The study was conducted in the Demarcated Area (DA) of Alicante, Spain, affected by X. fastidiosa subsp. multiplex sequence type ST6 (Velasco Amo et al. 2025) (Fig. 1). The region has a semi-arid Mediterranean climate with a mean annual temperature of ~ 18 °C and low annual precipitation (290–460 mm). Summers are hot and dry and winters are mild (Kelley et al. 2012; Valdes-Abellan et al. 2017). The landscape consists of Cretaceous limestone and marls, with elevations ranging from sea level to 1,400 m and is characterised by drought-adapted vegetation intermixed with cultivated areas, pine forests and coastal zones (Alfaro García et al. 2004; Duguy and Vallejo 2008; Valdes-Abellan et al. 2017). This mosaic configuration provides suitable conditions for the persistence and spread of Xf across managed and natural ecosystems.

Field surveys were conducted during June-July 2023 in 19 almond orchards located at the agriculture-forest interface, selected to capture a gradient of environmental conditions and invasion contexts. Orchards were initially classified as infected or uninfected, based on historical surveillance data provided by the regional Xylella monitoring network (GVA 2024). This classification reflects infection status rather than quantitative disease prevalence. In each orchard, six almond trees were selected and visually assessed for canopy defoliation and proportion of dead branches using a continuous 0–100% scale, following ICP Forests protocols adapted to agricultural systems (Eichhorn et al. 2016). Shrub cover was visually estimated as the percentage of ground covered by shrubs within the canopy projection area of each sampled almond tree.

Plant communities in the adjacent natural vegetation were surveyed along two parallel 50 m transects per orchard, separated by 10 m and starting at the orchard edge. Along transects, all woody and herbaceous species were recorded, irrespective of their host status for Xf, allowing a comprehensive assessment of plant diversity and community composition.

For each individual plant, distance to the orchard edge was recorded, and plant health was evaluated using leaf discoloration and dead branch percentage (0–100%), following ICP Forests guidelines (Eichhorn et al. 2016). Discoloration was used instead of defoliation because many non-crop species exhibit partial leaf chlorosis rather than complete leaf loss under stress, making discoloration a more appropriate and comparable indicator of canopy health across growth forms.

Plant material (mature leaves and branches) from almond trees and selected plant species in the adjacent vegetation was collected following EPPO standard PM7/24 (EPPO 2023), which recommends sampling xylem-rich tissues to maximise pathogen detection. In almond orchards, samples were collected from the six trees assessed for phytosanitary status. In the adjacent vegetation, the sampling targeted plant species previously identified as frequent hosts in the study area (GVA 2024). Potential host plants were searched for within a maximum distance of 50 m from the orchard edge in the area between the two natural vegetation transects. For each potential host species and orchard, three individual plants were recorded prioritising symptomatic individuals when present.

DNA extraction was performed using the cetyltrimethylammonium bromide (CTAB) method (EPPO 2023), including negative and positive controls (Velasco Amo et al. 2025). Detection of Xf was carried out using a duplex qPCR protocol (Bonants et al. 2019), which simultaneously incorporates the Harper et al. (2010) and Ouyang et al. (2013) assays targeting two independent genomic regions of Xf. A sample was considered positive when both protocols yielded positive amplifications. Reactions followed the EPPO protocol with minor modifications to reduce PCR inhibition. Samples with Cq ≤ 35 were considered positive, > 38 negative and values between 35 and 38 inconclusive and re-tested. Further methodological details are provided in Suppl. material 1: appendix S1. Given the known seasonal variability of Xf bacterial loads, molecular results were interpreted as infection status at the time of sampling, rather than absolute prevalence.

