Effect of flower identity and diversity on reducing aphid populations via natural enemy communities

Abstract Floral plantings are often used in agriculture to attract pollinator communities, but they also play an important role in recruiting and establishing natural communities for natural pest control. Inconsistent effects of floral plantings for pest control may be a result of an absence of mechanistic insights and a reliance on the idea that simply increasing flower diversity will benefit these services. A more tailored set of flower species may be needed to benefit the natural enemies through provision of nectar and alternative prey. We used an outside pot experiment to investigate the effect of three flower plants (Fagopyrum esculentum, Vicia faba, and Trifolium pratense) on reducing aphid pests on four different plant cultivars of barley (Hordeum vulgare), over two years. We grew the four cultivars of barley alone, next to a single flower or next to a mixture of flowers, and observed aphid and natural enemy colonization across the growing season. Aphid population sizes were reduced on all barley cultivars grown next to a flower with stronger pest suppression when all flowers were present. Each flower species recruited a different community of non‐barley aphids that, in turn, varied in their ability to establish the natural enemy populations and subsequently the ability to reduce barley aphid populations. Overall, increased pest suppression in the mixed treatments was a result of numerous weaker interactions between different flower, aphid, and natural enemy species, rather than a few dominant interactions. Natural enemy communities could be enhanced by incorporating flower species that vary in their ability to attract and host alternative prey (i.e., non‐pest aphids) as well as suitable nectar provisioning. We can use our knowledge of ecological interactions to tailor floral plantings to increase the effectiveness of pest control services.


| INTRODUC TI ON
Within the framework of Integrated Pest Management (IPM), understanding the ecology of the system is key to identifying the most suitable methods of pest control (Thomas, 1999). With estimated global yield losses of up to 40% from pests and pathogens (Savary et al., 2019), climate change-driven pest range expansion (Bebber et al., 2013), evolution of pesticide resistance (Whalon et al., 2008), and recent bans on many environmentally damaging pesticides (Bakker et al., 2020), alternative solutions to pest control are | 18435 ZYTYNSKA eT Al. needed. One commonly studied method is biological control by natural enemies, yet the effectiveness of this varies across systems and landscapes (Cohen & Crowder, 2017). In many agricultural systems, agri-environment schemes (AES) involving flower strips, banker plants, and intercropping are used to increase key ecosystem services provided by biodiversity, including pest regulation and pollination (Albrecht et al., 2020;Bommarco et al., 2013;Lichtenberg et al., 2017). Floral plantings are often used to attract pollinator communities, but they also play an important role in recruiting and maintaining natural enemy communities that consume insect pests (Haenke et al., 2009), under the framework of conservation biological control (Heimpel & Mills, 2017). Many natural enemies directly benefit from plant nectar, as the adults feed on nectar while the larvae feed on insects; provision of a nectar source can increase a parasitoid wasps' life span by up to 14.7-fold and increase their host-searching time from 3 days to 2 weeks (Russell, 2015). However, inconsistent outcomes of using floral plantings have hindered more widespread use (Albrecht et al., 2020;Hatt et al., 2020;Lowe et al., 2021).
In the absence of detailed mechanistic insights, it has been suggested that increasing diversity of plant species in flower strips, either species richness per se, or flower trait/functional diversity, can help to increase biocontrol (Balzan et al., 2014;Gurr et al., 2003). The argument is that increasing plant species richness is associated with increasing insect diversity including pest species natural enemies (Ebeling et al., 2018;Scherber et al., 2010) and that this increasing natural enemy diversity is associated with more effective biocontrol (Cardinale et al., 2003). In addition, a diversity of natural enemies will avoid selection for resistant pest populations; similar to resistance to pesticides, insects can also evolve resistance to specialized natural enemies either themselves or through interactions with microbial symbionts (McLean & Parker, 2020;Zytynska & Meyer, 2019). As a consequence, current flower mixtures for floral plantings often use a selection of local and native flower species that hope to provide various resources for increasing pollinators and natural enemy communities (Hatt et al., 2020). However, simply increasing flower species richness or plant functional diversity in fields does not necessarily translate into increased pest control (Albrecht et al., 2020;Hatt et al., 2017). Natural enemies are recruited and established on the noncrop flower plants (by nectar or alternative prey) but must spill over into the crop to consume the pest insects (Blitzer et al., 2012;Morandin & Kremen, 2013). If the increased plant diversity traps the natural enemies due to overabundant resources, the noncrop plants will compete with the crops for these services resulting in reduced crop pest control (Kremen et al., 2019). Similarly, nectar provision is important, but while nectar from many flowers enhances natural enemy host searching behavior, others have minimal effect (Bianchi & Wäckers, 2008;Russell, 2015). Thus, a more tailored combination of plants is required to recruit effective natural enemy communities (Tschumi et al., 2016).
Including a more mechanistic understanding of the ecology of pest and natural enemy species, such as their host ranges, may improve conservation biological control. Ideally, the noncrop plant would not host the pest insect (else it acts as a source or reservoir for new infestations) but rather host alternative prey insects that share a common natural enemy with the crop pest. Similar considerations can be taken for intercropping strategies. For example, the aphid Aphis fabae is a pest of beans and sugar beet yet does not feed on cereal crops, while cereal aphids (e.g., Sitobion avenae) do not feed on the beans or sugar beet (Blackman & Eastop, 2000).
Intercropping beans (with their nitrogen-fixing rhizobia) and cereals not only benefits nutrient cycling (Ofori & Stern, 1987) but also disease and pest reduction (Hansen et al., 2008;Zhang et al., 2019). This is because the host range of the natural enemies encompasses both groups of aphids.
Here, we investigated how species richness and the identity of Vicia faba that have been used in flower strips, as banker plants, and have been previously used as cover or intercropping species with cereal crops. Further, both Fagopyrum and Vicia produce harvestable yields as an additional source of income for farmers (Yang et al., 2009), while Trifolium, as a nitrogen (N)-fixing legume, benefits crop yield nitrogen fertilization of the soil (Thorsted et al., 2006). Lastly, there was no cross-over of aphid species from these three flower plant species to the barley plants. We hypothesized that: 1. Increased flower richness reduces aphid populations via increased natural enemy diversity (aphid suppression) 2. Flower identity alters the strength of aphid suppression effects due to varying abilities to recruit and establish natural enemy

