Citrus huanglongbing (HLB), or citrus greening, is considered the most devastating bacterial disease of citrus worldwide (Bové, 2006; McCollum and Baldwin, 2016; Jeger et al., 2023). The disease affects citrus trees, causing an overall decline which eventually leads to their death (Hendrichs et al., 2021). HLB is mainly associated with the phloem-limited bacteria Candidatus Liberibacter asiaticus (CLas), Ca. L. africanus (CLaf) and Ca. L. americanus (CLam). The CLas species is the most aggressive form of HLB and is widespread in Asia, North America, South America and Africa (Bové, 2006; Gasparoto et al., 2022). CLaf is only present in Africa, while CLam, initially identified in Brazil, is diminishing and being taken over by CLas (Teixeira et al., 2005; Gasparoto et al., 2012; Wang, 2020). HLB-associated Ca. Liberibacter spp. (CLs) can be transmitted by graft propagation (Bové, 2006), but their natural spread is primarily mediated by two psyllid vectors: the Asian citrus psyllid (ACP), Diaphorina citri Kuwayama (Hemiptera: Psyllidae), and the African citrus psyllid (AfCP), Trioza erytreae (Del Guercio, 1918) (Hemiptera: Triozidae), native to the Asian and African continents, respectively (Teixeira et al., 2005; Bové, 2006; da Graça et al., 2016; Alquézar et al., 2022).

The pathosystem of HLB involves the pathogen, the plant host and the psyllid vector, interacting in an environmental background that affects the biology of each component and its interactions (Stelinski, 2019). The three species of the genus Ca. Liberibacter associated with HLB have a persistent propagative relationship with these psyllid vectors (Carmo-Sousa et al., 2020). As these bacterial pathogens are confined to the phloem, acquisition happens when these psyllid vectors feed on the phloem sap of infected citrus plants, with inoculation occurring via salivation after invading the salivary glands. After acquisition, depending on the psyllid species, transmission can occur within 7-12 days (Canale et al., 2017). Infected adult and/or nymph vectors transmit the bacterium when feeding on healthy flushes, while emerging nymphs acquire it when feeding of them (da Graça et al., 2016; Shimwela et al., 2016). Once the psyllid vector becomes inoculative (able to inoculate plants), it retains this ability for weeks or its whole life, characterizing the transmission as persistent. The HLB-associated bacteria then develop in the tree phloem vascular system, and symptoms take at least four months to appear depending on the tree’s age (Coletta-Filho et al., 2010). The unpredictable lag period between the acquisition of the pathogen and transmission by the psyllid vectors, and for the development of visual symptoms, make HLB eradication by the visual detection of HLB-symptomatic trees problematic.

The CLas-ACP combination is widespread in Asia and America whereas CLaf-AfCP is the most prevalent in Africa (Alquézar et al., 2022). CLas is more heat tolerant than CLaf and can develop under relatively low humidity and high temperature conditions. Similarly, ACP is able to adapt to these same climatic conditions. Transmission of CLas is closely linked to insect development, where acquisition must occur with a higher efficiency during the nymphal stage (Inoue et al., 2009; Coletta-Filho et al., 2014). Thus, temperature ranges influence transmission rate and disease spread. For instance, in the USA, lower temperatures in California have been suggested to slow down the life cycle of ACP resulting in lower infection rates, whereas the higher temperature range in Florida is more favorable for ACP mating and continued disease progression (McRoberts et al., 2021; Igwe et al., 2022; Hosseinzadeh and Heck, 2023). Conversely, both CLaf and AfCP are sensitive to hot and dry conditions, and are thus restricted to areas with relatively cool and humid environments (Catling, 1969; Liu and Tsai, 2000; Bové, 2006; Dala-Paula et al., 2019). Both ACP and AfCP can transmit CLas and CLaf under experimental conditions (Reynaud et al., 2022), but more research is needed to better understand the epidemiological implications of the CLas/AfCP and CLaf/ACP combinations.