Climatic, topographic and landscape features of the orchards were obtained from public repositories (Table 1). Climatic variables (annual and range values of temperature, precipitation and potential evapotranspiration) were obtained from CHELSA v.2.1 (http://chelsa-climate.org/) at 30 arc-second resolution. These variables represent long-term climatic conditions relevant for invasion dynamics and host stress, rather than short-term weather during the sampling year. Topographic variables, including Compound Topographic Index (CTI) and Heat Load Index (HLI) (Evans et al. 2014), were derived from a 5 m Digital Elevation Model (CNIG, Spain; http://centrodedescargas.cnig.es). Landscape configuration was characterised within a 1000 m radius around each orchard using land-cover data from the Spanish Land Use and Land Cover Information System (SIOSE AR; http://centrodedescargas.cnig.es). We quantified forest area, shrub-land area and the Contagion Index (CONTAG) using the landscapemetrics R package (Hesselbarth et al. 2019), in R (R Core Team 2019). CONTAG measures the degree of aggregation and connectivity amongst land-cover patches, with lower values indicating more fragmented landscapes and higher values representing more aggregated landscape configurations (McGarigal et al. 2002). Throughout the manuscript, we therefore interpret lower CONTAG values as indicative of greater landscape fragmentation.

Plant diversity in adjacent vegetation was quantified using Shannon diversity index (H), calculated per orchard, based on transect data using the “vegan” package (Oksanen et al. 2022) in R. In addition, the abundance of potential Xf host species was explicitly quantified to distinguish between plant diversity effects and host availability.

GLMMs were used within a multi-model inference framework (Burnham and Anderson 2002), to disentangle the relative influence of invasion status (Xf infection) and environmental drivers on plant health across agricultural and semi-natural systems. Response variables included defoliation and dead branches in almond trees and discoloration and dead branches in adjacent vegetation.

Models assumed a log-normal error distribution and included orchard identity as a random effect. The GLMM framework is well suited to accommodate unbalanced sampling designs and was, therefore, appropriate for the unequal number of infected (n = 11) and uninfected (n = 8) orchards included in the study. Predictor variables were screened for collinearity (|r| > 0.7), selected, based on ecological relevance (Suppl. material 1: tables S1, S2) and standardised (mean = 0, SD = 1). Model selection followed an all-subsets approach using Akaike’s Information Criterion corrected (AICc), considering modo els within four AICc units of the best model (Hernández-Lambraño et al. 2018). Akaike weights were used to rank predictor importance and regression coefficients were estimated as weighted averages across models (Burnham and Anderson 2002).

All statistical analyses were performed in R. The lme function of the “nlme” package was used to fit the linear mixed-effects models (Pinheiro et al. 2020) and the MuMIn package was used to perform the multi-model selection (Barton 2022).

Across the 19 surveyed orchards, qPCR analyses confirmed the presence of Xf in 36 out of 114 sampled almond trees. Based on molecular detection at the time of sampling, orchards were classified as infected (n = 11) or uninfected (n = 8). In two orchards initially considered uninfected, based on historical surveillance data, at least one almond tree tested positive by qPCR, whereas two orchards previously classified as infected showed no Xf-positive trees during the survey. Orchard classification was, therefore, updated according to current infection status.

Defoliation levels in almond trees ranged from 11.6 ± 11.2% to 71.6 ± 15.0%, while dead branch percentages ranged from 1.66 ± 2.58% to 57.5 ± 25.6% (mean ± SD across orchards). At the individual tree level, mean defoliation was significantly higher in Xf-infected trees (53.9 ± 23.0%) than in uninfected trees (30.2 ± 23.5%). Similarly, infected trees showed a higher proportion of dead branches (37.0 ± 22.3%) than uninfected trees (25.6 ± 21.6%). Both differences were statistically significant (defoliation: t = –5.07, p < 0.001; dead branches: t = –2.61, p = 0.01).

In the adjacent vegetation, 353 plant individuals belonging to 22 species were sampled across all orchards (Table 2). Of these, 155 and 154 individuals were collected around uninfected and infected orchards, respectively. Fifteen plant species occurred in both orchard types (Suppl. material 1: fig. S1). Overall, 40 individuals (11.4%) tested positive for Xf by qPCR.

