Plant-Driven Effects of Wildflower Strips on Natural Enemy Biodiversity and Pest Suppression in an Agricultural Landscape in Hangzhou, China

Abstract

Agricultural intensification has led to biodiversity loss and compromised ecosystem services, necessitating sustainable pest management strategies. This study evaluates the efficacy of wildflower strips (WFS) in enhancing natural enemy communities and suppressing pest activity in rice-wheat rotation landscapes of eastern China. An experiment compared WFS (10-species mixtures) with natural grass strips (CK) across biodiversity, functional traits, and pest dynamics. WFS significantly increased parasitic wasp α-diversity (species richness: +195.5%, activity density: +362.0%) and suppressed pest (Armadillidium vulgare) populations by 68%, primarily through female-biased sex ratios and functional trait shifts. Key species like Lindenius mesopleuralis and Ectemnius continuus emerged as indicators of WFS habitats. Spider communities showed no β-diversity differentiation but exhibited functional guild shifts (e.g., web-building specialists). Plant community composition, particularly floral resource availability and phenological continuity, drove natural enemy assembly and pest regulation, outperforming the CK group in rare species conservation. Our findings highlight WFS as a precision tool for enhancing pest control through targeted plant selection and trait-mediated interactions. This study advances the understanding of habitat-driven pest regulation, providing a framework for optimizing ecological intensification in agroecosystems.

1. Introduction

Agricultural landscapes, as one of the most human-modified terrestrial ecosystems, are structurally defined by the spatial mosaic of productive farmlands (arable fields and orchards) and seminatural habitats (woodlands, fallows, and flower-rich meadows) [1]. These coupled “production-ecology” systems not only sustain food security but also maintain critical ecosystem services such as pollination and pest regulation, forming the foundation of ecological security [2,3]. However, global agricultural intensification has triggered a homogenization crisis, with seminatural habitats declining at an annual rate of 1.2% [4]. This process has reduced landscape connectivity by 37% [5], cascading into biodiversity loss and food web simplification [6]. Consequently, key ecosystem services have been severely compromised—pollinator declines have reduced global crop yields by 5–8%, while natural enemy depletion has increased annual pest control costs by $2.3 billion [7,8]. Reconstructing biodiversity while ensuring agricultural productivity has thus become central to achieving Sustainable Development Goals (SDG 2 and SDG 15) [9].

Landscape-level habitat management offers a promising solution. By establishing heterogeneous habitat patches, these interventions provide stepping-stone corridors and niche differentiation opportunities for biological communities [10]. Among these, wildflower strips—artificially established flower-rich margins—have emerged as a cost-effective measure due to their minimal land footprint and high ecological returns. Specifically, strategically designed mixtures of Asteraceae and Apiaceae species can deliver 3.7 times higher pollen resources per unit area compared to natural grasslands [11]. Under the European Common Agricultural Policy (CAP), wildflower strips have increased edge-dwelling natural enemy diversity by 41% and parasitoid abundance by 2.3-fold [12]. The underlying mechanisms involve three ecological processes: (1) resource provisioning through the phenological alignment of flowering periods to meet natural enemies’ nutritional demands [13]; (2) microhabitat modulation, where canopy shading reduces surface temperature by 4.2 °C, extending the diurnal activity of carabid beetles [14]; and (3) community assembly, where plant functional diversity drives functional group differentiation (e.g., niche partitioning between chewing and piercing-sucking insects) [15]. Nevertheless, current studies predominantly focus on single taxa, leaving multi-trophic interaction networks poorly understood [16].

Plant community design fundamentally drives functional divergence in wildflower strips. Two competing paradigms dominate; the high species richness hypothesis posits that plant diversity supports broader natural enemy functional groups via resource complementarity [17]—for instance, syrphid species richness strongly correlates with Asteraceae richness in 6-species mixtures [18]. Conversely, the keystone species hypothesis advocates a targeted use of Apiaceae plants (e.g., Foeniculum vulgare), whose volatile secondary metabolites (e.g., β-ocimene) specifically attract dominant natural enemies [19]. Preliminary evidence shows F. vulgare-enriched strips increase parasitoid visitation by 3.5× but risk non-target exposure through pheromone interference [20]. This underscores the need to model cascading responses of “plant functional traits → natural enemy assemblages → pest population dynamics” to elucidate ecological trade-offs.

Notably, existing research predominantly focuses on highly intensified Western agroecosystems [3], while East Asian landscapes—,where natural grassy margins already sustain significant natural enemy conservation remain understudied. These semi-natural edges harbor co-evolved plant-insect interaction networks that may outperform artificial wildflower strips in ecological stability [21]. Consequently, in regions with high baseline habitat heterogeneity (>18% seminatural habitat coverage), a critical question arises: Can wildflower strips surpass natural vegetation by strategically optimizing plant composition to reconfigure natural enemy functional guilds?

In recent years, the population of pillbugs (A. vulgare) has sharply increased in agricultural environments, particularly following the implementation of conservation agriculture practices such as reduced tillage, residue retention, and changes in crop management, which have created more favorable conditions for their survival, leading to a continuous population growth [22]. This ecological shift has made the damage caused by pillbugs to crops increasingly severe, particularly in humid environments, where pill-bugs not only directly feed on plants but also indirectly affect crop growth by competing for food resources with other pests, such as slugs (Deroceras reticulatum). However, traditional pesticide control measures are often ineffective, as pillbugs have shown resistance to many common pesticides, complicating pest management efforts [23]. Wildflower strips, as a sustainable agricultural landscape practice, can provide diverse ecological habitats and may have a positive impact on pillbug population control. Can appropriate landscape ecological agricultural measures, such as wildflower strips, alter the habitat and reproductive conditions of pillbugs, thereby indirectly reducing their damage to crops?

This study establishes a wildflower strip system in eastern China Integrating multi-scale monitoring (plant trait quantification, natural enemy guild dynamics, pest dispersal tracking), we aim to answer the following three questions:

(1)

How do wildflower strips alter the spatial distribution and diversity of natural enemy communities relative to natural grass strips?

(2)

What are the cascading effects of habitat-driven natural enemy changes on pest population dynamics?

