The Effects of Semi-Natural Habitat and Wildflower Plantings on Ecosystem Services, Bee Communities, and Tick Populations

The effects of semi-natural habitat and wildflower plantings on ecosystem services, bee communities, and tick populations

Christopher T. McCullough

Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of

Doctor of Philosophy In Horticulture

Megan O’Rourke - Chair Thomas Kuhar Sarah Karpanty Jacob Barney

May 8th, 2020 Blacksburg, Virginia

Keywords: biological control, natural habitat, yield, pollinators, lone star tick

Creative Commons, CC BY-NC

The effects of wildflower plots and diverse landscapes on ecosystem services, bee communities, and on-farm tick abundance

Christopher T. McCullough

ABSTRACT

Conservation of natural habitats and planting wildflower plots are two commonly promoted tactics to enhance pollination services and biological control of crop pests, which are ecosystem services that can improve agricultural outputs. There are several programs at various levels of government in the United States that landowners can use to defray the costs of implementing these conservation strategies. Studies of European Agricultural Environmental Schemes have shown these tactics to have positive outcomes for crop production. However, real-world applications of cost-sharing programs have not been evaluated in the United States on pollination services and biological control. Furthermore, these tactics may inadvertently perform ecosystem disservices, like increasing crop pests or creating habitat for disease vectors. In this study, we evaluated the effects of natural habit and wildflower plots on biological control, pollination services, bee communities, and tick populations in Eastern Virginia and Maryland. This research was conducted on 22 farms. 10 of these farms had wildflower plots that were designed by Natural Resource Conservation Service personnel, and implemented by cooperating farmers. Collards, strawberries, tomatoes, and squash were used as model systems. We measured pest density, sentinel egg predation, crop damage, seed pollination, biomass production, marketable crop yield, sampled the bee community, and recorded tick abundance in wildflower plots. Many of the measures of biological control and pollination services had idiosyncratic results in regards to the wildflower plots and natural habit in the landscape. However, the proportion of high quality yield for all four crops increased with increasing natural habitat in the landscape. Bee communities between sites with and without wildflower plots were not different. Bee abundance did increase at wildflower sites when natural habitat comprised a certain proportion of the habitat around the site. Ticks were sampled from wildflower plantings, but not in greater abundance compared to field margins. In this study, the effects of wildflower plots were overshadowed by the landscape effects of natural habitat. Government personnel that oversee these programs may need to consider the surrounding landscape when helping implement on-farm conservation measure like wildflower plots. Such measures, do not perform an ecosystem disservice in regards to ticks.

The effects of wildflower plots and diverse landscapes on ecosystem services, bee communities, and on-farm tick abundance

Christopher T. McCullough

GENERAL AUDIENCE ABSTRACT

Conservation of natural habitats and planting wildflower plots are two strategies to enhance pollination services and biological control of crop pests. These two ecosystem services are of needed to improve agricultural production without further damaging the environment. There are several programs at various levels of government in the United States that landowners can use to subsidize the costs of implementing these strategies. European studies have shown these government programs to be successful. However, these programs have not been evaluated in the United States on their ability to enhance pollination services and biological control. Furthermore, studies investigating potential ecosystem disservices these strategies. In this study, we evaluated the effects of natural habit and wildflower plots on biological control, pollination services, bee communities, and tick populations in Eastern Virginia and Maryland. This research was conducted on 22 farms. 10 of these farms had wildflower plots that were designed by Natural Resource Conservation Service personnel, and implemented by cooperating farmers. Collards, strawberries, tomatoes, and squash were used as model systems. We measured pest density, egg predation, crop damage, seed pollination, and yield. We also sampled the bee community, and recorded tick abundance in wildflower plots. There were no consistent trends for many measures of biological control and pollination services in response to the wildflower plots and natural habit. However, the proportion of high quality yield for all four crops increased with increasing natural habitat in the landscape. Bee communities between sites with and without wildflower plots were not different. Bee abundance did increase at wildflower sites when natural habitat comprised a certain proportion of the habitat around the site. Ticks were sampled from wildflower plantings, but not in greater abundance compared to field margins. In this study, the effects of wildflower plots were overshadowed by the landscape effects of natural habitat. Government personnel that oversee these programs may need to consider the surrounding landscape when helping implement on-farm conservation measure like wildflower plots. Such measures, do not perform an ecosystem disservice in regards to ticks.

Acknowledgements

There are many people that I need to thank for their contributions to the project, as well as my own development. First and foremost, is Megan, my adviser. Thank you for taking me on with this project, riding through my highs and lows, and holding me accountable to a higher standard. You have helped me realize the approach and forethought that is necessary for conducting research., as well as trying to help me improve my writing along the way.

Next, thank you to Gina for being on the Eastern Shore and handling so much of the logistical side of this work. It certainly helped ease the transition from Blacksburg to

Painter each year. I know I have to say on behalf of Laura, thanks for putting up with me for all those hours driving around in the truck. Thank you for the feedback and suggestions on different aspects of this work, and providing some inspiration to sample for ticks. I appreciated the various conversations about a plethora of things while riding in the truck.

Thank you to everyone else in the lab. Velva, baked goods aside, thanks for helping get stuff done through the university’s various systems, and running around to buy up all the coolers from our local retailers. Jennie, thanks for helping pin and wash those ~900 bees that were unearthed. I guess thanks for sharing your mom’s cookies, but her recipe could use some adjustments. I guess I ate them anyway. Mike, it is difficult to type sarcastically, so…Thanks.

Thank you to all the people who helped collect and process the data, Erika, Sarah,

Monique, Courtney, Brook, Brian, and Mika. Thank you to my committee members for the various roles you all have played. There was a lot of reshuffling with people coming

and going from the university. Thank you to all of our cooperators that let us do this work on their land. Thank you to Tom and Linda for being gracious hosts and local experts on all things related to the Eastern Shore. Thank you to Bob and Jane from the NRCS who helped implement the wildflower plots and find cooperators to work with. Thank you to

Don for providing us with the stink bug egg. Thank you Sam for providing species determinations for the bees, and showing me the collaborative spirit that exists in the bee conservation community.

Finally, thank you Laura. Thank you for putting up with the summers apart, but also visiting the Eastern Shore twice. Thank you for being supportive throughout this endeavor.

Table of contents Introduction ...... 1 References ...... 4 Chapter 1: Diverse landscapes increase marketable yields of four crops in the Eastern U.S...... 7 Introduction ...... 8 Materials and Methods ...... 10 Results ...... 15 Discussion ...... 17 References ...... 22 Supplemental Appendix 1.1...... 30 Supplemental Appendix 2 ...... 35 Chapter 2: Landscape context influences the bee conservation value of wildflower plantings ...... 52 Abstract ...... 52 Introduction ...... 53 Materials and Methods ...... 56 Results ...... 59 Discussion ...... 61 References ...... 65 Supplemental Appendix 2.1. List of plant species used in each mix for the wildflower treatment sites...... 82 Supplemental Appendix 2.2. AICc scores for scale selection, and results of other scales analyzed that were within 2 AICc points of the lowest scoring scale...... 84 Chapter 3: Conservation wildflower plantings do not enhance on-farm abundance of Amblyomma americanum (Ixodida:Ixodidae) ...... 98 Abstract ...... 98 Introduction ...... 98 Materials and Methods ...... 101 Results ...... 103 Discussion ...... 105 References ...... 109 Conclusion ...... 120 References ...... 123

List of figures Figure 1.1. The proportion of egg predation in response to the percent of semi-natural habitat (SNH) for three pests of vegetable crops. Tricopulsia ni on collards (A), Helicoverpa zea (B) and Halyomorpha halys (C) for tomatoes. Lines represent simple linear regression, and the shaded areas are the standard error. Two different colored dots indicate a significant effect of year. For A-C, N = 42, 38, and 44 for each panel respectively. NS = no significant effects detected...... 25 Figure 1.2. The average weekly abundance of four vegetable pests, and their response to the percent of semi-natural habitat (SNH) and wildflower plantings. P. rapae (A) and P. xylostella (B) were recorded on collards, and S. ornithogalli, (C) and Manduca spp. (D) on tomatoes. The p-value in panel (D) represents the significant effect of wildflower plots. Lines represent simple linear regression, and the shaded areas are the standard error. Two different colored dots indicate a significant effect of year. For panels A-D, N = 64, 88, 146, 110 for each panel respectively. NS = no significant effects detected...... 26 Figure 1.3. Various types of pest damage recorded on all four crops grown, and their response to wildflower plantings and percent of semi-natural habitat (SNH). The proportion of severely damaged collards (A). For tomatoes, the proportion of fruits damaged by chewing pests (B) and pests with piercing/sucking mouthparts (C). On strawberries, the damage caused by Lygus spp. (D), sap beetles (E), and Drosophila spp. (F). The probability of squash damage caused by squash vine borer (G) and the squash bug/Diabrotica spp. beetles (H). Data is shown on data scale. Lines represent simple linear regression, except panes (G) and (H). The shaded areas are the standard error. For panels A-H, N = 200, 119, 119, 237, 237, 237, 122, 106 for each panel respectively. NS = no significant effects detected...... 27 Figure 1.4. Response of metrics of pollination services. proportion of achenes pollinated on strawberries (A), proportion of pollen deficient strawberries (B), and the number of seed produced in squash (C), to the percent of semi-natural habitat (SNH) and wildflower plots. Data is shown on data scale. Lines represent simple linear regression, and the shaded areas are the standard error. For A-C, N = 213, 237,105 for each panel respectively. NS = no significant effects detected...... 28 Figure 1.5. The response of the proportion of the number of high quality harvested items of strawberry (A), squash (B), collards (C), and tomato (D) to the percent of semi-natural habitat (SNH) in the landscape at the 250-m scale. Data is shown on data scale. Lines represent simple linear regression, and the shaded areas are the standard error. For panels A-D, N = 237, 106, 200, and 113 for each panel respectively...... 29 Figure 2.1. NMDS plot of the of the bee communities sampled in 2017 (A) and 2018 (B) at control and wildflower fields. Circles represent 95% confidence intervals around the centroid of the sampled communities. (Stress values: 2017 = 0.18, 2018 = 0.18) ...... 70 Figure 2.2. The effects of the interaction of wildflower plots and the amount of SNH in the landscape on the Shannon-Wiener diversity index (A), bee species richness (B) and abundance (C). The scale presented has the lowest AICc score of the scales tested...... 71 Figure 2.3. The effects of the interaction of SNH in the landscape and wildflower plots on the abundance of three most common bee species sampled (A) A. virescens, (B) P.

vii

pruinosa, (C) M. bimaculatus. The scale presented has the lowest AICc score of the scales tested...... 72 Figure 3.1. Average number of A. americanum adults and nymphs sampled during each week of sampling in 2018 (A) and 2019 (B). The yearly average abundance of adults and nymphs sampled in 2018 (C) and 2019 (D). Means with the same letter are not significantly different. (Tukey’s Honestly Significant Difference) (p < 0.05)...... 112 Figure 3.2. Average number of A. americanum sampled from each sampling habit in 2018 (A) and 2019 (B). Means with the same letter are not significantly different. (Tukey’s Honestly Significant Difference) (p < 0.05) ...... 113 Figure C.1. Summary of the number and direction of the effects of all models analyzed by scale and response variable...... 124

viii

List of tables Table 2.1. List of species and the number caught over the duration of the study at wildflower and control sites...... 73 Table 2.2. Results of linear mixed effects models on the effect of wildflower plantings on the Shannon-Wiener diversity index and bee richness...... 76 Table 2.3. Results of generalized linear mixed effects model on the effect of wildflower plantings on the abundance of all bees and the three most abundant species sampled .... 77 Table 2.4. Results of linear mixed effects models testing the effects of wildflower plots and the amount of semi-natural habitat in the landscape on the species richness and Shannon-wiener diversity index of sampled bees. Scales presented had the lowest AICc of the three scales analyzed...... 78 Table 2.5. Results of generalized linear mixed effects models testing the effects of wildflower plots and the amount of semi-natural habitat in the landscape on total bee abundance and the three most abundant species sampled...... 79 Table 2.6. Results of model selection for factors affecting the effectiveness of wildflower plots for the Shannon-Wiener index, species richness, and abundance. The scale with the lowest average AICc is presented. All other scales are presented in Supplemental Appendix 2...... 80 Table 2.7. Results of model averaging for factors affecting the effectiveness of wildflower plots in promoting bee abundance at the 500-m scale. Other scales analyzed are presented in Supplemental Appendix 2...... 81 Table 3.1. Yearly averages (mean ± std. error) of environmental factors measured at each sampling location...... 114 Table 3.2. Results of generalized linear mixed models testing the interaction of life stage and habitat on the abundance of A. americanum. Adult abundance at wildflower plots is the reference category...... 115 Table 3.3. Model selection for environmental variables affecting the abundance of A. americanum adults. Models with interaction terms also include the main effects...... 116 Table 3.4. Parameter estimates of model averaging using selected models (habitat, habitat * vegetation height, and habitat * duff depth) of environmental variables affecting the abundance of A. americanum adults...... 117 Table 3.5. Model selection for environmental variables affecting the abundance of A. americanum nymphs. Models with interaction terms also include the main effects ...... 118 Table 3.6. Parameter estimates for the model used to analyze the effects of habitat and vegetation height on the abundance of A. americanum nymphs ...... 119

Introduction The human population of the world is expected to pass 9 billion by 2050 (FAO 2009). To meet the needs of the growing population more land could be used for agriculture or the intensity of agriculture could be increased. However, this would come at an extreme cost to the environment. Land used for agriculture would need to be increased by 70 million hectares by

2050 (FAO 2009). Nitrogen and phosphorus inputs would increase by 2.7 fold and 3.4 fold, respectively, from current levels (Tilman et al. 2001). Other inputs like water and pesticides would also require large increases (Tilman et al. 2001). Converting more land for agricultural production and increased external inputs would further exacerbate environmental problems.

Ecological intensification is an option that allows for a more sustainable and environmentally friendly increases in agricultural production. This paradigm seeks to enhance naturally occurring ecosystem services such as water filtration, fertilization, pollination, and pest control to lessen inputs while increasing yield to meet the food demands of a growing population (Tittonell 2014).

When 8% of arable land in 50 – 60 ha blocks was removed from production and planted with wildflowers and grasses, field level yield estimates were not different from the control where no land was removed from production (Pywell et al. 2015).

Creating or conserving natural habitats at the farm-scale and landscape scale are two well studies strategies to conserves and enhance ecosystem services like pest control and pollination services. These habitats benefit natural enemies and pollinators in multiple ways such as providing: additional food resources, overwintering habitat, nesting habitat, refuge from human disturbances, more favorable microclimates, and protection from predators (Landis et al. 2000,

Isaacs et al. 2009). An often used strategy to create natural habitat at the farm scale is by planting native wildflowers and grasses. Wildflower plantings have been shown to increase the abundance and richness of arthropod pollinators, as well as natural enemy abundance (Blaauw

and Isaacs 2015, Williams et al. 2015). This increase in beneficial arthropods can translate into yield gains (Tschumi et al. 2016), but not consistently (Nicholson et al. 2019).

Another frequently studied strategy is what effects do natural habitats have on enhancing ecosystems services, and are there benefits for farmers by conserving these habitats? Generally, more natural habitat surrounding farms is beneficial to pollinator communities and pollination services (Ricketts et al. 2008, Kennedy et al. 2013, Dainese et al. 2019). It is generally believed that natural habitats are beneficial for natural enemies and pest control services as well (Bianchi et al. 2006, Chaplin-Kramer et al. 2011, Veres et al. 2013, Rusch et al. 2016). However, recent reanalysis of these data have revealed that these responses are more idiosyncratic than a generalizable trend (Karp et al. 2018).

As research focuses on the benefits of habitat restoration and conservation, it is necessary to investigate the trade-offs that may occur between enhancing the desired ecosystem service and potential ecosystem disservices. Increasing water retention in wetlands on ranches, increased plant cover and reduced mosquito abundance; however, this also decreased forage plant abundance and decreased the species richness of fish and amphibians (Boughton et al. 2019).

These considerations are important, as people’s perceptions of the benefits and trade-offs of restoration projects impacts their support of these projects (Wilson et al. 2019). Being able to address the concerns of potential ecosystem disservices in light of the benefits may be important for the adoption of conservation efforts.

This research seeks to understand what effects wildflower plots. and SNH have on pest control services, pollination services, crop yield, and bee communities. In particular, do positive effects with regulating ecosystem service translate into gains with provisioning ecosystems

services? We also investigated a potential ecosystem disservice by sampling for Amblyomma americacanum in the wildflower plots.

References Bianchi, F. J. J. A., C. J. H. Booij, and T. Tscharntke. 2006. Sustainable pest regulation in agricultural landscapes: a review on landscape composition, biodiversity and natural pest control. Proc. R. Soc. Lond. B Biol. Sci. 273: 1715–1727. Blaauw, B. R., and R. Isaacs. 2015. Wildflower plantings enhance the abundance of natural enemies and their services in adjacent blueberry fields. Biol. Control. 91: 94–103. Boughton, E. H., P. F. Quintana‐Ascencio, D. G. Jenkins, P. J. Bohlen, J. E. Fauth, A. Engel, S. Shukla, G. Kiker, G. Hendricks, and H. M. Swain. 2019. Trade-offs and synergies in a payment-for-ecosystem services program on ranchlands in the Everglades headwaters. Ecosphere. 10: e02728. Chaplin-Kramer, R., M. E. O’Rourke, E. J. Blitzer, and C. Kremen. 2011. A meta-analysis of crop pest and natural enemy response to landscape complexity. Ecol. Lett. 14: 922– 932. Dainese, M., E. A. Martin, M. A. Aizen, M. Albrecht, I. Bartomeus, R. Bommarco, L. G. Carvalheiro, R. Chaplin-Kramer, V. Gagic, L. A. Garibaldi, J. Ghazoul, H. Grab, M. Jonsson, D. S. Karp, C. M. Kennedy, D. Kleijn, C. Kremen, D. A. Landis, D. K. Letourneau, L. Marini, K. Poveda, R. Rader, H. G. Smith, T. Tscharntke, G. K. S. Andersson, I. Badenhausser, S. Baensch, A. D. M. Bezerra, B. Caballero-Lopez, P. Cavigliasso, A. Classen, S. Cusser, J. H. Dudenhöffer, J. Ekroos, T. Fijen, P. Franck, B. M. Freitas, M. P. D. Garratt, C. Gratton, J. Hipólito, A. Holzschuh, L. Hunt, A. L. Iverson, S. Jha, T. Keasar, T. N. Kim, M. Kishinevsky, B. K. Klatt, A.-M. Klein, K. M. Krewenka, S. Krishnan, A. E. Larsen, C. Lavigne, H. Liere, B. Maas, R. E. Mallinger, E. M. Pachon, A. Martínez-Salinas, T. D. Meehan, M. G. E. Mitchell, G. A. R. Molina, M. Nesper, L. Nilsson, M. E. O’Rourke, M. K. Peters, M. Ple, D. de L. Ramos, J. A. Rosenheim, M. Rundlöf, A. Rusch, A. Sáez, J. Scheper, M. Schleuning, J. M. Schmack, A. R. Sciligo, C. Seymour, D. A. Stanley, R. Stewart, J. C. Stout, L. Sutter, M. B. Takada, H. Taki, G. Tamburini, M. Tschumi, B. F. Viana, C. Westphal, B. K. Willcox, S. D. Wratten, A. Yoshioka, C. Zaragoza-Trello, W. Zhang, Y. Zou, and I. Steffan-Dewenter. 2019. A global synthesis reveals biodiversity- mediated benefits for crop production. Sci. Adv. 14. FAO. 2009. Global Agriculture. (http://www.fao.org/fileadmin/templates/wsfs/docs/Issues_papers/HLEF2050_Global_Ag riculture.pdf). Isaacs, R., J. Tuell, A. Fiedler, M. Gardiner, and D. Landis. 2009. Maximizing arthropod- mediated ecosystem services in agricultural landscapes: the role of native plants. Front. Ecol. Environ. 7: 196–203. Karp, D. S., R. Chaplin-Kramer, T. D. Meehan, E. A. Martin, F. DeClerck, H. Grab, C. Gratton, L. Hunt, A. E. Larsen, A. Martínez-Salinas, M. E. O’Rourke, A. Rusch, K. Poveda, M. Jonsson, J. A. Rosenheim, N. A. Schellhorn, T. Tscharntke, S. D. Wratten, W. Zhang, A. L. Iverson, L. S. Adler, M. Albrecht, A. Alignier, G. M. Angelella, M. Z. Anjum, J. Avelino, P. Batáry, J. M. Baveco, F. J. J. A. Bianchi, K. Birkhofer, E. W. Bohnenblust, R. Bommarco, M. J. Brewer, B. Caballero-López, Y. Carrière, L. G. Carvalheiro, L. Cayuela, M. Centrella, A. Ćetković, D. C. Henri, A. Chabert, A. C. Costamagna, A. D. la Mora, J. de Kraker, N. Desneux, E. Diehl, T. Diekötter, C. F. Dormann, J. O. Eckberg, M. H. Entling, D. Fiedler, P. Franck, F. J. F. van Veen, T. Frank, V. Gagic, M. P. D. Garratt, A. Getachew, D. J. Gonthier, P.

B. Goodell, I. Graziosi, R. L. Groves, G. M. Gurr, Z. Hajian-Forooshani, G. E. Heimpel, J. D. Herrmann, A. S. Huseth, D. J. Inclán, A. J. Ingrao, P. Iv, K. Jacot, G. A. Johnson, L. Jones, M. Kaiser, J. M. Kaser, T. Keasar, T. N. Kim, M. Kishinevsky, D. A. Landis, B. Lavandero, C. Lavigne, A. L. Ralec, D. Lemessa, D. K. Letourneau, H. Liere, Y. Lu, Y. Lubin, T. Luttermoser, B. Maas, K. Mace, F. Madeira, V. Mader, A. M. Cortesero, L. Marini, E. Martinez, H. M. Martinson, P. Menozzi, M. G. E. Mitchell, T. Miyashita, G. A. R. Molina, M. A. Molina- Montenegro, M. E. O’Neal, I. Opatovsky, S. Ortiz-Martinez, M. Nash, Ö. Östman, A. Ouin, D. Pak, D. Paredes, S. Parsa, H. Parry, R. Perez-Alvarez, D. J. Perović, J. A. Peterson, S. Petit, S. M. Philpott, M. Plantegenest, M. Plećaš, T. Pluess, X. Pons, S. G. Potts, R. F. Pywell, D. W. Ragsdale, T. A. Rand, L. Raymond, B. Ricci, C. Sargent, J.-P. Sarthou, J. Saulais, J. Schäckermann, N. P. Schmidt, G. Schneider, C. Schüepp, F. S. Sivakoff, H. G. Smith, K. S. Whitney, S. Stutz, Z. Szendrei, M. B. Takada, H. Taki, G. Tamburini, L. J. Thomson, Y. Tricault, N. Tsafack, M. Tschumi, M. Valantin-Morison, M. V. Trinh, W. van der Werf, K. T. Vierling, B. P. Werling, J. B. Wickens, V. J. Wickens, B. A. Woodcock, K. Wyckhuys, H. Xiao, M. Yasuda, A. Yoshioka, and Y. Zou. 2018. Crop pests and predators exhibit inconsistent responses to surrounding landscape composition. Proc. Natl. Acad. Sci. E7863–E7870. Kennedy, C. M., E. Lonsdorf, M. C. Neel, N. M. Williams, T. H. Ricketts, R. Winfree, R. Bommarco, C. Brittain, A. L. Burley, D. Cariveau, L. G. Carvalheiro, N. P. Chacoff, S. A. Cunningham, B. N. Danforth, J.-H. Dudenhöffer, E. Elle, H. R. Gaines, L. A. Garibaldi, C. Gratton, A. Holzschuh, R. Isaacs, S. K. Javorek, S. Jha, A. M. Klein, K. Krewenka, Y. Mandelik, M. M. Mayfield, L. Morandin, L. A. Neame, M. Otieno, M. Park, S. G. Potts, M. Rundlöf, A. Saez, I. Steffan-Dewenter, H. Taki, B. F. Viana, C. Westphal, J. K. Wilson, S. S. Greenleaf, and C. Kremen. 2013. A global quantitative synthesis of local and landscape effects on wild bee pollinators in agroecosystems. Ecol. Lett. 16: 584–599. Landis, D. A., S. D. Wratten, and G. M. Gurr. 2000. Habitat Management to Conserve Natural Enemies of Arthropod Pests in Agriculture. Annu. Rev. Entomol. 45: 175–201. Nicholson, C. C., K. L. Ward, N. M. Williams, R. Isaacs, K. S. Mason, J. K. Wilson, J. Brokaw, L. J. Gut, N. L. Rothwell, T. J. Wood, S. Rao, G. D. Hoffman, J. Gibbs, R. W. Thorp, and T. H. Ricketts. 2019. Mismatched outcomes for biodiversity and ecosystem services: testing the responses of crop pollinators and wild bee biodiversity to habitat enhancement. Ecol. Lett. ele.13435. Pywell, R. F., M. S. Heard, B. A. Woodcock, S. Hinsley, L. Ridding, M. Nowakowski, and J. M. Bullock. 2015. Wildlife-friendly farming increases crop yield: evidence for ecological intensification. Proc R Soc B. 282: 20151740. Ricketts, T. H., J. Regetz, I. Steffan-Dewenter, S. A. Cunningham, C. Kremen, A. Bogdanski, B. Gemmill-Herren, S. S. Greenleaf, A. M. Klein, M. M. Mayfield, L. A. Morandin, A. Ochieng’, and B. F. Viana. 2008. Landscape effects on crop pollination services: are there general patterns? Ecol. Lett. 11: 499–515. Rusch, A., R. Chaplin-Kramer, M. M. Gardiner, V. Hawro, J. Holland, D. Landis, C. Thies, T. Tscharntke, W. W. Weisser, C. Winqvist, M. Woltz, and R. Bommarco. 2016. Agricultural landscape simplification reduces natural pest control: A quantitative synthesis. Agric. Ecosyst. Environ. 221: 198–204.

