Prospects for habitat management to suppress vegetable pests in Australia

Habitat management is an ecologically based approach to suppress pest densities, utilising properties of non‐crop vegetation to improve the impact of natural enemies or to directly affect pest behaviour. Research in this approach has escalated dramatically this century, extending to uptake in some crops, but adoption in Australia has been lower than overseas. Here, we address the need of the Australian vegetable sector to reduce reliance on insecticides by assessing the scope for habitat management in brassica (Brassicaceae), lettuce (Lactuca sativa) (Asteraceae), capsicum (Capsicum annuum) (Solanaceae), carrot (Daucus carota) (Apiaceae), French bean (Phaseolus vulgaris) (Fabaceae) and sweetcorn (Zea mays) (Poaceae) crops. Each crop is of major economic importance, and together, they represent contrasting botanical families and production systems that are associated with different arthropod complexes. We review studies of habitat management that are based on provision of shelter, nectar, alternative prey or pollen for natural enemies (top‐down effects) or changing pest behaviour (bottom‐up effects) through intercropping or trap crops. The likely utility of these approaches under Australian conditions is assessed, and recommendations are made to promote adoption and for adaptive research. Nectar‐ and pollen‐providing plants, such as alyssum (Lobularia maritima) (Brassicaceae), offer strong potential to promote natural enemies in multiple crops whilst trap crops, especially yellow rocket (Barbarea vulgaris) (Brassicaceae), have more targeted utility against diamondback moth, Plutella xylostella (Lepidoptera: Plutellidae), the most serious pest of brassicas. Opportunities for intercrops and banker plant species are also identified. Our recommendations serve as a platform for researchers and for farmer‐led studies to help realise the full potential of habitat management approaches in Australian vegetable production systems.


INTRODUCTION
Plant protection in agriculture in the 21st century is confronting immense challenges due to escalating resistance to available insecticidal compounds, restricted access to 'new' chemistries and health and environmental contamination concerns associated with older compounds (Zhang et al. 2011). More widely, agricultural intensification and extensive use of pesticides has led to biodiversity loss, disruption of ecosystem function, and compromised the delivery of ecosystem services including biological control of arthropod pests (Gagic et al. 2018;Sands 2018). Complex landscapes with natural or semi-natural habitats can promote beneficial arthropods (Landis et al. 2000;Gardiner et al. 2010), as they provide nectar resources, oviposition sites, alternative prey and hosts, physical shelter from crop disturbances and overwintering refuges (Bianchi et al. 2006). However, pest suppression is not an axiomatic outcome of increased landscape complexity (Karp et al. 2018), so local habitat management strategies are crucial.
Habitat management, a form of conservation biological control, has roots in traditional farming practices such as companion planting but, as a rigorous branch of pest management science, is a relatively new approach compared with classical or inundative biological control (Fiedler et al. 2008;Lu et al. 2015). Common habitat management tactics include provision of non-crop vegetation, such as field borders of flowering plants to provide nectar to parasitoids and intercropping with a secondary crop. Secondary crops can serve as donor habitats in which natural enemies feed, multiply and then move to parasitise or predate the target pests (Gurr et al. 2017). Habitat management includes the use of 'banker plants' which provide shelter and alternative prey to sustain natural enemies to provide continuity of pest suppression (Andorno & López 2014). Typically, banker plants host herbivores that are not pests of the crop but serve as an alternative food source for key natural enemies of focal pests (Huang et al. 2011). A less commonly used habitat management strategy is the use of trap crops to divert pests from high-value crops (Hokkanen 1991). A trap crop, which is naturally more attractive to a specific pest than the main crop as either a food source or an oviposition site, is planted next to the main crop to reduce pest pressure (Shelton & Badenes-Perez 2006). An ideal trap crop is a dead-end crop for the targeted pest; the adults are attracted to oviposit, but their offspring cannot survive (Idris & Grafius 1996). Trap crops function as a sink for pests, restricting the movement of pests to the main crop (Shelton & Badenes-Perez 2006).
There are many studies of habitat management, but most are limited to the principles, tools and tactics (Cuperus et al. 2000), with a minority dealing with their practical implementation (Schellhorn et al. 2009). Moreover, available information on habitat management for vegetable farms is further limited in its breadth and depth, especially in Australia. Most habitat management studies have been conducted in the USA and EU (Veres et al. 2013), but many have investigated cropping systems similar to those found in parts of Australia, so they are likely to provide valuable leads. Accordingly, this review considers opportunities for the development and use of habitat management in vegetable crops in Australia. We focus on crop species of major commercial importance: brassica vegetables (Brassica spp.) (Brassicaceae), lettuce (Lactuca sativa L.) (Asteraceae), capsicum (Capsicum annuum L.) (Solanaceae), carrot (Daucus carota L.) (Apiaceae), French beans (Phaseolus vulgaris L.) (Fabaceae), and sweet corn (Zea mays L.) (Poaceae). For each of these crops, we critically appraise overseas habitat management studies for applicability to Australian horticulture. Recommendations for use in Australia consider the strength of evidence from overseas including whether positive effects on natural enemies were observed to cascade to reduced pest numbers, lessened crop damage or economic benefit. We recognise that growers care most about the last-mentioned aspects of this series of effects, but we also consider the underlying mechanisms responsible for observed effects (such as predation rates, parasitism rates and reductions to pest immigration) for these are important in adapting and optimising interventions. From a practical perspective, we also consider whether the plant species which are used in overseas studies are available and allowable in Australia (due to their weed or exotic status).

