Human activities and climatic factors drive decoupling effects of parasitoids‐invasive agromyzid leafminers

Zhenan Jin; Haoxiang Zhao; Shiwei Yuan; Lan Huang; Weijie Wan; Sujie Du; Hongkun Huang; Fanhao Wan; Jianyang Guo; Wanxue Liu

doi:10.1002/ps.71050

. 2026 Jun 29;82(10):10124–10133. doi: 10.1002/ps.71050

Human activities and climatic factors drive decoupling effects of parasitoids‐invasive agromyzid leafminers

Zhenan Jin ^1,^†, Haoxiang Zhao ^1,^†, Shiwei Yuan ¹, Lan Huang ¹, Weijie Wan ¹, Sujie Du ¹, Hongkun Huang ², Fanhao Wan ¹, Jianyang Guo ^1,^✉, Wanxue Liu ^1,^✉

PMCID: PMC13569352 PMID: 42374621

Abstract

BACKGROUND

Human activities disrupt the parasitoid‐mediated control of invasive alien species by fragmenting habitats and intensifying temperature‐driven pest proliferation. However, the synergistic effects of changes in human‐modified landscapes and climatic gradients on parasitoid–herbivore interactions remain underexplored, limiting integrated pest management. Here, we mapped the habitat suitability and dispersal‐risk corridors of two invasive agromyzid leafminers (Liriomyza sativae and Liriomyza trifolii) and evaluated the effects of human activities and climate on the spatial association patterns of parasitoids‐agromyzid leafminers on Hainan Island.

RESULTS

The core risk patches and high‐risk dispersal corridors of the two species were concentrated in towns and cultivated crop‐growing areas in the northern plains and coastal regions, underscoring the key role of human activities in driving pest establishment and dispersal. As the temperature increased, human activities reduced the spatial association between agromyzid leafminers and parasitoids. In contrast, land use intensity strengthened their spatial association, likely indicating a trade‐off between habitat fragmentation and cropland resource diversity in the spatial association patterns of these two populations.

CONCLUSIONS

Our study reveals that human activities and climatic factors synergistically reduce spatial association patterns between parasitoids and invasive agromyzid leafminers. This pattern highlights a novel mechanism underlying pest–natural enemy mismatches and provides a conceptual basis for landscape‐level strategies to enhance biological control. © 2026 Society of Chemical Industry.

Keywords: dispersal risk corridors, ecological regulation, human‐mediated, land‐use intensity, Liriomyza sativae, Liriomyza trifolii, parasitoids, suitable habitats

Our results propose that in human‐modified landscapes, human activities (such as habitat fragmentation caused by urbanization) and climatic factors (mean annual temperature) act in synergy to disrupt the spatial association between parasitoids and invasive agromyzid leafminers. Importantly, this work provides a conceptual basis for developing landscape level strategies—such as optimizing cropping configurations—to mitigate decoupling and restore the ecological effectiveness of natural enemies.

graphic file with name PS-82-10124-g004.webp

1. INTRODUCTION

Human‐mediated land‐use changes and biological invasions are primary drivers of global agricultural productivity loss and terrestrial biodiversity decline. ¹ , ² Land‐use modifications promote survival and the establishment of invasive alien species (IAS) through the provision of optimal habitat environments and abundant food resources. ³ Agricultural expansion, urbanization, and infrastructure development reduce landscape heterogeneity by transforming natural habitats into monocultured croplands, industrial zones, and residential areas. ⁴ , ⁵ This transformation reduces the competitive advantage of native species and creates ‘ecological vacuums’ that facilitate the establishment of IAS. ⁴ , ⁵ For example, in intensified agricultural landscapes, highly fragmented riparian ecological infrastructure, degraded habitat quality, and intense human pressures drive high invasion rates of Linepithema humile. ⁶ Additionally, established IAS typically exhibit an aggregated spatial distribution to form core patches that act as dispersal hubs. ⁷ These patches may be connected by human corridors, such as roads and irrigation networks, accelerating dispersal across modified landscapes. ⁸ These resource gradients establish novel ecological niches, enabling IAS to overcome biogeographic barriers and establish ‘beachheads’ in new environments. ⁹ This synergy underscores the necessity to consider landscape structure when designing biological control and invasion‐management strategies.

The biological control of IAS using parasitoids represents a fundamental component of sustainable pest management; however, its efficacy is compromised by human disturbances. ¹⁰ The effectiveness of parasitoid‐mediated biological control is constrained by climate warming and phenological mismatch, habitat loss and fragmentation, and agricultural intensification, which collectively reduce resource availability, disrupt host–parasitoid synchrony and suppress parasitoid populations. ¹¹ , ¹² , ¹³ , ¹⁴ However, these stressors have rarely been evaluated in combination; this knowledge gap limits integrated pest management (IPM) in human‐modified landscapes during peak pest occurrence periods. Therefore, balancing the advantages of agricultural intensification and the conservation of natural enemies is crucial for sustainable pest management.

Global trade and high dispersal capacity have facilitated the spread of two closely related agromyzid leafminers (Liriomyza sativae and L. trifolii). ¹⁵ Oviposition punctures by females and larval leaf mining damage the structure of vegetable leaf tissues, thereby reducing photosynthesis and ultimately crop yield. ¹⁶ , ¹⁷ , ¹⁸ The highly polyphagous species L. sativae and L. trifolii were first discovered in the Hainan province in 1993 and 2006, respectively, and rapidly emerged as the most important pests of Vigna unguiculata, Luffa cylindrica, Cucumis sativus, and other local vegetable crops. ¹⁵ Previous studies have primarily focused on isolated ecological factors, including temperature‐dependent parasitoid efficacy under laboratory conditions and regional climate matching between agromyzid leafminers and their dominant parasitoids. ¹⁹ , ²⁰ , ²¹ , ²² However, the regional co‐occurrence patterns in landscapes shaped by varying climatic gradients and human pressures remain underexplored. Additionally, although the interaction between temperature and natural parasitoid enemies influences the populations of two species, the synergistic effects of the frequency of human activity and land‐use intensity on their occurrence remain under explored.

This study presents a novel ecological framework for understanding the invasion patterns of invasive agromyzid leafminers and their interactions with parasitoids, driven by human activities and climatic factors. This study aimed to elucidate the landscape interaction patterns of invasive agromyzid leafminers during their establishment and dispersal processes, as well as the risk management strategy for their outbreak processes. Specifically, by synergistically linking the spatial niches occupied by invasive agromyzid leafminers with the landscape connectivity of regional risk core patches, this study identified the dependence of the target species on different landscapes during establishment and dispersal. Consequently, we investigated whether human activities and climate jointly reduce parasitoid‐mediated top‐down control. We hypothesized that: (i) Two closely related invasive agromyzid leafminers exhibit high niche overlap and similar high‐risk distribution patterns in coastal and lowland areas; (ii) Human activities and climate factors (mean annual temperature) synergistically reduce the spatial association patterns of parasitoid populations on agromyzid leafminers, we defined this as the decoupling effect of parasitoids‐invasive agromyzid leafminers.