(3)

Which plant community traits (e.g., species richness, floral phenology) drive natural enemy assembly and pest regulation in wildflower strips?

2. Materials and Methods

2.1. Study Site

The experimental investigations were carried out at Qiaosi State-Owned Farm, a historic agricultural reclamation area located in the estuarine zone of the Qiantang River (120.3509° E, 30.3673° N; elevation 5.48 m, as shown in Figure 1). Characterized by a subtropical monsoon climate, the region experiences distinct seasons with a warm and humid annual climate. The mean annual temperature is approximately 16–17 °C. Summers are hot and rainy, with average temperatures in July and August exceeding 28 °C. Influenced by the Meiyu season and typhoons, these months witness concentrated precipitation, with the monthly average rainfall often surpassing 150 mm. Winters are relatively mild and dry, with the average temperature in January around 4–5 °C. The annual average precipitation is about 1300–1400 mm, predominantly occurring in spring and summer, while autumn and winter are relatively arid.

In terms of soil conditions, the area, formed through tidal flat reclamation, predominantly comprises silty clay with weak alkalinity. Despite its deep soil layers, the soil exhibits mediocre air permeability. Years of agricultural development, involving long-term cultivation and fertilization practices, have altered soil fertility, and certain areas have experienced a degree of salinization. These climatic and edaphic factors not only impact crop growth but also shape the unique local ecosystem, with pronounced seasonal variations, providing an abundant observational context for the study of ecosystem service functions.

Established through systematic tidal flat reclamation in 1950, this 72-year-old agricultural facility has developed distinctive geomorphological features, including extremely low topographic relief (elevation gradients <0.5‰), engineered hydrological networks connected to the Qiantang River system, and standardized plot configurations separated by artificial drainage channels. These geomorphic characteristics, combined with the intensive agricultural management measures implemented since 2005, have created an ideal controlled environment for researching landscape-ecosystem interactions, due to the minimized spatial heterogeneity and reduced confounding variables.

As a major horticultural production base in Hangzhou, the site’s ecological monitoring records demonstrate a gradual erosion of biodiversity, which is particularly evident in pollinator communities and predatory arthropod populations. This ecological degradation is correlated with significant declines in orchard productivity indicators, such as reduced fruit set rates and increased reliance on chemical pest control, indicating compromised ecosystem service functionality. The selection of this experimental site is strategically justified by its representative agricultural intensification pattern. The landscape simplification process has reached a threshold that allows for a clear observation of ecosystem service trade-offs. Moreover, the long-term institutional management framework ensures the consistent implementation of experimental protocols, while restricted access policies effectively isolate anthropogenic disturbance factors, creating favorable conditions for controlled ecological experiments.

2.2. Experimental Design

The experimental group consisted of artificially established wildflower strips along field edges, while the control group maintained natural vegetation strips, with four spatial replicates for each treatment. Each strip measured 80 m in length and 1.5 m in width (total area 120 m²), separated by over 1000 m between adjacent strips to eliminate spatial interference. Three biodiversity monitoring points (Consisted three designated sites along the strip and each point was equipped with one trap designed for sampling flying arthropods and two traps intended for capturing ground-dwelling arthropods) were positioned at 0 m, 40 m, and 80 m intervals within each strip, resulting in 24 independent observation units (4 strips × 3 points × 2 treatments).

The study site consists of various land use types, with seminatural habitats covering 18% of the total area, arable land accounting for approximately 60%, and construction land making up around 12%. The remaining area includes field roads, management facilities, and other minor components. The regional agricultural practice follows an annual rice-wheat rotation system. In the experimental year, all fields were planted with rice, and wheat was sown in the subsequent season as part of the crop rotation cycle. Regarding the layout of experimental blocks, this study utilized multiple rice fields. The study was conducted across six adjacent rice fields in the same region, each containing a paired wildflower strip and natural grass strip (control) to ensure consistent soil and microclimate conditions between treatments. Wildflower strips and grass blocks were established along the field margins (i.e., the terrace edges), directly adjacent to the cultivated area (distance from the crop field = 0 m). All fields in this study were planted with the same crop (rice) during the experimental period, ensuring consistency in agricultural management across blocks.

The species composition integrated domestic and international literature databases with germplasm availability, ultimately selecting 10 functionally complementary species: Trifolium pratense (10%, 0.6 g/m²), Salvia coccinea (12%, 0.72 g/m²), Coriandrum sativum (10%, 0.6 g/m²), Glycyrrhiza uralensis (10%, 0.6 g/m²), Myosotis arvensis (10%, 0.6 g/m²), Viola philippica (10%, 0.6 g/m²), Achillea millefolium (10%, 0.6 g/m²), Mentha canadensis (10%, 0.6 g/m²), Medicago sativa (12%, 0.72 g/m²), Malva arborea (6%, 0.36 g/m²). Wildflower strips were established on April 1, 2023, with a total seeding 255 rate is 6 g/m². Species selection followed five criteria: achieving phenological continuity through different-period blooming species; enhancing functional diversity; prioritizing native species; employing diverse color schemes for aesthetic enhancement; and selecting drought-tolerant varieties to minimize maintenance. We have prioritized the selection of native species that were accessible to us, along with some naturalized species commonly used in Chinese landscapes. No exotic species with distinct invasive characteristics were employed. Based on existing research, we have found no evidence that the selected plants exhibit significant invasive effects. Although we strongly advocate for the use of native plants, increasing their proportion in landscape design is a gradual process. Therefore, we have adopted a mixed-species strategy, coupled with active monitoring, to mitigate potential impacts from exotic invasive species. From the perspective of current research, we believe that, ideally, native species should be given preference whenever possible. To establish the wildflower strips, we applied three technical phases:

(i)

Pre-treatment: Glyphosate-isopropylamine (41% SL, 3.5 L/ha) was applied 28 ± 2 days prior, followed by 72-h waterlogging at 5 cm depth to induce weed germination, and subsequent rotary tillage at 15 cm depth. Soil amendment adjusted acidic soil pH to 6.0–6.5 using elemental sulfur and enriched organic matter to 2.3% with composted manure.

