Mosquito population dynamics

Entomological data were obtained from historical records owned by the Mosquito Control Service operating in the area. The monitoring network consisted of 20 trap locations deployed over a 20-year period, from 2001 to 2021, using a combination of BG traps, UV traps, and CO₂ traps (Fig 1). However, the 20 trap locations were not continuously active throughout the entire study period; their usage varied over time, influenced by operational requirements and logistical constraints. Mosquito sampling was typically conducted over 24-hour intervals, although in certain cases, trap exposure was extended several days or even up to a week. Although these data were not collected as part of a dedicated sampling design specifically for this study, we consider them representative of the diversity of landscapes and seasonal patterns observed over the past two decades.

Over the two decades, a total of 80,886 female mosquitoes were captured. Among them, 10,270 were identified as Ae. albopictus and 50,052 as Culex spp., including 45,078 Cx. pipiens, 1,256 Cx. modestus, 2,597 Cx. theileri, and 1,121 Culex sp.. Notably, the presence of Ae. albopictus was first recorded in the area in 2015. To analyze the spatial and temporal variability of these species, two separate models were developed: one for Culex spp. and the other for Ae. albopictus. To enhance reliability, we consolidated the three Culex spp. into a single group. This decision stemmed from practical limitations in species-level identification throughout the study period. Specifically, the identification of Cx. modestus and Cx. theileri was only implemented in the later years of our study, while earlier specimens were classified exclusively as Cx. pipiens. This temporal inconsistency in taxonomic resolution required the pooling approach to ensure comparable data across the entire study period and provide more robust estimates of actual WNV vector presence. Moreover, this approach offers a more ecologically coherent estimate of total vector abundance relevant to WNV transmission, especially in heterogeneous environments such as rice fields, where these species co-occur.

Meteorological data were obtained from two stations within the Catalan Meteorological Network [22]: W1 (Castelló d’Empúries) and U2 (Sant Pere Pescador) (Fig 1). Station W1, located closer to the study area, served as the primary source, while station U2 provided supplementary data in cases of missing values. The dataset included daily mean (Tmean), minimum (Tmin), and maximum (Tmax) temperature, daily mean relative humidity (MRH), and daily precipitation (PPT). The integration of these variables into predictive models incorporated both daily and cumulative metrics. Cumulative values were computed as averages over 7-, 14-, and 21-day periods preceding the trap placement date, excluding the exposure period. Notably, cumulative precipitation was determined by summing values rather than averaging. Additionally, growing degree days (GDD) were calculated on both daily and cumulative bases for the same timeframes as a measure of heat accumulation, using a thermal range of 10-30^°C suitable for mosquito development in Mediterranean climates.

Land cover data were extracted from the “2017 Land Use/Land Cover Map of Catalonia (MUCSC)”, obtained from official sources [23]. We established four main categories of interest: (1) urban areas, including roads, residential areas, and industrial zones; (2) herbaceous crops, comprising both irrigated and non-irrigated types; (3) wetlands, encompassing natural wetlands and continental water bodies due to their shared vegetation characteristics in the study area; and (4) rice fields. All other land cover types not falling into any of these primary groups were classified under a fifth category labeled as ‘Others’. We calculated the relative proportions of these four categories within a 250-meter radius influence buffer around the traps. Each trap was assigned a single predominant land cover category based on the dominant environment within its buffer. However, traps with rice fields within their buffer were directly classified as rice field-influenced, regardless of the extent of rice coverage. This adjustment was necessary because these traps are often situated along the field edges, so the buffer radius may not fully capture the influence of the rice fields on mosquito populations. Notably, this habitat classification was consistent with species-specific ecological patterns: Ae. albopictus was absent from rice field traps, Cx. modestus occurred exclusively in rice fields, Cx. theileri was found in rice fields and wetlands, and Cx. pipiens was present across all habitat types.