2. MATERIALS AND METHODS

2.1. Study area

The study area was Hainan Island (18°10′–20°10′ N, 108°37′–111°03′ E), with a total area of approximately 34 000 km². Hainan Island has a tropical monsoon maritime climate, which is characterized by small annual temperature variations and high average annual temperatures (16–25 °C). To adequately represent the distribution range of L. sativae and L. trifolii on Hainan Island and corresponding climatic conditions, we conducted a full‐island survey of species occurrence in 2020 (See Appendix, Table S1). As the central Wuzhi Mountain Range may restrict the dispersal of both species (Fig. 1), the survey sampling points were predominantly distributed in the plains surrounding Hainan Island, where higher average annual temperatures are suitable for their survival and reproduction. In contrast, a limited number of sampling points were distributed in mountainous and hilly regions owing to climatic conditions (e.g., lower temperatures) and topographical complexity, which limit the distribution of both species.

Study area and occurrence locations of Hainan province in China (2020).

2.2. Species data and predictor variables

We compiled global occurrence records for the two agromyzid leafminers from three sources (Supporting Information, Table S2): (i) field surveys conducted across mainland China in 2020; (ii) Global Biodiversity Information Facility (GBIF: https://www.gbif.org/, DOI: 10.15468/dl.zdwjf5 & 10.15468/dl.y9ane5) and iNaturalist Databases (https://www.inaturalist.org/); (iii) published field research papers from Web of Science (https://www.webofscience.com) and CNKI (https://www.cnki.net/). We reviewed the occurrence records from GBIF and INAT based on the scientific names and distribution environments of the species, excluding records of non‐target species and those located over water. To reduce sampling bias and spatial autocorrelation, we spatially thinned the dataset using ENMtools, retaining one occurrence record for the environmental characteristics associated with each grid cell in spatial distribution modeling. ²³ Finally, we obtained 1006 and 424 global occurrence records of L. sativae and L. trifolii, respectively (Supporting Information, Fig. S1; Supporting data [Link], [Link], [Link], [Link]).

We selected environmental variables that influence insect invasiveness across five categories to predict the suitable habitats for the two species: climatic conditions, terrain features, vegetation cover, land use intensity, and human activities (Supporting Information, Table S3). Global and regional bioclimatic data (5 and 1 km resolution, respectively) were sourced from the Chelsa database (https://chelsa-climate.org/), including seven bioclimatic factors related to arthropod survival and distribution. ²⁴ Regional specific data (1 km resolution), comprised elevation (https://www.earthdata.nasa.gov/), 2020 Normalized Difference Vegetation Index (NDVI; https://www.resdc.cn/), Human Footprint (HFP), ²⁵ and Land Use Intensity (LUI). ²⁶ These variables collectively characterize ecological niches critical for species establishment. ²⁷

2.3. Species distribution model

Conventional Species Distribution Models (SDMs) are frequently constrained by specific geographic boundaries, such as assessments limited to a single country or region, reducing the environmental range considered during model training. ²⁸ , ²⁹ This truncates the ecological niche of the species, reducing model accuracy and leading to biases in spatial predictions. ²⁹ , ³⁰ , ³¹ To address this limitation, we used a novel R package (sabinaNSDM) based on a spatially nested hierarchical approach (version R 4.4.3). ³² This method captures the ecological tolerance of species to large‐scale drivers (such as macroclimate) on a global scale while simultaneously incorporating important drivers of species distribution (such as landscape topographic features) on a regional scale. ³³ , ³⁴

Based on the principal component analysis (PCA) of all environmental covariates, we classified the values of the two principal components into quartiles and multiplied them to generate an overall stratification variable consisting of seven climate variable layers. Subsequently, 10 000 random background points were generated across the globe according to the area occupied by each stratum. ³² , ³⁵ We used an embedded regularization technique with a collinearity filtering algorithm combined with the Pearson correlation coefficient (set to 0.7) to select the most suitable variables. ³² The five modeling algorithms—Generalized Additive Model, Gradient Boosting Machine, Generalized Linear Model, Random Forest, and MaxEnt —were used to construct single‐scale models at both global and regional scales. ³² Model performance was evaluated through 20 repeated split‐sample validations (80% of the data as the training set and 20% as the test set). ³² We calibrated regional‐scale models by integrating global model outputs as an additional variable through covariate stratification and retained individual models with AUC ≥0.8 for ensemble model construction. ³² , ³⁶ The output of the ensemble model was based on the mean of probabilities across the selected models. Additionally, the model results were classified based on the maximum training sensitivity plus specificity (MTSS) threshold as follows: unsuitable areas (0–MTSS threshold), low (MTSS threshold–0.4), moderate (0.4–0.6), and high suitability areas (0.6–1.0). ³⁷

2.4. Niche characteristics comparison

The R‐based ecospat package (version R 4.4.2) was used to quantify, compare, and test the ecological niches of the two agromyzid leafminers on Hainan Island. This package employed Principal Component Analysis to reduce the dimensionality of environmental variables, transforming the original set of 11 correlated variables (Section 2.3) into two independent principal components (orthogonal environmental gradients). ³⁸ Based on the principal component scores, the ecological niche distribution data of both target species were mapped onto a regular grid system within the study area, which was defined by the minimum and maximum values of the principal component scores. Subsequently, the species occurrence probability for each grid cell was smoothed to estimate the density using Kernel Density Estimation. Schoener's D index (ranging from 0 to 1, with higher values indicating greater niche overlap) was used to quantify the global ecological niche overlap between the two species. ³⁹ Finally, we conducted equivalence and similarity tests on their occupied spatial niches to compare the differences in spatial niche characteristics between the two species. ³⁸

2.5. Risk core area identification

To prioritize the identification of core risk areas in heterogeneous landscapes, we combined the binary classification results from the NSDM model, which shows species distribution patterns, with the MSPA in the Guidos Toolbox to classify landscape elements. ⁴⁰ We examined the spatial geometry and connectivity of the landscape images by performing MSPA using the Guidos Toolbox. ⁴¹ This enabled us to measure, identify, and segment landscape patterns in pixel‐level data and classify them into different categories (Core, Islet, Perforation, Edge, Bridge, Loop, and Branch). ⁴¹ In the present study, a suitable habitat range was considered the foreground, whereas the other areas were classified as the background. Ultimately, the ‘core’ areas from the MSPA results were extracted as the risk core areas for the two species.

