{"paper_id":"15001b14-84ea-41fb-a0da-d2eec1e908af","body_text":"1 \n 1 \n 2 \n 3 \nTrait based assessment of the invasion potential of disease vector mosquitoes 4 \n 5 \n 6 \nRebecca Pabst1*, Carla A. Sousa1, César Capinha2,3  7 \n 8 \n 9 \n 10 \n1 Global Health and Tropical Medicine, GHTM, LA -REAL, Institute of Hygiene and Tropical 11 \nMedicine, IHMT, NOVA University Lisbon, Lisbon, Portugal. 12 \n2 Centre of Geographical Studies, Institute of Geography and Spatial Planning, University of 13 \nLisbon, Lisboa, Portugal. 14 \n3Associate Laboratory TERRA, Lisboa, Portugal.  15 \n 16 \n* Corresponding author 17 \nE-mail: pabst.rebecca@gmail.com  18 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n2 \nAbstract 19 \nMosquito-borne diseases pose a growing global health threat, largely driven by the human -20 \nmediated spread of vector species beyond their native regions. Although only a few mosquito 21 \nspecies historically established populations outside their native ranges, many have expanded 22 \nrapidly in recent decades.  Once established, these invaders are notoriously difficult to control, 23 \nemphasizing the need for proactive identification before human-mediated spread occurs. Here, we 24 \npresent a framework to anticipate invasion potential for 184 mosquito species of medical 25 \nimportance based on their ecological, life-history, and macroecological traits. We first compiled a 26 \ncomprehensive dataset of 26 traits characterizing each species. We then used random forest models 27 \nto relate these traits with the probability of species being introduced in new regions (before and 28 \nafter 1950, marking the onset of widespread trade globalization), and of establishment following 29 \nintroduction. Models achieved moderate to good predictive performance (AUC = 0.78 -0.85) and 30 \nrevealed that species native to Asia and Australia, adapted to human -made breeding sites, and 31 \ntolerant of climatic extremes are consistently more likely to be introduced and to establish in non-32 \nnative regions. Among species with no known invasion history, we identified 24 with higher 33 \npotential to become future spreaders, of which 1 7 also exhibit high establishment probabilities 34 \n(‘high-risk species’). These results show that invasion potential can be inferred, to some extent, 35 \nfrom intrinsic species traits and provide a quantitative basis for proactive surveillance, enabling 36 \nprioritization of species most likely to become introduced in the future.  37 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n3 \nAuthor Summary 38 \nMosquito-borne diseases threaten more than half of the world’s population and cause over 700,000 39 \ndeaths each year. Only a small share of mosquito species can spread these diseases, but some of 40 \nthem are moving into new regions where they have never been seen before . The spread of these 41 \nmosquitoes has led to increasing numbers of locally transmitted outbreaks in regions that 42 \npreviously, or in recent times, had no mosquito -borne diseases. Human activities like trade and 43 \ntravel help mosquitoes spread, and once they arrive, they are extremely difficult to eradicate . 44 \nTherefore, it is crucial to understand which species may spread in the future and to identify those 45 \nthat should be closely monitored to prevent their introduction and establishment. In this work, we 46 \nlinked species characteristics with their known invasion history to identify the factors driving their 47 \nintroduction and establishment in new regions. We found that species from Asia and Australia, 48 \ncapable of using human-made breeding sites, and tolerant of climatic extremes are most likely to 49 \nbecome invaders. We then used these findings to predict which species might spread next . We 50 \nidentified 24 species with high invasion potential, including 1 7 that also have high chances of 51 \nestablishing once introduced. These results demonstrate that invasion risk can be predicted from 52 \nmeasurable species traits, providing a framework to guide early -warning surveillance and 53 \nprioritize species for monitoring before they begin spreading and become widespread vectors of 54 \nhuman disease.  55 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n4 \nIntroduction 56 \nMosquito-borne diseases are a major threat to global public health, causing more than 700,000 57 \ndeaths annually and placing over half of the world’s population at risk of infection (1). This risk 58 \nis largely attributed to a small group of competent vector species capable of transmitting pathogens 59 \nsuch as dengue, Zika, chikungunya, malaria, and West Nile virus (2,3). Historically, many of these 60 \nspecies maintained relatively restricted distributions. However, in recent decades, the range of 61 \nseveral mosquito vectors has expanded at an unprecedented pace. Of the approximately 184 62 \nmosquito species, from which human pathogens have been isolated from wild-caught females (4), 63 \nby now 46 have already been introduced into regions outside their native ranges, with 28 confirmed 64 \nas having established populations (5). These introductions have expanded the species’ ranges, 65 \nsometimes into new continents and other fairway regions and enabled local disease transmission 66 \nin areas considered unsuitable or free of certain diseases, a reminder that while mosquitoes can 67 \nexist without transmitting pathogens, such pathogens cannot be without mosquitoes (6,7).  68 \nDespite receiving increasing attention, the drivers of mosquito introductions and 69 \nestablishment in non -native regions remain poorly understood. Some species, such as Aedes 70 \naegypti, have expanded globally since the 15th century (8), while others, like Aedes albopictus, 71 \nAedes japonicus and Anopheles stephensi, emerged as invasive non-native species (i.e., introduced 72 \nand established populations outside their native range; cf. ,9) only in recent decades (10–12). For 73 \nmany other mosquito species, however, there is no evidence that they were transported by humans 74 \nor successfully established themselves following introductions. This is intriguing because, in 75 \nseveral cases, such species have ecological traits, habitat use, or associations with humans that are 76 \nbroadly similar to those of invasive species (13). Hence, a key question in vector ecology and 77 \nprevention remains: why are some mosquito species being introduced and becoming established 78 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n5 \nin non -native regions while others remain restricted to their native ranges? Shedding light on 79 \ndrivers of these differences would be of key relevance for supporting surveillance and 80 \nintroduction-prevention efforts, allowing to help identify species more likely to become introduced 81 \nin the future. Similarly, in a global context where resources for surveillance are limited and the 82 \npool of known invasive species is likely incomplete (14), understanding these factors may also 83 \nhelp identify introduced or invasive species that remain undetected. To address this question, it is 84 \nnecessary to consider multiple factors shaping the propensity of mosquito species to be introduced 85 \nand established in non-native regions (15). The disposition to be transported is expected to depend 86 \nstrongly on intrinsic traits that mediate associations with human -traded commodities or transport 87 \nvectors. Classic examples include oviposition in human-made breeding sites such as used tires and 88 \nliving plant pots, which constitute major pathways for the spread of widespread species such as 89 \nAedes aegypti and Ae. albopictus (16). The breadth of a species’ geographic distribution is also 90 \nlikely to be relevant, with taxa occupying wide native ranges, particularly those overlapping 91 \nregions of high trade volume and openness, being more frequently exposed to transport 92 \nopportunities (17). Beyond introduction, establishment success likewise depends on species traits. 93 \nWater requirements and desiccation resistance influence survival during transit and the prevalence 94 \nof viable propagules upon arrival (18), while broader environmental tolerances increase the 95 \nlikelihood of encountering suitable conditions for colonization in non-native regions (e.g., salinity 96 \n19). Species with higher heat tolerances may also have a competitive advantage under increasingly 97 \nextreme temperatures (20). In order to provide warning lists to guide preventive biosecurity 98 \npolicies, trait-based approaches have recently emerged as valuable tools to identify species more 99 \nlikely to become introduced and invasive (e.g., 21 –23). By integrating variables representing 100 \necological and life-history traits and macroecological patterns into predictive frameworks, these 101 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n6 \napproaches assess which species are more or less likely to enter the global invasion pathway, while 102 \nalso highlighting the traits mediating interspecific variation in invasion potential. For example, Pili 103 \net al. (22) developed a global trait-based model and showed that to predict species establishment, 104 \nboth life -history and macroecological traits were important predictors and show that invasion 105 \nsuccess concentrates in species combining ecological flexibility with high probability of human 106 \nmediated transport. Similar trait-based assessments for disease vector mosquitoes, would provide 107 \ncritical advances in anticipating future biological invasions and the emergence of vector -borne 108 \ndiseases to protect naïve human populations. Revealing which mosquito species are predisposed 109 \nto introduction and establishment, would support surveillance planning, refine monitoring 110 \npriorities and advance our understanding of how ecological specialization, climate tolerance, and 111 \nhuman-mediated transport interact to shape global mosquito invasion risk. However, to date, such 112 \nan assessment has been constrained by the lack of comprehensive trait data, with most 113 \ncompilations restricted to a subset of species of known medical importance (24,25). Here, we take 114 \nadvantage of a newly compiled, extensive database covering 184 mosquito species capable of 115 \nnatural infection with human pathogens to perform a trait-based assessment of species introduction 116 \nand invasion potential. The trait dataset integrates species ecological, life -history, and 117 \nmacroecological data and to our knowledge is by far, the most taxonomically comprehensive 118 \navailable. Thus, this dataset provides a unique opportunity to test which characteristics of species 119 \nare associated with a higher or lower invasion potential. Specifically, we combined these data with 120 \ndata from a recent global assessment of mosquitos introduction records and non -native ranges 121 \n(Pabst et al., 2025) within a machine learning framework to: (i) identify species traits that predict 122 \ntheir propensity for introduction and establishment in non -native regions and, (ii) assess which 123 \ncurrently unintroduced or non-established species may pose future invasion risks. 124 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n7 \nMethods 125 \n2.1 Trait compilation 126 \nOur aim was to estimate the probability of introduction and establishment of mosquito species 127 \nvectors for human diseases. Species included in the analysis followed Pabst et al. (5) and were 128 \nlimited to mosquito vectors from which human pathogens were isolated from field-caught females, 129 \nverified by at least two literature sources (4) or by one source plus a disease association recorded 130 \nin Wilkerson et al. (26). For that purpose, we assembled a database of variables describing 131 \necological and life -history, as well as macroecological traits of these mosquitoes. Traits were 132 \nselected based on their potential relevance on human-mediated transport of species, their survival 133 \nduring transit, and their establishment in new environments (Table 1). We focused on variables 134 \nthat are relatively stable across environmental gradients and for which information was available 135 \nfor most species, thereby minimizing data gaps and ensuring general applicability. Most ecological 136 \nand life-history traits (e.g., oviposition behavior, desiccation resistance, salinity tolerance , flight 137 \nrange) were collected through literature research, and each species -trait combination was 138 \nreferenced to a specific literature source (Appendix S1). Macroecological variables (e.g., climatic 139 \nlimits) were derived from species distribution data and spatial environmental layers (CHELSA; 140 \n27,28) or obtained from specialized sources like species distribution from Wilkerson et al. (26) 141 \nand blood meal hosts from Soghigian et al. (29). The rationale for including these derived traits 142 \nand details on the procedures for their collection are described below.  143 \nTable 1. List of species traits collected for the 184 mosquito species of medical importance. 144 \nTrait Definition \nInformation recorded \nin variable \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n8 \nOviposition in non-human-\nmade breeding sites \nWhether a species lays eggs in natural meaning non-\nhuman-made breeding sites. Examples include tree holes, \nswamps, rivers. \nClasses: yes / no \nOviposition in human-\nmade breeding sites \nWhether a species lays eggs in human-made breeding sites. \nExamples include plastic container, vase, man-made cut \nbamboo that collects water, drainage ditch, rice field. \nClasses: yes / no \nOviposition in small \nbreeding sites \nWhether a species lays eggs in breeding sites smaller than 1 \nm2. Examples include tree holes, coconut shells, plastic \ncontainers, rockpools. \nClasses: yes / no \nOviposition in large \nbreeding sites \nWhether a species lays eggs in breeding sites larger than 1 \nm2. Examples include swamps, river margins, swimming \npools. \nClasses: yes / no \nFresh water Whether a species can develop in fresh water. Classes: yes / no \nBrackish water Whether a species can develop in brackish water. Classes: yes / no \nSalt water Whether a species can develop in salt water. Classes: yes / no \nEgg survives without water \nThe ability of eggs to survive without water for long \nperiods of time. Short-time means that eggs from dry soil \nsamples or under experimental conditions could still be \nincubated in some cases after up to 14 days. \nClasses: yes / no / short-\ntime \nDesiccation resistance \nAbility of eggs to resist desiccation (i.e. survive after \ndrying). \nClasses: yes / no \nOviposition strategy The way in which a species lays its eggs. \nClasses: In rafts / \nindividually / \nindividually but \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n9 \nclustered / clustered \nunderwater \nMinimum mean annual \ntemperatures of its native \nrange \nMinimum mean annual minimum temperature at which the \nspecies was observed. The extracted value corresponds to \nthe 97.5% percentile of mean annual minimum \ntemperatures at observation sites within countries where the \nspecies occurred naturally (in its native range).  \nContinuous value (°C) \nMaximum mean annual \ntemperatures of its native \nrange \nMaximum mean annual maximum temperature at which the \nspecies was observed. The extracted value corresponds to \nthe 97.5% percentile of mean annual maximum temperature \nat observation sites within countries where the species \noccurred naturally (in its native range).  \nContinuous value (°C) \nMinimum sum of annual \nprecipitation of its native \nrange \nMinimum sum of annual precipitation at which the species \nwas observed. The extracted value corresponds to the \n97.5% percentile of minimum sum of annual precipitation \nat observation sites within countries where the species \noccurred naturally (in its native range).  \nContinuous value (mm) \nMaximum sum of annual \nprecipitation of its native \nrange \nMaximum sum of annual precipitation at which the species \nwas observed. The extracted value corresponds to the \n97.5% percentile of maximum sum of annual precipitation \nat observation sites within countries where the species \noccurred naturally (in its native range).  \nContinuous value (mm) \nNative distribution range \n(whole country area) \nDistribution range of species based on the total area of \ncountries in which the species occurred in its native range. \nContinuous value (km2) \nNative to Africa Whether a species is native to the African continent.  Classes: yes / no \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n10 \nNative to Asia  Whether a species is native to the Asian continent.  Classes: yes / no \nNative to Australia  Whether a species is native to the Australian continent.  Classes: yes / no \nNative to Europe  Whether a species is native to the European continent.  Classes: yes / no \nNative to North America  \nWhether a species is native to the North American \ncontinent.  \nClasses: yes / no \nNative to South America  \nWhether a species is native to the South American \ncontinent.  \nClasses: yes / no \nProportion amphibian \nblood meal host \nProportion of bloodmeals derived from amphibian hosts. \nTogether, the amphibian, avian, mammalian, and reptilian \nvalues sum to 1.  \nProportion, between 0 \nand 1. \nProportion avian blood \nmeal host \nProportion of bloodmeals derived from avian hosts. \nTogether, the amphibian, avian, mammalian, and reptilian \nvalues sum to 1.  \nProportion, between 0 \nand 1. \nProportion mammalian \nblood meal host \nProportion of bloodmeals derived from mammalian hosts. \nTogether, the amphibian, avian, mammalian, and reptilian \nvalues sum to 1.  \nProportion, between 0 \nand 1. \nProportion reptilian blood \nmeal host \nProportion of bloodmeals derived from reptilian hosts. \nTogether, the amphibian, avian, mammalian, and reptilian \nvalues sum to 1.  \nProportion, between 0 \nand 1. \n 145 \n2.1.1 Country wide distribution data and native range delineation 146 \nNative ranges were delineated to capture each species’ pre -modern trade-driven distributions and 147 \nnatural ecological context and to avoid temporal overlap between predictors and invasion 148 \noutcomes. Furthermore, natural distribution determines invasion outcomes due to regional 149 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n11 \ndifferences in historical trade relations. Country -level data were obtained from Wilkerson et al. 150 \n(26) and the Walter Reed Biosystematics Unit (30). Each distribution was visually inspected, and 151 \nonly countries within the confirmed native range were retained. Calculated metrics included 152 \nnative-range area (sum of constituent country areas in km²) and native continent(s)  as binary 153 \nvariables. Including the latter accounts for the species’ origins.  