(ii)

Sowing: Drill methods were employed with 2–3 cm furrow depth and 30 cm row spacing, distributing hydro-primed seeds mixed with vermiculite substrate (1:3 ratio) at 4–6 g/m². During germination (0–21 days), drip irrigation maintained 18% ± 2% volumetric water content, followed by the farm’s irrigation system delivering 25 mm water twice weekly.

(iii)

Post-establishment management: Dynamic thinning was conducted to maintain plant density at 30–35 individuals/m², with strict prohibition of pesticides and fungicides. Invasive weeds were manually removed quarterly, maintaining a frequency of <5%. The total experiment flow can be seen in Figure 2.

2.3. Natural Enemies and Target Pest Survey

2.3.1. Selection of Target Taxa

In this study, spiders and parasitic/predatory wasps were selected as focal natural enemy taxa based on their pivotal agroecological functions [24,25,26]. These arthropod groups exhibit dual ecological significance as biodiversity bioindicators and biological control agents [27]. Spiders regulate multiple pest species through generalist predation strategies, whereas parasitic/predatory wasps provide host-specific control mechanisms [28]. Their standardized trapping and identification protocols enable quantitative cross-habitat comparisons of population dynamics.

The ecosystem service profiles of these natural enemies were investigated using the common pillbug (A. vulgare) as a model pest organism [29]. This isopod species demonstrates trophic plasticity under threshold density conditions, transitioning from detritivorous feeding to active phytophagy. Although typically functioning as decomposers at low population densities, these organisms exhibit significant population growth under humid conditions. At higher densities, they shift to feeding on tender plant tissues such as stems, leaves, and buds, thereby becoming agricultural pests. Their hard exoskeleton further complicates chemical control measures. Its crop-damaging potential is amplified by morphological adaptations (calcified tergites) and a humidity-dependent reproductive strategy. As the predominant pest species in our study region, A. vulgare’s population dynamics and interactions with natural enemy communities provide an ideal model system for examining integrated pest management (IPM) strategies mediated by wildflower strip interventions.

2.3.2. Methods for Biodiversity Monitoring

Wildflower strips monitoring conducted from 21 June to 9 August 2023, involving eight consecutive sampling cycles (7 days each) during peak flowering. Three sampling devices spaced 40 m apart were deployed per transect in both experimental and control groups. Each device comprised in both wildflower strip group (WFS group) and natural grass strip (CK group): (1) Aerial insect module: One tri-colored pan-trap (yellow, white, blue) mounted on welded rings, filled to two-thirds capacity with water and surfactant for parasitic wasp collection (above-ground height was approximately 1.2–1.3 m) [30,31]; (2) Ground-dwelling module: Two directional pitfall traps containing 200 mL saturated saline, flush-mounted with metal rain shields for spider and pillbug sampling [32,33]. All traps operated synchronously during weekly 7-day cycles. Specimens from tri-colored pans were pooled, while ground samples were directionally mixed by transect orientation. All samples were preserved in 75% ethanol and taxonomically identified to the species level through morphological analysis.

Pitfall trapping was chosen as the primary method due to its standardization, cost-effectiveness, and suitability for ground-active arthropods. Visual surveys indicated that ground-active spiders dominated the local arthropod community, while web-building spiders were less common. Although pitfall trapping might underrepresent web-builders, it remains a useful indicator of ground-based natural enemy communities. We recognize the limitations of relying on a single method and the biases inherent in pitfall traps. Integrating multiple methods (e.g., visual surveys, beat sampling, suction sampling) would provide a more comprehensive understanding. We have now discussed this limitation explicitly and suggested methodological diversification for future research.

2.4. Functional Traits

To better understand how natural enemy communities differ in their functional characteristics, this study analyzed key traits related to parasitism, movement, and reproduction in parasitic wasps and spiders. Specifically for parasitic wasps, we examined: (1) body size (average length), (2) hunting strategies, and (3) male-to-female ratios. These traits were chosen because prior studies have shown their direct connections to pest control effectiveness in agricultural systems. Complete data can be found in

Table 1. Natural Enemies Functional Traits.

Category	Functional Trait	Description	Data Unit
Wasps	Body Length	Average body length of wasps, randomly sampled from three individuals per species and averaged [34]	mm
	Hunting Type	Ratio of predatory individuals to parasitic individuals [35]	Parasitic = 1; Predatory = 0
	Gender	Ratio of female to male individuals, calculated independently for different species [36]	Number of females/Number of males
Spiders	Body Length	Average body length of spiders, randomly sampled from three individuals per species and averaged [37]	mm
	Ballooning Ability	Proportion of individuals with ballooning capability among the total population [38]	Ballooning species = 1; Other species = 0
	Hunting Type	Ratio of hunter individuals to web weaver individuals [39]	Hunters = 1; Web weavers = 0

.

2.5. Plant Survey

To investigate the regulatory effects of wildflower strip community structure on predator-prey systems, standardized quadrat sampling was implemented. Five 1 m × 1 m quadrats were systematically established around each experimental unit. Within each quadrat, we recorded vascular plant species composition (identified to species level), relative abundance, mean height, plant type (wildflower/non-wildflower), species-specific projective coverage, and total coverage. Projective coverage was calculated as the percentage of vertical projection area relative to quadrat area, with total coverage being the sum of individual species’ coverage values. All surveys were conducted during peak flowering periods. In this study, we approximated floral resources by visually estimating the number of open flowers within each full sampling plot. Specifically, at each sampling point, oservers recorded the total number of visible open flowers within the designated area, and flower density was used as a proxy indicator for floral resource abundance.

2.6. Data Analysis

To analyze the distribution patterns of natural enemy communities and target pests across different groups, this study first calculated the α-diversity of target natural enemy groups (spiders and wasps). Species richness (number of species), activity density (number of individuals), Shannon diversity, and Simpson diversity were used as measures of natural enemy α-diversity, while the population density (number of individuals) of the target pest was also calculated. Initially, t-tests were employed to compare the α-diversity of natural enemies and the activity density of target pests between control (natural grass strips) and wildflower strips [40].