VEGETABLE PRODUCTION IN AUSTRALIA: A BRIEF OVERVIEW
In Australia, agriculture has been expanding continuously since European settlement (Zalucki 2015). Australian horticultural production is now valued at over A$11 billion per annum, with $2.3 billion worth of export products each year (Hort Innovation 2018). Australian agricultural systems are often intensive, optimising the productivity of monocultures and rotations of specific crops with crop diversity limited to a few genetically homogeneous species (Zalucki 2015). Vegetable production systems have been criticised for heavy reliance on inputs, soil erosion, structural degradation and contamination, pesticide resistance and loss of biodiversity (Anderies et al. 2006;Sands 2018). Australian farmers, like others around the world, have begun to recognise the benefits of less intensive practices, partly reflecting guidance from governmental and non-governmental organisations (Edwards et al. 2012). Australian farming is increasingly adopting zero-till and stubble-retention tactics to protect soils and the environment, using more selective insecticides, and precision agriculture to rationalise input use (Pratley & Kirkegaard 2019).
Invertebrate pest control in Australia is heavily reliant on synthetic pesticides (Adamson et al. 2014). The widespread use of pesticides has led to reduced efficacy of some chemical compounds as a result of the evolution of pesticide resistance in some pest populations. For example, Plutella xylostella L. (Lepidoptera: Plutellidae) has become resistant to all classes of insecticides used to control this pest (Endersby et al. 2008;Furlong et al. 2013) and is ranked second in the Arthropod Pesticide Resistance Database (APRD) for the number of types of insecticide resistance (Furlong et al. 2013).

VEGETABLE CROPS
The vegetable crops at the focus of this review were selected to represent a range of botanical families as well as being among the most economically important species both in Australia (DAWE 2019) and globally (FAO 2018) (Table 1).

Brassica vegetables
Australian brassica vegetable production is valued at over A $300 million per annum (Horticulture Australia 2014a). Brassica vegetables are grown in a number of regions throughout Australia and, due to diverse climates, the industry supplies brassica vegetables throughout the year (Horticulture Australia 2014a). In Australia, P. xylostella is considered the most serious pest of brassicas and has developed resistance to a wide range of insecticides (Endersby et al. 2008;Rahman et al. 2010;Furlong et al. 2013). Cabbage aphid, Brevicoryne brassicae (L.) and green peach aphid, Myzus persicae (Sulzer) (Hemiptera: Aphididae) are also considered serious pests and can be found in all parts of Australia (Gu 2009

Sweetcorn
Sweetcorn is valued at more than $60 M per annum (Hort Innovation 2018) and its production has been expanding due to increasing domestic consumption, export demand and reduced import replacement (Hort Innovation 2018). Cotton bollworm, Helicoverpa armigera (Hübner) and native budworm, Helicoverpa punctigera (Wallengren) (Lepidoptera: Noctuidae) are considered the most damaging pests of sweetcorn (Table 1), reducing yield and market value by attacking the cobs. Feeding damage reduces the market value and renders the damaged crop vulnerable to secondary pest infestation and pathogen infection (Zalucki et al. 1986). Insecticide dependence for Helicoverpa spp. management has already led to the development of insecticide resistance to many insecticide groups (Murray et al. 2005). Fall armyworm, Spodoptera frugiperda (J. E. Smith) (Lepidoptera: Noctuidae), which is known to cause significant economic loss in sweetcorn in other parts of its now global range, has recently been identified in sweetcorn fields in Queensland, Northern Territory and Western Australia (DPIR & Secondary DPIR 2020; GRDC 2020) ( Table 1).

