Organic fertilisation enhances generalist predators and suppresses aphid growth in the absence of specialist predators

1. Biological control by natural enemies is a valuable ecosystem service. The preda tor community in a crop field is a combination of predators dwelling in the field and those moving into it from the surrounding landscape. The former is mainly affected by field management, the latter more by the composition of the sur rounding landscape. Yet, separate and combined effects of local and landscape management on pest suppression have seldom been investigated. We within an existing long- term field to the effects of local of organic manure or inorganic mineral fertilisation, and the the of different predator no generalist (wolf and (ladybirds). We by or under different fertilisation treatments, and how the magnitude of pest was affected by predator community


| INTRODUC TI ON
Pest suppression by predatory arthropods, such as ground carabid beetles and spiders, provides a crucial basis for crop protection (Chaplin-Kramer & Kremen, 2012). Semi-natural habitats in the landscape enhance predator communities in crop fields (Hendrickx et al., 2007;Holland et al., 2017), but the variability around the trend is large (Karp et al., 2018). Divergences could arise because of differences in predator's habitat preference and dispersal abilities (Meyer et al., 2019;Schalkwyk et al., 2020). Additionally, there is increasing evidence that soil management in the crop field can determine predator community composition (Eyre et al., 2012;Hanson et al., 2016;Marrec et al., 2015;Palmu et al., 2014). Managing soils to increase soil carbon and biodiversity benefits predator biomass (Riggi et al., 2017;Tamburini et al., 2016), but this needs confirmation for other management practices such as fertilisation. Furthermore, interactive effects of local and landscape factors on predator communities have rarely been investigated (Concepción et al., 2012;Zhao et al., 2014).
The use of fertilisers is currently ubiquitous. Fertiliser type can affect natural pest suppression (Riggi & Bommarco, 2019). Organic fertilisers can increase soil diversity of several taxa (Liu et al., 2016;Lori et al., 2017) including potential pest regulators such as carabids (Birkhofer, Bezemer, et al., 2008;Holland & Luff, 2000;Idinger et al., 1996). By providing resources to the detritus-based soil community, manure fertilisation provides alternative food to generalist predators, increasing their abundance and, thus, enforcing pest suppression (Leroux & Loreau, 2008), but it could also divert predators from eating the crop pest (Jaworski et al., 2013). Inorganic mineral fertilisers, in contrast, provide more directly accessible nutrients to the plant increasing plant nitrogen content, biomass and which can boost the herbivore growth rate (De Bruyn et al., 2002;Herencia et al., 2007;Mäder et al., 2002).
Despite these indications of significant fertiliser effects on natural pest suppression, most research on this topic lacks the connection between above-ground communities and fertilisation management.
Few comprehensive examinations have found that manure fertilisation boosted top-down predator suppression of aphids in barley fields (Birkhofer, Bezemer, et al., 2008;Riggi & Bommarco, 2019) likely caused by manure subsidising the detrital compartment and boosting generalist predators early in the season.
Carabids, spiders and ladybirds are common predators that often immigrate into crop fields from nearby semi-natural habitats.
However, several carabid species also overwinter within crop fields and their populations are known to be affected by crop type and management (Hanson et al., 2016;Holland et al., 2005;Palmu et al., 2014).
Thus, predator communities in crop fields are assembled by locally emerging and seasonally immigrating predators, making it difficult to separate local and landscape effects on pest suppression, and their interaction (Concepción et al., 2012). The presence of more soil prey in organically fertilised fields could attract more predators from the surrounding landscape. However, if the enrichment of the soil fauna drives a switch in the predator diet from pest to soil fauna or increases intra-guild predation, then we might see interactions between organic fertilisation and the composition of the predator community which could reduce pest suppression (Roubinet et al., 2018).
We used 2 × 2 × 2 m mesocosms in an oat field, to examine interactive effects on natural pest suppression between experimentally simulated combinations of predator guilds (generalists and specialists) moving to the crop field from the surrounding landscape, and three fertilisation treatments (no fertilisation, organic manure and inorganic mineral fertilisation). We explored whether fertilisation boosted local predator fauna, and how this impacted pest suppression when combined with a spillover of predators from the surrounding landscape of either specialist (ladybirds) and/or generalist predators (spiders). We aimed to investigate (a) the combined effects of fertilisation and predator immigration on aphid population suppression; (b) whether the strength of top-down and bottom-up suppression of aphids is mainly regulated by the fertilisation type or by the predator immigration; and, finally, (c) whether yield is affected by interactions between fertilisation and the immigrating predator community. We expected manure fertilisation to boost local soil fauna, benefitting local predators and translating into more efficient suppression of aphid densities. We also expected positive interactions between manure fertilisation and the inflow of predators from the surrounding landscape in suppressing aphid populations and in determining yield, by an increased top-down suppression.
in 1996 in the region of Västergötland, Sweden (58°20′38.0″N 13°06′13.5″E). The long-term field experiment (total size 8,200 m 2 ) was divided into four blocks. In each block, nine fertilisation treatments, including a no fertilisation control, had been applied every second year in the same plots, that is, block was the replication (see Appendix S1 in Supporting Information for more information about soil properties). The field has been sown with cereals every year since 1996. We carried out a mesocosm experiment in 2018, in which the entire experiment was sown with spring oats. We selected three out of the nine fertilisation treatments: manure (organic fertiliser), calcium nitrate (inorganic fertiliser) and a no fertilisation treatment (control; Appendix S1 in Supporting Information). In each experimental plot (6 × 18 m) within each block, that is, four replicates per treatment, we built three cages (2 × 2 × 2 m) covered by nylon net (0.6 × 1.6 mm mesh size), giving a total of 36 cages. Nets were placed on the frame and dug 20 cm providing complete exclusion and avoiding ground-dwelling predators entering from outside.
The cages were established at the beginning of May and emptied of predators using pitfall traps for 48 hr after the set-up.