Subsequently, this study conducted Principal Coordinates Analysis (PCoA) based on Bray-Curtis distances to examine the β-diversity of natural enemy communities across different experimental groups [41]. The significance of community composition differences was tested using PERMANOVA with 999 permutations [42]. Indicator species analysis (IndVal) was used to identify species responsible for differences in community composition. Additionally, this study calculated the community-weighted mean (CWM) [43] and RaoQ indices of functional traits [44] closely related to ecological processes for natural enemy groups. Kruskal–Wallis (K–W) tests were employed to compare differences in functional traits [45]. To explore the relationship between changes in pest density and variations in natural enemy communities and diversity levels within wildflower strips, this study modeled and analyzed the response relationship between pest density and various parameters of the natural enemy community composition (α-diversity, β-diversity—PCoA coordinates of the first two axes, and functional traits) using generalized linear models (GLMs) based on Poisson distributions or general linear models (LMs) [46]. Pest density served as the response variable, while natural enemy community composition parameters served as explanatory variables. To eliminate collinearity, factors were iteratively removed based on VIF values until all VIF values were below 5 [47]. Stepwise regression was then used to optimize the model [48], with the final optimized model selected. Furthermore, this study employed random forest models to rank the relative importance of different natural enemy species within communities, aiming to identify key enemy species significantly regulating pest density [49].

The study also analyzed the driving relationships between wildflower strip plant community composition factors (including α-diversity, β-diversity—PCoA coordinates of the first two axes, and floral resources) and natural enemy community α-diversity, functional trait composition factors, and target pest population activity density. The model construction method was similar to that of the linear models mentioned earlier. In addition, this study used distance-based redundancy analysis (db-RDA) with hierarchical partitioning to identify significant factors of wildflower strip vegetation composition influencing the β-diversity of spider and wasp communities, ranking them based on their relative importance [50].

All analyses were conducted using R version 4.2.7 [51], with visualizations created using the ggplot2 package. Diversity indices were calculated using the vegan package, t-tests, K–W tests, and normality tests (shapiro.test) were performed using the stats package. Distance matrices, PCoA, and PERMANOVA were also calculated using the vegan package [52]. IndVal analysis was conducted with the labdsv package [53], CWM and RaoQ indices were calculated using the FD package [54], and GLMs and stepwise regression were performed using the stats package. VIF values were calculated with the car package [55], and hierarchical partitioning was conducted using the rdacca.hp package [56].

3. Results

3.1. Overall Sample Collection

3.1.1. Spider Sample Collection

A total of 276 spider specimens were collected in this study, belonging to 7 families, 13 genera, and 21 species, with 151 individuals in the experimental group and 125 in the control group by pitfall traps. At the family level, Lycosidae (wolf spiders) dominated, comprising 266 individuals (85.3% of the total), followed by Titanoecidae (13 individuals, 4.1%), while minor contributions came from families such as Phrurolithidae and Linyphiidae. Genus-level analysis revealed Pardosa (wolf spiders) as the most abundant genus (175 individuals, 56.1%), followed by Arctosa (26 individuals, 8.3%), Piratula (25 individuals, 8.0%), and Trochosa (19 individuals, 6.1%). Genera including Nurscia and Phrurolithus were less represented. Species dominance was observed in Pardosa laura (63 individuals, 20.2%), with other common species being Pardosa tschekiangiensis, Piratula piratoides, Nurscia albofasciata, Arctosa depectinata, and Pardosa pseudoannulata.

3.1.2. Parasitic/Predatory Wasp Sample Collection

A total of 134 parasitic and predatory wasps were captured by pan-traps, including 105 in the experimental group and 29 in the control group. The family Crabronidae (digger wasps) showed the highest abundance (72 individuals, 53.7%), followed by Scoliidae (47 individuals, 35.1%). Minor families included Sphecidae, Eumenidae, Ichneumonidae, and Megachilidae. At the genus level, Scolia accounted for 47 individuals (35.1%), Lindenius for 40 (30.0%), and Ectemnius for 21 (15.7%), with low representation from genera such as Sceliphron, Oxybelus, and Larra. The dominant species was Lindenius mesopleuralis (40 individuals, 30.0%), followed by Scolia sp. 1 (27 individuals, 20.1%) and Ectemnius continuus (21 individuals, 15.7%). Rare species included Sceliphron madraspatanum, Larra fenchihuensis, Coelioxys pieliana, and Scolia watanabei.

Detailed information can be found in the Supplementary Materials.

3.2. Differences in Target Taxa Communities Between Wildflower vs. Natural Grass Strip

3.2.1. Alpha Diversity

As shown in

Table 2. Alpha diversity distribution between wildflower strips and grass strips.

Diversity Index	t	df	p	Mean in CK Group	Mean in WFS Group	95% CI	WFS/CK	Sig
Wasps
Species richness 1	−2.827	21.91	0.01	1.833	3.583	[−3.034, −0.466]	195.47%	**
Activity Density 1	−4.403	16.415	<0.001	2.417	8.75	[−9.377, −3.29]	362.02%	***
Shannon-Wiener Diversity	−2.613	20.429	0.016	0.438	0.998	[−1.006, −0.113]	227.85%	*
Simpson Diversity	−1.635	16.009	0.122	0.339	0.539	[−0.459, 0.059]	159.00%
Spiders
Species richness 1	0.428	17.0	0.674	1.59	1.51	[−0.340, 0.513]	94.97%
Activity Density 1	0.231	19.8	0.820	2.14	2.05	[−0.725, 0.905]	95.79%
Shannon-Wiener Diversity	0.857	16.4	0.404	1.21	1.02	[−0.287, 0.679]	84.30%
Simpson Diversity	1.42	13.5	0.178	0.655	0.514	[−0.073, 0.355]	78.47%
Target Pest control Service
Pest Activity Density	2.5995	13.718	0.021	525.25	168.25	[61.88, 652.12]	32.03%	*

¹ Log transformed; *** indicates p < 0.001, ** indicates 0.001 ≤ p < 0.01, and * indicates 0.01 ≤ p < 0.05. The same below.

and Figure 3, wildflower strips (WFS) significantly enhanced wasp alpha diversity compared to grass strips (CK). WFS exhibited higher species richness (t = −2.827, df = 21.91, p = 0.01; WFS/CK = 195.5%), activity density (t = −4.403, df = 16.42, p < 0.001; WFS/CK = 362.0%), and Shannon diversity (t = −2.613, df = 20.43, p = 0.016; WFS/CK = 227.9%), though Simpson diversity did not differ significantly (p = 0.122). For spiders, no significant differences were observed in species richness (p = 0.674), activity density (p = 0.820), Shannon (p = 0.404), or Simpson diversity (p = 0.178) between WFS and CK. Pest activity density was significantly reduced in WFS (t = 2.60, df = 13.72, p = 0.021), with WFS values 32.0% of CK (mean CK = 525.25 vs. WFS = 168.25; 95% CI [61.88, 652.12]).

