Agroecological farming uses crop and non-crop plant biodiversity to promote beneficial insects supplying pollination and biocontrol services to crops. Non-crop plants (sown or weeds) are integral to supporting these beneficial insect species interactions. How the uplift of biotic complexity by agroecological management (crop diversification, ecological infrastructure) influences mutualistic and antagonistic insect interactions regulating the reproduction of non-crop plants remains less understood. 

Using a pesticide-free farm-scale (125 ha) agroecological experiment, we tested how the individual reproduction of pollinator-dependent, non-crop plant species with different flowering phenology (Cyanus segetum, Centaurea jacea) and their mutualistic (pollinator) and antagonistic (seed herbivore–parasitoid) insect interactions were affected by agroecological practices. 

Seed set and species interactions of replicate C. segetum and C. jacea randomly introduced to field margins was correlated with floral resource heterogeneity at focal plant (e.g., flower display size), local community (floral richness/abundance driven by sown wildflower or grass margins), and local landscape (crop diversification, area of semi-natural habitat or mass flowering crops) scales.  

At the seasonal peak of non-crop floral diversity and abundance, antagonistic interactions weakly regulated C. segetum seed set with gains from pollinator activity predominating. Conversely, C. jacea, which flowered past the peak of non-crop floral diversity/abundance benefited from the promotion of seed herbivore parasitism and pollinator activity by the local landscape cover of semi-natural habitat and mass flowering crops.

Synthesis and applications. Agroecological management produced spatial and-temporal gradients in crop and non-crop floral resources that interacted to modify pollinator or seed herbivore-parasitoid interactions and seed set of Cyanus segetum and Centaurea jacea plants. The degree of phenological overlap between C. segetum and C. jacea flowering and floral resources in the local community or landscape dictated the type and level of exposure to insect interactions influencing reproduction. Design of agroecological practices to deliver pollination and biocontrol services must consider how effects will vary with species traits and the ensemble of mutualistic (pollination) and antagonistic (herbivory, parasitism) interactions governing non-crop plant reproduction. Agroecological management supporting beneficial insect interactions may feedback to help restore functional non-crop plant populations and associated biodiversity, potentially reducing the frequency of management interventions (e.g., re-sowing wildflower strips). 

Experimental design

The experiment was performed (2019) at the INRAE CA-SYS platform (Bretenière, France, 47°19’06.7”N 5°04’17.6”E), an arable farm-scale system experiment (125 ha). This aims to test pesticide-free, biodiversity-based agroecological management utilising diverse spatio-temporal crop rotations (wheat, barley, corn, soybeans, peas, chickpeas, lupins, mustard, rapeseed, sunflower), tillage regimes (±) and planned ecological infrastructure (permanent semi-natural habitat, wildflower or grass-legume strips) (see Appendix 1 and Fig. S1; Vanbergen et al., 2020; Petit et al., 2021).  Sixteen plots ≥150 m apart were established in the centre of wildflower (n = 10 plots) or grass-legume (n = 6 plots) field margins (species composition – Table S1) across the CA-SYS platform (Fig. S2). Into each plot, we randomly transplanted three triplets of C. segetum and C. jacea (nine randomly selected plants per species per plot = 288 plants), with individual plants within a triplet 50 cm apart and triplets separated by 1 m within a plot (Fig. S2). No permits or special licences were required for any of the fieldwork or sampling.

 Focal plant species

Cyanus segetum (syn. Centaurea cyanus L.) and Centaurea jacea L. were selected as focal plant species because they are widely distributed in Europe, depend on pollinators for reproductive success, and attract seed herbivores (Tephritidae; Diptera) and their parasitoids (Hymenoptera) (Steffan-Dewenter et al .2001; Ouvrard et al., 2018). C. segetum is an annual, pseudo-self-compatible, archaeophyte species flowering May-July with a segetal habit (i.e., it grows in cereal fields) and often included in wildflower seed mixtures. C. jacea is a self-incompatible perennial that flowers between June and October occurring in field margins and semi-natural areas. C. segetum seed (http://www.arbiotech.com), was germinated (22 ± 3°C; 16h light: 8h dark) and maintained in controlled environment cabinets (04/02/-04/03/2019 at 8-10°C 12h:12h; thereafter 15-18°C) until transplantation into field plots (11-12/03/2019). C. jacea replicates were field collected (07-13/03/2019) at the pre-reproductive stage (rosette ~10 mature leaves) from nearby locations (CA-SYS platform: 47°14’32.2”N 5°05’11.8”E; Dijon: 47°19’06.7”N 5°04’17.6”E; Champdôtre: 47°10’42.5”N 5°17’02.0”E) and transplanted into the plots (12-15/03/2019) after washing their roots free of soil.