| Aphid inoculation
To establish a pest population, we introduced aphids Rhopalosiphum padi (L.), reared on oat plants within temperature-controlled facilities, to the treatment cages between the tillering and stem elongation oat stages (Zadoks et al., 1974; 11th-16th of June). We inoculated the cages with aphids for 5 consecutive days. For the inoculation, 100 reared aphids were placed into Petri dishes and left inside the cages.
After 12 hr, plants were checked for aphid establishment. We repeated this step until reaching 100 successful aphid establishments per cage.

| Predator community treatment
Three predator community treatments were introduced to each plot, one per cage (+A+P), on 18th of June (Appendix S2 in Supporting Information), simulating three predator community assembly patterns: Firstly, to simulate a landscape effect of generalist predators moving into the crop field from the surrounding landscape, we introduced 45 adult wolf spiders (Pardosa sp.; Roubinet et al., 2018).
Secondly, to simulate a landscape effect of generalist and specialised predators moving into the field, we introduced 45 adult wolf spiders and four adult ladybirds Coccinella septempunctata (L.). Ladybirds were chosen to represent specialised predators because they feed primarily on aphids and not on soil fauna as do carabids and spiders (Roubinet et al., 2018). We selected the density of wolf spiders based on Pardosa agrestis densities in alfalfa (Kiss & Samu, 2000); grass-clover (Birkhofer, Bezemer, et al., 2008; and barley (Riggi & Bommarco, 2019), and ladybird densities based on densities in cereal fields (Elliott & Kieckhefer, 2000).
Thirdly, we simulated a lack of predators moving from adjacent habitats and introduced no predators to 12 cages. However, all treatments were expected to contain local ground-dwelling predators, such as carabids emerging from the soil as affected by the fertilisation treatment.
To separate top-down and bottom-up effects, we used two smaller exclosures within each large cage. These exclosures consisted of a cylinder of 30 cm diameter dug 10 cm into the soil and a mesh net (mesh size smaller than 0.6 × 1.6 mm) preventing grounddwelling predators from entering. Each small exclosure was 1.2 m high letting plants grow with minimum disturbance (Appendix S1 in Supporting Information). To measure plant performance without herbivore pressure, we established one exclosure with no aphids nor predators (−A−P). To measure the aphid population growth in the absence of predators, we established another small exclosure with aphids and no predators (+A−P; Appendix S2 in the Supporting Information). Each time we visited the cages, we checked the exclosures to ensure they were free from emerging predators.
We measured the temperature and humidity variation in cages

| Measurement of aphid growth
We measured aphid population growth by counting aphids in each cage 12 times between the 16th of June and the 3rd of August. At each count, we randomly selected 10 consecutive plants at five crop rows. For each group of 10 plants, we noted the number of aphids.
We also measured aphid the growth inside the predator exclosure (+A−P; total number of aphids/total number of plants) throughout the same period, but only six times minimising the risk of predators entering the exclosure.