The species with the highest incidence of qPCR-positive plants across both infected and uninfected almond orchards were Ulex parviflorus (37.5%), Rhamnus alaternus (17.5%), Cistus albidus (15.0%) and Salvia rosmarinus (12.5%). Amongst these, U. parviflorus, R. alaternus and C. albidus were detected in both infected and uninfected orchards (Table 2; Suppl. material 1: fig. S1). Notably, 87.5% of the 40 Xf-positive samples were obtained from areas adjacent to almond orchards infected by Xf, with the highest number of positive samples from Calicotome spinosa (5.0%), C. albidus (12.5%), C. salviifolius (5.0%), R. alaternus (15.0%), S. rosmarinus (12.5%) and U. parviflorus (32.5%). Conversely, only 12.5% of all qPCR-positive samples were collected from the natural vegetation adjacent uninfected almond orchards. In these areas, only C. albidus (2.5%), R. alaternus (2.5%), Santolina chamaecyparissus (2.5%) and U. parviflorus (5.0%) tested positive amongst the 15 potential host species analysed (Table 2; Suppl. material 1: fig. S1).

The Ouyang and Harper targets yielded different Cq values, with the Ouyang assay consistently producing lower Cq values than the Harper assay as previously shown in almond (Velasco Amo et al. 2025). In general, the Cq values of qPCR were lower for the samples obtained surrounding the Xf-infected almond orchards (16.3 to 38.0) indicating a higher inoculum concentration of the bacterium as compared to Cq values obtained for the plants sampled in areas surrounding non-infected almond orchards (32.7 to 36.3) (Table 2). Two samples, one from Bupleurum fruticescens and the other from R. lycioides, exhibited Ct values close to the detection limit (38.2 and 36.0) with some inconclusive results when repeated. Due to the destructive sampling, no additional plant material was left for confirmation. If those plant species are sampled again and confirmed positive through qPCR by the Official Plant Health Service of the Valencian Regional Community, these species will be considered new host plants for Xf at the Alicante DA (Table 2).

Discoloration rates of host plant species in adjacent vegetation did not significantly differ between Xf-infected and non-infected individuals (t = -1.72, p = 0.09). Similarly, dead branch rates showed no significant variation between groups (t = -1.21, p = 0.22).

GLMMs showed that defoliation and dead branch percentages in almond trees were associated with Xf infection status, shrub cover within orchards, abundance of Xf host plant species in adjacent vegetation, landscape configuration and climatic variables (Tables 3, 4). Neither response variable was directly related to the presence of Xf in adjacent vegetation.

Defoliation significantly increased with Xf infection, higher shrub cover, greater abundance of host species in adjacent vegetation and higher long-term variability in potential evapotranspiration, while it decreased with increasing landscape contagion and host species diversity in adjacent vegetation. Dead branch significantly increased with Xf infection, higher shrub cover and was negatively associated with annual mean of potential evapotranspiration. Significant interactions were detected between Xf infection and several environmental variables, including annual range of potential evapotranspiration, shrub cover, abundance of host species and CTI (all p < 0.001). Infected orchards exhibited higher defoliation under elevated conditions of annual mean of potential evapotranspiration, greater shrub cover, higher host abundance and lower soil water retention compared to uninfected orchards (Fig. 2).

In adjacent vegetation, leaf discoloration and dead branch rates were associated with plant diversity in adjacent vegetation, distance from orchard edges, abundance of host species and climatic variables, but not with Xf infection status of either the vegetation or the adjacent almond orchards (Tables 5, 6).

Discoloration increased with higher plant diversity, greater distance from orchards, higher host species abundance, higher temperatures and lower precipitation. Dead branch rates were positively related to distance from the orchard edge.

The detection of Xf in orchards previously classified as uninfected underscores the challenges of pathogen surveillance and highlights the importance of repeated molecular monitoring, given the temporal variability of bacterial detection and symptom expression (Velasco Amo et al. 2025). Our results further confirm that visual symptoms alone are insufficient to reliably infer infection status, as they may overlap with responses to other biotic and abiotic stressors. Beyond the cultivated system, the heterogeneous distribution of Xf detected in the surrounding natural vegetation reflects the complex spatial dynamics of this pathogen in Mediterranean landscapes (EFSA et al. 2019).

