Do Landscape Properties Mediate the Response of Herbivorous Pest and Natural Enemy Arthropods to Local Properties? Mst. Salma Ak

Do landscape properties mediate the response of herbivorous pest and natural enemy arthropods to local properties?

Mst. Salma Akter

A thesis submitted for the fulfilment of the requirements for the degree of Master of Philosophy

Faculty of Science, School of Agricultural and Wine Sciences Charles Sturt University

November 2020

Table of Contents

Table of Contents ...... i Certificate of Authorship ...... iii Acknowledgements ...... iv List of Figures ...... v List of Tables ...... viii List of Abbreviations ...... ix Publications prepared from this thesis ...... x Thesis structure ...... xi Thesis abstract ...... xii Chapter 1. Introduction and literature review ...... 1 1.1 Introduction ...... 1 1.2 Management of brassica vegetable pests ...... 2 1.3 Influence of local-scale factors on the arthropod community in the agroecosystem ...... 3 1.4 Significance of landscape factors on the arthropod community in the agroecosystem ...... 5 1.5 Relationship between local and landscape factors ...... 11 1.6 Research objectives ...... 12 1.7 Research Questions ...... 12 Chapter 2. Landscape context mediates the effects of local vegetation on in- field abundance of pests and natural enemies ...... 13 2.1 Introduction ...... 13 2.2 Materials and methods ...... 14 2.2.1 Sites: ...... 14 2.2.2 Arthropod data: ...... 15 2.2.3 Classification of adjacent land-uses: ...... 16 2.2.4 Measurement of landscape-scale properties: ...... 16 2.2.5 Statistical analyses: ...... 17 2.3 Results ...... 18 2.3.1 Arthropod abundance ...... 18 2.3.2 Effects of adjacent land-use on in-field arthropod abundance at field margin ...... 19 2.3.3 Effects of landscape properties on in-field arthropod abundance at field centre ...... 20 2.3.4 Influence of landscape properties on the adjacent land-use...... 22

2.4 Discussion ...... 27 2.4.1 Distinct effects of adjacent land-use on in-field arthropods at field margin ...... 28 2.4.2 Distinct effects of landscape properties on in-field arthropods at field centre ...... 29 2.4.3 Influence of landscape properties on adjacent land-use ...... 30 2.5 Conclusions ...... 32 Chapter 3. Landscape context affects the scope for local-scale habitat management to mediate in-field pest and natural enemy densities ...... 34 3.1 Introduction ...... 34 3.2 Materials and methods ...... 36 3.2.1 Sites: ...... 36 3.2.2 Floral resources: ...... 37 3.2.3 Experimental design: ...... 38 3.2.4 Arthropod sampling: ...... 38 3.2.5 Analysis of landscape factors: ...... 39 3.2.6 Statistical analyses: ...... 40 3.3 Results ...... 42 3.3.1 Abundance of arthropods ...... 42 3.3.2 Effect of landscape properties on pests and natural enemies ...... 42 3.3.3 Interaction between landscape and local drivers of pest and natural enemy numbers ...... 45 3.4 Discussion ...... 51 3.5 Conclusions ...... 55 Chapter 4. General Discussion and Conclusion ...... 57 4.1 Arthopod abundance ...... 58 4.2 Landscape properties affect field adjacent vegetation to mediate arthropod abundance ...... 58 4.3 Landscape properties affect responses of arthropods to local habitat management ...... 60 4.4 Inconsistent effect of landscape properties on pests ...... 61 4.5 Limitations ...... 62 4.6 Future directions ...... 62 4.7 Conclusions ...... 63 References ...... 64 Appendices ...... 78

Certificate of Authorship I am Mst. Salma Akter, hereby declare that this submission is my own work and that, to the best of my knowledge and belief, it contains no material previously published or written by another person nor material which to a substantial extent has been accepted for the award of any other degree or diploma at Charles Sturt University or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by colleagues with whom I have worked at Charles Sturt University or elsewhere during my candidature is fully acknowledged. I agree that this thesis be accessible for the purpose of study and research in accordance with the normal conditions established by the Executive Director, Library Services or nominee, for the care, loan and reproduction of theses.

Mst. Salma Akter ………………… Signature

Date: 18/11/2020

iii

Acknowledgements First and foremost, I am grateful to Almighty for his blessings and grace in my life and during this study. Then I would like to thank Australia Awards Scholarship authority to give me the opportunity.

I must acknowledge the people who helped along the journey. First, my sincere gratitude to my principal supervisor Professor Geoff Gurr for his guidance to initiate and complete this project with his immense knowledge, support, patience and motivation. The knowledge I gathered from him to conduct research, including how to be a humble person, will help me in my future endeavours. I would also like to extend my sincere appreciation to my co-supervisor Dr Olivia Reynolds who helped me with insightful comments and encouragements.

Thanks to Anne Johnson for her great support in my academic writing and encouragements during the last two years. Also thanks to Jian Liu for her assistance and advice.

I would like to acknowledge all the team members of the Field and Landscape Management to Support Beneficial Arthropods for IPM on Vegetable Farms (Hort Innovation project VG16062) research project, particularly, Syed Rizvi and Ahsanul Haque for their cooperation. Thanks to Simon McDonald of SPAN (The Spatial Analysis Unit) for his huge help to learn GIS and analyse the data. Also thanks to John Xia of QCU for statistical support.

I would like to convey my gratefulness to my brother and colleague Imtiaz Faruk Chowdhury and his family for the great support at the initial stage of my study.

I am very grateful to my friend Sunita Pandey and her family for the unconditional help during my study. I cannot forget another friend Pingyang Zhu for his help.

My earnest gratitude also extends to my parents and siblings for their continuous love and motivation. I would also like to thank my husband Md. Rafiqul Islam for his support. Finally, I would like to express my boundless love and gratefulness to my four years old daughter Raaida Shahreen Islam to be the companion of my anger, sorrow and happiness during this study.

List of Figures Figure 1.1 The Spatially Specific Proportional Area approach use to measure landscape composition. A high proportional area or composition of croplands (yellow areas) in 1 km area of the landscape from the focal crop field (star in the middle of the circles) indicates a relatively simple type of landscape. The figure has been sourced from (Schmidt et al., 2008)...... 7 Figure 1.2 Edge density in the landscape. The circles from left to right showing an increasing gradient of edge density of non-crop habitats (green areas). A crop field located in a landscape with high edge density on non-crop habitats (as in right circle) is expected to be more benefited in terms of biological control of herbivore pests than in a landscape with very low edge density of non-crop habitats (as in left circle). The figure has been obtained from (Bloomfield et al., 2020)...... 8 Figure 1.3 Hypothetical representation of landscape connectivity. The figure showing the same proportional area of non-crop vegetation, but the middle one showing higher connectivity with the focal crop field (green box). The figure has been sourced from (Perović et al., 2010)...... 10 Figure 1.4 Shows five radius spatial scales within 3 km area surrounding the focal field. Here the variation in amount of non-crop habitats (green and light green colored area) within each spatial scale is conspicuous. The in-field abundance of pests and natural enemies can be different based on this variation. The figure has been sourced from (Fahrig & Triantis, 2013)...... 11

Figure 2.1 Locations of data collection sites. The map was created in ArcGIS 10.6 (ESRI, 2017)...... 15 Figure 2.2 Positional effect sizes of adjacent land-uses on the distribution of arthropods in brassica fields (mean differences between field centre and field margins with 95% confidence interval). Only significant effects are presented (circles indicate pests and triangles indicate natural enemies). Negative values indicate adjacent land-uses that decrease the numbers of arthropods in the nearby margin compared with the field centre...... 20 Figure 2.3 Landscape composition of woody vegetation at 500 m scale significantly increased the abundance of cabbage aphid B. brassicae in the field centre...... 21 Figure 2.4 Landscape connectivity (cost-area metric of cost-distance analysis or total cost of the area at 1000 m scale for cost-ratio r6) significantly increased the abundance of green peach aphid M. persicae in brassica vegetable field centres...... 22 Figure 2.5 Association of landscape composition of woody vegetation (%) at 5000 m scale with the effect of the field adjacent woody vegetation on diamondback moth, P. xylostella abundance at the field margin...... 23 Figure 2.6 Association of the cost-area metric of cost-distance analysis or the cost of the landscape area in 500 m for the cost-ratio r9 with the effect of field adjacent pastures on the abundance of diamondback moth, P. xylostella at the field margin...... 24 Figure 2.7 Interactive effect between field adjacent land-uses and the cost- area metric of cost-distance analysis or the cost of the landscape area in 1000 m for the cost-ratio r9 significantly reduced the effect of adjacent land-uses v

(woodlands) on diamondback moth, P. xylostella abundance at the field margin. = adjacent brassica crops, = adjacent non-brassica crops, = adjacent pastures, and = adjacent woody vegetation...... 24 Figure 2.8 Association of the edge density (m/ha) of crop fields at 1000 m scale in the landscape with the effect of field adjacent non-brassica crops on the abundance of cabbage aphid B. brassicae at the field margin...... 25 Figure 2.9 Association of the edge density (m/ha) of woody vegetation at 500 m scale in the landscape with the effect of field adjacent woody vegetation on the abundance of ladybirds at the field margin...... 26 Figure 2.10 Association of the cost-path metric of cost-distance analysis or the lowest cumulative cost of the path to arrive the focal field from 1000 m away for the cost-ratio r5 with the effect of field adjacent woody vegetation on the abundance of lady beetles at the field margin...... 26 Figure 2.11 Interactive effect of adjacent land-uses with edge density of croplands at 2500 m scale that influenced the effects of adjacent land-uses on the abundance of lady beetles at the field margin. = adjacent brassica crops, = adjacent non-brassica crops, = adjacent pastures, and = adjacent woody vegetation...... 27

Figure 3.1 Location of the fields in four States in Australia. Numbers indicating the identity of individual sites. The map was created in ArcGIS 10.6 (ESRI, 2017)...... 37 Figure 3.2 A cabbage field with flowering plant strip to provide shelter and food resources to natural enemies of crop pests...... 38 Figure 3.3 Shows the layout of field trial used on each of the 12 sites. A = Alyssum, B = Buckwheat, C = Cornflower...... 39 Figure 3.4 Effect of landscape composition (%) of pasture at the spatial scale of 5000 m that significantly reduced the total pest abundance in focal field...... 43 Figure 3.5 Effect of landscape composition (%) of pasture on the abundance of diamondback moth P. xylostella within the crop field...... 43 Figure 3.6 Effect of composition (%) of woodland in landscape area of 1000 m that significantly increased in-field abundance of natural enemies...... 44 Figure 3.7 Effect of edge density (m/ha) of woodland in the landscape area of 5000 m which significantly increased in-field natural enemy abundance. .. 45 Figure 3.8 Effect of composition (%) of cropland in landscape area of 1000 m that significantly increased in-field natural enemy abundance (count per plants) in fields with flowering strips (dashed line) than without flowering strips or control treatment (solid line)...... 46 Figure 3.9 Effect of composition (%) of croplands in surrounding landscape at 1000 m scale. Natural enemy abundance (count per plant) was significantly higher in the first sampling position (corresponding first sampling position through flowering half and controlled half) than other interior positions. = first sampling position (solid line); = second sampling position 5 m away from field margin (dashed line below the solid line); = third sampling position 10 m away from field margin (long dashed line below the first dashed line); = fourth sampling position 15 m away from the field margin (long

dashed line above the dotted line); = fifth sampling position 20 m away from the field margin (dotted line)...... 47 Figure 3.10 Interactive effect of edge density (m/ha) of cropland in landscape area of 2500 m and local-scale intervention with flower strips (dashed line and control treatment (solid line) on in-field natural enemy abundance. .... 49 Figure 3.11 Effect of cost of landscape area of 1000 m for cost-ratio r2 in relation to local-scale interventions that significantly increased the in-field natural enemy abundance for flowering strips (dashed line) than control treatment (solid line)...... 50

Figure 4.1 Illustrates positions of focal field (location of local habitat manipulation and other local scale properties), and landscape in the agroecosystem...... 58

vii

List of Tables Table 2.1 Assigned cost-ratios used in the cost-distance analysis. Lowest cost (1) indicates highly favourable land-uses and highest cost (100) indicates highly unfavourable land-uses for a taxon to arrive a focal field from a source in the landscape...... 17

Table 3.1 Assigned cost-ratios used in the cost-distance analysis. Favourable land-uses need the lowest cost (1) and unfavourable land-uses need highest cost (100) for a taxon to arrive a focal field from a source in the landscape. Higher costs indicate dominancy of unfavourable land-uses...... 40 Table 3.2 Effect of composition of cropland at 1000 m scale on the within- field distribution of natural enemies when field margin was treated with flowering strips...... 47 Table 3.3 Effects of landscape and local habitat management on pest and natural enemy abundance. Significant results are presented in bold...... 51

viii

List of Abbreviations % Percentage M Metre Km Kilometre m/ha Metre per hectare GIS Geographic information system

Publications prepared from this thesis Chapter 2 Akter, S., Rizvi, S. Z. M., Haque, A., Reynolds, O. L., Furlong, M. J., Osborne, T., Mo, J., McDonald, S., Johnson, A. C., Gurr, G. M. 2020. Landscape context mediates the effects of local vegetation on in-field abundance of pests and natural enemies. Journal of Applied Ecology. (Submitted)

Chapter 3 Akter, S., Rizvi, S. Z. M., Haque, A., Reynolds, O. L., Furlong, M. J., Osborne, T., Mo, J., McDonald, S., Johnson, A. C., Gurr, G. M. 2020. Landscape context affects the scope for local-scale habitat management to mediate in-field pest and natural enemy densities. Agronomy for Sustainable Development (Submitted)

Thesis structure This thesis is composed of four chapters describe the studies on the impacts of landscape properties on herbivore pest and arthropod natural enemy communities in brassica vegetable fields located at different states in Australia. This thesis has been organized following the ‘Thesis by Publication’ format guidelines of Charles Sturt University. Chapter one is composed of a general introduction and the literature review regarding the impacts of local and landscape properties as well as their relationships to mediate pests and natural enemies in agroecosystem. Chapter two entitled “Landscape context mediates the effects of local vegetation on in-field abundance of pests and natural enemies” prepared to submit as a paper and investigates the extent to which properties of wider landscape influence the effectiveness of the diverse land-uses immediately adjacent to the focal crop fields to mediate the in-field abundance of pests and natural enemies. Chapter three is also prepared to submit as a paper which entitled “Landscape context affects the scope for local-scale habitat management to mediate in-field pest and natural enemy densities” to investigate the impacts of landscape properties on the local field scale habitat management with non- crop flowering plants to mediate the in-field abundance and distribution of pests and natural enemies. Chapter four comprises the general discussion, conclusions, limitations and future research directions.

Thesis abstract Intensification of agriculture to maximise crop production leads to disruption of ecosystem services such as of biological control of herbivore pests. Habitat management is a well-established practice at the local field scale to restore natural enemy-mediated pest suppression. Reflecting the vagile nature of arthropod pests and their natural enemies, however, it is crucial to also consider the potential impact on local habitat management strategies of surrounding landscape which may extend up to several kilometres from the focal crops. Recent research on habitat management emphasizes the inclusion of impacts of both local and landscape habitats to elucidate the ecological processes more precisely in herbivore pest management.

This thesis aimed to investigate how landscape properties affect the strength of local field scale properties in mediating the abundance of herbivore pests and their natural enemies in brassica vegetable fields located in diverse landscape types in Australia. The influence of overall landscape structure was explored by extracting data on three landscape properties: landscape composition, edge density, and landscape connectivity. The first part of this work considered the influence of landscape properties on the effects of immediately-adjacent land use on the in-field abundance of pests and natural enemies. The second part of this work explored the extent to which landscape properties influence the success of local-scale habitat management interventions with non-crop flowering plants.

Landscape properties were found to strongly influence the effects of several land-uses immediately adjacent to the focal brassica fields in enhancing natural enemy abundance and pest suppression. Relatively undisturbed or complex landscapes with higher composition and connectivity of non-crop habitats such as woody vegetation and pastures strengthened the effectiveness of similar habitats immediately adjacent to the focal field to suppress specialist brassica herbivore diamondback moth, Plutella xylostella. Landscape-scale edge density of croplands also strengthened the effectiveness of focal fields adjacent to non-host crops to suppress another specialist brassica herbivore, cabbage aphid Brevicoryne brassicae. These effects of landscape properties indicate non-natural enemy-mediated effects of these habitats in suppressing the specialist herbivores. However, the effect

xii

of field adjacent woody vegetation on enhance predatory ladybirds showed contrasting effects to different landscape properties; strengthened in crop dominated or relatively disturbed landscapes in terms of landscape connectivity and diminished in complex landscapes with high edge density of woody vegetation.

Field trials of habitat management with non-crop flowering plants suggest that such flowering resources have little effect in complex landscapes. Regardless of the presence of flowering strips, study fields in complex landscapes and with high edge density of woodlands and composition of pastures had a higher abundance of natural enemies and fewer pests, respectively.

Flowering resources enhance in-field abundance of natural enemies only in simple, crop dominated landscapes. However, the strength of benefit rapidly diminished with increasing distance from the flowers to the field interior. Responses of natural enemies to the flowering strips in crop dominated simple landscapes did not support the intermediate landscape complexity hypothesis. In Australia, very simple landscapes with greater area coverage of crop fields increased the strength of local-scale non-crop habitats such as woodlands or flower strips to enhance natural enemy population in focal fields.

Although landscape-scale woody vegetation promoted natural enemies and suppressed pests, inconsistent effects were also revealed. Promotion of in- field abundance of aphids (B. brassicae and Myzus persicae) occurred in landscapes dominated by woody vegetation.

Overall, this thesis highlighted the significance of both local and landscape scales in habitat management and the need for plans to consider both scales in order to maximize the natural enemy abundance and herbivore pest suppression in agroecosystems. Findings suggest the maintenance of resource habitats of natural enemies such as woody vegetation in close surroundings of crop fields in crop dominated simple landscapes to enhance natural enemy abundance. In a similar landscape circumstance, growers can consider the establishment of flowering strips in crop fields to obtain similar ecosystem services in absence of the field adjacent natural or semi-natural habitats.

xiii

However, in complex landscapes, the establishment of such flowering strips would be redundant.

xiv

Chapter 1. Introduction and literature review

1.1 Introduction Agricultural lands occupy around one-third of the world’s land area (FAO, 2018; Fitton et al., 2019) and are predicted to expand by approximately 12 percent by 2030 (FAO, 2003). This increase is considered essential for fulfilling the dietary requirement of the ever-growing human population (Fitton et al., 2019). This expansion is accompanied by increased simplification in agriculture which has known disadvantages on ecosystems and the services they provide (Landis, 2017). Ecosystem services in agriculture include natural enemy-mediated suppression of herbivore pests. In the last century, pesticides became a major part of agriculture which in turn increased the associated risks to the environment (pollutions and loss of biodiversity) and humans (health hazards) (Carvalho, 2017). Simplification of agriculture and detrimental consequences of pesticides have developed awareness of the need for sustainability and resilience in intensified agricultural systems to maximize crop production and prevent the loss of ecosystem services (Lamichhane, Dachbrodt-Saaydeh, Kudsk, & Messéan, 2016; Landis, 2017).