| Local generalist predators and soil fauna
To examine the soil fauna community (i.e. collembolans, mites and dipterans), we extracted two core soil samples per cage (5 cm diameter and 18 cm depth) three times (8th of May, 15th of July and 4th of August). Fauna extractions from soil cores were performed with Tullgren funnels and individuals were counted as the total amount of specimens per soil core sample.
We assessed two aspects regarding the carabid community. First, we estimated the community composition of emerging carabids just after the cages +A+P were built, but before aphids were inoculated.
After the cages +A+P were built, we removed all ground-dwelling predators using pitfall traps for 48 hr. These were not counted or re-introduced into the cages as the aim was to remove any predator from inside the cages before measuring 'true emergences' of soil predators. Once the cages +A+P were emptied, we quantified the ground beetle abundance emerging from the soil with three sampling rounds performed 2 weeks before aphid inoculation. Pitfall traps were kept open 24 hr and contained clay balls to provide the beetles with shelter and to keep them from escaping. Abundances of carabids were quantified in situ. All catches from the three rounds were released back into the respective cages.
We further estimated the density of ground-dwelling beetles during the rest of the experiment as a measurement of their activity.
We opened pitfall traps for 2 hr, for minimal disturbance, at three sampling rounds (26th of June, 11th and 24th of July) after aphid inoculation and quantified the density of carabids in situ. All catches were released back to the cages.

| Yield
To obtain the total biomass of the grain (caryopses) and the straw in each large cage, we hand-harvested 0.5 × 0.5 m of the crop as well as the total crop in both (+A−P) and (−A−P) exclosures (0.07 m 2 ). We oven-dried the samples at 70°C over 48 hr and weighed them.

| Statistical analysis
All data analyses were performed in R (version 3.6.2, R Core Team, 2014) using linear mixed models (lme4 package, Bates et al., 2015) and removing non-significant interactions with backward stepwise removal. For analysing differences between levels at each factor, we used post hoc Tukey's tests.

| Soil fauna
To examine the effect of soil fertilisation and predator community treatments on the total abundance of soil fauna, we built a model with sampling round (May, July and August), fertilisation treatment, predator community treatments, the interactions between fertilisation treatment and predator community treatment interactions as fixed factors, and cage nested within the block as random factor.
To compare the emergence of carabids between fertilisation treatments before adding the spiders and ladybirds, we calculated the total number of individuals emerging per cage before the aphids were inoculated. Effects on the total emergence of carabids were tested in a model with fertilisation treatment and date (every 3rd day during 9 days before the aphids were inoculated) as fixed factors, and cage nested within block as random factor.
To compare carabid densities among treatments during the experiment, that is, after the aphids were inoculated, we used the catches per cage and sampling date after the aphid inoculation. We assumed that the density of carabids during the experiment could be affected by the introduced predators via interference. Therefore, for the carabid density analysis we included both the fertilisation treatments and predator community treatments, their interaction and the sampling round (26th of June, 11th and 24th of July) as fixed factors, and cage nested within block as random factor.

| Herbivore growth rates
To examine aphid population growth, we first calculated aphid density as the average number of aphids per plant (tiller) in each cage.
To estimate aphid population growth during the exponential growth phase (Costamagna et al., 2007), we excluded the last two sampling days when aphid populations had begun to decrease. We tested the effect of the fertilisation treatment, predator community treatments and their interaction on the log-transformed aphid densities. For this model, we used day of sampling, and its interaction with fertiliser and predator community as a fixed factor, and cage nested within block as a random factor.

| Yield and plant biomass
To examine the effect of fertilisation and predator community treatments on crop yield, we first calculated grain (caryopsis) and straw biomass in each cage as weight per square metre. We then tested the effect of the fertilisation treatments and the predator community treatments on the total plant biomass and grain/ straw ratio using cage nested within block as random factor in the model.