Tilman, D., J. Fargione, B. Wolff, C. D’Antonio, A. Dobson, R. Howarth, D. Schindler, W. H. Schlesinger, D. Simberloff, and D. Swackhamer. 2001. Forecasting Agriculturally Driven Global Environmental Change. Science. 292: 281–284. Tittonell, P. 2014. Ecological intensification of agriculture — sustainable by nature. Curr. Opin. Environ. Sustain., SI: Sustainability governance and transformation. 8: 53–61. Tschumi, M., M. Albrecht, C. Bärtschi, J. Collatz, M. H. Entling, and K. Jacot. 2016. Perennial, species-rich wildflower strips enhance pest control and crop yield. Agric. Ecosyst. Environ. 220: 97–103. Veres, A., S. Petit, C. Conord, and C. Lavigne. 2013. Does landscape composition affect pest abundance and their control by natural enemies? A review. Agric. Ecosyst. Environ., Landscape ecology and biodiversity in agricultural landscapes. 166: 110–117. Williams, N. M., K. L. Ward, N. Pope, R. Isaacs, J. Wilson, E. A. May, J. Ellis, J. Daniels, A. Pence, K. Ullmann, and J. Peters. 2015. Native wildflower plantings support wild bee abundance and diversity in agricultural landscapes across the United States. Ecol. Appl. 25: 2119–2131. Wilson, K. A., K. J. Davis, V. Matzek, and M. E. Kragt. 2019. Concern about threatened species and ecosystem disservices underpin public willingness to pay for ecological restoration. Restor. Ecol. 27: 513–519.

Chapter 1: Diverse landscapes increase marketable yields of four crops in the Eastern U.S.

Abstract

Farms with wildflower plots and semi-natural habitat (SNH) surrounding them can benefit from multiple ecosystem services. These include increased pollination and biological control, which can contribute to increases in agricultural yields. Recent data syntheses, however, indicate that landscape-ecosystem service relationships can be highly unpredictable. Underlying this unpredictability may be the variety of research methodologies employed and the fact that there are relatively few landscape-scale studies that measure agricultural yields. We set out to examine a wide array of ecosystem services, including yields, across four different crops in highly controlled potted plant experiments on 22 farms across eastern Virginia and Maryland; ten of those farms were also planted with wildflower plots. We measured: biological control, pest densities, pollination, and marketable yields of collards, tomatoes, strawberries and winter squash. The presence of on-farm wildflower plots had very little influence on ecosystem service benefits and SNH in the landscape resulted in very few significant relationships with pest control or pollination. Marketable yields, however, increased consistently across all four crops with increasing amounts of semi-natural habitat in the surrounding landscapes. This consistency indicates that landscapes are having cumulative effects across multiple ecosystem services and over entire growing seasons that affect crop productivity. Studies focusing on particular ecosystem services that are measured intermittently during the growing season or that measure yields without controlling for differences in growing conditions among fields may not accurately reflect the benefits of semi-natural habitats for agricultural production.

Introduction

The effects of habitat diversity on and surrounding farms on pest control and pollination services have been extensively researched over the past few decades (1–3). Wildflower plantings and semi-natural habitats (SNH) around farms are generally believed to conserve beneficial organisms such as pollinators and natural enemies by providing nesting habitat, overwintering sites, supplemental food sources, and refuge from pesticides (4). These beneficial organisms, in turn, are expected to enhance pest control and pollination services and contribute to increased crop yields. Enhancing these relationships has the potential to meet the needs of farmers, while maintaining the quality of the environment (5).

Even though landscape-scale effects of SNH have been studied extensively over the last few decades, few generalizable trends have been empirically validated for pest control ecosystem services. A reanalysis of data from 132 studies found that broad generalizations are difficult to make about the effects of SNH on pest control services, even showing that pest abundance increases with SNH about 10% of time (6). When selecting subsets of the data by region, response, or predictor variable, the ability of the model to predict responses to SNH was only minimally improved (6). Further inconsistency may be due to the variety of response variables measured. For each of the three broad categories used by Karp et al. (6), abundance, activity, and yield, there were 27, over 100, and 18 different methods used to measure response variables, for each group respectively. In comparison to the uncertainty regarding the effects of SNH on pest control, increasing plant diversity on farm tends to consistently enhance natural enemy abundance and pest control (7–10). Planting native wildflowers is one common tactic to increase plant diversity. The effectiveness of wildflower plantings to enhance pest control may be modulated by the amount of SNH in the landscape (11). For example, aphid parasitism was

enhanced at wildflower sites relative to control sites when SNH in the landscape ranged from 50

-75% (12).

Pollination services generally have more positive responses to SNH on and surrounding farms than pest control. A meta-analysis of 39 studies found that bee species abundance and richness increased as the amount of SNH in the landscape increased (13). Having robust native pollinator communities is important for getting high quality and quantity crop yield, even when honey bees, Apis mellifera L., are present (14). Maintaining SNH in the landscape is important for continued delivery of pollination services, as pollinator visitation, richness, and fruit set decline with increasing distance from SNH (15, 16). Visitation rates to pollinator-dependent crops declined by 50% when only 0.6 km away from SNH (15). On-farm wildflower plantings show similar effects as landscape-scale SNH for boosting pollinator abundance, richness, and pollination services (17). As with pest control, the amount of SNH around wildflower plantings plays a role in their effectiveness to enhance pollinator communities and ecosystem services (18–

20). For example, bee visitation to strawberry flowers increased at wildflower sites compared to control sites when SNH made up 25-55% of the area around fields (18). Like pest control services, a variety of measures are used to quantify pollination services, which can complicate comparisons between studies. In a meta-analysis of 49 studies on landscape effect on pollination services, 21 different methods were used to quantify pollination services (21).

While the effects of habitat diversity on pest and pollination services have been relatively well researched, there are far fewer studies measuring impacts on crop yields, despite the clear importance to farmers (22). In one European review, only 5% of the studies measured crop yield impacts (23). In the reanalysis of data done by Karp et al. (6), only 53 of 359 response variables were crop yield measurements. This could be due to the difficulty in isolating the effects of SNH

from confounding factors that affect yield like soil type or field management (24). Growing crops in containers allows for better control of abiotic factors that affect plant growth (25), allowing greater emphasis to be placed on the biotic factors at work.

The objective of this research was to test the effects of on-farm wildflower plantings and landscape-scale SNH on multiple ecosystem services and crops in one field experiment. We measured biological control, pest densities, crop damage, pollination, and marketable yields in collards (Brassica oleracea), tomatoes (Solanum lycopersicu), strawberries (Fragaria x ananassa), and winter squash (Cucurbita maxima). We expected that on-farm wildflower plots and increasing SNH in the landscape would contribute to enhanced pest and pollination services.

Furthermore, we expected these enhanced ecosystem services to cascade into increased crop yields.

Materials and Methods

This research was conducted at 22 sites in Virginia and Maryland on the Delmarva

Peninsula and in Virginia Beach. The sites were selected to create a range of SNH in a 1-km area surrounding, ranging from 10 – 70% of the cover. Field sites were at least 2.5 km away from each other. Nine of these sites had wildflower plantings established in the spring of 2016. One site had a wildflower planting that was established in the spring of 2015. The average wildflower plot size was 2,360 m2, plot sizes ranged from 561 – 8,600 m2. More details on the location of the field sites, site characteristics, and wildflower mixes used are in supplemental appendix 1.1.

At each field site, collards (Brassica oleracea var. Top Bunch), tomatoes (Solanum lycopersicum var. Defiant), strawberries (Fragaria x ananassa var. Chandler), and winter squash

(Cucurbita maxima var. Gold Nugget) were grown in 190-litre plastic containers (Rubbermaid,

Atlanta, GA). Holes were drilled into the bottoms of the containers to allow adequate drainage.

Containers were filled with Sun Gro soil (BFG, Burton, OH). These four crops represent spring and summer season crops that are commonly grown in the study region. Collards and strawberries were grown at 20 sites both years. Tomatoes and squash were grown at 22 sites in

2017 and 19 sites in 2018. At wildflower fields, containers were placed next to the wildflower plot or approximately 2 m away from the wildflower plots.

All crops were started in a greenhouse before being transplanted into the field. Collards were transplanted during the week of May 15th in 2017 and May 7th in 2018. Five collards were grown at each site in one container. Collards were harvested during the week of June 19th in

2017 and June 11th in 2018. Tomatoes were transplanted into the containers during the week of

June 26 in 2017 and June 11th in 2018. Two tomato plants were grown at each site in 2017, and four in 2018. Two tomatoes were planted in each container. One container was used for tomatoes in 2017 and two in 2018. Strawberries were transplanted into the field during the week of April

3rd in 2017 and the week of April 4th in 2018. Eight plants were grown at each site, with four plants per container. Strawberry plants were removed and squash planted in the containers during the week of June 26th in 2017 and July 2nd in 2018. A total of four squash were grown at each site, with two plants per container. Individual plants at the field were the unit of measure, and field site was the replicate. Crops were grown following common production practices for the region (26).

Landscape data

Landcover data was retrieved from the 2017 data layer of USDA Cropscape, which has a

30-mpixel resolution. Cover classes were collapsed into three broad categories: agriculture,

semi-natural habitat, and developed. Open water/background classes were removed from the data. Semi-natural habitat comprises the cover classes: evergreen forest, deciduous forest, mixed forest, herbaceous wetlands, woody wetlands, and shrubland. Data were extracted for 250, 500,

1000-m radii around each field site, centered on the location of the bins. Data manipulation and extraction was done using ArcMap v10.5.1 (ESRI, Redlands, CA).

Biological control

Sentinel egg masses were used to quantify biological control. A common lepidopteran pest, cabbage looper, Trichoplusia ni (Hübner), eggs were used on collards; whereas, brown marmorated stink bug, Halyomorpha halys (Stål), and corn earworm, Helicoverpa zea (Boddie) eggs were used on tomatoes. Egg masses were glued to wax paper and fastened to the bottom sides of leaves with paper clips. Egg masses were left in the field for 48 hrs. Four egg masses of each species were deployed at each field. Two deployments of egg cards were done for collards and tomatoes each year. Eggs masses were deployed on collards during the weeks of June 6th and

June 9th in 2017, and May 28th and June 4th 2018. Egg masses were set out on tomatoes during the weeks of in July 17th and August 14th 2017, and July 16th and August 18th in 2018. The level of biological control was measured as the proportion of damaged or missing eggs after exposure.

Pests

Collards and tomatoes were scouted weekly for pests. Densities of the most abundant pests, imported cabbageworm, Pieris rapae L., and diamondback moth, Plutella xylostella (L.), were recorded on each collard plant. Similarly, yellowstriped armyworm, Spodoptera ornithogalli (Guenée) and hornworms, Manduca spp. [M. sexta (L.) and M. quinquemaculata

(Haworth)] were recorded for tomatoes.

Pest damage was recorded on all four crops. After harvest, damage was quantified as the proportion of harvested items damaged by the causative agent. On strawberries, damage caused by Lygus spp., Drosophila spp., and strawberry sap beetles was recorded. Strawberry sap beetles are a species complex consisting of Carpophilus lugubris Murray, Stelidota geminate (Say),

Glischrochilus quadrisignatus (Say). Damage to squash plants caused by squash vine borer,

Melitta curcurbitae (Harris) was recorded. Due to the sporadic nature of the damage caused by squash bug, Anasa tristis (DeGeer), and Diabrotica spp. beetles, they were quantified together.

Leaf damage on collards caused by arthropod pests with chewing mouthparts was recorded.

Leaves greater than 10 % defoliated were considered damaged (27). For tomatoes, piercing- sucking and chewing damage caused by arthropods to the fruit were recorded.

Pollination

Pollination efficiency was calculated on strawberries as the pollination ratio of a random subsample of all harvested berries at the field. The pollination ratio is the number of fertilized achenes divided by total achenes per berry. Successful pollination of achenes was determined by the size of the achene, as fertilized achenes are larger than unfertilized ones. The proportion of seriously damaged strawberries due to insufficient pollination was also measured. Pollen deficient berries are misshapen and have unfertilized achenes, compared to Lygus spp. damage where berries are misshapen but achenes can be fertilized. Pollination efficiency in squash was quantified as the total number of seeds in mature, harvested squash.

Yield

Ripe strawberries, ≥ 75% pink or red surface, were harvested as they appeared beginning in mid-May until fruit set ceased in mid-June. Squash fruit were harvested in the fall when the

connecting vine had senesced. Collards were once-over harvested five weeks from their transplant date. Tomato fruits were harvested once they were in the breaker stage. Tomatoes fruits were picked from mid-August to early September. Yield for these four crops was quantified in three ways, total harvested mass, grade 1 marketable fresh weight, and proportion of grade 1 marketable fruits or leaves divided by the total number of harvested fruit and leaves.

By analyzing yield this way, we were able to standardize the production potential of each plant, and reduce variability. Grading of the crops followed USDA guidelines (28–31).

Statistical analysis

All analyses were conducted with R Statistical Software v 3.5.2 (32). We examined the effects of wildflower planting and the proportion of SNH habitat on all measures of ecosystem services. Wildflower plot presence, the percent of SNH in the landscape, and their interaction were fixed effects. Year was included as a fixed effect, but was considered a nuisance variable.

Field was a random effect; for pest abundance and sentinel egg mass data, date was nested within field as a random effect. The analysis had two steps. First, for each response variable, models were constructed for each landscape scale (250, 500, 1000-m). Second, stepwise reduction was done on all models within 2 Akaike information criterion points, corrected for small sample size,

(AICc) of the scale with the lowest AICc value. The reduction process started by testing the interaction model, then the additive model, and then comparing single term models. If AICc values did not decrease, the reduction process was stopped. If AICc values did not decrease by more than two points, the more parsimonious model was selected. AICc values were obtained with the ‘AICc’ function in the package ‘MuMIn’ (33). All models were further analyzed for significant predictor variables (Supplemental Appendix 1.2). Final models had variance inflation factors (VIFs) under 6 (34). If VIFs were higher than 6, the term with the higher VIF was

dropped from the model. This occurred for all models analyzing the total harvested mass for each of the crops.

Most statistical models, unless noted, were linear mixed-effects models constructed using

‘lme’ in the package ‘LmerTest’ (35). Response variable data were transformed as necessary to improve the homoscedasticity of the variance. Seed production in squash and grade 1/marketable weight of collards, squash, and tomatoes were square root transformed. Arcsine transformations we applied to T. ni, H. halys, and H. zea egg predation, Lygus damage to strawberries, chewing and piercing-sucking damage to tomatoes, seed pollination in strawberries, and the proportion of pollen deformed strawberries. Data that could not be transformed to meet assumptions of normality and homoscedasticity of the variance were analyzed using generalized linear mixed models (GLMM) using ‘glmmadmb’ in the package ‘glmmADMB’ (36). The proportions of grade 1 squash harvested were analyzed using a GLMM with a beta distribution. The proportion of grade 1 harvested squash was transformed to shift values from 0 and 1 to meet the assumptions of the beta distribution using the equation [y(N-1)+ s]/N where N is the sample size, y is the observed response, and s had a chosen value of 0.5 (37). Squash vine borer and squash bug/cucumber beetle damage to squash were analyzed using logistic regression. Pest abundance data were analyzed using a negative binomial distribution. All figures depicting data are presented as simple linear regressions while the levels of significance are derived from the models described.

Results

A total of 30 sets of models were analyzed to test the response of each metric of ecosystem services to wildflower plantings and SNH. Model reduction resulted in 13 wildflower planting only models and 16 landscape only models. Only one model with the wildflower

planting by landscape interaction was analyzed. For the landscape only models, the 250-m scale was included for analysis 16 times, the 500-m scale 13 times, and the 1-km scale 12 times. The results of all models can be found in supplemental appendix 1.2.

The effects of on-farm wildflower plantings and landscape-scale SNH on biological control services was variable. Egg predation for T.ni, H. halys, and H. zea were not affected by the amount of SNH in the surrounding landscape at any scale considered (Fig. 1.1). No significant effects of SNH on the abundance of P. xylostella, P. rapae, and S. ornithogalli were detected (Fig. 2). Fields with wildflower plantings had significantly fewer Manduca spp. larvae than control fields (z = -1.97, p = 0.048) (Fig. 1.2).

On-farm wildflower plantings and landscape-scale SNH also had variable effects on pest damage to crops. The presence of wildflower plantings led to an increase in the proportion of berries damaged by Drosophila spp. (t = 2.35, p = 0.029) (Fig. 1.3). Strawberries damaged by

Lygus spp. trended towards being lower at sites with a wildflower planting (t = -2.05, p = 0.054)

(Fig. 1.3). Wildflower plantings had no effect on sap beetle damage (Fig 1.3). Increasing semi- natural habitat within a 1-km buffer increased the odds of squash vine borer damage (z = 2.41, p

= 0.016) (Fig 1.3), while no significant effects of landscapes were detected for squash bug/beetle damage (Fig 1.3.). As the amount of SNH increased in a 250-m radius, the proportion of damaged collards decreased (t = -2.33, p = 0.031) (Fig. 1.3). Chewing damage to tomato fruits showed no response landscape effects (Fig. 1.3). Piercing-sucking damage to tomatoes decreased as the amount of SNH increased at the 500-m scale (t = -3.56, p = 0.0001) (Fig 1.3).

Pollination was not significantly affected by wildflower plots or SNH in the landscape.

In strawberries, achene pollination ratio and the proportion of pollen deficient strawberries

showed no relationships with any habitat metrics measured (Fig 1.4.) Likewise, no significant effects were found for seed production in squash (Fig. 1.4).

Consistent relationships were found between SNH in the landscape and yields for all four crops. This pattern was apparent after controlling for the production potential of each individual plant. The proportion of SNH in the landscape at the 250-m scale had a positive effect on the proportion of grade 1 marketable yield per plant for collards (t = 2.33, p = 0.03) (Fig. 1.5), tomatoes (t = 2.91, p = 0.009), strawberries (t = 2.65, p = 0.015), and squash (z = 2.02, p = 0.04).

On the other hand, no significant effects of SNH or wildflower plots were detected for any crop on total harvested yield (Fig. 1.5), and no effects were detected for the weight of grade 1 marketable yield for collards, tomatoes, and strawberries (Fig S1.2.1.). More grade 1 marketable yield was harvested from squash plants at fields with wildflower plantings than control fields (t =

2.85, p = 0.01) (Fig.S1.2.2).

Discussion

Marketable yields of four crops increased with increasing amounts of SNH in the landscape at the 250-m scale. To our knowledge, this is the only study to report consistent significant yield increases in relation to natural habitat in the landscape for four different crops within the same study system. Fewer than 10 of the 53 responses that measured yield had a positive response to SNH, approximately half of the 53 responses used had no consistent response to SNH (6). Part of our result could be due to measuring yield as the proportion of high- quality harvested items rather than measuring by weight or other metrics. This standardizes yield across fields and crops, and lessens the variability associated with measures of weight and smaller sample sizes. In studies where researchers have to constantly visit multiple field sites, there may not be time or space to grow more plants. Using the proportion of high-quality yield

from crops grown in containers represents a good trade-off between reflecting the use of yield as a response and the logistical challenges associated with gathering yield data.

Growing plants in containers greatly homogenizes the growing environment across all the research sites and simplifies detecting relationships between landscapes and ecosystem services.

Potentially confounding factors that are labor intensive to measure or difficult to account for, like within field soil variability or legacy effects of management, can be avoided this way (24). Of the published papers used in two recent reanalyses of the effects of landscape factors on crop provisioning services and yield, none reported growing plants in containers (6, 21). Searching the literature on the landscape effects on pest control and pollination services, relatively few studies can be found that utilize plants in containers and report some metric of yield (19, 38–40). All four of the studies that do, focus on pollination services and report a positive effect of SNH in the landscape on some aspect of yield (19, 38–40). Using plants in containers is not a new idea

(1, 41), but rather one that needs more utilization as it can control for many exogenous factors affecting yields. Using crops in containers may limit the number of plants to take yield measurements from. Using plants in containers is meant to demonstrate the benefits of ecosystem services on enhancing the quality of crop yield, not serve as a variety trial.

Yield is an important metric that has generally been under investigated. Measuring yield in agroecological studies has been called for repeatedly for the last 40 years (42). A systematic review of European studies on the effects of SNH on ecosystem services used 270 studies published between 1988 – 2015 had only 13 studies that measured yield (23). Obtaining yield data, however, can be logistically more difficult than measuring snapshots of pest control and pollination services as many trips to each field site are necessary to keep crops alive to maturity.

Nevertheless, yield data are important because of the relevance to growers and clear links to

economics (22). In contrast, it is much less clear how a 10% increase in egg predation translates into economic returns; it will depend on the pest and crop and time of year.

Yield also represents the cumulative effects of many ecosystem services that occur throughout the growing season. It is important to understand the individual contributions of ecosystems service alone, but without considering the impacts of the combined effects of multiple ecosystem services throughout growing seasons, overall impacts on crop production may be missed. In an experimental study, the interaction of the high pollination and high pest control treatments resulted in greater yield in red clover than the main effect of either of the two ecosystem services alone (43). Increased fruit set may mean little if the plant does not survive to make it to market. Rather than designing two separate studies to measure a metric of pollination services and pest control, the success of these two ecosystem services could be inferred by measuring high quality yield of a pollinator dependent crop.

Furthermore, many measures of ecosystem services are a brief window into the conditions present at the time. Even with multiple sampling dates, these measures are often averaged over the duration of the study, as they were with this study. Averaging over a season simplifies analyses, but may mask significant interactions with variables of interest. For example, one study in California did not detect any significant effects of natural enemies or landscapes on aphid density, when aphid density was averaged over the field season but did find effects on certain weeks (44). By using crop yield as the focal measurement, researchers can utilize an indicator that synthesizes the impacts of multiple ecosystem service activities over entire growing seasons.

The ecosystem services that we measured were relatively unresponsive to wildflower plantings in contrast to previous studies that indicate they can have positive effects on ecosystem services, like increasing beneficial arthropod populations and reducing pest damage (8–10). Few effects of wildflower plantings may have been detected in this study because of the variable size and young age of the plots. The wildflower plots were planted where cooperators had space for them, resulting in a range of sizes from 560 to 12,140 m2. As habitat size increases, so does the number of flower blooms and, typically, beneficial arthropods (45, 46). Due to the relatively small size of the wildflower plots in this study, they may have been less important than the surrounding landscape for influencing pest and pollination ecosystem services (47). The age of the habitats may also have been a factor. Nine of the habitats were sampled in the second and third years after they were planted. Effects of habitats can take up to three years after planting to manifest themselves (48).

