Do non‐native plants contribute to insect declines?

1. With evidence of significant global insect declines mounting, urgent calls to mitigate such declines are also increasing. Efforts to reverse insect declines will only succeed, however, if we correctly identify and address their major causes.


Introduction
Over 30 years ago, E.O. Wilson made general but dire predictions about the ecological consequences of global insect declines that included the loss of flowering plants, terrestrial food web collapse, and its associated loss of animal diversity, as well as the end of rapid nutrient cycling (Wilson, 1987). Although the biomass of some groups may be declining less than others (e.g. moths: Macgregor et al., 2019;aquatic insects: van Klink et al., 2020), accumulating evidence from Germany (Hallmann et al., 2017), England (Conrad et al., 2006), Costa Rica (Janzen & Hallwachs, 2019), the Netherlands (van Strien et al., 2019), North America (Cameron et al., 2011;Forister et al., 2019), and other places, as well as global assessments (Dirzo et al., 2014; been a serious omission in the quest to understand and subsequently mitigate insect declines is the subject of this review. Given that: (a) there are so many species of insect herbivores that rely on particular plant lineages and, in turn, support a vast global richness of insect predators and parasitoids; (b) evolutionarily novel non-native vegetation is predicted to be poor at sustaining large and diverse populations of insect herbivores; and (c) non-natives have so widely diluted coevolved native plant communities in both human-dominated and 'natural' landscapes, we examine the evidence for and against the hypothesis that long term changes in the species composition of plant assemblages have contributed to local and global declines in the abundance and diversity of the insect communities dependent upon those assemblages.

Theoretical considerations
The hypothesis that non-native plants contribute to insect declines depends largely on the premise that host plant specialization in phytophagous insects, which represent the majority of insect diversity (Strong et al., 1984;Mitter et al., 1988) is the rule rather than the exception. Decades of research supports this view: the majority of plant-feeding insects have restricted relationships with plants that allow recognition and use of a just few closely related plant lineages (Ehrlich & Raven, 1964;Forister et al., 2015). When native plants are displaced in the landscape by non-native species, phytophagous insects typically do not recognise the novel host for feeding or oviposition, or may be unable to overcome novel plant defenses (Tallamy, 2004;Wagner & van Driesche, 2010;Bezemer et al., 2014;Litt et al., 2014;van Hengstum et al., 2014). The concurrent loss of native plant hosts and dominance of non-native plants can lead to local extirpation of phytophagous insects and thus to changes in the composition and structure of local food webs (Chew, 1981;Gratton & Denno, 2006;Bezemer et al., 2014;Sunny et al., 2015;Mitchell, 2018).

Host specialization
Insect-plant host relationships have been the focus, if not the impetus, for coevolutionary theory, starting with Ehrlich and Raven's (1964) seminal observations on host plant associations in the order Lepidoptera (butterflies, skippers, and moths). Ehrlich and Raven (1964) postulated that recognition of secondary plant compounds served as a primary driver for the distribution and diversity exhibited by phytophagous insects on their respective plant hosts. Consumption of plant tissues reduces host fitness and drives selection for the development of chemical, physical, or phenological defenses against herbivory, which consequently drives selection for phytophagous insects to develop traits to detect and tolerate plant defenses over time (Dethier, 1954;Fraenkel, 1959;Ehrlich & Raven, 1964;Holloway & Hebert, 1979). Coevolutionary theory assumed that such stepwise, reciprocal adaptations between insects and their hosts led to the specialised relationships we see today.