Nowadays, HLB is widespread in some of the major citrus producing areas worldwide, such as Argentina, Brazil, China, India, Mexico, South Africa and the USA, among others. No therapeutic treatments or citrus cultivars resistant to HLB are currently available. A three-pronged strategy (TPS), consisting of (i) planting of certified healthy citrus material; (ii) removal of inoculum sources-infected trees; and (iii) application of insecticide treatments to control psyllid populations, has been proposed to be key to managing citrus HLB (Bassanezi and Gottwald, 2009; Bové, 2012). Experiences in both Brazil and the USA have highlighted the futility to control HLB without removing infected trees (Alquézar et al., 2022). Consequently, the task of controlling psyllid vectors goes beyond managing current HLB outbreaks; it encompasses the critical mission of preventing future infections. By diminishing the population of infected trees and thereby reducing the reservoir of the bacterium, vector control becomes an essential part of the long-term management of this disease. In essence, vector control is a cornerstone of HLB management because it directly addresses the mode of disease transmission.

Nevertheless, effective control of vector populations can only be achieved through area-wide pest management (Gottwald et al., 2007; Bassanezi et al., 2013). This strategy consists in implementing pest management over extensive areas as a group rather than on individual orchards (Klassen, 2008). It therefore needs the coordination and teamwork of the whole industry including growers, advisers and governmental authorities, among others. This type of management is particularly effective for migratory pest species such as some insect vectors (Dalal et al., 2017; Drmić et al., 2017; Singerman et al., 2017; García-Figuera et al., 2022; International Atomic Energy Agency (IAEA), 2022). Successful cases of area-wide pest management consist of time-coordinated pesticide sprays or other means of control applied simultaneously in an extensive area (Bassanezi et al., 2013; Navarro-Llopis et al., 2014; García-Figuera et al., 2021b; International Atomic Energy Agency (IAEA), 2022).

Area-wide pest management has been attempted for CLas/ACP in Brazil, China, Mexico, Argentina, California, Florida and Texas with different aims (Bassanezi et al., 2013; McCollum and Baldwin, 2016; Singerman et al., 2017; McRoberts et al., 2019; García-Figuera et al., 2021b; Hendrichs et al., 2021; Yuan et al., 2021; Alquézar et al., 2022; García-Figuera et al., 2022). In areas where both the vector and the disease are widespread, area-wide management aims to reduce disease intensity and sustain fruit production by combining vector management with the elimination of affected trees and replanting with healthy ones from certified nurseries (Bassanezi et al., 2013; Singerman et al., 2017; García-Figuera et al., 2021b; Alquézar et al., 2022). However, in areas where the vector is widespread but the disease is still geographically restricted, the aim of area-wide management is to limit the spread of the disease to neighboring areas (Grafton-Cardwell and García-Figuera, 2018; García-Figuera et al., 2021b; Yuan et al., 2021; Alquézar et al., 2022).

Citrus-growing areas in Australia, New Zealand and the Mediterranean Basin countries are currently free of HLB (da Graça et al., 2016; McCollum and Baldwin, 2016; European and Mediterranean Plant Protection Organization (EPPO), 2022a; European and Mediterranean Plant Protection Organization (EPPO), 2022b; European and Mediterranean Plant Protection Organization (EPPO), 2022c; García-Figuera et al., 2022). However, in the European Union (EU), T. erytreae is currently spreading in the Iberian Peninsula and getting closer to the commercial citrus-growing areas of Spain and Portugal (Arenas-Arenas et al., 2019; Benhadi-Marín et al., 2020), and D. citri has recently been reported in Israel (European and Mediterranean Plant Protection Organization (EPPO), 2022) and Cyprus (European and Mediterranean Plant Protection Organization (EPPO), 2023). As indicated by Bové (2006), the introduction of the vector usually precedes subsequent outbreaks of HLB. In fact, during recent years the time lag between vector introduction and HLB outbreaks has been progressively reduced (Alquézar et al., 2022).

The HLB-associated CLs and their psyllid vectors ACP and AfCP have quarantine status in most countries of the Mediterranean Basin, including the EU, where CLs causing HLB are also included in the list of priority pests (European Union (EU), 2019a; European Union (EU), 2019b; Food and Agriculture Organization of the United Nations (FAO), 2022). These EU priority pests are the ones with the highest potential socioeconomic impact. They are subjected to additional mandatory phytosanitary measures including annual surveys, contingency and action plans, simulation exercises, and public information (European Union (EU), 2016; Aragón et al., 2022). Despite the phytosanitary regulations, over the last few years several interceptions of ACP at the EU borders have been reported. Furthermore, irregular imports of propagating plant material constitute another high-risk entry pathway (EFSA PLH Panel et al., 2021). This fact, together with the presence of AfCP in the Iberian Peninsula and ACP in Israel and Cyprus, represents a serious threat to the citrus-growing areas in the Mediterranean Basin (Gutierrez and Ponti, 2013).