| Top-down and bottom-up effects
To investigate differences between top-down and bottom-up effects on aphid densities and on plant biomass among the fertilisation and predator community treatments, we did the following calculations: Log Top-Down effect size on aphids (LTD aphids ) is measured by comparing aphid densities at the aphid growth peak in cages (+A+P) and the corresponding predators' exclosure (+A−P) at aphid peak density using: LTD aphids = ln Aphid density cage( + A + P) Aphid density exclosure( + A − P) . Using mixed models, we tested the effect of the fertilisation treatment, the predator community treatments and their interaction on the LTD aphid and LTD plant , and the effect of the fertilisation treatment on LBU aphids and LBU plant . Cage nested within block was included as random factor for LTD, while only block was used as random factor for modelling LBU as the fertilisation treatment was applied at the cage level.

| Soil fauna
Fertilisation and predator community treatments did not affect the total amount of soil mesofauna (other than carabids).

| Herbivore growth
Fertiliser type affected aphid population growth over time (Table 1b;

| Top-down and bottom-up effects
Top-down effect size on aphid densities (LTD aphids ) was affected by fertilisation (Table 1c; Figure 4a) and predator community (Table 1c; Figure 4b), but not by their interaction (Table 1c) Table 1c). Fertilisation treatments had no detectable bottom-up effect on the aphid population, that is, comparing F I G U R E 3 Fertiliser effect on (a) average total plant biomass (g/m 2 ); (b) average grain biomass (g/m 2 ); and (c) ratio of average grain over straw biomass (g/m 2 ) in the harvested area (cage) per fertilisation treatment ±95% confidence intervals. *indicates a significant difference (p < 0.05); ***indicates a significant difference (p < 0.001) between two treatments after post hoc analysis (n = 4) * *** *** (a) (b) (c) +A−P between fertilised and control treatments, but affected plant biomass, that is, comparing -A-P between fertilised and control treatments (Table 1c; Figure 5).

| D ISCUSS I ON
We show that fertiliser type and the composition of highly mobile predators moving into crop fields from the surrounding landscape can drastically reduce herbivorous pest populations. In particular, the combination of predators immigrating into the crop field from the surrounding landscape, and nutrient management with organic fertiliser enhance pest suppression via different mechanisms. Generally, predators drive top-down effects and fertilisation triggers bottomup control. However, fertilisation with manure also mediated topdown control by benefiting local predators overwintering in the soil.
Hence, local and landscape-level effects interplay in determining predator and herbivore abundances and crop yield.
The greatest aphid suppression occurred in plots fertilised with manure where specialist predators (ladybirds) were added. Overall, the aphid population growth was lower in the presence of predators (+A+P; locally emerging and immigrating from the landscape) than in their absence (+A−P), and this effect was independent of the fertilisation treatment. The suppression was particularly large for treatments supplemented with specialist predators (ladybirds) as compared with treatments with only locally emerging predators from the soil, or with locally emerging predators supplemented with spiders. This attests to a high suppressive capacity of ladybirds as a taxon that will typically emerge in large numbers after overwintering in refugia and immigrate to crop fields (Honek et al., 2017).
Spiders played a smaller role in suppressing aphids. We found no differences in aphid population growth rates between treatments with predators emerging from the soil and treatments with spiders added.
While spiders can provide efficient pest suppression (Michalko et al., 2019), their lack of influence in our experiment could be due to microhabitat mismatch between wolf spiders and the aphid species we studied. Wolf spiders are primarily ground-dwellers, while R. padi aphids are leaf-dwellers, resulting in low encounter rates.
Mineral and manure fertilisation strengthened bottom-up effects (LBU plant ). However, manure fertilisation also indirectly benefited top-down control of aphids by enhancing carabid beetle numbers emerging from the soil. Aphid population growth rates were lower in the manure-treated plots compared with mineral-treated and unfertilised plots. A likely explanation is that fertilisation with manure improves conditions for ground-dwelling predators by bolstering a diverse soil fauna, which provide alternate prey for carabids, and improves habitat structure in the soil (Riggi & Bommarco, 2019).
Alternative preys are important for sustaining carabids when crop pest populations are low, and for larval stages of the beetles residing in the soil (Leroux & Loreau, 2008;Liu et al., 2016). Several grounddwelling beetles are well-adapted to agricultural landscapes, and lay their eggs and overwinter in the crop field (Hanson et al., 2016;Kromp, 1999). Manure fertilisation might further improve soil texture or biochemistry benefitting the overwintering survival of adult carabids and their larvae (Lövei & Sunderland, 1996). Interestingly, we found that in manure-fertilised soils, ground-dwelling predators already present in the soil were, by themselves, able to control the aphid population growth (Figure 2).
Contrarily, we recorded the highest aphid densities and growth rates and the lowest top-down suppression in inorganically fertilised plots, consistent with the hypothesis that fast-growing plants are nutritious to herbivores and promote their population growth (Aqueel & Leather, 2011;De Bruyn et al., 2002). Overall, we also found the For an effective pest suppression, it is important to support both, predators immigrating into the crop fields from the surrounding landscape and the local predator fauna within the crop field.
Generalist predators already present in the crop field were, in organically fertilised soils, able to control aphid abundances by themselves.
However, if a crop field does not already contain high densities of local predators, management aiming to enhance immigration of specialist predators from the surrounding landscape becomes crucial for biological control. Ladybirds, such as Coccinella septempunctata, are habitat generalists and track food resources in the landscape. A management strategy to enhance ladybirds is to build a resource continuity in the landscape (Honek et al., 2017;Schellhorn et al., 2015).
Heterogeneous landscapes can in this way support a wider range of generalist predators, potentially leading to enhanced levels of natural pest control (Feit et al., 2019). We also need to consider species traits to understand the predation efficiency of assembled predator communities (Schmitz, 2017). For example, in retrospect, selecting leaf-dwelling spiders instead of ground-dwelling ones might have rendered a stronger top-down suppression. Further, aphids are not high-quality food items for some spider species which might switch to more nutritious soil prey (Toft, 2013).
Higher efficiency of specialised predators in reducing pest densities was not translated to an increase in plant biomass. Top-down processes are the main mechanisms explaining pest suppression in crops (Bianchi et al., 2006). However, strong bottom-up effects can mask this relationship under some circumstances. In our experiment, plants that received inorganic fertiliser had higher infestation levels than the ones receiving manure, and at the same time, inorganically fertilised had the highest total plant and straw biomass. However, no differences in plant biomass were found between fertilisation treatments in the predator exclosures (+A−P) and in the total exclosures (−A−P; Appendix S5 in Supporting Information). These nonsignificant differences point at few interesting possibilities. First, the lack of detectable differences among +A−P exclosures implies that plants receiving fertiliser (manure or inorganic) compensated for the biomass loss by herbivory by growing more biomass. Second, the lack of detectable differences among −A−P exclosures demonstrates that fertilisation type did not affect yields of not attacked plants.
An alternative is that inorganically fertilised plants are better able to compensate by growing more biomass when attacked by herbivores (Garratt et al., 2011;Mäder et al., 2002). Thus, it is plausible that both manure and inorganic fertilisation enhance plant biomass,