Although this line of reasoning dominated the literature for years, additional studies suggested that such stepwise, coevolutionary relationships may actually be rare (Janzen, 1980;Jermy, 1984). It was argued that herbivory by insects conferred variable effects on host plant fitness (Jermy, 1984;Crawley, 1989;Mauricio & Rausher, 1997;Fine et al., 2004) and the presence of secondary metabolites in plant tissues may not be a response to herbivore attack (Owen, 1990;Harvey & Purvis, 1991). Furthermore, most plant lineages are hosts for numerous herbivores, which would dilute selection pressures for defense against any specific attacker (Feeny, 1970;Feeny, 1976;Auerbach & Simberloff, 1988;Lau & Strauss, 2005;Agrawal et al., 2006;Pearse & Hipp, 2009). It has also been suggested that host specialization is not a product of coevolution at all; instead, specialised adaptations to particular plants, particularly cryptic coloration, is the result of selection pressures from predation and parasitism in space and time (Bernays, 1988;Bernays & Graham, 1988;Jermy, 1988;Bernays & Cornelius, 1989;Dyer, 1995). Finally, nutritional quality, digestibility, and availability of host plants, not just plant defenses, may play a significant role in the host plant preferences of phytophagous insects (Awmack & Leather, 2002). Nevertheless, coevolutionary theory served as a foundation to demonstrate associations between phytophagous insects and their hosts, with the need for further expansion and empirical evidence on larger scales (Janzen, 1980;Jermy, 1984;Futuyma & Agrawal, 2009).

Host specialization: Scope and scale
Understanding diet breadth in phytophagous insects is paramount in the context of interactions with non-native plants, as an insect's ability to recognise non-native species in the landscape as novel hosts will depend largely on its evolutionary experience. The diet of most insects is constrained to a single plant family in any one habitat or location, with dietary specialization even narrower both in many temperate lineages and hyper-diverse tropical lineages (e.g. Asteraceae, Fabaceae, Forister et al., 2015;Forister & Jenkins, 2017). In fact, diet specialization increases with decreasing latitudes, concurrent with theories of increased plant and animal diversity in the tropics (Hillebrand, 2004;Dyer et al., 2007;Anstett et al., 2016).
For example, in rainforests ecosystems of Papua New Guinea, the majority of phytophagous insects (>90%) specialise at the genus level (Novotny et al., 2002). This result is important as the potential for non-native plants to be recognised as novel hosts by insects may depend on what native confamilials are present in the landscape (Forister et al., 2015). Due to such specificity, the displacement of native plants by non-native species may have profound effects on phytophagous insect populations everywhere.
Of the taxa in global decline (Dirzo et al., 2014;Fox et al., 2014), by far the best studied taxon for diet breadth is Lepidoptera. Caterpillars are generally restricted to host plants at the genus or family level (Ehrlich & Raven, 1964;Futuyma, 1976;Thompson, 1998). Novotny et al. (2004) determined that the average caterpillar species in New Guinea rainforests feeds on three or fewer plant species; over 90% of these caterpillars are concentrated on a single plant host (but see Novotny et al., 2002). Globally, nearly 70% of caterpillar species develop on a single plant family (Forister et al., 2015). Thus, the displacement of native plants with non-native taxa is likely to contribute to declines in Lepidoptera.

Recognition and utilization of novel hosts
Although most phytophagous insects develop host specificity through evolutionary relationships with native plants, there are nonetheless uncommon cases where native insects utilise non-native plants as novel hosts (Bezemer et al., 2014;Litt et al., 2014;van Hengstum et al., 2014;Sunny et al., 2015). Because non-native plants are typically underutilised by native herbivores, it is generally assumed that they are 'released' from top-down pressures imposed by specialist herbivores that otherwise inhibit plant performance (i.e. Enemy release hypothesis, Hierro et al., 2005;Keane & Crawley, 2002). However, the opposite may also be occasionally true; non-native plants may lack defenses that repel native insects and thus become more susceptible to herbivore attack (e.g. agricultural crops, Colautti et al., 2004;Maron & Vilà, 2001). In cases where native phytophagous insects recognise susceptible non-natives, insects may expand their host range (Carroll & Boyd, 1992;Parker & Hay, 2005;Strauss et al., 2006;Carroll & Fox, 2007;Hull-Sanders et al., 2007), and even impede non-native plant performance (Parker et al., 2006;Suwa & Louda, 2012).