The Mediterranean Basin accounts for around 12% of the world’s citrus-growing areas (McCollum and Baldwin, 2016; Siverio et al., 2017; Food and Agriculture Organization of the United Nations (FAO), 2020). Citrus in the Mediterranean are cultivated mainly in coastal areas, forming a virtually continuous belt along the basin. Spain has the largest citrus-growing area in the Mediterranean Basin, with ∼23% of the total surface area (Food and Agriculture Organization of the United Nations (FAO), 2020). In Spain, commercial citrus crops cover nearly 300,000 ha with a production of ∼6.6 million tonnes, making it the main producer in the EU and the sixth in the world (Siverio et al., 2017; Food an Agriculture Organization of the United Nations (FAO), 2020). With a 25% share, Spain is the world’s leading exporter of fresh citrus, with nearly 60% of the production destined to foreign markets. The main citrus region in Spain is the Valencian Autonomous Community, with 54% of the total Spanish citrus-growing area, followed by Andalusia (28%) and the Region of Murcia (13%) (Ministerio de Agricultura, Pesca y Alimentación (MAPA), 2019; Food and Agriculture Organization of the United Nations (FAO), 2020; Ministerio de Agricultura, Pesca y Alimentación (MAPA), 2022). Furthermore, citrus in Spain also has an important sociocultural component and ornamental citrus are widely planted in residential areas (Duarte et al., 2016; Arenas-Arenas et al., 2018; Torregrosa et al., 2019).

The Spanish National Plant Protection Organization (NPPO) developed contingency plans for HLB and its vectors (Ministerio de Agricultura, Pesca y Alimentación (MAPA), 2021a; Ministerio de Agricultura, Pesca y Alimentación (MAPA), 2021b; Ministerio de Agricultura, Pesca y Alimentación (MAPA), 2023a). These contingency plans were recently evaluated in a simulation exercise (Aragón et al., 2022) and have been translated to a specific regulation to address the control and eradication of the AfCP and to prevent the introduction of the ACP and CLs (Ministerio de Agricultura, Pesca y Alimentación (MAPA), 2023b). Nevertheless, how to organize citrus growers in order to coordinate vector control through area-wide management is not specifically addressed in this contingency plan. The particularities of the citrus industry in the Mediterranean Basin, with small orchards (<1 ha on average in Spain) often managed independently by individual growers (Reig-Martínez and Picazo-Tadeo, 2004; Fernández-Zamudio et al., 2005; Morell-Monzó et al., 2021), increases their exposure to primary infections, namely those coming from neighboring areas (Bassanezi et al., 2013; Alquézar et al., 2022). The different characteristics of the citrus industry in the Mediterranean Basin limit the extrapolation of the area-wide management programs designed for other citrus-growing areas, thereby making the implementation of area-wide management a challenge. Moreover, area-wide management has not been attempted for AfCP control elsewhere, despite being the vector currently spreading in the Iberian Peninsula.

Considering that ACP and AfCP are now present in the Mediterranean Basin and to prepare for its potential spread to the main citrus-growing areas in the region, the objectives of this study are:

In addition, this proposal could provide valuable support to policy makers and the citrus industry in planning for risk management in the event of the spread of HLB vectors or the potential introduction of the disease into citrus-growing regions in the Mediterranean Basin. It can also be used by modelers to improve surveillance and management programs. By integrating epidemiological models with opinion dynamics models that incorporate the principles of area-wide management, it will be possible to determine how best to maximize the potential effectiveness of voluntary HLB control campaigns (Gottwald et al., 2014a; Milne et al., 2020).

The Valencian Autonomous Community (ES52 NUTS level 2 EU territorial unit for statistics) was selected as the study area. This region consists of three provinces: Alicante (ES521 NUTS 3), Castellón (ES522 NUTS 3) and Valencia (ES523 NUTS 3). The study area is the region with the largest extension dedicated to citrus in Spain, with characteristics similar to those of other citrus-growing areas in the Mediterranean Basin. Of the three provinces, Valencia has the biggest area devoted to citrus.