Much of the science relating ecosystem services to natural habitat in agricultural systems is directed at regulating services like biological control or pollination. While these are important considerations for ecologists, measuring ecosystem services that are important to growers and policy makers, such as yield, may increase the likelihood that the science of ecosystem services will reach decision makers who can implement science-driven land-use recommendations. This work indicates that natural habitat should be conserved extensively throughout agricultural landscapes to naturally improve crop yields for many farmers without the use of pesticides or efforts to artificially boost pollination. These results are consistent and unique, in part, due to growing crops in containers to minimize confounding abiotic factors and excess variability that can be found under natural growing conditions across many farms. Yield is a valuable metric that

ecologists should carefully measure when working in agroecosystems as it provides a season long picture of the impacts of multiple regulating ecosystem services.

Contributions: MEO provided the conceptual framework. GA and CTM collected the data. CTM preformed the analyses and wrote the original draft of the manuscript. MEO and GA provided revisions for the manuscript.

References 1. C. Thies, T. Tscharntke, Landscape structure and biological control in agroecosystems. Science 285, 893–895 (1999). 2. C. Kremen, N. M. Williams, R. L. Bugg, J. P. Fay, R. W. Thorp, The area requirements of an ecosystem service: crop pollination by native bee communities in California. Ecol. Lett. 7, 1109–1119 (2004). 3. I. Steffan-Dewenter, T. Tscharntke, Effects of habitat isolation on pollinator communities and seed set. Oecologia 121, 432–440 (1999). 4. G. M. Gurr, S. D. Wratten, D. A. Landis, M. You, Habitat management to suppress pest populations: progress and prospects. Annu. Rev. Entomol. 62, 91–109 (2017). 5. D. Kleijn, et al., Ecological intensification: bridging the gap between science and practice. Trends Ecol. Evol. 34, 154–166 (2019). 6. D. S. Karp, et al., Crop pests and predators exhibit inconsistent responses to surrounding landscape composition. Proc. Natl. Acad. Sci., E7863–E7870 (2018). 7. D. K. Letourneau, et al., Does plant diversity benefit agroecosystems? A synthetic review. Ecol. Appl. 21, 9–21 (2011). 8. M. V. Balzan, A.-C. Moonen, Field margin vegetation enhances biological control and crop damage suppression from multiple pests in organic tomato fields. Entomol. Exp. Appl. 150, 45–65 (2014). 9. B. R. Blaauw, R. Isaacs, Wildflower plantings enhance the abundance of natural enemies and their services in adjacent blueberry fields. Biol. Control 91, 94–103 (2015). 10. M. Tschumi, et al., Tailored flower strips promote natural enemy biodiversity and pest control in potato crops. J. Appl. Ecol. 53, 1169–1176 (2016). 11. T. Tscharntke, A. M. Klein, A. Kruess, I. Steffan-Dewenter, C. Thies, Landscape perspectives on agricultural intensification and biodiversity – ecosystem service management. Ecol. Lett. 8, 857–874 (2005). 12. M. Jonsson, et al., Experimental evidence that the effectiveness of conservation biological control depends on landscape complexity. J. Appl. Ecol. 52, 1274–1282 (2015). 13. C. M. Kennedy, et al., A global quantitative synthesis of local and landscape effects on wild bee pollinators in agroecosystems. Ecol. Lett. 16, 584–599 (2013). 14. L. A. Garibaldi, et al., Wild pollinators enhance fruit set of crops regardless of honey bee abundance. Science 339, 1608–1611 (2013). 15. T. H. Ricketts, et al., Landscape effects on crop pollination services: are there general patterns? Ecol. Lett. 11, 499–515 (2008). 16. L. A. Garibaldi, et al., Stability of pollination services decreases with isolation from natural areas despite honey bee visits. Ecol. Lett. 14, 1062–1072 (2011). 17. E. M. Venturini, F. A. Drummond, A. K. Hoshide, A. C. Dibble, L. B. Stack, Pollination reservoirs for wild bee habitat enhancement in cropping systems: a review. Agroecol. Sustain. Food Syst. 41, 101–142 (2017). 18. H. Grab, K. Poveda, B. Danforth, G. Loeb, Landscape context shifts the balance of costs and benefits from wildflower borders on multiple ecosystem services. Proc R Soc B 285, 20181102 (2018). 19. L. Herbertsson, et al., The impact of sown flower strips on plant reproductive success in Southern Sweden varies with landscape context. Agric. Ecosyst. Environ. 259, 127–134 (2018).

20. M. P. D. Garratt, D. Senapathi, D. J. Coston, S. R. Mortimer, S. G. Potts, The benefits of hedgerows for pollinators and natural enemies depends on hedge quality and landscape context. Agric. Ecosyst. Environ. 247, 363–370 (2017). 21. E. A. Martin, et al., The interplay of landscape composition and configuration: new pathways to manage functional biodiversity and agroecosystem services across Europe. Ecol. Lett., ele.13265 (2019). 22. R. Chaplin-Kramer, et al., Measuring what matters: actionable information for conservation biocontrol in multifunctional landscapes. Front. Sustain. Food Syst. 3, 60 (2019). 23. J. M. Holland, et al., Semi-natural habitats support biological control, pollination and soil conservation in Europe. A review. Agron. Sustain. Dev. 37, 31 (2017). 24. A. N. Kravchenko, G. P. Robertson, K. D. Thelen, R. R. Harwood, Management, topographical, and weather effects on spatial variability of crop grain yields. Agron. J. 97, 514 (2005). 25. D. Gibson, “Experimental treatments” in Methods in Comparative Plant Population Ecology, Second Edition - Oxford Scholarship, 2nd Ed., (Oxford University Press, 2014) (September 5, 2019). 26. C. A. Wyenandt, et al., Mid-Atlantic Commercial Vegetable Production Recommendations (Virginia Cooperative Extension, 2019) (February 5, 2020). 27. R. J. Cordero, T. P. Kuhar, J. Speese, III, R. R. Youngman, E. E. Lewis, Field efficacy of insecticides for control of lepidopteran pests on collards in virginia. Forage Grazinglands (2006) https:/doi.org/10.1094/PHP-2006-0105-01-RS (March 9, 2017). 28. USDA, “United States Standards for Grades of Collard Greens or Broccoli Greens” (2016) (April 23, 2020). 29. USDA, “United States Standards for Grades of Fresh Tomatoes” (1991) (April 23, 2020). 30. USDA, “United States Standards for Grades of Strawberries” (2006) (April 23, 2020). 31. USDA, “United States Standards for Grades of Fall and Winter Type Squash and Pumpkin” (1983) (April 23, 2020). 32. R Core Team, A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria, 2018). 33. K. Barton, MuMIn: Multi-Model Inference (2018). 34. R. M. O’Brien, A caution regarding rules of thumb for variance inflation factors. Qual. Quant. 41, 673–690 (2007). 35. A. Kuznetsova, P. B. Brockhoff, R. H. B. Christensen, lmerTest Package: Tests in Linear Mixed Effects Models. J. Stat. Softw. 82 (2017). 36. H. Skaug, D. Fournier, A. Nielsen, A. Magnusson, B. Bolker, Generalized Linear Mixed Models using “AD Model Builder” (2016). 37. M. Smithson, J. Verkuilen, A better lemon squeezer? Maximum-likelihood regression with beta-distributed dependent variables. Psychol. Methods 11, 54–71 (2006). 38. M. Mendoza-García, J. M. Blanco-Moreno, L. Chamorro, L. José-María, F. X. Sans, Patterns of flower visitor abundance and fruit set in a highly intensified cereal cropping system in a Mediterranean landscape. Agric. Ecosyst. Environ. 254, 255–263 (2018). 39. T. Bukovinszky, et al., Exploring the relationships between landscape complexity, wild bee species richness and reproduction, and pollination services along a complexity gradient in the Netherlands. Biol. Conserv. 214, 312–319 (2017).

40. M. Dainese, S. Montecchiari, T. Sitzia, M. Sigura, L. Marini, High cover of hedgerows in the landscape supports multiple ecosystem services in Mediterranean cereal fields. J. Appl. Ecol. 54, 380–388 (2017). 41. T. S. Woodcock, L. J. Pekkola, C. Dawson, F. L. Gadallah, P. G. Kevan, Development of a pollination service measurement (PSM) method using potted plant phytometry. Environ. Monit. Assess. 186, 5041–5057 (2014). 42. S. J. Risch, D. Andow, M. A. Altieri, Agroecosystem diversity and pest control: data, tentative conclusions, and new research directions. Environ. Entomol. 12, 625–629 (1983). 43. O. Lundin, H. G. Smith, M. Rundlof, R. Bommarco, When ecosystem services interact: crop pollination benefits depend on the level of pest control. Proc. R. Soc. B-Biol. Sci. 280, 20122243 (2013). 44. R. Chaplin-Kramer, P. de Valpine, N. J. Mills, C. Kremen, Detecting pest control services across spatial and temporal scales. Agric. Ecosyst. Environ. 181, 206–212 (2013). 45. B. R. Blaauw, R. Isaacs, Larger wildflower plantings increase natural enemy density, diversity, and biological control of sentinel prey, without increasing herbivore density. Ecol. Entomol. 37, 386–394 (2012). 46. B. R. Blaauw, R. Isaacs, Larger patches of diverse floral resources increase insect pollinator density, diversity, and their pollination of native wildflowers. Basic Appl. Ecol. 15, 701–711 (2014). 47. T. Tscharntke, et al., When natural habitat fails to enhance biological pest control – Five hypotheses. Biol. Conserv. 204, 449–458 (2016). 48. B. R. Blaauw, R. Isaacs, Flower plantings increase wild bee abundance and the pollination services provided to a pollination-dependent crop. J. Appl. Ecol. 51, 890–898 (2014).

Figure 1.1. The proportion of egg predation in response to the percent of semi-natural habitat (SNH) for three pests of vegetable crops. Tricopulsia ni on collards (A),

Helicoverpa zea (B) and Halyomorpha halys (C) for tomatoes. Lines represent simple linear regression, and the shaded areas are the standard error. Two different colored dots indicate a significant effect of year. For A-C, N = 42, 38, and 44 for each panel respectively. NS = no significant effects detected.

Figure 1.2. The average weekly abundance of four vegetable pests, and their response to the percent of semi-natural habitat (SNH) and wildflower plantings. P. rapae (A) and P. xylostella (B) were recorded on collards, and S. ornithogalli, (C) and Manduca spp. (D) on tomatoes. The p-value in panel (D) represents the significant effect of wildflower plots. Lines represent simple linear regression, and the shaded areas are the standard error. Two different colored dots indicate a significant effect of year. For panels A-D, N = 64, 88, 146, 110 for each panel respectively. NS = no significant effects detected.

NS NS

Figure 1.3. Various types of pest damage recorded on all four crops grown, and their response to wildflower plantings and percent of semi-natural habitat (SNH). The proportion of severely damaged collards (A). For tomatoes, the proportion of fruits damaged by chewing pests (B) and pests with piercing/sucking mouthparts (C). On strawberries, the damage caused by Lygus spp. (D), sap beetles (E), and Drosophila spp. (F). The probability of squash damage caused by squash vine borer (G) and the squash bug/Diabrotica spp. beetles (H). Data is shown on data scale. Lines represent simple linear regression, except panes (G) and (H). The shaded areas are the standard error.

For panels A-H, N = 200, 119, 119, 237, 237, 237, 122, 106 for each panel respectively. NS = no significant effects detected.

Figure 1.4. Response of metrics of pollination services. proportion of achenes pollinated on strawberries (A), proportion of pollen deficient strawberries (B), and the number of seed produced in squash (C), to the percent of semi-natural habitat (SNH) and wildflower plots. Data is shown on data scale. Lines represent simple linear regression, and the shaded areas are the standard error. For A-C, N = 213, 237,105 for each panel respectively. NS = no significant effects detected.

Figure 1.5. The response of the proportion of the number of high quality harvested items of strawberry (A), squash (B), collards (C), and tomato (D) to the percent of semi- natural habitat (SNH) in the landscape at the 250-m scale. Data is shown on data scale. Lines represent simple linear regression, and the shaded areas are the standard error. For panels A-D, N = 237, 106, 200, and 113 for each panel respectively.

Supplemental Appendix 1.1. Descriptions of crops, size, and overall management style at each site. List of plant species used in each mix for the wildflower treatment sites. Figure showing the location of the field sites used. Table S1.1.1. Description of crops grown, farm size, and overall management style for each site. WF plot Management Farm size SiteID State County Treatment Wildflower mix size (m2) style (acres) Cull VA Virginia Beach Control NA NA Conventional 250 Crops: tomatoes, pumpkins, corn, squash, beans, peas, eggplant Brook VA Virginia Beach Control NA NA Conventional 12 Crops: pumpkins, strawberries Flip VA Virginia Beach Wildflower Poorly-draining 1463.22 Conventional 20 Crops: strawberries, green peas, green beans, blackberries, potatoes, corn, tomatoes, brassicas Flan VA Virginia Beach Wildflower Well-draining 1024.72 Conventional 17 Crops: strawberries, sweet corn, soybeans, pumpkins, brassicas HRAREC VA Virginia Beach Control NA NA Conventional 70 Crops: strawberries, ornamental plants Organic, not Matt VA Northampton Control NA NA certified 22 Crops: corn, tomatoes, brassicas, kale, potatoes, amaranth Organic, not CopC VA Northampton Wildflower Well-draining 929.03 certified 2 Crops: wheat, tomato, celery, squash, cut flowers, snap beans, onion, potatoes, pears, chard, fennel, basil, brassicas NRCS (similar to Organic, not Stur VA Northampton Wildflower well-draining) 12140.61 certified 130 Crops: Wheat, soybeans, squash Sea VA Northampton Control NA NA Conventional 55 Crops: Soybeans, cotton Organic, not Qcove VA Northampton Control NA NA certified 80 Crops: sweet potatoes, squash ESAREC VA Accomack Wildflower Well-draining 1246.39 Conventional 226 Crops: wheat, soy, pumpkins, cotton, potatoes, sweet potato, corn, rape seed, barley, tomatoes, broccoli, cabbage, sweet corn, basil, snap beans, squash, lettuce, swiss chard, eggplant, peppers Pik VA Accomack Wildflower Well-draining 557.42 Conventional 45 Crops: squash, brassicas, beets, fennel, turnips, asparagus Organic, not Proot VA Accomack Wildflower Well-draining 1393.55 certified 49 Crops: heirloom lettuce, kale, chard, tomato, squash, pumpkins, mint Organic, not LaC VA Accomack Control NA NA certified <1 Crops: tomatoes, squash, beans

WF plot Management Farm size SiteID State County Treatment Wildflower mix size (m2) style (acres) vanD VA Accomack Control NA NA Conventional 100 Crops: horseradish, carrots, corn, soy Conventional UMES MD Somerset Wildflower Poorly-draining 2023.43 and Certified 46 Organic Crops: tomatoes, corn, peppers, ornamental plants, amaranth Organic, not Prov MD Wicomico Control NA NA certified 1.1 Crops: tomatoes, beans, squash, peppers, lettuce Organic, not Calli MD Wicomico Wildflower Well-draining 929.03 certified 1.5 Crops: squash, flowers, pumpkin, blackberries Conventional LESREC MD Wicomico Control NA NA and Organic, 214 not certified Crops: cantaloupes, watermelons, pumpkin, sunflower, lima beans, soybeans, corn, grapes Certified Pgar MD Wicomico Control NA NA organic 15 Crops: currants, blackberries, tomatoes, onions, kale, asparagus, string beans, mustard greens, squash, turnips Wri MD Wicomico Wildflower Well-draining 1486.45 Conventional 40 Crops: watermelon, sunflowers, squash, sweet corn, beans Phill MD Wicomico Control NA NA Conventional 190 Crops: Corn, soybeans, millet, wheat

Table S1.1.2. List of grass and forb mix seeded at fields with well-drained soils (N = 7) Common Name Scientific Name

Little Bluestem (G) Schizachyrium scoparium Splitbeard Bluestem (G) Andropogon ternarius Narrowleaf Mountain Pycnanthemum Mint (P) tenuifolium Plains Coreopsis (A) Coreopsis tinctoria Partridge Pea (A) Chamaecrista fasciculata Black-eyed Susan (B) Rudbeckia hirta Bergamot, Spotted (P) Monarda fistulosa Lanceleaf Coreopsis (P) Coreopsis lanceolata Maximilian Sunflower Helianthus maximilianii (P) Indian Blanket (A) Gaillardia pulchella Purple Coneflower (P) Echinacea purpurea G = grass, A = annual, B = biennial, P = perennial

Table S1.1.3. List of grass and forb mix seeded at fields with poorly-drained soils (N = 2) Common Name Scientific Name

Beaked Panicum (G) Panicum anceps Redtop Panicum (G) Panicum rigidulum Aster, Purple-stemmed Symphyotrichum (P) puniceum var. puniceum Sneezeweed, Common Helenium autumnale (P) Coreopsis, Plains (A) Coreopsis tinctoria Goldenrod, Wrinkleleaf Solidago rugosa (P) Joe Pye Weed, Spotted Eupatoriadelphus (P) fistulosus Partridge Pea (A) Chamaecrista fasciculata Rattlesnake Master (P) Eryngium yuccifolium Rosemallow (P) Hibiscus moscheutos Narrowleaf Sunflower Helianthus angustifolius (P) G = grass, A = annual, B = biennial P = perennial

Table S1.1.4. List of forb mix for NRCS wildflower plot planted in 2015 (N = 1) Common Name Scientific Name

Showy evening primrose Oenothera speciosa (P)* Indian Blanket (A) Gaillardia pulchella Maximilian Sunflower (P) Helianthus maximiliani Black-eyed Susan (B) Rudbeckia hirta Partridge Pea (A) Chamaecrista fasciculate Plains Coreopsis (A) Coreopsis tinctoria Lanceleaf Coreopsis (P) Coreopsis lanceolate Spotted Beebalm (P)* Monarda punctate Tickseed Sunflower (A)* Bidens aristosa G = grass, A = annual, B = biennial, P = perennial * not in well-drained mix

Figure S1.1.1. Map of field locations.

Supplemental Appendix 2. AICc values for the model reduction process. Outputs for all selected and reduced models using reverse stepwise reduction for every response variable measured. Models within 2 AICc points of the lowest scoring model are also presented. All models started by testing the effect of wildflower plantings, % SNH, and their interaction. Figures for total harvested mass and grade 1 yield by weight.

Table S1.2.1. Model selection table for each response variable Interaction model Best single Response variable Landscape scale AICc (AICc) term Predation of T. ni eggs (Collards) 1000 364.3 SNH 349.0* 500 363.7 SNH 348.7* 250 362.7 SNH 348.3* Predation of H. halys eggs (Tomatoes) 1000 5.3 SNH -11.9* 500 4.4 SNH -13.6* 250 5.2 SNH -13.4* Predation of H. zea eggs (Tomatoes) 1000 406.6 SNH 390.5* 500 406.5 SNH 390.2* 250 407.9 SNH 390.9* P. xylostella abundance (Collards) 1000 1535.9 SNH 1535.7* 500 1535.1 SNH 1534.1* 250 1536.7 SNH 1533.5* P. rapae abundance (Collards) 1000 1502.1 SNH 1500.3* 500 1504.0 SNH 1500.6* 250 1503.1 SNH 1499.8* S. ornithogalli abundance (Tomatoes) 1000 1211.8 SNH 1208.6* 500 1213.1 SNH 1209.6* 250 1212.6 SNH 1209.1* Manduca spp. abundance (Tomatoes) 1000 690.4 500 689.0 250 686.8 Wildflower 686.6* Chewing damaged to collards 1000 44.8 SNH 33.7* 500 46.9 250 45.4 SNH 29.9*

Response Scale Interaction AICc Best term AICc Chewing damage to tomato fruit 1000 49.6 SNH 36.2* 500 50.1 SNH 36.6* 250 49.6 SNH 36.2* Piercing-sucking damage to tomato fruit 1000 30.1 500 24.3 SNH 5.9* 250 28.8 Lygus spp. damaged strawberries 1000 -133.0 Wildflower -158.2* 500 -131.8 Wildflower -158.2* 250 -132.6 Wildflower -158.2* Drosophilia spp. damaged strawberries 1000 -261.8 Wildflower -286.2* 500 -261.6 Wildflower -286.2* 250 -260.3 Wildflower -286.2* Picnic beetle damaged strawberries 1000 -373.4 Wildflower -399.3* 500 -370.2 250 -371.6 Wildflower -399.3* Combined damage of squash bug and Diabrotica spp. beetles to squash 1000 101.9 SNH 98.1* 500 102.7 SNH 98.3* 250 102.3 SNH 97.9* Squash Vine Borer Damage 1000 173.8* 500 175.9 250 177.7 Proportion of Pollinated Achenes 1000 -103.7 Wildflower -130.5* 500 -102.9 Wildflower -130.5* 250 -104.3 Wildflower -130.5* Proportion of Pollen Deficient Strawberries 1000 -91.6 SNH -110.3* 500 -90.7 SNH -109.4* 250 -92.5 SNH -110.5* Number of Seeds Produced (Squash) 1000 528.1 Wildflower 513.3* 500 528.5 Wildflower 513.3* 250 529.9 Wildflower 513.3*

Response Scale Interaction AICc Best term AICc Total Mass Collards 1000 2510.4† SNH 2519.8* 500 2509.0† SNH 2520.0* 250 2508.8† SNH 2518.9* Total Mass Tomatoes 1000 1841.1† Wildflower 1850.2* 500 1839.2† Wildflower 1850.2* 250 1839.6† Wildflower 1850.2* Total Mass Strawberries 1000 2818.8† Wildflower 2890.3* 500 2818.4† Wildflower 2890.3* 250 2820.0† Wildflower 2890.3* Total Harvested Mass Squash 1000 1673.9† SNH 1688.7* 500 1674.9† SNH 1689.2* 250 1674.1† SNH 1687.6* Weight Grade 1 Collards 1000 1187.1 Wildflower 1179.0* 500 1189.8 250 1190.4 Weight Grade 1 Tomatoes 1000 816.5 500 813.8 Wildflower 812.4* 250 818.5 Weight Grade 1 Strawberries 1000 2414.6 Wildflower 2410.7* 500 2412.6 Wildflower 2410.7* 250 2415.2 Weight Grade 1 Squash 1000 948.6 Wildflower 945.0* 500 949.1 Wildflower 945.0* 250 953.2 Proportion of Grade 1 Harvested Collards 1000 44.8 SNH 33.7* 500 46.9 250 45.4 SNH 29.9* Proportion of Grade 1 Harvested Tomatoes 1000 27.0 500 22.7 SNH 9.5* 250 23.0 SNH 5.9*

Response Scale Interaction AICc Best term AICc Proportion of Grade 1 Harvested Strawberries 1000 -299.1 SNH -315.4* 500 -295.4 250 -299.6 SNH -316.6* Proportion of Grade 1 Harvested Squash 1000 -272.7 500 -273.7 SNH -275.7* 250 -275.1 SNH -276.4* * model analyzed † Predictors with variance inflation factors above 6 removed

Table S1.2.2. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the egg predation of T. ni on collards. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.678 0.102 6.7 21.6 <0.0001* Year 0.081 0.047 1.7 270.7 0.086 SNH 1000 -0.001 0.002 -0.5 18.9 0.642

Intercept 0.687 0.082 8.4 21.8 <0.0001* Year 0.081 0.047 1.7 270.7 0.083 SNH 500 -0.002 0.002 -0.7 18.2 0.462

Intercept 0.682 0.058 11.8 27.5 <0.0001* Year 0.083 0.047 1.8 270.8 0.077 SNH 250 -0.002 0.002 -1.2 18.7 0.232 *Significant at p<0.05

Table S1.2.3. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on H. halys egg predation. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.179 0.056 3.2 15.7 0.005* Year 0.093 0.025 3.7 248.9 <0.0001* SNH 1000 -0.0001 0.001 -0.3 14.1 0.753

Intercept 0.214 0.043 4.9 19.2 <0.0001* Year 0.091 0.025 3.6 249.2 <0.0001* SNH 500 -0.002 0.001 -1.4 15.6 0.180

Intercept 0.195 0.032 6.1 20.3 <0.0001* Year 0.093 0.025 3.7 250.0 <0.0001* SNH 250 -0.001 0.001 -1.5 14.2 0.160 *Significant at p<0.05

Table S1.2.4. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on H. zea egg predation. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.375 0.092 4.1 47.2 <0.0001* Year 0.153 0.046 3.3 293.5 0.001* SNH 1000 0.001 0.002 0.6 42.5 0.565

Intercept 0.372 0.072 5.1 51.0 <0.0001* Year 0.154 0.046 3.4 293.6 0.001* SNH 500 0.002 0.002 0.9 42.8 0.399