Alternatively, some herbivores have benefitted from the introduction of novel hosts. Rattlebox moths (Utetheisa ornatrix) exhibit higher oviposition rates and fitness on non-native congeners compared to their native hosts (Cogni, 2010). Several skippers and butterflies (e.g. common sootywing, Pholisora catullus, large marble, Euchloe ausonides, and anise swallowtail, Papilio zelicaon) benefit from the introduction of novel congeners in California, especially where anthropogenic factors have led to the loss of native host (Graves & Shapiro, 2003). Similarly, the palamedes swallowtail (Papilio palamedes) use less suitable but abundant camphor trees (Cinnamomum camphora) following outbreaks of disease on its preferred host (Chupp & Battaglia, 2014) and the federally protected Manduca blackburnei now uses tree tobacco in Hawaii (Rubinoff & San Jose, 2010). Although the potential exists for phylogenetically related plants to be recognised by phytophagous plants as novel hosts, it is by no means a guarantee (Burghardt et al., 2010); well after a century of colonization, non-native Piper trees were devoid of specialists that should have established from adjacent, native congeners (Novotny et al., 2003).
Given the close evolutionary link between herbivorous insects and their host plants, when nonnative plants dominate ecosystems, either through invasion or cultivation, they inevitably displace the host plants that are necessary to support sustainable insect populations. How well native insects perform on non-native plants is one important variable determining non-native plant impacts on insect declines.

Do non-native plants reduce insect populations?
To posit that non-native plants, either as invasive species, agroforestry crops, or widely planted ornamentals, are contributing to global or local insect declines, there must be clear evidence that insects directly requiring plant resources have lower fitness when using non-native plants, do not recognise them as viable host plants, or avoid them altogether. It is also necessary that reductions in herbivore numbers caused by non-native plants are not mitigated by density-dependent effects such as reduced competition. These questions have been an active area of research as they relate to the impacts of invasive plants and have been reviewed extensively several times in the last decade (Bezemer et al., 2014;Litt et al., 2014;van Hengstum et al., 2014;Yoon & Read, 2016). As predicted by theory (reviewed above), studies have repeatedly shown that phytophagous insect host plant affiliations are constrained by plant evolutionary history (Ødegaard et al., 2005;Weiblen et al., 2006;Janz, 2011) and that when native host plants are reduced within or removed from landscapes, insect herbivore populations on the whole are smaller and less diverse. For example, in one the most rigorous meta-analyses available, Yoon and Read (2016) examined the impact of non-native plants on Lepidoptera larval performance, survival, oviposition preference, abundance, and species richness in 76 studies. With few exceptions, caterpillars had higher survival and were larger when reared on native host plants, and plant communities invaded by non-native species had significantly fewer Lepidopterans and less species richness. Moreover, in 3 of 8 cases examined, non-native plants functioned as ecological traps, inducing female Lepidoptera to lay eggs on non-native plants that did not support successful larval development.
Non-native plants can also indirectly alter the abundance of native insects on native plants via their effects on the quality, abundance, and/or diversity of native plants (Powell et al., 2013), or on the structure of their habitat (Bezemer et al., 2014). Not only to invasive plants replace edible plant biomass with inedible vegetation (a direct impact), they also typically devastate the diversity of native plant communities on a local scale, which is the scale at which ecosystems function (Powell et al., 2013). Although this is an indirect effect, the global scale of plant invasions is so huge (see below) that the impact on insect populations is likely to enormous as well.
Although each of the above reviews concluded that non-native plants negatively impacted insects more often than not, insect responses to non-natives were not uniform, with some studies showing no effect and a few showing positive impacts from non-native plants. For example, in a review of 89 studies, Litt et al. (2014) found that phytophagous arthropod populations decreased in 48% of the studies but increased in 17%. Similarly, Bezemer et al. (2014) reviewed studies that found non-native plants to be toxic to native insects (e.g. Keeler & Chew, 2008;Ding & Blossey, 2009) as well as cases where non-natives were acceptable host plants (e.g. Harvey et al., 2010). Not surprisingly, the equivocal nature of these responses has led to controversy over how much non-native plants actually affect insect populations.