We employed Generalized Linear Mixed Models (GLMMs) to examine the relationship between mosquito counts and the explanatory variables. All climatic variables, including daily and cumulative measures, were included as predictors, while year, trap, and land cover variability were treated as random effects. Initially, trap type (BG, UV, CO₂) was also included as a random effect, but model comparisons revealed it contributed negligibly to the explained variance compared to individual trap effects. Consequently, only trap identity was retained in the model, which inherently accounts for both attractant type and site-specific factors. GLMMs were computed using the lme4 package [24]. Model selection was performed with the dredge() and model.sel() functions from the MuMIn package [25], incorporating a function to assess collinearity [26]. Additionally, collinearity was evaluated using the vif() function from the car package [27], applying a threshold value of 5. Once the best combination of variables was determined for each mosquito species, we transitioned to a Bayesian hierarchical modelling framework, using the brms package [28] to develop the final models and generate predictions. The Bayesian approach employed a zero-inflated negative binomial distribution to account for the excess zeros in the mosquito count data. Priors for the model parameters were specified as Cauchy distributions with location 0 and scaling factor 2.5 [29]. To assess model fit, we used the loo package [30] to conduct leave-one-out cross-validation and evaluate predictive accuracy.

To comprehensively analyze seasonal patterns, we generated daily predictions for the period 2015–2018. This timeframe aligns with the “Atlas of Nesting Birds of Catalonia”, ensuring consistency between mosquito predictions and available avian data. We constructed seasonal curves by aggregating daily counts into weekly sums, obtaining the total predicted mosquito counts per week. From a spatial perspective, we performed point-based predictions at 150-meter intervals, incorporating land cover information at each specific coordinate. For mapping purposes, we calculated the average daily mosquito counts over the four-year period for each hypothetical trap location.

WNV transmission dynamics in avian communities

Using a list of bird species documented in the PNAE [31], we selected 112 species, both resident and migratory, that are abundant during the mosquito season (from May to November). Spatial variability data were obtained from the “Atlas of Nesting Birds of Catalonia” [32], which provides detailed information on the distribution and abundance of selected species for the period 2015 to 2018. The Atlas data are organized using a grid system of km squares, with each square assigned a minimum and maximum abundance value. Further details on the methodologies and data used can be found in the methodology section of the “Atlas of Nesting Birds of Catalonia” [32].

To assess WNV transmission potential, we conducted a literature review of experimental infections in avian species present within the PNAE. While some studies use field molecular and seroprevalence data to identify competent species, this approach can be problematic as it considers exposure rather than the intrinsic ability to transmit the virus [33]. In contrast, experimental studies directly measure viremia levels, which serve as a more reliable proxy for viral transmission potential. We performed a systematic search in PubMed using the following search strategy: (Bird OR Avian) AND (“West Nile virus” OR WNV) AND (experiment OR vaccine OR infection OR viremia). The resulting studies were filtered to include only those involving at least one of the 112 species identified as relevant in the PNAE. A complete list of included studies is provided in S1 Text.

To identify species with a greater potential to transmit WNV in the area, we calculated a host competence index (H_comp) using data from experimental infections. This index quantifies the ability of a host to produce infectious mosquitoes, and is defined as the product of susceptibility (s), the proportion of birds that become infected as a result of exposure, infectiousness (i), the proportion of feeding mosquitoes that become infected after a viremic blood meal, and duration (d), the number of days of infectious viremia () [10]. However, since susceptibility tends to be equal to 1, the formula is often simplified to [10,34]. To calculate H_comp, we approximated i using viremia levels expressed in plaque-forming units (PFU)/ml, computed the daily average viremia levels for each species and subsequently calculated the area under the curve (AUC) above 10⁵ PFU/ml, the threshold required for mosquito infection [10]. This index allowed us to distinguish between reservoir species (competent in virus transmission, with an H_comp>0) and non-reservoir species (not competent in virus transmission, with an H_comp = 0).

To evaluate the potential role of these bird species as WNV hosts in the study area, we calculated a host capacity index (H_cap) using the formula [8,35,36], where I denotes the infection rate, and Ab indicates species abundance. This index estimates the number of mosquitoes that a particular species could potentially infect under hypothetical virus circulation. We obtained species abundances from the “Atlas of Nesting Birds of Catalonia” [32], selecting the maximum estimated value in each km square and calculating a total abundance value per species for the entire study area. Given the absence of detailed seroprevalence studies and vector-host contact rate estimates in the area, we assumed a constant value for I and excluded this parameter from the equation. Therefore, the final formula we employed is .

Finally, we analyzed the spatial patterns of potential virus amplification and dilution by calculating the proportion of reservoir avian species (P_R) in a km square lattice, using the formula:

Disease risk assessment: Interactions between vectors and hosts

Sensitivity analysis of WNV infection threshold

To assess the robustness of the avian WNV reservoir classification, we conducted a sensitivity analysis using an alternative 10⁴ PFU/ml infection threshold, compared to the established 10⁵ PFU/ml cutoff used in the primary analysis. By performing parallel analyses with identical methodologies for both thresholds, we assessed the impact of threshold selection on species classification and derived metrics. To further explore the implications of this alternative classification, we conducted a Principal Component Analysis (PCA) combined with k-means clustering, incorporating species abundance and host competence at the 10⁴ PFU/ml threshold.