Habitat management is an important approach in conservation biological control of herbivore pests that helps to achieve sustainable intensification in agroecosystem (González-Chang, Tiwari, Sharma, & Wratten, 2019). Conservation biological control involves alteration of the environment, preferably, through establishing plant diversity, to enhance natural enemies, reduce pest pressure and increase crop yield (Gurr, Wratten, Landis, & You, 2017; Shields et al., 2019). As an approach of conservation biological control, the purpose of habitat management is to promote the capability of natural enemies of herbivore pests provisioning necessary resources that have been defined as SNAP (shelter, nectar, alternate hosts, and pollen) (Gurr et al., 2017; Nilsson, Porcel, Swiergiel, & Wivstad, 2016). At the local field scale, habitat management involves maintenance of natural and semi-natural vegetation or planting of companion plants with focal crops to conserve natural enemies in agroecosystem (González-Chang et al., 2019). However, reflecting the mobile nature of herbivore pests and their arthropod natural enemies, the ecological services in agroecosystems are not only restricted to

the individual field or farm but also extended to wider landscapes (Cong, Smith, Olsson, & Brady, 2014). Thus, the impacts of habitats both at local field and wider landscape scales on pest and natural enemy arthropods need to be considered to achieve desirable pest management services (Chaplin- Kramer & Kremen, 2012). Disregarding the effect of any of the scales may fail to generate finite conclusions on arthropod response (Karp et al., 2018).

1.2 Management of brassica vegetable pests Vegetables under the plant family Brassicaceae are one of the leading crops worldwide (Prohens-Tomás, Nuez, & Carena, 2007). Cabbage (Brassica oleracea var. capitata), cauliflower (Brassica oleracea var. botrytis), brussels sprout (Brassica oleracea var. gemmifera), Chinese cabbage (Brassica rapa), kale (Brassica oleracea var. sabellica), and broccoli (Brassica oleracea var. Italica) are the most important vegetable crops of the genus Brassica. Brassica vegetables are considered as one of the largest sources of vitamins and minerals in human diets (Francisco et al., 2017). In Australia, among brassica vegetables, broccoli constitutes the highest production value of $230 million for 76,104 tons (Australian Horticulture Statistics Handbook Vegetables, 2017/18). Every year production of brassica vegetables is impacted by several insect pest species. Among the insect pests, lepidopterans are dominant, such as diamondback moth (Plutella xylostella), cabbage caterpillar (Pieris rapae), armyworm (Spodoptera litura), and cutworm (Agrotis ipsilon). Other than lepidopterous caterpillars, several species of aphids, whiteflies, and root flies cause extensive damage to brassica crops. Until now management of brassica insect pests is highly dependent on synthetic pesticides, especially in developing countries (Amoabeng et al., 2017). Repeated use of broad-spectrum pesticides leads to the development of resistance in brassica pests (Furlong, Wright, & Dosdall, 2013; Li et al., 2012). Moreover, the killing of natural enemies leads to developing pest resurgence and outbreak of secondary pests. To minimize the use of pesticides and achieve sustainability in pest management, habitat management practices considering local and landscape-scale circumstances need to be considered. Although distinctive effects of local and landscape scales on biological control of brassica vegetable pests have been studied previously, studies on

the interactive processes between the scales are limited (Chaplin-Kramer & Kremen, 2012).

1.3 Influence of local-scale factors on the arthropod community in the agroecosystem 1.3.1 Influence of field adjacent land-uses The in-field abundance and activities of natural enemy arthropods may vary with the types of land-uses immediately adjacent to the crop field margins. Generally, field adjacent non-crop vegetation is the corridor for natural enemies to reach the crop field. After observing such an effect in their study, Thomson and Hoffmann (2010) recommended the growers to establish woody vegetation adjacent to vineyard margins to enhance natural enemy abundance. Similarly, Zhao, Hui, He, and Ge (2013) reported the significance of presence of perennial alfalfa adjacent to wheat fields to enhance the abundance of aphid natural enemies. Natural enemies move from habitat to habitat in search of their resources which may be more abundant in relatively undisturbed annual and perennial habitats rather than the temporary arable crop fields characterised by high levels of disturbance, monocultures and pesticide use. Therefore, if crop fields are adjacent to such annual and perennial habitats, the chances of spillover of natural enemies from these habitats and suppression of pests during crop growing season are potentially high (Thomson & Hoffmann, 2010).

Conversely, uninterrupted fields of similar crops can increase the risk of higher invasion and local build-up of pests as a result of the concentration of their crop resource (Root, 1973). The presence of non-crop annual and perennial habitats can interrupt the arrival of herbivore pests and reduce their in-field abundance. For example, as an unsuitable host, the presence of woody vegetation adjacent to the crop field margins was found to interrupt the specialist brassica herbivore diamondback moth, Plutella xylostella, and subsequently reduced the in-field abundance in the study of Hu, Mitchell, and Okine (1997).

However, the mechanization and intensification of agriculture charactersed by large field sizes creates a scarcity of suitable natural or semi-natural adjacent in agroecosystem and emphasizes the necessity of habitat manipulation.

1.3.2 Influence of local-scale habitat manipulation Habitat manipulation with non-crop flowering plants has been established as one of the most effective techniques to enhance natural enemy abundance and activity in the agroecosystem (Gurr et al., 2017) as a supplement for the scarcity of suitable adjacent natural or semi-natural land-uses. Flowering plants promote ‘top-down’ pest suppression by natural enemies by provisioning nectar and pollen food resources (Gurr, Scarratt, Wratten, Berndt, & Irvin, 2004).

Buckwheat, Fagopyrum esculentum, and alyssum, Lobularia maritima are the most frequently studied flowering species so far and demonstrated as promising companion plants to enhance natural enemies and natural enemy- mediated pest suppression (Araj & Wratten, 2015; Brennan, 2016; Géneau, Wäckers, Luka, Daniel, & Balmer, 2012; Lee & Heimpel, 2005; Nilsson, Rännbäck, Anderson, & Rämert, 2012; Ribeiro & Gontijo, 2017). Other than the non-native flowering plants, some native flowering plants demonstrated promising outcomes in enhancing natural enemies in Australia (Pandey, Rahman, & Gurr, 2018). Along with attracting natural enemies these native plant species provided other significant ecological services such as increasing the abundance of diversified native and non-native pollinators (Pandey & Gurr, 2019)

The selection of correct companion flowering plants concerning crop types and associated arthropod community plays an important role in the success of habitat manipulation (Géneau et al., 2012). The flowering plants should attract natural enemies which depends on several factors, such as distance from sources, flower hue, reflected ultraviolet rays, seasonal variation, and the amount of nectar present in flowers (Hatt et al., 2017). Besides flower characteristics, selection should be based on synchrony of agronomic characteristics and antagonism with the main crop (Brennan, 2016) as well as the proper understanding of the trophic structure and dispersal of arthropods in the agroecosystem (Balmer et al., 2013; Ditner et al., 2013).

1.3.3 Within-field distribution of arthropods The abundance of pests and natural enemies may differ within the crop field as a consequence of the presence of field marginal natural and semi-natural habitats or habitat manipulation with non-crop flowering plants. Ecological

services such as natural enemy enhancement and pest suppression for the field marginal properties (either adjacent natural and semi-natural habitats or habitat management with companion plants) are significant mostly at the crop field margins than the interior (Anjum-Zubair, Schmidt-Entling, Querner, & Frank, 2010; Baillod, Tscharntke, Clough, Batáry, & Marini, 2017; Zhao et al., 2013). However, field margin properties also have the potential of enhancing pest abundance in crop fields observed by Plećaš et al. (2014) and Zhao et al. (2013) which is a real concern in habitat management planning. An important fact is that the in-field assemblage of pests and natural enemies may not always result from a spillover from field adjacent or marginal habitats; rather, arrival from other sources in the wider landscape surrounding the crop field (Baillod et al., 2017).

1.4 Significance of landscape factors on the arthropod community in the agroecosystem An increasing number of studies indicate that the landscape surrounding crop fields can exert a significant influence on the in-field abundance and activity of pests and natural enemies (Tscharntke et al., 2012). Heterogeneous land- uses in the landscape can provide resources for natural enemies such as during periods of pest (prey) scarcity (Gurr et al., 2018). This can facilitate arrival during the cropping season and promote natural enemy-mediated pest suppression especially early in the season. Landscape-scale habitats may also influence pest abundance, especially landscapes dominated by resource crop fields (Jonsson et al., 2015; Rand, Waters, Blodgett, Knodel, & Harris, 2014). Conversely, landscape-scale non-crop habitats may have direct contributions (non-enemy effects) in pest suppression without significant natural enemy abundance and activity (O'Rourke, Rienzo-Stack, & Power, 2011), however, research in this regard is limited .

The ecosystem service of pest suppression either by the natural enemy or non- natural enemy-mediated effects is not dependent only on the presence of heterogeneous habitats but also on the arrangement of the habitats in the landscape (Karp et al., 2018). Several landscape properties have been established as the elements of landscape structure or arrangement of the habitat in the landscape. The key properties of the landscape are, 1) landscape composition; 2) edge density; 3) landscape connectivity; and 4) spatial scales.

1.4.1 Landscape composition: Landscape composition is the proportional area of different habitats or land- uses within a landscape area (Chaplin-Kramer & Kremen, 2012). Complex landscapes with a high proportion of diversified habitats can more readily support diverse species of arthropods, especially, natural enemies compared to simple landscapes with a low proportion of diversified habitats (Bianchi, Booij, & Tscharntke, 2006; Tscharntke et al., 2008). The diversity of non- crop habitats in a complex landscape conveys more food sources and protection options for the natural enemies, and enhance natural enemy- mediated pest suppression in crop fields (Gonthier et al., 2014; Martin, Reineking, Seo, & Steffan-Dewenter, 2015).

The spatially specific proportional area approach (Figure 1.1) is a standard method to measure the composition or proportion of different land-uses within a given area of landscape (Schmidt, Thies, Nentwig, & Tscharntke, 2008). However, disregarding the impact of the arrangement of different land- uses in the landscape is the limitation of this approach (Perović, Gurr, Raman, & Nicol, 2010). The arrangements of the land-uses are different in every landscape, which emphasizes the consideration of the impacts of other landscape properties to complement the effect of the overall landscape.

Figure 1.1 The Spatially Specific Proportional Area approach use to measure landscape composition. A high proportional area or composition of croplands (yellow areas) in 1 km area of the landscape from the focal crop field (star in the middle of the circles) indicates a relatively simple type of landscape. The figure has been sourced from (Schmidt et al., 2008).

1.4.2 Edge density: Edge density is the boundary length of the land-use patches within a given area of landscape (Martin et al., 2019). Edges of enemy harbouring non-crop habitats serve as the interface between habitats and increase the interaction among organisms in a landscape (Bloomfield, McIntosh, & Lambin, 2020). Generally, edge density increases the fragmentation of natural habitats in landscape. Although habitat fragmentation is a threat to biodiversity (especially larger-bodied species) (Fletcher et al., 2018), intensified agroecosystem can benefit from increasing edge density of non-crop habitats. Increasing edges of non-crop habitats (left to right circles in Figure 1.2) in agroecosystem can conserve natural enemies (which are relatively small or minute bodied) and increase their spillover into the crop fields during the growing season (Martin et al., 2019).

Figure 1.2 Edge density in the landscape. The circles from left to right showing an increasing gradient of edge density of non-crop habitats (green areas). A crop field located in a landscape with high edge density on non-crop habitats (as in right circle) is expected to be more benefited in terms of biological control of herbivore pests than in a landscape with very low edge density of non-crop habitats (as in left circle). The figure has been obtained from (Bloomfield et al., 2020).

Edge density can be measured in different ways, including, kernel density estimation, minimum convex polygons, and simple density estimation. Kernel density estimates the successive density of patches in relation to distance (Cai, Wu, & Cheng, 2013). Although kernel density estimation is more demonstrative (Vizzari, 2011), the use of this approach in the ecological study is quite challenging because of the behavioural heterogeneity of organisms (Hoss, Guyer, Smith, & Schuett, 2010). Minimum convex polygons is an easier and effective approach for density analysis, however, the limitation of being biased makes this approach less usable in agroecological studies, particularly for conservation ecology (Burgman & Fox, 2003). Concerning the limitation of the kernel estimation and minimum convex polygon, simple density calculation is the most frequently used method in agroecological studies to measure the density of habitat patches. In this method, density is measured by calculating the total length of a habitat divided by the total selected area of the landscape (Bloomfield et al., 2020; Holzschuh, Steffan-Dewenter, & Tscharntke, 2010; Martin et al., 2019).

1.4.3 Landscape connectivity: The distribution of pests and natural enemies in the agroecosystems depends on landscape connectivity (Goodwin & Fahrig, 2002). Landscape connectivity describes the extent to which the land-uses are connected and influence the movement of arthropods (Pither & Taylor, 1998). Landscapes with the same proportional area or composition of different land-uses may 8

facilitate or hinder the movement of arthropods based on how the land-uses are connected (Figure 1.3). Connectivity of crop fields with the non-crop source habitats of natural enemies can facilitate their movement to arrive the crop fields. Conversely, landscape connectivity of crop fields with the sources of pests can enhance the arrival of pest species rather than natural enemies. For example, the connectivity of rice fields with water bodies which are sources of rice water weevil (Lissorhoptrus oryzophilus) observed to facilitate the movement of the pest in the rice ecosystem over a long distance in a study by Wang, Wu, Shang, and Cheng (2011). Landscape-scale habitat manipulation may complement the connectivity of crop field with supporting habitats of natural enemies in the fragmented agricultural landscapes (Perović et al., 2010) and prevent the arrival of pests in the crop fields. Besides structural connectivity, the dispersal ability of a species determines the functional connectivity within a given area of landscape (Maguire, James, Buddle, & Bennett, 2015; Perović et al., 2010).

Cost-distance analysis is a promising approach, which is demonstrated as more reliable than another well-established approach Euclidean distance analysis, to measure the movement of arthropods based on the connectivity among the land-uses in the landscape (Chardon, Adriaensen, & Matthysen, 2003). The method involves assigning costs for the different land-uses in the landscape based on relative favorability of arthropods (Perović et al., 2010). Generally, unfavorable land-uses are assigned high costs, and favorable land- uses are assigned low costs. The cost-distance analysis demonstrated to have more explanatory power over landscape composition to understand the impact of landscape arrangement on the arthropod community, which is crucial in landscape-scale habitat management (Perović et al., 2010).

Figure 1.3 Hypothetical representation of landscape connectivity. The figure showing the same proportional area of non-crop vegetation, but the middle one showing higher connectivity with the focal crop field (green box). The figure has been sourced from (Perović et al., 2010).

1.4.4 Spatial scales: In ecological studies, spatial scales indicate the extent of ecological services available within a given area or space (Hein, van Koppen, de Groot, & van Ierland, 2006). To achieve a desirable level of pest management services in agroecosystem, it is essential to identify the spatial scales (Figure 1.4) at which pests and natural enemies get the highest benefit to assemble in the crop field. Arthropods communicate differently to the crop and non-crop habitats at various spatial scales (Gonthier et al., 2014). Generally, natural enemies respond positively to the non-crop habitats at smaller spatial scales whereas pests respond to crops at smaller scales (Bischoff et al., 2016). Supporting non-crop vegetation in smaller scales may promote quick colonization of natural enemies in the field, which can facilitate effective pest management (Heimoana et al., 2017).

Figure 1.4 Shows five radius spatial scales within 3 km area surrounding the focal field. Here the variation in amount of non-crop habitats (green and light green colored area) within each spatial scale is conspicuous. The in-field abundance of pests and natural enemies can be different based on this variation. The figure has been sourced from (Fahrig & Triantis, 2013).

However, the response of arthropods also depends on their dispersal ability, mobile species are capable of traveling a larger spatial scale while less vagile species will be strongly affected at smaller scales (Carriere et al., 2012; Perović et al., 2010; Zaller, Moser, Drapela, Schmöger, & Frank, 2008). Scale dependency also aligns with the specialization of species. For example, Chaplin-Kramer, O'Rourke, Blitzer, and Kremen (2011) argued that generalist natural enemies are correlated to both higher and lower scales, but specialists are only with lower scales in the landscapes.

1.5 Relationship between local and landscape factors The amount of non-crop habitats and their arrangement in the landscape regulates the extent of ecological services of pest management in a crop field (Chaplin-Kramer, de Valpine, Mills, & Kremen, 2013). For promoting ecological services in the agricultural lands, it has been suggested to preserve at least 20% non-crop natural or semi-natural habitat around a crop field (Beduschi, Kormann, Tscharntke, & Scherber, 2018; Loos et al., 2019; Tscharntke et al., 2012). However, in intensified agricultural landscapes it is challenging to maintain 20% non-crop habitats and highlights the necessity of maintenance of local-scale diversity to disrupt the pest assemblage as well as facilitate natural enemy abundance (Chaplin-Kramer & Kremen, 2012). Furthermore, responses of pests and natural enemies to landscape properties

can be inconsistent if local-scale effects are not included in landscape studies (Karp et al., 2018). Thus, to maximize the ecosystem delivered pest suppression, impacts of both local and landscape factors need to be considered (Chaplin-Kramer & Kremen, 2012; Saqib et al., 2020).

The interaction between local and landscape factors is proposed to vary with the region where the crop fields are located and for the impacts of diverse landscape properties (Karp et al., 2018; Tscharntke et al., 2012). Though landscape composition is the key feature to measure landscape structure, other properties such as edge density and connectivity can aid in measuring the overall landscape structure.

1.6 Research objectives The primary objective of this thesis was to investigate the influence of multiple landscape properties on local-scale factors to mediate pest and natural enemy abundance in brassica vegetable fields located at multiple states in Australia. More specifically, this work was designed to investigate if crop-adjacent vegetation and landscape properties have distinctive and interactive effects that significantly affect the in-field densities of pests and natural enemies of brassica vegetables. Another specific objective was to investigate the impacts of landscape properties on the success of local habitat manipulation with non-crop flowering strips.