Lettuce
Lettuce is grown in all states of Australia and valued at over $147 M per annum (Hort Innovation 2018). The currant lettuce aphid, Nasonovia ribisnigri (Mosley) (Hemiptera, Aphididae), and M. persicae are the major pests and can significantly damage the crop by feeding and vectoring viruses, such as the lettuce mosaic virus (Barrière et al. 2014;Horticulture Australia 2014d). Nasonovia ribisnigri tends to be found under the wrapper leaves that protect them from insecticides (Kift et al. 2004), which can lead to crop rejection at market due to the contamination (Table 1).

Carrot
Carrots are mostly produced in Queensland in winter and in Tasmania in summer months and have a value of over $215 M per annum (Hort Innovation 2018). In Australia, willow-carrot aphid, Cavariella aegopodii (Scopoli) (Hemiptera: Aphididae) and western flower thrips, Frankliniella occidentalis (Pergande) (Thysanoptera: Thripidae) are considered the major pests of carrot (Horticulture Australia 2014c) and they inflict yield loss by feeding and by transmitting several plant pathogenic viruses (Table 1).

French bean
In Australia, French bean is produced in all states, with a value of more than $77 M per annum (Hort Innovation 2018). Bean pod borer, Maruca vitrata (Fabricius) (Lepidoptera: Crambidae) is the major pest (Turner 1978) which feeds on all developmental stages of the crop (Sharma 1998). Helicoverpa spp. can also cause significant damage in beans by feeding on leaves, flowers and developing pods (Zalucki et al. 1986) (

INSECTICIDE RESISTANCE: A MAJOR CHALLENGE FOR INSECT PEST MANAGEMENT IN AUSTRALIA
Intensive use of synthetic pesticides often leads to changes in pest complexes, including development of pesticide resistance in pests, secondary pest outbreaks (Gurr et al. 2005) and the disruption of organisms in higher trophic levels, resulting in reduced pest suppression (Stark et al. 2007). For example, M. persicae has become resistant to a wide range of insecticides (Bass et al. 2014), making it one of the most widely and strongly resistant species worldwide (www.pesticideresistance.org). In Australia, as in other parts of the world, broad-spectrum pesticides are part of standard agricultural pest management and are often used prophylactically as 'insurance', rather than as means to control pest and pathogen outbreaks (Lamine et al. 2010). Pesticide use in Australia has risen dramatically in recent decades, increasing from less than 0.4 kg/ha on average to over 1.0 kg/ ha across all croplands between 1990 and 2016 (FAO 2019). Reliance on frequent and indiscriminate use of broad-spectrum pesticides has led to a sometimes ineffective and unsustainable situation resulting from the disruption of natural pest regulation that can be afforded by the functional diversity within arthropod assemblages in crop fields (Furlong et al. 2004;Moonen & Bàrberi 2008). More sustainable pest control strategies are needed to overcome the negative impacts of pesticides on the environment and human health, effects on natural habitat and emerging pesticide resistant pests. Beneficial insects can play an important role in pesticide resistance management as they target prey/hosts irrespective of the pests' degree of resistance or resistance mechanism and thus can help to slow down the resistance selection process (Gurr et al. 2017).

ECOLOGICAL BASIS FOR NATURAL PEST SUPPRESSION
There is strong evidence that the population density of insect herbivores tends to reach higher levels in simple agroecosystems compared with more diverse systems (Root 1973;Horne et al. 2008). In early work, Root (1973) proposed two possible mechanisms for this: (1) 'the resource concentration hypothesis', which considers that herbivorous pests more easily locate and then stay in and reproduce in large patches (monocultures) of their host plants, and (2) 'the enemy's hypothesis', which proposes that the predators and parasitoids of herbivorous pests are more effective at controlling pest populations in more diverse systems. These mechanisms of pest suppression are not mutually exclusive; they can operate simultaneously.