Intercept 0.399 0.052 7.7 60.9 <0.0001* Year 0.152 0.046 3.3 293.6 0.001* SNH 250 0.001 0.002 0.7 41.7 0.474 *Significant at p<0.05

Table S1.2.5. Results of generalized linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the weekly abundance of P. xylostella on collards. Model presented were selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept 0.585 0.546 0.55 0.28 Year -1.149 0.176 -6.5 <0.0001* SNH 1000 -0.001 0.013 -0.1 0.93

Intercept 0.987 0.408 2.4 0.016* Year -1.158 0.176 -6.6 <0.0001* SNH 500 -0.159 0.0124 -1.3 0.201

Intercept 0.831 0.289 2.9 0.004* Year -1.148 0.176 -6.5 <0.0001* SNH 250 -0.014 0.009 -1.5 0.138 *Significant at p<0.05

Table S1.2.6. Results of generalized linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the weekly abundance of P. rapae on collards. Models presented were selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept -2.760 0.649 -4.2 <0.0001* Year 1.625 0.148 10.9 <0.0001* SNH 1000 0.010 0.016 0.6 0.54

Intercept -2.280 0.496 -4.6 <0.0001* Year 1.624 0.148 10.9 <0.0001* SNH 500 -0.004 0.014 -0.3 0.77

Intercept -2.174 0.352 -6.2 <0.0001* Year 1.626 0.148 10.9 <0.0001* SNH 250 -0.011 0.012 -0.9 0.34 *Significant at p<0.05

Table S1.2.7. Results of generalized linear mixed effects models on effect of the amount of semi-natural habitat in the landscape on the weekly abundance of S. ornithogalli on tomatoes. Models presented selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept 0.174 0.656 0.3 0.79 Year -0.969 0.172 -5.6 <0.0001* SNH 1000 -0.272 0.017 -1.6 0.10

Intercept -0.255 0.518 -0.5 0.62 Year -0.960 0.172 -5.6 <0.0001* SNH 500 -0.203 0.016 -1.2 0.21

Intercept -0.455 0.355 -1.3 0.20 Year -0.949 0.172 -5.5 <0.0001* SNH 250 -0.017 0.012 -1.5 0.14 *Significant at p<0.05

Table S1.2.8. Results of generalized linear mixed effects model on the effect of wildflower plots on the weekly abundance of Manduca spp. larvae on tomatoes. Model presented was selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept -1.587 0.384 -4.1 <0.0001* Year -0.249 0.259 -0.9 0.336 Wildflower plot -1.011 0.512 -1.9 0.048* *Significant at p<0.05

Table S1.2.9. Results of linear mixed effects models on the effect of the amount of semi- natural habitat in the landscape on the proportion of insect damaged collard leaves. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.568 0.083 6.8 22.6 <0.0001* Year 0.319 0.033 9.7 185.6 <0.0001* SNH 1000 -0.004 0.003 -1.4 20.1 0.17

Intercept 0.555 0.055 10.1 25.6 <0.0001* Year 0.320 0.033 9.8 186.5 <0.0001* SNH 250 -0.004 0.002 -2.3 20.2 0.031* *Significant at p<0.05

Table S1.2.10. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the proportion of tomatoes damaged by chewing pests. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.630 0.126 5.0 23.7 <0.0001* Year -0.129 0.083 -1.5 20.6 0.136 SNH 1000 -0.003 0.003 -0.9 19.6 0.334

Intercept 0.583 0.101 5.8 26.1 <0.0001* Year -0.128 0.083 -1.5 20.6 0.138 SNH 500 -0.002 0.003 -0.7 19.6 0.451

Intercept 0.576 0.073 7.9 32.5 <0.0001* Year -0.124 0.083 -1.5 20.7 0.151 SNH 250 -0.003 0.002 -1.3 19.0 0.215 *Significant at p<0.05

Table S1.2.11. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the proportion of tomatoes damaged by pests with piercing-sucking mouthparts. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.374 0.066 5.7 22.7 <0.0001* Year 0.160 0.058 2.7 20.2 0.009* SNH 500 -0.007 0.002 -3.6 19.3 0.001* *Significant at p<0.05

Table S1.2.12. Results of linear mixed effects model on the effect of wildflower plots on the proportion of Lygus spp. damaged strawberries harvested. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.269 0.025 10.6 30.7 <0.0001* Year -0.139 0.021 -6.6 229.2 <0.0001* Wildflower plot -0.068 0.033 -2.1 19.6 0.054 *Significant at p<0.05

Table S1.2.13. Results of linear mixed effects model on the effect of pollinator habitats on the proportion of Drosophila spp. damaged strawberries harvested. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept -0.040 0.021 -1.8 27.9 0.076 Year 0.156 0.016 9.8 227.5 <0.0001* Wildflower plot 0.076 0.029 2.6 19.5 0.017* *Significant at p<0.05

Table S1.2.14. Results of linear mixed effects model on the effect of pollinator habitats on the proportion of sap beetle damaged strawberries harvested. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.108 0.015 7.3 33.0 <0.0001* Year 0.034 0.013 2.7 229.8 0.007* Wildflower plot 0.024 0.019 1.3 20.5 0.216 *Significant at p<0.05

Table S1.2.15. Results of logistic regression models on the effect of the amount of semi-natural habitat in the landscape on the odds of by squash bug and Diabrotica spp. beetles damaging fruit. Models presented were selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept 0.50 0.13 3.7 <0.0001* Year -0.38 0.08 -4.3 <0.0001* SNH 1000 -0.001 -0.001 -0.08 0.45

Intercept 0.46 0.11 4.2 <0.0001* Year -0.38 0.09 -4.2 <0.0001* SNH 500 -0.002 0.002 -0.6 0.54

Intercept 0.46 0.1 4.7 <0.0001* Year -0.38 0.09 -4.5 <0.0001* SNH 250 -0.002 0.002 -0.9 0.39 *Significant at p<0.05

Table S1.2.16. Results of logistic regression model on the effect of the amount of semi-natural habitat in the landscape and wildflower plots on the potential of squash vine borer damaged fruit. Model presented was selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept 0.13 0.20 0.7 0.5 Year -0.20 0.09 -2.3 0.02* SNH 1000 0.014 0.009 2.4 0.02* Wildflower plot 0.16 0.38 0.4 0.66 Interaction -0.013 0.01 -1.5 0.15 *Significant at p<0.05

Table S1.2.17. Results of linear mixed effects model on the effect of wildflower plots on the ratio of pollinated achenes of strawberries. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.710 0.021 33.9 33.6 <0.0001* Year -0.109 0.023 -4.7 209.9 <0.0001* Wildflower plot 0.019 0.025 0.7 15.4 0.462 *Significant at p<0.05

Table S1.2.18. Results of linear mixed effects models on effect of the amount of semi- natural habitat in the landscape on proportion of pollen deficient strawberries. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.400 0.038 10.6 22.6 <0.0001* Year -0.266 0.023 -11.4 231.6 <0.0001* SNH 1000 -0.001 0.001 -1.2 18.1 0.249

Intercept 0.375 0.023 16.1 36.4 <0.0001* Year -0.266 0.023 -11.5 231.6 <0.0001* SNH 500 -0.001 0.001 -0.9 18.9 0.336

Intercept 0.392 0.031 12.7 27.1 <0.0001* Year -0.267 0.023 -11.5 231.6 <0.0001* SNH 250 -0.001 0.001 -1.3 18.4 0.216 *Significant at p<0.05

Table S1.2.19. Results of linear mixed effects model on the effect of wildflower plots on the number of seeds produced by squash Estimate std.error t-value df p-value Intercept 15.20 0.51 29.7 33.7 <0.0001* Year -6.18 0.55 -11.3 102.6 <0.0001* Wildflower plot -0.37 0.53 -0.7 17.9 0.489 *Significant at p<0.05

Table S1.2.20. Results of linear mixed effects models on the effects of the amount of semi-natural habitat in the landscape on the total harvested mass of collards. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 169.58 45.58 3.7 20.3 0.001* Year 82.53 18.40 4.5 188.3 <0.0001* SNH 1000 0.56 1.12 0.5 18.1 0.62

Intercept 202.81 36.17 5.6 22.2 <0.0001* Year 82.13 18.42 4.4 188.2 <0.0001* SNH 500 -0.43 1.07 -0.4 18.4 0.69

Intercept 213.48 24.84 8.6 26.3 <0.0001* Year 81.65 18.38 4.4 188.6 <0.0001* SNH 250 -1.09 0.81 -1.3 18.3 0.193 *Significant at p<0.05

Table S1.2.21. Results of linear mixed effects model on the effect wildflower plots on the total harvested mass of tomatoes. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 1862.93 181.42 10.3 20.7 <0.0001* Year -703.32 107.74 -6.5 96.9 <0.0001* Wildflower plot -346.68 251.89 -1.4 15.6 0.19 *Significant at p<0.05

Table S1.2.22. Results of linear mixed effects model on the effect of pollinator habitat on the total harvested mass of strawberries. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 209.99 18.31 11.5 26.6 <0.0001* Year 88.51 13.74 6.4 228.2 <0.0001* Wildflower plot 29.36 24.65 1.2 18.5 0.25 *Significant at p<0.05

Table S1.2.23. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the total harvested mass of squash. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 464.15 219.56 2.1 28.1 0.03* Year 414.50 154.97 2.7 102.9 0.009* SNH 1000 3.71 4.94 0.7 19.5 0.45

Intercept 637.56 178.49 3.3 30.3 0.001* Year 415.95 155.64 2.7 102.9 0.008* SNH 500 -1.37 4.76 -0.3 17.0 0.77

Intercept 708.15 142.79 4.9 37.9 <0.0001* Year 421.07 153.82 2.7 102.9. 0.007* SNH 250 -5.54 3.71 -1.5 14.9 0.14 *Significant at p<0.05

Table S1.2.24. Results of linear mixed effects model on the effect of wildflower plots on the harvested marketable mass of collards. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 10.22 1.03 9.9 22.5 <0.0001* Year -2.82 0.61 -4.6 186.1 <0.0001* Wildflower plot 2.09 1.35 1.5 19.7 0.14 *Significant at p<0.05

Table S1.2.25. Results of linear mixed effects model on the effect wildflower plots on the harvested grade 1 yield of tomatoes. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 28.70 2.27 12.6 25.2 <0.0001* Year -4.77 1.54 -3.1 94.6 0.003* Wildflower plot -2.35 3.10 -0.8 17.9 0.45 *Significant at p<0.05

Table S1.2.26. Results of linear mixed effects model on the effect of pollinator habitat on the mass (g) of grade 1 strawberries harvested. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 45.44 5.69 7.9 32.7 <0.0001* Year 21.42 5.09 4.2 230.1 <0.0001* Wildflower plot 9.51 7.33 1.3 19.7 0.19 *Significant at p<0.05

Table S1.2.27. Results of linear mixed effects model on the effect of wildflower plots on the harvested grade 1 mass of squash. Model presented was selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 9.52 1.78 5.3 22.8 0.572 Year -7.34 2.32 -3.1 114.3 0.002* Wildflower plot 6.64 2.33 2.8 13.9 0.01* *Significant at p<0.05

Table S1.2.28. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the proportion of marketable collard leaves harvested. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.471 0.110 4.3 21.2 <0.0001* Year -0.320 0.033 -9.8 185.3 <0.0001* SNH 1000 0.002 0.003 0.6 19.8 0.548

Intercept 0.445 0.055 8.1 25.6 <0.0001* Year -0.320 0.033 -9.8 186.5 <0.0001* SNH 250 0.004 0.002 2.3 20.2 0.031* *Significant at p<0.05

Table S1.2.29. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the proportion of grade 1 tomatoes harvested. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.339 0.065 5.2 23.0 <0.0001* Year 0.109 0.041 2.6 97.0 0.009* SNH 500 0.006 0.002 2.9 17.4 0.01*

Intercept 0.392 0.044 8.9 34.0 <0.0001* Year 0.100 0.041 2.4 96.9 0.016* SNH 250 0.005 0.001 4.0 16.2 0.001* *Significant at p<0.05

Table S1.2.30. Results of linear mixed effects models on the effect of the amount of semi-natural habitat in the landscape on the proportion of grade 1strawberries harvested. Models presented were selected after stepwise reduction and model selection. Estimate std.error t-value df p-value Intercept 0.233 0.015 15.4 37.8 <0.0001* Year -0.064 0.015 -4.3 231.6 <0.0001* SNH 1000 0.001 0.000 2.1 19.8 0.047*

Intercept 0.203 0.024 8.3 23.4 <0.0001* Year -0.064 0.015 -4.3 231.7 <0.0001* SNH 250 0.001 0.001 2.3 18.7 0.03* *Significant at p<0.05

Table S1.2.31. Results of generalized linear mixed effects models using a beta distribution on the effect of the amount of semi- natural habitat in the landscape on the proportion of grade 1 squash harvested. Models presented were selected after stepwise reduction and model selection. Estimate std.error z-value p-value Intercept -1.49 0.36 -4.1 <0.0001* Year 0.29 0.25 1.1 0.25 SNH 1000 0.013 0.01 1.7 0.092

Intercept -1.39 0.31 -4.5 <0.0001* Year 0.27 0.25 1.0 0.28

SNH 500 0.015 0.008 1.8 0.07

Intercept -1.27 0.25 -5.0 <0.0001*

Year 0.29 0.25 1.2 0.25 SNH 250 0.013 0.006 2.0 0.043* *Significant at p<0.05

NS NS

Figure S1.2.1. The total harvested mass for strawberry (A), squash (B), collards (C), and tomato (D), and their response to SNH or pollinator refuges. Data is shown on data scale. Lines represent simple linear regression, and the shaded areas are the standard error. For A-D, N = 238, 106, 200, and 119 for each panel respectively. NS = no significant effects detected.

NS NS

Figure S1.2.2. The response of the weight of high quality yield of strawberries (A), squash (B), collards (C), and tomatoes (D) to SNH and wildflower plantings. Data is shown on data scale. Lines represent simple linear regression, and the shaded areas are the standard error. For panels A-D, N = 237, 106, 200, and 113 for each panel respectively. NS = no significant effects detected.

Chapter 2: Landscape context influences the bee conservation value of wildflower plantings

Abstract

Pollination provided by bees is a critical ecosystem service for maintaining and enhancing agricultural production. However, bee populations are at risk from multiple stressors such as habitat loss, pesticides exposure, and disease. Planting wildflower plot on farms to provide habitat and resources for bees is a commonly used mitigation strategy. In many instances, government programs are available for private-landowners to subsidize the installation of these plots. Semi-natural habitat (SNH) elements in the landscape are also important for conserving bee populations and the amount of SNH in the landscape may alter the effectiveness of wildflower plots in conserving bee communities. In this study, we tested the effectiveness of wildflower plots and interactions with SNH in the landscape for promoting bee abundance and richness. Bee surveys were conducted over two years at 22 sites in eastern Virginia and

Maryland. Wildflower plots, averaging 0.22 ha in size, were installed and maintained by cooperators at 10 of the sites. In total, 5,384 bees were identified from 85 species. Wildflower plots did not alter or enhance bee communities. Bee abundance and richness had nonlinear responses to increasing SNH in the landscape surrounding the sites, albeit at different scales. The positive effects for richness and abundance peaked when SNH was approximately 40% of the landscape. Similar to predictions of the intermediate landscape hypothesis, fewer bees were sampled from wildflower habitats where the surrounding landscape had more SNH. Results indicate that small wildflower plots in the Eastern U.S. do not enhance bee conservation on the scale studied and that conserving SNH across the landscape is a more important conservation strategy.

Introduction

Pollination services provided by pollinators are a vital ecosystem service for the production of food and fiber, animal pollination improves yield in 39 of the 57 leading crops in the world (Klein et al. 2007). Diverse and abundant pollinator communities are needed to maintain and enhance crop yields (Woodcock et al. 2019). Native bees, in particular, can provide sufficient pollination for many pollinator-dependent crop species (Winfree et al. 2007, 2008), and can further enhance pollination in the presence of managed species like Apis mellifera L.

(Garibaldi et al. 2013). However, the continued delivery of pollination services is at risk as bee populations face threats from pesticides, habitat loss, pests, and pathogens (Potts et al. 2010,

Goulson et al. 2015).

Creating on-farm pollinator refuges by planting wildflowers is one commonly used strategy to mitigate the decline of pollinators and conserve pollination services (Williams et al.

2015, Venturini et al. 2017). In 2014, a presidential memorandum was issued in the United

States to enhance or restore seven million acres of land for pollinator habitat (US EPA 2014).

The U.S. Department of Agriculture’s (USDA) Natural Resources Conservation Service (NRCS) manages 13 programs that can be utilized to conserve and create pollinator habitat on privately owned agricultural lands (NRCS 2015). The Environmental Quality Incentives Program has been used to pay for pollinator conservation efforts on over 16,000,000 acres from 2009 - 2018

(Vaughn and Skinner 2015, USDA 2018).

Given the political, ecological and agronomic importance of conserving pollinators, it is imperative to understand the effectiveness of these conservation programs. One study on an

NRCS administered program to conserve Karner blue butterflies, Lycaeides melissa samuelis

(Lepidoptera: Lycaenidae) found that restored farmland sites did not increase Karner blue

populations but did increase overall butterfly richness compared to native prairie sites (Kleintjes

Neff et al. 2017). The Conservation Reserve Program (CRP) pays landowners to take land out of agricultural production and plant cover of conservation value (USDA 2019). This program is currently implemented on more than 20 million acres (USDA 2019). Studies in North Dakota and Texas found that native habitats had greater bee abundance than CRP land (Otto et al. 2017,

Begosh et al. 2020). In Colorado, CRP land enhanced with wildflowers did not increase bee abundance or richness relative to control CRP sites dominated by grasses (Arathi et al. 2019).

The effectiveness of CRP land may be limited by the wide variation in management practices, size, or age (Otto et al. 2017). Studies focused on the NRCS pollinator habitat are lacking on the effectiveness of these publicly available and subsidized programs to conserve and promote bee communities on private lands, particularly on farms.

The context of the surrounding landscape may moderate the effectiveness of farm-scale measures to promote bee communities and pollination services. Increasing semi-natural habitat

(SNH) in the landscape has been shown to increase bee species richness, abundance, and pollinator visitation rates to nearby crops (Ricketts et al. 2008, Kennedy et al. 2013). Agricultural land surrounding farms, on the other hand, has been shown to have negative effects on bee species richness and pollination services (Dainese et al. 2019, Grab et al. 2019). Farm-scale conservation strategies are thought to be most effective when located in simplified landscapes, areas where SNH is 1 - 20% of the surrounding landscape (Tscharntke et al. 2005). Outside of this range, landscapes either have too few species to conserve or already have species rich communities (Tscharntke et al. 2005). Studies that tested this interaction have found that increasing SNH limits the effectiveness of on-farm measures (Scheper et al. 2013, Garratt et al.

2017, Grab et al. 2018, Herbertsson et al. 2018, Krimmer et al. 2019). One study (Grab et al.

2018), which investigated this interaction using non-linear analyses, found wildflower plots had a positive effect on bee visitation relative to control plots where SNH was between ~ 25 - 50% of the surrounding landscape.

Other factors such as bloom density, flower diversity, plot size, or the time since the plot was planted may alter the effectiveness of wildflower plots. Blaauw and Isaacs (Blaauw and

Isaacs 2014a) found that the number of flower species blooming at the time of sampling increased wild bee abundance and richness. It has also been shown that floral area or bloom abundance of wildflowers in mixes can also have positive effects on wild bee abundance and richness (Tuell et al. 2008, Balzan et al. 2014, Williams et al. 2015). Wild bee richness and abundance had a positive response to increasing wildflower planting size; however, this effect plateaued as the 10-m2 plot has similar responses as the100-m2 plot (Blaauw and Isaacs 2014b).

There can also be a time lag from when the wildflower plot is planted to when effects are detected, with Blaauw and Isaacs (2014a) detecting increases in bee abundance only three years after planting wildflowers Accounting for these factors may help refine expectations for pollinator conservation from pollinator habitats.

The objectives of this study were to: 1) assess the effectiveness of NRCS designed wildflower mixes for conserving bees in eastern Virginia and Maryland; 2) understand how SNH interacts with wildflower plots and bee communities; and 3) investigate what factors are impacting the effectiveness of the wildflower plots to conserve native bees. We expected that wildflower plots would enhance bee communities relative to control fields by increasing species richness and abundance. Furthermore, we expected that wildflower plots will be affected by the amount of SNH in the landscape and the amount of floral resources provided by the plots.

Materials and Methods

Field sites

Bee communities were sampled at 20 sites in 2017 and 2018. Two control sites were changed from 2017 to 2018, for a total of 22 sites. These sites were located on the Delmarva

Peninsula in Virginia and Maryland and within the locality of Virginia Beach, Virginia. The majority of these sites were small-scale diversified farms that sold direct-to-market. Only two of these farms were certified ‘USDA Organic’, but others utilized a diversity of management strategies. Of the sites sampled, nine of them had wildflower plots planted in the spring of 2016.

Two different wildflower mixes were used to match soil drainage characteristics at each field.

Seven fields were planted with a mix adapted for well-draining soils, and two were planted with a mix for poorly-draining soil. The mixes included grasses and forbs that were perennials, annuals, and biennials (Supplemental appendix 1). Plant species chosen were based on the recommendation of the local NRCS private lands biologist to provide continuous blooms throughout the growing season (B. Glennon, personal communication). For details on establishment at the well-draining sites, see (Angelella and O’Rourke 2017). One site was added to the study that had a wildflower plot installed in the spring of 2015. A total of 9 wildflower species were used at this site, 7 of which were used in the well-draining mix (Supplemental appendix 2.1). Wildflower plots ranged in size from 561 - 8600 m2.

Bee sampling

At each site, bees were sampled using blue-vane traps and pan traps. Pan traps were

7.5cm by 7.5 cm plastic dishes (Rubbermaid, Atlanta, GA) that were painted with UV-reflective yellow, blue or white (Blick Art Materials, Galesburg, IL). All pan traps were placed on the ground. Blue-vane traps (Springstar Inc., Woodinville, WA) were attached to metal stakes with

the base of the trap 50 cm above the ground. At each field site, a total of 12 traps were placed, three of each pan trap color and three blue-vane traps. Traps were spaced 5 m apart along transects in an alternating sequence of trap type. At fields with wildflower plots, traps were placed along the edge of the wildflower plot. At control fields, traps were placed along field edges or roads. All traps were filled with water and a drop of non-scented dish soap. Traps were left in the field for 48 hrs. Three rounds of sampling took place each year. Sampling was done during the weeks of May 15th, June 19th, and August 14th in 2017. Sampling was conducted during the weeks of June 4th, July 23rd, and August 13th in 2018. Collected bees were processed and identified after the conclusion of the field season. Final species determinations were provided by Sam Droege of the United States Geological Survey. Vouchers are deposited in the

Virginia Tech Entomology Collection.

Habitat Measurements

Bloom density was measured during the same weeks as bee sampling. Bloom density was recorded within three 1-m2 quadrats randomly placed within wildflower plots. Landscape data were obtained from the 2017 data layer from USDA Cropscape. The cover classes of deciduous forest, evergreen forest, mixed forest, shrubland, woody wetland, and herbaceous wetland were aggregated into semi-natural habitat. The percent area of semi-natural habitat was calculated within 250, 500, and 1,000-m radii buffers surrounding each field site. At the 1,000-m radius, the range of SNH surrounding was 10.5 - 69.2% for control sites and 21 - 59.9% for wildflower sites.

Statistical analysis

For all analyses, Apis mellifera L. and bees that could not be identified to species were excluded. Bees not identified to species, ~ 1.5 % of all bee sampled, either have unresolved taxonomy or were in too poor of condition for species determination.

A nonmetric multidimensional scaling (NMDS) plot was created to visualize the bee communities sampled at control and wildflower sites each year. The NMDS was performed using the Bray-Curtis dissimilarity matrix. To test for differences between bee communities at wildflower and control fields, a permutation multivariate analysis of variance (PERMANOVA) was performed. This was done using the functions MetaMDS and anisom in the ‘R’ package

‘Vegan’ (Oksanen et al. 2018). All analyses were performed in ‘R’ Studio v3.5.2 (R Core Team

2018).