154 \n 155 \n2.1.2 Climate data  156 \nOur trait-based predictive framework requires understanding the climatic tolerances that determine 157 \na species’ survival, development, and reproductive cycles. Temperature constrains processes, such 158 \nas gonotrophic cycle duration, body size, and fecundity  (31,32), while precipitation shapes larval 159 \nhabitat availability by influencing standing water body formation and habitat persistence (33). To 160 \ncharacterize each species’ climatic niche, we extracted native-range climate values from CHELSA 161 \nv2.1 layers (27,28) at 30 arc-second resolution (~1 km at the equator). We assembled occurrence 162 \nrecords (1980–2024) from GBIF (34) (via the ‘rgbif’ package; Chamberlain et al. 2012) and from 163 \nVectorMap (30), two global sources of mosquito data. We then cleaned the records for coordinate 164 \naccuracy, duplicates, and outliers using the ‘CoordinateCleaner’ package (Zizka 2017). Points 165 \nwere restricted to one per grid cell within the species’ native range and overlaid with monthly 166 \nmaximum and minimum temperature, and precipitation layers, to calculate annual averages 167 \n(temperature) and totals (precipitation). Climatic tolerance thresholds were defined by the 2.5th 168 \nand 97.5th percentiles to minimize errors from anomalous records caused by errors in 169 \ngeoreferencing or species identification.  170 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n12 \n2.1.3 Host preference 171 \nMosquitoes’ dispersal potential and habitat use are shaped by the availability of their preferred 172 \nblood meal hosts. Generalist, anthropophilic and mammal-feeding species often occupy urban or 173 \nperi-urban environments (35), facilitating unintentional human transport, whereas specialised 174 \nornithophilic species are typically associated with  forested or wetland settings (36). To include 175 \nthis in our model we obtained blood meal host data from Soghigian et al. (29), who compiled 176 \nmolecular blood meal analyses quantifying feeding proportions on mammals, birds, reptiles, and 177 \namphibians, providing a standardized host use metric across taxa. 178 \n 179 \n2.2 Imputation of missing data 180 \nDespite our efforts of data compilation, several species still had incomplete trait information. To 181 \nensure the reliability of model estimates, we only included species with at least 75% data 182 \ncompleteness and at least 3 unique occurrence points  in our modelling (n=1 69 out of 184). For 183 \nkept species, the data gaps were imputed using the ‘missForest’ algorithm (37), which iteratively 184 \npredicts missing values via random forest. To account for imputation uncertainty, we repeated the 185 \nimputation 100 times, producing 100 complete datasets. Imputation accuracy was evaluated using 186 \nout-of-bag (OOB) error estimates (normalized root mean square error (NRMSE) for continuous 187 \nvariables; proportion falsely classified (PFC) for categorical variables). 188 \n 189 \n2.3 Modeling framework and response variables 190 \nWe used a random forest (RF) modeling framework to estimate the probability of introduction and 191 \nestablishment of mosquitoes that transmit diseases to humans,  based on the fully impute d trait 192 \ndatasets. RF is a nonparametric ensemble learning method that handles continuous, categorical, 193 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n13 \nand partially redundant predictors with high robustness to overfitting (38,39). The algorithm builds 194 \nmultiple decision trees from bootstrap samples and aggregates their predictions, resulting in stable, 195 \nnonlinear models with strong predictive performance. RF models perform well with sparse or 196 \nnoisy ecological data and offer interpretable measures such as variable importance and partial 197 \ndependence plots (38,40). These strengths have led to their widespread application in ecological 198 \nand epidemiological studies (41–44). Our models require two sets of input components: A binary 199 \nresponse variable representing a species' invasion status, and a set of predictor variables describing 200 \nmosquito ecological, life -history, and macroecological traits. We developed four RF models to 201 \nassess the invasion potential of mosquito species. The first model estimated the probability of a 202 \nspecies being introduced outside its native range, and the second the probability of the species 203 \nbeing introduced after 1950. The third and fourth models evaluated the probability of species 204 \nestablishing themselves outside their native range (i.e., forming self -sustaining populations after 205 \ntheir introduction), considering either all records, or only those establishments occurring after 1950 206 \nrespectively. The year 1950 was used as a threshold because it marks the onset of the major 207 \nacceleration of biological invasions associated with the post -World War II globalization of trade 208 \n(45). In each model the response variable was binary, coded as “Yes” when species met the 209 \ncriterion and “No” otherwise. Introduction and establishment dates were derived from Pabst et al. 210 \n(5). Each of these four models was repeated independently 100 times, based on the 100 separate 211 \nimputation datasets, to generate ensemble predictions that capture prediction uncertainty. 212 \n 213 \n2.4 Variable selection 214 \nTo minimize redundancy and collinearity among predictors, pairwise Pearson correlations were 215 \ncalculated for numerical variables, removing one of each pair with |r| > 0.7 (46). Multicollinearity 216 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n14 \nwas further evaluated using the variance inflation factor (VIF) with the ‘vifstep’ function from the 217 \n‘usdm’ R package (47), keeping only variables with a VIF < 5 (48). As a result, we excluded the 218 \nvariables: Oviposition in human -made breeding sites and proportion of avian blood meal hosts. 219 \nFollowing this process, in each model we used the same non-redundant combination of continuous, 220 \nbinary, and categorical predictors. The correlation structure of the final numeric variables is shown 221 \nin Figure 1. In addition to models that included the full set of ecological, life -history, and 222 \nmacroecological traits, we also fitted models using only ecological and life -history variables to 223 \nassess the stability and consistency of species intrinsic traits as drivers of introduction and 224 \nestablishment. In the main manuscript, we present only the results of the full -variable models, as 225 \nthese performed better and reflect current real-world scenarios; results from the reduced variable 226 \nset are provided in the Appendix S2.  227 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n15 \n228 \nFigure 1. Pairwise Pearson correlation matrix of all numerical predictor variables kept after correlation analysis. 229 \n 230 \n2.5 Model fitting and cross-validation 231 \nFor each response variable, the 100 imputed datasets produced in section 2.2 were used to train 232 \n100 separate RF models having the same parameters, implemented in the ‘ ranger’ package (49). 233 \nFollowing standard practice, each model was built with 1,000 trees to ensure prediction stability. 234 \nEach tree used four randomly selected predictors at each split, following the common rule of using 235 \nthe rounded square root of the total number of predictor variables in classification models. A 236 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n16 \nminimum node size of five was specified to reduce overfitting and to ensure that each terminal 237 \nnode contained a sufficient number of observations for stable mean estimates (50). 238 \nModel performance was evaluated through leave -one-out cross -validation (LOOCV) (51), in 239 \nwhich each species was iteratively withheld from the training set, the model is fitted to the 240 \nremaining species, and predictions are generated for the excluded species. This was repeated until 241 \nevery species had been left out once, resulting in a full set of out -of-sample predictions. Model 242 \nperformance was evaluated by comparing predicted probabilities with observed outcomes across 243 \nall species, using the area under the receiver operating characteristic curve (AUC; 52) as a measure 244 \nof discrimination ability. This approach is robust and mimics a real -world situation where the 245 \ninvasion potential of a species is assessed based on what has been observed for other species. 246 \nWhile AUC is a robust, threshold -independent metric for model evaluation and comparison, it 247 \ndoes not indicate the probability cutoff at which classification performance is maximized. In our 248 \ncontext, identifying such a threshold is important because it allows us to identify which species 249 \nwere correctly or incorrectly classified by the models. To address this, we used the True Skill 250 \nStatistic (TSS; 53) , which provides the probability threshold that maximizes classification 251 \naccuracy by balancing sensitivity and specificity. The optimal threshold of each RF model was 252 \nidentified by maximizing TSS with the function ‘ecospat.max.tss()’ from the ‘ecospat’ package 253 \n(54). Predicted probabilities were then averaged across all 100 RF model repetitions to obtain 254 \nstable ensemble estimates of introduction and establishment probabilities. 255 \n 2.6 Variable importance and trait effects 256 \nFor each of the four response variables, we additionally ran 100 full RF models with all species to 257 \ncalculate contributions of the predictors using permutation -based importance scores, which 258 \nrepresent the decrease in model accuracy after shuffling each variable (38). These values were then 259 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n17 \naveraged to obtain stable estimates of variable importance at the ensemble -level. Variables that 260 \nshowed consistent influence (importance ≥ 0.01 across models) were further explored (50) using 261 \npartial dependency plots generated from the averaged ensemble predictions, using the 262 \nFeatureEffect$new() function in the ‘iml’ package  (55). 