1.7 Research Questions The overarching questions of this thesis were: 1. What are the influences of landscape properties on the efficacy of adjacent land use on in-field abundance of pest and natural enemy arthropods in brassica vegetables? 2. To what extent do landscape properties influence the success of in-field habitat manipulation interventions to deliver pest suppression in brassica vegetables?

Chapter 2. Landscape context mediates the effects of local vegetation on in-field abundance of pests and natural enemies

2.1 Introduction Agricultural intensification reduces land-use diversity in farming landscapes (Tscharntke, Klein, Kruess, Steffan-Dewenter, & Thies, 2005). Protecting or restoring landscape diversity can enhance the abundance of natural enemies and biological pest suppression in crop fields (Tscharntke et al., 2012). Importantly, however, whilst some land-uses, especially non-crop vegetation can benefit natural enemies (Blitzer et al., 2012; Perović et al., 2010), this is not always the case (Karp et al., 2018). Indeed, enhanced pest abundance can result even with a high proportion of non-crop habitats (Plećaš et al., 2014). The inconsistent effects of landscape composition are, in part, a reflection of the importance of other landscape properties such as edge density and connectivity. For example, Martin et al. (2019) found landscape composition and edge density (boundary length of patches) of non-crop habitats affected the strength of ecosystem services including pest suppression. The connectivity of the land-uses is another important landscape property that affects the movement of pest and natural enemy arthropods within the agricultural landscape (Blitzer et al., 2012; Rösch, Tscharntke, Scherber, Batáry, & Osborne, 2013). In the absence of sufficient connectivity, a high composition of supporting land-uses may not facilitate the arrival of natural enemies in crop fields (Perović et al., 2010).

Karp et al. (2018) proposed that disregarding the influence of local scale properties may be one of several possible reasons for inconsistency in the responses of pests and natural enemies to non-crop habitats in the landscape. Local-scale properties include the presence of adjacent non-crop habitats that might support natural enemies with food and shelter and facilitate spillover into agricultural fields during the cropping season (Anjum-Zubair et al., 2010; Heimoana et al., 2017). However, adjacent land-uses have also been reported to have negative effects, increasing abundance of pests rather than natural enemies (Zhao et al., 2013), or having no measurable influence (Fusser, Pfister, Entling, & Schirmel, 2017). Such inconsistency in the effects of local vegetation may result from the influence of landscape-scale properties (Tscharntke et al., 2012). Local-scale effects can benefit from landscape-scale

heterogeneity (Chaplin-Kramer & Kremen, 2012) but not in all cases (Fusser et al., 2017). Accordingly, though it has been established that landscape factors potentially affect the efficacy of local properties, the nature and strength of interactions are poorly understood. Addressing this knowledge gap has applied importance in the field of habitat management to maximise the ecosystem service of biological pest control. Here we integrate arthropod data from a survey of 24 Australian brassica vegetable fields, extending across multiple states and seasons, with information on adjacent land use and wider landscape properties. The primary objective of the study was to examine the respective and interactive effects of field adjacent land-uses and landscape properties on the abundance of arthropods within the focal fields.

2.2 Materials and methods 2.2.1 Sites: Sites were located across the southern half of Australia, representing the main brassica vegetable production regions (Figure 2.1). Data were collected from one focal field on each of 24 farms, each located at least 10 km from other surveyed farms to ensure spatial independence (Thiele & Markussen, 2012). The brassica vegetables grown in these fields were cabbage (Brassica oleracea var. capitata), cauliflower (Brassica oleracea var. botrytis), broccoli (Brassica oleracea var. italica) and Brussels sprouts (Brassica oleracea var. gemmifera). The focal farms had followed their regular management practices to grow the crops. Six farms of the total 24 farms had followed organic farming practice and 18 farms followed conventional practices.

Figure 2.1 Locations of data collection sites. The map was created in ArcGIS 10.6 (ESRI, 2017).

2.2.2 Arthropod data: Between February 2018 and January 2019 (summer of 2017-2018, winter of 2018 and summer of 2018-2019), arthropod abundance was assessed in the central zone of each focal field and 2 m from the edge of each margin representing the types of adjacent land use. At each sampling point, 10 plants were randomly selected and a visual search was carried out for arthropods. All surveys were non-destructive in nature and conducted on dry days between 10.00 and 16.00 hours rather than extending to the more extreme times of early morning, late afternoon or night. The logistics of data capture across multiple sites that were remote from the university campus precluded a narrower time window being used. All sides of leaves, stems, flowers and fruits of each plant were hand-searched for the presence of any arthropods (pests and beneficial). Arthropods were identified in situ to appropriate taxonomic levels and recorded as a point count data.

2.2.3 Classification of adjacent land-uses: The land use adjacent to each field was visually identified and recorded. These were classified as brassica cropland, cropland of other species (non- brassica crops), woody vegetation (trees, shelterbelts, riparian regions and shrubs), pastures (usually dominated by cultivated or uncultivated grasses but occasionally of lucerne (Medicago sativa)) or water bodies. These classifications were to keep consistency with the landscape-scale land-use classifications.

2.2.4 Measurement of landscape-scale properties: Landscape properties were measured using ArcGIS 10.6 (ESRI, 2017) from the Dynamic Land Cover Dataset (DLCD) sourced from Lymburner et al. (2010). Three landscape properties; landscape composition, edge density and connectivity were analysed at the spatial scales of 250 m, 500 m, 1000 m, 2500 m and 5000 m radii from the field. Landscape composition or percentage of different land-uses within a spatial scale was measured using the spatially specific proportional area approach (Schmidt et al., 2008). Land- uses were similarly classified as adjacent land-use analysis except it was not possible to partition cropland into brassica and non-brassica crops. Edge density of each land-use was calculated according to Martin et al. (2019) and is defined as the total length of one land-use type divided by the total landscape area of a given spatial scale. The unit of edge density was metre per hectare. A ‘cost-distance’ analysis was utilized to measure landscape connectivity of land-uses that determine the movement of an arthropod from sources within the selected spatial scales to the focal crop field (Perović et al., 2010). Ten sets of cost-ratios (Table 1.1) were tested using the ‘cost-distance’ tool in ArcGIS 10.6. The values in each of these ratios represented the hypothetical costs to a given arthropod taxon of moving through each type of habitat. Two metrics of ‘cost-distance’ analysis were used for further statistical analysis. The summation of costs of all cells in the cost raster (digitized aerial image layer) of a spatial scale represents the ‘cost-area’ metric. The cost-area for each spatial scale indicates the overall connectivity of that scale. The second metric was cost-path, representing the lowest cumulative cost to reach the crop field (destination) from sources within a spatial scale.

Table 2.1 Assigned cost-ratios used in the cost-distance analysis. Lowest cost (1) indicates highly favourable land-uses and highest cost (100) indicates highly unfavourable land-uses for a taxon to arrive a focal field from a source in the landscape. Land-use types Cost-ratios r1 r2 r3 r4 r5 r6 r7 r8 r9 r10 Water bodies 4 4 20 20 20 1 1 1 100 100 Croplands 2 4 10 20 10 1 1 1 1 100 Pastures 1 1 2 2 1 2 2 100 10 4 Woody 1 1 1 1 1 4 2 20 20 1 vegetation

2.2.5 Statistical analyses: 1) The effects of adjacent land-uses on in-field abundance of arthropods were measured by comparing the arthropod abundance in the crop margin nearest each type of adjacent land use (sum of arthropod count over 10 plants) with equivalent abundance in the same field’s centre (that is, margin minus centre). The comparison between margin and centre of all fields gave a positional effect size for which negative values indicated a lower abundance of arthropods at the field margins indicating a suppressive effect of that adjacent land use and vice-versa. The statistical significance of each positional effect size was tested using one-sample t-test comparing arthropod density in field centres with each type of edge (de Araújo & do Espírito-Santo Filho, 2012). This analysis was conducted for each type of adjacent land use and arthropod taxon to test the null hypothesis that numbers were consistent between field centres and edges. 2) To assess the effects of landscape properties with the minimum possible influence of adjacent land use, arthropod data from only the field centres were regressed against landscape properties as fixed factors in a generalized linear mixed model (GLMM) with season of sampling as a random factor in R 3.5.2 software. A negative binomial distribution was used to resolve the issue of overdispersion. Stepwise regression was followed to select the predictor variables of landscape composition and edge density of different land-use types at each of the five spatial scales. The best regression models were selected based on the lowest Akaike information criterion (AIC) value (Appendix 1.1). Similarly, the cost-ratios of the cost-distance analysis were selected for both cost-area and cost-path metrics at each spatial scale.

This regression analysis investigated whether landscape properties significantly affect the abundance of pests and natural enemies of brassica vegetables at the centre of the focal fields. Exploration of the interaction between landscape factors and adjacent land uses employed the positional effect sizes calculated to quantify the differences between field margins and centres. For each land use and arthropod taxon, positional effect sizes were regressed against the composition of similar land-use type at each of the five spatial scales in the wider landscape and interaction with adjacent land uses. The predictors of edge density were regressed similarly. The predictors of metrics of the cost- distance analysis were selected following stepwise regression. The positional effect sizes for some taxa were negative which were not acceptable in the regression of count data. To fulfil the principle of the regression the negative values were transformed to positive values. For the transformation, the lowest negative value of a particular taxon was added as constant to the differences of that taxon for all 24 sites. Due to the transformation, the lowest negative value turns to zero and other negative values turn to positive values. The transformed data were used as the response variable in the regression model. Generalized linear mixed model (GLMM) analysis with a negative binomial distribution and season of sampling as a random factor was used (Bolker et al., 2009). After regression analysis, the models were back-transformed by deducting the same lowest negative value. To confirm the correctness of the prediction of regression analysis, one-sample t-test was performed again on the transformed back regression model. The regression models that gave similar predicted means as the sample mean were accepted as final models. This analysis was to test the hypothesis that the strength of the effects of field adjacent land-uses on the abundance of arthropods in the crop margins is significantly influenced by landscape-scale properties. 2.3 Results 2.3.1 Arthropod abundance A total of 811 pests and 355 natural enemies were recorded from the 24 brassica vegetable fields. Brassica specialist Plutella xylostella was the most abundant pest species (178 individuals) along with a high abundance of generalist herbivore Myzus persicae (155 individuals). Other brassica specialists were Pieris rapae and Brevicoryne brassicae. Most abundant 18

natural enemy taxon was the predatory lady beetles (105 individuals) (Appendix 1.2).

2.3.2 Effects of adjacent land-use on in-field arthropod abundance at field margin There were robust effects of adjacent land-use, despite aggregation of data from fields dispersed over a wide geographical area and sampling in multiple seasons of the year. Presence of pasture was associated with lower abundance of the major pests, diamondback moth (Plutella xylostella), cabbage white butterfly (Pieris rapae), whiteflies (Aleyrodidae) and green peach aphid (Myzus persicae) in margins of the adjacent brassica crops compared to the field centre (p-value = 0.035, 0.034, 0.022, 0.006, respectively) (Figure 2.2). Woody vegetation similarly reduced P. xylostella (p-value = 0.013) and whiteflies (p-value = 0.017) but had the opposite effect on predatory ladybirds (Coccinellidae), increasing their abundance in the adjacent crop (p-value < 0.001). The presence of an adjacent brassica crop was significantly associated with a lower abundance of whiteflies in the focal brassica crop (p-value = 0.014). Presence of non-brassica crops reduced abundance of cabbage aphid (Brevicoryne brassicae) (p-value = 0.005) and predatory ladybirds in the field margins of the focal crop (p-value = 0.038) (Figure 2.2).

Figure 2.2 Positional effect sizes of adjacent land-uses on the distribution of arthropods in brassica fields (mean differences between field centre and field margins with 95% confidence interval). Only significant effects are presented (circles indicate pests and triangles indicate natural enemies). Negative values indicate adjacent land-uses that decrease the numbers of arthropods in the nearby margin compared with the field centre.

2.3.3 Effects of landscape properties on in-field arthropod abundance at field centre The presence of woody vegetation within 500 m of the focal field increased abundance of B. brassicae in the centre of focal fields (Est. = 0.025 ± 0.016, p-value = 0.005, Figure 2.3). This significant influence of landscape composition of woody vegetation was revealed in comparison to composition of pastures (p-value = 0.177) and crops (p-value = 0.073) at the same spatial scale (Appendix 1.1).

For the other major aphid pest species, M. persicae, the cost-area metric of landscape connectivity was significantly associated with increased abundance in the centres of focal fields. The cost-ratio r6, that assigned croplands as the most favourable land-use, pastures half as favourable and woody vegetation n half again (Table 1.1), applied at a landscape radius of 1000 m, best explained in-field abundances of this pest (Est. = 0.040 ± 0.010, p-value <0.001, Figure 2.4). Cost ratio r6 generated this outcome when regressed along with ratio r6 at 500 m radius (p-value = 0.373) and r3 at 1000 m radius (p-value = 0.591) (Appendix 1.1). No taxa were significantly affected by the cost-path metric for the measurement of landscape connectivity, or by the edge densities of any land-uses at any spatial scale.

Figure 2.3 Landscape composition of woody vegetation at 500 m scale significantly increased the abundance of cabbage aphid B. brassicae in the field centre.

Figure 2.4 Landscape connectivity (cost-area metric of cost-distance analysis or total cost of the area at 1000 m scale for cost-ratio r6) significantly increased the abundance of green peach aphid M. persicae in brassica vegetable field centres.

2.3.4 Influence of landscape properties on the adjacent land-use For P. xylostella, there were significant effects of landscape-scale properties on the positional effect sizes caused by adjacent land-uses. Abundance of woody vegetation within 5 km of focal fields significantly strengthened the suppressive effect of adjacent woody vegetation on P. xylostella (Est. = - 0.025 ± 0.008, p-value = 0.002, Figure 2.5 and Appendix 1.1). A significant strengthening of the effect of adjacent pastures on P. xylostella was also evident in the negative association of adjacent pastures and the cost- area metric of cost-distance analysis. The cost-ratio r9, that assigned crops as the most favourable land-use (lowest cost), pastures 10-fold less favourable and woodland 20-fold less favourable (higher cost), was significant at the 500 m scale. As the cost of landscape area to arrive focal fields increased, for the higher costs of pastures and woodlands, the suppressive effect of adjacent pastures strengthened (Est. = -0.005 ± 0.002, p-value = 0.005, Figure 2.6 and Appendix 1.1). Similarly, the interactive effect of adjacent land-uses with cost-area metric of cost-distance analysis for the same cost-ratio r9 significantly strengthened the suppressive effects of adjacent woody vegetation on P. xylostella (Figure 2.7).

Figure 2.5 Association of landscape composition of woody vegetation (%) at 5000 m scale with the effect of the field adjacent woody vegetation on diamondback moth, P. xylostella abundance at the field margin.

Figure 2.6 Association of the cost-area metric of cost-distance analysis or the cost of the landscape area in 500 m for the cost-ratio r9 with the effect of field adjacent pastures on the abundance of diamondback moth, P. xylostella at the field margin.

Figure 2.7 Interactive effect between field adjacent land-uses and the cost- area metric of cost-distance analysis or the cost of the landscape area in 1000 m for the cost-ratio r9 significantly reduced the effect of adjacent land-uses (woodlands) on diamondback moth, P. xylostella abundance at the field margin. = adjacent brassica crops, = adjacent non-brassica crops, = adjacent pastures, and = adjacent woody vegetation.

Edge density of land-uses in the landscape, though not significant in isolation, significantly influenced positional effect sizes evident for B. brassicae and predatory ladybirds. High edge density of cropland in the landscape at the 1 km scale strengthened the effect of the field adjacent non-brassica crops in suppressing the abundance of B. brassicae (Est. = -0.003 ± 0.001, p-value = 0.039, Figure 2.8 and Appendix 1.1).

Figure 2.8 Association of the edge density (m/ha) of crop fields at 1000 m scale in the landscape with the effect of field adjacent non-brassica crops on the abundance of cabbage aphid B. brassicae at the field margin.

For predatory ladybirds, whilst numbers were positively affected by the presence of adjacent woody vegetation (Figure 2.2), the effect diminished with increasing edge density of woody vegetation within 500 m (Est. = -0.011 ± 0.003, p-value < 0.001, Figure 2.9). However, an increase in the positional effect sizes of adjacent woody vegetation for these predators was evident in the association with the cost-path metric for the cost-ratio r5 at the scale of 1000m (Est. = 0.040 ± 0.008, p < 0.001, Figure 2.10). For r5, woody vegetation and pastures were assigned to be 10-fold as favourable as croplands and 20-fold as favourable as water bodies. The enhancing effect of adjacent woody vegetation on the predatory ladybirds also was unaffected by most types of adjacent land-uses but strongly promoted by woody vegetation in the landscape (Figure 2.11).

Figure 2.9 Association of the edge density (m/ha) of woody vegetation at 500 m scale in the landscape with the effect of field adjacent woody vegetation on the abundance of ladybirds at the field margin.

Figure 2.10 Association of the cost-path metric of cost-distance analysis or the lowest cumulative cost of the path to arrive the focal field from 1000 m away for the cost-ratio r5 with the effect of field adjacent woody vegetation on the abundance of lady beetles at the field margin.

Figure 2.11 Interactive effect of adjacent land-uses with edge density of croplands at 2500 m scale that influenced the effects of adjacent land-uses on the abundance of lady beetles at the field margin. = adjacent brassica crops, = adjacent non-brassica crops, = adjacent pastures, and = adjacent woody vegetation.

2.4 Discussion This study demonstrates multiple, strong effects of immediately adjacent land-uses. We observed significant suppressive effects on abundance of all major pest taxa in the crop margins adjacent to at least one type of adjacent land-use (such as pasture or woody vegetation) compared with crop centres. The only natural enemy taxon to respond significantly was ladybirds which, consistent with the observed suppression of pests, were more numerous in crop margins adjacent to woody vegetation, the land use associated with suppressed diamondback moth and whitefly. Landscape-scale properties, in contrast, had few distinct effects on arthropods, with only the two aphid species significantly increased. Notably, however, there were many instances of landscape properties affecting the strength of the adjacent land-uses, highlighting the importance of this study.