Linear mixed-effects models were used to test for differences between wildflower and control fields on species richness and the Shannon-Wiener diversity index. The effects of wildflower plots on aggregate bee abundance and for each of the three most abundant species were tested with generalized linear mixed-effects models with a negative binomial distribution.

For all models, wildflower plot by year interaction and their main effects were fixed effects, with field as a random effect. Abundance, richness, and the Shannon-Wiener diversity index were summed across each year. Linear mixed-effects models were analyzed with the lmer function in the package ‘LmerTest’ (Kuznetsova et al. 2017). Generalized linear mixed-effects models were performed with the glmmadmb function in the package ‘GLMMADMB’ (Skaug et al. 2016).

Bee abundance, the abundance of the three most abundant species, species richness, and the Shannon-Wiener diversity index were the response variables used to test the interaction between wildflower plots and the amount of SNH in the landscape at 250, 500, and 1,000-m radii scales. Model types and random effects are the same as previously described. The landscape by

wildflower plot interaction, their main effects, and a quadratic term of landscape were the fixed effects. The linear and quadratic landscape terms were centered to reduce collinearity. Year was included as a covariate. The scale with the lowest AICc value is presented. Other scales within 2

AICc points are included in supplemental appendix 2.2.

Bloom abundance, plot area, and % area of SNH were used as predictors of wildflower plot quality. Bee abundance, species richness, and the Shannon-Wiener diversity index were used as response variables. A multi-model approach was used to test what factors had an effect on plot effectiveness. Seven a priori models were tested, each factor alone, the two-way interactions between each factor, and an intercept only model to serve as a null comparison. Year was included as a covariate in each model, and field was treated as a random effect. Starting with the model with the most weight, models were included to construct a 95% confidence set of models for model averaging (Harrison, Donaldson, et al. 2018). Multi-model analyses were done using the package ‘MuMIn’ (Barton 2018).

Results A total of 5,384 bees were identified from 85 species (Table 2.1.). A total of 2,151 bees were caught in 2017, and 3,233 were caught in 2018. Agapostemon virescens Fabricius

(Hymenoptera:Halictidae), Peponapis pruinosa (Say) (Hymenoptera:Apidae), and Melissodes bimaculatus Lepeletier (Hymenoptera:Apidae) were the three most abundant species sampled, accounting for 49% of the total bees.

Bee communities at farms with and without wildflower plots were the same. The 95% confidence intervals around the bee communities sampled at control and wildflower fields in

2017 and 2018 greatly overlap each other (Figure 2.1.). PERMANOVA quantitatively verifies that the bee communities sampled in 2017 and 2018 did not differ between control and

wildflower treatments (2017, p = 0.78; 2018, p = 0.66). No differences were detected between wildflower and control fields for species richness and the Shannon-wiener diversity index (Table

2.2.). Overall bee abundance was not different between wildflower and control fields (Table.

2.3.). No treatment by year interactions were detected for any bee community metric (Table 2.2.

& 2.3.). Wildflower plots did not affect the abundance of A. virescens or P. pruinosa, but, significantly fewer M. bimaculatus were sampled from wildflower fields compared to control fields (Table 2.3.).

Landscape effects were detected for species richness, aggregate bee abundance, A. virescens abundance, and M. bimaculatus abundance. When conditioned over landscape wildflower plots had an effect on P. pruinosa and M. bimaculatus abundance. No effects were detected for the Shannon-Wiener index. SNH in the landscape at the 250-m scale has a positive effect on species richness that peaks when SNH is approximately 40% of the land cover before declining (Table 2.4., Fig 2.2.). Aggregate bee, A. virescens, and M. bimaculatus abundance increased with increasing SNH in the landscape at the 1,000-m scale (Table 2.5., Fig 2.2 & 2.3.).

This positive effect peaked then became negative when SNH made up about 40% of the landscape. A landscape by wildflower plot interaction was also significant for aggregate bee abundance, abundance was greater at wildflower sites than control sites when SNH was between

20 – 30% of the surrounding landscape. When conditioned over landscape, P. pruinosa are more abundant at wildflower sites compared to control sites (z = 2.32, p = 0.02), and M. bimaculatus was more abundant at control sites than wildflower sites (z = -1.11, p = 0.03) (Fig 2.3.).

In examining the factors that affect the effectiveness of the wildflower plots, the intercept only models were the only ones selected for species richness and the Shannon-Wiener index

(Table 2.6. and Supplemental Appendix 2.2). Each scale analyzed for bee abundance selected

multiple factors, including the intercept only models (Table 2.6. and Supplemental Appendix

2.2). Percent SNH in the landscape was the top factor at the 500-m scale. SNH in the landscape had a significant negative effect on bee abundance at wildflower plots (Table 2.7.).

Discussion When considered by themselves, the wildflower plots did not have an effect on the bee community. This contrasts with much of the published literature showing that wildflower plots and other restoration efforts are beneficial to bee communities (Williams et al. 2015, Venturini et al. 2017, Tonietto and Larkin 2018, Nicholson et al. 2019). Year by wildflower plot interactions were not approaching significance, indicating that the lack of differences is not likely caused by plot age. It may be that the small-scaled diverse nature of many of the sites were already harboring robust bee communities, mitigating any potential wildflower effects. Having more unmanaged areas on farms and using less pesticides leads to having more speciose and abundant bee communities (Nicholson et al. 2017). Increasing plant diversity on farms is shown to increase the abundance and richness of pollinators (Lichtenberg et al. 2017).

The amount of SNH in the landscape around wildflower plots modulated their effectiveness in promoting bee communities. Keeping with the intermediate landscape hypothesis, we found that there is a range of SNH in the landscape, 20 - 30%, that wildflower plots enhanced bee abundance. This range is higher than the range proposed by Tscharntke et al.

(2005) of 1 - 20%, but better aligns with the range of effectiveness observed in New York of 25-

55% SNH (Tscharntke et al. 2005, Grab et al. 2018). When considering only the wildflower sites, increasing SNH in the landscape led to decreased bee abundance. A meta-analysis of on- farm actions to promote pollinators showed diminished effectiveness as SNH increased (Scheper et al. 2013). The success of on-farm actions in simplified landscapes is due to the ability to

provide resources in an area that is not entirely devoid, but lacking in resources (Scheper et al.

2013). Stronger effects of wildflower plots may be detected if plot locations can be situated in areas where SNH is less than 20% of the landscape, which we were unable to do in this study.

In addition to landscape context surrounding wildflower plots, the quality of the plots themselves could be a factor. Plot area and bloom density were included in model averaging, but were never significant. Bloom density and plot area can be important factors in determining the effectiveness of the plot (Blaauw and Isaacs 2014b, Krimmer et al. 2019), but landscape effects may be stronger (Kleijn et al. 2018). Time since planting was also not a factor, as no changes were seen from year to year for these wildflower plots. The inclusion of intercept only models indicates there could be other factors affecting the plots. Crop production practices at wildflower sites could be impacting the effectiveness of the wildflower plots. P. pruinosa was three times more abundant on farms that practiced no-till than farms that used tillage (Shuler et al. 2005).

Increased pesticide usage and more mowing of non-crop areas decreased on-farm bee abundance and richness (Nicholson et al. 2017).

The three most abundant species sampled, A. virescens, P. pruinosa, and M. bimaculatus, did not have a positive response to the wildflower plots. All three of these species are often sampled from or associated with agriculturally-dominated landscapes (Wheelock et al. 2016,

Buchanan et al. 2017, Harrison, Gibbs, et al. 2018, Collado et al. 2019), which likely already provide all the resources these species need. A. virescens is a generalist that has been documented on more than 50 species of plants (Pickering et al. 2020). P. pruinosa is a squash specialist that evolved with humans and squash, as it followed the cultivation of squash out of

Central America and across the rest of the continent (López-Uribe et al. 2016). M. bimaculatus has shown a strong preference for agricultural land and avoided forests in a survey using 15

years of geo-referenced bee data (Collado et al. 2019). M. bimaculatus and P. pruinosa are two of top 10 most commonly reported bee species foraging on pollinator dependent crops in agricultural systems (Klein et al. 2018). These three species are likely providing the bulk of the pollination services in study area (Winfree et al. 2015). With the potential to be enhanced by increasing species richness (Winfree et al. 2018, Woodcock et al. 2019) by conserving SNH with

250 m of the farm.

Variable farm management of wildflower plots may have affected our findings. The wildflower plots in this study represent a real-world situation where grower-cooperators utilized a government program to create plantings; all but one of the wildflower plots were planted and managed by grower-cooperators. Farmers of two well-draining wildflower sites prepared the seedbed with tillage and an herbicide compared to the other farmers who used only tillage. This difference in preparation resulted in more wildflower biomass at the sites where herbicides were used (Angelella and O’Rourke 2017). It was also recommended that the plots be mowed during the dormant season to promote growth for the following year. Two sites were never mowed after planting, and another was only mowed once after establishment. This lack of mowing may have altered the effectiveness of the plots as it led to increased weed pressure and woody plant encroachment (Angelella, personal observation). Engaging with farmers about their experiences in using government conservation programs can lead to better conservation outcomes rather than simply paying them (McCracken et al. 2015).

When considered alone, wildflower plantings did not enhance the bee communities sampled. However, when SNH in the landscape was considered, there is a range in which the wildflower plantings enhanced bee abundance. The three most abundant bee species sampled in this study are well adapted to agriculture and may provide sufficient pollination services, even in

the absence of pollinator plots. Further studies are needed in real world situations to better understand the effectiveness of government programs to restore pollinator habitat in the United

States. Understanding how to better engage with farmers to best plant and maintain plots may increase the effectiveness of these plots and help curb pollinator declines.

References Angelella, G. M., and M. E. O’Rourke. 2017. Pollinator habitat establishment after organic and no-till seedbed preparation methods. HortScience. 52: 1349–1355. Arathi, H. S., M. W. Vandever, and B. S. Cade. 2019. Diversity and abundance of wild bees in an agriculturally dominated landscape of eastern Colorado. J. Insect Conserv. Balzan, M. V., G. Bocci, and A.-C. Moonen. 2014. Augmenting flower trait diversity in wildflower strips to optimise the conservation of arthropod functional groups for multiple agroecosystem services. J. Insect Conserv. 18: 713–728. Barton, K. 2018. MuMIn: Multi-Model Inference. Begosh, A., L. M. Smith, S. T. McMurry, and J. P. Harris. 2020. Influence of the Conservation Reserve Program (CRP) and playa wetlands on pollinator communities in the Southern High Plains, USA. J. Environ. Manage. 256: 109910. Blaauw, B. R., and R. Isaacs. 2014a. Flower plantings increase wild bee abundance and the pollination services provided to a pollination-dependent crop. J. Appl. Ecol. 51: 890–898. Blaauw, B. R., and R. Isaacs. 2014b. Larger patches of diverse floral resources increase insect pollinator density, diversity, and their pollination of native wildflowers. Basic Appl. Ecol. 15: 701–711. Buchanan, A., J. Gibbs, L. Komondy, and Z. Szendrei. 2017. Bee community of commercial potato fields in Michigan and Bombus impatiens visitation to neonicotinoid-treated potato plants. Insects. 8: 30. Collado, M. Á., D. Sol, and I. Bartomeus. 2019. Bees use anthropogenic habitats despite strong natural habitat preferences. Divers. Distrib. 25: 924–935. Dainese, M., E. A. Martin, M. A. Aizen, M. Albrecht, I. Bartomeus, R. Bommarco, L. G. Carvalheiro, R. Chaplin-Kramer, V. Gagic, L. A. Garibaldi, J. Ghazoul, H. Grab, M. Jonsson, D. S. Karp, C. M. Kennedy, D. Kleijn, C. Kremen, D. A. Landis, D. K. Letourneau, L. Marini, K. Poveda, R. Rader, H. G. Smith, T. Tscharntke, G. K. S. Andersson, I. Badenhausser, S. Baensch, A. D. M. Bezerra, B. Caballero-Lopez, P. Cavigliasso, A. Classen, S. Cusser, J. H. Dudenhöffer, J. Ekroos, T. Fijen, P. Franck, B. M. Freitas, M. P. D. Garratt, C. Gratton, J. Hipólito, A. Holzschuh, L. Hunt, A. L. Iverson, S. Jha, T. Keasar, T. N. Kim, M. Kishinevsky, B. K. Klatt, A.-M. Klein, K. M. Krewenka, S. Krishnan, A. E. Larsen, C. Lavigne, H. Liere, B. Maas, R. E. Mallinger, E. M. Pachon, A. Martínez-Salinas, T. D. Meehan, M. G. E. Mitchell, G. A. R. Molina, M. Nesper, L. Nilsson, M. E. O’Rourke, M. K. Peters, M. Ple, D. de L. Ramos, J. A. Rosenheim, M. Rundlöf, A. Rusch, A. Sáez, J. Scheper, M. Schleuning, J. M. Schmack, A. R. Sciligo, C. Seymour, D. A. Stanley, R. Stewart, J. C. Stout, L. Sutter, M. B. Takada, H. Taki, G. Tamburini, M. Tschumi, B. F. Viana, C. Westphal, B. K. Willcox, S. D. Wratten, A. Yoshioka, C. Zaragoza-Trello, W. Zhang, Y. Zou, and I. Steffan-Dewenter. 2019. A global synthesis reveals biodiversity- mediated benefits for crop production. Sci. Adv. 14. Garibaldi, L. A., I. Steffan-Dewenter, R. Winfree, M. A. Aizen, R. Bommarco, S. A. Cunningham, C. Kremen, L. G. Carvalheiro, L. D. Harder, O. Afik, I. Bartomeus, F. Benjamin, V. Boreux, D. Cariveau, N. P. Chacoff, J. H. Dudenhöffer, B. M. Freitas, J. Ghazoul, S. Greenleaf, J. Hipólito, A. Holzschuh, B. Howlett, R. Isaacs, S. K. Javorek, C. M. Kennedy, K. M. Krewenka, S. Krishnan, Y. Mandelik, M. M. Mayfield, I. Motzke, T. Munyuli, B. A. Nault, M. Otieno, J. Petersen, G. Pisanty, S. G. Potts, R. Rader, T. H. Ricketts, M. Rundlöf, C. L. Seymour, C. Schüepp, H.

Szentgyörgyi, H. Taki, T. Tscharntke, C. H. Vergara, B. F. Viana, T. C. Wanger, C. Westphal, N. Williams, and A. M. Klein. 2013. Wild pollinators enhance fruit set of crops regardless of honey bee abundance. Science. 339: 1608–1611. Garratt, M. P. D., D. Senapathi, D. J. Coston, S. R. Mortimer, and S. G. Potts. 2017. The benefits of hedgerows for pollinators and natural enemies depends on hedge quality and landscape context. Agric. Ecosyst. Environ. 247: 363–370. Goulson, D., E. Nicholls, C. Botias, and E. L. Rotheray. 2015. Bee declines driven by combined stress from parasites, pesticides, and lack of flowers. Science. 347: 1255957– 1255957. Grab, H., M. G. Branstetter, N. Amon, K. R. Urban-Mead, M. G. Park, J. Gibbs, E. J. Blitzer, K. Poveda, G. Loeb, and B. N. Danforth. 2019. Agriculturally dominated landscapes reduce bee phylogenetic diversity and pollination services. 4. Grab, H., K. Poveda, B. Danforth, and G. Loeb. 2018. Landscape context shifts the balance of costs and benefits from wildflower borders on multiple ecosystem services. Proc R Soc B. 285: 20181102. Harrison, T., J. Gibbs, and R. Winfree. 2018. Phylogenetic homogenization of bee communities across ecoregions. Glob. Ecol. Biogeogr. 27: 1457–1466. Harrison, X. A., L. Donaldson, M. E. Correa-Cano, J. Evans, D. N. Fisher, C. E. D. Goodwin, B. S. Robinson, D. J. Hodgson, and R. Inger. 2018. A brief introduction to mixed effects modelling and multi-model inference in ecology. PeerJ. 6: e4794. Herbertsson, L., A. M. Jönsson, G. K. S. Andersson, K. Seibel, M. Rundlöf, J. Ekroos, M. Stjernman, O. Olsson, and H. G. Smith. 2018. The impact of sown flower strips on plant reproductive success in Southern Sweden varies with landscape context. Agric. Ecosyst. Environ. 259: 127–134. Kennedy, C. M., E. Lonsdorf, M. C. Neel, N. M. Williams, T. H. Ricketts, R. Winfree, R. Bommarco, C. Brittain, A. L. Burley, D. Cariveau, L. G. Carvalheiro, N. P. Chacoff, S. A. Cunningham, B. N. Danforth, J.-H. Dudenhöffer, E. Elle, H. R. Gaines, L. A. Garibaldi, C. Gratton, A. Holzschuh, R. Isaacs, S. K. Javorek, S. Jha, A. M. Klein, K. Krewenka, Y. Mandelik, M. M. Mayfield, L. Morandin, L. A. Neame, M. Otieno, M. Park, S. G. Potts, M. Rundlöf, A. Saez, I. Steffan-Dewenter, H. Taki, B. F. Viana, C. Westphal, J. K. Wilson, S. S. Greenleaf, and C. Kremen. 2013. A global quantitative synthesis of local and landscape effects on wild bee pollinators in agroecosystems. Ecol. Lett. 16: 584–599. Kleijn, D., T. E. W. Linders, A. Stip, J. C. Biesmeijer, F. L. Wäckers, and T. Bukovinszky. 2018. Scaling up effects of measures mitigating pollinator loss from local- to landscape- level population responses. Methods Ecol. Evol. 9: 1727–1738. Klein, A.-M., V. Boreux, F. Fornoff, A.-C. Mupepele, and G. Pufal. 2018. Relevance of wild and managed bees for human well-being. Curr. Opin. Insect Sci. 26: 82–88. Klein, A.-M., B. E. Vaissière, J. H. Cane, I. Steffan-Dewenter, S. A. Cunningham, C. Kremen, and T. Tscharntke. 2007. Importance of pollinators in changing landscapes for world crops. Proc. R. Soc. B Biol. Sci. 274: 303–313. Kleintjes Neff, P., C. Locke, and E. Lee-Mӓder. 2017. Assessing a farmland set-aside conservation program for an endangered butterfly: USDA State Acres for Wildlife Enhancement (SAFE) for the Karner blue butterfly. J. Insect Conserv. 21: 929–941. Krimmer, E., E. A. Martin, J. Krauss, A. Holzschuh, and I. Steffan-Dewenter. 2019. Size, age and surrounding semi-natural habitats modulate the effectiveness of flower-rich agri-

environment schemes to promote pollinator visitation in crop fields. Agric. Ecosyst. Environ. 284: 106590. Kuznetsova, A., P. B. Brockhoff, and R. H. B. Christensen. 2017. lmerTest Package: Tests in Linear Mixed Effects Models. J. Stat. Softw. 82. Lichtenberg, E. M., C. M. Kennedy, C. Kremen, P. Batáry, F. Berendse, R. Bommarco, N. A. Bosque‐Pérez, L. G. Carvalheiro, W. E. Snyder, N. M. Williams, R. Winfree, B. K. Klatt, S. Åström, F. Benjamin, C. Brittain, R. Chaplin‐Kramer, Y. Clough, B. Danforth, T. Diekötter, S. D. Eigenbrode, J. Ekroos, E. Elle, B. M. Freitas, Y. Fukuda, H. R. Gaines‐Day, H. Grab, C. Gratton, A. Holzschuh, R. Isaacs, M. Isaia, S. Jha, D. Jonason, V. P. Jones, A.-M. Klein, J. Krauss, D. K. Letourneau, S. Macfadyen, R. E. Mallinger, E. A. Martin, E. Martinez, J. Memmott, L. Morandin, L. Neame, M. Otieno, M. G. Park, L. Pfiffner, M. J. O. Pocock, C. Ponce, S. G. Potts, K. Poveda, M. Ramos, J. A. Rosenheim, M. Rundlöf, H. Sardiñas, M. E. Saunders, N. L. Schon, A. R. Sciligo, C. S. Sidhu, I. Steffan‐Dewenter, T. Tscharntke, M. Veselý, W. W. Weisser, J. K. Wilson, and D. W. Crowder. 2017. A global synthesis of the effects of diversified farming systems on arthropod diversity within fields and across agricultural landscapes. Glob. Change Biol. 23: 4946–4957. López-Uribe, M. M., J. H. Cane, R. L. Minckley, and B. N. Danforth. 2016. Crop domestication facilitated rapid geographical expansion of a specialist pollinator, the squash bee Peponapis pruinosa. Proc. R. Soc. B Biol. Sci. 283: 20160443. McCracken, M. E., B. A. Woodcock, M. Lobley, R. F. Pywell, E. Saratsi, R. D. Swetnam, S. R. Mortimer, S. J. Harris, M. Winter, S. Hinsley, and J. M. Bullock. 2015. Social and ecological drivers of success in agri-environment schemes: the roles of farmers and environmental context. J. Appl. Ecol. 52: 696–705. Nicholson, C. C., I. Koh, L. L. Richardson, A. Beauchemin, and T. H. Ricketts. 2017. Farm and landscape factors interact to affect the supply of pollination services. Agric. Ecosyst. Environ. 250: 113–122. Nicholson, C. C., K. L. Ward, N. M. Williams, R. Isaacs, K. S. Mason, J. K. Wilson, J. Brokaw, L. J. Gut, N. L. Rothwell, T. J. Wood, S. Rao, G. D. Hoffman, J. Gibbs, R. W. Thorp, and T. H. Ricketts. 2019. Mismatched outcomes for biodiversity and ecosystem services: testing the responses of crop pollinators and wild bee biodiversity to habitat enhancement. Ecol. Lett. ele.13435. NRCS. 2015. Using 2014 Farm Bill Programs for Pollinator Conservation. Oksanen, J., F. G. Blanchet e, M. Friendly, R. Kindt, P. Legendre, D. McFlinn, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. Stevens, E. Szoecs, and H. Wagner. 2018. vegan: Community Ecology Package. Otto, C. R. V., S. O’Dell, R. B. Bryant, N. H. Euliss, R. M. Bush, and M. D. Smart. 2017. Using publicly available data to quantify plant–pollinator interactions and evaluate conservation seeding mixes in the northern Great Plains. Environ. Entomol. 46: 565–578. Pickering, J., J. Ascher, and S. Droege. 2020. Apoidea, Bees; Apoid Wasps. Discov. Life. (https://www.discoverlife.org/20/q?search=Apoidea). Potts, S. G., J. C. Biesmeijer, C. Kremen, P. Neumann, O. Schweiger, and W. E. Kunin. 2010. Global pollinator declines: trends, impacts and drivers. Trends Ecol. Evol. 25: 345– 353. R Core Team. 2018. A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Ricketts, T. H., J. Regetz, I. Steffan-Dewenter, S. A. Cunningham, C. Kremen, A. Bogdanski, B. Gemmill-Herren, S. S. Greenleaf, A. M. Klein, M. M. Mayfield, L. A. Morandin, A. Ochieng’, and B. F. Viana. 2008. Landscape effects on crop pollination services: are there general patterns? Ecol. Lett. 11: 499–515. Scheper, J., A. Holzschuh, M. Kuussaari, S. G. Potts, M. Rundlöf, H. G. Smith, and D. Kleijn. 2013. Environmental factors driving the effectiveness of European agri- environmental measures in mitigating pollinator loss – a meta-analysis. Ecol. Lett. 16: 912–920. Shuler, R. E., T. H. Roulston, and G. E. Farris. 2005. Farming practices influence wild pollinator populations on squash and pumpkin. J. Econ. Entomol. 98: 790–795. Skaug, H., D. Fournier, A. Nielsen, A. Magnusson, and B. Bolker. 2016. Generalized Linear Mixed Models using “AD Model Builder.” Tonietto, R. K., and D. J. Larkin. 2018. Habitat restoration benefits wild bees: A meta- analysis. J. Appl. Ecol. 55: 582–590. Tscharntke, T., A. M. Klein, A. Kruess, I. Steffan-Dewenter, and C. Thies. 2005. Landscape perspectives on agricultural intensification and biodiversity – ecosystem service management. Ecol. Lett. 8: 857–874. Tuell, J. K., A. K. Fiedler, D. Landis, and R. Isaacs. 2008. Visitation by wild and managed bees (Hymenoptera: Apoidea) to eastern U.S. native plants for use in conservation programs. Environ. Entomol. 37: 707–718. US EPA, O. 2014. Federal Pollinator Health Task Force: EPA’s Role. (/pollinator- protection/federal-pollinator-health-task-force-epas-role). USDA. 2018. Environmental Quality Incentives Program (EQIP) | Farm Bill Report (FY 2009 through FY 2018) | NRCS. (https://www.nrcs.usda.gov/Internet/NRCS_RCA/reports/fb08_cp_eqip.html). USDA. 2019. Conservation Reserve Program Fact Sheet. United States Department of Agriculture. Vaughn, M., and M. Skinner. 2015. Using 2014 Farm Bill Programs for Pollinator Conservation (Biology Technical Note No. No. 78). United States Department of Agriculture. Venturini, E. M., F. A. Drummond, A. K. Hoshide, A. C. Dibble, and L. B. Stack. 2017. Pollination reservoirs for wild bee habitat enhancement in cropping systems: a review. Agroecol. Sustain. Food Syst. 41: 101–142. Wheelock, M. J., K. P. Rey, and M. E. O’Neal. 2016. Defining the insect pollinator community found in Iowa corn and soybean fields: implications for pollinator conservation. Environ. Entomol. 45: 1099–1106. Williams, N. M., K. L. Ward, N. Pope, R. Isaacs, J. Wilson, E. A. May, J. Ellis, J. Daniels, A. Pence, K. Ullmann, and J. Peters. 2015. Native wildflower plantings support wild bee abundance and diversity in agricultural landscapes across the United States. Ecol. Appl. 25: 2119–2131. Winfree, R., J. R. Reilly, I. Bartomeus, D. P. Cariveau, N. M. Williams, and J. Gibbs. 2018. Species turnover promotes the importance of bee diversity for crop pollination at regional scales. Science. 359: 791–793. Winfree, R., J. W. Fox, N. M. Williams, J. R. Reilly, and D. P. Cariveau. 2015. Abundance of common species, not species richness, drives delivery of a real-world ecosystem service. Ecol. Lett. 18: 626–635.