There are two reasons to urge caution when interpreting the reports of host range expansions to non-native plants that are scattered across the literature. First, these reports do not represent a balanced survey of host use across all species. When host range expansions in particular species are detected, it is news-worthy and they are published. In contrast, examples of insects unable to use introduced plants are predicted by theory and therefore not newsworthy, studied, or written up for publication. Who, for example, would study whether monarch butterflies could develop on Crepe myrtle (Lagerstromia indica) and what journal would publish such foregone results? Because of their exceptional nature, there is a danger of overestimating the prevalence of non-native host use and worse, underestimating the negative impacts of non-natives on insect populations.
Host range expansions are exceptions, which means more species cannot use non-natives for growth and reproduction than can. In one area of northern California, 34% of the butterfly species rely on non-native hosts because their closely related native hosts have been extirpated (Graves & Shapiro, 2003). Yet even in this most celebrated example of insects expanding to non-native plants, it is likely that 66% of the butterflies in this region would be expected to suffer population declines where their hosts are declining due to encroachment by non-natives.
An important source of variation in the results of studies examining non-native plant impacts on insects is the context of the host plant associations studied. Insects are associated with plants in a number of contexts: as folivores, wood eaters, detritivores, pollinators, frugivores, and seed-eaters; as herbivores with chewing or sucking mouthparts; and as host plant specialists or generalists. Non-of these contexts are equivalent and thus they cannot be lumped when reporting results. That is, the context in which the impacts of non-native plants on insects are examined is critical to the ramifications of the results. When the context within which studies were performed is considered, the following patterns emerge; 1 Insects with chewing (mandibulate) mouthparts are typically more susceptible to defensive secondary metabolites contained in leaf vacuoles than are insects with sucking (haustelate) mouthparts that tap into poorly defended xylem or phloem fluids (Verhoeven et al., 2009). Thus, sucking insects find novel non-native plants to be acceptable hosts more often than do chewing species. Indeed, failure to recognise that all members of the second trophic level do not interact with plants in the same way has been a source of misinterpretation and controversy in many studies of non-native plants. Not surprisingly, some studies of detritivorous invertebrates (e.g. Sax, 2002) and frugivorous birds (e.g. Gleditsch & Carlo, 2011) found no negative impact from invasive plants, while studies of insect folivores and the insectivores that eat them (e.g. Flanders et al., 2006; , 2010;Narango et al., 2018;Richard et al., 2018) usually find large impacts. Considering that there are more than 4.5 times as many mandibulate insect herbivores as haustellate species (Table 1), there is reason for concern when non-native plants replace native hosts; the largest guild of insect herbivores is also the most vulnerable to non-native plants and the most valuable to insectivores. 2 Although both woody and herbaceous non-natives decrease the overall abundance of insects, the impact of woody species is stronger (Daehler, 2005;van Hengstum et al., 2014). This could be because woody native species, on average, support more species of phytophagous insects (Tallamy & Shropshire, 2009), so their displacement by non-natives will have greater impacts. 3 Because plants in closely related lineages often share defensive chemicals and phenology with native hosts, insects are more likely to accept non-native congeners or con-familial species as novel hosts than non-natives that do not share an evolutionary history with native host plants (Connor et al., 1980;Hill & Kotanen, 2009;Pearse & Hipp, 2009;Burghardt et al., 2010;Lombardero et al., 2012;Burghardt & Tallamy, 2013;Pearse & Altermatt, 2013). Acceptance of related non-natives by insects is far from universal, however. A highly controlled common garden experiment comparing insect use of 18 congeneric pairings found that non-native congeners of native species reduced insect abundance and species richness by 68% (Burghardt et al., 2010;Burghardt & Tallamy, 2013). 4 Insects with a narrow diet breadth (host plant specialists) are less likely to develop on evolutionarily novel non-natives than are insects with broader diet breadth (Bertheau et al., 2010;Pearse, 2011). Because there are far more species of host plant specialist than generalists (Forister et al., 2015), generalists will not prevent serious declines in species richness and abundance when native plants are replaced by non-natives. Even when populations of generalists are compared on native and non-native plants, non-natives cause significant reductions in species richness and abundance (Ballard et al., 2013). In fact, despite evidence of broad host plant use, generalists are often locally specialised on particular plant lineages and thus may function more like specialists than host lists accumulated across their range suggestion (Fox & Morrow, 1981;Tallamy et al., 2010). 