The higher frequency of Xf-positive plants in adjacent vegetation to infected orchards is consistent with a potential spillover from cultivated to semi-natural habitats, as previously suggested for Xf pathosystems (Sicard et al. 2018). Species, such as U. parviflorus, R. alaternus and Cistus spp., showed comparatively higher infection rates, supporting their potential role as pathogen reservoirs in Mediterranean ecosystems (Olmo et al. 2021; Mesmin et al. 2022). Lower Cq values in vegetation near infected orchards suggest higher bacterial loads, which may increase the likelihood that xylem-feeding insect vectors acquire the bacterium, while feeding on those infected plants (Almeida and Nunney 2015; Sicard et al. 2018), although transmission processes were not directly assessed in this study. While some detections remained inconclusive due to proximity to the qPCR detection threshold, these cases highlight the need for continued surveillance to better define the host range of Xf in the Alicante demarcated area. Overall, our findings support the relevance of integrating orchard- and landscape-scale monitoring to improve risk assessment and management of Xf spread.

Our models reveal that Xf infection was a primary factor explaining defoliation and dieback in almond trees, a direct consequence of xylem vessel occlusion and subsequent hydraulic failure (Camino et al. 2021; Landa et al. 2022). However, the impact of Xf was significantly modulated by abiotic stress. Orchards with higher drought variability and those located in areas with lower soil water retention showed exacerbated symptoms when infected. This synergy between drought and Xf infection aligns with findings in other pathosystems and supports the “invasion opportunity” hypothesis, where abiotic stress weakens host defences and facilitates greater damage by an invasive agent (De Pascali et al. 2022; Gomez-Gallego et al. 2022).

Contrary to the expectation that plant diversity might dilute pathogen pressure, we found no buffering effect of overall plant diversity on almond tree health. Instead, the abundance of Xf host plants in the adjacent vegetation was positively associated with increased defoliation in infected orchards. This result challenges the generalised notion of diversity-mediated resistance in the context of generalist plant pathogens (Gilbert and Parker 2023). For a pathogen with a broad host range such as Xf, a higher abundance of potential host plants is likely to increase landscape-level inoculum availability and provide a continuous habitat for insect vectors, thereby amplifying impacts on the focal crop (Johnson et al. 2015; Cornara et al. 2019). Interestingly, we did not find a direct link between Xf detection in the adjacent vegetation and almond damage, suggesting that the indirect role of host plants as vector reservoirs may be more critical than their role as direct pathogen reservoirs in this system (Gilbert and Parker 2023). Furthermore, disaggregated landscape and higher shrub cover within orchards were associated with greater almond damage. Fragmented landscapes (i.e. lower contag index) likely increase edge exposure and microclimatic stress, while shrub encroachment due to lack of crop management may intensify competition for water – a particularly limiting resource in Mediterranean ecosystems (Caldeira et al. 2015). These findings underscore that the impact of an invasive pathogen is deeply embedded in the landscape matrix considering both environmental gradients and management.

In contrast to almond orchards, plant health in adjacent vegetation was primarily associated with abiotic stress gradients. Temperature and water availability emerged as key correlates of discoloration and branch dieback, consistent with well-documented drought-driven hydraulic constraints in Mediterranean ecosystems (Pratt et al. 2014).

Interestingly, higher plant diversity and greater distance from orchard edges were associated with increased symptom expression. The positive effect of distance may reflect the fact that plants located further from orchards occur within denser, less-managed semi-natural communities, where competition for limited water resources is stronger under Mediterranean conditions (Sardans and Peñuelas 2013). Likewise, higher plant diversity may intensify competition for scarce resources during drought, potentially increasing plant susceptibility to abiotic stress rather than to Xf infection. Although Xf presence was not directly related to vegetation health, the abundance of potential host species was positively associated with discoloration, suggesting that vegetation composition may influence plant performance indirectly by increasing competition for limited resources or by providing more suitable habitat for xylem-feeding insect vectors, irrespective of local pathogen detection (Caldeira et al. 2015; Mesmin et al. 2022).