263 \n2.7 Identifying species with high invasion potential  264 \nUsing predictions from the LOOCV procedure, we identified species with high potential for future 265 \ninvasion. We first selected taxa not currently introduced but predicted to have a probability of 266 \nintroduction exceeding the TSS -defined threshold. These were considered as potential future 267 \nspreaders. We then cross-checked these species against predictions from the establishment model 268 \nand retained those exhibiting simultaneously high probabilities of introduction and establishment, 269 \nas species of highest invasion concern. All analyses were performed in R (56,57), with scripts and 270 \nreproducible code provided in the Appendix S3. 271 \nResults 272 \n3.1 Model performance  273 \nA comprehensive dataset was compiled with sufficient data for 169 mosquito species characterized 274 \nby 24 ecological, life -history and macroecological variables, including eight continuous, three 275 \ncategorical and thirteen one -hot encoded traits.  Missing values were imputed with low error, 276 \naveraging <0.01 % for continuous traits and 3.3 % for categorical traits.  Random forest models 277 \napplied to this dataset yielded cross-validated performances with AUC values indicating moderate 278 \n(0.78) to good (0.81 and 0.85) performance (Table 2). The introduction models exhibited higher 279 \nspecificity than sensitivity, suggesting stronger performance in identifying non‐introduced species, 280 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n18 \nwhile the establishment models displayed the opposite pattern, reflecting improved recognition of 281 \nspecies capable of successful establishment. 282 \nOverall, the establishment models slightly outperformed the introduction models, reflecting 283 \ngreater consistency in identifying traits linked to persistence after arrival. The models correctly 284 \nidentified 35.3 of 46 species known to be introduced beyond their native ranges and 2 4.1 of 29 285 \nestablished species. For the post -1950 subset, 2 9.8 of 41 introduced and 2 1.6 of 26 established 286 \nspecies were accurately predicted (Figure 2). 287 \nTable 2. LOOCV random forest model performance metrics (mean ± SD) based on 100 replicates predicting species 288 \nintroduction and establishment probabilities. 289 \nResponse  OOB error Threshold \ntss \nAccuracy Sensitivit\ny \nSpecificity Precision F1 AUC \nIntroduced 0.14±0 0.30±0.01 0.80±0.01 0.78±0.02 0.80±0.02 0.60±0.02 0.65±0.01 0.82±0 \nIntroduced after 1950 0.15±0 0.26±0.03 0.76±0.03 0.76±0.06 0.76±0.05 0.51±0.04 0.58±0.01 0.78±0 \nEstablished 0.11±0 0.17±0.01 0.77±0.01 0.85±0.02 0.75±0.02 0.42±0.01 0.56±0.01 0.85±0 \nEstablished after 1950 0.11±0 0.16±0.01 0.74±0.01 0.84±0.01 0.73±0.02 0.36±0.01 0.50±0.01 0.81±0 \n 290 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n19 \n 291 \nFigure 2. LOOCV random forest model performance. (a) Introduction model (AUC = 0. 82); (b) Introduced species 292 \nafter 1950 (AUC = 0.78); (c) Establishment model (AUC = 0.8 5); (d) Established species after 1950 (AUC = 0.8 1). 293 \nPanels show, from left to right, mean predicted probability distributions, mean ± SD confusion matrices, and ROC 294 \ncurves from 100 replicates with the average shown in black. 295 \n 296 \n3.2 Variable importance and partial dependency plots 297 \nPermutation importance analysis identified native range in Asia as the strongest predictors of 298 \nmosquito introduction risk, with mean decrease in accuracy of 0.02 7±0.001 (Figure 3A). 299 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n20 \nAdditional influential variables included oviposition in human-made breeding sites (0.018±0.001), 300 \nnative range in Australia (0.016±0.001), native range in North America (0.011±0.001),  maximum 301 \nand minimum precipitation within native ranges (0.015±0.001 and 0.00 3±0.001), thermal 302 \ntolerance limits (maximum and minimum temperatures at 0.0 10±0.001 and 0.01 1±0.001), and 303 \noverall native distribution area (0.015±0.001). 304 \nIn models restricted to species introduced since 1950, origin in Asia remained the predominant 305 \npredictor (0.021±0.001), along with increased importance of climatic variables such as maximum 306 \nprecipitation, maximum and minimum temperature (0.015±0.001, 0.011±0.001 and 0.011±0.001), 307 \nconsistent with tropical origin of recent introductions, originating in Australia (0.01 5±0.001) and 308 \noviposition in human-made breeding sites (0.011±0.001), (Figure 3B).  309 \nFor establishment probability in both models, maximum precipitation emerged as the top predictor 310 \n(0.023±0.001 and 0.02 1±0.001), followed by Asian origin (0.020±0.001 and 0.014±0.001) 311 \n(Figures 3C-D). Results based only on ecological and life -history traits were similar in terms of 312 \nidentified and ranking of important variables and the predicted probabilities for the introduction 313 \nand establishment of species. Major changes were general lower model performance and the 314 \nemergence of several species predicted as potential spreaders that were not identified by the full 315 \nmodels based on all variables (Appendix S2). 316 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n21 \n 317 \nFigure 3. Variable importance rankings from the full random forest models, shown as mean decrease in accuracy 318 \n(points) ± standard deviation, across 100 model replications. Higher values indicate greater influence in predicting (a) 319 \nintroduction, (b) introduction after 1950, (c) establishment, and (d) establishment after 1950. 320 \n 321 \nIntroduction probabilities increased for species native to Asia and Australia, those using 322 \nhuman-made breeding sites, those from regions with higher maximum precipitation in their native 323 \nrange and those with wider distribution in original range.  The probability decreased for species 324 \nnative to North America. The relationship with minimum temperature was bimodal, with elevated 325 \nprobabilities for species from both cold and warm minimum temperature regimes, whereas species 326 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n22 \nfrom intermediate temperature ranges were less likely to be introduced (Figure 4A). Although for 327 \nspecies introduced after 1950, the maximum temperature of the regions of origin is more important 328 \nthan the distribution area, the patterns remain largely consistent. (Figure 4B). 329 \nEstablishment probability is largely driven by maximum precipitation in their native range 330 \nand Asian origin (Figures 4C-D).  331 \n 332 \nFigure 4. Partial dependence plots illustrating the marginal effects of predictors with importance ≥ 0.01 on predicted 333 \nprobabilities, averaged (green) across 100 random forest model replications (grey). Panels show (a) introduction, (b) 334 \nintroduction after 1950, (c) establishment, and (d) establishment after 1950. 335 \n 336 \n3.3 Predicted spreaders and discrepancies  337 \nThe introduction model identified 2 4 mosquito species without known invasion history to have 338 \ntraits consistent with unintentionally transported and introduced species (Table 3). These species 339 \nshare ecological profiles of known invaders, including human altered environments, geographic 340 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n23 \norigin, and climatic characterization of their native range, underscoring their potential as emerging 341 \ninvaders warranting targeted surveillance.  342 \nAmong them 1 7 “high-risk” species have attributes consistent with species that also 343 \nestablished non-native populations. These include Culex vishnui, Anopheles culicifacies, Culex 344 \nunivittatus, Aedes lineatopennis, Anopheles amictus, Culex theileri, Mansonia septempunctata, 345 \nCulex torrentium, Anopheles hyrcanus, Culex perexiguus, Culex poicilipes, Anopheles claviger, 346 \nCulex nigripalpus, Anopheles plumbeus, Culiseta inornata, Aedes geniculatus, and Anopheles 347 \npseudopunctipennis. 348 \n  On the other hand, seven species that have attributes consistent with accidentally 349 \nintroduced species do not share characteristics with species that were established. Such species 350 \nmay survive transport but fail to overcome ecological or climatic constraints necessary for 351 \nsustained establishment. These species include Anopheles rufipes, Anopheles fluviatilis, 352 \nCoquillettidia linealis, Coquillettidia richiardii, Anopheles sergentii, Anopheles pulcherrimus, and 353 \nAnopheles quadrimaculatus.  354 \nEleven species with known invasion history were not captured by our introduction model 355 \n(predicted probabilities < 0.3). The same applies to five species in the establishment model. Most 356 \nof these species exhibited transient detection or failure to establish self -sustaining populations, 357 \nsuggesting stochastic introduction events or influences from drivers beyond trait-based predictors, 358 \nsuch as chance transport or climatic anomalies.  