2.6. Landscape connectivity analysis

Considering the potential for human‐assisted dispersal of Diptera into diverse landscape environments, ⁴² we incorporated regional landscape characteristics (elevation and NDVI), human disturbance variables (HFP and LUI) (see Section 2.3), and dispersal requirements (road distance). To avoid weight imbalance resulting from dimensional differences across various landscape characteristics, we normalized each resistance factor layer and rescaled the values to a numerical range of 0–1000 (Supporting Information, Table S4). The resistance weights for the five variables were determined using the entropy weight method, and a comprehensive resistance surface was constructed in ArcGIS (Supporting Information, Table S4). ¹⁶

Using the constructed resistance surface, the linkage pathway tool from the Linkage Mapper Toolkit in the ArcGIS platform was used to identify potential dispersal risk corridors (DRCs) between the core risk zones. ⁴³ This tool identifies adjacent core areas, constructs a core area network using proximity and distance data (including Euclidean and cost‐weighted distances), calculates the least‐cost path (LCP), and synthesizes least‐cost corridors into a comprehensive map. ⁴⁴ , ⁴⁵ To further optimize the connectivity of the DRCs, we used a Pinchpoint Mapper tool to detect connection hotspots. This tool is based on circuit theory through the simulation of the flow of electrons and current conduction to model the potential migration paths for species. ⁴³ During the analysis, we performed calculations based on the user‐defined cost‐weighted distance (CWD), using 5 km as the cutoff distance value and selecting the all‐to‐one mode. ⁴⁶ The analysis enabled the current to flow from all source nodes (i.e., core habitat patches) to each ground node, iteratively generating cumulative current density maps. Areas exhibiting high current densities were identified as critical risk pinch points, which are essential for maintaining connectivity across the entire dispersal risk corridor. ⁴³ , ⁴⁷ In addition, centrality scores for risk core areas and different corridors were generated using the Centrality Mapper tool to highlight the importance of the main risk core areas and DRCs in maintaining connectivity across the network and were categorized into seven categories using the natural breakpoint method. ⁴⁰

2.7. Driving effect analysis

We calculated the population density of two agromyzid leafminers and their parasitoids in each sampling plot, defined as the emergence abundance of each species divided by the total leaf area. The tropical climate of Hainan Island provides favorable optimal geographical conditions for their sympatric coexistence and they exhibit similar biological traits and host preferences. Consequently, the two species commonly co‐occurred within the same sample, making it difficult to identify the exact host origin of emerging parasitoids. By aggregating the abundances of agromyzid leafminers and parasitoids across all samples, we aimed to elucidated the regulatory effects of different environmental variables on the ‘agromyzid leafminers–parasitoids’ system from a population perspective.

We applied structural equation modeling (SEM) to examine the relationships among the ‘agromyzid leafminers–parasitoids’ system, and key abiotic and biotic factors. We hypothesized that parasitoid abundance, together with human activity, land use intensity, and mean annual temperature, directly and indirectly alter the field abundance of agromyzid leafminers populations (Supporting Information, Fig. S2). All predictor variables exhibited low collinearity, with variance inflation factors (VIF) < 5 for all parameters. Three environmental variables were standardized using Z‐scores to facilitate comparisons of effect sizes among predictors, while biotic factors, including the abundances of parasitoids and agromyzid leafminers populations, were log‐transformed to satisfy the assumption of residual normality. ⁴⁸ The initial model was then refined by sequentially removing non‐significant pathways until only statistically significant paths were retained. The final model fit was evaluated using Fisher's C statistic and P‐value in the ‘piecewiseSEM’ package (version R 4.4.2). ⁴⁹ For the statistically significant interaction paths detected in SEM, we employed the ‘interaction’ package (version R 4.4.2) to perform linear regression analyses.

3. RESULTS

3.1. Suitable habitats and spatial ecological niche comparison

The ensemble model showed reliable predictive accuracy (AUC > 0.9, TSS > 0.75) compared with that of individual algorithm approaches (Supporting Information, Table S5). Covariate model analysis revealed human activities as the dominant driver of establishment risk at local scales, while global climatic factors showed reduced influence (Supporting Information, Table S6). On Hainan Island, L. sativae inhabited larger suitable areas (6200 km², 22% of the island) than those of L. trifolii (5000 km², 18%), with 6411 km² of overlapping habitat containing 1819 km² (28% of the overlapping habitat) of the high‐suitability areas. Both species predominantly colonized coastal lowlands with intensive human activities, exhibiting higher suitability in high HFP regions (Fig. 2(A)–(C); Supporting Information, Fig. S3).

Suitable habitat distribution and ecological niche comparison of the two agromyzid leafminers in Hainan: Suitable habitat distributions for *Liriomyza sativae* (A) and *L. trifolii* (B) categorized by suitability levels (lowly/moderately/highly); Coupled suitability pattern of the both species (C); Ecological niche overlap scenario of the both species: Niche overlap (red) *versus* exclusive niches (green: *L. sativae*; yellow: *L. trifolii*) (D); Niche similarity (P < 0.05) (E) and equivalency (P > 0.05) tests (F) based on *Schoener's D* metric.

Despite significant niche overlap between the two species (Schoener's D = 0.80), L. sativae occupied broader ecological space than that of L. trifolii (Fig. 2(D)). Niche similarity tests demonstrated that the niches of both species were similar and equivalent (P < 0.05 for similarity, P > 0.05 for equivalence) (Fig. 2(E), (F)). Kernel density analysis highlighted shared preferences for higher temperatures, low elevations, and areas with elevated human footprint and land‐use intensity for both species (Supporting Information, Fig. S4).

3.2. Core risk areas and connectivity hotspots

The dispersal risk corridor establishment results for two species are shown in Fig. 3. The L. sativae dispersal network comprised 34 corridors (total length of 1730.06 km), while that of L. trifolii had exhibited 37 corridors (1783.28 km), with 1256.56 km of overlapping corridors predominantly distributed in low altitude northern and coastal plains with high levels of human activity (Fig. 3(A)–(C)). Risk core patch analysis identified 21 patches (538.18 km²) for L. sativae and 23 patches (562.70 km²) for L. trifolii, sharing a 527.17 km² overlap predominantly concentrated in coastal zones, with minor inland patches in Danzhou and Ding'an cities (Fig. 3(A)–(C)). Notably, two high centrality patches per species were localized in urban Danzhou and Sanya (Supporting Information, Fig. S5a–d), functioning as critical connectivity hubs, the of which disruption could fragment network stability. Spatial analysis demonstrated divergent regional connectivity patterns: southern coastal risk patches showed dispersed, isolated configurations owing to limited corridor linkages, whereas northern plains exhibited aggregated, interconnected clusters supported by increased patch associations (Supporting Information, Fig. S5A,B).