HABITAT MANAGEMENT
Landscape complexity, as conferred by the preservation or reintroduction of non-crop habitat, can have a positive impact on the abundance and diversity of natural enemies in a system (Fiedler et al. 2008;Macfadyen et al. 2015;Parry et al. 2015). However, the effects are inconsistent (Karp et al. 2018) and we remain with a far from complete understanding of how landscape characteristics might be exploited to achieve long-term pest suppression (Tscharntke et al. 2012). This challenge is compounded by the practical considerations that landscape effects can operate at scales of several kilometres and are often promoted by slow-growing woody vegetation (Perović et al. 2010). These spatial and temporal factors make it challenging for the manager of an extensive property and still more difficult if management requires active cooperation by multiple neighbours. Accordingly, there is a great interest in habitat management strategies that farmers can employ at a smaller scale (e.g. farm or field) and use annual plants (Landis et al. 2000;Gurr et al. 2005) to maintain the population of relevant natural enemies. Especially valuable is the identification of plant species which can provide benefits to the beneficial insects in a selective manner, denying benefit to key pest species (Baggen & Gurr 1998;Gurr et al. 1998). Whilst the present review draws from successful overseas studies, a limitation of the available literature is that studies of improved efficacy of natural enemies and reduced pest numbers often do not measure effects on plant yield or, especially, the economics of production (Gurr et al. 2016;Johnson et al. 2020). This deficiency weakens the value proposition to growers (who care most about yield and profit and less about natural enemy densities) and has likely been a factor in the limited levels of uptake despite large numbers of research studies. To date, the most widely adopted forms of habitat management are nectar plant borders to rice fields in Asia (Lu et al. 2015;Gurr et al. 2016) and the 'push-pull' system in East African maize (Khan et al. 2010). Against this background, we examined the global literature, focusing primarily on field studies of habitat management that were conducted in our selected crops, to identify the strategies that offer the best scope for adoption in Australian horticultural vegetable crops.

LOCATION AND FILTERING OF LITERATURE
Web of Science (Institute of Scientific Information) was searched using the following search terms: habitat management, habitat manipulation, pest management, natural enemies, conservation biological control, conservation biocontrol and beneficial insect, with each of these terms linked to the common or scientific name of each of our focal crop species. Results were filtered to retain field studies in which plants were purposefully established to (1) improve availability of shelter, nectar, alternative prey or pollen for natural enemies (top-down effect) or (2) change arthropod pest behaviour (bottom-up effect) including intercropping or trap crops. Laboratory studies and reviews were excluded, unless important in revealing a mechanism for observed field effects.
Pitfall traps and entomological nets were used to assess arthropods' densities. (Alomar et al. 2008) The abundance of predatory Syrphidae increased by 43%.
L. sativa were visually inspected to assess the abundance of N. ribisnigri and Syrphidae.
(  P. vulgaris were sampled and arthropods were counted. Larvae of E. kraemeri were reared in the laboratory to verify the parasitism. (Francis et al. 1976) The abundance of Orius spp. increased up to two-fold.
Yield loss due to thrips damage was reduced up to 60%. P. vulgaris was visually assessed for damage. Arthropods were visually counted. (Nyasani et al. 2012) (Continues)  Aphis gossypii population reduced to 50%.
Sticky traps and polythene sleeve traps baited with pheromone lures were used to determine the population densities of pest and beneficial arthropods. (Aswathanarayanareddy et al. 2006) Pests' infestation reduced by 81%.
Trap crop was sprayed with insecticide which reduced the infestation of Zonosemata electa (Say) (Diptera: Tephritidae) up to 98%.
Z. electa were captured using ammonia-baited traps.
Sentinel C. annuum were used to assess the parasitism of A. gossypii by L. testaceipes. (Rodrigues et al. 2001) Banker plants can effectively sustain populations of natural enemies during overwintering and/or after harvesting of the main crop and can provide greater continuity of pest suppression (Huang et al. 2011). For example, Savoy cabbage (Brassica oleracea var. sabauda L.) (Brassicaceae) can serve as a banker plant when sown 1 month before the principal crop, cauliflower (B. oleracea var. botrytis L.) (Brassicaceae); the parasitoids, Diaeretiella rapae (McIntosh) (Hymenoptera, Aphidiinae) were promoted by early arriving Brevicoryne brassicae and Myzus persicae (Freuler et al. 2003) (Table 2). Banker plant systems have, however, received relatively little attention compared to other habitat management strategies, despite their potential to improve biological control efficacy (Frank 2010).
Trap crops such as Indian mustard (Brassica juncea (L.) Czern.) have been shown to attract gravid P. xylostella, which resulted in higher oviposition on trap crops than on the focal crop (B. oleracea var. alba L.) (Åsman 2002). Trap crops functioned as a sink for P. xylostella larvae, as larval survival was reduced significantly on trap crops. In a laboratory study, P. xylostella laid significantly more eggs on yellow rocket and Barbarea verna (Mill.) Asch. (Brassicaceae) than on cabbage (B. oleracea var. capitata L.) (Badenes-Perez et al. 2014). Kale (B. oleracea var. acephala L.) borders around cabbage fields resulted in significantly less P. xylostella eggs on cabbage which reduced pesticide input by 62% (Mitchell et al. 2000) (Table 2).
If a flowering trap crop shows the same effectiveness as it does when it is not flowering, this would open the possibility to use it to then attract and nourish parasitoids. For instance, a laboratory study showed that flowering and nonflowering yellow rocket were equally attractive to ovipositing P. xylostella (Lu et al. 2004), while, in another study, flowering yellow rocket attracted significantly greater number of Diadegma insulare (Cresson) (Hymenoptera: Ichneumonidae), a parasitoid of P. xylostella (Idris & Grafius 1997). Integrating flowering yellow rocket in cauliflower has resulted in greater numbers of generalist predators and higher rates of P. xylostella parasitism (Badenes-Perez et al. 2017) ( Table 2).