Mosquito population dynamics

The mosquito population models identified key climatic parameters that best explained the variability in species abundance: for Ae. albopictus, the 21-day cumulative maximum temperature (Tmax21), the 7-day cumulative mean relative humidity (MRH7), and the 21-day cumulative precipitation (PPT21); for Culex spp., the mean daytime temperature (Tmean), the 7-day cumulative mean relative humidity (MRH7), and the 21-day cumulative precipitation (PPT21) (S1 Table).

The seasonal patterns of Ae. albopictus and Culex spp. (Fig 2) showed qualitatively similar trends but revealed remarkable differences. Trap count estimates indicated substantially higher abundance values for Culex spp. compared to Ae. albopictus. Both species reached peak abundance during the first week of August. However, Ae. albopictus maintained a more prolonged plateau, extending from late June to mid-October, whereas Culex spp. exhibited a more pronounced peak followed by a steeper decline afterwards. Raw trap count data and model predictions for both species are provided in S2 Text.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Temporal patterns of (a) Ae. albopictus and (b) Culex spp. populations from 2015 to 2018, including the mean annual variability for each species.

Average predicted mosquito counts per trap are presented on a weekly basis, calculated as the sum of daily predictions.

https://doi.org/10.1371/journal.pntd.0013447.g002

The average spatial distribution predictions of Ae. albopictus and Culex spp. over the same period (2015-2018) were strikingly different (Fig 3). Ae. albopictus predominantly occupied urban areas, with some presence around riverbeds and wetlands, and minimal presence in rice fields and agricultural areas. Conversely, Culex spp. thrived in natural and rural areas, showing a major presence in wetlands and rice fields, while exhibiting lower occurrence in urban settings.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Spatial distribution of (a) Ae. albopictus and (b) Culex spp. populations.

The visualization employs a color gradient based on the Natural Jenks classification method, highlighting spatial distribution patterns of both species across different land cover types. Each pixel represents the predicted number of mosquitoes that would be captured in a hypothetical trap on an average day during the period 2015–2018, at a grid resolution of 150 meters.

https://doi.org/10.1371/journal.pntd.0013447.g003

WNV transmission dynamics in avian communities

Among the 112 avian species selected as most abundant in the area during the mosquito season (Table S2), we found data on WNV experimental infections for 15 species. Fig 4 shows the viremia curves for these 15 species across all reviewed experiments, with an average curve calculated for each species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Experimental WNV viremia curves for PNAE bird species.

Green lines represent individual experiment curves, while the solid black line depicts the average viremia curve. The dashed red line indicates the threshold (10⁵ PFU/ml) for mosquito infection.

https://doi.org/10.1371/journal.pntd.0013447.g004

Only six of these species (Alectoris rufa, Anas platyrhynchos, Nycticorax nycticorax, Passer domesticus, Pica pica, and Sturnus vulgaris) exhibited an H_comp>0 in our literature review (Fig 4), indicating that their average viremia curves exceed the 10⁵ PFU/ml infection threshold at some point. Consequently, these species were classified as reservoir hosts, capable of both transmitting and amplifying the virus. When considering H_cap, which integrates both the species intrinsic competence (H_comp) and abundance, P. domesticus emerged as the most significant contributor to WNV transmission in the study area, with an H_cap approximately 20 times greater than that of the second-ranked species (Table 1). Conversely, abundant Columbiformes in the area, such as Columba livia and Streptopelia decaocto, were classified as non-reservoir hosts (Fig 4) and thus acted as WNV dilutors within the avian community.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Host competence (), abundance (Ab), and host capacity () for bird species in the study area.

https://doi.org/10.1371/journal.pntd.0013447.t001

Building on the identification of reservoir and non-reservoir hosts, we examined the spatial distribution of WNV amplification and dilution phenomena across the study area, considering how variations in avian community composition influence potential transmission dynamics. Fig 5(a) presents a spatial map (1 × 1 km resolution) illustrating the proportion of reservoir species relative to the total abundance of the 15 reservoir and non-reservoir species (Table 1). Cells with higher proportions, predominantly located in rice fields and agricultural areas, are likely to experience WNV amplification due to the predominance of reservoir species. In contrast, cells with lower proportions, particularly in urban areas, suggest a greater abundance of non-reservoir species, which may correspond to potential virus dilution zones.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Maps of the study area, divided into km grid cells, illustrating the proportion of WNV reservoir birds and the WNV risk index.