3.2.2. Beta Diversity

Principal Coordinates Analysis (PCoA) of beta diversity revealed significant compositional dissimilarities in wasp assemblages between wildflower strip habitats and natural grassland controls (Figure 4). This spatial differentiation was statistically validated through PERMANOVA (F = 2.236, p = 0.028).

Notably, population densities increased substantially across most surveyed species, with particularly pronounced benefits observed for rare taxa. Two species, L. mesopleuralis (IndVal = 0.82, p = 0.003) and E. coninutus (IndVal = 0.76, p = 0.012), emerged as statistically robust ecological indicators for wildflower strip habitats. Furthermore, several species including E. picta and C. sinensis were exclusively detected in wildflower strips.

Contrastingly, arachnid communities exhibited no significant β-diversity differentiation between habitat types (PERMANOVA: F = 1.127, p = 0.294). ISA results similarly failed to identify any spider taxa with significant indicator values for wildflower strips (maximum IndVal = 0.41, p > 0.05 across all species). For beta diversity, PCoA analysis results (Figure 3) suggest partial differentiation in the species composition of wasps within the wildflower strip groups, a process confirmed by PERMANOVA results as statistically significant (F = 2.236, p = 0.028). For instance, L. mesopleuralis and E. coninutus emerged as significant indicator species for wildflower strips. Additionally, species not found in natural grassland strips, such as E. picta and C. sinensis, were present in wildflower strips.

For the spider groups, we did not observe significant differentiation between the wildflower strips and the natural grassland control groups. Similarly, we found no significant indicator species within the wildflower strips.

3.2.3. Functional Traits (CWM)

Analysis of functional trait dynamics revealed that wildflower strip implementation moderately enhanced spider aerial activity and predation rates; however, these increases did not reach statistical significance (p > 0.05). (Figure 5).

3.3. The Suppression Effect of Natural Enemies on Target Pests in WFS Habitats

3.3.1. Wasps

As shown in

Table 3. GLM-Based mechanisms of natural enemies regulation of target pests in wildflower strips.

Groups	Term	Estimate	Std. Error	Z	p	Sig
Wasps	Intercept	4.492	0.632	6.775	<0.001	***
	Gender (Female Rate)	−3.277	1.3499	2.135	0.0327	*
Spiders	Intercept	2.3031	2.8006	0.791	0.4289
	Hunting Type (Predatory individuals Rate)	5.0522	1.984	2.228	0.0259	*
	PCoA1	1.724	0.7966	1.891	0.0586

*** indicates p < 0.001 and * indicates 0.01 ≤ p < 0.05.

, post-model optimization analyses revealed that pest suppression (targeting pillbugs) by wasp communities in wildflower strips operates through a univariate regulatory mechanism, specifically driven by the proportion of female individuals within the community. Quantitatively, each 1% increase in female proportion reduced pest activity density by 1.93–4.81 individual units.

To delineate the biological sources of this suppression effect, random forest validation identified three key species within the floral provisioning habitat: active hunting specialists including L. mesopleuralis and L. fenchihuensis, alongside the obligate parasitic species E. trilobus. Crucially, these taxa collectively exhibited female-dominated population structures, with sex-specific predation strategies potentially constituting the primary mechanism underlying observed pillbug population control.

3.3.2. Spiders

Consistent with patterns in wasp communities, spider alpha diversity indices showed no significant correlation with pillbug (A. vulgare) abundance. The functional composition of spider communities, rather than their taxonomic diversity, was identified as the primary determinant of pillbug suppression. Although web-weaving spiders comprised only a small fraction of the community, an increased proportion of these species was significantly associated with reduced pillbug abundance. While community composition analysis along the PCoA2 axis suggested potential regulatory effects through niche differentiation, this relationship only reached marginal statistical significance.

Random forest modeling revealed three spider species with disproportionate influence on pillbug activity density: the web-building specialist P. piratoides, the ambush predator A. depectinata, and the generalist feeder Mermessus sp. 1 (all p < 0.05). Their distinct predation strategies likely synergistically contribute to multidimensional pest control within the ecosystem.

3.4. Responses of Natural Enemies and Target Pests to Wildflower Strip Plant Communities

3.4.1. Wasps

At the α-diversity level, wildflower strip vegetation composition exerted differential impacts on wasp population metrics (as shown in

Table 4. Effects of Wildflower Strip Plant Community Composition on Wasp Alpha Diversity based on the GLM model.

Diversity	Plant Explanatory Variables	Estimate	Std. Error	Z	p	Sig
Species Richness	Community Composition (PCoA1)	−0.783	0.441	−1.777	0.076
Activity Density	Species Richness	0.255	0.144	1.779	0.075
	Total Coverage	−0.195	0.112	−1.744	0.081
Shannon Diversity	Simpson Diversity	−4.049	2.507	−1.615	0.141
	Community Composition (PCoA1)	−1.282	0.362	−3.542	0.006	**
Simpson Diversity	Simpson	−1.910	1.071	−1.784	0.108
	Community Composition (PCoA1)	−0.534	0.155	−3.451	0.007	**
Mean Body Length	Total Coverage	−1.364	0.771	−1.769	0.120
	Simpson Diversity	84.482	54.181	1.559	0.163
	Community Composition (PCoA1)	7.049	4.881	1.444	0.192
	Community Composition (PCoA2)	−15.865	7.614	−2.084	0.076

** indicates 0.001 ≤ p < 0.01.

). The primary findings revealed: (1) Plant community composition (PCoA1 axis) showed a marginally significant negative effect on species richness (p = 0.076), suggesting a subtle decline in wasp diversity with increasing PCoA1 values; (2) Activity density exhibited dual marginal effects—positive correlation with plant species richness (p = 0.075) vs. negative association with total vegetation coverage (p = 0.081).