5 Any reduction in the abundance and diversity of insect herbivores will cause a concomitant reduction in the insect predators and parasitoids of those herbivores. Although the logic here is irrefutable and has some support (Harvey, 2005;Narango et al., 2018), few studies have attempted to measure natural enemy reductions where invasive plants are common. Predaceous arthropods decreased in only 44% of the studies examined by Litt et al. (2014). These results could reflect the typical prey of spiders, the most abundant arthropod predators in terrestrial habitats. Web-spinning spiders are generalist predators that target flying insects more often produced by detritus and aquatic habitats than living plants. No studies have traced the fate of parasitoid communities linked to phytophagous insects when native plants are replaced by non-natives. Yet the vast majority of parasitoids are highly specialised on particular host lineages (Vinson, 1998;Smith et al., 2006;Forbes et al., 2018). It follows, then, that if a host lineage decreases in abundance due to non-native plants, so will its parasitoid complex. 6 Studies comparing native plants that support very few phytophagous insects to non-native plants are less likely to find differences in phytophagous insect communities than studies comparing non-natives to native plants that host dozens of species. Native plants differ by orders of magnitude in their ability to host phytophagous insects (Tallamy & Shropshire, 2009). In the mid-Atlantic region of North America, for example, oaks (Quercus spp.) host 557 Lepidoptera species, while tulip trees (Lireodendron tulipifera) host only 21 species and Yellowwood (Cladrastus kentuckea) does not serve as host for any Lepidoptera. Thus, a comparison of the impacts of a congeneric non-native such as Norway maple will depend a great deal on the choice of native plants with which it is compared. 7 Insects that feed on well-defended living tissues such as leaves, buds, and seeds are less likely to be able to include non-natives in their diets than are insects that develop on undefended tissues like wood, fruits, and nectar. Although this hypothesis has never been formally tested, the ease with which introduced wood borers like emerald ash borer (Agrilus planipennis), Sirex woodwasp (Sirex noctilio), Asian long-horned beetle (Anoplophora glabripennis), the polyphagous shot hole borer (Euwallacea sp), redbay ambrosia beetle (Xyleborus glabratus), and many other species of introduced bark beetles have included North American trees in their diets, as well as the extraordinary polyphagy of frugivores like the brown marmorated stinkbug (Halyomorpha halys) support this notion (Baranchikov et al., 2008;Fraedrich et al., 2008;Haack et al., 2010;Eskalen et al., 2013;Paap et al., 2018).

How pervasive are non-native plants?
Given the poor performance of most insect herbivores on non-natives, the widespread use and invasive behaviour of many non-native plants increases the potential scale of impact on insect abundance and diversity. Although plants have always distributed themselves around the globe, the increased temporal and spatial mobility of humans has resulted in an extraordinary increase in the rate of plant movements (Vitousek et al., 1997a) and most species' introductions have happened in the last 200 years (van Kleunen et al., 2018). Habitat is rapidly being converted from coevolved native ecosystems into novel assemblages of plants and animals, making the conversion of native plant communities into plant assemblages dominated by non-native species one of the most ubiquitous threats to biodiversity today (Johnson, 2007;Radeloff et al., 2015). The introduction of non-native plants has completely transformed the composition of present-day plant communities in both natural and human-dominated ecosystems around the globe (McKinney, 2004;Dolan et al., 2011) and the magnitude of introductions is staggering. An estimated 13 168 plant species (about 3.9% of global vascular flora) have been introduced and naturalised beyond their native ranges as a result of human activity (van Kleunen et al., 2015) with more than 3300 species of non-native plants established in the continental U.S. alone (Qian & Ricklefs, 2011). Where they are abundant, non-native plants can dominate plant biomass, and also reduce native plant taxonomic, functional and phylogenetic diversity, and heterogeneity, further exacerbating their impact (Walker et al., 2009;Beaury et al., 2019;Padullés Cubino et al., 2019). The global dominance of non-native plants species in natural and human-dominated systems is due to introductions from three primary sources: invasion, agriculture & agroforestry products, and horticultural ornamental plants.

