany, 71(3):375–387.

Loram, A., Warren, P., Thompson, K., & Gaston, K. (2011). Urban domestic gardens: the effects of human interventions on garden composition. Environmental Management, 48(4), 808.

Loiselle, B., V. Sork, J. Nason, and C. Graham. (1995). Spatial Genetic Structure of a Tropical Understory Shrub, Psychotria ofﬁcinalis (Rubiaceae). American Journal of Botany, 82: 1420–1425.

Lowe, A. (2005). Population genetics of neotropical trees focus issue. Heredity, 95: 243– 245.

Lowe, A. J., Cavers, S., Boshier, D., Breed, M. F., & Hollingsworth, P. M. (2015). The resilience of forest fragmentation genetics—no longer a paradox—we were just looking in the wrong place. Heredity, 115(2), 97-99. 112

Lowenstein, D.M., Matteson, K.C., Xiao, I., Silva, A.M. and Minor, E.S., (2014). Humans, bees, and pollination services in the city: the case of Chicago, IL (USA). Biodiversity and conservation, 23(11), pp.2857-2874.

Lowenstein, D.M. and Minor, E.S. (2016). Diversity in flowering plants and their characteristics: integrating humans as a driver of urban floral resources. Urban Ecosystems, 19: 1735–48.

Lowenstein, D. M., Matteson, K. C., & Minor, E. S. (2019). Evaluating the dependence of urban pollinators on ornamental, non-native, and ‘weedy’ floral resources. Urban Ecosystems, 22(2), 293-302.

Luck, G. W., and G. C. Daily. (2003). Tropical countryside bird assemblages: Richness, composition, and foraging differ by landscape context. Ecological Applications, 13: 235–247.

MacIvor, J. S., Cabral, J. M., & Packer, L. (2014). Pollen specialization by solitary bees in an urban landscape. Urban ecosystems, 17(1), 139-147.

Magrach, A., W. F. Laurance, A. R. Larrinaga, and L. Santamaria. (2014). Meta-analysis of the effects of forest fragmentation on interspeciﬁc interactions. Conservation Biology, 28: 1342–1348.

Makino, T. T., K. Ohashi, and S. Sakai. (2007). How do floral display size and the density of surrounding flowers influence the likelihood of bumble bee revisitation to a plant? Functional Ecology, 21: 87–95.

Marshall, T. C., J. B. K. E. Slate, L. E. B. Kruuk, and J. M. Pemberton. (1998). Statistical conﬁdence for likelihood-based paternity inference in natural populations. Molecular Ecology, 7: 639–655.

Matteson, K. C., Ascher, J. S., & Langellotto, G. A. (2008). Bee richness and abundance in New York City urban gardens. Annals of the Entomological Society of America, 101(1), 140-150.

Matteson, K. C., & Langellotto, G. A. (2009). Bumble bee abundance in New York City community gardens: implications for urban agriculture. Cities and the Environment (CATE), 2(1), 5.

McFrederick, Q. S., & LeBuhn, G. (2006). Are urban parks refuges for bumble bees Bombus spp.(Hymenoptera: Apidae)?. Biological conservation, 129(3), 372-382.

113

McKinney, M. L. (2002). Urbanization , Biodiversity , and Conservation. BioScience, 52:883–890.

Memmott, J., Waser, N. M., & Price, M. V. (2004). Tolerance of pollination networks to species extinctions. Proceedings of the Royal Society of London. Series B: Biological Sciences, 271(1557), 2605-2611.

Memmott, J., Craze, P. G., Waser, N. M., & Price, M. V. (2007). Global warming and the disruption of plant–pollinator interactions. Ecology letters, 10(8), 710-717.

Michener, C.D. (2000). Bees of the World. Johns Hopkins University Press, Baltimore, MD.

Millar, M. A., Coates, D. J., & Byrne, M. (2014). Extensive long-distance pollen dispersal and highly outcrossed mating in historically small and disjunct populations of Acacia woodmaniorum (Fabaceae), a rare, banded iron formation endemic. Annals of botany, 114(5), 961-971.