The resulting area-wide management zones in our study were defined as Pest Management Areas (PMAs). To facilitate regionalization of the PMAs, the study area was gridded into 24,048 cells of 1 km² using the European Environment Agency (EEA) reference grid downloaded from its official website. This grid is in the ETRS89 Lambert Azimuthal Equal Area (LAEA) projection coordinate reference system (ETRS89/ETRS-LAEA (EPSG:3035)) with a coding system identifying each grid cell with a unique ID (Peifer, 2011).

The original 24,048-cell grid was filtered to identify cells with commercial citrus coverage using a georeferenced dataset from the Spanish Agricultural Plots Geographic Information System (SIGPAC). The SIGPAC dataset provided comprehensive information on the size and shape of commercial citrus orchards and was intersected with the reference grid of the study area. Finally, 7,021 of the original 24,048 cells were identified as cells with commercial citrus coverage. Thus, the grid of the study area was defined by 7,021 cells of 1 km2.

Based on previous studies in other citrus areas (Díaz-Padilla et al., 2014; Gottwald et al., 2014a; Gottwald et al., 2014b), the study area was characterized through different risk factors than can affect the introduction and spread of HLB vectors and, consequently, the disease epidemiology (see Supplementary Material A for further details). For this purpose, different georeferenced datasets were exploited to extract data on climatic variables, commercial citrus orchards, population census data and the network of the main transportation corridors. The extracted data were transformed to the ETRS89/ETRS-LAEA (EPSG:3035) reference system using R software version 3.6.0, https://www.R-project.org and aggregated at the cell level into the grid of the study area.

This study presents a risk-based approach to defining management areas for HLB vectors. The rationale of this approach is based on the regionalization of citrus-growing areas into Pest Management Areas (PMAs) that share a similar overall risk to the introduction and spread of insect vectors but considering spatial constraints to make it feasible from the viewpoint of implementation. The size and location of PMAs were estimated using the Ward-like hierarchical algorithm developed by Chavent et al. (2018). This spatial aggregation algorithm provides relevant elements for the grouping of areas under similar risk and allows the incorporation of geographical information to address the clustering process, which in the context of vector-borne plant diseases is a fundamental point to support the operational implementation (Bassanezi et al., 2013; Yuan et al., 2021). The Valencian Autonomous Community (ES52 NUTS 2), Spain, was selected as the study area because it is the largest citrus-growing region in Europe and is representative of others in the Mediterranean Basin. Specifically, PMAs were designed and optimized for “La Ribera Alta” and “El Camp de Túria” counties.

To date, several proposals for area-wide management of HLB have been developed, framed in different epidemiological settings  (Bové, 2006; Manjunath et al., 2010; Kumagai et al., 2013; Flores-Sánchez et al., 2017; Badaracco et al., 2021; García-Figuera et al., 2021a). In general, the rationale behind these approaches considers regional risk factors, estimated based on epidemiological variables such as cultivar susceptibility, host density, and vector presence and abundance (Gottwald et al., 2014a; Flores-Sánchez et al., 2017). Some of them also incorporate other factors related to human factors, such as the availability of infrastructure, human and economic resources in the citrus-producing regions, as in Mexico (Flores-Sánchez et al., 2017), or the social willingness to participate, as in California (McRoberts et al., 2019; García-Figuera et al., 2022). Nevertheless, although previous research has shown improved HLB control with the area-wide strategy (Bassanezi et al., 2013; García-Figuera et al., 2021a), the standard settings for determining the size and location of area-wide management have not been established (Flores-Sánchez et al., 2017).

Also, different recommendations have been suggested concerning the area-wide size to manage HLB. Using empirical data, Bové (2012) proposed an area size of 500 ha in Brazil, whereas sizes in California range from 4,000 to 20,000 ha (Flores-Sánchez et al., 2017). Nonetheless, the farming systems in Brazil and California are usually composed of large orchards. This contrasts with the citrus-growing areas in the Mediterranean Basin like Spain, where farms are smaller and divided up into orchards of <1 ha often managed independently by individual growers (Fernández-Zamudio et al., 2006). Thus, PMAs aggregation will probably be more challenging from an extension point of view.