Winfree, R., N. M. Williams, J. Dushoff, and C. Kremen. 2007. Native bees provide insurance against ongoing honey bee losses. Ecol. Lett. 10: 1105–1113. Winfree, R., N. M. Williams, H. Gaines, J. S. Ascher, and C. Kremen. 2008. Wild bee pollinators provide the majority of crop visitation across land-use gradients in New Jersey and Pennsylvania, USA. J. Appl. Ecol. 45: 793–802. Woodcock, B. A., M. P. D. Garratt, G. D. Powney, R. F. Shaw, J. L. Osborne, J. Soroka, S. A. M. Lindström, D. Stanley, P. Ouvrard, M. E. Edwards, F. Jauker, M. E. McCracken, Y. Zou, S. G. Potts, M. Rundlöf, J. A. Noriega, A. Greenop, H. G. Smith, R. Bommarco, W. van der Werf, J. C. Stout, I. Steffan-Dewenter, L. Morandin, J. M. Bullock, and R. F. Pywell. 2019. Meta-analysis reveals that pollinator functional diversity and abundance enhance crop pollination and yield. Nat. Commun. 10: 1481.

Figure 2.1. NMDS plot of the of the bee communities sampled in 2017 (A) and 2018 (B) at control and wildflower fields. Circles represent 95% confidence intervals around the centroid of the sampled communities. (Stress values: 2017 = 0.18, 2018 = 0.18)

Figure 2.2. The effects of the interaction of wildflower plots and the amount of SNH in the landscape on the Shannon-Wiener diversity index (A), bee species richness (B) and abundance (C). The scale presented has the lowest AICc score of the scales tested.

Predicted bee abundance bee Predicted

Figure 2.3. The effects of the interaction of SNH in the landscape and wildflower

plots on the abundance of three most common bee species sampled (A) A. virescens, (B) P. pruinosa, (C) M. bimaculatus. The scale presented has the lowest AICc score of the scales tested.

Table 2.1. List of species and the number caught over the duration of the study at wildflower and control sites. Family Species Wildflower Control Halictidae Agapostemon sericeus 5 1 Agapostemon splendens 7 9 Agapostemon texanus 1 1 Agapostemon virescens 736 1050 Augochloropsis metallica metallica 3 0 Augochloropsis metallica fulgoda 0 1 Augochlora pura 8 4 Augochlorella aurata 134 46 Augochlorella gratiosa 0 3 Halictus confusus 20 5 Halictus parallelus 0 2 Halictus poeyi/ligatus 108 46 Halictus rubicundus 0 11 Lasioglossum admirandum 1 8 Lasioglossum birkmanni 0 1 Lasioglossum bruneri 10 37 Lasioglossum callidum 66 56 Lasioglossum coreopsis 5 5 Lasioglossum cressoni 1 3 Lasioglossum flordanum 0 1 Lasioglossum hitchensi 31 43 Lasioglossum illinoense 1 1 Lasioglossum imitatum 1 1 Lasioglossum leucocomum 0 18 Lasioglossum lustrans 3 0 Lasioglossum nelumbonis 1 0 Lasioglossum oblongum 2 8 Lasioglossum pectorale 49 10 Lasioglossum pilosum 74 91 Lasioglossum rozeni 1 0 Lasioglossum tegulare 43 46 Lasioglossum trigeminum 48 20 Lasioglossum versatum 49 17 Lasioglossum vierecki 1 0 Lasioglossum weemsi 0 6 Lasioglossum zephyrum 1 0 Andrenidae Andrena carlini 0 1 Andrena nasonii 1 1 Andrena perplexa 2 0

Family Species Wildflower Control Calliopsis andreniformis 56 6 Apidae Anthophora abrubta 7 2 Bombus bimaculatus 7 2 Bombus fervidus 0 1 Bombus griseocollis 11 2 Bombus impatiens 24 36 Bombus pensylvanicus 108 86 Ceratina calcarta 14 28 Ceratina dupla 39 12 Ceratina floridana 0 2 Ceratina mikmaqi 0 3 Ceratina strenua 3 1 Eucera hamata 55 81 Eucera rosae 1 4 Florilegus condiguns 3 0 Holcopasites calliopsidis 0 1 Melissodes agilis 1 0 Melissodes bimaculatus 112 282 Melissodes comptoides 106 127 Melissodes comumunis 29 51 Melissodes desponsus 0 1 Melissodes denticulatus 1 0 Melissodes trinodis 82 11 Peponapis pruinosa 293 155 Ptilothrix bombiformis 55 57 Svastra atripes 35 45 Svastra obliqua 6 1 Triepeolus concavus 1 0 Triepeolus lunatus 0 1 Triepeolus simplex 2 0 Xenoglossa strenua 1 6 Xylocopa virginica 13 14 Colletidae Hylaeus ornatus 1 0 Megachilidae Heriades leavitti/variolosa 1 0 Hoplitis pilosifrons 3 1 Hoplitis producta 1 0 Hoplitis truncata 1 0 Megachile brevis 7 1 Megachile campanulae 1 1 Megachile mendica 3 1

Family Species Wildflower Control Megachile rotundata 5 1 Osmia collinsae 1 3 Osmia pumila 4 0 Osmia sandhouseae 3 0 Osmiia georgica 1 0 Stelis lateralis 1 0

Table 2.2. Results of linear mixed effects models on the effect of wildflower plantings on the Shannon-Wiener diversity index and bee richness. Response Parameter Estimate std.error df t-value p-value Shannon Intercept 1.92 0.16 34.3 12.2 <0.0001* Year 0.03 0.19 20.2 0.14 0.88 Wildflower 0.16 0.22 33.5 0.70 0.48 Year*Wildflower 0.08 0.26 18.4 0.32 0.75 Richness Intercept 13.2 1.59 34.5 8.28 <0.0001* Year 5.18 2.06 20.5 2.51 0.02* Wildflower 2.99 2.55 35.1 1.32 0.19 Year*Wildflower -0.88 2.88 18.5 -0.31 0.76 *Significant at p<0.05

Table 2.3. Results of generalized linear mixed effects model on the effect of wildflower plantings on the abundance of all bees and the three most abundant species sampled Species Effect Estimate std.error z-value p-value All bees Intercept 4.28 0.24 17.5 <0.0001* Year 0.66 0.22 2.92 0.003* Wildflower 0.29 0.35 0.85 0.39 Year*Wildflower -0.30 0.31 -0.97 0.33 A. virescens Intercept 1.77 0.66 2.70 0.007* Year 0.91 0.35 2.64 0.008* Wildflower 0.32 0.92 0.34 0.73 Year*Wildflower -0.69 0.47 -1.48 0.14 P. pruinosa Intercept 1.46 0.65 2.23 0.03* Year -0.98 0.64 -1.54 0.12 Wildflower 0.29 0.86 0.34 0.73 Year*Wildflower 1.18 0.85 1.40 0.16 M. bimaculatus Intercept 2.23 0.54 4.07 <0.0001* Year 0.33 0.77 0.43 0.67 Wildflower -1.93 0.79 -2.43 0.02* Year*Wildflower 1.53 1.06 1.45 0.15 *Significant at p<0.05

Table 2.4. Results of linear mixed effects models testing the effects of wildflower plots and the amount of semi-natural habitat in the landscape on the species richness and Shannon-wiener diversity index of sampled bees. Scales presented had the lowest AICc of the three scales analyzed. Statistic Model term Estimate std.error df t-value p-value Richness Intercept 13.9 1.34 25.9 10.3 <0.0001* Year 4.96 1.43 18.2 3.45 0.003* Wildflower 1.72 1.66 13.7 1.04 0.31 SNH 250 0.29 0.13 14.8 2.23 0.04* WF * SNH 250 -0.04 0.09 13.4 -0.55 0.59 SNH 2502 -0.004 0.002 13.8 -2.21 0.04* Shannon Intercept 1.85 0.14 25.2 12.7 <0.0001* Year 0.06 0.13 19.8 0.49 0.62 Wildflower 0.22 0.20 17.3 1.13 0.27 SNH 1000 -0.04 0.02 16.8 -1.51 0.15 WF * SNH 1000 0.017 0.014 15.9 1.22 0.24 SNH 10002 0.0004 0.0003 16.6 1.34 0.20 *Significant at p<0.05

Table 2.5. Results of generalized linear mixed effects models testing the effects of wildflower plots and the amount of semi-natural habitat in the landscape on total bee abundance and the three most abundant species sampled. Species Term Estimate std.error z-value p-value All bees Intercept 4.53 0.18 24.7 <0.0001* Year 0.53 0.18 3.05 0.002* Wildflower -0.06 0.22 -0.25 0.80 SNH 1000 0.13 0.03 4.12 <0.0001* WF * SNH 1000 -0.04 0.02 -2.55 0.01* SNH 10002 -0.001 0.0001 -3.72 0.0002* A. virescens Intercept 1.61 0.68 2.35 0.02* Year 0.56 0.26 2.16 0.03* Wildflower 0.20 0.90 0.22 0.83 SNH 1000 0.27 0.13 1.99 0.04* WF * SNH 1000 0.02 0.07 0.34 0.73 SNH 10002 -0.004 0.001 -2.18 0.03* P. pruinosa Intercept 0.82 0.59 1.40 0.16 Year -0.39 0.44 -0.89 0.37 Wildflower 1.88 0.81 2.32 0.02* SNH 500 -0.15 0.08 -1.81 0.07 WF * SNH 500 -0.06 0.05 -1.19 0.24 SNH 5002 0.002 0.002 1.69 0.09 Melissodes bimaculatus Intercept 1.77 0.40 4.40 <0.0001* Year 1.17 0.48 2.40 0.02* Wildflower -1.11 0.50 -2.21 0.03* SNH 1000 0.17 0.06 2.62 0.01* WF * SNH 1000 -0.01 0.03 -0.17 0.87 SNH 10002 -0.002 0.001 -2.85 0.004* *Significant at p<0.05

Table 2.6. Results of model selection for factors affecting the effectiveness of wildflower plots for the Shannon-Wiener index, species richness, and abundance. The scale with the lowest average AICc is presented. All other scales are presented in Supplemental Appendix 2. Response Scale (m) Model k logLik AICc Δ AICc Weight Shannon 1000 Intercept only 4 -11.36 33.38 - 0.969† Avg. blooms 5 -13.42 41.12 7.737 0.02 % SNH 5 -14.08 42.46 9.073 0.01 Plot area 5 -20.17 54.62 21.241 0 Avg. blooms * % SNH 7 -23.05 69.44 36.06 0 Avg. blooms * area 7 -32.26 87.85 54.473 0 Plot area * % SNH 7 -32.93 89.19 55.803 0 Richness 250 Intercept only 4 -57.24 125.15 - 0.917† % SNH 5 -58.52 131.32 6.175 0.042 Avg. blooms 5 -58.55 131.39 6.242 0.04 Plot area 5 -62.77 139.83 14.678 0.001 Avg. blooms * % SNH 7 -63.3 149.94 24.786 0 Plot area * % SNH 7 -70.82 164.97 39.817 0 Avg. blooms * area 7 -72.33 167.99 42.845 0 Abundance 500 % SNH 5 -107.5 229.29 - 0.836† Avg. blooms * % SNH 7 -105.89 235.12 5.834 0.045† Intercept only 4 -112.3 235.26 5.969 0.042† Avg. blooms 5 -110.74 235.76 6.474 0.033 Plot area 5 -111.2 236.68 7.394 0.021 Plot area * % SNH 7 -106.73 236.79 7.496 0.02 Avg. blooms * area 7 -108.56 240.46 11.172 0.003 K is the number of parameters in the model † models selected for averaging

Table 2.7. Results of model averaging for factors affecting the effectiveness of wildflower plots in promoting bee abundance at the 500-m scale. Other scales analyzed are presented in Supplemental Appendix 2. Term Estimate std.error z-value p-value

Intercept 5.72 0.53 10.7 <0.0001* Year 0.36 0.25 1.45 0.14 % SNH -0.037 0.01 2.85 0.004* Avg. blooms 0.68 0.04 1.46 0.14 % SNH * Blooms -0.002 0.001 1.27 0.20 * significant at p < 0.05

Supplemental Appendix 2.1. List of plant species used in each mix for the wildflower treatment sites.

Table S2.1.1 Grass and forb mix for well-drained soils (N = 7) Common Name Scientific Name

Little Bluestem (G) Schizachyrium scoparium Splitbeard Bluestem (G) Andropogon ternarius Narrowleaf Mountain Mint Pycnanthemum tenuifolium (P) Plains Coreopsis (A) Coreopsis tinctoria Partridge Pea (A) Chamaecrista fasciculata Black-eyed Susan (B) Rudbeckia hirta Bergamot, Spotted (P) Monarda fistulosa Lanceleaf Coreopsis (P) Coreopsis lanceolata Maximilian Sunflower (P) Helianthus maximilianii Indian Blanket (A) Gaillardia pulchella Purple Coneflower (P) Echinacea purpurea G = grass, A = annual, B = biennial, P = perennial

Table S2.1.2 Grass and forb mix for poorly-drained soils (N = 2) Common Name Scientific Name

Beaked Panicum (G) Panicum anceps Redtop Panicum (G) Panicum rigidulum Aster, Purple-stemmed (P) Symphyotrichum puniceum var. puniceum Sneezeweed, Common (P) Helenium autumnale Coreopsis, Plains (A) Coreopsis tinctoria Goldenrod, Wrinkleleaf (P) Solidago rugosa Joe Pye Weed, Spotted (P) Eupatoriadelphus fistulosus Partridge Pea (A) Chamaecrista fasciculata Rattlesnake Master (P) Eryngium yuccifolium Rosemallow (P) Hibiscus moscheutos Narrowleaf Sunflower (P) Helianthus angustifolius G = grass, A = annual, B = biennial P = perennial

Table S2.1.3 Forb mix for replacement well-drained field (N = 1) Common Name Scientific Name

Supplemental Appendix 2.2. AICc scores for scale selection, and results of other scales analyzed that were within 2 AICc points of the lowest scoring scale.

Table S2.2.1. Model selection for the best scale to analyze the effects of wildflower plots and SNH in the landscape. The best model was selected based on AICc scores. All models within two points of lowest score were analyzed. Response Scale AICc Shannon-Wiener 1000 104.97* diversity index 500 107.38 250 109.75

Species richness 1000 266.38 500 265.40* 250 264.33*

Total bee abundance 1000 455.99* 500 463.19 250 466.37

Agapostemon virecens 1000 332.79* abundance 500 333.85 250 331.04*

Peponapis pruinosa 1000 255.36* abundance 500 253.48* 250 253.83*

Melissodes bimaculatus 1000 247.84* abundance 500 249.20* 250 249.7* * scales analyzed

Table S2.2.2. Results of linear mixed effects models testing the effects of wildflower plots and the amount of semi-natural habitat in the landscape on the species richness and Shannon-Wiener diversity index of sampled bees. Response Term Estimate std.error df t-value p-value Richness Intercept 13.7 1.48 25.1 9.28 <0.0001* Year 4.77 1.42 18.7 3.69 0.003* Wildflower 2.58 2.03 15.7 1.27 0.22 SNH 500 0.16 0.22 15.4 0.73 0.48 WF * SNH 500 -0.13 0.13 14.3 -0.93 0.37 SNH 5002 -0.002 0.003 15.2 -0.65 0.53

Intercept 13.9 1.34 25.9 10.3 <0.0001* Year 4.96 1.43 18.2 3.45 0.003* Wildflower 1.72 1.66 13.7 1.04 0.31 SNH 250 0.29 0.13 14.8 2.23 0.04*† WF * SNH 250 -0.04 0.09 13.4 -0.55 0.59 SNH 2502 -0.004 0.002 13.8 -2.21 0.04* Shannon Intercept 1.85 0.14 25.2 12.7 <0.0001* Year 0.06 0.13 19.8 0.49 0.62 Wildflower 0.22 0.20 17.3 1.13 0.27 SNH 1000 -0.04 0.02 16.8 -1.51 0.15† WF * SNH 1000 0.017 0.014 15.9 1.22 0.24 SNH 10002 0.0004 0.0003 16.6 1.34 0.20 *Significant at p<0.05 † Scale with lowest AICc

Table S2.2.3. Results of generalized linear mixed effects models testing the effects of wildflower plots and the amount of semi-natural habitat in the landscape on the three most abundant species sampled. Species Term Estimate std.error z-value p-value A. virescens Intercept 1.61 0.68 2.35 0.02* Year 0.56 0.26 2.16 0.03* Wildflower 0.20 0.90 0.22 0.83 SNH 1000 0.27 0.13 1.99 0.04*† WF * SNH 1000 0.02 0.07 0.34 0.73 SNH 10002 -0.004 0.001 -2.18 0.03*

Intercept 1.54 0.67 2.31 0.02* Year 0.59 0.26 2.23 0.03* Wildflower 0.09 0.88 0.11 0.92 SNH 250 0.09 0.06 1.47 0.14 WF * SNH 250 0.08 0.05 1.44 0.15 SNH 2502 -0.003 0.001 -2.41 0.02*

P. pruinosa Intercept 1.20 0.57 2.10 0.04* Year -0.26 0.42 -0.61 0.53 Wildflower 0.98 0.76 1.29 0.20 SNH 1000 0.07 0.11 0.65 0.51 WF * SNH 1000 -0.08 0.05 -1.56 0.12 SNH 10002 -0.001 0.001 -0.40 0.68

Intercept 0.82 0.59 1.40 0.16 Year -0.39 0.44 -0.89 0.37 Wildflower 1.88 0.81 2.32 0.02* SNH 500 -0.15 0.08 -1.81 0.07† WF * SNH 500 -0.06 0.05 -1.19 0.24 SNH 5002 0.002 0.002 1.69 0.09

Intercept 0.86 0.58 1.49 0.14 Year -0.30 0.42 -0.71 0.48 Wildflower 1.50 0.74 2.03 0.04* SNH 250 -0.09 0.05 -1.69 0.09 WF * SNH 250 -0.03 0.04 -0.71 0.48 SNH 2502 -0.001 0.001 1.60 0.11

Species Term Estimate std.error z-value p-value Melissodes bimaculatus Intercept 1.77 0.40 4.40 <0.0001* Year 1.17 0.48 2.40 0.02* Wildflower -1.11 0.50 -2.21 0.03* SNH 1000 0.17 0.06 2.62 0.01*† WF * SNH 1000 -0.01 0.03 -0.17 0.87 SNH 10002 -0.002 0.001 -2.85 0.004*

Intercept 1.59 0.41 3.88 0.0001* Year 1.16 0.54 2.15 0.03* Wildflower -0.80 0.63 -1.28 0.20 SNH 500 0.01 0.07 0.17 0.86 WF * SNH 500 0.02 0.04 0.59 0.57 SNH 5002 -0.001 0.001 -0.84 0.40

Intercept 1.58 0.40 3.99 <0.0001* Year 1.12 0.52 2.15 0.03* Wildflower -0.70 0.52 -1.36 0.18 SNH 250 -0.03 0.04 -0.73 0.46 WF * SNH 250 0.02 0.03 0.92 0.36 SNH 2502 -0.0002 0.0006 -0.37 0.71 *Significant at p<0.05 † Scale with lowest AICc

Table S2.2.4. Results of model selection for factors affecting the effectiveness of wildflower plots in conserving bee species richness. Scale (m) Model k logLik AICc Δ AICc Weight 1000 Intercept only 4 -57.24 125.15 - 0.913† % SNH 5 -58.42 131.13 5.986 0.046 Avg. blooms 5 -58.55 131.39 6.242 0.04 Plot area 5 -62.77 139.83 14.678 0.001 Avg. blooms * % SNH 7 -62.65 148.64 23.486 0 Plot area * % SNH 7 -71.61 166.55 41.401 0 Avg. blooms * area 7 -72.33 167.99 42.845 0 500 Intercept only 4 -57.24 125.15 - 0.871† % SNH 5 -57.7 129.69 4.541 0.09 Avg. blooms 5 -58.55 131.39 6.242 0.038 Plot area 5 -62.77 139.83 14.678 0.001 Avg. blooms * % SNH 7 -61.38 146.09 20.944 0 Plot area * % SNH 7 -71.41 166.15 40.998 0 Avg. blooms * area 7 -72.33 167.99 42.845 0 250 Intercept only 4 -57.24 125.15 - 0.917† % SNH 5 -58.52 131.32 6.175 0.042 Avg. blooms 5 -58.55 131.39 6.242 0.04 Plot area 5 -62.77 139.83 14.678 0.001 Avg. blooms * % SNH 7 -63.3 149.94 24.786 0 Plot area * % SNH 7 -70.82 164.97 39.817 0 Avg. blooms * area 7 -72.33 167.99 42.845 0 k the number of parameters in the model. † Models selected

Table S2.2.5. Results of model selection for factors affecting the effectiveness of wildflower plots as measured by the Shannon-Wiener Diversity Index. Scale (m) Model k logLik AICc Δ AICc Weight 1000 Intercept only 4 -11.36 33.38 - 0.969† Avg. blooms 5 -13.42 41.12 7.737 0.02 % SNH 5 -14.08 42.46 9.073 0.01 Plot area 5 -20.17 54.62 21.241 0 Avg. blooms * % SNH 7 -23.05 69.44 36.06 0 Avg. blooms * area 7 -32.26 87.85 54.473 0 Plot area * % SNH 7 -32.93 89.19 55.803 0 500 Intercept only 4 -11.36 33.38 - 0.974† Avg. blooms 5 -13.42 41.12 7.737 0.02 % SNH 5 -14.79 43.86 10.476 0.005 Plot area 5 -20.17 54.62 21.241 0 Avg. blooms * % SNH 7 -22.57 68.48 35.101 0 Avg. blooms * area 7 -32.26 87.85 54.473 0 Plot area * % SNH 7 -33.27 89.87 56.493 0 250 Intercept only 4 -11.36 33.38 - 0.977† Avg. blooms 5 -13.42 41.12 7.737 0.02 % SNH 5 -15.44 45.16 11.783 0.003 Plot area 5 -20.17 54.62 21.241 0 Avg. blooms * % SNH 7 -20.87 65.08 31.696 0 Avg. blooms * area 7 -32.26 87.85 54.473 0 Plot area * % SNH 7 -34.36 92.05 58.673 0 k is the number of parameters in the model. † Models selected