Miller-Rushing, A. J., Høye, T. T., Inouye, D. W., & Post, E. (2010). The effects of phenological mismatches on demography. Philosophical Transactions of the Royal Society B: Biological Sciences, 365(1555), 3177-3186.

Miller-Struttmann, N. E., Geib, J. C., Franklin, J. D., Kevan, P. G., Holdo, R. M., Ebert-May, D., ... & Galen, C. (2015). Functional mismatch in a bumble bee pollination mutualism under climate change. Science, 349(6255), 1541-1544.

Mitchell, R. J., J. D. Karron, K. G. Holmquist, and J. M. Bell. (2004). The inﬂuence of Mimulus ringens ﬂoral display size on pollinator visitation patterns. Functional Ecology, 18: 116–124.

Mitchell, R. J., W. G. Wilson, K. G. Holmquist, and J. D. Karron. (2013). Influence of pollen transport dynamics on sire profiles and multiple paternity in flowering plants. PloS One, 8: e76312.

Molano-Flores, B., S. D. Hendrix, and S. B. Heard. (1999). The effect of population size on stigma pollen load, fruit set, and seed set in Allium stellatum Ker. (Liliaceae). International Journal of Plant Sciences, 160:753-757.

Morgan, J. W. (1999). Effects of population size on seed production and germinability in an endangered, fragmented grassland plant. Conservation Biology, 13:266-273.

Morandin, L. A., & Kremen, C. (2013). Bee preference for native versus exotic plants in restored agricultural hedgerows. Restoration Ecology, 21(1), 26-32.

114

Nattero, J., Malerba, R., Medel, R., & Cocucci, A. (2011). Factors affecting pollinator movement and plant fitness in a specialized pollination system. Plant Systematics and Evolution, 296(1-2), 77-85.

Nei, M. (1972). Genetic Distance between Populations. American Naturalist, 106: 283– 292.

Noreika, N., Bartomeus, I., Winsa, M., Bommarco, R., & Öckinger, E. (2019). Pollinator foraging flexibility mediates rapid plant-pollinator network restoration in semi-natural grasslands. Scientific reports, 9(1), 1-11.

Van Nieuwstadt, M., and C. R. Iraheta. (1996). Relation between size and foraging range in stingless bees (Apidae, Meliponinae). Apidologie, 27: 219–228.

Odell, E., Raguso, R. A., & Jones, K. N. (1999). Bumblebee foraging responses to variation in floral scent and color in snapdragons (Antirrhinum: Scrophulariaceae). American Midland Naturalist, 257-265.

Ohashi K, Thomson JD, D’Souza D (2007) Trapline foraging by bumble bees: IV. Optimization of route geometry in the absence of competition. Behavioral Ecology, 18:1–11.

Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D., ... & Stevens, M. H. H. (2018). vegan: Community Ecology Package. R package version 2.5-2. 2018.

Ollerton, J., R. Winfree, And S. Tarrant. (2011). How many ﬂowering plants are pollinated by animals? Oikos 120: 321–326.

Paker, Y., Yom-Tov, Y., Alon-Mozes, T., & Barnea, A. (2014). The effect of plant richness and urban garden structure on bird species richness, diversity, and community structure. Landscape and Urban Planning, 122, 186-195.

Pauls, S. U., Nowak, C., Bálint, M., & Pfenninger, M. (2013). The impact of global climate change on genetic diversity within populations and species. Molecular ecology, 22(4), 925-946.

Pawelek, J. C., Frankie, G. W., Thorp, R. W., & Przybylski, M. (2009). Modification of a community garden to attract native bee pollinators in urban San Luis Obispo, California.

Peakall, R. and P. E. Smouse (2006). GENALEX 6: Genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6: 288–295. 115

Pelabon, C., E. Albertsen, M. Falahati-Anbaran, J. Wright, and W. S. Armbruster. (2015). Does multiple paternity affect seed mass in angiosperms? An experimental test in Dalechampia scandens. Journal of Evolutionary Biology, 28: 1719–1733.