Diversity metrics analysis demonstrated significant negative impacts of plant composition (PCoA1) on both Shannon diversity (p = 0.006) and Simpson diversity (p = 0.007), confirming floral species assembly as the principal driver of wasp biodiversity shifts. Morphologically, plant composition (PCoA2 axis) marginally influenced mean body length (p = 0.076), implying that vegetation-mediated resource allocation affects wasp allometry. Crucially, no significant vegetation effects were detected on functional traits like sex ratio or predation strategies.

At the β-diversity level, vegetation parameters collectively explained 48.1% of wasp community variation, highlighting strong environmental filtering. Key drivers included (1) Vegetation coverage as the dominant factor for parasitic wasp assemblages (independent explanatory power = 16.47%, total = 18.76%); (2) Balanced spatial effects from wildflower composition (combined explanatory power = 16.68% across both axes); (3) Weak contributions from plant richness (6.58%) and floral resource availability (6.7%). This hierarchy confirms that structural parameters outweigh biomass metrics in explaining community-level variation.

3.4.2. Spiders

At the α-diversity level, vegetation parameters exhibited contrasting effects on spider assemblages (as shown in

Table 5. Effects of Wildflower Strip Plant Community Composition on Spider Alpha Diversity based on GLM model.

Diversity	Plant Explanatory Variables	Estimate	Std. Error	Z	p	Sig
Species Richness	Total Coverage	−0.358	0.130	−2.745	0.006	**
	Simpson Diversity	18.667	9.178	2.034	0.042	*
	Community Composition (PCoA2)	−3.050	1.280	−2.382	0.017	*
	Floral Resources	0.010	0.005	2.032	0.042	*
Activity Density	Species Richness	−1.212	0.373	−3.250	0.001	**
	Simpson Diversity	21.180	9.405	2.252	0.024	*
	Community Composition (PCoA1)	−1.323	0.668	−1.980	0.048	*
Shannon Diversity	Total Coverage	−0.298	0.131	−2.266	0.058
	Simpson Diversity	15.967	10.247	1.558	0.163
	Community Composition (PCoA2)	−2.891	1.319	−2.192	0.065
	Floral Resources	0.009	0.006	1.545	0.166
Simpson Diversity	Total Coverage	−0.117	0.069	−1.687	0.136
	Simpson Diversity	6.848	5.419	1.264	0.247
	Community Composition (PCoA2)	−1.332	0.698	−1.909	0.098
	Floral Resources	0.004	0.003	1.195	0.271
Mean Body Length	Simpson Diversity	−6.004	4.074	−1.474	0.179
	Community Composition (PCoA2)	−2.343	1.237	−1.895	0.095
	Floral Resources	0.013	0.009	1.564	0.156

** indicates 0.001 ≤ p < 0.01, and * indicates 0.01 ≤ p < 0.05.

). Total coverage exerted significant negative impacts on species richness (p = 0.006), whereas Simpson diversity demonstrated positive associations with richness (p = 0.042). Community composition along the PCoA2 axis (p = 0.017) and floral resource availability (p = 0.042) synergistically enhanced species richness. Activity density showed sensitivity to multiple factors, both species richness (p = 0.001) and Simpson diversity (p = 0.024) negatively correlated with this metric, paralleled by inhibitory effects from PCoA1-driven community structure (p = 0.048). Although total coverage (p = 0.058) and PCoA2 composition (p = 0.065) approached significance in reducing Shannon diversity, these relationships remained marginally non-significant under conventional thresholds.

Transitioning to β-diversity analysis, wildflower strip vegetation parameters explained 26.8% of spider community variation, substantially lower than observed in hymenopteran systems. Notably, PCoA1 vegetation gradients exerted disproportionate influence on parasitic wasp assemblages (independent explanatory power = 9.62%; net contribution = 8.62%), while PCoA2 components showed negligible effects on spider communities. Among secondary factors, variable importance followed this hierarchy: Floral Resources (7.31%) > Species Richness (5.61%) > Total Coverage (4.92%), indicating resource availability outweighs structural metrics in shaping spider β-diversity patterns.

3.4.3. Target Pest

Analysis of vegetation-pest dynamics revealed significant suppressive effects of floral resource availability on target pest density (β = −0.34, p < 0.001), with tripling of floral resources corresponding to a 28–41% reduction in pest counts. While plant species richness showed no statistically significant independent effect (p = 0.071), its interaction with floral resources amplified pest suppression efficacy by 19% (95% CI: 5–32%), suggesting complementary biocontrol mechanisms through resource dilution and natural enemy enhancement, as shown in

Table 6. Effects of Wildflower Strip Plant Community Composition on Target Pest Density based on the GLM model.

Diversity	Plant Explanatory Variables	Estimate	Std. Error	Z	p	Sig
Target Pest Density	Species Richness	−0.325	0.180	−1.804	0.071
	Floral Resources	−0.018	0.005	−3.688	0.000	***

*** indicates p < 0.001.

.

4. Discussion

4.1. Wildflower Strips Change Natural Enemy Communities in a Rice-Wheat Agricultural Landscape

This study aimed to explore the impact of wildflower strips (WFS) on natural enemy communities in agricultural landscapes. The hypothesis was that the establishment of wildflower strips would significantly enhance the diversity of natural enemies, particularly parasitoid wasps. These findings are consistent with previous studies and support the idea that wildflower strips can act as a valuable tool for promoting biological pest control [57,58]. However, the stability of spider communities may have made them less responsive to short-term habitat changes, particularly during the early establishment of wildflower strips [59].

In terms of key results, wildflower strips significantly enhanced parasitoid wasp populations and pest control services, particularly by supporting rare taxa and increasing the density of certain species. In contrast, no significant changes were observed in spider communities, which suggests that the effects of wildflower strips on natural enemy community structure are not uniform across all taxa [60].