2.4.1 Distinct effects of adjacent land-use on in-field arthropods at field margin Field adjacent pasture, woody vegetation, and non-brassica crops all reduced the abundance of specialist brassica pests in the adjacent brassica crop margin. This effect was evident across pest families and feeding guilds: P. xylostella and P. rapae (lepidopteran chewing feeders) and B. brassicae (hemipteran sucking feeder). This effect is consistent with the resource concentration hypothesis (Root, 1973) whereby the establishment and persistence of specialist herbivores on their host plant may be disrupted by the presence of nearby non-host plants and operates separately to any effect mediated by natural enemies. Resource concentration effects can operate by several mechanisms including the non-host plants serving as a barrier to the arrival of specialist herbivores on host plants. This is most conspicuously possible in the observed effects of woody vegetation, large and perennial plants that would potentially block visual and olfactory cues as well as physical movement. A similar assumption has been made by Kross et al. (2020), however, there is no mechanistic study on this aspect to support the claim. Associated with this, and likely to have been important across multiple types of non-crop vegetation, is the fact that these will not have supported populations of the pest. Although the adults of brassica specialist herbivores oviposit on cultivated and uncultivated brassica plants, and these can support larval development (Y.-Z. Chen, Lin, Wang, Yeh, & Hwang, 2004; Idris & Grafius, 1996), these plants are scarce in woody vegetation patches. Accordingly, spillover of pests from these patches will have been negligible. Whiteflies also exhibited reduced abundance in the crop margins adjacent to woody and pasture vegetation but the in-field assessment protocol used in this study did not allow discrimination of the specialist whitefly Aleyrodes proletella, and more polyphagous species such as Trialeurodes vaporariorum and Bemisia tabaci. Accordingly, it is not possible to unambiguously attribute this effect to resource concentration effects because non-brassica specialist or generalist herbivores may have been hosted by some plant species in the non- crop vegetation (McKenzie, Anderson, & Villarreal, 2004) from which spillover may have subsequently occurred.

Notwithstanding the operation of resource concentration hypothesis effects, natural enemy effects also are likely to have operated at this local scale 28

because ladybirds (a taxon of predators collectively attacking all brassica pests) were significantly more numerous in brassica crop margins adjacent to woody vegetation. This effect may have occurred if the woody vegetation provided resources, such as pollen, alternative prey and shelter, that led to a high density of ladybirds and so constituted donor habitat from which these predators colonised nearby crop fields when prey became available (Heimoana et al., 2017; Thomson & Hoffmann, 2013). Notably, ladybirds were reduced in numbers when the adjacent land-use was a non-brassica crop, possibly reflecting the more ephemeral nature of this vegetation and generally poorer consistency of availability in shelter and food resources compared with perennial woody vegetation and pasture (Wiedenmann & Smith Jr., 1997).

2.4.2 Distinct effects of landscape properties on in-field arthropods at field centre Landscape compositions with high proportions of woody vegetation at the 500 m radius scale were associated with increased abundance of B. brassicae in field centres. In broad agreement, numbers of the other aphid species, M. persicae were increased by high cost area values from the landscape connectivity analysis. In that analysis, the relative non-favourability assigned to woodland was four-fold that of the favourability of crops, and double that assigned to pastures. Thus, landscapes with high cost had relatively abundant woodland. This apparent promotion of aphids in crops by aspects of woodland predominance in the landscape is contrary to the often held belief that non- crop vegetation in the landscape leads to herbivore suppression, potentially via enhanced natural enemy activity (Bianchi et al., 2006). However, the general notion has been eroded by the recent meta-analysis of Karp et al. (2018) and these authors suggested that pest suppression was not axiomatic with non-crop vegetation. In this present study, the higher in-field abundance of B. brassicae in fields set in woody vegetation-rich landscapes could have resulted from the paucity of natural enemies and their resources, such as shelter, provided by the relative structural complexity and botanical diversity of woody plant communities as per an assumption in Tscharntke et al. (2016). Whilst this effect is counter to the observation in the present study, as woody vegetation immediately adjacent to crop fields promotes ladybirds at field margins, scale effects could account for this. Woody vegetation adjacent to crops would, for example, allow ladybirds to diurnally commute to crops to 29

forage for prey yet return to woody vegetation for shelter from environmental extremes. Such movement by ladybirds from crop field to adjacent shelterbelt was observed in a study with wheat by (Dong, Ouyang, Lu, & Ge, 2015) where in-shelterbelt abundance of the beetles was associated with landscape- scale woody vegetation, however, in-crop abundance relied on landscape- scale diversity. That is, the movement of the predatory beetles across the crop and non-crop ecotone depends on landscape attributes and may account for the high abundance of aphids in field centre.

2.4.3 Influence of landscape properties on adjacent land-use Testing for interactions between landscape and field-adjacent land-uses involved analysing the influence of landscape properties on the positional effect sizes that were calculated by comparing arthropod numbers in field centres with field margins. For the specialist herbivore of brassica vegetables, P. xylostella, the suppressive effect of adjacent woody vegetation was enhanced (i.e., larger effects) by the presence of woody vegetation in the wider landscape to the remarkably large scale of 5km from the focal field. Broad agreement with this positive effect of landscape on local-scale effects was evident for the cost distance analysis and interaction term of cost distance analysis with local land-uses in suppressing P. xylostella abundance in the focal crops. In that analysis, woody vegetation was assigned a cost 20-fold greater than cropland (with pasture cost at 10 and water bodies at 100). Landscapes with high cost had larger effects. That is, a landscape dominated by land-uses that were strongly unfavourable to this specialist herbivore (Hu et al., 1997) had exaggerated differences in in-field numbers of this pest, with much lower abundance in the margins close to woody vegetation.

For predatory lady beetles, effect of the adjacent woody vegetation in enhancing the abundance was magnified by the interaction of field adjacent woody vegetation with edge density of landscape-scale croplands and cost area metric of cost-distance analysis. In that cost-ratio of cost-distance analysis a low cost (1) was assigned to woody and pasture vegetation and much higher costs (10) for crops and (20) for water bodies which indicates the dominancy of croplands. That is, the increasing amount of croplands in the landscape increased the effect of the adjacent woody vegetation to enhance natural enemies. This linear relationship observed here (and for other

response variables) is inconsistent with the intermediate landscape- complexity hypothesis proposed by (Tscharntke et al., 2012). According to that hypothesis, local-scale effects were assumed to be strongest in moderately complex landscapes (1-20% non-crop area), but diluted in extremely simple (<1% non-crop area) or complex (>20% non-crop area) landscapes because the former lacked donor habitats from which natural enemies could respond to local features whilst the latter had an abundance of resource-rich habitats that made additional resource patches redundant. The intermediate landscape-complexity hypothesis speculated that the threshold of moderately complex landscapes can vary based on the region where the crop field is located (Tscharntke et al., 2012). Notwithstanding the possibility that thresholds may differ in Australia compared with European agricultural landscapes in which the hypothesis was generated, the fact that all relationships observed in the present study were linear suggests no support for the intermediate landscape-complexity hypothesis.

Conversely, the enhancing effect of adjacent woody vegetation on predatory lady beetles was diminished by the high edge density of woody vegetation in the wider landscape. That is, brassica fields located in settings with a large amount of woody vegetation within 500m diluted the effect of adjacent woody vegetation. The effect of edge density of woody vegetation indicated increasing landscape complexity, which is consistent with the intermediate landscape-complexity hypothesis, where it has been proposed that effectiveness of local-scale factors will not noticeable in complex landscapes (Tscharntke et al., 2012).

For B. brassicae, suppressive effect of field adjacent non-brassica crops were strengthened by high edge density of croplands in the landscape at 1000 m scale suggesting non-enemy effects on this specialist brassica herbivore. The host specificity of B. brassicae may limit their ability to permeate the edges of other non-host vegetation (Gabryś & Pawluk, 1999; Macfadyen & Muller, 2013) and reinforced the effect of the field adjacent non-brassica crops. However, Bukovinszky, Potting, Clough, van Lenteren, and Vet (2005) reported that the winged B. brassicae can move across the landscape irrespective of patches of host or non-host plants. One possible reason for the contradiction between this present study and Bukovinszky et al. (2005) may

be the variation of sampling season and procedures between the studies, as adult winged B. brassicae is formed only during the migrating season in search for new hosts (Shaw, 1970) and B. brassicae colonies composed of different forms (winged or wingless) or life stages (immature to adult) were sampled only once from the sites in this present study.

2.5 Conclusions The key finding of this study was the existence of strong influences of landscape properties on the effect of local land-uses that can shape the in- field abundance of pest and natural enemy arthropods. Initially, comparing the numbers of arthropods in field margins with the field centre, suppressive effects of diverse adjacent land-uses were observed on several key pests which is presumably the consequence of resource concentration effect. The enhanced abundance of a generalist natural enemy taxon (lady beetles) by woody vegetation adjacent to focal field indicated the contribution of natural enemies in suppressing the pests. Secondly, assessment of distinctive effects of landscape properties revealed relatively fewer effects but a consistently increased abundance of two species of aphids at the field centre for the high amount of landscape-scale woody vegetation which indicates the inconsistent effect of landscape when local effects are excluded. Finally, the impact of landscape properties on effects of adjacent land-uses demonstrated both pest suppression and enhancement of natural enemies. Suppression of P. xylostella was evident as the strength of adjacent woody vegetation and pastures was magnified in complex landscapes suggesting non-natural enemy-mediated effect of the land-uses. Non-enemy effect in suppressing another specialist herbivore B. brassicae resulted from the influence of landscape-scale croplands on the effect of field adjacent non-brassica crops suggesting the necessity of crop diversity in agricultural landscape. The enhancing effect of adjacent woody vegetation on predatory lady beetles was magnified in crop dominated simple landscapes. However, this effect of adjacent woody vegetation had contrasting responses to landscape properties, both diminished and magnified in complex landscapes in terms of edge density and connectivity, respectively.

Among the interactions between adjacent land-uses and landscape properties, strong impacts were observed for suppressing two specialist brassica

herbivores and promoting one generalist natural enemy. Conversely, increased abundance of pests by landscape properties was observed when the effects of adjacent land-uses were disregarded. The findings of this study demonstrate that disregarding the effects of local drivers such as immediately- adjacent land use can result in poor capacity to understand the responses of arthropods to landscape properties. More studies combing local and landscape drivers are needed to comprehend the interactive processes on pests and natural enemies that can improve the biological control in agroecosystem.

Chapter 3. Landscape context affects the scope for local-scale habitat management to mediate in-field pest and natural enemy densities 3.1 Introduction Intensification of agriculture has reduced natural and semi-natural habitats of agricultural landscapes that are reservoirs of the natural enemies of herbivore pests (Gurr, Wratten, Landis, & You, 2017). The scarcity of resource habitats and increased use of pesticides at the local field scale also adversely affects natural enemies (Saqib et al., 2020). Remediating these landscape with local scale interventions is currently a major area of research activity (Johnson et al., 2020). Natural enemy enhancement and pest suppression can be delivered by in-field habitat management via the addition of plants such as nectar- containing non-crop flowering plants (Gurr et al., 2016). However, selection of appropriate flowering plants is important given that flowering plants can also be advantageous to herbivore pests at the adult stage, especially to lepidopteran adults (Chen et al., 2020). Moreover, the enhanced abundance of natural enemies for the flowering plants may not make a sufficient contribution to suppressing pest abundance in crop fields (Berndt, Wratten,

& Scarratt, 2006). If the wider landscape surrounding the crop field is a greater source of pests than natural enemies, the biological control services delivered by the floral resources can be disrupted (Tscharntke et al., 2016).

Accordingly, the success of local-scale management practices is potentially affected by properties of the wider landscape surrounding the crop fields

(Phillips & Gardiner, 2016). Similarly, disregarding local-scale management may account for the inconsistent responses of pests and enemies in landscape studies (Karp et al., 2018).

According to the ‘intermediate landscape-complexity hypothesis’, local-scale floral resources are predicted to be relatively ineffective in landscapes with

extremely high or low complexity, because these have abundant resources for natural enemies or such poor resource availability that enemies are locally scarce, respectively (Tscharntke et al., 2012). Thus, landscapes of moderate complexity, with sufficient natural or semi-natural non-crop habitat to support an assemblage of natural enemies, are hypothesised to benefit most from the addition of floral resource strips that provide a significant local addition to already available resources. The intermediate landscape-complexity hypothesis was supported in the New Zealand investigation of Jonsson et al.

(2015) in kale to the extent that the effect of flower strips increased over a range of landscape compositions ranging from 2-48% annual crops.

Importantly, however, the response of aphid and diamondback moth parasitism was linear over this range with no indication of the relationship reaching horizontal asymptote in the simpler landscapes. A USA study conducted in soybean fields, had more widely contrasting landscapes, from less than 10% to over 70% semi-natural habitat (Woltz, Isaacs, and Landis

(2012). Coccinellid predator numbers responded to the presence of buckwheat border strips but this occurred in all landscapes irrespective of their composition and there was no interaction to support the intermediate landscape complexity hypothesis.

Whilst composition is the most frequently studied property of landscapes, other properties such as edge density (length of habitat patches per unit area) and connectivity (the extent of connection among different land-use types in a defined area of landscape that can facilitate or hinder the arrival of a species into crop field from a source) are known to influence arthropod movement

(Perović, Gurr, Raman, & Nicol, 2010; Tscharntke et al., 2012). However, the influence of edge density and connectivity on the strength of local effects

is even less well understood than is the influence of landscape composition.

Accordingly, this study aimed to investigate the effects of all three of these major landscape properties on local-scale habitat management with flowering plants that were established to promote biological pest control in brassica vegetable fields. The study was designed to test the following specific hypotheses: (1) irrespective of local management, the abundance of pest and natural enemy arthropods within crop fields is affected by landscape properties; (2) local and landscape factors interact to drive net effects on in- field abundance of pests and natural enemies.

3.2 Materials and methods 3.2.1 Sites: Experimental trials were established in 12 brassica vegetable fields in the south-east of Australia; New South Wales, Victoria, South Australia, and

Queensland (Figure 3.1). Brassica vegetables grown in these crop fields were cabbage (Brassica oleracea var. capitata), wombok (Brassica rapa), and

Brussels sprouts (Brassica oleracea var. gemmifera).

The crops surveyed were all managed according to the normal agronomic practices of the host farms. Accordingly, the results reflect the outcome of habitat manipulation being adopted by farmers as a complement to their usual management (rather than, for example, being a potential substitute)

Figure 3.1 Location of the fields in four States in Australia. Numbers indicating the identity of individual sites. The map was created in ArcGIS 10.6 (ESRI, 2017).

3.2.2 Floral resources: Three flowering species, sweet alyssum, Lobularia maritima, buckwheat,

Fagopyrum esculentum, and cornflower, Centaurea cynanus were selected as the local-scale management interventions. These plants were selected based on their significant influence on enhancing natural enemies reported in previous studies (Ditner et al., 2013; Géneau et al., 2012; Orre, Wratten,

Jonsson, Simpson, & Hale, 2013; Ribeiro & Gontijo, 2017).

Figure 3.2 A cabbage field with flowering plant strip to provide shelter and food resources to natural enemies of crop pests.

3.2.3 Experimental design: A 58m length of each crop field was divided into two 24 m areas that were randomly allocated to flowering strips or control treatment with 10 m buffer area in between (Figure 3.3). The flower strip contained three plant species whilst the control strip remained unplanted and did not contain blooming weeds during the study. The sub plots of flowering strip consists of alyssum

(Lobularia maritima), buckwheat (Fagopyrum esculentum), and cornflower

(Centaurea cyanus)), each replicated four times per field.

3.2.4 Arthropod sampling: Arthropod abundance was sampled visually during summer 2020 on ten plants at five points (Figure 3.3) progressively into the field at the mid part of the treatment and strips. Visual counting was carried out between 10.00 and

16.00 hours in the daytime. The number of arthropods recorded on ten plants at each sampling position was pooled together to avoid frequency of too many zeros (Baillod et al., 2017). Arthropods were identified up to species level wherever possible and also categorized as total pests and total natural

enemies. Fields were surveyed when the flowering plants were at the peak of their blooming and the vegetables were started maturity.

Figure 3.3 Shows the layout of field trial used on each of the 12 sites. A = Alyssum, B = Buckwheat, C = Cornflower.

3.2.5 Analysis of landscape factors: A Dynamic Land Cover Dataset (DLCD) obtained from Lymburner et al.

(2010) was used in ArcGIS 10.6 (ESRI, 2017) to measure the landscape properties at the spatial scales of 250 m, 500 m, 1000 m, 2500 m, and 5000 m radii from each field. Land uses were classified as cropland, woody vegetation (trees, shelterbelts, riparian regions, and shrubs), pastures, or water bodies. Landscape composition is the percentage of the land-use types at each spatial scale and assessed by using the spatially specific proportional area approach (Schmidt et al., 2008). Edge density is the ratio of the length of one land-use type and the total landscape area of a given spatial scale. Edge density of each land-use type at each spatial scale was assessed in metre per hectare according to the following formula stated in Bloomfield et al. (2020).

Landscape connectivity of land-uses was measured using a cost-distance

analysis similar to Perović et al. (2010). The cost-distance analysis involves assigning different costs or resistant values, which generated comparative cost-ratios (Chardon et al., 2003), to the land-uses within a spatial scale to measure the connectivity or hindrance that a species may face when assembling the destination (focal crop field) from a source within the spatial scale. Ten different sets of cost-ratios (Table 3.1) were used in the ‘cost- distance’ tool in ArcGIS 10.6. Two metrics of ‘cost-distance’ analysis were used for further statistical analysis. The summation of costs of all cells in the cost raster (digitized image layer) of a spatial scale represents the ‘cost-area’ metric. The cost-area for each spatial scale indicates the overall connectivity of that scale. Another metric was the cost-path, represents the path which taken the lowest cumulative cost to reach the crop field (destination) from the source within a spatial scale.

Table 3.1 Assigned cost-ratios used in the cost-distance analysis. Favourable land-uses need the lowest cost (1) and unfavourable land-uses need highest cost (100) for a taxon to arrive a focal field from a source in the landscape. Higher costs indicate dominancy of unfavourable land-uses. Land-use types Cost-ratios r1 r2 r3 r4 r5 r6 r7 r8 r9 r10 Water bodies 4 4 20 20 20 1 1 1 100 100 Croplands 2 4 10 20 10 1 1 1 1 100 Pastures 1 1 2 2 1 2 2 100 10 4 Woody 1 1 1 1 1 4 2 20 20 1 vegetation

3.2.6 Statistical analyses: The effects of landscape properties and two-way interaction between landscape properties and local habitat management with flowering strips and without flowering strips (control treatment) on arthropod abundance within focal crop fields were assessed (Woltz, Isaacs, & Landis, 2012). The number of arthropods recorded on the brassica plants within the field treated with

flower strips compared to without flower strips was regressed with landscape variables at each of the five spatial scales using generalized linear mixed models (GLMM) with negative binomial distribution and field identity as a random factor in R version 3.5.2 and package ‘lme4’. Arthropod data were square-root transformed to achieve normality before fitting in models

(Baillod et al., 2017). Five different models were run for each arthropod species as well as for total pests and total natural enemies. The models were ranked based on the Akaike information criterion (AIC) to select the final model that generated the lowest AIC value (Appendix 1.4). The level of significance and test statistics were calculated using type-III ANOVA of the final regression models. This statistical analysis was to investigate the hypothesis that the extent to which local habitat management with non-crop flowering plants influence the in-field abundance of pests and natural enemies depend on the landscape properties.