359 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n24 \nTable 3. Predicted potential spreaders. Mosquito species with a model -predicted introduction probability above the 360 \naverage TSS threshold (≥ 0.3) but no confirmed introduction outside their native range. Species in bold also show 361 \nhigh predicted establishment probabilities, indicating elevated overall invasion risk. 362 \nSpecies True Label Predicted Probability ± SD Predicted Label \nCulex vishnui No 0.89 ± 0.01 Yes \nAnopheles culicifacies No 0.70 ± 0.01 Yes \nAedes lineatopennis No 0.57 ± 0.01 Yes \nCulex univittatus No 0.57 ± 0.02 Yes \nAnopheles amictus No 0.56 ± 0.01 Yes \nCulex theileri No 0.55 ± 0.01 Yes \nMansonia septempunctata No 0.50 ± 0.01 Yes \nCulex torrentium No 0.46 ± 0.01 Yes \nAnopheles hyrcanus No 0.45 ± 0.01 Yes \nCulex perexiguus No 0.45 ± 0.02 Yes \nCulex poicilipes No 0.45 ± 0.01 Yes \nAnopheles rufipes No 0.43 ± 0.02 Yes \nAnopheles claviger No 0.42 ± 0.01 Yes \nCulex nigripalpus No 0.42 ± 0.02 Yes \nAnopheles plumbeus No 0.42 ± 0.01 Yes \nCuliseta inornata No 0.40 ± 0.01 Yes \nAnopheles fluviatilis No 0.40 ± 0.01 Yes \nAedes geniculatus No 0.40 ± 0.01 Yes \nAnopheles pseudopunctipennis No 0.37 ± 0.01 Yes \nCoquillettidia linealis No 0.36 ± 0.01 Yes \nCoquillettidia richiardii No 0.32 ± 0.01 Yes \nAnopheles sergentii No 0.32 ± 0.01 Yes \nAnopheles pulcherrimus No 0.31 ± 0.01 Yes \nAnopheles quadrimaculatus No 0.31 ± 0.01 Yes \n 363 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n25 \n3.4 Occurrence of possible spreaders  364 \nBelow we show the current distributions of species that have been identified as high -risk species 365 \nobtained from Wilkerson et al. (26), meaning species not yet detected outside their native ranges 366 \nwith high probability of introduction and establishment (Figure 5).  367 \n 368 \n 369 \nFigure 5. Current distribution of species with high invasion potential, meaning high probability to be introduced as 370 \nwell as to become established. Green shading shows countries where the species are currently reported (26).  371 \n 372 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n26 \nDiscussion 373 \nOur findings suggest that the introduction and establishment of non-native mosquitoes are driven, 374 \nto some extent, by some ecological, life -history, and macroecological characteristics. Once 375 \nestablished, vector mosquitoes can profoundly alter disease transmission dynamics, or introduce 376 \npathogens into previously unaffected regions (3). Responding to their global spread requires 377 \npredictive tools that go beyond local surveillance and include proactive prevention and risk 378 \nassessment strategies. Our modelling approach achieved moderate to good predictive ability, stable 379 \nacross model repetitions, which indicates that the invasion potential of species can be inferred to 380 \nsome extent from trait data alone. Unlike previous efforts that modelled the potential distribution 381 \nof species already introduced (58,59), our framework indicates potential invaders before they 382 \nspread. Specifically, of the 169 species analyzed, 24 species with no prior invasion history received 383 \nintroduction probabilities equal to or higher than for known introduced species, including 1 7 384 \nspecies with a high probability of  also becoming established. These species are predominantly 385 \nnative to Asia, originated in regions with high precipitation, tend to have broad climatic tolerances, 386 \nhave a wide distribution and are adapted to human-modified environments. 387 \nA main result of our analysis is the consistent importance of native biogeographic origin 388 \nin shaping invasion potential. Species native to Asia and Australia consistently ranked as the most 389 \nlikely to be introduced and to establish outside their native ranges, whereas species from Africa, 390 \nthe Americas or Europe showed substantial lower importance and probabilities. This pattern 391 \nreflects Asia’s role as principal source region for invasive species overall (60), including invasive 392 \ninsects (61), and recently introduced mosquitoes (5). In Asia, high human population density, 393 \nintensive containerized trade, and rapid economic expansion create both the propagule pressure 394 \nand the disturbed habitats that favor human-commensal vectors. Following World War II, used 395 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n27 \naircraft tires were shipped from Asia and the Pacific region back to the United States as part of 396 \npostwar recovery and logistics operations because rubber remained a valuable commodity (62), 397 \ninadvertently facilitating the long-distance transport of mosquito eggs and larvae from that region. 398 \nToday, the globalization of plant and materials trade continues to provide suitable pathways. For 399 \ninstance, China is by far the leading bamboo exporter (63), and ornamental -plant producers 400 \nincreasingly outsource cultivation to tropical developing countries such as Thailand to reduce 401 \nlabor, land, and infrastructure costs (64). According to the World Bank, China alone accounted for 402 \nroughly 32% of global container port traffic in 2022 (65). Ongoing large-scale initiatives such as 403 \nChina’s Belt and Road Initiative (BRI) further expand trade and transport networks involving more 404 \nthan 120 countries (66), and the thawing of the Northern Sea Route through the Arctic Ocean (67), 405 \npotentially opens opportunities for the spread of non-native mosquito species from Asia to Europe. 406 \nAustralia's significance arises mainly from the fact that most of its native mosquito species were 407 \nintroduced to the Pacific region and New Zealand. Yet, there are several examples of long-distance 408 \nintroductions of mosquitoes native to Australia, including Aedes vexans in Hawaii (68), Anopheles 409 \nsubpictus in the Netherlands (69), and Aedes notoscriptus  in California (70). The latter, first 410 \ndetected in Los Angeles County in 2014, shows how integration into global trade networks, 411 \ncombined with climatic similarities between native and introduced regions, can lower 412 \nenvironmental barriers and facilitate rapid establishment. In that sense, native continent acts as an 413 \nintegrative factor, capturing evolutionary history, trade exportation intensity, and environmental 414 \nmatching simultaneously. However, global trade networks continuously evolve, potentially 415 \nreshaping introduction pathways and altering the relative importance of certain origins, routes, and 416 \ngoods (17,45). Nevertheless, our results suggest that the intrinsic traits of a species remain 417 \nrelatively stable predictors, independent of macroecological factors (Appendix S2). 418 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n28 \nConcerning probability of species introduction, another main predictor identified is the use 419 \nof human-made breeding sites. Species that oviposit in artificial habitats, plastic vessels, plant pots, 420 \nrice fields, drainage systems or discarded tires, showed markedly higher probabilities of being 421 \ntransported and introduced than species restricted to natural water bodies. This finding aligns with 422 \nclassic invasion ecology, in which propagule pressure and human association are primary 423 \ndeterminants of transport success (15,71). Container breeding has the advantage that eggs and 424 \nlarvae persist in environments closely associated with human activity, trade and travel, from cargo 425 \nholds to used -tire shipments (72). This pattern, documented e.g., for Ae. albopictus  and Ae. 426 \njaponicus (10,72,73), is generalizable: nearly all species predicted with high probabilities of both 427 \nintroduction and establishment use human -made breeding sites. This underscores that invasion 428 \npotential is closely coupled with the degree of adaptation to human -modified landscapes. Global 429 \nurban expansion and the proliferation of disposable containers continue to multiply these breeding 430 \nsites, providing abundant breeding opportunities for species with suitable ecological strategies 431 \n(74,75). 432 \nClimatic predictors had a complementary influence, particularly on establishment success. 433 \nThe maximum annual precipitation of the native range was by far the strongest predictor of 434 \nprobability of species establishment, with probabilities rising sharply for species native from areas 435 \nreceiving approximately 1,500 mm yr⁻¹. The relationship between minimum temperature and 436 \nintroduction probability was bimodal: species from both cold and warm native ranges were more 437 \nlikely to be introduced than those from mild climates. This dual pattern might just again stand as 438 \na proxy for the species native region, or it suggests that both cold-adapted and tropical mosquitoes 439 \npossess distinct mechanisms enabling survival during transport, either tolerance to cold or 440 \nresistance to dehydration. Species from thermally extreme regions on the other hand were more 441 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n29 \nlikely to establish, with over half of the successfully established species originating from areas 442 \nexceeding 34 °C in average maximum temperature. Again, the climate data used here reflect the 443 \nconditions within  the species’ native ranges . H owever, there are several species were able to 444 \nestablish under environmental conditions that differ from those in their native ranges. This applies, 445 \nfor example, to Ae. albopictus, which was initially thought unlikely to establish beyond its native 446 \nrange in tropical and subtropical regions of Southeast Asia, requiring a lengthy adaptation to new 447 \necological conditions and to be constrained by competition with local mosquito species  (76). 448 \nNevertheless, it rapidly adapted to colder conditions (77,78), outperformed presumed competitors 449 \n(79,80) and expanded into temperate regions worldwide. Hence, predictions of invasion and 450 \nestablishment potential based solely on native-range climates should be interpreted with caution. 