Plant invasion
Invasive plants are regarded as a major threat to biodiversity and ecosystem function (Czech et al., 2000;Pyšek et al., 2020). The dispersal and spread of invasive plants has been driven by global trade networks and colonialism (Chapman et al., 2017;van Kleunen et al., 2018). At least 1/6 of the globe is highly vulnerable to plant invasions which includes areas of biodiversity hotspots (Stohlgren et al., 2003;Early et al., 2016).
In many ecosystems, invasive flora can be substantial components of floral diversity. By extrapolating data from USDA forest service inventory plots, Miller et al. (2008) estimated that 9% of SE U.S. forests are covered by just 33 common invasive plant species. New technology has facilitated surveys over larger spatial areas; for example, using remote sensing, Bradley et al. (2015) found that in a 28 000 km 2 area in northern Nevada, monocultures of non-native cheatgrass (Bromus tectorum) expanded from 14% of the area to 29% in less than 30 years. Even in protected lands, invasive flora is a major threat to ecological function and few areas are completely free of invaders (Pyšek et al., 2020). For example, in some island systems, invasive species can be 50-70% of the species in the ecosystem (Vitousek et al., 1997b). Despite management efforts, evidence points toward invasive plants increasing in abundance, especially in protected areas (Pyšek et al., 2020).
Introductions of non-native plants can introduce other non-native organisms that alter the composition of native plant communities and thus the insects that depend on them. The horticultural and agricultural plant trade has been a leading pathway for invasive pests and pathogens. For example, the commercial sale of Chinese chestnut (Castanea mollissima) introduced chestnut blight (Cryphonectria parasitica), which completely transformed >70 million km 2 of eastern deciduous forest by the loss of the iconic American chestnut (Castanea dentata) and is believed to have caused the extinction of five specialist insect herbivores that fed on Castanea (Wagner & van Driesche, 2010); the blight's effects on non-specialist taxa are unknown. Similarly, Hemlock woolly adelgid (Adelges tsugae) was imported with ornamental Japanese hemlocks and has destroyed most southern populations of eastern hemlock (Tsuga canadensis) populations along with the insects dependent upon them (Havill et al., 2014).

Agriculture and agroforestry
The need for fast-growing colonizing trees in agroforestry and restoration (e.g. Acacia and Albizia) has also increased the use of non-native species, many of which have escaped to dominate nearby native forests (Schroth et al., 2004). At least 25% of the world's planted forests are composed of non-native tree species (Lombardero et al., 2012); for example, one fourth of Portugal's forestland (900 000 hectares) is planted in Eucalyptus (Ames, 2017). At least 118 exotic tree species have naturalised in Puerto Rico and compete with native species in natural stands (Francis & Liogier, 1991); the African rubber tree, Funtumia elastic, a species that is invasive in >30 countries and completely dominates secondary forests of many tropical Pacific islands, is now one of the most common trees in Puerto Rico  as well. Grassland communities are also vulnerable to invasion by non-native plants, particularly from adjacent agricultural land. Aided by purposeful introductions for cattle as well as excess nutrients from agriculture, grasslands around the globe are six times more likely to be dominated by non-native plants than are other ecosystems (Seabloom et al., 2015).