Our results support a shift towards integrated, landscape-level management strategies for Xf that move beyond reactive eradication of infected crops. In Mediterranean agroecosystems, ecological impact appears to emerge from the interaction between pathogen presence, host availability and environmental stress, rather than from infection status alone. This highlights the need for management frameworks that explicitly incorporate landscape configuration, vegetation structure and climate context into phytosanitary decision-making (Plantegenest et al. 2007; Cornara et al. 2019).

First, management and surveillance efforts should prioritise orchards located in high-risk environments, particularly those embedded in drought-prone and fragmented landscapes, with high shrub cover and a high abundance of alternative Xf host species in the adjacent vegetation. Such conditions may amplify impacts by increasing water stress and facilitating vector persistence and movement, thereby elevating invasion pressure at the crop–natural interface (Caldeira et al. 2015; De Pascali et al. 2022; Mesmin et al. 2022). Second, our findings suggest that the adjacent vegetation should be considered an integral component of Xf management. While complete removal of semi-natural vegetation is neither feasible nor ecologically desirable, strategic management of key reservoir species, such as U. parviflorus and Cistus spp. and the reduction of dense shrub cover within and immediately adjacent to orchards may help limit vector habitat suitability and reduce local pathogen pressure (Caldeira et al. 2015; Mesmin et al. 2022). Such targeted interventions align with landscape-ecological approaches that seek to reduce connectivity for pests and pathogens without compromising broader biodiversity values (Plantegenest et al. 2007; Loureiro et al. 2024). Third, adaptation to increasing climatic stress should be regarded as a central component of Xf management under ongoing climate change. Practices that mitigate drought stress, such as optimised irrigation, soil management aimed at improving water retention or the selection of drought-resilient rootstocks may indirectly enhance host resistance and reduce Xf-related symptoms, potentially weakening the interaction between abiotic stress and pathogen impact (Prgomet et al. 2020; De Pascali et al. 2022).

Finally, our results challenge the assumption that increasing plant diversity at the agriculture–natural interface inherently buffers disease risk. In the case of generalist, polyphagous invaders, such as Xf, promoting plant diversity per se does not appear to be a reliable management strategy. Instead, management should focus on the composition and abundance of competent host species within the landscape, as host identity and density are more directly linked to invasion dynamics and vector ecology than diversity metrics alone (Johnson et al. 2015; Gilbert and Parker 2023).

This study provides a spatially explicit snapshot of abiotic and biotic factors associated with Xf impact during an active invasion phase. However, several limitations should be acknowledged. First, the single-season sampling design does not capture the temporal dynamics of Xf titers, which are known to fluctuate seasonally and amongst years (Sicard et al. 2018). This temporal variability likely explains part of the discrepancy between prior monitoring data and our qPCR results and highlights the importance of constant surveillance to better characterise invasion trajectories. Second, although Xf infection was confirmed using specific qPCR assays, the visual symptoms recorded (defoliation and discoloration) can also be induced by other biotic agents or abiotic stressors. While our models attribute a significant portion of the observed damage to Xf infection and its interaction with environmental factors, we cannot fully exclude the contribution of other, undiagnosed stressors. Third, host-plant abundance was quantified using transect-based measures that serve as a proxy for reservoir size; future studies could explicitly integrate vector population densities, movement patterns and transmission assays to mechanistically link reservoir abundance with crop infection risk (Cornara et al. 2019; Bodino et al. 2021).

Despite these limitations, our landscape-scale approach identifies key environmental and biological correlates of Xf impact and provides a robust foundation for risk assessment in Mediterranean agroecosystems. Future research should combine multi-year monitoring, vector ecology and experimental manipulation of water stress and host plant communities to test the causal relationships suggested here. Such integrative efforts will be essential for developing predictive frameworks and resilient management strategies for Xf and other invasive, polyphagous pathogens under global change.