Sweetcorn
Habitat management research in sweetcorn is largely dominated by interventions using flowering strips and trap crops with large potential benefits. Plants such as buckwheat, cowpea (Vigna unguiculata (L.) Walp.) (Fabaceae) and sunn hemp (Crotalaria juncea L.) (Fabaceae) have been extensively used to enhance the efficacy of a wide range of natural enemies (Manandhar & Wright 2015) (Table 2). Flowering strips have been reported to lower pest pressure by providing nectar and pollen to natural enemies but also to enhance pest 'fitness' (Duffield & Steer 2006). Sunflower (Helianthus annuus L.) (Asteraceae) strips around a sweetcorn field have been reported to significantly increased the abundance of natural enemies (Jones & Gillett 2005) but seem ultimately to be unsuitable because they are a preferred oviposition and feeding site for Helicoverpa spp. (Zalucki et al. 1986;Duffield & Steer 2006).
A related approach, 'attract and reward', combines the use of flowering plants with applications of herbivore-induced plant volatiles (HIPVs) to attract natural enemies to the area (Gurr et al. 2017;Furlong et al. 2018), and this is one case where small-scale evaluations have shown promise in sweetcorn in Australia (Simpson et al. 2011a). The combination of habitat manipulation with plant volatiles can be an effective approach because of benign nature of both strategies to beneficial arthropods and the principle that strategies that support ecological functions in multiplerather than singleways are preferable (Gentz et al. 2010). For example, sweetcorn sprayed with synthetic HIPVs and at the same time surrounded by buckwheat strips significantly elevated the abundance of natural enemies which, in turn, reduced the larval population of Helicoverpa spp. up to 74% (Simpson et al. 2011a) (Table 2). Despite this promise, however, the commercial partner involved in that research launched an HIPV-based product (Eco Oil) which has been commercially successful as a stand-alone technology rather than being combined with habitat management. This is likely to reflect the familiarity of growers with spray-on products and the ease of use compared with the inherent complexities of habitat management.
A particular form of trap cropping, the 'push-pull' system deployed in East Africa (Khan et al. 2006), provides control of Lepidoptera pests such as spotted stem borer, Chilo partellus (Swinhoe) (Lepidoptera: Crambidae). The approach makes use of preferred oviposition hosts such as Napier grass (Pennisetum purpureum Schumach) and (Brachiaria cv. Mulato II) (Poaceae) to 'pull' gravid moths from the maize crop. Larvae hatching from eggs laid on these grassy borders do not survive, making the trap crop a dead-end host. This effect is complemented by a 'push' from growing pest-repellent molasses grass (Melinis minutiflora Beauv.) (Poaceae) within the maize field which also attract parasitoids (Khan et al. 2001;Khan et al. 2006;Khan et al. 2010). Other candidate plants including black mustard (Brassica nigra L.) (Brassicaceae) (Rea et al. 2002), guinea grass (Panicum maximum Jacq.) (Poaceae) (Koji et al. 2007) and Desmodium spp. (Fabaceae) (Midega et al. 2018) have been extensively used as trap crop in maize fields against a wide range of pests (Table 2). Whilst the push-pull system has proven successful in East African maize, it has not been adapted for or trialled in sweetcorn (Khan et al. 2010) (Table 2).