Table S2.2.6. Results of model selection for factors affecting the effectiveness of wildflower plots in promoting bee abundance. Scale (m) Model k logLik AICc Δ AICc Weight 1000 % SNH 5 -109.75 233.8 - 0.457† Intercept only 4 -112.3 235.26 1.463 0.22† Avg. blooms 5 -110.74 235.76 1.968 0.171† Plot area 5 -111.2 236.68 2.888 0.108 Avg. blooms * area 7 -108.56 240.46 6.666 0.016 Plot area * % SNH 7 -108.73 240.79 6.994 0.014 Avg. blooms * % SNH 7 -108.73 240.79 6.998 0.014 500 % SNH 5 -107.5 229.29 - 0.836† Avg. blooms * % SNH 7 -105.89 235.12 5.834 0.045† Intercept only 4 -112.3 235.26 5.969 0.042† Avg. blooms 5 -110.74 235.76 6.474 0.033 Plot area 5 -111.2 236.68 7.394 0.021 Plot area * % SNH 7 -106.73 236.79 7.496 0.02 Avg. blooms * area 7 -108.56 240.46 11.172 0.003 250 Intercept only 4 -112.3 235.26 - 0.351† Avg. blooms 5 -110.74 235.76 0.505 0.273† Plot area 5 -111.2 236.68 1.425 0.172† % SNH 5 -111.26 236.8 1.539 0.163 Avg. blooms * area 7 -108.56 240.46 5.203 0.026 Avg. blooms * % SNH 7 -109.5 242.34 7.083 0.01 Plot area * % SNH 7 -110.14 243.62 8.359 0.005 K is the number of parameters in the model † models used for averaging

Table S2.2.7. Results of model averaging for factors affecting the effectiveness of wildflower plots in promoting bee abundance. Scale (m) Term Estimate std.error z-value p-value 1000 Intercept 5.25 1.04 5.03 <0.0001* Year 0.38 0.30 1.22 0.22 % SNH -0.033 -0.014 2.32 0.02* Avg. blooms 0.018 0.014 1.62 0.11 500 Intercept 5.72 0.53 10.7 <0.0001* Year 0.36 0.25 1.45 0.14 % SNH -0.037 0.01 2.85 0.004* Avg. blooms 0.68 0.04 1.46 0.14 % SNH * Blooms -0.002 0.001 1.27 0.20 250 Intercept 4.40 0.39 10.9 <0.0001* Year 0.46 0.37 1.42 0.15 Avg. blooms 0.018 0.01 1.62 0.11 Plot area 0.0001 0.00008 1.40 0.16 * significant at p < 0.05

Table S2.2.8. Results of model selection for factors affecting the effectiveness of wildflower plots in promoting A. virecens abundance. Scale (m) Model k logLik AICc Δ AICc Weight 1000 Intercept only 4 -75.84 162.35 0 0.555† Plot area 5 -75.07 164.43 2.082 0.196† Avg. blooms 5 -75.67 165.63 3.275 0.108† % SNH 5 -75.78 165.85 3.495 0.097 Avg. blooms * Plot area 7 -72.1 167.54 5.186 0.041 Plot area * % SNH 7 -75.05 173.44 11.09 0.002 Avg. blooms * % SNH 7 -75.52 174.38 12.029 0.001 500 Intercept only 4 -75.84 162.35 0 0.523† Plot area 5 -75.07 164.43 2.082 0.185† % SNH 5 -75.3 164.89 2.537 0.147† Avg. blooms 5 -75.67 165.63 3.275 0.102 Avg. blooms * Plot area 7 -72.1 167.54 5.186 0.039 Plot area * % SNH 7 -74.67 172.68 10.328 0.003 Avg. blooms * % SNH 7 -75 173.34 10.983 0.002 250 Intercept only 4 -75.84 162.35 0 0.539† Plot area 5 -75.07 164.43 2.082 0.19† Avg. blooms 5 -75.67 165.63 3.275 0.105† % SNH 5 -75.71 165.7 3.345 0.101† Avg. blooms * Plot area 7 -72.1 167.54 5.186 0.04 Plot area * % SNH 7 -72.69 168.72 6.366 0.022 Avg. blooms * % SNH 7 -75.26 173.85 11.496 0.002 K is the number of parameters in the model † models used for averaging

Table S2.2.9. Results of model averaging for factors affecting the effectiveness of wildflower plots in promoting A. virecens abundance. Scale (m) Term Estimate std.error z-value p-value 1000 Intercept 1.39 0.78 2.48 0.013* Year 0.18 0.27 0.67 0.49 Plot area 0.00007 0.0002 0.39 0.69 Avg. blooms -0.0009 0.005 0.18 0.85 500 Intercept 2.19 1.20 1.82 0.07 Year 0.20 0.27 0.77 0.44 Plot area 0.0003 0.0002 1.18 0.24 % SNH -0.05 0.05 0.99 0.32 250 Intercept 2.00 0.84 2.39 0.02* Year 0.18 0.28 0.68 0.49 Plot area 0.0003 0.0002 1.18 0.24 Avg. blooms -0.007 0.01 0.57 0.57 % SNH -0.016 0.03 0.48 0.63 * significant at p < 0.05

Table S2.2.10. Results of model selection for factors affecting the effectiveness of wildflower plots in promoting P. pruinosa abundance. Scale (m) Model k logLik AICc Δ AICc Weight 1000 Intercept only 4 -67.05 144.76 0 0.399† Avg. blooms 5 -65.82 145.93 1.167 0.223† % SNH 5 -66.04 146.36 1.593 0.18† Plot area 5 -66.68 147.66 2.891 0.094† Avg. blooms * Plot area 7 -62.19 147.7 2.939 0.092 Plot area * % SNH 7 -64.74 152.82 8.053 0.007 Avg. blooms * % SNH 7 -65.1 153.54 8.776 0.005 500 Intercept only 4 -67.05 144.76 0 0.416† Avg. blooms 5 -65.82 145.93 1.167 0.232† % SNH 5 -66.27 146.83 2.064 0.148† Plot area 5 -66.68 147.66 2.891 0.098† Avg. blooms * Plot area 7 -62.19 147.7 2.939 0.096 Plot area * % SNH 7 -65.04 153.41 8.644 0.006 Avg. blooms * % SNH 7 -65.41 154.15 9.387 0.004 250 Intercept only 4 -67.05 144.76 0 0.4† Avg. blooms 5 -65.82 145.93 1.167 0.223† % SNH 5 -66.14 146.56 1.795 0.163† Plot area 5 -66.68 147.66 2.891 0.094† Avg. blooms * Plot area 7 -62.19 147.7 2.939 0.092 Plot area * % SNH 7 -63.78 150.9 6.138 0.019 Avg. blooms * % SNH 7 -64.52 152.37 7.601 0.009 K is the number of parameters in the model † models used for averaging

Table S2.2.11. Results of model averaging for factors affecting the effectiveness of wildflower plots in promoting P. pruinosa abundance. Scale (m) Term Estimate std.error z-value p-value 1000 Intercept 2.05 1.66 1.24 0.22 Year 0.38 0.75 0.51 0.61 Avg. blooms 0.038 0.02 1.44 0.15 Plot area -0.0002 0.0002 0.81 0.42 % SNH -0.05 0.04 1.36 0.17 500 Intercept 1.81 1.30 1.39 0.16 Year 0.38 0.75 0.53 0.59 Avg. blooms 0.38 0.02 1.44 0.15 Plot area -0.0002 0.0002 0.81 0.42 % SNH -0.05 0.04 1.21 0.22 250 Intercept 1.71 1.06 1.61 0.11 Year 0.40 0.75 0.53 0.59 Avg. blooms 0.38 0.02 1.44 0.15 Plot area -0.0002 0.0002 0.81 0.42 % SNH -0.03 0.02 1.44 0.15

Table S2.2.12. Results of model selection for factors affecting the effectiveness of wildflower plots in promoting M. bimaculatus abundance. Scale (m) Model k logLik AICc Δ AICc Weight 1000 Intercept only 4 -49.64 109.95 0 0.529† % SNH 5 -48.67 111.63 1.683 0.228† Plot area 5 -49.17 112.63 2.681 0.138† Avg. blooms 5 -49.58 113.45 3.498 0.092 Avg. blooms * Plot area 7 -47.81 118.96 9.007 0.006 Avg. blooms * % SNH 7 -48.19 119.71 9.76 0.004 Plot area * % SNH 7 -48.36 120.05 10.101 0.003 500 Intercept only 4 -49.64 109.95 0 0.554† % SNH 5 -48.93 112.15 2.197 0.185† Plot area 5 -49.17 112.63 2.681 0.145† Avg. blooms 5 -49.58 113.45 3.498 0.096 Avg. blooms * % SNH 7 -47.31 117.95 7.998 0.01 Avg. blooms * Plot area 7 -47.81 118.96 9.007 0.006 Plot area * % SNH 7 -48.45 120.22 10.272 0.003 250 Plot area * % SNH 7 -42.3 107.93 0 0.613† Intercept only 4 -49.64 109.95 2.023 0.223† % SNH 5 -49.13 112.55 4.625 0.061† Plot area 5 -49.17 112.63 4.704 0.058 Avg. blooms 5 -49.58 113.45 5.521 0.039 Avg. blooms * % SNH 7 -47.51 118.34 10.416 0.003 Avg. blooms * Plot area 7 -47.81 118.96 11.029 0.002 K is the number of parameters in the model † models used for averaging

Table S2.2.13. Results of model averaging for factors affecting the effectiveness of wildflower plots in promoting M. bimaculatus abundance. Scale (m) Term Estimate std.error z-value p-value 1000 Intercept 0.60 1.25 0.48 0.63 Year 1.82 0.57 3.19 0.001* Plot area 0.0001 0.0001 0.88 0.38 % SNH -0.04 0.03 1.35 0.18 500 Intercept 0.37 0.94 0.39 0.69 Year 1.83 0.57 3.19 0.001* Plot area 0.0001 0.0001 0.88 0.38 % SNH -0.04 0.03 1.13 0.26 250 Intercept -1.87 1.77 1.05 0.29 Year 1.87 0.51 3.69 0.0002* Plot area 0.002 0.0004 3.95 <0.0001* % SNH 0.1 0.05 1.96 0.04* Plot area * % SNH -0.00007 0.00002 3.84 0.001* * significant at p < 0.05

Chapter 3: Conservation wildflower plantings do not enhance on-farm abundance of Amblyomma americanum (Ixodida:Ixodidae)

Abstract

Planting wildflowers is a commonly suggested measure to conserve pollinators. While beneficial for pollinators, plots of wildflowers may be inadvertently performing an ecosystem disservice by providing suitable habitat for arthropod disease vectors like ticks. The lone star tick, Amblyomma americanum (L)., is a medically important tick species that might be able to utilize wildflower plantings as suitable habitat. In this two-year study, ticks were sampled with dry-ice baited traps from wildflower plots, weedy field margins, and forested areas to determine if wildflower plantings were increasing the on-farm abundance of A. americanum. Abiotic and biotic environmental variables were also measured to better understand which factors are affecting A. americanum abundance. We found no more A. americanum in wildflower plots than in weedy field margins. Forested areas harbored the greatest number of A. americanum sampled. The height of vegetation in the sampled habitats was a significant factor in determining A. americanum abundance. Depending on the sampled habitat and life stage, this relationship can be positive or negative. The relationship to vegetation height is likely related to the behavior of the white-tailed deer and the questing success of A. americanum. Overall, wildflower plots do not pose an increased risk of exposure to A. americanum on farms.

Introduction

The lone star tick, Amblyomma americanum (L). (Ixodida:Ixodidae), is an aggressively biting species that is a nuisance to humans and a pest of livestock (Semtner et al. 1971). It has been gaining attention as an important vector of human diseases including erlichiosis, tuleramia, and heartland virus (Goddard and Varela-Stokes 2009, Savage et al. 2016). More recently, A.

americanum has been implicated in triggering red meat anaphylaxis caused by the sugar galactose-훼-1,3-galactose that is injected during feeding (Commins et al. 2011). Further adding to its importance, the range of A. americanum has been expanding northward from the southeastern United States, and is likely to continue moving northward under current climate change conditions (Springer et al. 2014, 2015).

Amblyomma americanum is the most abundant tick species sampled in southeastern

Virginia (Nadolny et al. 2014). In Virginia, adult and nymph A. americanum are most active from late April to mid-July, and larvae are active from August to October (Sonenshine and Levy

1971). A. americanum requires three blood meals throughout its lifetime, one to molt from larva to nymph, another to molt from nymph to adult, and the final one to reach sexual maturity. A. americanum utilizes a wide variety of hosts like small mammals and birds, but white-tailed deer,

Odocoileus virginianus (Zimmerman), is considered the primary host (Kollars et al. 2000,

Paddock and Yabsley 2007). A. americanum primarily disperses opportunistically on hosts, but have been documented questing up to 5 meters in response to CO2 plumes in mark-recapture studies (Kensinger and Allan 2011).

A. americanum spend the majority of their life off-host, subjecting them to abiotic conditions in the environment. They protect themselves against desiccation with cuticular wax deposits to inhibit water loss, and can absorb moisture directly from the air (Semtner and Hair

1973, Needham and Teel 1991, Yoder et al. 1997). A. americanum quest during times of the day when temperatures are high and relative humidity is low (Schulze et al. 2001, Schulze and

Jordan 2003), and generally seek environments that experience low temperature variation and have high relative humidity (Randolph 2013). Forests provide such favorable conditions for A.

americanum to survive and are generally more favorable habitats compared with grasslands

(Stein et al. 2008, Mangan et al. 2018).

Habitat manipulations can affect the abundance of A. americanum by altering host behavior and microclimates. Areas where the invasive shrub Amur honeysuckle, Lonircera maackii (Rupr.) Herder, had been removed, had lower A. americanum densities than areas with the shrub due to the preference of white-tailed deer for areas with L. maackii (Allan et al. 2010).

A. americanum mortality was higher in plots with the invasive Japanese stiltgrass, Microstegium vimineum (Trin.) compared to plots without the plant (Civitello et al. 2008); plots with M. vimineum had higher temperatures and lower humidities than control plots (Civitello et al. 2008).

The westward expansion of the eastern red cedar, Juniperus virginiana L., in Oklahoma is believed to be facilitating a similar westward expansion of A. americanum by providing both better environmental conditions for the tick and white-tailed deer (Noden and Dubie 2017).

Concurrent with concerns about the changes in tick vectored diseases are concerns about pollinator declines. One of the primary drivers of pollinator decline is habitat loss, in conjunction with pesticide exposure and diseases (Potts et al. 2010). One mitigation strategy is the planting of wildflower plots to provide resources for bees (Venturini et al. 2017). Wildflower plots can increase the on-farm diversity and abundance of the bee species (Venturini et al. 2017). These plots are often planted in unmanaged areas of farms. The installation of these plots is subsidized by government programs, like the Environmental Quality Incentives Program in the United

States (Vaughn and Skinner 2015). From 2009 - 2018, this program has helped pay for habitat management that is beneficial for pollinators on over 16,000,000 acres of land (Vaughn and

Skinner 2015, USDA 2018).

100

However, it is unknown whether wildflower plots provide favorable habitats for ticks and could potentially increase the risk of exposure to people and animals to the lone star tick.

Wildflower plantings can replace unmanaged field margins with potentially more favorable habitat that is more suitable for A. americanum. Areas where the invasive shrub Japenese barberry, Berberis thundbergii de Candolle, has established had more favorable microclimates for ticks compared to control plots (Williams and Ward 2010). Furthermore, the management of these plantings requires that they be mowed each year during the dormant season (Glennon

2015). This annual mowing has the potential build-up a duff layer that provides a critical microclimate that ticks needs to survive. Given that the removal of plants can reduce A. americanum populations (Allan et al. 2010, Noden and Dubie 2017), does the addition of plants aid A. americanum populations? The purpose of this study was to determine if on-farm wildflower plots can serve as quality habitat for the lone star tick.

Materials and Methods

Study Area

Tick surveys were conducted at 10 farms in eastern Virginia and Maryland in 2018, and nine of the same farms in 2019. Nine of these farms had wildflower plots installed during the spring of 2016; one wildflower plot was seeded in the spring of 2015. Wildflower plot size ranged from 561 – 8,600 m2, with an average size of 2,360 m2. For details on the plant mixes and establishment of the wildflower plots, see Angelella and O’Rourke (2017).

Sampling

Ticks were sampled with dry-ice traps. Traps were 8-liter coolers, measuring 33 cm × 24 cm × 22 cm (Igloo Coolers, Katy, Texas) with 13 mm diameter holes drilled into each side.

101

Traps were loaded with 2 kg of dry ice. A 5-cm band of Shurtape Indoor/Outdoor tape was placed around the outside of the trap (Hickory, NC). Three habitats were sampled at each farm: the wildflower plot, a weedy field margin, and nearby forest. One dry-ice trap was placed in each sampling habitat at each location on each sampling date; traps were placed at least 10 m apart.

They were set between 9:00 am and 12:00 pm, and left in the field for 24 hrs. Each field was sampled once per month from April to July in 2018 and 2019. All ticks caught were placed in

95% ethanol and later identified as nymphs or adults.

Environmental Variables

Temperature and relative humidity at the soil surface were recorded every half-hour for

24 hrs in each habitat at each field site during each round of sampling using a Hobo U23 Pro V2 data logger (Onset; Bourne, MA). Vegetation height was measured during each sampling date.

The height of herbaceous vegetation was measured at the trap location and 4 m away from the trap in each cardinal direction (Smith 2008). Duff depth was measured from these same locations after the last round of sampling as the distance from the bare soil to the top of the organic matter on the ground.

Statistical Analysis

To determine if wildflower plots increased lone star tick abundance relative to weedy field margins, a generalized linear mixed model fit to a negative binomial distribution was used.

The interaction of sampling habitat and tick life stage and their main effects were fixed factors.

Field and sampling date nested in field were random effects. Tukey’s Honestly Significant

Difference (HSD) was used to test for differences among means. The analyses were done in ‘R’ version 3.5.2. (R Core Team 2018). The package ‘glmmADMB’ was used to test the generalized

102

linear mixed models (Skaug et al. 2016). The package ‘emmeans’ was used for multiple comparison (Lenth 2018).

To investigate the effects of the measured environmental variables on tick abundance, a multimodel approach was used. A set of 10 a priori models was created with environmental variables that are commonly associated with A. americanum. The variables tested were: habitat, duff depth, average vegetation height, average relative humidity, and temperature standard deviation. Each variable was tested alone along with their main effects and interaction with habitat. An intercept only model was included as a null model. This was done separately for nymphs and adults. Generalized linear mixed models with a negative binomial distribution were used. Field was treated as a random effect, with year and sampling date nested within field.

Models within 4 AICc points of the top model were selected and averaged (Harrison et al. 2018).

To reduce the likelihood of including uninformative parameters, the changes in AICc values were compared relative to changes in log likelihood values (Leroux 2019). These analyses were done separately for nymphs and adults, as they have different responses to the environment (Van

Horn et al. 2018). The package ‘MuMIn’ was used for model selection and averaging (Barton

2018).

Results We collected 1,165 nymphs and 566 adult A. americanum over the two years of sampling. A. americanum nymph abundance peaked in June of both years (Fig. 3.1.). A peak in adult abundance was seen in May in 2018, and April in 2019 (Fig. 3.1.) On average, more nymphs than adults were sampled in 2018 (z = 2.91, p = 0.003). This effect was not statistically significant in 2019 (z = 1.85, p = 0.06) (Fig. 3.1.).

103

Of the total 1,731 A. americanum sampled, 164 were taken from wildflower plots, 302 from weedy field margins, and 1,265 from forest locations. Forests consistently had the thickest duff layers, shortest vegetation, and most stable temperatures of the three habitat types sampled

(Table 3.1). No interaction between life stage and habitat was detected either year of study

(Table 3.2.). There was no difference in A. americanum abundance between wildflower plots and weedy field margins in 2018 and 2019 (Fig. 3.2.). A. americanum nymphs and adults were sampled most often from forest plots (Fig. 3.2.), with significantly fewer A. americanum collected in wildflower than forest plots in both years (z = -5.23, p < 0.0001; z = -3.89, p =

0.003).

The model with the interaction of habitat type and vegetation height was the top model for predicting adult A. americanum abundance (Table 3.3). The models containing habitat only and the interaction between habitat and duff depth were within 4 AICc points of the top model and included in model averaging (Table 3.3). Vegetation height had a significant different effect in weedy field margins compared to the other habitats. As the height of vegetation increased in weedy field margins, the abundance of A. americanum decreased (z = 2.4, p = 0.02) (Table 3.4).

No effects were detected for duff depth (Table 3.4).

Similar to adult A. americanum, the habitat by vegetation height model was the best predictor of nymph abundance. All other models were more than 4 AICc points higher (Table

3.5). As vegetation height increased in forest samples, so did A. americanum nymph abundance

(z = 3.4, p = 0.001) (Table 3.6). This contrasts to wildflower plots and weedy field margins, where nymph abundance decreases with increasing vegetation height (z = -2.08, p = 0.04) (Table

3.6).

104

Discussion

In this study, wildflower plots planted for pollinator conservation did not inadvertently constitute an ecosystem disservice by increasing A. americanum abundance. While A. americanum were sampled from wildflower plots, they harbored fewer A. americanum than weedy field margins. Therefore, wildflower plots do not pose an increased risk of augmenting on-farm A. americanum abundance. Similar to how the wildflower plots could potentially increase A. americanum abundance, so could leaf litter dumped in forested habitat after removal from residential lawns. However, dumping leaf litter removed from lawns did not increase A. americanum abundance relative to forested plots (Jordan and Schulze 2020). To our knowledge, this is the first study looking at the specific interaction of wildflower plots and tick abundance.

Vegetation height is playing a role in the differences in the adult A. americanum abundance between the sampling habitats. These changes could be related to the questing success of A. americanum adults on taller vegetation. Adult A. americanum could have had higher success rates in finding a host in weedy field margins compared to the other habitats sampled.

With taller vegetation, adult A. americanum would have more area to utilize to try and attach to larger hosts like white-tailed deer. Many of the weedy field margins sampled were a transition zone from agricultural areas to the forested ones. These transition areas are frequented by white- tailed deer as they move from areas of cover to open areas as part of their diurnal movement

(Beier and McCullough 1990). With greater success in finding hosts, fewer A. americanum would be available to sample.

A. americanum nymph abundance decreased with taller vegetation in wildflower plots and weedy field margins, but increased with taller vegetation in forested areas. Decreases in nymph abundance could be following a similar pattern as adults; as vegetation height increases

105

so did questing success. The increase of A. americanum nymph abundance with increasing vegetation height in forested areas could be explained by the preference of white-tailed deer to use dense vegetation in forests for bedding sites (Beier and McCullough 1990). With female white-tailed deer having a strong preference for their home range, they could be frequenting these bedding areas giving A. americanum larvae easy access to their host and creating hotspots of new molted nymphs. Larger wildflower plots could potentially provide enough cover for deer to use as bedding sites, creating a similar situation. However, given the smaller size of the plots used in this study this likely was not occurring. A study with Peromyscus mice found that 64 % of the nests surveyed had Ixodes scapularis present, and 87% of all larval ticks present had taken a blood meal (Larson et al. 2020). Specific sites that hosts remain immobile in and frequently visit can be attractive for immature stages of ticks.

Duff depth was selected as a factor for adult A. americanum, but was not significant with model averaging. A similar result was seen in Missouri, where duff layer depth was selected as a factor, but not a significant in determining adult A. americanum abundance (Van Horn et al.

2018). However, duff depth is an important environmental factor for the survival of adult A. americanum, as it a critical microclimate for preventing desiccation. A. americanum adults are sensitive to moisture loss, and will seek out moist microclimates after losing only 10 - 15 % of their body weight to counteract desiccation (Hair et al. 1975). While the duff layer is important, its presence may be all that matters. A. americanum were collected from areas that had shallower duff layers than I. scapularus, but never from areas that had no duff layer (Schulze et al. 2002).

Better quantification of the microclimate of the duff layer may also help detect effects on A. americanum abundance, as the duff layer can be 2 - 3° C cooler than the ambient air temperature

(Schulze and Jordan 2003).

106

The result that wildflower plots are not increasing on-farm A. americanum is encouraging for both pollinator conservation and vector control, further studies are needed to verify the results in other geographic areas and with different mixes of wildflowers. The mix of wildflower species in a pollinator habitat may have a large influence on the behaviors of tick hosts. A study in Florida observed that white-tailed deer browsed on all 11 wildflower species used in their pollinator mixes (DeGroote et al. 2011). Of the species tested by Degroote et al.(2011), two species were present in this study: Coreopsis lanceolata L. and Rudbeckia hirta L., which were the least and 5th least browsed species, respectively. If wildflower mixes have species that are attractive to white-tailed deer, tick abundance is likely to increase.

Different tick species may also have varied levels of attraction to the same habitat.