Focal plant chemical quality

To quantify the individual capacity for reproduction (seed set) according to their uptake of soil nitrate or ammonium, we took a random sample of mature leaves (~5 g) prior to the onset of flowering (13/05/2019) from each C. jacea/C. segetum. After oven drying (48h, 80°C) and milling (diameter ≤ 80 µm, re-dried 80°C, 24h), we used a Thermo Scientific FLASH 2000 Organic Elemental Analyzer™ to quantify the C/N content in 4-6 mg of these ground tissue samples. Sample injection and oxidisation (O₂ under helium flow at 950°C) followed by reduction (Nox) and removal of excess O₂/H2O (Cu at 750°C/anhydron) yielded N₂, and CO₂. Gas chromatography (porapak column 40°C in stationary phase) separated and detected (catharometer) the component N, CO₂ and He. Integrated examination of signal peaks and calibration curves allowed determination of % N and % C dry weight (g).

Mass flowering crops, non-crop vegetation and semi-natural habitat

We quantified the abundance and species richness of the non-crop (dicotyledon) plants once per month (from May to August) in 16 2m x 100m transects along field borders centred on each focal plant plot (Fig. S2). Non-crop floral species richness and floral abundance (Table S1-S2) were recorded in six quadrats (2m x 50cm) systematically placed at 20m intervals along the transect. Flowering plant species (cumulative count) were identified (Appendix 2 for keys) and the total number of inflorescences (individual flower/umbel/spike/capitulum) per quadrat was derived for all species per plot per sampling period (Fig. S2).

Within a radius of 300m of each plot, we quantified the local landscape composition (ArcGIS Pro 10.8) as: i) mass flowering crop species richness; and the proportional area of ii) mass flowering crops and iii) semi-natural habitats (woodland, hedges, grass and wildflower strips, vegetation along pathways/tracks) (Table S2-S3). The occurrence of tillage in the fields adjacent to the focal plant plots was pre-determined by the CA-SYS experimental design (Table S2, Fig. S2).

Mean non-crop floral species richness and floral abundance, the proportional area and species richness of mass flowering crops were calculated separately for C. segetum and C. jacea to coincide with their species-specific flowering periods and represent their phenological overlap with crop and non-crop floral resources (Fig. 1).

Focal plant biomass, flower head production and seed set

After flowering (16/07/2019 for C. segetum and 09/09/2019 for C. jacea), the focal plant flower stalks were harvested and placed in muslin bags. After seed head insects emerged (below), we measured the dry weight of plant biomass (g), the number of flower heads (capitula), and the count (n) and mass (g) of seeds per replicate plant. A priori we expected these plant metrics to be highly correlated (Appendix 3 and Fig. S4, all P < 0.001). Consequently, we chose seed yield (count) as a direct measure of reproductive potential that also indicated other aspects of focal plant performance.

Pollinator visitation  

Both focal plants and transects were observed (09:30-17:30, dry weather, little wind, ≥14°C) for insect pollinators (mainly Hymenoptera and Diptera, with a single Lepidopteran) fortnightly (C. segetum: late May to mid-July; C. jacea: mid-July-September). Sampling effort was standardised by observing pollinator visitation for a fixed duration of 30 minutes (15 minutes each per focal plot and transect). The order of sampling (plot + transect) on each date was randomized to avoid introducing a systematic bias due to the time of day. Pollinator species observed legitimately visiting a flower (contact with stamen/carpel, nectar or pollen feeding) were captured, killed, and stored (70% ethanol) until identification (ZEISS Stemi 2000-C microscope, see Appendix 2 for standard keys, Table S4).

Observations of focal plants provided the number of pollinator individuals and species per focal individual over the season. We supplemented this with transect data (10 surveys) giving pollinator abundance and species richness on non-focal C. segetum (and hybrids with horticultural varieties) or C. jacea during the season. This assumed that pollinators foraging on C. segetum/C. jacea in transects could have visited focal plants (c.f. directly observed interactions) and so comprised a potential pool of visitors active in the vicinity (100m) of our focal plants. Therefore, pollinator abundance and species richness were the sum of insect visits and cumulative count of different species recorded per focal plant individual and focal plant species per transect.

Seed herbivores and parasitoids.

After a minimum of 2 months storage (20 ± 3°C), seed herbivores (Tephritidae; Diptera) and their parasitoids (Hymenoptera) emerged within the muslin bags containing the harvested focal plant capitula.  Tephritid seed herbivores and parasitoids were counted and identified using standard keys (Appendix 2) to the highest taxonomic resolution possible (Table S5) (ZEISS Stemi 2000-C). Seed herbivory rate was estimated as the proportional count of seed herbivores per total number of seeds (n/N) per plant. Parasitism rate was the proportional count of parasitoids (n/N) per total potential hosts – estimated as the sum of herbivores and parasitoids emerging per plant and assuming a 1:1 host-parasitoid relationship with no hyperparasitism (Vanbergen et al., 2006, 2007).