The visualization employs an equal-interval classification scheme. (a) The proportion of WNV reservoir birds was calculated as the abundance of reservoir birds divided by the total abundance of reservoir and non-reservoir birds. The map represents amplification and dilution effects on WNV transmission dynamics based on the bird community composition. (b) The WNV risk index was calculated as the product of predicted mean Culex spp. mosquito counts and the proportion of WNV reservoir birds in each cell, assuming that mosquito biting patterns are proportional to host abundance. The resulting values were normalized on a scale from 0 to 1, reflecting the relative potential for mosquito-reservoir interactions. Darker areas indicate greater risk of WNV transmission due to the convergence of high vector and reservoir abundance.

https://doi.org/10.1371/journal.pntd.0013447.g005

Disease risk assessment: Interaction between vectors and hosts

Aedes-borne disease risk assessment.

Our analysis revealed significant temporal synchronization between vector abundance, seasonal human population dynamics, and imported cases of dengue, Zika, and chikungunya. The integration of quarterly FTE seasonal population data with predicted Ae. albopictus abundance showed a pronounced concurrent peak in both metrics during July, August, and September (Fig 6(a)), emphasizing the temporal overlap of increased host and vector densities during this period.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Comparison of Ae. albopictus seasonal dynamics with human seasonal population and imported dengue, Zika, and chikungunya cases.

The red line in both panels illustrates the predicted weekly average mosquito counts per trap. (a) Grey bars represent the quarterly mean Full-Time Equivalent (FTE) seasonal human population in the Alt Empordà region. (b) Stacked bars represent the weekly average number of imported cases throughout the year, both for the entire region of Catalonia and for the provinces of Barcelona and Girona.

https://doi.org/10.1371/journal.pntd.0013447.g006

Analysis of the temporal relationship between vector abundance and imported cases identified clear seasonal patterns (Fig 6(b)). The temporal dynamics of imported cases in Catalonia closely mirrored those observed in the province of Barcelona. The peak in imported cases occurred during week 36 (first week of September), with the highest importation period extending from week 32 to week 42. Although there was a clear correlation between vector abundance and case importation patterns, we observed a notable 4-week lag between their respective maximum peaks.

Sensitivity analysis of WNV infection threshold

Reducing the threshold from 10⁵ to 10⁴ PFU/ml led to the reclassification of three species (Columba livia, Streptopelia decaocto, and Coturnix coturnix) from non-reservoirs to reservoirs (Fig A and Table A in S3 Text). Although these species now technically qualify as reservoirs, they exhibit relatively low viremia levels and brief infectious periods, indicating a limited capacity to infect mosquito vectors. Consistently, PCA and clustering analyses grouped them with non-reservoir species, indicating restricted functional competence (Fig B in S3 Text).

Spatially, lowering the threshold to 10⁴ resulted in a more homogeneous community structure across the study area. Nevertheless, WNV risk patterns remained largely unchanged, with urban areas still exhibiting lower risk due to reduced mosquito abundance (Fig C in S3 Text).

Spatial and temporal dynamics of mosquito populations

Understanding arboviral transmission pathways requires an ecological perspective, which is central to the One Health approach. This framework [38] highlights the intricate interconnections between humans, disease-carrying mosquitoes, reservoir hosts, and the environments they share. Crucially, the One Health approach underscores the need to investigate mechanisms across multiple scales, identifying local environmental variability and population dynamics as key drivers of mosquito-borne disease emergence in ecologically sensitive regions.

Our analysis identified key environmental drivers shaping Ae. albopictus and Culex spp. population dynamics, revealing shared patterns and species-specific differences. Among these factors, temperature emerged as the most significant driver of population variability for both taxa, showing a positive correlation with mosquito abundance. However, previous studies have demonstrated that extreme temperatures can significantly reduce mosquito survival and reproduction [39,40]. This dual role of temperature is particularly relevant in the context of ongoing climate change: while rising temperatures may initially enhance vector abundance and accelerate viral development within mosquitoes, they may also, under extreme scenarios, push conditions beyond optimal thermal limits and disrupt phenological patterns [41]. Rainfall, in contrast, exhibited a weaker correlation, likely due to extensive water management practices in the area. Irrigation, human-mediated wetland flooding, and agricultural practices provide suitable habitats for mosquito reproduction, reducing the reliance of both Ae. albopictus and Culex spp. on natural precipitation.