Within a 1-ha forested plot, A. americanum and Ixodes scapularis Say were distributed between two different sets of habitat conditions related to each species’ tolerance to desiccation (Schulze et al. 2002). A. americanum was found in areas with a more open canopy and less shrubby understory compared to I. scapularis (Schulze et al. 2002). If dense stands of ground cover develop within wildflower plots, they could attract rodent hosts of I. scapularis (Machtinger and

Li 2019), potentially increasing I. scapularis abundance. This could be more problematic if conservation efforts are focused on the use of longer-lived woody plants instead of perennial and annual wildflowers.

A. americanum were sampled from within wildflower plots, but they are not increasing the risk of exposure to A. americanum relative to weedy field margins. The role of hosts in moving A. americanum into the wildflower plots is an important factor in the success of A. americanum colonizing these habitats and deserves further study. Understanding how the hosts

107

of A. americanum utilize wildflower plots for cover and their preferences for different wildflower species as food could further inform the risks posed by wildflower plots.

Contributions: CTM provided the conceptual framework. MEO improved the experimental design. GA and CTM collected the data. CTM wrote the original draft of the manuscript. MEO and GA provided revisions for the manuscript.

108

References Allan, B. F., H. P. Dutra, L. S. Goessling, K. Barnett, J. M. Chase, R. J. Marquis, G. Pang, G. A. Storch, R. E. Thach, and J. L. Orrock. 2010. Invasive honeysuckle eradication reduces tick-borne disease risk by altering host dynamics. Proc. Natl. Acad. Sci. 107: 18523–18527. Angelella, G. M., and M. E. O’Rourke. 2017. Pollinator Habitat Establishment after Organic and No-till Seedbed Preparation Methods. HortScience. 52: 1349–1355. Barton, K. 2018. MuMIn: Multi-Model Inference. Beier, P., and D. R. McCullough. 1990. Factors Influencing White-Tailed Deer Activity Patterns and Habitat Use. Wildl. Monogr. 3–51. Civitello, D. J., S. L. Flory, and K. Clay. 2008. Exotic Grass Invasion Reduces Survival of Amblyomma americanum and Dermacentor variabilis Ticks (Acari: Ixodidae). J. Med. Entomol. 45: 867–872. Commins, S. P., H. R. James, L. A. Kelly, S. L. Pochan, L. J. Workman, M. S. Perzanowski, K. M. Kocan, J. V. Fahy, L. W. Nganga, E. Ronmark, P. J. Cooper, and T. A. E. Platts-Mills. 2011. The relevance of tick bites to the production of IgE antibodies to the mammalian oligosaccharide galactose-α-1,3-galactose. J. Allergy Clin. Immunol. 127: 1286-1293.e6. DeGroote, L. W., H. K. Ober, J. H. Aldrich, J. G. Norcini, and G. W. Knox. 2011. Susceptibility of Cultivated Native Wildflowers to Deer Damage. Southeast. Nat. 10: 761–771. Glennon, R. 2015. Habitat Establishment and Mangement Recommendations for Pollinator Habitat on Well-Drained Soil. USDA Natural Resources Conservation Service. Goddard, J., and A. S. Varela-Stokes. 2009. Role of the lone star tick, Amblyomma americanum (L.), in human and animal diseases. Vet. Parasitol. 160: 1–12. Hair, J. A., J. R. Sauer, and K. A. Durham. 1975. Water Balance and Humidity Preference in Three Species of Ticks. J. Med. Entomol. 12: 37–47. Harrison, X. A., L. Donaldson, M. E. Correa-Cano, J. Evans, D. N. Fisher, C. E. D. Goodwin, B. S. Robinson, D. J. Hodgson, and R. Inger. 2018. A brief introduction to mixed effects modelling and multi-model inference in ecology. PeerJ. 6: e4794. Jordan, R. A., and T. L. Schulze. 2020. Artificial Accumulation of Leaf Litter in Forest Edges on Residential Properties via Leaf Blowing Is Associated with Increased Numbers of Host-Seeking Ixodes scapularis (Acari: Ixodidae) Nymphs. J. Med. Entomol. Kensinger, B. J., and B. F. Allan. 2011. Efficacy of Dry Ice-Baited Traps for Sampling Amblyomma americanum (Acari: Ixodidae) Varies with Life Stage but Not Habitat. J. Med. Entomol. 48: 708–711. Kollars, T. M., J. H. Oliver, L. A. Durden, and P. G. Kollars. 2000. Host Associations and Seasonal Activity of Amblyomma americanum (Acari: Ixodidae) in Missouri. J. Parasitol. 86: 1156–1159. Larson, R. T., X. Lee, T. Zembsch, G. M. Bron, and S. M. Paskewitz. 2020. Immature Ixodes scapularis (Acari: Ixodidae) Collected from Peromyscus leucopus (Rodentia: Cricetidae) and Peromyscus maniculatus (Rodentia: Cricetidae) Nests in Northern Wisconsin. J. Med. Entomol. 57: 304–307. Lenth, R. 2018. emmeans: Estimated Marginal Means, aka Least-Squares Means. Leroux, S. J. 2019. On the prevalence of uninformative parameters in statistical models applying model selection in applied ecology. PLOS ONE. 14: e0206711.

109

Machtinger, E. T., and A. Y. Li. 2019. Tick control bait box use by Peromyscus spp. influenced by habitat placement but raises questions on disease ecology. Ecosphere. 10. Mangan, M. J., S. A. Foré, and H.-J. Kim. 2018. Ecological modeling over seven years to describe the number of host-seeking Amblyomma americanum in each life stage in northeast Missouri. J. Vector Ecol. 43: 15. Nadolny, R. M., C. L. Wright, D. E. Sonenshine, W. L. Hynes, and H. D. Gaff. 2014. Ticks and spotted fever group rickettsiae of southeastern Virginia. Ticks Tick-Borne Dis. 5: 53–57. Needham, G. R., and P. D. Teel. 1991. Off-Host Physiological Ecology of Ixodid Ticks. Annu. Rev. Entomol. 36: 659–681. Noden, B. H., and T. Dubie. 2017. Involvement of invasive eastern red cedar (Juniperus virginiana) in the expansion of Amblyomma americanum in Oklahoma. J. Vector Ecol. 42: 178–183. Paddock, C. D., and M. J. Yabsley. 2007. Ecological Havoc, the Rise of White-Tailed Deer, and the Emergence of Amblyomma americanum-Associated Zoonoses in the United States, pp. 289–324. In Childs, S.R.S.J.E., Mackenzie, P.J.S., Richt, V.M.O.J.A. (eds.), Wildl. Emerg. Zoonotic Dis. Biol. Circumst. Consequences Cross-Species Transm., Current Topics in Microbiology and Immunology. Springer Berlin Heidelberg. Potts, S. G., J. C. Biesmeijer, C. Kremen, P. Neumann, O. Schweiger, and W. E. Kunin. 2010. Global pollinator declines: trends, impacts and drivers. Trends Ecol. Evol. 25: 345– 353. R Core Team. 2018. A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Randolph, S. 2013. Ecology of Non-nidicolous Ticks, pp. 3–38. In Biol. Ticks. Oxford University Press. Savage, H. M., M. S. Godsey, N. A. Panella, K. L. Burkhalter, D. C. Ashley, R. R. Lash, B. Ramsay, T. Patterson, and W. L. Nicholson. 2016. Surveillance for Heartland Virus (Bunyaviridae: Phlebovirus ) in Missouri During 2013: First Detection of Virus in Adults of Amblyomma americanum (Acari: Ixodidae). J. Med. Entomol. 53: 607–612. Schulze, T. L., and R. A. Jordan. 2003. Meteorologically Mediated Diurnal Questing of Ixodes scapularis and Amblyomma americanum (Acari: Ixodidae) Nymphs. J. Med. Entomol. 40: 395–402. Schulze, T. L., R. A. Jordan, and R. W. Hung. 2001. Effects of Selected Meteorological Factors on Diurnal Questing of Ixodes scapularis and Amblyomma americanum (Acari: Ixodidae). J. Med. Entomol. 38: 318–324. Schulze, T. L., R. A. Jordan, and R. W. Hung. 2002. Effects of Microscale Habitat Physiognomy on the Focal Distribution of Ixodes scapularis and Amblyomma americanum (Acari: Ixodidae) Nymphs. Environ. Entomol. 31: 1085–1090. Semtner, P. J., and J. A. Hair. 1973. The Ecology and Behavior of the Lone Star Tick (Acarina: Ixodidae) V. Abundance and seasonal distribution in different habitat types. J. Med. Entomol. 10: 618–628. Semtner, P. J., D. E. Howell, and J. A. Hair. 1971. The Ecology and Behavior of the Lone Star Tick (Acarina: Ixodidae) I. The relationship between vegetative habitat type and tick abundance and distribution in Cherokee Co., Oklahoma. J. Med. Entomol. 8: 329–335. Skaug, H., D. Fournier, A. Nielsen, A. Magnusson, and B. Bolker. 2016. Generalized Linear Mixed Models using “AD Model Builder.”

110

Smith, M. A. 2008. Robel pole technique and data interpretation ( No. MP-110.10). University of Wyoming Extension. Sonenshine, D. E., and G. F. Levy. 1971. The Ecology of the Lone Star Tick, Amblyomma Americanum (L.), In Two Contrasting Habitats in Virginia (Acarina: Ixodidae). J. Med. Entomol. 8: 623–635. Springer, Y. P., L. Eisen, L. Beati, A. M. James, and R. J. Eisen. 2014. Spatial Distribution of Counties in the Continental United States with Records of Occurrence of Amblyomma americanum (Ixodida: Ixodidae). J. Med. Entomol. 51: 342–351. Springer, Y. P., C. S. Jarnevich, A. J. Monaghan, R. J. Eisen, and D. T. Barnett. 2015. Modeling the Present and Future Geographic Distribution of the Lone Star Tick, Amblyomma americanum (Ixodida: Ixodidae), in the Continental United States. Am. J. Trop. Med. Hyg. 93: 875–890. Stein, K. J., M. Waterman, and J. L. Waldon. 2008. The effects of vegetation density and habitat disturbance on the spatial distribution of ixodid ticks (Acari: Ixodidae). Geospatial Health. 2: 241–252. USDA. 2018. Environmental Quality Incentives Program (EQIP) | Farm Bill Report (FY 2009 through FY 2018) | NRCS. (https://www.nrcs.usda.gov/Internet/NRCS_RCA/reports/fb08_cp_eqip.html). Van Horn, T. R., S. A. Adalsteinsson, K. M. Westby, E. Biro, J. A. Myers, M. J. Spasojevic, M. Walton, and K. A. Medley. 2018. Landscape Physiognomy Influences Abundance of the Lone Star Tick, Amblyomma americanum (Ixodida: Ixodidae), in Ozark Forests. J. Med. Entomol. 55: 982–988. Vaughn, M., and M. Skinner. 2015. Using 2014 Farm Bill Programs for Pollinator Conservation. 18. Venturini, E. M., F. A. Drummond, A. K. Hoshide, A. C. Dibble, and L. B. Stack. 2017. Pollination reservoirs for wild bee habitat enhancement in cropping systems: a review. Agroecol. Sustain. Food Syst. 41: 101–142. Williams, S. C., and J. S. Ward. 2010. Effects of Japanese Barberry (Ranunculales: Berberidaceae) Removal and Resulting Microclimatic Changes on Ixodes scapularis (Acari: Ixodidae) Abundances in Connecticut, USA. Environ. Entomol. 39: 1911–1921. Yoder, J. A., M. E. Selim, and G. R. Needham. 1997. Impact of feeding, molting and relative humidity on cuticular wax deposition and water loss in the lone star tick, Amblyomma americanum. J. Insect Physiol. 43: 547–551.

111

Figure 3.1. Average number of A. americanum adults and nymphs sampled during each week of sampling in 2018 (A) and 2019 (B). The yearly average abundance of adults and nymphs sampled in 2018 (C) and 2019 (D). Means with the same letter are not significantly different. (Tukey’s Honestly Significant Difference) (p < 0.05).

112

Figure 3.2. Average number of A. americanum sampled from each sampling habit in 2018 (A) and 2019 (B). Means with the same letter are not significantly different. (Tukey’s Honestly Significant Difference) (p < 0.05)

113

Table 3.1. Yearly averages (mean ± std. error) of environmental factors measured at each sampling location. Year Habitat Veg. height Duff depth RH Temp.SD (dm) (cm) (ºC) 2018 Wildflower 4.5 ± 0.5 0.62 ± 0.1 86.1 ± 1.0 5.8 ± 0.3 Weedy margin 4.2 ± 0.4 1.56 ± 0.1 85.4 ± 1.1 6.1 ± 0.3 Forest 2.2 ± 0.4 3.38 ± 0.2 85.7 ± 1.3 3.4 ± 0.2 2019 Wildflower 6.7 ± 0.6 1.09 ± 0.1 84.5 ± 1.1 6.3 ± 0.4 Weedy margin 5.1 ± 0.5 1.27 ± 0.1 85.1 ± 1.0 6.3 ± 0.4 Forest 1.2 ± 0.1 3.87 ± 0.1 82.2 ± 1.4 3.7 ± 0.2

114

Table 3.2. Results of generalized linear mixed models testing the interaction of life stage and habitat on the abundance of A. americanum. Adult abundance at wildflower plots is the reference category. Year Term Estimate Std.err Z value P > z 2018 Intercept -2.05 0.58 -3.55 0.0003* Weedy margin 0.95 0.44 2.19 0.03* Forest 2.71 0.40 6.79 <0.0001* Life stage 1.21 0.42 2.91 0.004* Weedy margin: Life stage -0.33 0.54 -0.61 0.54 Forest: Life stage -0.94 0.51 -1.86 0.06

2019 Intercept -1.63 0.58 -2.79 0.005* Weedy margin 1.13 0.48 2.34 0.02* Forest 2.09 0.45 4.62 <0.0001* Life stage 0.91 0.49 1.85 0.06 Weedy margin: Life stage -0.97 0.64 -1.49 0.13 Forest: Life stage 0.001 0.63 0.00 0.99 * significant at p < 0.05

115

Table 3.3. Model selection for environmental variables affecting the abundance of A. americanum adults. Models with interaction terms also include the main effects. Model k logLik AICc Δ AICc Weight Habitat * Vegetation Height 10 -346.94 714.88 - 0.539† Habitat 7 -350.89 716.29 1.407 0.266† Habitat * Duff Depth 10 -348.75 718.51 3.628 0.088† Duff depth 6 -353.46 719.3 4.416 0.059 Habitat * Temperature SD 10 -349.91 720.83 5.944 0.028 Habitat * Relative Humidity 10 -350.21 721.44 6.552 0.02 Vegetation Height 6 -373.38 759.13 44.25 0 Temperature SD 6 -377 766.39 51.504 0 Relative Humidity 6 -382.34 777.06 62.172 0 Intercept only 5 -383.59 777.45 62.564 0 † model selected for averaging k is the number of parameters in the model

116

Table 3.4. Parameter estimates of model averaging using selected models (habitat, habitat * vegetation height, and habitat * duff depth) of environmental variables affecting the abundance of A. americanum adults. Term Estimate Adjusted SE Z value P > z Intercept -1.93 0.64 3.00 0.003* Weedy Margin 1.48 0.83 1.78 0.08 Forest 2.25 0.69 3.28 0.001* Vegetation Height 0.05 0.07 0.80 0.42 Vegetation Height * Weedy Margin -0.26 0.11 2.35 0.019* Vegetation Height * Forest 0.07 0.14 0.49 0.62 Duff depth -0.26 0.42 0.63 0.53 Duff depth * Weedy Margin 0.63 0.55 1.16 0.25 Duff depth * Forest 0.61 0.47 1.29 0.20 * significant at p < 0.05

117

Table 3.5. Model selection for environmental variables affecting the abundance of A. americanum nymphs. Models with interaction terms also include the main effects. Model k logLik AICc Δ AICc Weight Habitat * Vegetation Height 10 -422.78 866.57 - 0.985† Habitat 7 -430.55 875.6 9.031 0.011 Habitat * Duff Depth 10 -429.32 879.65 13.078 0.001 Habitat * Relative Humidity 10 -429.41 879.83 13.258 0.001 Habitat * Temperature SD 10 -429.42 879.84 13.272 0.001 Duff depth 6 -436.94 886.26 19.688 0 Vegetation Height 6 -445.33 903.03 36.46 0 Temperature SD 6 -448.95 910.29 43.718 0 Relative Humidity 6 -452.4 917.18 50.612 0 Intercept only 5 -454.4 919.07 52.502 0 † model selected for averaging k is the number of parameters in the model.

118

Table 3.6. Parameter estimates for the model used to analyze the effects of habitat and vegetation height on the abundance of A. americanum nymphs. Term Estimate Std. Error Z value P > z Intercept -0.55 0.77 -0.71 0.48 Weedy Margin 0.70 0.58 1.22 0.22 Forest 0.77 0.48 1.61 0.11 Vegetation Height -0.13 0.06 -2.08 0.04* Vegetation Height * Weedy Margin -0.03 0.11 -0.23 0.82 Vegetation Height * Forest 0.54 0.16 3.37 0.001* * significant at p < 0.05

119

Conclusion

This research investigated the effects of enhancing on-farm habitat diversity and the amount of semi-natural habitat (SNH) in that landscape at different scales on a variety of ecosystem services. The amount of SNH surrounding the farms was the predominant factor affecting ecosystem services and bee communities within the study area. Fewer wildflower models were selected and fewer significant effects were detected than landscape effects (Figure

C.1). Overall, each scale of SNH in the landscape was analyzed roughly the same number of times across different ecosystem services (Figure C.1). SNH at each scale had a positive effect of on some aspect of the ecosystem services measured. For example, the 1,000-m scale had a positive effect on bee abundance, the 500-m scale showed reduction in pest damage to tomatoes, and the 250-m scale had a positive effect on bee species richness. However, when selecting the best scale for analysis, the 250-m was selected the most often (Figure C.1). The most impactful and consistent result was that yield quality improved for all four crops at the 250-m scale. This finding may be key to helping convince private landowners of the benefits of habitat conservation in the United States, as the majority of the land in the United States is privately owned (Wunderlich 1978, Chaplin-Kramer et al. 2019).

When considering wildflower plots by themselves, few effects were found. This could be related to size and quality of the plots, their proximity to SNH, and the amount of SNH in the landscape (Scheper et al. 2013, Krimmer et al. 2019). The wildflower plots may not have provided enough resources to draw beneficial arthropods out of other habitats to them. When significant wildflower effects were detected, there was no consistent trend towards being beneficial or not. For example, fewer Manduca spp. were found at wildflower fields, but a higher proportion of strawberries were infested with Drosophila spp. larvae. Notably, the wildflower

120

plots did not enhance on-farm population of Amblyomma americanum, the lone star tick. Further study on the dynamics of the plant communities within wildflower plots could lead to better management recommendations for landowners to increase the effectiveness of these plots.

A prominent theme that emerges for future research is the geographic scale of study of the effects of SNH on ecosystem services. Within chapters 1 and 2, different ecosystem service metrics consistently responded to the SNH at the 250-m scale. The inclusion of the 250-m scale in all landscape only model indicates that nearby SNH has the strongest impact on the ecosystem services measured. This strong selection for smaller landscape scales could highlight the important role that smaller arthropods with relatively low dispersal abilities like, mites, thrips, micro-hymenopteran parasitoids, or Lassioglosum spp. bees, play in providing ecosystem services. It may also imply that to effectively incorporate habitat diversity on agricultural lands, larger fields may need to be broken up to maximize the benefits that were detected at the 250-m scale. Scaling up the size of the farms used would be important, as the average size of the farm in this study was 29 ha. Pywell et al. (2015) provides an intermediate step, showing positive yield gains on 50-60 ha plots. However, the average farm size in the United States is 179 ha, six-times the average size used in this study (USDA 2019).

The consistent yield response to SNH, but inconsistent effects of the other ecosystem service measures implies there are other processes occurring that were not captured by the metrics used in this study. High quality yield can be reflective of other regulating ecosystem services, and it would require a thorough study to capture all of the ecosystem services that may be contributing to yield. While tomatoes do not require pollinators, they benefit from the buzz pollination provided by bees, particularly Bombus spp. bees (Greenleaf and Kremen 2006).

Strawberries have a diverse pest and natural enemy community that is responsive to wildflower

121

plots and SNH (Grab et al. 2018). Other animals like birds or mammals may also provide ecosystem services that may go unaccounted for or be difficult to measure (Lindell et al. 2018,

Nyffeler et al. 2018). Ecologists may continue to study all the interactions that contribute to yield, but they also need to start measuring yield as well. Yield is economically important and relevant to many stakeholders, and serves to integrate an array of ecosystem services over the course of the growing season (Chaplin-Kramer et al. 2019).

122

References Chaplin-Kramer, R., M. O’Rourke, N. Schellhorn, W. Zhang, B. E. Robinson, C. Gratton, J. A. Rosenheim, T. Tscharntke, and D. S. Karp. 2019. Measuring what matters: actionable information for conservation biocontrol in multifunctional landscapes. Front. Sustain. Food Syst. 3: 60. Grab, H., K. Poveda, B. Danforth, and G. Loeb. 2018. Landscape context shifts the balance of costs and benefits from wildflower borders on multiple ecosystem services. Proc R Soc B. 285: 20181102. Greenleaf, S. S., and C. Kremen. 2006. Wild bee species increase tomato production and respond differently to surrounding land use in Northern California. Biol. Conserv. 133: 81–87. Krimmer, E., E. A. Martin, J. Krauss, A. Holzschuh, and I. Steffan-Dewenter. 2019. Size, age and surrounding semi-natural habitats modulate the effectiveness of flower-rich agri- environment schemes to promote pollinator visitation in crop fields. Agric. Ecosyst. Environ. 284: 106590. Lindell, C., R. A. Eaton, P. H. Howard, S. M. Roels, and M. E. Shave. 2018. Enhancing agricultural landscapes to increase crop pest reduction by vertebrates. Agric. Ecosyst. Environ. 257: 1–11. Nyffeler, M., Ç. H. Şekercioğlu, and C. J. Whelan. 2018. Insectivorous birds consume an estimated 400–500 million tons of prey annually. Sci. Nat. 105: 47. Pywell, R. F., M. S. Heard, B. A. Woodcock, S. Hinsley, L. Ridding, M. Nowakowski, and J. M. Bullock. 2015. Wildlife-friendly farming increases crop yield: evidence for ecological intensification. Proc R Soc B. 282: 20151740. Scheper, J., A. Holzschuh, M. Kuussaari, S. G. Potts, M. Rundlöf, H. G. Smith, and D. Kleijn. 2013. Environmental factors driving the effectiveness of European agri- environmental measures in mitigating pollinator loss – a meta-analysis. Ecol. Lett. 16: 912–920. USDA. 2019. Farms and Land in Farms 2018 Summary. United States Department of Agriculture. Wunderlich, G. 1978. Facts about U.S. landownership (Agriculture Information Bulletin No. 422). United States Department of Agriculture.

123

Figure C.1. Summary of the number and direction of the effects of all models analyzed by scale and response variable. Panel A shows the number of final models analyzed by landscape, wildflower, or the interaction of the two scales. Panel B shows the number of times each scale of the landscape only models were analyzed. Panel C shows the number of times a positive (+), negative (-), or no response (0) was seen in all the models analyzed by broad grouping of ecosystem services. Bees is the aggregate response for abundance, species richness, Shannon-Wiener Index, and PERMANOVA for the wildflower only and interaction analyses. Biocontrol is the response of egg predation for Trichopulsia ni, Helicoverpa zea, Halyomorpha halys. Pest regulation includes abundance of the dominant pests observed and pest damage to crops. The dominant pests observed were Plutella xylostella, Pieris rapae, Spodoptera ornithogalli, and Manduca spp. Pest damage included: Lygus spp., Drosophila spp., and sap beetles on strawberries; Melitta curcurbitae, Anasa tristis, and Diabrotica spp. beetles on squash; chewing damage on collards; and piercing-sucking and chewing damage to tomatoes. A positive response in this category indicates that fewer pests were observed or that pest damage was reduced. Pollination services category is composed of the pollination of strawberry achenes, the proportion of pollen deficient strawberries, and the number of squash seeds produced. Ticks is the abundance of ticks sampled from wildflower plots. Yield entails all the metrics used to measure yield for collards, tomatoes, strawberries, and squash. The metrics measured were total mass, weight of grade 1, proportion of the number of grade 1 harvested fruit or leaves. 124