Philpott, S. M., & Bichier, P. (2017). Local and landscape drivers of predation services in urban gardens. Ecological Applications, 27(3), 966-976.

Plascencia, M., & Philpott, S. M. (2017). Floral abundance, richness, and spatial distribution drive urban garden bee communities. Bulletin of entomological research, 107(5), 658- 667.

Pope, N. S., & Jha, S. (2018). Seasonal food scarcity prompts long-distance foraging by a wild social bee. The American Naturalist, 191(1), 45-57.

Quistberg, R. D., Bichier, P., & Philpott, S. M. (2016). Landscape and local correlates of bee abundance and species richness in urban gardens. Environmental Entomology, 45(3), 592-601.

Rabeling, S. C., Lim, J. L., Tidon, R., Neff, J. L., Simpson, B. B., & Pawar, S. (2019). Seasonal variation of a plant-pollinator network in the Brazilian Cerrado: Implications for community structure and robustness. PloS one, 14(12), e0224997. Rafferty, N. E., CaraDonna, P. J., & Bronstein, J. L. (2015). Phenological shifts and the fate of mutualisms. Oikos, 124(1), 14-21.

Rafferty, N. E., CaraDonna, P. J., & Bronstein, J. L. (2015). Phenological shifts and the fate of mutualisms. Oikos, 124(1), 14-21.

Rafferty, N. E. (2017). Effects of global change on insect pollinators: multiple drivers lead to novel communities. Current Opinion in Insect Science, 23, 22-27.

Ramos, S. L. F., G. Dequigiovanni, A. M. Sebbenn, M. T. G. Lopes, P. Y. Kageyama, J. L. V. De Mace^Do, M. Kirst, and E. A. Veasey. (2016). Spatial genetic structure, genetic diversity, and pollen dispersal in a har- vested population of Astrocaryum aculeatum in the Brazilian Amazon. BMC Genetics, 17: 63.

Ramos–Jiliberto, R., de Espanés, P. M., Franco–Cisterna, M., Petanidou, T., & Vázquez, D. P. (2018). Phenology determines the robustness of plant–pollinator networks. Scientific reports, 8(1), 1-10.

Requier, F., Odoux, J. F., Tamic, T., Moreau, N., Henry, M., Decourtye, A., and Bretagnolle, V. (2015). Honey bee diet in intensive farmland habitats reveals an unexpectedly high flower richness and a major role of weeds. Ecological Applications, 25(4), 881-890.

116

Rhodes, M. K., J. B. Fant, and K. A. Skogen. (2017). Pollinator identity and spatial isolation inﬂuence multiple paternity in an annual plant. Molecular Ecology, 26: 4296–4308.

Ritchie, A. D., Ruppel, R., & Jha, S. (2016). Generalist behavior describes pollen foraging for perceived oligolectic and polylectic bees. Environmental entomology, 45(4), 909-919.

Roubik, D. W. (1989). Ecology and natural history of tropical bees. Cambridge University Press, Cambridge, UK.

Roubik, D., and M. Aluja. (1983). Flight ranges of Melipona and Trigona in tropical forest. Journal of the Kansas Entomological Society, 56: 217–222.

Rudd, H., Vala, J., & Schaefer, V. (2002). Importance of backyard habitat in a comprehensive biodiversity conservation strategy: a connectivity analysis of urban green spaces. Restoration ecology, 10(2), 368-375.

Sahli, H., and J. Conner. (2007). Visitation, effectiveness, and efﬁciency of 15 genera of visitors to wild radish, Raphanus raphanistrum (Brassicaceae). American Journal of Botany, 94: 203–209.

Saifuddin, M., & Jha, S. (2014). Colony-Level Variation in Pollen Collection and Foraging Preferences Among Wild-Caught Bumble Bees (Hymenoptera: Apidae), Environmental Entomology, 43(2), 393–401.