451 \nFinally, we also found that widely distributed species are also more likely to be introduced 452 \ninto new regions. This may translate into increased propagule pressure as greater distribution 453 \nincreases the chances of introduction. This was however not observed for established species, 454 \nwhich could indicate a filter for species establishment that is based more on intrinsic traits or 455 \nadaptation to new environments (20,81). 456 \nOverall, our results indicate that mosquito invasion potential is best explained by the 457 \ninteraction between life -history flexibility and macroecological breadth. Adaptation to human 458 \nenvironments and occurrence in regions that are export hubs for commodities associated with 459 \nhuman-made oviposition sites  that promote dispersal and introduction. In turn, broad climatic 460 \ntolerance, particularly to extremes of rainfall and temperature, determines whether colonization 461 \nresults in successful establishment. Together, these traits define a functional profile of likely 462 \ninvaders: human -adapted species, often native to Asia and/or Australia, capable of exploiting 463 \nhuman-made breeding sites , tolerant of precipitation and thermal extremes  and with a wide 464 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n30 \ndistribution. These findings extend previous qualitative insights  (18) to a quantitative global 465 \nframework, go beyond the few medically very important vectors (e.g., 59,72,82,83), and provide 466 \ninsights into the invasi ve potential of lesser -known species. The  results also demonstrate that 467 \ninvasion potential in mosquitoes can be forecasted to some extent from measurable ecological, 468 \nlife-history, and macroecological traits. 469 \nBeyond identifying traits of importance, our models identified 24 species with no recorded 470 \ninvasion history but trait profiles closely matching those of known invaders. Of these, 1 7 species 471 \nwere also predicted with high probabilities of establishment, marking them as priority candidates 472 \nfor early surveillance. For example, Culex vishnui,  the species with the highest predicted 473 \nintroduction probability and likely to establish is native to South and East Asia, where it thrives in 474 \nrice fields, ground pools, and small artificial containers (26). It has been found naturally infected 475 \nwith multiple arboviruses, including Japanese encephalitis and West Nile virus (84,85). 476 \nAnopheles culicifacies, ranking second in our high risk species list, is an important malaria 477 \nvector, particularly in South and Southwest Asia. (86). Predominantly anthropophilic but 478 \noccasionally zoophilic, it can take multiple blood meals per gonotrophic cycle (26) increasing its 479 \npotential for pathogen transmission. 480 \nAedes lineatopennis, native to the Australasian and Oriental regions, was first described 481 \nfrom the Philippines (26,87). It has been found naturally infected with Japanese encephalitis  (88) 482 \nand Middelburg virus (89), and is considered a potential vector for Ross River and Murray Valley 483 \nencephalitis viruses (90,91).  484 \nOur results provide valuable insights, but their interpretation should also take certain 485 \nconsiderations into account. First, trait data for some mosquitoes remains partly incomplete. 486 \nSimilar to other studies (21,22,92), we used a trait imputation approach to overcome this limitation. 487 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n31 \nTo minimize bias, we included only species with at least 75% trait completeness, and only 488 \nvariables that were known for at least 75% of the species. Nonetheless, the imputation process may 489 \nobscure subtle ecological differences that influence invasion potential. Second, because our 490 \napproach is species trait–centric, it does not account for the local conditions species encounter at 491 \nintroduction sites, such as climate, biotic interactions, or vector control measures that can influence 492 \ninvasion outcomes  (93,94). Furthermore, environmental change can modify the current 493 \ndistribution of suitable habitats, impairing species establishment success. Thus, our framework 494 \nshould be regarded as an initial profiling tool for species with invasion potential, to be 495 \ncomplemented by geographically explicit assessments of establishment risk. Finally, it is important 496 \nto recognize that the predictors used in our models are not static. Global trade networks may 497 \ncontinue to evolve and other regions of origin may become sources for species exports. However, 498 \nresults based solely on ecological and life -history traits gave similar predicted probabilities of 499 \nspecies introduction and establishment showing the resilience of our findings  (Appendix S2) .  500 \nThus, although our results may be temporally bounded, our framework provides a robust baseline 501 \nfor identifying species with elevated invasion potential under present -day ecological and trade 502 \nconditions and are most relevant for near-term invasion assessments. 503 \n5 Conclusion 504 \nOverall, our results indicate that the invasion potential of mosquitoes can be partially predicted 505 \nfrom intrinsic biological and ecological traits alone. Species displaying traits like being native to 506 \nAsia and Australia, using human-made breeding sites, from regions with climate extremes can be 507 \nexpected to be those posing higher risk of invasion in non -native regions in the near future. By 508 \nidentifying these species and their characteristic profiles, our approach underscores the potential 509 \nfor proactive, trait-based surveillance strategies that extend beyond the currently recognized vector 510 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n32 \nspecies. Trait -based frameworks such as the one presented here can support early -warning 511 \nsystems, guide allocation of surveillance resources, and ultimately reduce the risk of novel vector-512 \nborne disease emergence. Once invasive mosquitoes succeed in establishing populations, their 513 \ncontrol and eradication become exceedingly challenging (95,96). In this context, preventing 514 \nintroductions remains the most efficient and economically viable strategy. 515 \nAcknowledgments 516 \nRP and CAS gratefully acknowledge the support of the Portuguese Foundation for Science and 517 \nTechnology (FCT) for funds to the R&D Unit Global Health and Tropical Medicine 518 \n(UIDB/04413/2025) and the Associated Laboratory in Translation and Innovation Towards Global 519 \nHealth REAL (LA/P/0117/2020). RP acknowledges funding from FCT (PRT/BD/153694/2021;  520 \nhttps://doi.org/10.54499/PRT/BD/153694/2021) and thanks the AIR Center for their support. CC 521 \nacknowledges funding from FCT through InvaSTOP grant 522 \n(https://doi.org/10.54499/2023.12533.PEX) and the support from FCT through funds to 523 \nCEG/IGOT Research Unit (https://doi.org/10.54499/UID/00295/2025).  524 \nReferences 525 \n1. WHO. Vector-borne diseases. World Health Organization. 2024 [cited 2025 Sept 3]; Available 526 \nfrom: https://www.who.int/news-room/fact-sheets/detail/vector-borne-diseases 527 \n2. Kraemer MU, Sinka ME, Duda KA, Mylne AQ, Shearer FM, Barker CM, et al. The global 528 \ndistribution of the arbovirus vectors Aedes aegypti and Ae. albopictus. Jit M, editor. eLife. 529 \n2015 June 30;4:e08347.  530 \n3. Farooq Z, Segelmark L, Rocklöv J, Lillepold K, Sewe MO, Briet OJT, et al. Impact of climate 531 \nand Aedes albopictus establishment on dengue and chikungunya outbreaks in Europe: a time-532 \nto-event analysis. The Lancet Planetary Health. 2025 May 1;9(5):e374–83.  533 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n33 \n4. Yee DA, Dean Bermond C, Reyes -Torres LJ, Fijman NS, Scavo NA, Nelsen J, et al. Robust 534 \nnetwork stability of mosquitoes and human pathogens of medical importance. Parasites 535 \nVectors. 2022;15(216):1–9.  536 \n5. Pabst R, Sousa CA, Essl F, García -Rodríguez A, Liu D, Lenzner B, et al. Global invasion 537 \npatterns and dynamics of disease vector mosquitoes. Nature Communications. 2025;  538 \n6. Weaver SC, Reisen WK. Present and future arboviral threats. Antiviral Research. 2010 539 \nFeb;85(2):328–45.  540 \n7. Ryan SJ, Carlson CJ, Mordecai EA, Johnson LR. Global expansion and redistribution of 541 \nAedes-borne virus transmission risk with climate change. Han BA, editor. PLoS Negl Trop 542 \nDis. 2019 Mar 28;13(3):e0007213.  543 \n8. Powell JR, Gloria-Soria A, Kotsakiozi P. Recent History of Aedes aegypti : Vector Genomics 544 \nand Epidemiology Records. BioScience. 2018 Nov 1;68(11):854–60.  545 \n9. Soto I, Balzani P, Carneiro L, Cuthbert RN, Macêdo R, Serhan Tarkan A, et al. Taming the 546 \nterminological tempest in invasion science. Biological Reviews. 2024;99(4):1357–90.  547 \n10. Peyton EL, Campbell SR, Candeletti TM, Romanowski M, Crans WJ. Aedes (Finlaya) 548 \nJaponicus Japonicus (Theobald), A New Introduction into the United States. Journal of the 549 \nAmerican Mosquito Control Association. 1999;15(2):238–41.  550 \n11. Swan T, Russell TL, Staunton KM, Field MA, Ritchie SA, Burkot TR. A literature review of 551 \ndispersal pathways of Aedes albopictus across different spatial scales: implications for vector 552 \nsurveillance. Parasites & Vectors. 2022 Aug 27;15(303):1–13.  553 \n12. WHO. Vector alert: Anopheles stephensi  invasion and spread in Africa and Sri Lanka 554 \n[Internet]. 