Modeling the effects of vector species composition and temperature on the risk of dengue virus Authors: Suzanne L Robertson, Rebeca de Jesús Crespo, Emma Beck, Lauren Beuerle, Patt Martin, Regan Stambaugh, Michael A Robert Published: 2026 ABSTRACT AbstractThe risk of vector-borne disease is highly dependent on the local community composition of hosts and vectors, as well as the means by which disease is introduced into a susceptible population. The mosquito species Aedes aegypti and Aedes albopictus are both vectors of dengue virus, but differ in their biting preferences and ability to transmit the disease. The two species compete for habitat at the larval stage and their spatial distributions are highly heterogeneous, due in part to variability in factors such as temperature and resource quality affecting the outcome of competition and the resulting abundance of each species. In addition to affecting vector population dynamics, temperature also strongly affects multiple aspects of the disease transmission process. We present the basic reproduction number R0 for a deterministic temperature-dependent transmission model between humans, Ae. aegypti, and Ae. albopictus, then develop a stochastic continuous-time Markov chain transmission model and determine the probability of disease extinction P0 for introduction by exposed or infectious humans, Ae. aegypti, or Ae. albopictus. We explore how both R0 and P0 depend on a number of variables, including temperature, vector species composition, vector-host ratio, and mosquito biting behavior. We discuss our results in the context of changes in climate and neighborhood-level spread of mosquito populations and dengue. Keywords: Stochastic epidemic model · Continuous-time Markov chain · Multitype branching process · Vector-borne disease · Aedes aegypti · Aedes albopictus Introduction For vector-borne diseases, such as dengue, risk of infection in humans depends strongly on the type and abundance of mosquito species present in an area, which in turn depends on environmental factors such as temperature. Aedes aegypti and Aedes albopictus are two vectors of dengue that share habitat requirements but differ in their biting behavior and ability to spread disease. These two mosquito species also differ in responses to temperature which impacts behavioral and life history traits differently for each species and plays a role in determining their distribution (Reinhold et al. 2018). Distributions of both species overlap globally, and in these regions of overlap, Ae. aegypti and Ae. albopictus compete for resources and habitat, primarily artificial containers, at the larval stage (Kraemer et al. 2019; Laporta et al. 2023; Ding et al. 2018). There is significant heterogeneity in their spatial distributions; Ae. aegypti is typically associated with hotter, more urban areas while Ae. albopictus is associated with cooler, more vegetated areas (O’dell et al. 2025). Ae. albopictus has effectively displaced Ae. aegypti from many locations due to its generally superior competitive ability at the larval stage (O’Meara et al. 1995; Hawley et al. 1987; Lounibos 2007; Moore et al. 1999; Fader 2016). However, Ae. aegypti has a higher tolerance to heat and is able to survive at higher temperatures than Ae. albopictus (Reinhold et al. 2018). Ae. aegypti are often found in urban heat islands, areas of cities with little to no vegetation and high heat index (de Jesús Crespo R, Rogers RE 2021; Sangermano et al. 2025). Beyond survival, temperature also impacts many aspects of the mosquito life-cycle and disease transmission process – fecundity, larval development rates, length of the gonotrophic cycle, disease incubation rates, and transmission rates (Reinhold et al. 2018; Mordecai et al. 2017). The presence of vegetation can lower temperatures, and also affect the basal resources available to mosquito larvae. Resource quality in turn impacts competition between Ae. aegypti and Ae. albopictus (Juliano 2010; Juliano et al. 2004; Yee et al. 2004; Camara et al. 2016; Lounibos and Juliano 2018). In particular, high quality resources alleviate the strong competitive effect of Ae. albopictus on Ae. aegypti and can promote coexistence of the two species (Beck et al. 2025). Within urban areas, heterogeneity in the type and amount of vegetation drives micro-environments differing in temperature and larval resources, which can result in heterogeneous spatial distributions of Ae. aegypti and Ae. albopictus, as was recently observed in New Orleans, Louisiana (de Jesús Crespo R, Rogers RE 2021). These micro-environments also exhibit significant variation in other factors influencing the vector-host ratio and vector-borne disease transmission, such as human population density and economic status. While both Ae. aegypti and Ae. albopictus can transmit disease to humans, Ae. albopictus has been suggested to be a less competent vector of dengue, as well as other diseases such as Zika virus, due to lower transmission probabilities per bite (Whitehorn et al. 2015; Lozano-Fuentes et al. 2019; Vega-Rúa et al. 2014). Ae. albopictus is also a more opportunistically biting species than Ae. aegypti and may bite alternative hosts if available, resulting in a lower frequency of feeding on human hosts (Lambrechts et al. 2010; Sivan et al. 2015). In contrast, Ae. aegypti might take multiple bites per gonotrophic cycle, with each bite potentially spreading disease (Scott et al. 1993; Rezza 2012). Beck et al. (2025) studied the effects