Dispersal risk corridors (DRCs) for (A) *L. sativae* and (B) *L. trifolii* and (C) spatial coupling relationship between the two DRCs.

3.3. Risk dispersal corridors connectivity hotspot identification

We delineated DRCs for the two species on Hainan Island based on the flow magnitude of each corridor. The total areas of the DRCs for L. sativae and L. trifolii were 7672.79 and 7641.47 km², respectively (Fig. 4(A), (B)). High flow densities were observed in coastal regions. Similarly, corridor centrality analysis revealed that corridors with higher centrality scores were mainly concentrated in the southern region, including Sanya, Ledong, Dongfang, Lingshui, Wanning, and Qionghai. Monitoring and disrupting these DRCs may reduce the connectivity between the core risk areas of the two agromyzid leafminers across Hainan Island. The distribution of these corridors was closely associated with transportation networks (provincial and national highways) and agricultural cultivation areas. These areas form continuous ‘islands’ or elongated ‘pathways’ within human‐mediated heterogeneous landscapes (Fig. 4(C), (D), (I), (II), (III)).

Importance assessment of dispersal risk corridors (DRCs) for optimization: current density of the DRCs for (A) *L. sativae* and (B) *L. trifolii*; flow centrality scores of different DRCs for (C) *L. sativae* and (D) *L. trifolii*; satellite imagery of regions with the DRCs exhibiting higher centrality scores.

3.4. Driving effect

SEM indicated that the population abundance of the two agromyzid leafminers was directly positive, driven by abiotic factors (mean annual temperature) and biotic factors (parasitoid abundance) (Fig. 5(A)). Additionally, compared with human footprint and land use intensity, the combined effect of mean annual temperature and parasitoids exhibited a more important role in determining the population outbreak process of the two agromyzid leafminers (Fig. 5(B)). The interaction between mean annual temperature and parasitoid abundance revealed that the strength of the parasitoid–agromyzid leafminers association declined with increasing temperature (Fig. 5(A), (B), (C)). Moreover, the land‐use intensity showed a positive effect on parasitoids with increasing temperature, indirectly promoting the strength of the parasitoid–agromyzid leafminers association (Fig. 5(A), (B), (D)). Conversely, under elevated temperature conditions, the human footprint was negatively associated with parasitoid abundance under warmer conditions, indirectly reducing the strength of the parasitoid–agromyzid leafminers association (Fig. 5(A), (B), (E)).

Piecewise structural equation model (piecewiseSEM) showing the direct and indirect effects of abiotic factors (human footprint (HFP), land use intensity (LUI), and mean annual temperature (MAT)) and biotic factors (parasitoids abundance) on the abundance of agromyzid leafminers. Red solid lines and green arrows represent positive and negative relationships (P < 0.05), respectively. Partial plots were used to visualize the effects of selected significant pathway variables (A); Direct and indirect effects of multiple predictor variables on the abundance of agromyzid leafminers (B); Interaction effect of MAT on the effect of parasitoids abundance on agromyzid leafminers abundance; (D‐E) Interaction effect of MAT on the effect of HFP and LUI on parasitoids abundance. [Correction added on 08 July 2026, after first online publication: Figure 5 has been updated.]

4. DISCUSSION

Our study demonstrates that human disturbance and temperature are associated with invasion risk and altered parasitoid–agromyzid leafminers coupling. Human activities (such as habitat fragmentation caused by urbanization) and higher temperature act in synergy to reduce the spatio‐temporal coupling between parasitoids and invasive agromyzid leafminers. This synergistic decoupling reduces the natural regulatory capacity of parasitoids and increases the likelihood of pest outbreaks. Our found highlights that pest–natural enemy mismatches may result from the interaction between climate and human activity rather than from either factor alone and also provides a basis for landscape‐level strategies, such as optimizing crop configurations, to reduce decoupling and restore the ecological effectiveness of natural enemies.

4.1. Multi‐scale influence of environmental predictors on habitat suitability

The climatic covariates, such as temperature predict variation along continental‐scale environmental gradients and are quantifiable across spatial scales. ⁵⁰ However, the distribution of IAS is increasingly influenced by human activities, which frequently vary unpredictably across different scales at the regional level. ⁵¹ Therefore, incorporating fine‐scale human activity data can improve model performance, particularly in intensively managed and human‐disturbed habitats. ⁵⁰ , ⁵² Our objective was to investigate the influence of regional factors, such as human activity and land‐use intensity, on establishment risk on Hainan Island within climatically suitable areas rather than building a climate‐only SDM. In the present study, global temperatures significantly affected the fundamental ecological niches of the two agromyzid leafminers. However, at the regional scale, the effect of fundamental ecological niches on the establishment risk of the two species was reduced, with human activity playing a predominant role. This highlights the importance of regional factors in estimating habitat suitability, ⁵³ , ⁵⁴ and underscores the necessity of integrating global and regional factors when predicting invasive species. ⁵⁵ Furthermore, the co‐occurrence of the two species in areas with higher temperatures, lower altitudes, and dense human activity was prominent, suggesting that urbanization and agricultural expansion may facilitate their establishment and dispersal by altering their potential habitat range.

4.2. Human activity mediated dispersal patterns and connectivity hotspots

Human‐mediated dispersal reconfigures IAS spatial networks and influences their spread across landscapes. ⁵⁶ , ⁵⁷ This pattern was evident in their habitat affinity of both agromyzid leafminers, which were more common in crop‐dominated and settlement‐associated landscapes, such as urban and rural settlements. This affinity for human‐mediated landscape is expected to lead to human‐assisted dispersal patterns associated with specific habitats. ⁵⁸ Urbanization and agricultural expansion have intensified habitat fragmentation on Hainan Island, particularly in low‐altitude northern and southern regions such as Haikou and Sanya. ⁵⁹ The corridor analysis revealed that the DRCs of the two agromyzid leafminers were predominantly concentrated in cultivated crop areas and along transportation routes. In the southern coastal hilly regions, corridors exhibit elongated distribution patterns owing to topographic constraints, whereas the northern plains, characterized by dense agricultural landscapes, form extensive and interconnected DRCs. This indicates that human‐driven landscape fragmentation and low‐altitude topographic factors are common mediators of dispersal. Network hubs and central patches can strongly influence dispersal because numerous movement pathways pass through these structures. ⁶⁰ Evaluating management based on the network structure (centrality metrics) can reduce the number of unchecked but potentially infectious connections. Ashander et al. identified optimal IAS management strategies by investigating the centrality metrics of IAS flow across 58 counties in Minnesota, USA, at the landscape scale. ⁶¹ Among the risk hubs identified in our study, both species exhibited high centrality scores in Danzhou and Sanya. Compared to that of the northern plains, the southern coastal areas showed higher corridor flow density, probably owing to the combined topographic constraints and intensive human activity.