Similarly, the abundance of arthropods at each sampling position in figure 3.2

(i.e. summation of abundance of arthropods counted on 20 plants at corresponding sampling points through the flowering half and control half) was regressed with landscape variables at each spatial scale and interaction of landscape variables with local interventions, using GLMM with negative binomial distribution and field identity as a random factor. This analysis was to examine the effect of landscape properties at each spatial scale on the distribution of pests and natural enemies at different sampling positions keeping the effect of local interventions aside. Furthermore, vertical sampling positions in figure 2 through flowering and control halves were analysed separately against landscape properties and interaction with local interventions following a similar analysis procedure. This statistical analysis

was to investigate the distribution of pest and natural enemy arthropods within crop fields in relation to local and landscape properties.

3.3 Results 3.3.1 Abundance of arthropods In the survey a total of 5967 arthropods were recorded from the 12 focal fields. Pest abundance (4658 individuals) was dominant over the beneficials

(1309 individuals). Pest community composed of mostly Plutella xylostella

(1243 individuals), including Pieris rapae (73), whiteflies (861) and jassids

(88). Lady beetles (422) and predatory spiders (381) were the frequent beneficial taxa (Appendix 1.3).

3.3.2 Effect of landscape properties on pests and natural enemies Pest abundance in crop fields was significantly reduced as composition of pastures within 5 km of focal fields increased (Figure 3.4). However, there was no significant interaction between landscape-scale pasture composition and local habitat management in suppressing the pest abundance (Table 3.3).

There was a similar response of the specialist brassica herbivore diamondback moth Plutella xylostella to landscape composition of pastures at a 5 km scale

(Figure 3.5). None of the other landscape properties of other land-uses at any spatial scale revealed any significant influence on the abundance of any other pest species sampled from the focal fields.

Figure 3.4 Effect of landscape composition (%) of pasture at the spatial scale of 5000 m that significantly reduced the total pest abundance in focal field.

Figure 3.5 Effect of landscape composition (%) of pasture on the abundance of diamondback moth P. xylostella within the crop field.

In contrast to the response of pests, an increased abundance of natural enemies was observed in landscapes with increasing complexity as reflected in the composition of woodlands in the landscape at the 1 km spatial scale

(Figure 3.6 and Table 3.3). Increasing edge density of woodland within 5 km of the focal field revealed a similar effect on natural enemy populations

(Figure 3.7). Interaction of the woody vegetation with local habitat management was not significant at any spatial scale to enhance the in-field abundance of natural enemies.

None of the natural enemies sampled in the focal field responded significantly at the species level to any of the landscape predictors or their interaction with local interventions.

Figure 3.6 Effect of composition (%) of woodland in landscape area of 1000 m that significantly increased in-field abundance of natural enemies.

Figure 3.7 Effect of edge density (m/ha) of woodland in the landscape area of 5000 m which significantly increased in-field natural enemy abundance.

3.3.3 Interaction between landscape and local drivers of pest and natural enemy numbers Natural enemy populations were strongly influenced by the interaction of landscape properties with local-scale habitat management in simple landscapes dominated by croplands. Natural enemy abundance was strongly elevated by flowering strips in landscapes with high composition of cropland at the 1 km scale (Figure 3.8). This effect diminished as cropland composition declined from close to 70% cropland to the extent that natural enemy numbers were comparable to those in the control (no flower) treatment when cropland composition fell to below 40%. In the control treatment, natural enemy numbers were consistent across the full range of cropland compositions.

When natural enemy numbers were partitioned according to the spatial distance from the flower strips, there was a clear trend such that counts close to the flower strip were strongly affected by cropland composition whilst

sampling positions progressively further from the strip showed a less strong response (Figure 3.9, Table 3.2).

Figure 3.8 Effect of composition (%) of cropland in landscape area of 1000 m that significantly increased in-field natural enemy abundance (count per plants) in fields with flowering strips (dashed line) than without flowering strips or control treatment (solid line).

Figure 3.9 Effect of composition (%) of croplands in surrounding landscape at 1000 m scale. Natural enemy abundance (count per plant) was significantly higher in the first sampling position (corresponding first sampling position through flowering half and controlled half) than other interior positions. = first sampling position (solid line); = second sampling position 5 m away from field margin (dashed line below the solid line); = third sampling position 10 m away from field margin (long dashed line below the first dashed line); = fourth sampling position 15 m away from the field margin (long dashed line above the dotted line); = fifth sampling position 20 m away from the field margin (dotted line).

Table 3.2 Effect of composition of cropland at 1000 m scale on the within- field distribution of natural enemies when field margin was treated with flowering strips. Within field sampling positions Estimated Std. z Pr(>|z|) coefficients Error value Sampling position 1 (0 m from 0.024392 0.010 2.328 0.0199 flower margin) Sampling position 2 (5 m from 0.019195 0.010 1.805 0.0710 flower margin) Sampling position 3 (10 m from 0.013510 0.011 1.193 0.2328 flower margin) Sampling position 4 (15 m from 0.009746 0.011 0.850 0.3953 flower margin) Sampling position 5 (20 m from 0.011256 0.011 0.956 0.3390 flower margin)

Edge density of cropland, like cropland composition, had a significant effect on in-field abundance of natural enemies though at the larger scale of 2.5 km

(Figure 3.10). In the case of natural enemy numbers in the flower strip treatment area, there was a positive response to cropland edge density with numbers declining in more complex landscapes with progressively less crop to a level comparable with the numbers in the control (no flower strip) areas.

In these control areas, natural enemy numbers showed the opposite trend to numbers in the flower strip areas: declining as the edge density of the landscape increased.

Landscape connectivity also affected the response of natural enemies to the local presence of a flowering strip (Figure 3.11). For the set of cost ratios in which cropland and water bodies had a cost ratio four-fold greater than woodland or pasture (i.e. assuming that they were four times less conducive to the movement of natural enemies) natural enemy numbers in the flower strip areas showed a positive response. In landscapes relatively costly for natural enemy movement, flower strips had strong effects on numbers which declined as landscape connectivity fell, reaching a value comparable to that in the control (no flower strip) treatment. Numbers in the control treatment were consistent over the full range of landscape cost.

Figure 3.10 Interactive effect of edge density (m/ha) of cropland in landscape area of 2500 m and local-scale intervention with flower strips (dashed line and control treatment (solid line) on in-field natural enemy abundance.

Figure 3.11 Effect of cost of landscape area of 1000 m for cost-ratio r2 in relation to local-scale interventions that significantly increased the in-field natural enemy abundance for flowering strips (dashed line) than control treatment (solid line).

Table 3.3 Effects of landscape and local habitat management on pest and natural enemy abundance. Significant results are presented in bold. Chisq Pr(>Chisq)

Total pest abundance Pasture composition at 5000 m 7.592 0.005 Pasture composition at 5000m*Treatments 0.395 0.529 P. xylostella abundance Pasture composition at 5000 m 10.891 0.001 Pasture composition at 5000 m*Treatments 0.156 0.692 Total natural enemy abundance Cropland composition at 1000 m 0.045 0.830 Cropland composition at 1000 m*Treatments 9.048 0.002 Total natural enemy abundance Cropland edge density at 2500 m 0.592 0.441 Cropland edge density at 2500m*Treatments 8.151 0.004 Total natural enemy abundance Cost area at 1000 m for cost-ratio r2 0.013 0.905 Cost area at 1000 m for cost-ratio r2*Treatments 6.393 0.035 Total natural enemy abundance Woodland composition at 1000 m 5.138 0.023 Woodland composition at 1000 m*Treatments 0.782 0.376 Total natural enemy abundance Woodland edge density at 5000 m 11.455 0.001 Woodland edge density at 5000m*Treatments 0.317 0.573

3.4 Discussion The twelve landscapes in the present study ranged from simple, highly disturbed in nature with strong representation of croplands to much more complex and less disturbed areas with more woodland and pasture. Over this range of landscape types there were significant direct effects on pests and natural enemies as well as strong evidence that landscape properties affected the responses of arthropods to in-crop flower strips. At the largest spatial scale, extending 5 km from focal fields, increasing composition of pasture negatively affected in-field pest populations irrespective of local-scale habitat

manipulation. This effect was evident also for the key pest in this system, P. xylostella, which is a monophagous specialist on Brassicaceae hosts. A possible explanation for the P. xylostella effect is non-enemy effects of pastures is the availability of brassica weeds and volunteers in and around pastures was poorer than the proportion of brassica hosts in other land uses such as croplands (Hu et al., 1997). Whilst such an effect cannot be discounted, since detailed data on botanical composition is not available, it is likely that natural enemies were acting on P. xylostella and other pests in this study. Relatively complex and undisturbed landscapes with high composition and edge density of woodlands had strongly increased in-field abundance of natural enemies irrespective of local-scale habitat management with flowering strips. The influence of woody vegetation on total natural enemy populations in terms of landscape composition and edge density agreed with the effect of landscape composition of non-crop habitats observed in the study of Schmidt, Roschewitz, Thies, and Tscharntke (2005) and Woltz et al. (2012) on ground-dwelling predatory spiders and predatory ladybirds, respectively.

Simple landscapes dominated by croplands in terms of composition (and less pasture and woodland) significantly enhanced the abundance of natural enemies in the presence of flower strips. This enhancement was not apparent in the control treatment lacking a flower strip. Accordingly, whilst proportion of cropland in the landscape did not have a direct overall effect on in-field numbers of arthropods, in the presence of local-scale habitat management, more powerful enhancement of natural enemies occurred in moderately simple landscapes with 60-70% cropland than in moderately complex landscapes with 30-40% cropland. This range extends that in the work by

(Jonsson et al., 2015) in which a broadly similar effect of flower strips was

evident. Natural enemy numbers strongly enhanced by strips in their landscapes with crops comprised up to 48% but no such effects was evident in more complex landscapes with lower (as low as 2%) cropland. Whilst

(Jonsson et al., 2015) considered this result to be consistent with the intermediate landscape-complexity hypothesis (Tscharntke et al., 2012), it is noteworthy that the response they reported was linear rather bell shaped. The latter relationship would suggest that natural enemy responses peaked in landscapes of intermediate complexity and were attenuated in the simpler landscapes of their observed range. Accordingly, had the New Zealand kale crop work by (Jonsson et al., 2015) included still simpler landscapes with greater proportions of cropland, weakened responses of natural enemies to flower strips is predicted by the intermediate complexity hypothesis. The present study in Australian brassica vegetables, extend the proportion of cropland to close to 69% but still does not show diminution of response or even an approach to horizontal asymptote in the linear relationships.

Accordingly, the range of landscape types in which flower strips offer scope to promote conservation biological control is indicated to be wide.

The extension of effect of local-scale flower strips beyond the threshold of intermediate complexity hypothesis in crop dominated landscape may be a reflection of the ‘landscape-moderated concentration and dilution hypothesis’ of Tscharntke et al. (2012), which proposed that changes in landscapes

(revegetation or fragmentation) can have temporary and strong ecological effects to drive remarkably larger amount of arthropods. As a consequence, even very simple landscapes that were strongly dominated by crops with sufficient resources and donor habitat to support the adequately numerous assemblage of natural enemies responded strongly to the addition of a flower

strip to a crop. Conversely, in landscapes with extended area of the flowering resources, such as mass-flowering crops may dilute the ecosystem services which has been observed by Holzschuh et al. (2016) in European landscapes.

Aside from building on earlier studies of the effects of landscape composition on local management, the present study also extended to consider landscape connectivity and edge density. Against both of these landscape attributes, a trend similar to that observed for composition was apparent. Landscapes with low edge density of crops and a high cost, reflective of low amounts of woodland and pasture) both led to stronger responses of natural enemies to local flower strips. This enhancement was not apparent in the control treatment lacking a flower strip.

Natural enemy distribution within focal fields was significantly higher at the first sampling position immediately adjacent to flower strips and fell with increasing distance from the flowers, a response consistent with the strips serving as a source of nectar and pollen for natural enemies (Géneau et al.,

2012; Lu et al., 2014). The reduced abundance of natural enemies in subsequent interior positions is presumably a consequence of distance decay

(Boetzl, Schuele, Krauss, Steffan‐Dewenter, & Marini, 2020) and potentially changes in microclimatic conditions (Bianchi et al., 2006). Thus, although establishment of flowering plants at local-scale can enhance the natural enemy abundance, the spatial extent of benefit in the present study was extremely limited, differing from work in other systems where practically useful longer-range effects were evident (Gurr et al., 2016). Moreover, enhancement of natural enemies by flowering strips led to no measurable effects on pest densities. Rather, pests in general, and P. xylostella specifically, were affected by landscape characteristics independently of local 54

management. Complex landscapes can increased the natural enemy abundance and pest suppression stability within the crop fields regardless of local-scale properties, an effect in broad agreement with (Boetzl et al., 2020;

Rusch et al., 2016). This result also emphasizes that in complex landscapes ecological services are may be ubiquitous, reducing the additional value of local-scale management such as flower strips (Boetzl et al., 2020; Tscharntke et al., 2012). Furthermore, the distribution of natural enemies may depend on the dispersal ability and taxon-specific characteristics of the species (Gallé et al., 2019; Perović et al., 2010).

3.5 Conclusions The overarching aim of this study was to investigate the scope for local-scale habitat management with flowering plants to can shape the pest and natural enemy abundance in crop fields located in contrasting landscape structures.

Complex landscape structure with substantial amounts of less disturbed habitat of pasture and woodlands reduced the pest abundance and increased natural enemy abundance, respectively, regardless of local habitat management with flowering strips. These effects suggest the less possibility of habitat management with flowering plants to be effective in complex landscapes. Local-scale habitat management with flowering strips observed to increase natural enemy abundance in crop fields located in simple landscapes dominated by croplands. Thus, ecosystem services delivered by non-crop flowering plants such as natural enemy enhancement is maximised in simple landscapes. Findings strongly suggest the consideration of landscape structure before the implementation of such flowering strips at the local-scale. Payback for time and money invested in establishing flowering strips is unlikely in complex landscapes that might be benchmarked as less

than 50% cropland. Above this threshold, progressively higher levels of benefit are likely to accrue.

Chapter 4. General Discussion and Conclusion

The primary aim of this thesis was to evaluate the impact of landscape properties on local, field scale properties in mediating the assemblage of herbivore pests and natural enemies in brassica vegetable fields. Across a range of Australian landscapes, three key landscape properties (landscape composition, edge density, and landscape connectivity) were investigated to determine the effect of overall landscape structure. Firstly, the impact of the landscape properties on the effect of immediately adjacent land uses on in- field numbers of pests and natural enemies (Figure 4.1) was investigated (Chapter 2). The presence of natural or semi-natural vegetation, that is a habitat of natural enemies, immediately adjacent to crop fields is considered advantageous for natural enemy-mediated pest suppression (Thomson & Hoffmann, 2013). However, conservation of natural or semi-natural resources in close vicinity of the crop fields is uncommon in intensified agricultural landscapes, particularly in regions with limited land area where all the available space is used for cropping. Therefore, interventions such as local- scale habitat management using fast-growing nectar containing flowering plants can supplement the absence of the natural enemy resources adjacent to the crop fields. Following this, the impacts of landscape properties on the success of the local-scale habitat management by establishing flower strips (Figure 4.1) were investigated in Chapter 3. This concluding chapter considers implications of the key findings from the research in relation to current knowledge in habitat management at local and landscape scales. The limitations of the thesis and the recommendation for further investigations on the local and landscape-scale ecological processes are also discussed.

Figure 4.1 Illustrates positions of focal field (location of local habitat manipulation and other local scale properties), and landscape in the agroecosystem.

4.1 Arthopod abundance In both the research chapters (Chapters 2 and 3) the abundance of pest and natural enemy arthropods were quite low, which could be a subject of argument. However, the precision and significance of the results indicated the sufficient statistical power of the sample sizes.

4.2 Landscape properties affect field adjacent vegetation to mediate arthropod abundance Pest suppression, especially, suppression of specialist herbivores is more likely in complex landscapes with a greater amount of natural or semi-natural habitats. According to previous studies, pest suppression services delivered by natural or semi-natural habitats in the landscape are primarily due to enhanced natural enemy abundance and activity (Perez‐Alvarez, Nault, & Poveda, 2018; Rand et al., 2014). Significant reductions of two brassica specialist herbivores Plutella xylostella and Brevicoryne brassicae were revealed in Chapter 2, but this was not correlated with natural enemy abundance or activity. Reduction of P. xylostella was due to the effects of woody vegetation and pastures immediately adjacent to the crop when landscape-scale woody vegetation (in terms of landscape composition and connectivity) and pastures (in terms of landscape connectivity) strengthened 58

their effects. Suppression of B. brassicae was evident when the effects of field adjacent non-host crops were strengthened by the croplands in the landscape (in terms of edge density).

The non-significant response of natural enemy populations to the similar landscape-scale circumstances suggests non-natural enemy-mediated effects of these land-uses to suppress specialist pests (Peterson, Davis, Higley, & Fernandes, 2009; Root, 1973). Increasing land-use complexity in the landscape may augment pest suppression by manipulating pest behaviour (Podeva, Diaz, Espinosa, Obregon, & Ramirez, 2019) by reducing the concentration (Rand et al., 2014) and disrupting phytochemical cues (Finch & Collier, 2012) of resource crops. Although specialist herbivores are capable of detecting the phytochemical properties of a particular crop type (Poelman, Galiart, Raaijmakers, van Loon, & van Dam, 2008), they are susceptible to disruption by phytochemical cues from other plants. Thus, maintenance of land-use complexity at local and landscape scales can reduce the invasion of specialist herbivores.