Lettuce
The use of flowering strips with lettuce has been explored in several studies. Nectar-rich, non-crop flowering plants in the field margins such as alyssum, cornflower, common vetch (Vicia sativa L.) and lupins (Lupinus hispanicus Boiss. & Reut.) (Fabaceae), and corn daisy (Chrysanthemum segetum (L.) Fourr.) (Asteraceae), increased the abundance of generalist predators (mostly hoverflies, ladybeetles and Orius spp.), promoting the biological control of Nasonovia ribisnigri and Frankliniella occidentalis (Pascual-Villalobos et al. 2006;Alomar et al. 2008), and M. persicae (Chaney 1998). Flower strips maintained the population of Orius spp. during the crop-free period (Alomar et al. 2008) (Table 2). The use of multiple plant species, such as alyssum, cornflower, common vetch, lupins, Indian chrysanthemum (Chrysanthemum indicum L.) (Asteraceae), chamomile (Anthemis arvensis L.) (Asteraceae) and clover species in lettuce, increased abundance of natural enemies three-fold, which, in turn, significantly decreased the abundance of N. ribisnigri (Skirvin et al. 2011). The decrease in the aphid population was greatest close to the flowering strips with the effect rapidly decaying with distance from the strips, such that there was little effect at a distance of 10 m from the flowers (Alomar et al. 2008;Skirvin et al. 2011). Other candidate plants such as coriander (Coriandrum sativum L.) (Apiaceae) and dill have also been used successfully in lettuce to support a wide range of natural enemies (Pascual-Villalobos et al. 2006;Alomar et al. 2008) ( Table 2).
Intercropping carrot with onion significantly reduced the abundance of Cavariella aegopodii and P. rosae by promoting carabids (Carabidae) and rove beetles (Staphylinidae) (Uvah & Coaker 1984). In another study, intercropping carrot with subterranean clover significantly reduced root damage to carrots from P. rosae and Pythium spp. which increased the marketable yield of carrot up to 120% (Theunissen & Schelling 2000).
Trap cropping is also practised in carrot; Cotes et al. (2018) identified two fast-growing carrot cultivars ('Calibra' and 'Bolero'), as trap crops to manage carrot psyllid. Ovipositing females preferred to lay eggs on these more phenologically advanced carrot cultivars (Rygg 1977) and the numbers of eggs laid on the main carrot crop was significantly reduced compared with carrot without a trap crop (Cotes et al. 2018) ( Table 2).