Schenau, E., & Jha, S. (2017). High levels of male diploidy but low levels of genetic structure characterize Bombus vosnesenskii populations across the Western US. Conservation Genetics, 18(3), 597-605.

Schleuning, M., Fründ, J., Klein, A. M., Abrahamczyk, S., Alarcón, R., Albrecht, M., ... & Dalsgaard, B. (2012). Specialization of mutualistic interaction networks decreases toward tropical latitudes. Current biology, 22(20), 1925-1931.

Schleuning, M., Fruend, J., Schweiger, O., Welk, E., Albrecht, J., Albrecht, M., ... & Boehning-Gaese, K. (2016). Ecological networks are more sensitive to plant than to animal extinction under climate change. Nature communications, 7(1), 1-9.

Schweiger, O., Biesmeijer, J. C., Bommarco, R., Hickler, T., Hulme, P. E., Klotz, S., ... & Petanidou, T. (2010). Multiple stressors on biotic interactions: how climate change and alien species interact to affect pollination. Biological Reviews, 85(4), 777-795.

Shapiro, A. M. (2002). The Californian urban butterfly fauna is dependent on alien plants. Diversity and Distributions, 8(1), 31-40.

117

Shimamoto, C. Y., Padial, A. A., da Rosa, C. M., & Marques, M. C. (2018). Restoration of ecosystem services in tropical forests: A global meta-analysis. PloS one, 13(12), e0208523.

Shwartz, A., Turbé, A., Simon, L., & Julliard, R. (2014). Enhancing urban biodiversity and its influence on city-dwellers: An experiment. Biological Conservation, 171, 82-90.

Simonit, S., and Perrings, C. (2013). Bundling ecosystem services in the Panama Canal watershed. Proceedings of the National Academy of Sciences, 110(23), 9326-9331.

Smith, R. M., Thompson, K., Hodgson, J. G., Warren, P. H., & Gaston, K. J. (2006). Urban domestic gardens (IX): composition and richness of the vascular plant flora, and implications for native biodiversity. Biological conservation, 129(3), 312-322.

Snow, D. (1965). A possible selective factor in the evolution of fruiting seasons in tropical forest. Oikos, 15: 274–281.

Soares, R. G. S., Ferreira, P. A., & Lopes, L. E. (2017). Can plant-pollinator network metrics indicate environmental quality?. Ecological Indicators, 78, 361-370.

Songer, M., Aung, M., Senior, B., DeFries, R., & Leimgruber, P. (2009). Spatial and temporal deforestation dynamics in protected and unprotected dry forests: a case study from Myanmar (Burma). Biodiversity and Conservation, 18(4), 1001-1018.

Southon, G. E., Jorgensen, A., Dunnett, N., Hoyle, H., & Evans, K. L. (2017). Biodiverse perennial meadows have aesthetic value and increase residents’ perceptions of site quality in urban greenspace. Landscape and Urban Planning, 158, 105-118.

Steffan-Dewenter, I., Münzenberg, U., Bürger, C., Thies, C., & Tscharntke, T. (2002). Scale‐ dependent effects of landscape context on three pollinator guilds. Ecology, 83(5), 1421- 1432.

Steffan-Dewenter, I., Klein, A. M., Gaebele, V., Alfert, T., & Tscharntke, T. (2006). Bee diversity and plant-pollinator interactions in fragmented landscapes. Specialization and generalization in plant-pollinator interactions, 387-410.

Stout, J. C. (2000). Does size matter? Bumblebee behavior and the pollination of Cystisus scoparius L. (Fabaceae). Apidologie, 31: 129–139.

Sushinsky, J. R., Rhodes, J. R., Possingham, H. P., Gill, T. K., & Fuller, R. A. (2013). How should we grow cities to minimize their biodiversity impacts?. Global change biology, 19(2), 401-410.