Substantial amounts of woody vegetation in the landscape can support natural enemies and enhance their in-field abundance (Bianchi et al., 2005; Perović et al., 2010; Thomson & Hoffmann, 2010; Thomson et al., 2010). However, complex landscape with a relatively high amount of woody vegetation may dilute the effect of woody vegetation at the local-scale or immediately adjacent to the crop field to enhance the natural enemy abundance in crop fields (Tscharntke, Rand, & Bianchi, 2005). Such an outcome was revealed in Chapter 2, when increasing landscape complexity with higher edge density of woody vegetation diminished the effects of adjacent woody vegetation on predatory ladybirds. Conversely, the effects of adjacent woody vegetation on predatory lady beetles were increased with the increasing amount of crop fields (in terms of connectivity) in the landscape. This finding suggests that local-scale non-crop habitats are most effective to deliver the ecosystem services of natural enemy enhancement in crop dominated simple landscapes (Batary, Baldi, Kleijn, & Tscharntke, 2011).

4.3 Landscape properties affect responses of arthropods to local habitat management Habitat management by establishing non-crop flowering plants at the local, field scale (Chapter 3) suggested that increasing complexity in the landscape can override the need to provide floral resources to suppress pest populations (Chaplin-Kramer & Kremen, 2012; Tscharntke, Rand, et al., 2005). Increased amount of pastures in the landscape significantly reduced the in-field abundance of total pest populations regardless of local-scale habitat management with non-crop flowering plants. Among the total pest populations, P. xylostella revealed a similar response to landscape-scale pastures. This response indicates that habitat management with flowering strips is not useful in terms of pest suppression when surrounding landscapes have high land-use complexity, which provides partial support of the intermediate landscape complexity hypotheses of Tscharntke et al. (2012) which proposed that for the availability of resources ecosystem services are high everywhere in complex landscapes which may make local-scale management practices redundant (but see below). While all the three flowering plants experimented in this study were previously demonstrated to attract or subsidize P. xylostella adults to some extent (Badenes-Perez, Reichelt, Gershenzon, & Heckel, 2014; Y. Chen et al., 2020; De Groot, Winkler, & Potting, 2005), the interaction with landscape composition of pasture had no significant influence on the pest. Possible explanations can be the disruption of resource concentration effect (Root, 1973) or the erosion of chemical cues released by the experimental flowering plants, to attract the adult P. xylostella, for the increasing amount of pastures in the landscape (Tajmiri, Fathi, Golizadeh, & Nouri-Ganbalani, 2017). Although local-scale habitat management with floral resources is not strong in complex landscapes, other habitat management practices such as planting of trap plants have been proposed for complex landscapes to enhance pest suppression and crop yield (Podeva et al., 2019). The results of Chapter 2 in terms of suppressing specialist herbivore P. xylostella and B. brassicae are in line with the findings of Podeva et al. (2019) and emphasize the maintenance of natural or semi-natural habitats, or crop diversity at local field scale for suppressing specialist herbivore pests when the surrounding landscapes are complex.

Habitat management with flowering strips in Chapter 3 promoted natural enemy abundance in crop dominated landscapes. Natural enemy abundance was significantly increased in fields with flowering strips along the margin compared to control treatment of no flowering plants in simple landscapes dominated by an increasing amount of croplands. The relative scarcity of resource habitats in crop dominated landscapes bound the natural enemy populations within the limited resources and the provision of local-scale resources such as woody vegetation or flowering plants become more effective to draw natural enemies in crop fields (Jonsson et al., 2015; Tscharntke, Klein, et al., 2005).

The relationships of crop dominated landscapes and local-scale properties to enhance natural enemies are not precisely consistent with the intermediate landscape complexity hypothesis of Tscharntke et al. (2012). According to the intermediate landscape complexity hypothesis, landscapes with intermediate complexity have the potential to make the local-scale resources successful which reaches the peak when landscapes have 50% complexity (50% crop and 50% non-crop habitats). The findings of Chapters 2 and 3 of this thesis indicate that the threshold of intermediate landscape complexity may fluctuate based on the regions where the focal field is located (Tscharntke et al., 2012). In Australian landscapes, the threshold of intermediate landscape complexity has been revealed to be extended beyond the proposed threshold of the intermediate landscape complexity hypothesis.

4.4 Inconsistent effect of landscape properties on pests The findings in Chapters 2 and 3 highlighted that pest suppression is stronger in complex landscapes, particularly, with an increasing amount of pastures and woody types of habitats. However, in Chapter 2 of this thesis, landscapes reversed effects on the abundance of two aphid species (B. brassicae and Myzus persicae) sampled from field centres where the effects of local-scale properties (adjacent land-uses) were low and regression models were devoid of local-scale effects. Increasing amounts of woody vegetation increased the abundance of these aphid species at the centres of the focal fields.

Such inconsistent effect of landscapes with and without considering local- scale effects was proposed by Karp et al. (2018) and evident in the study of Saqib et al. (2020). Karp et al. (2018) discussed that local-scale properties can 61

conceal the effect of wider landscape properties and highlighted the necessity of including local-scale factors in landscape models to justify the understanding of overall ecological processes in herbivore pest management.

4.5 Limitations In Chapter 2, information on distance from centre to each type of adjacent land-use was not available, however, the minimum size of the fields was 20×20 m2.

Recent studies demonstrate that natural enemies can travel further than assumed before (Heimpel, 2019). Consideration of spatial scales larger than the highest scale (i.e. 5000 m) considered in this thesis could better explain the dispersal of natural enemies in the landscape and arrival in the crop fields. However, to maintain adequate sample size and spatial independence among the focal sites, a maximum of 5000 m spatial scale was considered. Furthermore, lower spatial scales, as investigated by Thomson and Hoffmann (2013), than the minimum spatial scale (250 m) considered in this thesis could also provide significant directions on natural enemy dispersal of the less vagile species across the landscape.

Neither Chapter 2 nor Chapter 3 included the contribution of natural enemy enhancement and pest suppression in crop yield. The sites located in different states in Australia usually have diverse climatic conditions, ranging from tropical to temperate, which may have impacts on the arthropod community in focal fields. Impacts of the climatic factors were not included in this thesis.

4.6 Future directions As landscape-scale properties demonstrate strong influences on the local- scale properties to enhance natural enemy abundance and suppress pest populations in brassica vegetable fields, future investigations should consider the effects of both the scales. More specific investigation using the mark- recapture technique (Perović, Gurr, Simmons, & Raman, 2011), advanced geostatistical tools such as kriging (Sciarretta & Trematerra, 2006) or molecular techniques (Heimoana et al., 2017) would provide better directions in habitat management at local and landscape scales. In addition, impacts of other factors such as farming practices (organic vs conventional), pollinators’

services, climatic and soil nutritional conditions, and their contribution towards crop yield need to be considered.

4.7 Conclusions Landscape structure has been established as a significant factor that influences the strength of pest management in agroecosystems. For the maximization of these services of wider landscapes, the effects of local-scale properties need to be included (Karp et al., 2018). This thesis established significant impacts of landscape properties in relation to local-scale properties in mediating pest and natural enemy populations in brassica vegetable fields. The results reveal that landscape properties can strongly influence the local- scale properties to enhance natural enemy abundance and pest suppression. Before implementing local-scale habitat management approaches, particularly, with non-crop flowering resources, it is essential to recognize the properties of wider landscapes.

Habitat management with flowering plants or revegetation of farmscape with woody vegetation is demonstrated to be significant in enhancing natural enemy abundance only in crop dominated landscapes. In complex landscapes, natural enemy enhancement and pest suppression are more ubiquitous and the implementation of flowering or woody resources is largely redundant.

References Amoabeng, B., Asare, K., Asare, O., Mochiah, M., Adama, I., Fening, K., & Gurr, G. M. (2017). Pesticides Use and Misuse in Cabbage Brassica oleracea var. capitata L. (Cruciferae) Production in Ghana: The Influence of Farmer Education and Training. Journal of Agriculture and Ecology Research International, 10(1), 1-9. doi:10.9734/jaeri/2017/30128 Anjum-Zubair, M., Schmidt-Entling, M. H., Querner, P., & Frank, T. (2010). Influence of within-field position and adjoining habitat on carabid beetle assemblages in winter wheat. Agricultural and Forest Entomology, 301-306. doi:10.1111/j.1461-9563.2010.00479.x Araj, S.-E., & Wratten, S. D. (2015). Comparing existing weeds and commonly used insectary plants as floral resources for a parasitoid. Biological Control, 81, 15-20. doi:10.1016/j.biocontrol.2014.11.003 Australian Horticulture Statistics Handbook Vegetables. (2017/18). Retrieved from https://www.horticulture.com.au/growers/help-your- business-grow/research-reports-publications-fact-sheets-and- more/australian-horticulture-statistics-handbook/ Badenes-Perez, F. R., Reichelt, M., Gershenzon, J., & Heckel, D. G. (2014). Using plant chemistry and insect preference to study the potential of Barbarea (Brassicaceae) as a dead-end trap crop for diamondback moth (Lepidoptera: Plutellidae). Phytochemistry, 98, 137-144. doi:10.1016/j.phytochem.2013.11.009 Baillod, A. B., Tscharntke, T., Clough, Y., Batáry, P., & Marini, L. (2017). Landscape-scale interactions of spatial and temporal cropland heterogeneity drive biological control of cereal aphids. Journal of Applied Ecology, 54(6), 1804-1813. doi:10.1111/1365-2664.12910 Balmer, O., Pfiffner, L., Schied, J., Willareth, M., Leimgruber, A., Luka, H., & Traugott, M. (2013). Noncrop flowering plants restore top-down herbivore control in agricultural fields. Ecol Evol, 3(8), 2634-2646. doi:10.1002/ece3.658 Batary, P., Baldi, A., Kleijn, D., & Tscharntke, T. (2011). Landscape- moderated biodiversity effects of agri-environmental management: a meta-analysis. Proceedings of the Royal Society B: Biological Sciences, 278(1713), 1894-1902. 64

Beduschi, T., Kormann, U. G., Tscharntke, T., & Scherber, C. (2018). Spatial community turnover of pollinators is relaxed by semi-natural habitats, but not by mass-flowering crops in agricultural landscapes. Biological Conservation, 221, 59-66. doi:10.1016/j.biocon.2018.01.016 Bianchi, F. J. J. A., Booij, C. J., & Tscharntke, T. (2006). Sustainable pest regulation in agricultural landscapes: a review on landscape composition, biodiversity and natural pest control. Proc Biol Sci, 273(1595), 1715-1727. doi:10.1098/rspb.2006.3530 Bianchi, F. J. J. A., van Wingerden, W. K. R. E., Griffioen, A. J., van der Veen, M., van der Straten, M. J. J., Wegman, R. M. A., & Meeuwsen, H. A. M. (2005). Landscape factors affecting the control of Mamestra brassicae by natural enemies in Brussels sprout. Agriculture, Ecosystems & Environment, 107(2-3), 145-150. doi:10.1016/j.agee.2004.11.007 Bischoff, A., Pollier, A., Lamarre, E., Salvadori, O., Cortesero, A.-M., Le Ralec, A., . . . Jaloux, B. (2016). Effects of spontaneous field margin vegetation and surrounding landscape on Brassica oleracea crop herbivory. Agriculture, Ecosystems & Environment, 223, 135-143. doi:10.1016/j.agee.2016.02.029 Blitzer, E. J., Dormann, C. F., Holzschuh, A., Klein, A.-M., Rand, T. A., & Tscharntke, T. (2012). Spillover of functionally important organisms between managed and natural habitats. Agriculture, Ecosystems & Environment, 146(1), 34-43. doi:10.1016/j.agee.2011.09.005 Bloomfield, L. S. P., McIntosh, T. L., & Lambin, E. F. (2020). Habitat fragmentation, livelihood behaviors, and contact between people and nonhuman primates in Africa. Landscape Ecology, 35(4), 985-1000. doi:10.1007/s10980-020-00995-w Boetzl, F. A., Schuele, M., Krauss, J., Steffan‐Dewenter, I., & Marini, L. (2020). Pest control potential of adjacent agri‐environment schemes varies with crop type and is shaped by landscape context and within‐ field position. Journal of Applied Ecology, 57(8), 1482-1493. doi:10.1111/1365-2664.13653 Bolker, B. M., Brooks, M. E., Clark, C. J., Geange, S. W., Poulsen, J. R., Stevens, M. H., & White, J. S. (2009). Generalized linear mixed

models: a practical guide for ecology and evolution. Trends Ecol Evol, 24(3), 127-135. doi:10.1016/j.tree.2008.10.008 Brennan, E. B. (2016). Agronomy of strip intercropping broccoli with alyssum for biological control of aphids. Biological Control, 97, 109- 119. doi:10.1016/j.biocontrol.2016.02.015 Bukovinszky, T., Potting, R. P. J., Clough, Y., van Lenteren, J. C., & Vet, L. E. M. (2005). The role of pre- and post- alighting detection mechanisms in the responses to patch size by specialist herbivores. OIKOS, 109, 435-446. Burgman, M. A., & Fox, J. C. (2003). Bias in species range estimates from minimum convex polygons: implications for conservation and options for improved planning. Animal Conservation, 6(1), 19-28. doi:10.1017/s1367943003003044 Cai, X., Wu, Z., & Cheng, J. (2013). Using kernel density estimation to assess the spatial pattern of road density and its impact on landscape fragmentation. International Journal of Geographical Information Science, 27(2), 222-230. doi:10.1080/13658816.2012.663918 Carriere, Y., Goodell, P. B., Ellers-Kirk, C., Larocque, G., Dutilleul, P., Naranjo, S. E., & Ellsworth, P. C. (2012). Effects of local and landscape factors on population dynamics of a cotton pest. PLoS One, 7(6), e39862. doi:10.1371/journal.pone.0039862 Carvalho, F. P. (2017). Pesticides, environment, and food safety. Food and Energy Security, 6(2), 48-60. doi:10.1002/fes3.108 Chaplin-Kramer, R., de Valpine, P., Mills, N. J., & Kremen, C. (2013). Detecting pest control services across spatial and temporal scales. Agriculture, Ecosystems & Environment, 181, 206-212. doi:10.1016/j.agee.2013.10.007 Chaplin-Kramer, R., & Kremen, C. (2012). Pest control experiments show benefits of complexity at landscape and local scales. Ecological Applications, 22(7), 1936-1948. doi:10.1890/11-1844.1 Chaplin-Kramer, R., O'Rourke, M. E., Blitzer, E. J., & Kremen, C. (2011). A meta-analysis of crop pest and natural enemy response to landscape complexity. Ecol Lett, 14(9), 922-932. doi:10.1111/j.1461- 0248.2011.01642.x

Chardon, J. P., Adriaensen, F., & Matthysen, E. (2003). Incorporating landscape elements into a connectivity measure: a case study for the Speckled wood butterfly (Pararge aegeria L.). Landscape Ecology, 18(6), 561-573. doi:10.1023/a:1026062530600 Chen, Y.-Z., Lin, L., Wang, C.-W., Yeh, C.-C., & Hwang, S.-Y. (2004). Response of Two Pieris (Lepidoptera: Pieridae) Species to Fertilization of a Host Plant. Zoological Studies, 43(4), 778-786. Chen, Y., Mao, J., Reynolds, O. L., Chen, W., He, W., You, M., & Gurr, G. M. (2020). Alyssum (Lobularia maritima) selectively attracts and enhances the performance of Cotesia vestalis, a parasitoid of Plutella xylostella. Scientific Reports, 10(1), 1-9. Cong, R.-G., Smith, H. G., Olsson, O., & Brady, M. (2014). Managing ecosystem services for agriculture: Will landscape-scale management pay? Ecological Economics, 99, 53-62. doi:10.1016/j.ecolecon.2014.01.007 de Araújo, W. S., & do Espírito-Santo Filho, K. (2012). Edge effect benefits galling insects in the Brazilian Amazon. Biodiversity and Conservation, 21(11), 2991-2997. doi:10.1007/s10531-012-0333-z De Groot, M., Winkler, K., & Potting, R. (2005). Testing the potential of White mustard (Sinapis alba) and Sweet alyssum (Lobularia maritima) as trap crops for the Diamondback moth Plutella xylostella. Proc. Neth. Entomol. Soc. Meet, 16, 117-123. Ditner, N., Balmer, O., Beck, J., Blick, T., Nagel, P., & Luka, H. (2013). Effects of experimentally planting non-crop flowers into cabbage fields on the abundance and diversity of predators. Biodiversity and Conservation, 22(4), 1049-1061. doi:10.1007/s10531-013-0469-5 Dong, Z., Ouyang, F., Lu, F., & Ge, F. (2015). Shelterbelts in agricultural landscapes enhance ladybeetle abundance in spillover from cropland to adjacent habitats. BioControl, 60(3), 351-361. doi:10.1007/s10526- 015-9648-5 Fahrig, L., & Triantis, K. (2013). Rethinking patch size and isolation effects: the habitat amount hypothesis. Journal of Biogeography, 40(9), 1649- 1663. doi:10.1111/jbi.12130 FAO. (2003). World agriculture: towards 2015/2030. An FAO perspective. Retrieved from http://www.fao.org/3/y4252e/y4252e.pdf 67

FAO. (2018). Land use indicators. In F. a. A. O. o. t. U. N. FAOSTAT (Ed.). Finch, S., & Collier, R. H. (2012). The influence of host and non-host companion plants on the behaviour of pest insects in field crops. Entomologia Experimentalis et Applicata, 142(2), 87-96. doi:10.1111/j.1570-7458.2011.01191.x Fitton, N., Alexander, P., Arnell, N., Bajzelj, B., Calvin, K., Doelman, J., . . . Smith, P. (2019). The vulnerabilities of agricultural land and food production to future water scarcity. Global Environmental Change, 58. doi:10.1016/j.gloenvcha.2019.101944 Fletcher, R. J., Didham, R. K., Banks-Leite, C., Barlow, J., Ewers, R. M., Rosindell, J., . . . Haddad, N. M. (2018). Is habitat fragmentation good for biodiversity? Biological Conservation, 226, 9-15. doi:10.1016/j.biocon.2018.07.022 Francisco, M., Tortosa, M., Martínez-Ballesta, M. d. C., Velasco, P., García- Viguera, C., & Moreno, D. A. (2017). Nutritional and phytochemical value of Brassica crops from the agri-food perspective. Annals of Applied Biology, 170(2), 273-285. doi:10.1111/aab.12318 Furlong, M. J., Wright, D. J., & Dosdall, L. M. (2013). Diamondback moth ecology and management: problems, progress, and prospects. Annu Rev Entomol, 58, 517-541. doi:10.1146/annurev-ento-120811- 153605 Fusser, M. S., Pfister, S. C., Entling, M. H., & Schirmel, J. (2017). Effects of field margin type and landscape composition on predatory carabids and slugs in wheat fields. Agriculture, Ecosystems & Environment, 247, 182-188. doi:10.1016/j.agee.2017.06.030 Gabryś, B., & Pawluk, M. (1999). Acceptability of different species of Brassicaceae as hosts for the cabbage aphid. In S. J. Simpson, A. J. Mordue, & J. Hardie (Eds.), Proceedings of the 10th International Symposium on Insect-Plant Relationships (pp. 105-109). Dordrecht: Springer Netherlands. Gallé, R., Happe, A. K., Baillod, A. B., Tscharntke, T., Batáry, P., & MacIvor, J. S. (2019). Landscape configuration, organic management, and within‐field position drive functional diversity of spiders and carabids. Journal of Applied Ecology, 56(1), 63-72. doi:10.1111/1365-2664.13257 68