4.3. Combined regulation of the outbreak of agromyzid leafminers by biotic and abiotic factors

SEM indicated that temperature and natural enemy (parasitoid) abundance were the key factors regulating the population outbreak of the agromyzid leafminers. Land use intensity and temperature interact, indicating that under increasing temperatures, the inhibitory effect of high‐intensity agricultural activities on parasitoid abundance is reduced. This may be attributed to the potential of interspersing semi‐natural habitats in areas of high land use to facilitate biological control as temperatures increase (such as intercropping or succession of multiple host plants in peri‐urban or internal self‐sustaining areas), providing agromyzid leafminers with more food resources or suitable microhabitats, thus enhancing the positive regulatory effect of parasitoid populations on their populations. ⁶² According to the resource specialization hypothesis, increasing field plant diversity attracts species specialized in specific resources, thereby enhancing diversity at higher trophic levels. ⁶² Diverse‐field plants provide more refuge, microhabitats, and alternative food resources (hosts/prey, pollen, and nectar) for generalist herbivores, promoting the growth of natural enemy populations and strengthening the biological control of herbivorous pests. ⁶³ , ⁶⁴ Therefore, in intercropping and relay cropping systems, the abundance of generalist parasitoids in agricultural ecosystems is enhanced owing to the increased resources that provide stronger natural defenses against pest control. ⁶⁵

In contrast, an increase in the human footprint under increasing temperatures notably reduces parasitoid abundance, generating synergistic stressors associated with ‘temperature‐human disturbance’ that exacerbates the risk of leafminer outbreaks. This may be because, along with increasing temperatures, intensified habitat destruction or pollution resulting from human activities, such as urbanization or road expansion, leads to fragmented landscapes that counteract their original positive effects. Additionally, high‐frequency human activities in settlements can severely reduce natural enemy populations and disrupt the ecosystem balance, thereby reducing pest control effectiveness in adjacent areas. ⁶⁶ Fragmented landscapes resulting from urbanization or road expansion may restrict the movement of natural enemies, thus reducing their ability to effectively control pest populations. ⁶⁶ In addition, increased landscape heterogeneity reduces host plant diversity across patches, further limiting the host selectivity of parasitoids. ⁶⁷ , ⁶⁸ Influenced by chemical control in agricultural management, insecticides sprayed manually may drift from nearby fields and negatively affect parasitoids in different patches, particularly in regions dominated by agriculture where pesticide use is more prevalent. ⁶⁹

The contrasting effects of HFP and LUI indicate that multiple aspects of human modification should not be treated interchangeably. The human footprint may capture habitat fragmentation and road‐associated pressure, whereas land‐use intensity may indicate crop‐resource availability in human‐modified landscapes.

4.4. Integrated management strategies

From 1980 to 2018, the areas of cultivated land, grassland, and forest on Hainan Island decreased, while the area of construction land increased by 1.2 times. ⁷⁰ Substantial progress for integrated post‐invasion pest management can be achieved by restricting urban expansion encroaching on natural habitats, reducing pesticide overuse and unnecessary land development, as well as utilizing ecological compensation mechanisms to promote the sustainable management of agricultural landscapes. ¹³ , ⁶⁶ Diversification of field vegetation is an effective strategy for enhancing biodiversity and ecological functions. ⁷¹ Promoting intercropping and covering crops to increase crop diversity can maintain natural enemy populations. In high temperature regions and seasons, priority should be directed toward preserving edge habitats within crop planting areas to provide refuge and migration corridors for parasitoid wasps. ⁷² Policy based tools, such as incorporating ‘crop diversification’ into the ‘Hainan Province Implementation Plan for Chemical Pesticide and Fertilizer Reduction’, can facilitate positive changes. In addition, managing invasive alien species based on network centrality metrics enables long term management outcomes that approach or achieve optimal effectiveness. ⁶¹ In this study, we identified the core risk areas and key DRCs for the two agromyzid leafminers using network centrality metrics. We observed that the core risk areas were located in Danzhou and Sanya, whereas the key risk corridors were concentrated along highways in southern Hainan. Real time monitoring of the two agromyzid leafminers in these regions, combined with the establishment of physical barriers, such as vegetation buffers, along transportation routes and field boundaries, can effectively disrupt the connectivity between core risk areas and reduce the stability of their dispersal networks.

5. CONCLUSION

On Hainan Island, the high risk establishment areas and dispersal risk corridors were concentrated in the northern plains and coastal regions, aligning with the distribution of transportation networks and cultivated crop areas, highlighting the role of human activity in facilitating the establishment and dispersal of IAS. The increasing temperatures reduced the abundance of parasitoids, forming a pest proliferation feedback loop. During biocontrol efforts, solely increasing the parasitoid abundance did not enhance the regulation of the agromyzid leafminers. Fragmented landscapes and human activities disrupt the ecological regulation of parasitoids. In practice, we recommend protecting edge habitats and maintaining diversified landscapes and host plant planting patterns, with a focus on monitoring and controlling high centrality hubs, such as Danzhou and Sanya. Incorporating these strategies into agricultural ecological policies can buffer the direct impact of climate change on outbreaks of agromyzid leafminers and sustain biological control. Our study demonstrates that human activities and climatic factors (mean annual temperature) synergistically reduce the spatial association between parasitoids and invasive agromyzid leafminers. This work provides a new conceptual basis for strengthening biological control through landscape level strategies.

CONFLICT OF INTEREST

The authors declare they have no conflict of interest.

AUTHOR CONTRIBUTIONS

Zhenan Jin, Haoxiang Zhao and Wanxue Liu conceived the idea and designed methodology; Shiwei Yuan and Lan Huang performed the insect sampling; Weijie Wan conducted the identifications; Zhenan Jin analysed the data; Zhenan Jin and Haoxiang Zhao led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

Supporting information

Table S1. The species of parasitic wasps and their total abundance emerging from the two agromyzid leafminers during the 2020 survey of Hainan.

Table S2. Overview of data sources for recorded occurrence points of the two agromyzid leafminers.