Based on these findings, it can be inferred that wildflower strips provide an important tool for ecological farming, particularly in enhancing pest control services. However, the effectiveness of wildflower strips may be influenced by cultivation practices, habitat quality, and landscape scale, which warrants further research to optimize their design and management [61]. Future studies should explore how different management strategies, such as cutting frequency [62] and plant diversity optimization [12], can affect the ecological benefits of wildflower strips.

This study also has some limitations. First, it focused primarily on parasitoid wasps and spider communities, without exploring the effects on other predator or pest species. Different natural enemies may respond differently to habitat changes, and future re-search should expand the scope to include a wider range of ecological groups, particularly considering the potential impact of wildflower strips on other pest control groups [61]. Furthermore, the management of wildflower strips, including cutting regimes and plant species diversity, may have a significant impact on the results. Future research should systematically assess the effects of different management strategies on ecosystem services and validate these findings in larger agricultural landscapes. Thirdly, the study was constrained by a limited number of experimental replicates and a relatively short duration of only one year. Weather patterns during this period could have deviated significantly from long-term norms, potentially skewing plant development, organismal responses, and the overall experimental outcomes. To mitigate these limitations, we have initiated multi-year experiments across multiple sites. These longitudinal investigations will systematically capture seasonal and interannual fluctuations, thereby providing a more robust dataset for understanding ecological dynamics and ensuring greater accuracy in our findings.

Theoretically, this study supports the view that wildflower strips enhance natural enemy communities and provides further evidence for the existing theory that habitat changes and resource provision can effectively enhance pest control services. Particularly, wildflower strips have been shown to increase the population density and diversity of certain natural enemies, offering dual benefits for biodiversity conservation and pest management [57]. This provides strong theoretical support for the sustainable development of ecological agriculture, particularly in reducing chemical inputs and promoting agricultural intensification through ecological means.

4.2. Wildflower Strips Provide Pest Control Services in Agricultural Landscapes

This study systematically explores the direct and indirect effects of wildflower strips (WFS) on pest control services, revealing multidimensional mechanisms through which WFS regulate pest populations by altering predator community structure and functional traits. The results align with findings by some scholars [57,63], demonstrating that WFS enhance pest suppression via optimized resource utilization by natural enemies. For instance, the increased proportion of female wasps correlated with reduced activity density of pillbugs (A. vulgare) (1.93–4.81 individuals reduction per 1% female ratio), resonating with Hoffmann et al. (2018)’s observations on sex-specific foraging efficiency in predatory wasps [64]. Additionally, the functional composition of spider communities (e.g., web-builders vs. active hunters) exerted stronger effects on pest control than taxonomic diversity, supporting Blaauw and Isaacs (2015) hypothesis that functional traits outweigh diversity metrics [17]. However, contrary to Pollier et al. (2019)’s expectations, no significant link was found between overall predator diversity and pest suppression, suggesting that niche differentiation among key functional species (e.g., Piratula piratoides’ web-building strategy and Arctosa depectinata’s ambush behavior) may dominate over community-level diversity in WFS habitats [60].

Indirect effects of WFS may involve allelochemical interactions or microhabitat modifications. Although not directly measured here, Toivonen et al. (2018) reported that secondary metabolites from perennial plants could disrupt pest behavior [65]—a mechanism potentially synergizing with predator effects in WFS through specific floral compounds (e.g., Asteraceae or Apiaceae secretions) [66]. Spatial heterogeneity in pest suppression was evident, with the strongest effects within 10 m of WFS, consistent with Mei et al. (2021)’s “resource-distance decay” model [59]. However, landscape context (e.g., seminatural habitat coverage) may modulate this pattern by altering predator dispersal efficiency [67].

The limitations of this study include an insufficient resolution of long-term dynamics and allelochemical pathways. For example, plant succession in WFS may reshape enemy-pest interaction networks [68], while metabolomic validation of allelochemical dynamics is needed. Furthermore, extending findings to multi-trophic scenarios (e.g., pollinator-pest-enemy trade-offs) would better assess WFS’s holistic agroecosystem services [69]. Theoretically, this study advances the “trait-mediated ecosystem service” framework, challenging biodiversity-centric models and informing precision habitat design [70]. Practically, WFS optimization requires integration with regional landscape heterogeneity and farm management (e.g., pesticide regimes) to balance conservation and production goals [50]. Future research should prioritize the trait-based selection of floral species and quantify landscape-scale synergies between WFS and seminatural habitats to maximize ecological intensification potential.

4.3. Wildflower Strip Drive Natural Enemies-Pest Relationship via Changing Plant Community

This study demonstrates that wildflower strip vegetation communities significantly shape the spatial patterns of natural enemy-pest relationships in agricultural landscapes through structural and functional mechanisms. While confirming the foundational role of plant diversity in supporting natural enemies [71,72], the results further reveal that the interaction between plant composition and resource allocation regulates arthropod communities beyond simple linear “diversity-service” correlations. For instance, the negative response of parasitic wasp communities to vegetation coverage (β = −0.34, p < 0.001) and the positive dependence of spiders on plant functional traits (PCoA2 axis effect p = 0.017) highlight differentiated response patterns of different natural enemy groups to vegetation parameters. This finding partially supports the “habitat partitioning hypothesis” [73] but challenges the universality of traditional homogeneous landscape management assumptions.

The results deviate from expectations in two aspects: First, the inhibitory effect of vegetation coverage on spider diversity contradicts Triquet et al. (2022)’s conclusion about cover crops promoting ground-dwelling predators [74], potentially due to microclimate alterations or reduced prey accessibility under dense vegetation [75]. Second, the synergistic pest suppression effect between plant richness and floral resources (19% enhancement) extends the complementary mechanism theory proposed by Yanget al. (2021) [76], indicating that resource heterogeneity can simultaneously achieve pest control through dilution effects and natural enemy functional enhancement. Unlike Clemente-Orta et al. (2020)’s findings in maize fields [77], this study shows that structural vegetation parameters (e.g., PCoA1 axis) explain natural enemy β-diversity (48.1%) more significantly than landscape composition metrics, suggesting that local habitat design may partially offset the negative impacts of regional landscape simplification.