Géneau, C. E., Wäckers, F. L., Luka, H., Daniel, C., & Balmer, O. (2012). Selective flowers to enhance biological control of cabbage pests by parasitoids. Basic and Applied Ecology, 13(1), 85-93. doi:10.1016/j.baae.2011.10.005 Gonthier, D. J., Ennis, K. K., Farinas, S., Hsieh, H. Y., Iverson, A. L., Batary, P., . . . Perfecto, I. (2014). Biodiversity conservation in agriculture requires a multi-scale approach. Proc Biol Sci, 281(1791), 20141358. doi:10.1098/rspb.2014.1358 González-Chang, M., Tiwari, S., Sharma, S., & Wratten, S. D. (2019). Habitat Management for Pest Management: Limitations and Prospects. Annals of the Entomological Society of America. doi:10.1093/aesa/saz020 Goodwin, B. J., & Fahrig, L. (2002). How does landscape structure influence landscape connectivity. OIKOS, 99(552-570). Gurr, G. M., Lu, Z., Zheng, X., Xu, H., Zhu, P., Chen, G., . . . Catindig, J. L. (2016). Multi-country evidence that crop diversification promotes ecological intensification of agriculture. Nature Plants, 2(3), 1-4. Gurr, G. M., Reynolds, O. L., Johnson, A. C., Desneux, N., Zalucki, M. P., Furlong, M. J., . . . You, M. (2018). Landscape ecology and expanding range of biocontrol agent taxa enhance prospects for diamondback moth management. A review. Agronomy for Sustainable Development, 38(3). doi:10.1007/s13593-018-0500-z Gurr, G. M., Scarratt, S. L., Wratten, S. D., Berndt, L., & Irvin, N. (2004). Ecological engineering, habitat manipulation and pest management. In Ecological Engineering for Pest Management : Advances in Habitat Manipulation for Arthropods. Australia: CSIRO Publishing. Gurr, G. M., Wratten, S. D., Landis, D. A., & You, M. (2017). Habitat Management to Suppress Pest Populations: Progress and Prospects. Annu Rev Entomol, 62, 91-109. doi:10.1146/annurev-ento-031616- 035050 Hatt, S., Uyttenbroeck, R., Lopes, T., Chen, J. L., Piqueray, J., Monty, A., & Francis, F. (2017). Effect of flower traits and hosts on the abundance of parasitoids in perennial multiple species wildflower strips sown within oilseed rape (Brassica napus) crops. Arthropod-Plant Interactions, 12(6), 787-797. doi:10.1007/s11829-017-9567-8 69

Heimoana, V., Pilkington, L. J., Raman, A., Mitchell, A., Nicol, H. I., Johnson, A. C., & Gurr, G. M. (2017). Integrating spatially explicit molecular and ecological methods to explore the significance of non- crop vegetation to predators of brassica pests. Agriculture, Ecosystems & Environment, 239, 12-19. doi:10.1016/j.agee.2017.01.008 Heimpel, G. E. (2019). Linking parasitoid nectar feeding and dispersal in conservation biological control. Biological Control, 132, 36-41. Hein, L., van Koppen, K., de Groot, R. S., & van Ierland, E. C. (2006). Spatial scales, stakeholders and the valuation of ecosystem services. Ecological Economics, 57(2), 209-228. doi:10.1016/j.ecolecon.2005.04.005 Holzschuh, A., Dainese, M., Gonzalez-Varo, J. P., Mudri-Stojnic, S., Riedinger, V., Rundlof, M., . . . Steffan-Dewenter, I. (2016). Mass- flowering crops dilute pollinator abundance in agricultural landscapes across Europe. Ecol Lett, 19(10), 1228-1236. doi:10.1111/ele.12657 Holzschuh, A., Steffan-Dewenter, I., & Tscharntke, T. (2010). How do landscape composition and configuration, organic farming and fallow strips affect the diversity of bees, wasps and their parasitoids? J Anim Ecol, 79(2), 491-500. doi:10.1111/j.1365-2656.2009.01642.x Hoss, S. K., Guyer, C., Smith, L. L., & Schuett, G. W. (2010). Multiscale Influences of Landscape Composition and Configuration on the Spatial Ecology of Eastern Diamond-backed Rattlesnakes (Crotalus adamanteus). Journal of Herpetology, 44(1), 110-123. doi:10.1670/08-277.1 Hu, G. Y., Mitchell, E. R., & Okine, J. S. (1997). Diamondback moth (Lepidoptera: Plutellidae) in cabbage: Influence of initial immigration sites on population distribution, density and larval parasitism. Journal of Entomological Science, 32(1), 56-71. doi:10.18474/0749-8004- 32.1.56 Idris, A. B., & Grafius, E. (1996). Effects of wild and cultivated host plants on oviposition, survival, and development of diamondback moth (Lepidoptera: Plutellidae) and its parasitoid Diadegma insulare (Hymenoptera: Ichneumonidae). Environmental Entomology, 25(4), 825-833. 70

Jonsson, M., Straub, C. S., Didham, R. K., Buckley, H. L., Case, B. S., Hale, R. J., . . . Clough, Y. (2015). Experimental evidence that the effectiveness of conservation biological control depends on landscape complexity. Journal of Applied Ecology, 52(5), 1274-1282. doi:10.1111/1365-2664.12489 Karp, D. S., Chaplin-Kramer, R., Meehan, T. D., Martin, E. A., DeClerck, F., Grab, H., . . . Martínez-Salinas, A. (2018). Crop pests and predators exhibit inconsistent responses to surrounding landscape composition. Proceedings of the national academy of sciences, 115(33), E7863- E7870. Kross, S. M., Martinico, B. L., Bourbour, R. P., Townsend, J. M., McColl, C., & Kelsey, T. R. (2020). Effects of field and landscape scale habitat on insect and bird damage to sunflowers. Frontiers in Sustainable Food Systems, 4(40). doi:10.3389/fsufs.2020.00040 Lamichhane, J. R., Dachbrodt-Saaydeh, S., Kudsk, P., & Messéan, A. (2016). Toward a reduced reliance on conventional pesticides in European agriculture. Plant Disease, 100(1), 10-24. Landis, D. A. (2017). Designing agricultural landscapes for biodiversity- based ecosystem services. Basic and Applied Ecology, 18, 1-12. doi:10.1016/j.baae.2016.07.005 Lee, J. C., & Heimpel, G. E. (2005). Impact of flowering buckwheat on Lepidopteran cabbage pests and their parasitoids at two spatial scales. Biological Control, 34(3), 290-301. doi:10.1016/j.biocontrol.2005.06.002 Li, Z., Zalucki, M. P., Bao, H., Chen, H., Hu, Z., Zhang, D., . . . Feng, X. (2012). Population Dynamics and “Outbreaks” of Diamondback Moth (Lepidoptera: Plutellidae) in Guangdong Province, China: Climate or Failure of Management? Journal of Economic Entomology, 105(3), 739-752. doi:10.1603/ec11384 Loos, J., Batáry, P., Grass, I., Westphal, C., Bänsch, S., Baillod, A. B., . . . Tscharntke, T. (2019). Vulnerability of Ecosystem Services in Farmland Depends on Landscape Management. In Atlas of Ecosystem Services (pp. 91-96). Lu, Z. X., Zhu, P. Y., Gurr, G. M., Zheng, X. S., Read, D. M., Heong, K. L., . . . Xu, H. X. (2014). Mechanisms for flowering plants to benefit 71

arthropod natural enemies of insect pests: prospects for enhanced use in agriculture. Insect Sci, 21(1), 1-12. doi:10.1111/1744-7917.12000 Lymburner, L., Tan, P., Mueller, N., Thackway, R., Lewis, A., Thankappan, M., . . . Senarath, U. (2010). The National Dynamic Land Cover Dataset. In. Macfadyen, S., & Muller, W. (2013). Edges in Agricultural Landscapes: Species Interactions and Movement of Natural Enemies. PLOS ONE, 8(3), e59659. doi:10.1371/journal.pone.0059659 Maguire, D. Y., James, P. M. A., Buddle, C. M., & Bennett, E. M. (2015). Landscape connectivity and insect herbivory: A framework for understanding tradeoffs among ecosystem services. Global Ecology and Conservation, 4, 73-84. doi:10.1016/j.gecco.2015.05.006 Martin, E. A., Dainese, M., Clough, Y., Báldi, A., Bommarco, R., Gagic, V., . . . Kovács‐Hostyánszki, A. (2019). The interplay of landscape composition and configuration: new pathways to manage functional biodiversity and agroecosystem services across Europe. Ecology letters, 22(7), 1083-1094. Martin, E. A., Reineking, B., Seo, B., & Steffan-Dewenter, I. (2015). Pest control of aphids depends on landscape complexity and natural enemy interactions. PeerJ, 3, e1095. doi:10.7717/peerj.1095 McKenzie, C. L., Anderson, P. K., & Villarreal, N. (2004). An extensive survey of Bemisia tabaci (Homoptera: Aleyrodidae) in agricultural ecosystems in Florida. The Florida Entomologist, 87(3), 403-407. Nilsson, U., Porcel, M., Swiergiel, W., & Wivstad, M. (2016). Habitat manipulation- as a pest management tool in vegetable and fruit cropping systems, with the focus on insects and mites. Sweden: SLU, EPOK - Centre for Organic Food & Farming. Nilsson, U., Rännbäck, L. M., Anderson, P., & Rämert, B. (2012). Herbivore response to habitat manipulation with floral resources: a study of the cabbage root fly. Journal of Applied Entomology, 136(7), 481-489. doi:10.1111/j.1439-0418.2011.01685.x O'Rourke, M. E., Rienzo-Stack, K., & Power, A. G. (2011). A multi‐scale, landscape approach to predicting insect populations in agroecosystems. Ecological Applications, 21(5), 1782-1791.

Orre, G. U. S., Wratten, S. D., Jonsson, M., Simpson, M., & Hale, R. (2013). ‘Attract and reward’: Combining a herbivore-induced plant volatile with floral resource supplementation – Multi-trophic level effects. Biological Control, 64(2), 106-115. doi:10.1016/j.biocontrol.2012.10.003 Pandey, S., & Gurr, G. M. (2019). Conservation biological control using Australian native plants in a brassica crop system: seeking complementary ecosystem services. Agriculture, Ecosystems & Environment, 280, 77-84. doi:10.1016/j.agee.2019.04.018 Pandey, S., Rahman, A., & Gurr, G. M. (2018). Australian native flowering plants enhance the longevity of three parasitoids of brassica pests. Entomologia Experimentalis et Applicata, 166(4), 265-276. doi:10.1111/eea.12668 Perez‐Alvarez, R., Nault, B. A., & Poveda, K. (2018). Contrasting effects of landscape composition on crop yield mediated by specialist herbivores. Ecological Applications, 28(3), 842-853. Perović, D. J., Gurr, G. M., Raman, A., & Nicol, H. I. (2010). Effect of landscape composition and arrangement on biological control agents in a simplified agricultural system: A cost–distance approach. Biological Control, 52(3), 263-270. doi:10.1016/j.biocontrol.2009.09.014 Perović, D. J., Gurr, G. M., Simmons, A. T., & Raman, A. (2011). Rubidium labelling demonstrates movement of predators from native vegetation to cotton. Biocontrol Science and Technology, 21(10), 1143-1146. doi:10.1080/09583157.2011.607232 Peterson, R. K., Davis, R. S., Higley, L. G., & Fernandes, O. A. (2009). Mortality risk in insects. Environmental Entomology, 38(1), 2-10. Pither, J., & Taylor, P. D. (1998). An experimental assessment of landscape connectivity. OIKOS, 83, 166-174. Plećaš, M., Gagić, V., Janković, M., Petrović-Obradović, O., Kavallieratos, N., Tomanović, Ž., . . . Ćetković, A. (2014). Landscape composition and configuration influence cereal aphid–parasitoid–hyperparasitoid interactions and biological control differentially across years. Agriculture, ecosystems & environment, 183, 1-10.

Podeva, K., Diaz, M. F., Espinosa, S., Obregon, D., & Ramirez, A. (2019). Landscape complexity and elevation affect the effectiveness of a local pest-management practice. Global Ecology and Conservation, 20, 1- 12. doi:10.1016/j.gecco.2019.e00763 Poelman, E. H., Galiart, R. J. F. H., Raaijmakers, C. E., van Loon, J. J. A., & van Dam, N. M. (2008). Performance of specialist and generalist herbivores feeding on cabbage cultivars is not explained by glucosinolate profiles. Entomologia Experimentalis et Applicata, 127(3), 218-228. doi:10.1111/j.1570-7458.2008.00700.x Prohens-Tomás, J., Nuez, F., & Carena, M. J. (2007). Vegetables I : Asteraceae, Brassicaceae, Chenopodicaceae, and Cucurbitaceae. New York, NY, UNITED STATES: Springer. Rand, T. A., Waters, D. K., Blodgett, S. L., Knodel, J. J., & Harris, M. O. (2014). Increased area of a highly suitable host crop increases herbivore pressure in intensified agricultural landscapes. Agriculture, Ecosystems & Environment, 186, 135-143. Ribeiro, A. L., & Gontijo, L. M. (2017). Alyssum flowers promote biological control of collard pests. BioControl, 62(2), 185-196. doi:10.1007/s10526-016-9783-7 Root, R. B. (1973). Organization of a plant-arthropod assiciation in simple and diverse habitats: the fauna of collards (Brassica oleracea). Ecological Monographs, 43(1), 95-124. Rösch, V., Tscharntke, T., Scherber, C., Batáry, P., & Osborne, J. (2013). Landscape composition, connectivity and fragment size drive effects of grassland fragmentation on insect communities. Journal of Applied Ecology, 50(2), 387-394. doi:10.1111/1365-2664.12056 Rusch, A., Chaplin-Kramer, R., Gardiner, M. M., Hawro, V., Holland, J., Landis, D., . . . Bommarco, R. (2016). Agricultural landscape simplification reduces natural pest control: A quantitative synthesis. Agriculture, Ecosystems & Environment, 221, 198-204. doi:10.1016/j.agee.2016.01.039 Saqib, H. S. A., Chen, J., Chen, W., Pozsgai, G., Akutse, K. S., Ashraf, M. F., . . . Gurr, G. M. (2020). Local management and landscape structure determine the assemblage patterns of spiders in vegetable fields. Scientific Reports, 10(1), 1-11. 74

Schmidt, M. H., Roschewitz, I., Thies, C., & Tscharntke, T. (2005). Differential effects of landscape and management on diversity and density of ground-dwelling farmland spiders. Journal of Applied Ecology, 42(2), 281-287. doi:10.1111/j.1365-2664.2005.01014.x Schmidt, M. H., Thies, C., Nentwig, W., & Tscharntke, T. (2008). Contrasting responses of arable spiders to the landscape matrix at different spatial scales. Journal of Biogeography, 35, 157-166. doi:10.1111/j.1365- 2699.2007.01774.x Sciarretta, A., & Trematerra, P. (2006). Geostatistical characterization of the spatial distribution of Grapholita molesta and Anarsia lineatella males in an agricultural landscape. Journal of Applied Entomology, 130(2), 73-83. doi:10.1111/j.1439-0418.2006.01034.x Shaw, M. J. P. (1970). Effects of population density on alienicolae of Aphis fabae Scop. Annals of Applied Biology, 65(2), 191-196. doi:10.1111/j.1744-7348.1970.tb04578.x Shields, M. W., Johnson, A. C., Pandey, S., Cullen, R., González- Chang, M., Wratten, S. D., & Gurr, G. M. (2019). History, current situation and challenges for conservation biological control. Biological Control, 131, 25-35. doi:10.1016/j.biocontrol.2018.12.010 Tajmiri, P., Fathi, S. A. A., Golizadeh, A., & Nouri-Ganbalani, G. (2017). Strip-intercropping canola with annual alfalfa improves biological control of Plutella xylostella (L.) and crop yield. International Journal of Tropical Insect Science, 37(3), 208-216. Thiele, J., & Markussen, B. (2012). Potential of GLMM in modelling invasive spread. CAB Reviews: Perspectives in Agriculture, Veterinary Science, Nutrition and Natural Resources, 16(7). doi:10.1079/PAVSNNR20127016 Thomson, L. J., & Hoffmann, A. A. (2010). Natural enemy responses and pest control: Importance of local vegetation. Biological Control, 52(2), 160-166. doi:10.1016/j.biocontrol.2009.10.008 Thomson, L. J., & Hoffmann, A. A. (2013). Spatial scale of benefits from adjacent woody vegetation on natural enemies within vineyards. Biological Control, 64(1), 57-65. doi:10.1016/j.biocontrol.2012.09.019

Thomson, L. J., McKenzie, J., Sharley, D. J., Nash, M. A., Tsitsilas, A., & Hoffmann, A. A. (2010). Effect of woody vegetation at the landscape scale on the abundance of natural enemies in Australian vineyards. Biological Control, 54(3), 248-254. doi:10.1016/j.biocontrol.2010.05.018 Tscharntke, T., Bommarco, R., Clough, Y., Crist, T. O., Kleijn, D., Rand, T. A., . . . Vidal, S. (2008). Reprint of “Conservation biological control and enemy diversity on a landscape scale” [Biol. Control 43 (2007) 294–309]. Biological Control, 45(2), 238-253. doi:10.1016/s1049- 9644(08)00082-0 Tscharntke, T., Karp, D. S., Chaplin-Kramer, R., Batáry, P., DeClerck, F., Gratton, C., . . . Larsen, A. (2016). When natural habitat fails to enhance biological pest control–Five hypotheses. Biological Conservation, 204, 449-458. Tscharntke, T., Klein, A. M., Kruess, A., Steffan-Dewenter, I., & Thies, C. (2005). Landscape perspectives on agricultural intensification and biodiversity - ecosystem service management. Ecology letters, 8(8), 857-874. doi:10.1111/j.1461-0248.2005.00782.x Tscharntke, T., Rand, T. A., & Bianchi, F. J. J. A. (2005). The landscape context of trophic interactions: insect spillover across the crop— noncrop interface. Annales Zoologici Fennici, 42(4), 421-432. Retrieved from http://www.jstor.org/stable/23735887 Tscharntke, T., Tylianakis, J. M., Rand, T. A., Didham, R. K., Fahrig, L., Batary, P., . . . Westphal, C. (2012). Landscape moderation of biodiversity patterns and processes - eight hypotheses. Biol Rev Camb Philos Soc, 87(3), 661-685. doi:10.1111/j.1469-185X.2011.00216.x Vizzari, M. (2011). Spatio-Temporal Analysis Using Urban-Rural Gradient Modelling and Landscape Metrics. Paper presented at the Computational Science and Its Applications - ICCSA 2011. Wang, Z., Wu, J., Shang, H., & Cheng, J. (2011). Landscape connectivity shapes the spread pattern of the rice water weevil: a case study from Zhejiang, China. Environ Manage, 47(2), 254-262. doi:10.1007/s00267-010-9595-y Wiedenmann, R. N., & Smith Jr., J. W. (1997). Attributes of natural enemies in ephemeral crop habitats. Biological control, 10(1), 16-22. 76