ECOSYSTEM DISSERVICES IN HABITAT MANAGEMENT
Habitat management can have unwanted negative impacts on ecosystem functions, which generate 'ecosystem disservices' as a result of detrimental direct effects on the focal crop through increased competition for water, light and nutrients or by allelopathic effects. For example, when alfalfa (Medicago sativa L.) (Fabaceae) was grown as living mulch in a soybean field, the natural enemy abundance increased by 45% when compared to monocultured soybean, resulting in delayed establishment of Aphis glycines (Matsumura) (Hemiptera: Aphididae), but the alfalfa consumed nutrients, resulting a soybean yield reduction of 26% (Schmidt et al. 2007). An additional risk is the potential for introduced vegetation to become weedy (Gurr et al. 2016). For example, if not managed, common vetch may pose a weed threat (Jursík & Holec 2009;Lockowandt et al. 2019).
Habitat management can also lead to indirect effects on the focal crop by enhancement of species other than the targeted natural enemies, which, in turn, may increase pest pressure (Zhang et al. 2007;Gurr et al. 2017). For instance, an Australian laboratory study found larval development and adult longevity of Epiphyas postvittana (Walker) (Lepidoptera: Tortricidae), a major polyphagous pest, were significantly improved in the presence of flowering starflower (Borago officinalis L.) (Boraginaceae) and buckwheat, which demonstrates the importance of identifying selective flowering plants that can only be exploited by the natural enemies (Begum et al. 2006). Though laboratory studies do not necessarily mean that pests will derive benefit from such plants under field conditions, they can lead to the identification of 'safer bet' plant species that cannot be utilised by pests even under non-choice conditions. This concept is evident in another Australian study, in which coriander, buckwheat and B. officinalis significantly increased the parasitism of potato moth, Phthorimaea operculella (Zeller) (Lepidoptera: Gelechiidae) by Copidosoma koehleri Blanchard (Hymenoptera: Encyrtidae), but coriander and buckwheat were also fed upon by P. operculella. Accordingly, only B. officinalis was identified as a 'selective food plant' (Baggen & Gurr 1998) and favoured for later field testing. In a contrasting system, buckwheat cover crops in vineyards attracted 27 times more beneficial arthropods compared with the vineyard without irrigated buckwheat field, but the density of pests, Erythroneura elegantula Osborn and Erythroneura variabilis Beamer (Hemiptera: Cicadellidae) was elevated up to 240% (Irvin et al. 2016).