Table S3. The 11 predictor variables related to the survival and distribution of invasive alien insects.

Table S4. The weight and direction of each resistance factor is determined, and the standardized resistance factors are reassigned and classified into four equal intervals.

Table S5. The accuracy of individual models (Generalized additive model, Boosted regression tree, Generalized linear model, Maximum entropy model and Random forest) and ensemble models based on true skill statistics (TSS) and area under the receiver operating characteristic curve (AUC).

Table S6. The importance values of different predictor variables for the predictive results of the NSDM for Liriomyza sativae and L. trifolii are as follows:

Fig. S1. Global distribution maps of two agromyzid leafminers: (A) Liriomyza sativae; (B) L. trifolii.

Fig. S2. The conceptual model illustrates the regulatory pathways by which climate, human activities, and land‐use intensity shape the association patterns between parasitoids and leafminers.

Fig. S3. Non‐linear responses for suitability to elevation (A) and human footprint (HFP) (B) derived from generalized additive models.

Fig. S4. The kernel density plots of the 12 variables (Bio1, Bio2, Bio5, Bio6, Bio12, Bio13, Bio14, HFP, LUI, NDVI & Altitude) comparing the separate environmental occupancy space between Liriomyza sativae and L. trifolii (green: L. sativae; yellow: L. trifolii; brown: both).

Fig. S5. Dispersal risk networks monitoring and recommendation patterns for Liriomyza sativae and L. trifolii on Hainan Island. (A) and (B): Abstract representation of dispersal risk corridors and risk core areas as nodes and links based on complex network theory; (a), (b), (c), and (d) depict landscape details from satellite imagery, highlighting localized features.

PS-82-10124-s001.docx^{(4.5MB, docx)}

Data S1. Supporting Information.

PS-82-10124-s002.csv^{(24.5KB, csv)}

Data S2. Supporting Information.

PS-82-10124-s003.csv^{(10.4KB, csv)}

Data S3. Supporting Information.

PS-82-10124-s005.csv^{(2.2KB, csv)}

Data S4. Supporting Information.

PS-82-10124-s006.csv^{(1.9KB, csv)}

Data S5. Supporting Information.

PS-82-10124-s004.csv^{(2.2KB, csv)}

ACKNOWLEDGEMENTS

This research was funded by the National Key R&D Program of China (2023YFC2605200) and the Agricultural Science and Technology Innovation Program (ASTIP) (CAAS‐ZDRW202505).

Contributor Information

Jianyang Guo, Email: guojianyang@caas.cn.

Wanxue Liu, Email: liuwanxue@caas.cn.

DATA AVAILABILITY STATEMENT

The data that supports the findings of this study are available in the supplementary material of this article.