2022 [cited 2023 Sept 13]. Available from: https://www.who.int/publications -555 \ndetail-redirect/9789240067714 556 \n13. Vezzani D. Review: Artificial container-breeding mosquitoes and cemeteries: a perfect match. 557 \nTropical Medicine & International Health. 2007;12(2):299–313.  558 \n14. Bebber DP, Field E, Gui H, Mortimer P, Holmes T, Gurr SJ. Many unreported crop pests and 559 \npathogens are probably already present. Glob Chang Biol. 2019 Aug;25(8):2703–13.  560 \n15. Catford JA, Jansson R, Nilsson C. Reducing redundancy in invasion ecology by integrating 561 \nhypotheses into a single theoretical framework. Diversity and Distributions. 2009 562 \nJan;15(1):22–40.  563 \n16. Becker N, Pluskota B, Kaiser A, Schaffner F. Exotic Mosquitoes Conquer the World. In: 564 \nMehlhorn H, editor. Arthropods as Vectors of Emerging Diseases [Internet]. Berlin, 565 \nHeidelberg: Springer; 2012. p. 31 –60. Available from: https://doi.org/10.1007/978 -3-642-566 \n28842-5_2 567 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n34 \n17. Capinha C, Essl F, Porto M, Seebens H. The worldwide networks of spread of recorded alien 568 \nspecies. Proc Natl Acad Sci USA. 2023 Jan 3;120(1):1–10.  569 \n18. Juliano SA, Lounibos LP. Ecology of invasive mosquitoes: effects on resident species and on 570 \nhuman health: Invasive mosquitoes. Ecology Letters. 2005;8(5):558–74.  571 \n19. Kengne P, Charmantier G, Blondeau -Bidet E, Costantini C, Ayala D. Tolerance of disease -572 \nvector mosquitoes to brackish water and their osmoregulatory ability. Ecosphere. 573 \n2019;10(10):e02783.  574 \n20. Renault D, Laparie M, McCauley SJ, Bonte D. Environmental Adaptations, Ecological 575 \nFiltering, and Dispersal Central to Insect Invasions. Annu Rev Entomol. 2018 Jan 576 \n7;63(1):345–68.  577 \n21. Fournier A, Penone C, Pennino MG, Courchamp F. Predicting future invaders and future 578 \ninvasions. Proc Natl Acad Sci USA. 2019 Apr 16;116(16):7905–10.  579 \n22. Pili AN, Leroy B, Measey JG, Farquhar JE, Toomes A, Cassey P, et al. Forecasting potential 580 \ninvaders to prevent future biological invasions worldwide. Global Change Biology. 581 \n2024;30(7):e17399.  582 \n23. Biancolini D, Rondinini C. Global Enhancers and Constraints of Alien Range Size in 583 \nMammals: The Roles of Species Attributes, Invasion History and Ecological Contexts. Global 584 \nEcol Biogeogr [Internet]. 2025 July [cited 2025 July 15];34(7). Available from: 585 \nhttps://onlinelibrary.wiley.com/doi/10.1111/geb.70081 586 \n24. Johnson LR, Cator L, Rund SSC, Ryan S, Huxley PJ, Pawar S. VecTraits Explorer [Internet]. 587 \nUniversity of  Notre Dame; 2023 [cited 2025 Oct 21]. Available from: 588 \nhttps://vectorbyte.crc.nd.edu/vectraits-explorer 589 \n25. Da Re D, Andreo V, San Miguel T, Blaha M, Rosà R, Rizzoli A, et al. AedesTraits: A global 590 \ndatabase of temperature–dependent trait responses in Aedes mosquitoes [Internet]. Biology; 591 \n2025 [cited 2025 Oct 21]. Available from: https://ecoevorxiv.org/repository/view/8919/ 592 \n26. Wilkerson RC, Linton YM, Strickman D. Mosquitoes of the World [Internet]. 1st edition. 593 \nBaltimore, Maryland: Johns Hopkins University Press; 2021. 1332 p. Available from: 594 \nhttps://doi.org/10.1353/book.79680 595 \n27. Karger DN, Conrad O, Böhner J, Kawohl T, Kreft H, Soria -Auza RW, et al. Data from: 596 \nClimatologies at high resolution for the earth’s land surface areas. Version 1.2. Dryad Digital 597 \nRepository. 2017;  598 \n28. Karger DN, Conrad O, Böhner J, Kawohl T, Kreft H, Soria-Auza RW, et al. Climatologies at 599 \nhigh resolution for the earth’s land surface areasCHELSA V2.1 (current) [Internet]. EnviDat; 600 \n2021 [cited 2025 Sept 8]. p. 2.1 KB. Available from: 601 \nhttps://www.envidat.ch/#/metadata/chelsa-climatologies 602 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n35 \n29. Soghigian J, Sither C, Justi S, Morinaga G, Cassel B, Vitek C, et al. Phylogenomics reveals 603 \nthe history of host use in mosquitoes. Nature Communications. 2023 Oct 6;14.  604 \n30. WRBU. (Walter Reed Biosystematics Unit), VectorMap Data Portal. VectorMap website, 605 \nhttps://vectormap.si.edu/. 2024; Available from: 17.12.2024 606 \n31. Ciota AT, Matacchiero AC, Kilpatrick AM, Kramer LD. The Effect of Temperature on Life 607 \nHistory Traits of Culex Mosquitoes. J Med Entomol. 2014 Jan 1;51(1):55–62.  608 \n32. Barr JS, Estevez-Lao TY, Khalif M, Saksena S, Yarlagadda S, Farah O, et al. Temperature and 609 \nage, individually and interactively, shape the size, weight, and body composition of adult 610 \nfemale mosquitoes. Journal of Insect Physiology. 2023 July 1;148:104525.  611 \n33. Newman EA, Feng X, Onland JD, Walker KR, Young S, Smith K, et al. Defining the roles of 612 \nlocal precipitation and anthropogenic water sources in driving the abundance of Aedes aegypti, 613 \nan emerging disease vector in urban, arid landscapes. Sci Rep. 2024 Jan 24;14(1):2058.  614 \n34. GBIF. Global Biodiversity Information Facility - Free and open access to biodiversity data 615 \n[Internet]. 2025 [cited 2025 Apr 17]. Available from: https://www.gbif.org/ 616 \n35. Wilke ABB, Wilk-da-Silva R, Marrelli MT. Microgeographic population structuring of Aedes 617 \naegypti (Diptera: Culicidae). Sekaran SD, editor. PLoS ONE. 2017 Sept 20;12(9):e0185150.  618 \n36. Chathuranga W g. d., Karunaratne S h. p. p., Fernando B r., De Silva WAPP. Diversity, 619 \ndistribution, abundance, and feeding pattern of tropical ornithophilic mosquitoes. Journal of 620 \nVector Ecology. 2018;43(1):158–67.  621 \n37. Stekhoven DJ, Bühlmann P. MissForest—non-parametric missing value imputation for mixed-622 \ntype data. Bioinformatics. 2012 Jan 1;28(1):112–8.  623 \n38. Breiman L. Random Forests. Machine Learning. 2001 Oct 1;45(1):5–32.  624 \n39. Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning [Internet]. New York, 625 \nNY: Springer New York; 2009. (Springer Series in Statistics). Available from: 626 \nhttps://doi.org/10.1007/978-0-387-84858-7 627 \n40. Freeman EA, Moisen GG, Coulston JW, Wilson BT. Random forests and stochastic gradient 628 \nboosting for predicting tree canopy cover: comparing tuning processes and model 629 \nperformance. Can J For Res. 2016 Mar;46(3):323–39.  630 \n41. Cutler DR, Edwards Jr. TC, Beard KH, Cutler A, Hess KT, Gibson J, et al. Random Forests 631 \nfor Classification in Ecology. Ecology. 2007;88(11):2783–92.  632 \n42. Pless E, Saarman NP, Powell JR, Caccone A, Amatulli G. A machine -learning approach to 633 \nmap landscape connectivity in Aedes aegypti with genetic and environmental data. Proc Natl 634 \nAcad Sci USA. 2021 Mar 2;118(9):e2003201118.  635 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n36 \n43. Arora AK, Sim C, Severson DW, Kang DS. Random Forest Analysis of Impact of Abiotic 636 \nFactors on Culex pipiens and Culex quinquefasciatus Occurrence. Front Ecol Evol. 2022 Jan 637 \n27;9.  638 \n44. Lippi CA, Mundis SJ, Sippy R, Flenniken JM, Chaudhary A, Hecht G, et al. Trends in 639 \nmosquito species distribution modeling: insights for vector surveillance and disease control. 640 \nParasites Vectors. 2023 Aug 28;16(1):302.  641 \n45. Seebens H, Essl F, Dawson W, Fuentes N, Moser D, Pergl J, et al. Global trade will accelerate 642 \nplant invasions in emerging economies under climate change. Global Change Biology. 643 \n2015;21(11):4128–40.  644 \n46. Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Carré G, et al. Collinearity: a review of 645 \nmethods to deal with it and a simulation study evaluating their performance. Ecography. 646 \n2013;36(1):27–46.  647 \n47. Naimi B. usdm: Uncertainty Analysis for Species Distribution Models [Internet]. 2023 [cited 648 \n2024 July 10]. Available from: https://cran.r-project.org/web/packages/usdm/index.html 649 \n48. Neter J, Wasserman W, Kutner MH. Applied Linear Regression Models. Richard D. Irwin, 650 \nInc.; 1983.  651 \n49. Wright MN, Ziegler A. ranger: A Fast Implementation of Random Forests for High 652 \nDimensional Data in C++ and R. Journal of Statistical Software. 2017 Mar 31;77:1–17.  653 \n50. Genuer R, Poggi JM. Random Forests with R [Internet]. Cham: Springer International 654 \nPublishing; 2020 [cited 2025 Oct 16]. (Use R!). Available from: 655 \nhttp://link.springer.com/10.1007/978-3-030-56485-8 656 \n51. Wenger SJ, Olden JD. Assessing transferability of ecological models: an underappreciated 657 \naspect of statistical validation. Methods in Ecology and Evolution. 2012;3(2):260–7.  658 \n52. Fielding AH, Bell JF. A review of methods for the assessment of prediction errors in 659 \nconservation presence/absence models. Envir Conserv. 1997 Mar;24(1):38–49.  660 \n53. Allouche O, Tsoar A, Kadmon R. Assessing the accuracy of species distribution models: 661 \nprevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology. 662 \n2006;43(6):1223–32.  663 \n54. Broennimann O, Di Cola V, Guisan A. ecospat: Spatial Ecology Miscellaneous Methods 664 \n[Internet]. The R Foundation; 2014 [cited 2025 July 17]. (CRAN: Contributed Packages). 665 \nAvailable from: https://CRAN.R-project.org/package=ecospat 666 \n55. Casalicchio G, Molnar C, Schratz P. iml: Interpretable Machine Learning [Internet]. The R 667 \nFoundation; 2018 [cited 2025 July 29]. (CRAN: Contributed Packages). Available from: 668 \nhttps://CRAN.R-project.org/package=iml 669 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n37 \n56. Posit team. RStudio: Integrated Development Environment for R. Posit Software, PBC,   670 \nBoston, MA [Internet]. 