Mechanistically, the sensitivity of parasitic wasps to vegetation coverage may reflect trade-offs between flight capacity and resource search strategies [78], while spider dependence on plant functional traits could relate to microhabitat construction and prey trapping effects [79]. Key limitations include the single-season observation window hindering the capture of legacy effects from vegetation succession [80] and the lack of quantification of allelochemical regulation on natural enemy behavior [81]. Future research should integrate molecular dietary analysis and long-term monitoring to decipher the chemical ecological mechanisms and temporal dynamics of vegetation-natural enemy interactions.

4.4. Wildflower Strips Could Provide a Path to Balancing Agriculture and Ecology in China

This study focuses on the application of wildflower strips in ecological agriculture, aiming to explore effective approaches for balancing agricultural production and ecological conservation [17]. The findings demonstrate that wildflower strips designed with native plant communities exhibit significant advantages in enhancing biodiversity and pest control, which holds substantial implications for the development of ecological agriculture [5]. In China, with the advancement of ecological agriculture, wildflower strips have gradually been incorporated into land consolidation projects. However, previous pilot projects predominantly utilized commercial seed mixes containing invasive species, resulting in limited success in increasing biodiversity and suppressing pests. In contrast, the use of native plant communities in this study successfully improved the populations of natural enemies and pest control efficiency, which can be attributed to the excellent adaptability of native species to local ecological conditions.

The transformation in farmers’ attitudes towards wildflower strips is a crucial element of this research. At the outset, farmers were apprehensive about the extra labor demands and the possible competition for resources that wildflower strips might entail [82]. However, following their involvement in the trials and firsthand observation of tangible advantages, such as decreased pesticide usage and heightened pollinator activity, their acceptance of wildflower strips notably grew [83]. This change underscores the indispensable role of practical validation in the dissemination of new technologies and lays a robust groundwork for the widespread implementation of wildflower strips.

To achieve the widespread adoption of wildflower strips, concerted efforts from multiple stakeholders are essential. Policy support and agricultural extension services serve as key drivers for its development [12,84]. In the long term, regarding wildflower strips as crucial ecological products and facilitating the conversion of ecological benefits into economic returns represent the core strategies for their long-term sustainability. On the one hand, economic gains can be generated through the development of value-added products. For instance, honey beekeeping can be developed using nectar-rich wildflowers to produce high-quality honey [85]. On the other hand, the landscape value of wildflower strips can be leveraged to promote rural tourism, with farmers increasing their income through guided tours, accommodation services, and handicraft sales. These economic incentives not only compensate farmers for their initial investments but also motivate them to maintain wildflower strips actively, establishing a virtuous cycle of “ecological improvement–increased economic returns–continuous ecological investment” [86]. This cycle not only contributes to enhancing ecological services, including soil conservation, water purification [87], and carbon sequestration, but also promotes human well-being and sustainable development, achieving a balance among agricultural, environmental, and community needs [88]. Despite the achievements of this research on wildflower strips, numerous challenges remain. In terms of policy coordination, there is a need to further integrate resources from various departments and formulate more comprehensive supportive policies. Regarding farmer education, it is essential to strengthen technical training and publicity to enhance the farmers’ understanding of the ecological and economic benefits of wildflower strips. Additionally, future research should focus on optimizing the design of wildflower strips and improving their adaptability in diverse environments.

Many perennial wildflowers typically do not flower in the first growing season after sowing, as observed in our study. To enhance first-year flowering, we considered methods such as vernalization, seed priming, and sowing in autumn or winter [89]. Despite using a standard wildflower mix adapted to local conditions, limited flowering occurred for some species. This highlights the importance of long-term management for wildflower strips, as it takes time for the plants to fully establish and reach their full ecological potential. Therefore, while first-year flowering can be improved with the right strategies, achieving the full functionality of wildflower strips requires sustained care and management over multiple seasons [90].

4.5. Limitations

While this study offers valuable insights into how wildflower strips (WFS) and natural grass strips (CK) affect natural enemies and pest suppression, several limitations must be noted. First, the experiment was conducted at a single site over one year, limiting generalizability. Longer-term, multi-site studies are needed to capture broader ecological dynamics [11]. Second, only one wildflower species mixture was tested, without examining how varying plant compositions influence natural enemy responses. Third, the use of pitfall and pan traps may have introduced sampling bias, underrepresenting groups like web-building spiders; more diverse sampling methods are recommended. Fourth, spillover effects into adjacent crop fields were not assessed, leaving the impact on actual pest control and crop yield uncertain [12]. Lastly, this study focused on only two natural enemy groups and one pest species, without evaluating broader trophic interactions or potential reductions in pesticide use. These limitations highlight the need for more comprehensive, long-term, and landscape-level research.

5. Conclusions

This study advances our understanding of how wildflower strips (WFS) reconfigure agroecosystem functionality in East Asian landscapes, bridging biodiversity conservation and pest regulation. By integrating trait-mediated mechanisms and landscape context, we demonstrate that WFS selectively enhances natural enemy communities, particularly parasitic wasps, through floral resource optimization rather than broad biodiversity increases. Female-biased sex ratios in wasp communities emerged as a critical driver of pest suppression, with vegetation design—specifically the phenological continuity of floral resources and functional plant traits—outperforming natural habitats in supporting rare species and stabilizing predator-prey dynamics. In contrast, spider communities exhibited resilience to short-term habitat changes, relying on functional guild shifts (e.g., web-building specialists) rather than taxonomic restructuring. These findings highlight the importance of precision habitat design, where strategic plant selection (e.g., Apiaceae species emitting female-attracting volatiles) and phenological alignment can amplify pest control services.

The current research is constrained by its single-year experimental design, with observations of plants and associated organisms conducted within the same year. This short timeframe may have prevented insects from establishing stable populations in new micro-habitats, limiting the comprehensiveness of the results. Although the current findings show promising trends, caution is needed when interpreting them. Future research should replicate the experiment over multiple years to understand long-term ecological processes like vegetation succession and pest-enemy feedback, and accurately assess the lasting effects of WFS. Meanwhile, long-term monitoring, exploration of multi-trophic interactions, molecular validation of allelochemical pathways, and landscape-scale experiments are also essential to refine WFS design for global agroecological intensification.