REFERENCES

1. Millard J, Outhwaite CL, Kinnersley R, Freeman R, Gregory RD, Adedoja O et al., Global effects of land‐use intensity on local pollinator biodiversity. Nat Commun 12:2902 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
2. Marino C, Leroy B, Latombe G and Bellard C, Exposure and sensitivity of terrestrial vertebrates to biological invasions worldwide. Glob Change Biol 30:e17607 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Tassin J and Kull CA, Facing the broader dimensions of biological invasions. Land Use Policy 42:165–169 (2015). [Google Scholar]
4. Borden JB and Flory SL, Urban evolution of invasive specie. Front Ecol Environ 19:1540–9295 (2021). [Google Scholar]
5. González‐Moreno P, Diez JM, Ibáñez I, Font X and Vilà M, Plant invasions are context‐dependent: multiscale effects of climate, human activity and habitat. Divers Distrib 20:720–731 (2014). [Google Scholar]
6. Zina V, Duarte G, Fonseca A, Conde S, Ferreira MT, Franco JC et al., Land use system, invasive species and shrub diversity of the riparian ecological infrastructure determine the specific and functional richness of ant communities in Mediterranean river valleys. Ecol Indic 145:109613 (2022). [Google Scholar]
7. Germain RM, Jones NT and Grainger TN, Cryptic dispersal networks shape biodiversity in an invaded landscape. Ecology 100:e02738 (2019). [DOI] [PubMed] [Google Scholar]
8. Mills DH and McKinney ML, Climate change and jump dispersal drive invasion of the rosy Wolfsnail (Euglandina rosea) in the United States. Sustainability 16:1929 (2024). [Google Scholar]
9. Zhao ZH, Carey JR and Li ZH, The global epidemic of Bactrocera pests: mixed‐species invasions and risk assessment. Annu Rev Entomol 69:219–237 (2024). [DOI] [PubMed] [Google Scholar]
10. Khamis FM and Ajene IJ, Human influences on parasitoid wasps' biocontrol of invasive insect pest species in Africa. Curr Opin Insect Sci 68:101300 (2025). [DOI] [PubMed] [Google Scholar]
11. Harvey JA, Heinen R, Armbrecht I, Basset Y, Baxter‐Gilbert JH, Bezemer TM et al., International scientists formulate a roadmap for insect conservation and recovery. Nat Ecol Evol 4:174–176 (2020). [DOI] [PubMed] [Google Scholar]
12. Richardson JL, McCoy EP, Parlavecchio N, Szykowny R and Beech‐Brown E e a, Increasing rat numbers in cities are linked to climate warming, urbanization, and human population. Sci Adv 11:eads6782 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Haan NL, Zhang Y and Landis DA, Predicting landscape configuration effects on agricultural pest suppression. Trends Ecol Evol 35:175–186 (2020). [DOI] [PubMed] [Google Scholar]
14. Soti V, Thiaw I, Debaly ZM, Sow A, Diaw M, Fofana S et al., Effect of landscape diversity and crop management on the control of the millet head miner, Heliocheilus albipunctella (Lepidoptera: Noctuidae) by natural enemies. Biol Control 129:115–122 (2019). [Google Scholar]
15. Gao YL, Reitz SR, Wei Q, Yu W, Zhang Z and Lei ZR, Local crop planting systems enhance insecticide‐mediated displacement of two invasive Leafminer Fly. PLoS One 9:e92625 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
16. Wang XF, Zhu MN, Zhang YR, Ma JH, Zhou JT, Chen T et al., Ecological security pattern construction of the three‐river source region basedon source‐resistance‐corridor. Acta Ecol Sin 44:4609–4623 (2024). [Google Scholar]
17. Parrella MP, Jones VP, Youngman RR and Lebeck LM, Effect of leaf mining and leaf stippling of Liriomyza spp. on photosynthetic rates of chrysanthemum. Ann Entomol Soc Am 78:90–93 (1985). [Google Scholar]
18. Johnson MW, Welter SC, Toscano NC, Ting P and Trumble JT, Reduction of tomato leaflet photosynthesis rates by mining activity of Liriomyza sativae (Diptera: Agromyzidae). J Econ Entomol 76:1061–1063 (1983). [Google Scholar]
19. Ye FY, Yang YM, Zhang YB, Pan LT, Yefremova Z, Yang LY et al., The thelytokous strain of the parasitoid Neochrysocharis formosa outperforms the arrhenotokous strain in reproductive capacity and biological control of agromyzid leafminers. Pest Manag Sci 79:729–740 (2023). [DOI] [PubMed] [Google Scholar]
20. Xuan JL, Xiao Y, Ye FY, Zhang YB, Tao SX, Guo JY et al., High temperatures do not decrease biocontrol potential for the host‐killing parasitoid Neochrysocharis formosa (Hymenoptera: Eulophidae) on agromyzid leafminers. J Integr Agric 21:1722–1730 (2022). [Google Scholar]
21. Wang H, Reitz SR, Xiang J, Smagghe G and Lei Z, Does temperature‐mediated reproductive success drive the direction of species displacement in two invasive species of Leafminer Fly? PLoS One 9:e98761 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
22. Maino JL, Pirtle EI, Baudrot V, Ridland PM and Umina PA, Forecasting the potential distribution of invasive leafminer pests, Liriomyza spp. (Diptera: Agromyzidae) and their natural enemies. Austral Entomol 62:118–130 (2023). [Google Scholar]
23. Warren DL, Glor RE and Turelli M, ENMTools: a toolbox for comparative studies of environmental niche models. Ecography 33:607–611 (2010). [Google Scholar]
24. De Meyer M, Robertson MP, Mansell MW, Ekesi S, Tsuruta K, Mwaiko W et al., Ecological niche and potential geographic distribution of the invasive fruit fly Bactrocera invadens (Diptera, Tephritidae). B Entomol Res 100:35–48 (2010). [DOI] [PubMed] [Google Scholar]
25. Mazzariello J and Gassert F, Global 100m Terrestrial Human Footprint (HFP‐100) [Dataset]. Dryad (2023).
26. Zhang XT, Zhang DX, Wang XL, Peng JF, Li YA and Li SJ, A dataset of 1km‐grid comprehensive indexes of land use degree density in China. China Sci Data 9:272–279 (2020). [Google Scholar]
27. Tantowijoyo W and Hoffmann AA, Variation in morphological characters of two invasive leafminers, Liriomyza huidobrensis and L. sativae, across a tropical elevation gradient. J Insect Sci 11:69 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
28. Scherrer D, Esperon‐Rodriguez M, Beaumont LJ, Barradas VL and Guisan A, National assessments of species vulnerability to climate change strongly depend on selected data sources. Divers Distrib 27:1367–1382 (2021). [Google Scholar]
29. Guisan A, Chevalier M, Adde A, Zarzo‐Arias A, Goicolea T, Broennimann O et al., Spatially nested species distribution models (N‐SDM): an effective tool to overcome niche truncation for more robust inference and projections. J Ecol 113:1588–1605 (2025). [Google Scholar]
30. Chevalier M, Zarzo‐Arias A, Guélat J, Mateo RG and Guisan A, Accounting for niche truncation to improve spatial and temporal predictions of species distributions. Front Ecol Evol 10:944116 (2022). [Google Scholar]
31. Goicolea T, Adde A, Broennimann O, García‐Viñas JI, Gastón A, Aroca‐Fernández MJ et al., Hierarchical species distribution models to overcome niche truncation in National‐Scale Studies. Ecography 2025:e07328 (2025). [Google Scholar]
32. Mateo RG, Morales‐Barbero J, Zarzo‐Arias A, Lima H, Gómez‐Rubio V and Goicolea T, sabinaNSDM: an R package for spatially nested hierarchical species distribution modelling. Methods Ecol Evol 15:1796–1803 (2024). [Google Scholar]
33. Mateo RG, Mokany K and Guisan A, Biodiversity models: what if unsaturation is the rule? Trends Ecol Evol 32:556–566 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
34. Mateo RG, Aroca‐Fernández MJ, Gastón A, Gómez‐Rubio V, Saura S and García‐Viñas JI, Looking for an optimal hierarchical approach for ecologically meaningful niche modelling. Ecol Model 409:108735 (2019). [Google Scholar]
35. Barbet‐Massin M, Jiguet F, Albert CH and Thuiller W, Selecting pseudo‐absences for species distribution models: how, where and how many? Methods Ecol Evol 3:327–338 (2012). [Google Scholar]
36. Adde A, Rey PL, Brun P, Külling N, Fopp F, Altermatt F et al., N‐SDM: a high‐performance computing pipeline for nested species distribution modelling. Ecography 2023:e06540 (2023). [Google Scholar]