RECOMMENDATIONS FOR HABITAT MANAGEMENT IN AUSTRALIAN AGROECOSYSTEMS
Distilling the global literature, our recommendations for habitat management in Australian vegetable crop systems are based on selecting plant species that (1) are present and readily available in Australia, (2) are well adapted to the climate in the relevant Australian vegetable production districts, (3) flower quickly and for long enough to cover the focal crop's vulnerable period to pest attack (Fig. 1) and (4) do not lead to ecosystem disservices. As is clear from the foregoing account of predominately international studies, habitat manipulation normally involves the use of a single tactic (e.g. trap cropping or banker plants) and using a single, intervention species, e.g. alyssum alone rather than seed mixes of multiple flowering plants (though this is not always the case). The plants used in overseas studies represent a tiny fraction of the plant kingdom, and a wider range of species needs to be investigated and guiding principles for their use developed. A recent attempt at resolving this issue has been to assess candidate flowering plants from the perspective of their ecological traits (rather than their taxonomy) (Zhu et al. 2020). Whilst that work generated some generalisable findings, e.g. plants with compound umbel or raceme inflorescences and shallow corollas showed positive influence on parasitoid longevity (Zhu et al. 2020), it was hampered by the lack of available data for many ecological traits that are likely to be important.
Secondary to the foregoing selection criteria, we recommend plant species that have shown effective in overseas work against pest species that are present and of importance in Australia. For example, Napier grass, guinea grass and molasses grass used as trap crops in African maize (Khan et al. 2010) (Table 2) were not considered for use in Australian sweetcorn because they are active against corn stem borer, which is not present in Australia. Finally, we highlighted the strategies that did not employ plant species with weedy potential. For example, Indian mustard used for controlling P. xylostella, or Desmodium intortum for Spodoptera frugiperda have weed status in Australia (Oram et al. 2005).
Alyssum, buckwheat, cornflower and dill significantly reduced densities of insect pests by enhancing the impact of natural enemy communities in brassica and lettuce overseas (Lavandero et al. 2005;Ribeiro & Gontijo 2017). In a laboratory assay, alyssum has been shown to be a potential habitat management candidate as it enhances the performance of Cotesia vestalis (Haliday) (Hymenoptera: Braconidae), but P. xylostella derives no benefit from its flowers . Similarly, buckwheat, cowpea and sunn hemp have potential to manage Helicoverpa spp. (Lepidoptera: Noctuidae) and thrips (Frankliniella occidentalis and Frankliniella williamsi Hood) (Thysanoptera: Thripidae) in sweetcorn (Jones & Gillett 2005;Manandhar & Wright 2015), whilst buckwheat and basil can be recommended to control lepidopteran pests, and sunflower to control Halyomorpha halys in capsicum fields (Skirvin et al. 2007;Bickerton & Hamilton 2012). Though H. halys is not established in Australia, its detection in imported goods is quite common (Horwood et al. 2019), indicating a need to minimise opportunities for it to become established and to have strategies for mitigation should it establish. These flowering plants are inexpensive, well adapted to a wide range of Australian climate, and some make effective cover crops or are economically important as secondary crops (Ngouajio et al. 2003;Björkman & Shail 2013). These plants also flower quickly and for long enough to cover the focal crop's vulnerable period to pest attack (Fig. 1). More generally, the blooming period of habitat management plants can be manipulated for optimal effects by the timing of planting. No ecosystem disservices were reported among studies on our targeted vegetables, but this could reflect the lack of comprehensive evaluation (Schellhorn et al. 2009;Gagic et al. 2018). Therefore, we recommend that additional pilot studies should be conducted by researchers and farmers before these plant species are widely promoted. An additional caveat is that economic factors have been little investigated in habitat management research (Shields et al. 2019;Johnson et al. 2020) so the benefit : cost ratios of each technique need evaluation. Practical considerations, such as capacity to accommodate flower strips in irrigation rows rather than occupying productive, crop growing space, will influence the ratio of benefit to cost as well as general ease with which habitat management plants can be established and maintained. Intercropping is attractive as a habitat management option because it involves the production of a secondary crop that can be harvested for profit. It has been well explored in brassica with a diverse range of other crops such as onion, black pepper, tomato, barley and yellow clover and can be effective in management of P. xylostella (Bach & Tabashnik 1990;Hooks & Johnson 2002;Bukovinszky et al. 2004). Similarly, intercropping in sweetcorn can employ mung beans and French beans (Nyasani et al. 2012;Tian et al. 2012) and carrot intercropped with onion reduced the abundance of Cavariella aegopodii whilst increasing the population of beneficial carabids and rove beetles (Uvah & Coaker 1984). When capsicum was intercropped with onion or garlic (Aswathanarayanareddy et al. 2006), the pests' population was significantly reduced. These intercrop species are well established in many Australian vegetable production districts (Fig. 1) and can be adopted more readily than in broadacre crops because vegetable production often involves relatively small areas, often as sequentially planted strips with high edge to area ratios. Most intensive producers, however, focus on optimising the productivity of monocultures and intercropping involves a level of additional labour requirement and complexity of management and marketing. Further, some intercropping systems such as onion-brassica, garlic-brassica or garlic-sweetcorn may require capital investment for new farm machinery for bulb-type crops (Heisswolf 2004).
Trap cropping avoids the aforementioned agronomic complexities because the trap crop is usually closely related to the focal crop and is usually not harvested. Among trap crops used successfully overseas and meriting experimentation in Australia are yellow rocket, Chinese cabbage and collards to manage P. xylostella infestation in brassica vegetables (Badenes-Perez et al. 2017). Yellow rocket is a dead-end host for P. xylostella larvae (Idris & Grafius 1996) and, whilst it has weedy potential (Tahvanainen & Root 1970;MacDonald & Cavers 1991), we recommend evaluation in Australia because it is biennial, offering scope for use as an annual (Fig. 1) to serve as a trap crop in the year of sowing followed by destruction before any risk of setting seed (Badenes-Perez et al. 2004).
Banker plants have received relatively little attention compared to other habitat management strategies. We consider the major opportunity to be the use of sorghum for supressing pests and harbouring natural enemies of capsicum pests, an approach now being trialled by growers in Western Australia (Rizvi's personal observation).

CONCLUSION
There is increasing interest in habitat management among growers, researchers and governmental organisations around the world (Landis 2017). A number of studies illustrating the advantages of increased agricultural complexity and its relationship to ecosystem functions and ecosystem services have chiefly come from Europe, North America, with a few from Asia (Gurr et al. 2016;Gagic et al. 2018). For each of our targeted vegetable crops, we identified successful examples of habitat management techniques such as flowering nectar plant strips and groundcovers, intercropping, trap plants and banker plants. Most commonly reported were positive effects on the diversity and abundance of beneficial insects, but many studies also had cascading benefits on pest densities and crop damage. Whilst ecosystem disservices that lead to negative outcomes for growers are possible outcomes of habitat management studies, such risks can be managed by using the recommendations made in this review as a starting point for pilot studies. Accordingly, the habitat management strategies that appear most promising for use in the targeted vegetable crop systems identified in this review will serve as a platform for future, farmer participatory research and development.