Woltz, J. M., Isaacs, R., & Landis, D. A. (2012). Landscape structure and habitat management differentially influence insect natural enemies in an agricultural landscape. Agriculture, Ecosystems & Environment, 152, 40-49. doi:10.1016/j.agee.2012.02.008 Zaller, J. G., Moser, D., Drapela, T., Schmöger, C., & Frank, T. (2008). Insect pests in winter oilseed rape affected by field and landscape characteristics. Basic and Applied Ecology, 9(6), 682-690. doi:10.1016/j.baae.2007.10.004 Zhao, Z.-H., Hui, C., He, D.-H., & Ge, F. (2013). Effects of position within wheat field and adjacent habitats on the density and diversity of cereal aphids and their natural enemies. BioControl, 58(6), 765-776. doi:10.1007/s10526-013-9536-9

Appendices Appendix 1.1 Statistical information on the outcomes of the models presented in Chapter 2. Outcome variables Fixed independent variables Estimate Standard Z value Wald Pr(>|z|) AIC error chisq Abundance of Cabbage Composition of pasture at 500 m scale 0.14994 0.11109 1.350 1.8218 0.17710 77.0 aphids at the field centre Composition of woody vegetation at 500 m 0.02503 0.01546 2.820 7.9504 0.00481 scale Composition of crop at 500 m scale 0.02877 0.01605 1.793 3.2141 0.07301 Abundance of Green peach Landscape connectivity for cost ratio r6 at 0.039684 0.009564 4.149 17.2180 3.33e-05 87.5 aphids at the field centre 1000m Landscape connectivity for cost ratio r6 at -0.029656 0.031996 -0.891 0.8136 0.37345 500m Landscape connectivity for cost ratio r3 at -0.016613 0.030903 -0.538 0.5160 0.59107 1000m Effect of adjacent woody Composition of woody vegetation at 250 m 0.011953 0.006875 1.739 3.0229 0.08210 179.7 vegetation on diamond back scale moth Composition of woody vegetation at 500 m -0.019604 0.012243 -1.985 3.9081 0.06284 scale Composition of woody vegetation at 1000 m 0.005610 0.008168 0.687 0.4718 0.49217 scale Composition of woody vegetation at 2500 m 0.020856 0.011294 1.847 3.4104 0.06478 scale Composition of woody vegetation at 5000 m -0.025011 0.008194 -3.052 9.3168 0.00227 scale 78

Effect of adjacent woody Landscape connectivity for r9 at 500 m scale -0.004991 0.001790 -2.788 7.7721 0.00531 157.7 vegetation on diamond back Landscape connectivity for r9 at 250 m scale 0.014912 0.039407 0.318 2.0729 0.05458 moth Effect of adjacent Interaction between adjacent brassica crops -0.000106 0.0001651 -0.646 9.8072 0.51823 664.8 vegetation on diamond back and landscape connectivity for r9 at 1000 m moth scale Interaction between adjacent non-brassica -0.000327 0.0001675 -1.953 0.05080 crops and landscape connectivity for r9 at 1000 m scale Interaction between adjacent pasture crops - 0.0001655 -1.066 0.28637 and landscape connectivity for r9 at 1000 m 0.0001765 scale Interaction between adjacent woody - 0.0001716 -2.826 0.00472 vegetation crops and landscape connectivity 0.0004850 for r9 at 1000 m scale Interaction between adjacent brassica crops 0.0016730 0.0025348 0.660 7.4056 0.50924 and landscape connectivity for r7 at 500 m scale Interaction between adjacent non-brassica 0.0037524 0.0025660 1.462 0.14364 crops and landscape connectivity for r7 at 500 m scale Interaction between adjacent woody 0.0022054 0.0025402 0.868 0.38528 vegetation crops and landscape connectivity for r7 at 500 m scale

Interaction between adjacent woody 0.0057108 0.0025966 1.199 0.06785 vegetation crops and landscape connectivity for r7 at 500 m scale Effect of adjacent non- Edge density of crops at 500 m scale 0.019759 0.011029 1.792 3.2101 0.0732 104.6 brassica crops on cabbage Edge density of crops at 1000 m scale -0.003295 0.001597 -2.063 4.2550 0.0391 aphids Edge density of crops at 2500 m scale 0.019399 0.015789 0.015789 1.5096 1.2290 Effect of adjacent woody Edge density of woody vegetation at 250 m 0.056405 0.021650 1.605 3.7873 0.097181 117.7 vegetation on lady beetles scale Edge density of woody vegetation at 500 m -0.010551 0.002738 -3.854 14.8496 0.000116 scale Edge density of woody vegetation at 1000 m 0.022486 0.021525 1.045 1.0913 0.296191 scale Edge density of woody vegetation at 2500 m -0.033786 0.060813 -0.556 0.3087 0.578504 scale Edge density of woody vegetation at 5000 m 0.013723 0.041752 0.329 0.1080 0.742403 scale Effect of adjacent woody Landscape connectivity for ratio r1 at 500 m 0.031581 0.016210 1.948 5.273 0.051483 98.432 vegetation on lady beetles scale Landscape connectivity for ratio r5 at 1000 m 0.040185 0.007629 5.267 9.408 0.001632 scale Landscape connectivity for ratio r9 at 1000 m 0.001543 0.010339 0.149 0.686 0.881451 scale Landscape connectivity for ratio r1 at 5000 m 0.020856 0.011294 1.847 3.483 0.061478 scale

Landscape connectivity for ratio r2 at 2500 m 0.011953 0.006875 1.739 2.774 0.082190 scale Effect of adjacent Interaction between adjacent brassica crops 0.000804 0.0030428 0.264 19.639 0.791 408.0 vegetation on lady beetles and edge density of croplands at 2500 m scale Interaction between adjacent non-brassica -0.000828 0.0031529 -0.263 0.793 crops and edge density of croplands at 2500 m scale Interaction between adjacent pasture crops 0.000273 0.0030977 0.088 0.930 and edge density of croplands at 2500 m scale Interaction between adjacent woody 0.010546 0.0026663 3.955 7.64e-05 vegetation crops and edge density of croplands at 2500 m scale

Appendix 1.2 Information on arthropod abundance recorded in the survey for chapter 2. The abundance of most dominated arthropods at different positions within the 24 brassica vegetable sites is presented here. Arthropod taxa Abundance Abundance at the field margins adjacent to land-uses at the field Pastures Woody Brassica Non-brassica crops centre vegetation crops Plutella xylostella 107 52 17 72 37 Pieris rapae 5 1 3 4 0 Aleyrodes proletella 47 13 16 10 25 Brevicoryne brassicae 33 18 16 15 7 Myzus persicae 57 12 59 53 31 Lady beetles 14 8 83 11 3

Appendix 1.3 Information on arthropod abundance recorded in the survey for Chapter 3 a total of 5967 arthropods were recorded from 12 brassica fields. Individual arthropod recorded more than 50 in number is presented here. Arthropods Recorded number Pests Diamond back moth 1243 Cabbage white butterfly 73 Whitefly 861 Jassid 88 Leaf miner 59 Beneficials Lady beetle 422 Predatory spider 381 Wasp 132 Brown lace wing 131 Carabid beetle 67

Appendix 1.4 Statistical models with summary of the final model included in Chapter 3. Figures Models AIC Fig. 4a M1 Total pests ~ Crop composition at 250 m + Crop composition at 500 m + Crop composition at 1000 m + Crop 921.2 composition at 2500 m + Crop composition at 5000 m + (1|Field ID) M2 Total pests ~ Woodland composition at 250 m + Woodland composition at 500 m + Woodland composition at 1000 914.7 m + Woodland composition at 2500 m + Woodland composition at 5000 m + (1|Field ID) M3 Total pests ~ Pasture composition at 250 m + Pasture composition at 500 m + Pasture composition at 1000 m + 910.1 Pasture composition at 2500 m + Pasture composition at 5000 + (1|Field ID) M4 Total pests~ Water composition at 250 m + Water composition at 500 m + Water composition at 1000 m + Water 913.0 composition at 2500 m + Water composition at 5000 m + (1|Field ID) M5 Total pests ~ Crop composition at 5000 m + Woodland composition at 5000 m + Pasture composition at 5000 898.9 m + Water composition at 5000 m + (1|Field ID) Fig. 4b M6 Diamondback moth ~ Crop composition at 250 m + Crop composition at 500 m + Crop composition at 1000 m + 712.6 Crop composition at 2500 m + Crop composition at 5000 m + (1|Field ID) M7 Diamondback moth ~ Woodland composition at 250 m + Woodland composition at 500 m + Woodland composition 717.5 at 1000 m + Woodland composition at 2500 m + Woodland composition at 5000 m + (1|Field ID) M8 Diamondback moth ~ Pasture composition at 250 m + Pasture composition at 500 m + Pasture composition at 1000 702.6 m + Pasture composition at 2500 m + Pasture composition at 5000 + (1|Field ID) M9 Diamondback moth ~ Water composition at 250 m + Water composition at 500 m + Water composition at 1000 m + 708.1 Water composition at 2500 m + Water composition at 5000 m + (1|Field ID) M10 Diamondback moth ~ Crop composition at 5000 m + Woodland composition at 5000 m + Pasture composition 694.5 at 5000 m + Water composition at 5000 m + (1|Field ID) Fig. 5a M11 Total natural enemies ~ Crop composition at 250 m + Crop composition at 500 m + Crop composition at 1000 m + 803.4 Crop composition at 2500 m + Crop composition at 5000 m + (1|Field ID)

M12 Total natural enemies ~ Woodland composition at 250 m + Woodland composition at 500 m + Woodland 799.3 composition at 1000 m + Woodland composition at 2500 m + Woodland composition at 5000 m + (1|Field ID) M13 Total natural enemies ~ Pasture composition at 250 m + Pasture composition at 500 m + Pasture composition at 802.9 1000 m + Pasture composition at 2500 m + Pasture composition at 5000 + (1|Field ID) M14 Total natural enemies ~ Water composition at 250 m + Water composition at 500 m + Water composition at 1000 m 801.1 + Water composition at 2500 m + Water composition at 5000 m + (1|Field ID) M15 Total natural enemies ~ Crop composition at 1000 m + Woodland composition at 1000 m + Pasture composition at 800.4 1000 m + Water composition at 1000 m + (1|Field ID) Fig. 5b M16 Total natural enemies ~ Crop edge density at 250 m +Crop edge density at 500 m + Crop edge density at 1000 m + 807.0 Crop edge density at 2500 m + Crop edge density at 5000 m + (1|Field ID) M17 Total natural enemies ~ Woodland edge density at 250 m + Woodland edge density at 500 m + Woodland edge 789.3 density at 1000 m + Woodland edge density at 2500 m + Woodland edge density at 5000 m+ (1|Field ID) M18 Total natural enemies ~ Pasture edge density at 250 m +Pasture edge density at 500 m + Pasture edge density at 801.4 1000 m + Pasture edge density at 2500 m + Pasture edge density at 5000 m + (1|Field ID) M19 Total natural enemies ~ Water edge density at 250 m + Water edge density at 500 m + Water edge density at 1000 m 801.1 + Water edge density at 2500 m + Water edge density at 5000 m+ (1|Field ID) M20 Total natural enemies ~ Crop edge density at 5000 m + Woodland edge density at 5000 m + Pasture edge density at 798.2 5000 m + Water edge density at 5000 m + (1|Field_ID) Fig. 6a M21 Total natural enemies ~ Crop composition at 250 m *Treatments+ Crop composition at 500 m *Treatments+ 768.2 Crop composition at 1000 m *Treatments+ Crop composition at 2500 m *Treatments+ Crop composition at 5000 m *Treatments+ (1|Field ID) M22 Total natural enemies ~ Woodland composition at 250 m *Treatments+ Woodland composition at 500 m 787.1 *Treatments+ Woodland composition at 1000 m *Treatments+ Woodland composition at 2500 m *Treatments+ Woodland composition at 5000 m *Treatments+ (1|Field ID)

M23 Total natural enemies ~ Pasture composition at 250 m *Treatments+ Pasture composition at 500 m *Treatments+ 786.1 Pasture composition at 1000 m *Treatments+ Pasture composition at 2500 m *Treatments+ Pasture composition at 5000 m *Treatments+ (1|Field ID) M24 Total natural enemies ~ Water composition at 250 m *Treatments+ Water composition at 500 m *Treatments+ 785.8 Water composition at 1000 m *Treatments+ Water composition at 2500 m *Treatments+ Water composition at 5000 m *Treatments+ (1|Field ID) M25 Total natural enemies ~ Crop composition at 1000 m 0*Treatments+ Woodland composition at 1000 m 779.7 *Treatments+ Pasture composition at 1000 m *Treatments+ Water composition at 1000 m *Treatments +(1|Field ID) Fig. 7 M26 Total natural enemies ~ Crop edge density at 250 m *Treatments +Crop edge density at 500 m *Treatments + 763.4 Crop edge density at 1000 m*Treatments + Crop edge density at 2500 m *Treatments + Crop edge density at 5000 m*Treatments + (1|Field ID) M27 Total natural enemies ~ Woodland edge density at 250 m *Treatments + Woodland edge density at 500 m 787.3 *Treatments + Woodland edge density at 1000 m *Treatments + Woodland edge density at 2500 m*Treatments + Woodland edge density at 5000 m *Treatments + (1|Field ID) M28 Total natural enemies ~ Pasture edge density at 250 m*Treatments +Pasture edge density at 500 m*Treatments + 790.6 Pasture edge density at 1000 m *Treatments + Pasture edge density at 2500 m*Treatments + Pasture edge density at 5000 m *Treatments + (1|Field ID) M29 Total natural enemies ~ Water edge density at 250 m *Treatments + Water edge density at 500 m*Treatments + 785.8 Water edge density at 1000 m*Treatments + Water edge density at 2500 m *Treatments + Water edge density at 5000 m*Treatments + (1|Field ID) M30 Total natural enemies ~ Crop edge density at 2500 m *Treatments + Woodland edge density at 2500 m*Treatments 786.7 *Treatments+ Pasture edge density at 2500 m*Treatments + Water edge density at 2500 m *Treatments + (1|Field ID) Fig. 8 M31 Total natural enemies ~ Cost of landscape area at 1000 m for cost-ratio r2 *Treatments + (1|Field ID) 782.9

Appendix 1.5 Summary of ANOVA of final models included in Chapter 3. Models Analysis of Deviance Table (Type III Wald chi-square tests) M3 Chisq Df Pr(>Chisq) Pasture composition at 250 m 0.0921 1 0.76154 Pasture composition at 500 m 0.0301 1 0.862324 Pasture composition at 1000 m 4.0634 1 0.051704 Pasture composition at 2500 m 6.0440 1 0.014153 Pasture composition at 5000 m 6.6605 1 0.004708

M5 Crop composition at 5000 m 4.0622 1 0.04385 Woodland composition at 5000 m 6.3427 1 0.00608 Pasture composition at 5000 m 7.5923 1 0.00452 Water composition at 5000 m 0.8084 1 0.36861

M8 Pasture composition at 250 m 0.354 1 0.55187 Pasture composition at 500 m 1.4031 1 0.236202 Pasture composition at 1000 m 6.6747 1 0.009779 Pasture composition at 2500 m 10.9135 1 0.000955 Pasture composition at 5000 m 13.027 1 0.000307

M10 Crop composition at 5000 m 0.3725 1 0.541622 Woodland composition at 5000 m 6.8051 1 0.00909 87

Pasture composition at 5000 m 10.8911 1 0.000628 Water composition at 5000 m 0.6832 1 0.408474

M12 Woodland composition at 250 m 0.0126 1 0.9105 Woodland composition at 500 m 0.7661 1 0.3814 Woodland composition at 1000 m 5.1384 1 0.0232 Woodland composition at 2500 m 1.4455 1 0.2293 Woodland composition at 5000 m 0.0127 1 0.9102

M17 Woodland edge density at 250 m 2.9299 1 0.08695 Woodland edge density at 500 m 1.4492 1 0.22865 Woodland edge density at 1000 m 0.0297 1 0.86314 Woodland edge density at 2500 m 0.0111 1 0.9162 Woodland edge density at 5000 m 11.1984 1 0.00073

Crop composition at 250 m 1.8067 1 0.1789 M21 Crop composition at 500 m 0.7137 1 0.39822 Crop composition at 1000 m 0.0454 1 0.83014 Crop composition at 2500 m 0.0248 1 0.87483 Crop composition at 5000 m 3.1872 1 0.12874 Crop composition at 250 m *Treatments 1.1594 1 0.28159 Crop composition at 500 m*Treatments 1.5831 1 0.20832 Crop composition at 1000 m*Treatments 9.0483 1 0.00183 88

Crop composition at 2500 m*Treatments 1.2765 1 0.25854 Crop composition at 5000 m*Treatments 2.2844 1 0.13068

M26 Crop edge density at 250 m 0.0203 1 0.8867 Crop edge density at 500 m 0.2725 1 0.6017 Crop edge density at 1000 m 0.0105 1 0.9184 Crop edge density at 2500 m 0.5924 1 0.4411 Crop edge density at 5000 m 1.1991 1 0.2735 Crop edge density at 250 m*Treatments 0.2932 1 0.5882 Crop edge density at 500 m*Treatments 0.3281 1 0.5668 Crop edge density at 1000 m*Treatments 0.0749 1 0.7843 Crop edge density at 2500 m*Treatments 8.1515 1 0.0043 Crop edge density at 5000 m*Treatments 0.3228 1 0.5699