2023; Available from: http://www.posit.co/ 671 \n57. R Core Team. R: A Language and Environment for Statistical Computing [Internet]. Vienna, 672 \nAustria: R Foundation for Statistical Computing:; 2023 [cited 2023 Oct 31]. Available from: 673 \nhttp://www.R-project.org/ 674 \n58. Früh L, Kampen H, Kerkow A, Schaub GA, Walther D, Wieland R. Modelling the potential 675 \ndistribution of an invasive mosquito species: comparative evaluation of four machine learning 676 \nmethods and their combinations. Ecological Modelling. 2018 Nov 24;388:136–44.  677 \n59. Kraemer MUG, Reiner RC, Brady OJ, Messina JP, Gilbert M, Pigott DM, et al. Past and future 678 \nspread of the arbovirus vectors Aedes aegypti  and Ae. albopictus . Nat Microbiol. 2019 679 \nMay;4(5):854–63.  680 \n60. Turbelin AJ, Malamud BD, Francis RA. Mapping the global state of invasive alien species: 681 \npatterns of invasion and policy responses. Global Ecology and Biogeography. 2017;26(1):78–682 \n92.  683 \n61. Roques A, Shi J, Auger-Rozenberg MA, Ren L, Augustin S, Luo Y qing. Are Invasive Patterns 684 \nof Non-native Insects Related to Woody Plants Differing Between Europe and China? Front 685 \nFor Glob Change. 2020 Jan 15;2.  686 \n62. Pratt Jr. JJ, Hexerick RH, Harrison JB, Haber L. Tires as a Factor in the Transportation of 687 \nMosquitoes by Ships. Military Surgeon. 1946;99(6).  688 \n63. World Bank. WITS—World Integrated Trade Solution. 2019 [cited 2025 Oct 30]. Bamboos 689 \nexports by country in 2019. Available from: 690 \nhttps://wits.worldbank.org/trade/comtrade/en/country/ALL/year/2019/tradeflow/Exports/part691 \nner/WLD/product/140110 692 \n64. Hongpakdee P, Suzuki N. Current Status: The Global and Thai Flower and Ornamental Plant 693 \nProduction and Opportunities for Developing the Ornamental Plant Industry in the Lao PDR. 694 \nThe Formation of Cooperative Network for bottom -up Approach in Rural Community 695 \nDevelopment of Lao PDR. 2025;  696 \n65. The World Bank. Container port traffic (TEU: 20-foot equivalent units) (IS.SHP.GOOD.TU). 697 \nWorld Development Indicators [Internet]. 2022 [cited 2025 Oct 29];UN Conference on Trade 698 \nand Development (UNCTAD). Available from: 699 \nhttps://data.worldbank.org/indicator/IS.SHP.GOOD.TU 700 \n66. Liu X, Blackburn TM, Song T, Li X, Huang C, Li Y. Risks of Biological Invasion on the Belt 701 \nand Road. Current Biology. 2019 Feb;29(3):499-505.e4.  702 \n67. Zhao H, Hu H, Lin Y. Study on China -EU container shipping network in the context of 703 \nNorthern Sea Route. Journal of Transport Geography. 2016 May;53:50–60.  704 \n68. Joyce CR, Nakagawa PY. Aedes vexans nocturnus (Theobald) in Hawaii. 1963;(2).  705 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n38 \n69. Ibáñez-Justicia A, Smitz N, den Hartog W, van de Vossenberg B, De Wolf K, Deblauwe I, et 706 \nal. Detection of Exotic Mosquito Species (Diptera: Culicidae) at International Airports in 707 \nEurope. International Journal of Environmental Research and Public Health. 2020 708 \nJan;17(3450):1–19.  709 \n70. Metzger ME, Wekesa JW, Kluh S, Fujioka KK, Saviskas R, Arugay A, et al. Detection and 710 \nEstablishment of Aedes notoscriptus (Diptera: Culicidae) Mosquitoes in Southern California, 711 \nUnited States. J Med Entomol. 2022 Jan;59(1):67–77.  712 \n71. Lockwood JL, Cassey P, Blackburn T. The role of propagule pressure in explaining species 713 \ninvasions. Trends in Ecology & Evolution. 2005 May;20(5):223–8.  714 \n72. Benedict MQ, Levine RS, Hawley WA, Lounibos LP. Spread of The Tiger: Global Risk of 715 \nInvasion by The Mosquito Aedes albopictus . Vector -Borne and Zoonotic Diseases. 2007 716 \nMar;7(1):76–85.  717 \n73. Lounibos LP. Invasions by Insect Vectors of Human Disease. Annu Rev Entomol. 718 \n2002;47(1):233–66.  719 \n74. Townroe S, Callaghan A. British Container Breeding Mosquitoes: The Impact of Urbanisation 720 \nand Climate Change on Community Composition and Phenology. PLOS ONE. 2014 Apr 721 \n23;9(4):e95325.  722 \n75. Perrin A, Glaizot O, Christe P. Worldwide impacts of landscape anthropization on mosquito 723 \nabundance and diversity: A meta-analysis. Global Change Biology. 2022;28(23):6857–71.  724 \n76. Watson MS. Aedes (Stegomyia) albopictus  (Skuse): A Literature Review. Miscellaneous 725 \nPublication. 1967;22:1–38.  726 \n77. Roiz D, Neteler M, Castellani C, Arnoldi D, Rizzoli A. Climatic Factors Driving Invasion of 727 \nthe Tiger Mosquito (Aedes albopictus) into New Areas of Trentino, Northern Italy. Baylis M, 728 \neditor. PLoS ONE. 2011 Apr 15;6(4):e14800.  729 \n78. Medley KA, Westby KM, Jenkins DG. Rapid local adaptation to northern winters in the 730 \ninvasive Asian tiger mosquito Aedes albopictus: A moving target. Journal of Applied Ecology. 731 \n2019;56(11):2518–27.  732 \n79. Livdahl TP, Willey MS. Prospects for an Invasion: Competition Between Aedes albopictus 733 \nand Native Aedes triseriatus. Science. 1991;253(5016):189–91.  734 \n80. Yang B, Borgert BA, Alto BW, Boohene CK, Brew J, Deutsch K, et al. Modelling distributions 735 \nof Aedes aegypti and Aedes albopictus using climate, host density and interspecies 736 \ncompetition. PLOS Neglected Tropical Diseases. 2021 Mar 25;15(3):e0009063.  737 \n81. Marcolin F, Branco P, Santos J, Reino L, Santana J, Ribeiro J, et al. Species traits and invasion 738 \nhistory as predictors of freshwater fish invasion success in Europe. Management of Biological 739 \nInvasions. 2025;16(1):277–94.  740 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n39 \n82. Cuthbert RN, Darriet F, Chabrerie O, Lenoir J, Courchamp F, Claeys C, et al. Invasive 741 \nhematophagous arthropods and associated diseases in a changing world. Parasites Vectors. 742 \n2023 Aug 17;16(291):1–17.  743 \n83. Longbottom J, Walekhwa AW, Mwingira V, Kijanga O, Mramba F, Lord JS. Aedes albopictus 744 \ninvasion across Africa: the time is now for cross-country collaboration and control. The Lancet 745 \nGlobal Health. 2023 Feb;S2214109X23000463.  746 \n84. Hubálek Z, Halouzka J. West Nile Fever –a Reemerging Mosquito -Borne Viral Disease in 747 \nEurope. Emerg Infect Dis. 1999 Oct;5(5):643–50.  748 \n85. Le Flohic G, Porphyre V, Barbazan P, Gonzalez JP. Review of Climate, Landscape, and Viral 749 \nGenetics as Drivers of the Japanese Encephalitis Virus Ecology. Johansson MA, editor. PLoS 750 \nNegl Trop Dis. 2013 Sept 12;7(9):e2208.  751 \n86. Zaim M, Subbarao SK, Manouchehri AV, Cochrane AH. Role of Anopheles culicifacies s.l. 752 \nand An. pulcherrimus in malaria transmission in Ghassreghand (Baluchistan), Iran. J Am Mosq 753 \nControl Assoc. 1993 Mar;9(1):23–6.  754 \n87. Swain S, Sharma G, Chittora S, Suman DS. Molecular confirmation of a new mosquito record 755 \nAedes lineatopennis from Odisha along with a comprehensive update of mosquito (Diptera: 756 \nCulicidae) fauna in India. Biologia. 2025 Oct 1;80(10):2801–11.  757 \n88. Vythilingam I, Oda K, Mahadevan S, Abdullah G, Thim CS, Hong CC, et al. Abundance, 758 \nparity, and Japanese encephalitis virus infection of mosquitoes (Diptera:Culicidae) in Sepang 759 \nDistrict, Malaysia. J Med Entomol. 1997 May;34(3):257–62.  760 \n89. CDC USC for DC and P. Arbovirus catalog. 1985.  761 \n90. Kay BH, Carley JG, Fanning ID, Filippich C. Quantitative Studies of the Vector Competence 762 \nof Aedes Aegypti, Culex Annulirostris and Other Mosquitoes (Diptera: Culicidae) with 763 \nMurray Valley Encephalitis and Other Queensland Arboviruses1. J Med Entomol. 1979 Sept 764 \n12;16(1):59–66.  765 \n91. van den Hurk AF, Nisbet DJ, Foley PN, Ritchie SA, Mackenzie JS, Beebe NW. Isolation of 766 \nArboviruses from Mosquitoes (Diptera: Culicidae) Collected from the Gulf Plains Region of 767 \nNorthwest Queensland, Australia. J Med Entomol. 2002 Sept 1;39(5):786–92.  768 \n92. El-Barougy RF, Dakhil MA, Halmy MW, Gray SM, Abdelaal M, Khedr AHA, et al. Invasion 769 \nrisk assessment using trait -environment and species distribution modelling techniques in an 770 \narid protected area: Towards conservation prioritization. Ecological Indicators. 2021 771 \nOct;129:107951.  772 \n93. Hutchinson ML, Darsie RF, Spichiger SE, Jones GE, Naguski EA. Annotated Checklist of the 773 \nMosquitoes of Pennsylvania Including New State Records. Journal of the American Mosquito 774 \nControl Association. 2008 Mar;24(1):1–5.  775 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n40 \n94. Kay BH, Russell RC. Mosquito eradication: the story of killing “Campto.” Collingwood, Vic: 776 \nCSIRO Publishing; 2013.  777 \n95. PAHO W. Status of Aedes aegypti  eradication in the Americas. Pan American Health 778 \nOrganization and World Health Organization. 1967;XVII Meeting.  779 \n96. Roiz D, Pontifes PA, Jourdain F, Diagne C, Leroy B, Vaissière AC, et al. The rising global 780 \neconomic costs of invasive Aedes mosquitoes and Aedes-borne diseases. Science of The Total 781 \nEnvironment. 2024 July 10;933:1–11.  782 \n 783 \nSupporting information captions 784 \nAppendix S1: Traits data set. Data set generated with all species traits combinations and 785 \ncorresponding sources of information will be made available upon acceptance of article 786 \nAppendix S2: Additional model results. Model results of models using only ecological and life-787 \nhistory traits, not macroecological traits, 3 Figures, 2 Tables 788 \nAppendix S3: R Code for analysis. Compressed folder containing the R script used to perform 789 \nthe analyses, along with a simplified version of the input dataset  will be made available upon 790 \nacceptance of the article . The original global mosquito introduction dataset can be accessed at: 791 \nhttps://doi.org/10.5281/zenodo.15731141. 792 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted January 5, 2026. ; https://doi.org/10.64898/2026.01.05.697723doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}