Horticulture
Because of the popularity of non-native plants in landscaping, horticulture is a major source of non-native plants in both cultivated and natural ecosystems (van Kleunen et al., 2015). Non-native plants tend to be aggregated in areas with high human densities (McKinney, 2001), high international trade, and robust horticultural industries, highlighting the importance of these regions for facilitating introductions to wider geographical areas (van Kleunen et al., 2015;Dawson et al., 2017). There are estimates that 50-70% of invasive and naturalised species are a direct result of intentional horticultural introductions (Dehnen-Schmutz & Touza, 2008;Richardson & Rejmánek, 2011). Even though many invasive plant species are regulated in the U.S. and included on 'do-not-plant lists,' these problematic taxa are abundantly sold in horticulture; 61% of non-native plants on the invasive species list are still sold in all lower 48 states (Beaury et al., 2020). Gardens retain ideal conditions and stable sources of non-native species that have the potential to become invasive if climate or environmental conditions shift (Dukes & Mooney, 1999;Smith et al., 2020).
One land use with high dominance of non-native species is in urbanised areas which are predicted to cover 5-20% of earth's habitable land mass by 2030 (Seto et al., 2012). Non-native plants are particularly abundant in urban areas because of the strong preference of non-native plants in horticulture (McKinney, 2004;Dolan et al., 2011;Avolio et al., 2015). Moreover, non-natives frequently escape cultivation into disturbed and fragmented areas which characterise urban areas (Vilà et al., 2007;van Kleunen et al., 2019).
Because of these and other factors, the majority of flora are non-native in most cities (Qian & Ricklefs, 2006;Avolio et al., 2015Avolio et al., , 2018Zeeman et al., 2017). For example, more than 40% of plant species are non-native in the most urbanised areas of New York and New Jersey (Aronson et al., 2014). Although urban areas are often characterised by habitat loss relative to natural areas, if the plant biomass that is retained or replaced is strongly dominated by non-native species, this further amplifies the loss the native habitats within the developed matrix (Niinemets & Peñuelas, 2008;Goddard et al., 2010;Lerman & Warren, 2011;Aronson et al., 2014;Lepczyk et al., 2017;Avolio et al., 2018) as well as further degradation of adjacent natural areas due to invasion (Maskell et al., 2006;Duguay et al., 2007).
In horticulture and ecological circles alike, concern has focused primarily on invasive species with the assumption that if a plant is not invasive, it does not cause ecological problems. Indeed, the majority of ornamental species have not become invasive (Reichard & White, 2001), leading land managers and the public to deem these species acceptable for plantings. There are two problems with this reasoning. First, ornamental plants that are distributed by the millions across landscapes represent the first trophic level wherever they are planted. Even if they never develop invasive behaviour, they have not replaced the ecological functions of the native plants that used to support insect populations. Second, there is no guarantee that an ornamental species that is well-behaved today will not become invasive in the future. Many invasive plants experienced a lag phase during which they were benign or overlooked members of plant communities before being recognised as invasive (Essl et al., 2011). Altered conditions and increased introductions from climate and land use change may further enable species to escape cultivation and become invasive; thus, non-native plant dominance is predicted to increase in magnitude over the next decades (Hellmann et al., 2008;Bradley et al., 2012;Smith et al., 2020).

Designing robust experiments
One cannot accurately determine the role of non-native plants in insect declines if the experiments purportedly measuring how non-native plants impact insect abundance and diversity are not designed properly. To our knowledge there have been no studies explicitly designed to measure impacts of non-native plants on insect populations over ecological time frames and/or across landscape-level spatial scales. At present, we can only extrapolate from the results of existing short-term studies performed at local spatial scales, often with one or more design limitations.