To implement our proposal, the study area was characterized by several risk factors at a grid resolution of 1 km² using data on climatic suitability, conventional, organic, abandoned and residential citrus areas, and transportation corridors. These risk factors have been associated with the introduction and spread of ACP, and in consequence with HLB epidemiology (Halbert et al., 2010; Richards et al., 2013; Gottwald et al., 2014a; Martini et al., 2015; Bayles et al., 2017; Flores-Sánchez et al., 2017). Our proposal considers the same approach for AfCP. While there have been extensive studies of ACP and AfCP in other regions, there is limited knowledge about both vectors’ biology, spread capacity and host range in the Mediterranean Basin. Likewise, the Valencian Autonomous Community remains free from HLB and from both ACP and AfCP and so it is uncertain how the vectors and the disease will behave.

The risk factors included in our study do not consider the wind effects (i.e., direction and speed) although previous studies on ACP spread dynamics and area-wide management often consider this factor to predict vector distribution in an invading scenario from an entry point. Antolínez et al. (2022) indicated that ACP adults are likely to spread across the prevailing wind direction and that wind speeds higher than 48 km/h are able to dislodge adults from citrus leaves and trigger passive long-distance spread events. Moreover, some studies have hypothesized that severe meteorological wind-related episodes, such as hurricanes in Florida, greatly influenced the major long-distance spread of ACP in Florida (Johnston et al., 2019; Monzó and Vanaclocha, 2023). Nevertheless, the availability of accurate wind data at large scales on a regional resolution is scarce given that wind observations vary on small space and time scales due to the fact that wind is affected by the local terrain, vegetation and buildings. Furthermore, recent studies have shown that other factors, such as orchard layout, have a stronger influence on HLB incidence and spread than wind (Benhadi-Marín et al., 2021; Primiano et al., 2023). Other risk factors integrated in previous studies, such as packing houses and nurseries (Gottwald et al., 2014a; Flores-Sánchez et al., 2017), were not included either. The relevance of these factors in the vector distribution also remains unclear. Similarly, host data (variable and cultivar) was not included because of the lack of available data on cultivar preference and susceptibility.

Nevertheless, the proposed regionalization methodology for estimating the size and location of PMAs ensures sufficient flexibility to update the default risk factors considered or to add new ones as more information about the insect vectors/disease become available or as the epidemiological setting evolves. Similarly, although in our study two options have been proposed for estimating the overall risk, i) averaging its influence, and 2) giving greater importance to the risk posed by the presence of abandoned and organic orchards than to the rest of the risks, other criteria can be accommodated. In the same way, further refinements may be implemented for the risk factor normalization in the presence of extreme values.

The approach proposed in our study for defining the size and location of PMAs makes use of a hierarchical clustering algorithm that allows the introduction of spatial constraints (Chavent et al., 2018). This kind of clustering methodology is usually called regionalization. The term regionalization was defined by Guo (2008) as the process of aggregating a set of spatial entities into a reduced number of regions in a way that a predefined objective function is optimized. Spatial contiguity and homogeneity are the basic criteria underpinning regionalization. Spatial contiguity requires that spatial connectivity or spatial tightness should be met. Beyond the ClustGeo (Chavent et al., 2018), there are several important regionalization algorithms such as SKATER (Assunção et al., 2006) or REDCAP (Guo, 2008). However, ClustGeo was chosen because it enables the definition of the spatial relationship between the units intended for grouping through the assessment of similarity between geographical coordinates, without the necessity for these units to satisfy explicit spatial connectivity requirements, as, for instance, demanded by SKATER (Assunção et al., 2006). This functionality offered by ClustGeo was the decisive factor that led us to opt for this algorithm, given the specific spatial arrangement characteristics of our study area concerning the units to be grouped.