| Experimental setup
Plants were germinated and grown in standard potting soil (Floragard B Pot Medium-Coarse, pH 5.6, NPK 1-0.6-1.2) with no added fertilizer. In 2017, the plants were grown for 3 weeks inside a plant growth room (18°C 16:8 L:D), while in 2018, the plants were grown outside under a rain cover and mesh to avoid insect/rodent damage before the start of the experiment. The plants were placed outside on the 16th May 2017 and 29th May 2018 (delayed due to bad weather) and allowed to grow until harvest 60 days later (mid-end of July). The plants were exposed to all weather conditions, and all insects could colonize. The plants were watered when needed (daily during warmer summer days) or excess water removed from pot saucers after rain. As the plants grew larger, sticks were used to support the Fagopyrum plants but no other plants needed the extra support.

| Data collection
In 2017, data were collected twice weekly, whereas in 2018, this was done once weekly. The weather for the experiment was similar for the two years, but with higher initial temperatures in 2018 due to a de- seed yield was collected. For the insect community, we identified all aphids to species and counted abundance for each plant. Natural enemies observed on the different plants were identified at least to family but to genus where possible. Ladybirds were split into ladybird larvae and ladybird adults (but eventually combined for analyses). Parasitoid wasps were identified through the mummies that are formed when the wasp develops within the aphid host; the aphid mummy color indicates the genus of parasitoid wasp. A set of aphids and parasitoids were collected to confirm identification (either by morphology or molecular methods using universal COI primers following Gossner et al. (2016)).

| Data analysis
To compare the 2017 (16 observation days) and 2018 (8 observation days) data sets for aphid numbers, we used peak aphid population size (i.e., the maximum number of aphids counted during one observation day). Aphid peak population is an informative variable since timing of arrival and growth rates can differ among species, due to preceding weather and the ability of natural enemies to control outbreaks (among other factors). Here, it allowed us to compare the data across the years, species, and treatments. We also analyzed the data across the season using repeated measures, but it did not provide further information than considering cumulative factors and thus is not presented.
All data were analyzed in R 3.6.3 (R Core Team, 2020) in R Studio 1.2.5033 (RStudio Team, 2020). We used generalized linear models with quasi-Poisson error on count data (aphid number, natural enemy abundance, tiller number, flower number) and linear models with normal error distributions for continuous data (natural enemy diversity, barley biomass, barley relative growth rate, seed mass) to measure the effects of year, barley cultivar, and flower treatment (flower richness as well as all treatments that include flower identity). All models contained the experimental blocking factors indicating the replicate (Block) and distance from the meadow border (row).
Further, covariates were used where necessary, predominantly barley biomass to account for plant size, and total number of unwinged aphids to account for aphid resource.
We also used Structural Equation Modelling (SEM) in the R package "piecewise" (Lefcheck, 2016), building two models for each year using linear mixed effect models with block and row as ran-