| Application and synthesis
We show that manure fertilisation increased local carabid emergence from the soil. This enhanced the top-down herbivore suppression reducing their population growth. Inorganic fertilisers promoted bottom-up forces by directly increasing plant biomass, thereby compensating for increased herbivore density. When the local predator community was not able to suppress the aphid population (in the inorganic and no fertilisation treatments), predators moving into the crop fields became essential for pest control.
Overall, the most effective top-down suppression of herbivorous crop pests was attained by enhancing the local predator community by adding organic fertilisers and specialist predators moving from the surrounding landscape. Managing landscapes to promote mobile and specialist predators emerges as particularly important for crop fields receiving no manure amendments. We suggest the active promotion of both local predators within the crop field and mobile predators in landscapes to secure crop protection and yields.

ACK N OWLED G EM ENTS
The authors are grateful to David Åhlén, Carol Högfeldt and Gerard Malsher for their valuable help with the fieldwork. They thank Sofia Delin and Erik Zakrisson for technical assistance at the long-term trials. They thank the two anonymous reviewers whose suggestions helped improve this manuscript. They acknowledge the Faculty of Natural Resources and Agricultural Sciences at the Swedish University of Agricultural Sciences for funding a faculty professor starting grant to R.B., and the long-term field experiment in Lanna.
Funding was further provided by the Swedish research council FORMAS (2016-01168) to R.B. and T.R.

AUTH O R S ' CO NTR I B UTI O N S
G.A., L.R., K.M., T.M. and R.B. provided the ideas and designed the methodology; G.A. collected the data, analysed the data and led the writing of the manuscript. All the authors contributed critically to the drafts and gave final approval for publication.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available via the Dryad Digital Repository https://doi.org/ 10.5061/dryad.3n5tb 2rgp (Aguilera et al., 2021).