37. Thuiller W, Guéguen M, Renaud J, Karger DN and Zimmermann NE, Uncertainty in ensembles of global biodiversity scenarios. Nat Commun 10:1446 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
38. Di Cola V, Broennimann O, Petitpierre B, Breiner FT, D'Amen M, Randin C et al., ecospat: an R package to support spatial analyses and modeling of species niches and distributions. Ecography 40:774–787 (2017). [Google Scholar]
39. Broennimann O, Fitzpatrick MC, Pearman PB, Petitpierre B, Pellissier L, Yoccoz NG et al., Measuring ecological niche overlap from occurrence and spatial environmental data. Glob Ecol Biogeogr 21:481–497 (2012). [Google Scholar]
40. Zhou S, Yin HW, Kong FH, Wu W, Sun H, Su J et al., Enhancing ecological network establishment with explicit species information and spatially coordinated optimization for supporting urban landscape planning and management. Landscape Urban Plan 248:105079 (2024). [Google Scholar]
41. Vogt P, Riitters KH, Estreguil C, Kozak J, Wade TG and Wickham JD, Mapping spatial patterns with morphological image processing. Landsc Ecol 22:171–177 (2007). [Google Scholar]
42. Elsensohn JE, Wolford S, Tabb A and Leskey T, Experimental evidence supports the ability of spotted lanternfly to hitchhike on vehicle exteriors as a mechanism for human dispersal. R Soc Open Sci 11:240493 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
43. McRae BH, Dickson BG, Keitt TH and Shah VB, Using circuit theory to model connectivity in ecology, evolution, and conservation. Ecology 89:2712–2724 (2008). [DOI] [PubMed] [Google Scholar]
44. Bu H, McShea WJ, Wang D, Wang F, Chen YP et al., Not all forests are alike: the role of commercial forest in the conservation of landscape connectivity for the giant panda. Landsc Ecol 36:2549–2564 (2021). [Google Scholar]
45. Dutta T, Sharma S, McRae BH, Roy PS and DeFries R, Connecting the dots: mapping habitat connectivity for tigers in central India. Reg Environ Change 16:53–67 (2016). [Google Scholar]
46. Washington Wildlife Habitat Connectivity Working Group (WHCWG) , Washington Connected Landscapes Project: Statewide Analysis. Washington Departments of Fish and Wildlife, and Transportation, Olympia, WA: (2010). [Google Scholar]
47. McRae BH, Hall SA, Beier P and Theobald DM, Where to restore ecological connectivity? Detecting barriers and quantifying restoration benefits. PLoS One 7:e52604 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
48. Sanczuk P, Verheyen K, Lenoir J, Zellweger F, Lembrechts JJ, Rodríguez‐Sánchez F et al., Unexpected westward range shifts in European forest plants link to nitrogen deposition. Science 386:193–198 (2024). [DOI] [PubMed] [Google Scholar]
49. Lefcheck JS, piecewiseSEM: piecewise structural equation modelling in r for ecology, evolution, and systematics. Methods Ecol Evol 7:573579 (2016). [Google Scholar]
50. Requena‐Mullor JM, Maguire KC, Shinneman DJ and Caughlin TT, Integrating human factors into regional‐scale species distribution models—a novel application in the imperiled sagebrush biome. Glob Chang Biol 25:3844–3858 (2019). [DOI] [PubMed] [Google Scholar]
51. Hellmann C, Werner C and Oldeland J, A spatially explicit dual‐isotope approach to map regions of plant‐plant interaction after exotic plant invasion. PLoS One 11:e0159403 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
52. Anikó C, Zoltán B, György K, Barbara L, Gábor Ó, Tamás R et al., Tree plantations are hot‐spots of plant invasion in a landscape with heterogeneous land‐use. Agric Ecosyst Environ 226:88–98 (2016). [Google Scholar]
53. Pearson RG and Dawson TP, Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Glob Ecol Biogeogr 12:361–371 (2003). [Google Scholar]
54. Milbau A, Stout JC, Graae BJ and Nijs I, A hierarchical framework for integrating Invasibility experiments incorporating different factors and spatial scales. Biol Invasions 11:941–950 (2009). [Google Scholar]
55. Lázaro‐Lobo A, Wessely J, Essl F, Moser D and Jiménez‐Alfaro B, Combining hierarchical distribution models with dispersal simulations to predict the spread of invasive plant species. Glob Ecol Biogeogr 34:e70026 (2025). [Google Scholar]
56. Bullock JM, Bonte D, Pufal G, da Silva Carvalho C, Chapman DS, García C et al., Human‐mediated dispersal and the rewiring of spatial networks. Trends Ecol Evol 33:958–970 (2018). [DOI] [PubMed] [Google Scholar]
57. Banks NC, Paini DR, Bayliss KL and Hodda M, The role of global trade and transport network topology in the human‐mediated dispersal of alien species. Ecol Lett 18:188–199 (2015). [DOI] [PubMed] [Google Scholar]
58. Mothes CC and Searcy CA, Comparative landscape genomics reveals species‐specific spatial patterns and suggests human‐aided dispersal in a global hotspot for biological invasions. Biol Invasions 26:4059–4074 (2024). [Google Scholar]
59. Lei JR, Chen YQ, Li LM, Chen ZZ, Chen XH, Wu TT et al., Spatiotemporal change of habitat quality in Hainan Island of China based on changes in land use. Ecol Indic 145:10970 (2022). [Google Scholar]
60. Morel–Journel T, Assa CR, Mailleret L and Vercken E, Its all about connections: hubs and invasion in habitat networks. Ecol Lett 22:313–321 (2019). [DOI] [PubMed] [Google Scholar]
61. Ashander J, Kroetz K, Epanchin‐Niell R, Phelps NBD, Haight RG and Dee LE, Guiding large‐scale management of invasive species using network metrics. Nat Sustainability 5:762–769 (2022). [Google Scholar]
62. Zhang J, You SJ, Niu DS, Guaman KGG, Wang A, Saqib H et al., Landscape composition mediates suppression of major pests by natural enemies in conventional cruciferous vegetables. Agric Ecosyst Environ 316:107455 (2021). [Google Scholar]
63. Root RB, Organization of a plant‐arthropod association in simple and diverse habitats: the fauna of collards (Brassica Oleracea). Ecological monographs 43:95–124 (1973). [Google Scholar]
64. Shameer KS, Nasser M, Mohan C and Hardy ICW, Direct and indirect influences of intercrops on the coconut defoliator Opisina arenosella . J Pest Sci 91:259–275 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
65. Jeavons E, van Baaren J, Le Ralec A, Buchard C, Duval F, Llopis S et al., Third and fourth trophic level composition shift in an aphid–parasitoid–hyperparasitoid food web limits aphid control in an intercropping system. J Appl Ecol 59:300–313 (2022). [Google Scholar]
66. Blank L, How urban proximity shapes agricultural pest dynamics: a review. Pest Manag Sci 81:2704–2711 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
67. Pak D, Iverson AL, Ennis KK, Gonthier DJ and Vandermeer JH, Parasitoid wasps benefit from shade tree size and landscape complexity in Mexican coffee agroecosystems. Agric Ecosyst Environ 206:21–32 (2015). [Google Scholar]
68. Gagic V, Tscharntke T, Dormann CF, Gruber B, Wilstermann A and Thies C, Food web structure and biocontrol in a four‐trophic level system across a landscape complexity gradient. Proc R Soc B Biol Sci 278:2946–2953 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
69. Matthew GE, Mitchell EH, Tsuruda M, Gonzalez A and Bennett EM, Contrasting responses of soybean aphids, primary parasitoids, and hyperparasitoids to forest fragments and agricultural landscape structure. Agric Ecosyst Environ 326:107752 (2022). [Google Scholar]
70. Lei JR, Chen ZZ, Chen XH, Li YL and Wu TT, Spatio‐temporal changes of land use and ecosystem services value in Hainan Island from 1980 to 2018. Acta Ecol Sin 40:4760–4773 (2020). [Google Scholar]
71. Barnes AD, Scherber C, Brose U, Borer ET, Ebeling A, Gauzens B et al., Biodiversity enhances the multitrophic control of arthropod herbivory. Sci Adv 6:eabb6603 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
72. Dominik C, Seppelt R, Horgan FG, Settele J and Václavík T, Landscape composition, configuration, and trophic interactions shape arthropod communities in rice agroecosystems. J Appl Ecol 55:2461–2472 (2018). [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials