{"paper_id":"17ef14af-c28f-47ba-bcf8-8c33c4d24d43","body_text":"Quantifying vector diversion effects in zoonotic systems: A\nmodelling framework for arbovirus transmission between reservoir\nand dead-end hosts\nEmma L Fairbanks1,*, Matthew Baylis2, Janet M Daly3, and Michael J Tildesley1\n1The Zeeman Institute for Systems Biology & Infectious Disease Epidemiology Research, Mathematics\nInstitute and School of Life Sciences, University of Warwick, Coventry, CV4 7AL, UK\n2Institute of Infection, Veterinary and Ecological Sciences, Faculty of Health and Life Sciences, University\nof Liverpool, Liverpool, L, UK\n3One Virology - Wolfson Centre for Global Virus Research, School of Veterinary Medicine and Science,\nUniversity of Nottingham, Loughborough, LE12 5RD, UK\n*Corresponding author: emma-louise.fairbanks@warwick.ac.uk\nAbstract\nVector-borne disease transmission involves complex interactions between vectors, reser-\nvoir hosts and dead-end hosts. We present a mathematical model for the vectorial ca-\npacity that incorporates multiple host types and their interactions, focusing specifically\non West Nile virus transmission byCulex pipiens mosquitoes. Our model integrates\nclimate-dependent parameters affecting vector biology with vector control interventions\nto predict transmission potential under various scenarios. We demonstrate how vec-\ntor control interventions targeting one host type can significantly impact transmission\ndynamics across all host populations. By examining the effects of different vector con-\ntrol tool modes of action (repellency, preprandial killing, disarming and postprandial\nkilling), we develop target product profiles that minimise unintended consequences of\nvector control. Notably, we identify the optimal intervention characteristics needed to\nprevent repellency on dead-end hosts from inadvertently increasing transmission among\nreservoir hosts. This research provides valuable insights for public health officials de-\nsigning targeted vector control strategies and offers a flexible modelling framework that\ncan be adapted to other vector-borne diseases with complex host dynamics.\nKey words: Culex, UK, Vectorial capacity, West Nile virus, Vector control interventions,\nHost-vector interactions, Climate, Target product profiles, Repellency, Basic reproduction\nnumber, Seasonal transmission\n1\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nAuthor summary\nMosquitoes that spread diseases like West Nile virus don’t just bite one type of animal—they\nfeed on birds, humans, and other mammals. This creates a complex web of disease transmis-\nsion that current prevention strategies often overlook. We developed a mathematical model\nto understand what happens when mosquito control methods target different types of hosts\nin this network.\nOur research reveals a surprising and concerning finding: when people use repellents to pro-\ntect themselves from mosquitoes, those mosquitoes don’t simply disappear—they redirect\nto birds instead. Since birds are the main animals that can spread West Nile virus to other\nmosquitoes, this redirection can actually increase disease transmission in the bird population\nby up to 23%. More infected birds ultimately means more infected mosquitoes and higher\nrisk for humans.\nHowever, we also identified a solution. We found that if repellent products could kill just 2%\nof the mosquitoes they encounter before those mosquitoes find alternative hosts, this would\neliminate the harmful redirection effect. Our work provides specific guidelines for developing\nbetter mosquito control products and helps public health officials understand the broader\nconsequences of different intervention strategies. This framework can be applied to other\nmosquito-borne diseases beyond West Nile virus.\n2\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\n1 Introduction\nWhile many mathematical models for predicting the risk of disease transmission consider\nthe effects of climate and vector control on the number of vectors [1–15], sometimes the\nvectorial capacity is not considered. Vectorial capacity, a measure of the potential of a\nvector population to transmit a pathogen, is defined as the total number of potentially\ninfectious bites that would arise from all vectors biting a single infectious host on a single\nday [16]. This not only considers the number of vectors, but also their ability to transmit a\npathogen.\nTherefore, vectorial capacity is the product of both the number of vectors and the relative\nvector capacity (rVC), defined as the total number of potentially infectious bites that would\narise from a single vector biting a single host on a single day. The basic reproductive number,\nthe expected number of cases to arise from a single infectious case in a susceptible popula-\ntion, can be expressed as the product of the vectorial capacity and duration of the host’s\ninfectious period [17].\nWe previously modelled the rVC for viruses transmitted byCulicoides [18], considering how\nclimatic variables affect the rate at which vectors feed, expected lifespan of vectors and the\nextrinsic incubation period (EIP), defined as the time required for a vector to become infec-\ntious after consuming an infected blood meal. The model also considers the effects of vector\ncontrol interventions at different stages during the mosquito feeding cycle. Additionally, the\nmodel considers host selection, which is influenced by the availability of different host types\nand the vector’s preferences for these hosts.\nHowever, there are some limitations to this model. For example, the model only consid-\ners one host species. While this may be informative in localised populations with only one\ncompetent host present, it is not possible to consider interactions between host species. In\nsome cases, a disease may be mainly transmitted by a host which shows less severe clinical\nsigns than other hosts or dead-end hosts (hosts which can be infected but cannot infect).\nHosts which can both be infected and transmit a pathogen are often referred to as reservoir\nhosts. These hosts can amplify disease transmission, whereas infection of a dead-end host\nhas no downstream effects of transmission. In these cases it is important to consider the\ninteractions between hosts.\nVector control interventions targeting one host type can significantly impact transmission\ndynamics across all host populations. When control measures such as repellents or insecti-\ncides are applied to one host type, mosquito vectors may be diverted to alternative hosts,\npotentially altering disease transmission patterns.\nIn this study, we extend this model to consider the rVC for transmission of pathogens\nbetween host species. Here, we include reservoir hosts and dead-end hosts. As an example,\nwe model West Nile virus (WNV) transmission byCulex pipiens, where interactions between\nreservoir hosts (birds) and dead-end hosts (humans and other mammals) determine overall\ntransmission intensity.\n3\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\n2 Methods\n2.1 Model development\nThe baseline model for simulating the dynamics in a single host species is described in full\ndetail in Fairbanks et al. [18]. Here, the rVC on dayt was calculated as the probability the\nvector feeds on the host of interest on dayt and survives until the end of that day multiplied\nby the probability it infects another of the same host type every day after, given it survives\nuntil that day. The model considered the climatic dependence of the rates of gonotrophic\ncycle completion, vector mortality and pathogen EIP completion. It can either be simulated\nusing daily climatic data or, alternatively, for a constant temperature or set durations of\ngonotrophic cycle completion, vector lifespan and EIP completion.\nIn this study, we extend this model to account for the interactions between vectors and\ndifferent host types. This multiple host model is described in detail in the Supplementary\ninformation. It can be utilised to calculate the transmission potential from each reservoir\nhost species to each other host species, including dead-end hosts. For interpretability, we\ndefine the basic reproduction number (R0) as the product of the rVC, the infectious period\nand the expected number of vectors per host [17, 18].\nWithin the modelling framework, vectors that encounter hosts can experience several end-\npoints: they might successfully feed, be killed before (preprandial mortality), become dis-\narmed (prolongedbloodfeeding inhibition) orabandon the current host tosearch foranother.\nVectors that succeed in feeding may still be killed after the blood meal (postprandial mor-\ntality).\nThe model simulations and subsequent data processing were conducted in R, with visualisa-\ntions generated using the ggplot2 package [19] within the RStudio integrated development\nenvironment [20].\n2.2 Model application\nWe simulate the model for West Nile virus transmission byCx. pipiens. The model param-\neters were set following a comprehensive literature review. The rate of gonotrophic cycle\ncompletion and vector life-span p day were parameterised by Shocket et al. [21] using data\nfrom four [22–25] and three [24, 26, 27]Cx. pipiens studies, respectively. For the EIP, we\nuse the model from Vollans et al. [28], which was parameterised using Bayesian methods and\ndata from 2145 mosquitoes. This parameterisation includes the probability of transmission\nfrom host to vector and therefore we setρh→v = 1 for reservoir hosts. Vollans et al. [28] found\nthat the gamma and Weibull distributions performed similarly. Here, we favour the gamma\ndistribution since it has a property of being able to be broken up into multiple exponential\ndistributions. This allows for consideration of daily fluctuations in temperature.\nWhen estimating the number of vectors per host required forR0 > 1, we consider avian\nreservoir hosts with infectious periods of 7 and 10 days [29]. An infectious period of 7 days\n4\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nmay be applicable to geese [30], passerines [31] and owls [32], and 10 days is applicable to\nraptors [33] and turkeys [34].\n2.2.1 Comparing the transmission potential between host types\nGriep et al. [35] reviewed blood meal sources from 26,857Culex species, including 2062Cx.\npipiens in palearctic regions. They found that 68.3% percent had fed on avian hosts, and\n14.1% and 17.2% fed on human and non-human mammal hosts, respectively. We consider\nthe rVC from birds (the reservoir hosts) to human and other mammal hosts (assuming these\nare a dead-end hosts). As not all mammals are hosts of WNV this is likely to represent an\nupper bound. We calculate the rVC for temperatures between 10 and 32°C using these fixed\nblood selection values, however these are likely to vary between settings [36].\nThe probability of transmission from vector to host is assumed to be the same for all hosts,\ngiving ρh→v = 0.74 for all three hosts [37]. We also assume that vectors bite only once per\nfeeding cycle and that there are no vector control tools present.\n2.2.2 Simulation at a larger spatial scale\nWe then simulate the model considering daily temperature time series from the UK to pro-\nduce maps showing the patterns in transmission potential. Temperature for 2019–2023 was\nextracted from the HadUK-Grid gridded average climate observations for 5 km grid squares\nof the UK [38]. For each year, we calculate the rVC each day. To summarise the results,\nwe pool estimates for each month across days and year to calculate the monthly minimum,\nmean and maximum. The UK represents an ideal case study for demonstrating applicability\nof the model, as it encompasses considerable climatic variation across a relatively compact\ngeographical area, allowing for examination of diverse transmission dynamics.\n2.2.3 Considering the effect of vector-control tools on different host types\nWhen considering how vector control affects the transmission potential, we will consider\ntwo host types; a reservoir (birds) and another dead-end host. We vary the percentage of\nvectors feeding on the reservoir host and assume the rest feed on a dead-end host. We\nsimulate the model for settings where, in the absence of vector control, the vector would se-\nlect a reservoir host 20%, 50% or 80% of the time to capture a range of demographic settings.\nInitially, we will consider the effects of a tool which only repels mosquitoes, therefore they\ncontinue host-seeking and may bite another host of any type. We vary the repellent effect\nassuming a 0–100% reduction in the rate of blood feeding. The coverage is also varied from\n0-100% of the target host type having the tool. We assume that if a host type has access\nto a tool they always use it when exposed to mosquitoes. When analysing the effects of\nvector-control we fix the temperature at the optimal value for transmission, as calculated in\nSection 2.2.1.\nFinally, we investigate how other vector control characteristics (modes of action) can coun-\nteract the potential negative effects of repellents. When repellents are applied to dead-end\n5\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nhosts, they may inadvertently increase transmission by redirecting mosquitoes towards reser-\nvoir hosts. For each percentage of the reduction in the rate of biting investigated (0-100%),\nwe therefore calculate the percentage of the repelled mosquitoes which would need to follow\neach alternative mode of action (preprandial killing, disarming, or postprandial killing) to\nmitigate this increased transmission risk.\n3 Results\n3.1 Model development\nOur study extends a mathematical model of the rVC that integrates climatic factors and\nvector control interventions, introducing interactions between vectors and multiple types of\nhost. This approach enables consideration of secondary and dead-end hosts in the analysis\nallowing for consideration of how control tools on one host type affect transmission potential\nto other host types, as demonstrated in the following results.\n3.2 Comparing the transmission potential to host types\nAs expected, the magnitude of the rVC for each type of host corresponds to the likelihood\na vector selects the host, with hosts fed upon more often receiving a larger portion of the\ndisease burden (Figure 1). The maximum rVC is when the temperature is 24.5°C. This\ncorresponds to a rVC of 0.061, 0.013 and 0.015 for birds, humans and other mammals, re-\nspectively.\nFor a bird species with an infectious period of 7 days, the vectors per host required for\nR0 > 1 in birds, humans and other hosts would be 2.34, 10.99 and 9.52, respectively. For\nbird species with an infectious period of 10 days the required number of vectors per host\nwould be 1.64, 7.69 and 6.67 for birds, humans and other hosts, respectively.\n3.3 Large scale simulation\nUsing daily mean temperature data, we produced maps of the mean monthly rVC (Fig-\nure 2). This allows for comparisons both spatially and temporally. Further details regarding\nmonthly rVC values and corresponding vector population thresholds can be found in Sup-\nplementary table S3.\nThe modelling outputs reveal considerable spatial heterogeneity. We observe that the rVC\nin Scotland and Northern Ireland is relatively low when compared with Wales and England,\nwith England having the largest predicted rVC on average. The model predicts the largest\nrVC in south-east England, however, similar values are observed across the south of England,\nin South Wales and as far north as Merseyside and parts of Yorkshire.\nThe estimates demonstrate pronounced seasonal variations in rVC with significant impli-\ncations for disease transmission dynamics. June exhibits the peak rVC (0.016), situated\nwithin a broader high-transmission window extending from May to August. During this\n6\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nFigure 1: The relative vectorial capacity from birds (the reservoir host) to each host type at\nconstant temperatures.\nFigure 2: Monthly mean predicted relative vectorial capacity for birds (the reservoir host)\ncalculated for mean daily temperature estimates for 2019–2023.\n7\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nsummer period, the required vector population threshold for disease establishment (R0 > 1)\nremains biologically plausible, ranging from approximately 6 to 15 vectors per host for an\navian reservoir host with a 10-day infectious period.\nIn contrast, the winter months (November to March) show markedly reduced transmission\npotential, with numerous locations experiencing a rVC of zero. The maximum rVC in De-\ncember is approximately 1500 times lower than the June peak, corresponding to an extraor-\ndinarily high vector population threshold (9371.54 vectors per host) that renders sustained\ntransmission biologically unlikely given typical winter vector population densities. April and\nSeptember-October emerge as critical transition periods, characterised by rapidly changing\nrVC as environmental conditions shift.\n3.4 Considering the effect of vector-control tools on different host\ntypes\nFor interpretation, it is important to consider that, given a fixed number of vectors per host\nwithin a population, we would expect the vectorial capacity to increase or decrease propor-\ntionally with the rVC. This assumes that the size of the vector population is limited by the\ncarrying capacity, which is a common assumption when modelling the impact of interven-\ntions on vectorial capacity [39–41].\nFigure 3 shows that while repelling vectors from a reservoir host always reduces the rVC of\nthe reservoir and dead-end hosts, repelling vectors from dead-end hosts can increase the rVC\nof reservoir hosts. As coverage of the tool and repellency increase we see a larger decrease\nin rVC for both host types when the tool is given to a reservoir host or in dead-end hosts\nwhen the tool is given to a dead-end host, and a larger increase in the rVC in a reservoir\nhost when the tool is given to a dead-end host.\nWhen targeting dead-end hosts, the reduction in the rVC of dead-end hosts tells us that\neach mosquito that feeds on an infectious reservoir is less likely to cause an infection in the\ndead-end host. However, the unintentional increase in vectorial capacity in reservoir hosts\nsuggests that there will be more infected reservoir hosts, and therefore potentially more\nvectors becoming infected. In our scenario, where the model is parameterised to describe\nWNV transmission byCx. pipiens, we see an increase in vectorial capacity of up to 23.4%.\n3.5 Target product profiles for vector-control tools\nFor a given level of repellency for a tool used on a dead-end host, Figure 4 shows the es-\ntimated magnitude of the other modes of action required to avoid the increase in vectorial\ncapacity for reservoir hosts. Tools which preprandially kill at least 2% ofCx. pipiens of\nthe repelled vectors counteract these unintended negative effects for all levels of repellency.\nHowever, for tools which repel almost all (> 96%) mosquitoes feeding on dead-end hosts,\npostprandially killing or disarming cannot counteract all of the negative effects of the repel-\n8\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nFigure 3: The estimated relative change in vectorial capacity for the reservoir and dead-end\nhost when a tool is used by a (a) reservoir or (b) dead-end host for tools which reduce the\nrate of biting at different population coverage levels.\n9\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nFigure 4: Magnitude of each mode of action (MoA) required to counteract the increased\nvectorial capacity in reservoir hosts when a tool repells hosts from dead-end hosts.\nlency.\nPostprandial mortality is more effective than disarming, however its effects rapidly decrease\nwhen over 80% of vectors are repelled. The percentage of mosquitoes which need to be\npostprandially killed, given 80% or 90% of mosquitoes are repelled, is 8% or 18%, further\nincreasing to 38% when 95% of mosquitoes are repelled. This is because the vectors must\nfeed for postprandial killing to be effective – the fewer vectors that feed, of those that do\nfeed a higher percentage need to be killed.\n4 Discussion\nTo our knowledge, this is the first model that quantifies the rVC between multiple host\ntypes independently. Understanding these complex interactions is crucial for developing ef-\nfective vector control strategies, as interventions focused solely on protecting dead-end hosts\ncould inadvertently increase transmission among reservoir hosts, potentially elevating overall\ndisease risk. By modelling these relationships, we can identify optimal intervention charac-\nteristics that minimise unintended consequences of vector control and maximise public health\nbenefits.\nThe optimal temperature for WNV transmission was estimated to be 24.5°C, in line with\nprevious estimates [21, 42]. This temperature represents the thermal optimum where rVC\nreaches its peak due to the combined effects of accelerated mosquito development, short-\nened gonotrophic cycle, and reduced EIP, while still maintaining sufficient vector longevity.\nFurthermore, this optimal temperature falls within the summer temperature range of many\ntemperate regions, including parts of the UK, explaining the seasonal patterns observed in\nour spatial analysis. For the UK, modelled seasonal and spatial patterns in rVC suggest\nthat early warning systems, disease surveillance and control strategies would benefit from\ntargeted implementation, with intensified efforts during the high-transmission window and\n10\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nin locations exhibiting elevated rVC.\nOur model predicts higher rVC values in urban areas, particularly London. Urban areas\ntypically experience temperatures 1–3°C higher than surrounding rural environments [43].\nThis urban amplification can be attributed to the urban heat island effect, which creates\nmicroclimatic conditions that are more favourable forCx. pipiens, potentially extending the\ntransmission season and intensifying WNV risk in temperate, metropolitan areas. Addition-\nally, recent research indicates that whileCx. pipiens larvae are more abundant in urban\nresidential areas due to the prevalence of artificial breeding habitats such as storm drains,\nadult mosquitoes often show higher densities in urban parks where they find more suitable\nmicroclimatic conditions and diverse hosts [44]. While our model focuses primarily on rVC,\nit is important to acknowledge that actual disease transmission risk is the product of both\nrVC and the number of vectors per host. Mosquito population dynamics are influenced by\nnumerous factors beyond temperature, including rainfall patterns and habitat availability.\nThe heterogeneity in vector abundance can significantly modulate transmission risk even in\nareas with high rVC.\nHost selection of vectors depends on host preference and host availability [18], therefore is\nlikely to vary across settings. Experimental studies have demonstrated thatCx. pipiens\nprefer some avian species over others [45]. However, these preferences are modulated by\nenvironmental conditions and host availability in field settings, leading to spatial and tem-\nporal variation in feeding patterns [? ]. Such heterogeneity in host selection across different\nlandscapes significantly impacts WNV transmission potential.\nThis modelling framework is flexible for application to other vectors and pathogens. Due to\na reduction in wildlife habitats, we are observing an increase in the risk of human exposure to\nemerging and existing zoonotic pathogens [46]. An increase in transmission between wildlife\nreservoirs and human hosts may also have implications for disease elimination. Currently,\nnumbers of human zoonotic malaria cases are increasing globally and there are currently no\ncontrol measures that target wildlife reservoirs [46, 47].\nMany health agencies suggest using repellent to reduce the risk of vector-borne disease [48–\n50]]. This work shows that for zoonotic diseases this may not always be the case. More\ninvestigation is needed into the effects of increasing vectorial capacity on disease transmis-\nsion. The modes of action of vector control tools change according to the resistance level of\nvectors and insecticide dose [39, 40]. Insecticide resistance has been reported inCx. pipi-\nens in Europe [51] and globally [52]. Therefore, more data on the responses of local vector\npopulations to insecticides should be gathered to ensure robust evaluation within different\nsettings. This modelling framework can be used to assess whether these tools are suitable\nfor use on dead-end hosts, and suggest target product profiles of additional tools for use.\nRepellents that also kill or disarm mosquitoes are more likely to avoid unintentional increases\nin vectorial capacity in reservoir hosts, with preprandial killing being the most effective. This\nstudy provides the first quantitative estimates of the magnitude of unique modes of action\nof tools required to reduce disease transmission. These quantitative thresholds offer valuable\n11\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\ntargets for product developers and public health officials designing next-generation vector\ncontrol strategies. In our WNV example, if 2% of the reduction in biting is due to prepran-\ndial killing, vectorial capacity will be reduced in all hosts. However, this may vary across\ndifferent settings and vector populations [41].\nOur mathematical modelling presented here advances the understanding of host-vector-\npathogen interactions. This work demonstrates the critical importance of considering entire\ntransmission systems when implementing vector control, as interventions targeting one host\npopulation can significantly influence transmission throughout the ecological network. This\nunderscores the need for integrated approaches to vector control that consider both personal\nprotection and community-level impacts. The model is a valuable tool for assessing trans-\nmission risk and optimising control interventions in the face of emerging zoonotic threats\nand changing climate conditions.\nConflict of interest The authors declare no conflict of interest.\nEthical approval The authors confirm that the ethical policies of the journal, as noted\non the journal’s author guidelines page, have been adhered to. No ethical approval was\nrequired.\nAuthor contributions\nEmma L Fairbanks:Conceptualisation, Methodology, Software, Validation, Formal anal-\nysis, Investigation, Data Curation, Writing - Original Draft, Writing - Review & Editing,\nVisualisation, Funding acquisition;Matthew Baylis:Conceptualisation, Writing - Review\n& Editing;Janet M Daly:Conceptualisation, Writing - Review & Editing, Funding acqui-\nsition;Michael J Tildesley:Conceptualisation, Methodology, Writing - Review & Editing,\nFunding acquisition\nFinancial support The authors were supported by the Horserace Betting Levy Board\n(vet/prj/809). MJT is funded on a joint BBSRC/EEID grant (BB/T004312/1).\nData Availability Statement Code to simulate the model will be made available in a\nGitHub package and achieved using Zenodo to provide a DOI upon publication. This is\nprovided for review in VectorialCapacity_MultipleHosts.R.\nReferences\n[1] AJ Worton, RA Norman, L Gilbert, and RB Porter. GIS-ODE: linking dynamic\npopulation models with GIS to predict pathogen vector abundance across a coun-\ntry under climate change scenarios. J R Soc Interface, 21(217):20240004, 2024. doi:\n10.1098/rsif.2024.0004.\n[2] LHV Franklinos, DW Redding, TCD Lucas, R Gibb, I Abubakar, and KE Jones. Joint\nspatiotemporal modelling reveals seasonally dynamic patterns of Japanese encephalitis\n12\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\nvector abundance across India. PLoS Negl Trop Dis , 16(2):e0010218, 2022. doi: 10.\n1371/journal.pntd.0010218.\n[3] Modellingthemonthlyabundanceof Culicoides bitingmidgesinnineeuropeancountries\nusing random forests machine learning.Parasit Vectors, 13(1):194, 2020. doi: 10.1186/\ns13071-020-04053-x.\n[4] J Liu-Helmersson, A Brännström, MO Sewe, JC Semenza, and J Rocklöv. Estimat-\ning past, present, and future trends in the global distribution and abundance of the\narbovirus vector Aedes aegypti under climate change scenarios.Front Public Health, 7:\n148, 2019. doi: 10.3389/fpubh.2019.00148.\n[5] MC Cecere, LI Rodríguez-Planes, GM Vazquez-Prokopec, U Kitron, and RE Gürtler.\nCommunity-based surveillance and control of chagas disease vectors in remote rural\nareas of the Argentine Chaco: A five-year follow-up.Acta Trop, 191:108–115, 2019. doi:\n10.1016/j.actatropica.2018.12.038.\n[6] RS McCann, JP Messina, DW MacFarlane, MN Bayoh, JE Gimnig, E Giorgi, and\nED Walker. Explaining variation in adult Anopheles indoor resting abundance: the\nrelative effects of larval habitat proximity and insecticide-treated bed net use.Malar J,\n16(1):288, 2017. doi: 10.1186/s12936-017-1938-1.\n[7] HE Brown, R Barrera, AC Comrie, and J Lega. Effect of temperature thresholds on\nmodeled Aedes aegypti (Diptera: Culicidae) population dynamics. J Med Entomol, 54\n(4):869–877, 2017.\n[8] DA Ewing, CA Cobbold, BV Purse, MA Nunn, and SM White. Modelling the effect\nof temperature on the seasonal population dynamics of temperate mosquitoes.J Theor\nBiol, 400:65–79, 2016. doi: 10.1016/j.jtbi.2016.04.008.\n[9] G Marini, P Poletti, M Giacobini, A Pugliese, S Merler, and R Rosà. The role of\nclimatic and density dependent factors in shaping mosquito population dynamics: the\ncase of Culex pipiens in northwestern Italy. PLoS One, 11(4):e0154018, 2016. doi:\n10.1371/journal.pone.0154018.\n[10] C Christiansen-Jucht, K Erguler, CY Shek, M-G Basáñez, and PE Parham. Mod-\nelling Anopheles gambiae ss population dynamics with temperature-and age-dependent\nsurvival. Int J Environ Res Public Health , 12(6):5975–6005, 2015. doi: 10.3390/\nijerph120605975.\n[11] C Talla, D Diallo, I Dia, Y Ba, J-A Ndione, AA Sall, A Morse, A Diop, and M Diallo.\nStatistical modeling of the abundance of vectors of West african Rift Valley fever in\nBarkédji, Senegal. PLoS One, 9(12):e114047, 2014. doi: 10.1371/journal.pone.0114047.\n[12] PE Parham, D Pople, C Christiansen-Jucht, S Lindsay, W Hinsley, and E Michael.\nModeling the role of environmental variables on the population dynamics of the\nmalaria vector Anopheles gambiae sensu stricto . Malar J, 11:271, 2012. doi: 10.1186/\n1475-2875-11-271.\n13\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\n[13] J Wang, NH Ogden, and H Zhu. The impact of weather conditions onCulex pipiens\nand Culex restuans (Diptera: Culicidae) abundance: a case study in Peel region.J Med\nEntomol, 48(2):468–475, 2011. doi: 10.1603/me10117.\n[14] KP Paaijmans, SS Imbahale, MB Thomas, and W Takken. Relevant microclimate for\ndetermining the development rate of malaria mosquitoes and possible implications of\nclimate change. Malar J, 9:196, 2010. doi: 10.1186/1475-2875-9-196.\n[15] B Schaeffer, B Mondet, and S Touzeau. Using a climate-dependent model to pre-\ndict mosquito abundance: application to Aedes (Stegomyia) africanus and Aedes\n(Diceromyia) furcifer (Diptera: Culicidae). Infect Genet Evol, 8(4):422–432, 2008. doi:\n10.1016/j.meegid.2007.07.002.\n[16] C Garrett-Jones. The human blood index of malaria vectors in relation to epidemiolog-\nical assessment. Bull World Health Organ , 30(2):241, 1964.\n[17] SPC Brand and MJ Keeling. The impact of temperature changes on vector-borne\ndisease transmission: Culicoides midges and bluetongue virus. J R Soc Interface, 14\n(128):20160481, 2017. doi: 10.1098/rsif.2016.0481.\n[18] EL Fairbanks, JM Daly, and MJ Tildesley. Modelling the influence of climate and\nvector control interventions on arbovirus transmission.Viruses, 16(8):1221, 2024. doi:\n10.3390/v16081221.\n[19] HWickham, WChang, LHenry, TLPedersen, KTakahashi, CWilke, KWoo, HYutani,\nD Dunnington, and RStudio. Create Elegant Data Visualisations Using the Grammar\nof Graphics, 2022. URL https://ggplot2.tidyverse.org.\n[20] RStudio Team. Rstudio: Integrated Development Environment for R . RStudio, PBC.,\nBoston, MA, 2020. URLhttp://www.rstudio.com/.\n[21] Marta S Shocket, Anna B Verwillow, Mailo G Numazu, Hani Slamani, Jeremy M Cohen,\nFadoua El Moustaid, Jason Rohr, Leah R Johnson, and Erin A Mordecai. Transmission\nof West Nile and five other temperate mosquito-borne viruses peaks at temperatures\nbetween 23 c and 26 c.Elife, 9:e58511, 2020. doi: 10.7554/eLife.58511.\n[22] J Li, G Zhu, H Zhou, J Tang, and J Cao. Effect of temperature on the development of\nCulex pipiens pallens. Chin J Vector Biol Control , 28:35–37, 2017.\n[23] DJ Madder, GA Surgeoner, and BV Helson. Number of generations, egg production,\nand developmental time ofCulex pipiens and Culex restuans (Diptera: Culicidae) in\nsouthern Ontario.J Med Entomol, 20(3):275–287, 1983. doi: 10.1093/jmedent/20.3.275.\n[24] Jordan E Ruybal, Laura D Kramer, and AM Kilpatrick. Geographic variation in the\nresponse ofCulex pipiens life history traits to temperature.Parasit Vectors, 9:116, 2016.\ndoi: 10.1186/s13071-016-1402-z.\n14\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\n[25] A Tekle. The physiology of hibernation and its role in the geographical distribution of\npopulations of theCulex pipiens complex. Am J Trop Med Hyg , 9:321–330, 1960. doi:\n10.4269/ajtmh.1960.9.321.\n[26] SS Andreadis, OC Dimotsiou, and M Savopoulou-Soultani. Variation in adult longevity\nof Culex pipiens f. pipiens , vector of the West Nile virus.Parasitol Res, 113:4315–4319,\n2014. doi: 10.1007/s00436-014-4152-x.\n[27] AT Ciota, AC Matacchiero, AM Kilpatrick, and LD Kramer. The effect of temperature\non life history traits of Culex mosquitoes. J Med Entomol , 51(1):55–62, 2014. doi:\n10.1603/me13003.\n[28] M Vollans, J Day, S Cant, J Hood, AM Kilpatrick, LD Kramer, A Vaux, J Medlock,\nT Ward, and RS Paton. Modelling the temperature dependent extrinsic incubation\nperiod of West Nile virus using Bayesian time delay models.J Infect, 89(6):106296,\n2024. doi: 10.1016/j.jinf.2024.106296.\n[29] V Gamino and U Höfle. Pathology and tissue tropism of natural West Nile virus\ninfection in birds: a review.Vet Res, 44(1):39, 2013. doi: 10.1186/1297-9716-44-39.\n[30] C Banet-Noach, L Simanov, and M Malkinson. Direct (non-vector) transmission\nof West Nile virus in geese. Avian Pathol , 32(5):489–494, 2003. doi: 10.1080/\n0307945031000154080.\n[31] DA LaPointe, EK Hofmeister, CT Atkinson, RE Porter, and RJ Dusek. Experimental\ninfection of hawaiiamakihi (Hemignathus virens) with West Nile virus and competence\nof a co-occurring vector,Culex quinquefasciatus: Potential impacts on endemic hawaiian\navifauna. J Wildl Dis , 45(2):257–271, 2009. doi: 10.7589/0090-3558-45.2.257.\n[32] NM Nemeth, DC Hahn, DH Gould, and RA Bowen. Experimental West Nile virus\ninfection in eastern screech owls (Megascops asio). Avian Dis, 50(2):252–258, 2006. doi:\n10.1637/7466-110105R1.1.\n[33] U Ziegler, J Angenvoort, D Fischer, C Fast, M Eiden, AV Rodriguez, S Revilla-\nFernández, N Nowotny, JG de la Fuente, M Lierz, and MH Groschup. Pathogenesis\nof West Nile virus lineage 1 and 2 in experimentally infected large falcons.Vet Micro-\nbiol, 161(3-4):263–273, 2013. doi: 10.1016/j.vetmic.2012.07.041.\n[34] DE Swayne, JR Beck, and S Zaki. Pathogenicity of West Nile virus for turkeys.Avian\nDis, pages 932–937, 2000. doi: 10.2307/1593067.\n[35] JS Griep, E Grant, J Pilgrim, O Riabinina, M Baylis, M Wardeh, and MSC Blagrove.\nMeta-analyses ofCulex blood-meals indicates strong regional effect on feeding patterns.\nPLoS Negl Trop Dis, 19(1):e0012245, 2025. doi: 10.1371/journal.pntd.0012245.\n[36] Emma L Fairbanks, Michael J Tildesley, and Janet M Daly. A systematic review quan-\ntifying host feeding patterns of culicoides species responsible for pathogen transmission.\nbioRxiv, pages 2024–07, 2024. doi: 10.1101/2024.07.25.605155.\n15\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\n[37] Marjorie J Wonham, Mark A Lewis, Joanna Rencławowicz, and P Van den Driess-\nche. Transmission assumptions generate conflicting predictions in host–vector dis-\nease models: a case study in West Nile virus. Ecol Lett, 9(6):706–725, 2006. doi:\n10.1111/j.1461-0248.2006.00912.x.\n[38] D Hollis, M McCarthy, M Kendon, T Legg, and I Simpson. HadUK-grid‚ÄîA new\nUK dataset of gridded climate observations.Geosci Data J, 6(2):151–159, 2019. doi:\n10.1002/gdj3.78.\n[39] EL Fairbanks, M Saeung, A Pongsiri, E Vajda, Y Wang, DJ McIver, JH Richardson,\nA Tatarsky, NF Lobo, SJ Moore, A Ponlawat, T Chareonviriyaphap, A Ross, and\nN Chitnis. Inference for entomological semi-field experiments: Fitting a mathemati-\ncal model assessing personal and community protection of vector-control interventions.\nComput Biol Med , 168:107716, 2024. doi: 10.1016/j.compbiomed.2023.107716.\n[40] EL Fairbanks, MM Tambwe, J Moore, A Mpelepele, NF Lobo, R Mashauri, N Chitnis,\nand SJ Moore. Evaluating human landing catches as a measure of mosquito biting and\nthe importance of considering additional modes of action.Sci Rep, 14(1):11476, 2024.\ndoi: 10.1038/s41598-024-61116-0.\n[41] Y Wang, N Chitnis, and EL Fairbanks. Optimizing malaria vector control in the Greater\nMekong Subregion: a systematic review and mathematical modelling study to identify\ndesirable intervention characteristics. Parasit Vectors, 17(1):162, 2024. doi: 10.1186/\ns13071-024-06234-4.\n[42] J Heidecke, J Wallin, P Fransson, P Singh, H Sjödin, PC Stiles, M Treskova, and\nJ Rocklöv. Uncovering temperature sensitivity of West Nile virus transmission: Novel\ncomputational approaches to mosquito-pathogen trait responses. PLoS Comput Biol,\n21(3):e1012866, 2025. doi: 10.1371/journal.pcbi.1012866.\n[43] NB Grimm, SH Faeth, NE Golubiewski, CL Redman, J Wu, X Bai, and JM Briggs.\nGlobal change and the ecology of cities. Science, 319(5864):756–760, 2008. doi: 10.\n1126/science.1150195.\n[44] LKrol, MLangezaal, LBudidarma, DWassenaar, EADidaskalou, KTrimbos, MDellar,\nPM van Bodegom, GW Geerling, and M Schrama. Distribution ofCulex pipiens life\nstages across urban green and grey spaces in Leiden, The Netherlands.Parasit Vectors,\n17(1):37, 2024. doi: 10.1186/s13071-024-06120-z.\n[45] JE Simpson, CM Folsom-O’Keefe, JE Childs, LE Simons, TG Andreadis, and MA Diuk-\nWasser. Avian host-selection byCulex pipiens in experimental trials.PLoS One, 4(11):\ne7861, 2009. doi: 10.1371/journal.pone.0007861.\n[46] F Keesing and RS Ostfeld. Impacts of biodiversity and biodiversity loss on zoonotic\ndiseases. Proc Natl Acad Sci USA , 118(17):e2023540118, 2021. doi: 10.1073/pnas.\n2023540118.\n16\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint \n\n[47] KM Fornace, CJ Drakeley, KA Lindblade, J Jelip, and K Ahmed. Zoonotic malaria\nrequires new policy approaches to malaria elimination.Nat Commun, 14(1):5750, 2023.\ndoi: 10.1038/s41467-023-41546-6.\n[48] Centers for Disease Control and Prevention. Preventing West Nile Virus, 2024.\nURL https://www.cdc.gov/west-nile-virus/prevention/index.html. Accessed:\n15/05/2025.\n[49] Canadian Centre for Occupational Health and Safety. West Nile Virus, 2021.\nURL https://www.ccohs.ca/oshanswers/diseases/westnile.html. Accessed:\n15/05/2025.\n[50] European Centre for Disease Prevention and Control. West Nile virus season in\nfull swing in Europe, 2024. URL https://www.ecdc.europa.eu/en/news-events/\nwest-nile-virus-season-full-swing-europe . Accessed: 15/05/2025.\n[51] S Vereecken, A Vanslembrouck, IM Kramer, and R Müller. Phenotypic insecticide\nresistance status of theCulex pipiens complex: a european perspective.Parasit Vectors,\n15(1):423, 2022. doi: 10.1186/s13071-022-05542-x.\n[52] JG Scott, MH Yoshimizu, and S Kasai. Pyrethroid resistance in Culex pipiens\nmosquitoes. Pestic Biochem Physiol , 120:68–76, 2015. doi: 10.1016/j.pestbp.2014.12.\n018.\n17\n.CC-BY 4.0 International licenseavailable under a \nwas 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 preprint (whichthis version posted July 31, 2025. ; https://doi.org/10.1101/2025.07.24.666561doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}