118

Thompson, K., Austin, K. C., Smith, R. M., Warren, P. H., Angold, P. G., & Gaston, K. J. (2003). Urban domestic gardens (I): putting small‐scale plant diversity in context. Journal of Vegetation Science, 14(1), 71-78.

Timmermann A, Oberhuber J, Bacher A, Esch M, Latif M, Roeckner E. (1999). Increased El Niño frequency in a climate model forced by future greenhouse warming. Nature, 398, 694 – 697.

Traveset, A., Tur, C., & Eguíluz, V. M. (2017). Plant survival and keystone pollinator species in stochastic coextinction models: role of intrinsic dependence on animal- pollination. Scientific reports, 7(1), 1-10.

Tscharntke T., Brandl R. (2004). Plant-insect interactions in fragmented landscapes. Annual Review of Entomology, 49:405- 430.

Tur, C., Castro-Urgal, R., & Traveset, A. (2013). Linking plant specialization to dependence in interactions for seed set in pollination networks. PloS one, 8(10), e78294.

Turo, K. J., & Gardiner, M. M. (2019). From potential to practical: conserving bees in urban public green spaces. Frontiers in Ecology and the Environment, 17(3), 167-175.

Vaudo, A. D., Tooker, J. F., Grozinger, C. M., & Patch, H. M. (2015). Bee nutrition and floral resource restoration. Current Opinion in Insect Science, 10, 133-141.

Violle, C., B. Enquist, B. Mcgill, L. Jiang, C. Albert, C. Hulshof, V. Jung, and J. Messier. (2012). The return of the variance: Intraspeciﬁc variability in community ecology. Trends in Ecology and Evolution, 27: 244– 52.

Vekemans X., and Hardy O.J. (2004) New insights from fine-scale spatial genetic structure analyses in plant populations. Molecular Ecology, 13,921-935.

Vranckx, G., H. Jacquemyn, B. Muys, and O. Honnay. (2011). Meta-analysis of susceptibility of woody plants to loss of genetic diversity through habitat fragmentation. Conservation Biology, 26: 228–237.

Walther‐Hellwig, K., & Frankl, R. (2000). Foraging habitats and foraging distances of bumblebees, Bombus spp.(Hym., Apidae), in an agricultural landscape. Journal of Applied Entomology, 124(7‐8), 299-306.

Wenner, A. M., D. E. Meade, and L. J. Friesen. (1991). Recruitment, search behavior, and ﬂight ranges of honey bees. American Zoologist, 31: 768–782.

119

Werrell P.A., Langellotto G.A., Morath S.U., Matteson K.C. (2009) The influence of garden size and floral cover on pollen deposition in urban community gardens. Cities and the Environment.

Westrich, P. (1996). Habitat requirements of central European bees and the problems of partial habitats. In Linnean Society Symposium Series (Vol. 18, pp. 1-16). Academic Press Limited.

Williams, N. M., Cariveau, D., Winfree, R., & Kremen, C. (2011). Bees in disturbed habitats use, but do not prefer, alien plants. Basic and Applied Ecology, 12(4), 332-341.

Winfree, R., Williams, N. M., Gaines, H., Ascher, J. S., & Kremen, C. (2008). Wild bee pollinators provide the majority of crop visitation across land‐use gradients in New Jersey and Pennsylvania, USA. Journal of applied ecology, 45(3), 793-802.

Wright, SJ and O Calderón. (2006). Seasonal, El Niño and longer-term changes in flower and seed production in a moist tropical forest. Ecology Letters, 9: 35-44.

Wright, S. J., Calderón, O., & Muller‐Landau, H. C. (2019). A phenology model for tropical species that flower multiple times each year. Ecological Research, 34(1), 20-29.

Zurbuchen, A., L. Landert, J. Klaiber, A. Muller, S. Hein, and S. Dorn. (2010). Maximum foraging ranges in solitary bees: Only few individuals have the capability to cover long foraging distances. Biological Conservation, 143: 669–676.

120

Permanent email: [email protected] This dissertation was typed by Megan O’Connell.

121