These regionalization techniques are used in several disciplines, including sociology, economics, urban planning, politics and health to identify areas with similar characteristics and to get useful information to support policy-making (Camêlo Aguiar et al., 2020; Guo et al., 2022). However, to our knowledge, this is the first time that this methodology has been used in plant health to group areas with a similar level of risk of vector introduction and spread in order to address a coordinated management and, additionally, on an open-source implementation. The greatest potential of this algorithm in the frame of area-wide management is that it makes it possible to consider the spatial attributes of the units to be clustered and to modulate the importance of spatial aggregation versus the “risk” in the design of the solutions. This importance of spatial aggregation is defined in the implementation offered (i.e., the R package ClustGeo) through the setting of the α parameter. Specifically, when α = 0 the spatial dissimilarities are not taken into account and when α = 1 the “non-spatial attributes” distances are not considered and only the spatial distances are taken into account (Chavent et al., 2018; Camêlo Aguiar et al., 2020).

The Clustgeo algorithm presents a marked variability in the design of the results depending on the setting of parameter k (i.e., maximum number of clusters or PMAs allowed) and depending on the value of parameter α. Specifically, the choice of the maximum number of PMAs (i.e., clusters) was defined assuming a maximum PMA size of 25 cells based on expert criteria, but this parameter can be adjusted by the end user or optimized based on different criteria. Furthermore, a sensitivity analysis was performed to evaluate the effect of three different values of α. For this purpose, in addition to the graphical results, several internal validation measures were computed to quantify compactness and/or separation as well as the trade-off between the loss of risk homogeneity and the loss of spatial aggregation (Leskovec et al., 2014; Kassambara, 2017). For α = 0.9 under the four configurations evaluated (OR_1,acp , OR_2,acp , OR_1,afcp , OR_2,afcp ), the results obtained in both counties, “La Ribera Alta” and “El Camp de Túria”, showed the greatest spatial aggregation in the design of the location of PMAs, which is fundamental for their implementation from a practical point of view. Furthermore, with α = 0.9 the solutions also supported at least 72.4% homogeneity in terms of overall risk.

In sum, this PMAs design proposal, addressing two overall risk estimates for both ACP and AfCP, offers a comprehensible, accessible and adaptable approach both to estimating the risk factors associated with the introduction and spread of the two HLB vectors and to establishing their corresponding management areas (PMAs). Although it was implemented at NUTS2 level due to administrative reasons, this framework can be upscaled at country and even at Mediterranean Basin levels to implement a standard management and minimize the possible migration of vectors among areas. Thus, the development of this PMAs design proposal means having a preventive tool at the service of the citrus industry. Its use can be extended beyond the coordination of vector/disease management, for instance, to the field of epidemiological surveillance as well as to the context of education and awareness-raising.

Nevertheless, future work would be needed to verify the citrus data at orchard level and its classification as conventional, organic, abandoned, residential, etc. Indeed, the locations of organic and abandoned citrus orchards were randomly allocated to reach the total area reported by Ministerio de Agricultura, Pesca y Alimentación (MAPA) (2019; 2020). The isotherm for citrus growth and population census data were used to estimate the residential citrus areas. Thus, it would also be necessary to update the default risk factors considered, and to add new ones when more information about the insect vectors/disease become available after an eventual introduction in the study area. Additionally, social willingness toward the adoption of PMAs should be properly addressed. Previous experiences from regions with the presence of disease and vector have demonstrated that the only successful action to control the spread of HLB is to control the vector on a large spatial scale (Singerman et al., 2017; Milne et al., 2018; Alquézar et al., 2022). Area-wide programs in countries with the presence of HLB typically rely on voluntary adoption (García-Figuera et al., 2021a; García-Figuera et al., 2022). To achieve voluntary adoption, in our case, the advantages of the coordinated treatments should be effectively communicated and make stakeholders trust the risk manager to keep them motivated and involved in the program (Klassen, 2008; Singerman et al., 2017; McRoberts et al., 2019; Monzó and Stansly, 2020; García-Figuera et al., 2021b; Pavone, 2022).

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author. Data curation and formal analysis have been implemented in the R statistical programming. All the generic code used in the research is open-source and available at https://zenodo.org/records/10079629.

AG: Data curation, Writing – original draft, Writing – review & editing. RB: Conceptualization, Writing – review & editing. WL: Conceptualization, Writing – review & editing. PV: Methodology, Writing – original draft, Writing – review & editing. AV: Conceptualization, Data curation, Funding acquisition, Methodology, Project administration, Supervision, Writing – original draft, Writing – review & editing. EL: Data curation, Writing – original draft, Conceptualization, Formal analysis, Investigation, Methodology, Software, Writing – review & editing.