We suggest the following 10 guidelines should be considered when designing or interpreting robust investigations of the impact of non-native plants on insect populations: should be investigated in both natural, managed, and working agricultural landscapes. Managed, non-agricultural landscapes (e.g. urban, timber, etc.), which now occupy nearly half of the conterminous U.S. and enormous parts of Europe, can and must support viable insect populations in the future if we are to curb the loss of insects, so understanding how non-natives impact insects in such landscapes is essential. Similarly, agricultural lands must be made more insect friendly by returning native plant communities wherever possible as nearly have of terrestrial earth is now in some form of agriculture (Kremen & Merenlender, 2018) 8 Because both native and non-native plants vary by orders of magnitude in their ability to support insect herbivores, native vs non-native comparisons are often oversimplified. The identity of native and non-native species chosen for an experiment, as well as the species composition of the matrix in which they reside, may heavily influence the results. The most accurate comparisons will thus employ realistic assemblages of both productive and unproductive native and non-native species in a given system. 9 The spatial arrangement of the experimental plants will also impact results and should be controlled or accounted for in all treatments. A native plant surrounded by non-natives may support a different insect community composition and abundance than a native surrounded by other natives (Clem & Held, 2018). The converse may also be true. Similarly, host plants surrounded or adjacent to different landscape matrices (e.g. urban, forest, agriculture) may support different insect communities due to variation from rates of colonization and dispersal.

Concluding remarks
Understanding insect declines requires a careful review of the evidence that (1) insect populations in many parts of the world are, in fact, declining and (2) such declines are being driven by several anthropogenic causes. Whether insects are declining in undisturbed natural areas is still a matter of debate (e.g. Harris et al., 2019;Janzen & Hallwachs, 2019;Crossley et al., 2020), but there is less controversy about declines caused by light pollution, development, industrial agriculture, and pesticides in human-dominated landscapes (Wagner, 2020). In this review, we have presented evidence that non-native plants, both invasive species and widely used ornamental plants, have disrupted specialised evolutionary relationships between insect herbivores and their native host plants over such large areas that non-natives should also be considered a threat to insect populations. When deciphering the contribution non-native plants have made to global insect declines, the question should not be 'Can we document single species use of non-native plants?' but rather 'How do the majority of species and thus insect populations in aggregate respond to the replacement of native host plants with non-native plants?' The preponderance of evidence suggests that more insect species than not suffer when the abundance and diversity of native host plants are reduced by non-native species, and that mandibulate folivores, the most speciose insect taxa, suffer more than less species-rich insects with sucking mouthparts. The reduction of mandibulate folivores such as caterpillars has outsized negative impacts on animal diversity for two reasons; (1) caterpillars are critical dietary components of most terrestrial birds species and their loss from food webs directly reduces bird fitness (Narango et al., 2017(Narango et al., , 2018, and (2) caterpillars serve as hosts for insect parasitoids, perhaps the most speciose guild of metazoan animals on the planet (Forbes et al., 2018).
Critical gaps in our knowledge remain. For example, we know little about how to extrapolate the results of local studies to non-native plant biomass at larger spatial scales, and to our knowledge, few studies have compared patterns across different ecosystems and biomes using systematic methodology. Similarly, we know little about the impacts of non-native plants on insect populations over long ecological time scales. To address the variance in sensitivity to non-native plants, we need a better understanding of how various insect guilds and functional groups differ in their responses. Finally, our knowledge of the impacts of non-natives on different trophic levels and food webs is scanty at best.
Despite these knowledge gaps, enough evidence already exists to implicate non-native plants as a factor in insect declines, and one that can be mitigated by property owners and managers worldwide relatively easily. Non-native plants are not the ecological equivalents of native plants, yet they have replaced native plant communities as ornamental species, agroforestry products, and invasive species across the globe. Given that more than 160 million hectares of arable land are nonnegotiably dedicated to non-native crop plants in the U.S. alone (USDA NASS, 2019), how we respond to the increasing abundance of non-native plants in areas outside of agriculture may determine how well we can sustain ecologically vibrant insect herbivore populations, as well as the myriad bird and parasitoid species that depend on them in the future. We suggest that curbing the spread and use of non-native plants at local, national, and international scales will be a necessary and effective way to reduce insect declines.