Seasonal trends were similar for both taxa, with peak abundances observed between May and November, aligning with previous phenological studies conducted in Spain [42,43]. Despite these shared seasonal patterns, their spatial distributions revealed striking differences. Ae. albopictus is strongly associated with urban environments, relying on temporary, human-made water collections for reproduction [44,45]. In contrast, Culex spp. display broader ecological adaptability, thriving across a wide range of habitats, including wetlands, agricultural landscapes, and, to a lesser extent, urban areas [5,46]. Rice fields, in particular, provide ideal breeding grounds for several Culex spp., supporting large mosquito populations that play various roles in the transmission of mosquito-borne diseases [5].

Given that WNV transmission is sustained by multiple Culex mosquito species, it is essential to consider species-specific ecological roles when assessing disease dynamics. In our study, we were compelled to group several Culex spp., namely Cx. pipiens, Cx. modestus, and Cx. theileri, due to insufficient data for developing specific models for the latter two. Nonetheless, our data suggested distinct patterns among these species. Cx. modestus, predominantly found in rice fields, may contribute significantly to the WNV enzootic cycle due to its ornithophilic feeding behavior [47,48]. Cx. theileri, which also breeds in rice fields and wetlands, feeds predominantly on mammalian hosts and could have a secondary role in the epizootic cycle [47,49]. In contrast, Cx. pipiens, with its versatile feeding habits, is likely involved in both enzootic and epizootic cycles, potentially bridging transmission between avian and mammalian hosts [49–51]. In addition, Cx. pipiens was by far the most abundant mosquito species in our study area, supporting its potential leading role in local WNV transmission. This aligns with observations from other European regions, where Cx. pipiens consistently emerges as the primary WNV vector [52,53], although notable exceptions have been reported [54]. Finally, the anthropophilic Ae. albopictus, recognized as a competent vector for WNV under laboratory conditions [55,56], has also been found naturally infected in southwestern France [52], suggesting a potential involvement in urban spillover if occasional avian blood-feeding occurs.

Amplification and dilution effects driven by avian host communities

The analysis of the avian host community yielded significant insights into species spatial variability and WNV transmission pathways. However, it also presented several limitations that, while constraining our findings, underscore important knowledge gaps. First, the bird abundance data lacked temporal variability, restricting our ability to assess how seasonal fluctuations in bird communities might influence transmission dynamics. Second, experimental WNV infection data were available for only 15 of the 112 bird species identified as ecologically relevant in the study area, limiting the scope of our analysis to this subset. Although this limited representation simplifies the broader ecological reality, it still provides a reasonable approximation for understanding how the composition of the avian community determines WNV transmission potential, as many of the reviewed species are among the most abundant in the area. Third, across these 15 species, the number of experimental studies varied considerably, resulting in greater uncertainty in viremia profiles for species with limited data. Fourth, we applied a binary classification of host infectiousness, categorizing species as competent or non-competent based on whether their mean viremia exceeded a predefined threshold. While conservative, this approach may overlook the nonlinear—often exponential—relationship between viremia and mosquito infection probability, where modest increases in viral load can lead to disproportionately higher transmission potential [57]. Finally, our analysis did not account for critical aspects of avian biology such as higher viremia levels in nestlings, brood size and timing, lifespan, population turnover, and migratory behavior, which can significantly modulate host competence and viral amplification patterns [58,59].

Despite these limitations, our analysis underscores the pivotal role of bird abundance and community composition in shaping WNV dynamics through amplification and dilution effects. These phenomena, driven by the ratio of reservoir to non-reservoir species in a given area, critically influence virus circulation and spillover potential [11,60]. In our study area, P. domesticus emerged as a key contributor to virus amplification due to its high host competence and widespread abundance, consistent with other European studies that have identified this species as a major amplifier of WNV transmission [61,62]. Conversely, C. livia and S. decaocto, characterized by lower competence, may primarily act as diluting agents, as suggested in previous studies [9]. Consequently, the balance between amplification and dilution is likely driven by the relative proportions of these species across different landscapes. This is particularly evident in urban areas, where the predominance of Columbiformes over P. domesticus likely reduces the proportion of reservoir species, thereby diminishing the overall host capacity and limiting transmission potential.