| Aphid and natural enemy species
In 2017, we counted 1694 winged aphids and 25,057 unwinged aphids over 16 days of data collection (twice per week). Almost half of the unwinged aphids counted (12,307)

| Plant growth and seed mass
Seed mass was increased for barley plants next to any flower compared to no flower (Table 1; Figure 4); however, flower identity did not alter seed mass (F 2,149 = 0.61, p = .922; Figure 4) or other traits.
Similarly, barley plants next to flowers (presence/absence) experienced a higher relative growth rate, but a decrease in final biomass, and no change in the number of tillers (stems) ( Table 1). All plant traits varied across the years also dependent on barley cultivar (two-way interactions; Table 1), while for plant relative growth rate (height), this interaction also included flower treatment (Table 1). These effects were not driven by the increased natural enemy diversity on mixtures, abundance of aphids, or natural enemy groups (all p > .05).

| Community assembly and aphid suppression
There was a time-lag from initial aphid colonization until the natural enemy community was established (~10 days) and natural enemy numbers remained highest on the mixed flower treatments for the duration of the experiment in both years (Figure 2g In contrast, there was some potential for facilitation observed between Praon sp. and Aphelinus sp. parasitoids on barley and Vicia plants (Figure 5c).

| DISCUSS ION
Our results clearly show that providing natural enemies with a se- That diversity begets diversity is not a new finding (Palmer & Maurer, 1997;Scherber et al., 2010;Snyder & Tylianakis, 2012), but our work shows that plant identity is likely an important factor for effective biocontrol in crop systems. While we hypothesized that the flowers themselves would be important for general natural enemy recruitment, as they offer a nectar resource for many of the adult parasitoid/predators, it was the variety of nonbarley aphids on these plants that was key for aphid suppression.
Our modeling approach showed that the different plants were to perform its own function by establishing a different community through occupying different feeding, spatial or temporal niches (Snyder, 2019;Snyder & Tylianakis, 2012). However, simply increasing plant species or functional diversity may not result in the promotion of these beneficial interactions (Hatt et al., 2017).
The aim of floral plantings is to establish natural enemy populations that then spill over to the crop plants (Blitzer et al., 2012;Morandin & Kremen, 2013). This must occur for the duration of the cropping season otherwise natural enemies may experience competition for prey resources that leads them to move away from the area (Snyder & Tylianakis, 2012). Alternatively, if the floral planting hosts too much prey resource, the natural enemies will never spill over into the crop (Kremen et al., 2019). In some circumstances, increasing the prey resource not only stops the natural enemy spill over but can also build trophic complexity in the system. This can result in reduced pest suppression due to Distance to flower strip is a commonly significant variable for pest control effectiveness (Albrecht et al., 2020) and habitat preferences may explain why natural enemies stay closer to the edges rather than move further into the field. The distances used in the current experiment are obviously much shorter than those in the field and therefore further work is needed to optimize natural enemy movement into the crop. This also shows the potential for using these plants as an intercrop rather than as flower strips on the border of fields, allowing for more in-field benefits. Identifying what is needed by the various natural enemies in an agroecosystem can help to identify those flower species that can provide these requirements, for example, by using simulations based on empirical data (Bianchi & Wäckers, 2008).
Under integrated pest management schemes, the ecology of the system drives the decision making for pest control strategies.
By knowing the common pest insect, noncrop (resource) insects, and natural enemies in a given area, we can begin to design effective flower mixtures to enhance natural pest control. For example, Gontijo et al. (2018) found that nocturnal biocontrol of aphids by predators was hampered by intercropping with a plant that benefitted overall abundance of natural enemies. Thus, flowers must provide the resource when needed but not hinder pest control efforts by other species in the system. Choosing plants with variable growth rates, flowering times, and growth habits can promote the establishment of a diverse natural enemy community.
In conclusion, we identified many weak interactions that together led to stronger suppression of aphids on potted barley plants that were grown next to a flower with even stronger aphid suppression when all flowers were present. The flowers were chosen for their previous use in flower strips, as banker plants and potential for intercropping. In this system, Fagopyrum grew fast, flowered before the others and established an early nonbarley aphid population that recruited initial populations of natural enemies, followed by Vicia, which flowered soon after. Lastly, Trifolium plants flowered late but rather than flower resource, they provided additional alternative prey resources and shelter for the natural enemies in the latter part of the experiment. By understanding the colonization of various crop plant and flower plants by aphid species and the shared natural enemies, we can begin to tailor floral plantings to enhance biological control effectiveness in field systems.

ACK N OWLED G M ENTS
We thank the project students involved in data collection and daily plant watering. SEZ was supported by a British Ecological Society

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available at https://doi.org/10.5061/dryad.pk0p2 ngps .