However, WNV transmission dynamics are far more complex and cannot be reduced to straightforward ratios involving only a few species. Avian communities are highly diverse, and each species likely plays a distinct role in the transmission cycle. Komar et al. (2003) [10] suggested that viremia profiles often align within phylogenetically related avian groups, allowing for broader generalizations. Passeriformes, which dominate our study area in terms of abundance, are widely recognized as the most competent group for WNV transmission [10,63]. Aside from Passer domesticus, other small Passeriformes, such as Passer montanus and Serinus serinus, may also contribute to virus amplification. In contrast, Columbiformes, Gruiformes, Psittaciformes, and Galliformes generally exhibit lower viremia levels [10], indicating that some abundant species within these groups may instead play a role in WNV dilution. It is worth noting, though, that the relationship between viremia and host infectiousness represents a continuous gradient rather than a strict dichotomy [33]. While we employed the 10⁵ PFU/ml threshold for classification purposes, species with viremia levels close to this value (e.g., Columbiformes) may occasionally contribute to both amplification and dilution processes, depending on individual variability, age, and vector-host interactions [33]. Even so, transmission efficiency generally becomes substantial above the 10⁵ PFU/ml threshold, supporting its use for identifying key reservoir species [10].

Adding further complexity, the capacity of each species to amplify or dilute the virus is not solely determined by its host competence and abundance, but also by mosquito feeding preferences [60,64,65]. For instance, Cx. pipiens has shown a distinct preference for Turdus merula [64], a moderately abundant species in our study area that could also play a role in WNV transmission. This hypothesis is supported by (1) the high antibody prevalence observed in Andalusia [54], (2) the species’ phylogenetic proximity to Turdus migratorius—a highly competent host in the United States [10,60,66]—and (3) its well-documented role as an efficient amplifier for Usutu virus in Europe [61,67]. Additionally, Cx. pipiens has exhibited feeding preferences for P. pica and P. domesticus [62,64], further underscoring the potential contributions of these species to local WNV amplification.

Disease risk assessment

The analysis of vector and host populations revealed key factors influencing arboviral transmission risk. In the case of WNV, we identified rural and natural wet environments (e.g., rice fields, wetlands) as the areas of highest risk for viral amplification, where both mosquito abundance and avian community composition create optimal conditions for the enzootic cycle. This pattern is consistent with previous studies that have identified irrigated croplands and wetlands as important predictors of WNV circulation in the Mediterranean [7,46]. However, the proximity of urban centers to these areas increases the likelihood of human exposure, as both birds and mosquitoes can facilitate viral spillover into urban populations [7].

In contrast, Aedes-borne viruses, such as DENV, CHIKV, and ZIKV, pose a more direct threat in urban areas, where their primary vector, Ae. albopictus, is most prevalent. This species reaches its phenological peak between July and October, coinciding with a seasonal influx of tourists and increased pathogen importation from endemic regions. Although imported cases in Girona are few and do not exhibit a clear seasonal pattern, the strong connectivity with Barcelona, driven by secondary residences and tourist movements to northern coastal areas, likely facilitates the movement of pathogens between the two provinces. The convergence of higher vector abundance, human density, and the importation of cases from endemic areas can substantially elevate the risk of local transmission in the region [68,69].

This study enhances our understanding of the complex interactions between environmental factors, mosquito populations, avian hosts, and humans at a local scale. Our results provide valuable insights for public health management by identifying priority areas for pathogen surveillance and vector control. These results are essential not only for guiding future research efforts but also for informing public health strategies aimed at mitigating the emergence and spread of mosquito-borne diseases in the region.

Future work

Our approach represents a foundational step in mapping arboviral risk across space and time. One of its main contributions is highlighting key data gaps and outlining a roadmap for future research to improve the accuracy and comprehensiveness of transmission models. To advance this goal, upcoming studies should examine vector and host populations at finer spatial and temporal scales, evaluating how environmental and ecological factors influence transmission dynamics. From the vector perspective, exploring Cx. modestus and Cx. theileri populations is crucial, particularly in rice fields, where these species might significantly contribute to WNV transmission. Regarding avian hosts, incorporating seasonal variability and habitat connectivity estimates based on bird movement will offer deeper insights into how population fluctuations impact transmission across diverse environments. Lastly, further research into mosquito blood-feeding preferences will provide a more comprehensive view of vector-host interactions.