of temperature and resource quality on the outcome of competition between Ae. aegypti and Ae. albopictus and determined the resulting risk of infectious disease using a deterministic ordinary differential equation model for dengue transmission between humans and two vector species. Risk was measured by the basic reproduction number R0, the expected number of secondary infections introduced into a susceptible population by a typical infected individual. They found that the outcome of competition (coexistence or competitive exclusion) and the resulting abundance of each species was dependent on both resource quality and the relative values of temperature-dependent fecundity and mortality rates. In turn, the value of R0 depended on vector composition, temperature-dependent vector mortality rates, temperature-dependent disease-related parameters (host-vector and vector-host transmission probabilities, extrinsic incubation period length), and vector biting behavior. The baseline biting rate for each species assumed all bites were on humans, but alternative biting behaviors were also explored. With a deterministic model, disease introduction either leads to an outbreak or the disease dies out depending on whether R0 is above or below 1. For a stochastic model where variables are integer valued, an outbreak does not necessarily occur upon disease introduction even when R0>1, and the probability that the disease dies out prior to resulting in a major outbreak depends on how the disease is introduced into the population. If the disease is introduced by an exposed individual, they may die prior to becoming infectious, and infectious individuals may recover or die prior to transmitting the disease. In regions presently naïve to dengue and related arboviruses, such as the southeastern United States, the primary means by which dengue is introduced into susceptible populations is exposed or infectious travelers returning from areas abroad where dengue is endemic. While Ae. aegypti and Ae. albopictus have limited flight ranges (Honório et al. 2003), it is also possible that disease could be introduced on a local neighborhood scale by exposed or infectious mosquitoes. The probability of vector-borne disease extinction for different means of disease introduction was previously explored by Allen and van den Driessche (2013) and Maliyoni (2020) in models with a single species of host and vector, finding the probability of extinction for introduction by an exposed mosquito to exceed that of an infectious mosquito and the probability of extinction for introduction by an infectious host typically in between. Lloyd et al. (2007) developed methodology for determining disease invasion probabilities in a model with an arbitrary number of host and vector species, focusing on the case of 1 vector and 2 host species, where hosts are identical except for their abundance and vector biting rates. They found the highest probability of disease extinction to occur when vectors have no host preference and the disease is introduced by an infectious vector. Horton and Robertson (2024) also explored a model with transmission between 2 hosts and a single vector, finding the probability of an outbreak was typically greater when the disease was introduced by an infectious vector compared to an infectious host, since vectors were always assumed to find a host to bite while the likelihood of a single host being bitten was dependent upon both the vector-host ratio and vector biting preferences. In this paper we extend the results of Beck et al. (2025) and explore how the local risk of dengue transmission depends on the composition of Ae. aegypti and Ae. albopictus, temperature, biting behavior, and how the disease is introduced into the population. Rather than explicitly modeling competition between vectors, we explore how dengue risk, as measured by the basic reproduction number and probability of disease extinction, changes with the overall vector abundance and the relative abundance of each species at different temperatures and for different assumptions about vector biting rates on humans. In the following sections, we present a deterministic transmission model that is a modification of Beck et al. (2025) with directly input vector abundance along with the basic reproduction number R0, then derive a stochastic continuous-time Markov chain (CTMC) model and compute the theoretical probabilities of disease extinction resulting from disease introduction by each type of infected individual. We show how disease risk depends on total vector abundance, the relative abundance of Ae. aegypti versus Ae. albopictus, temperature, biting behavior, and the type of infected individual introducing the disease. Deterministic Model Herein, we first describe the deterministic ordinary differential equations (ODE) model that underlies the stochastic modeling framework we describe in the next section. This mechanistic model is critical for the development of the continuous-time Markov chain model (CTMC) that we implement later in the study. A stage-structured deterministic model for vector-borne disease transmission incorporating competition between Ae. aegypti and Ae. albopictus at the juvenile stage and transmission between adult mosquitoes and humans was developed in Beck et al. (2025). The values of competition coefficients were assumed to depend upon the quality of available resources, while all other vector-related parameters were assumed to vary with temperature (fecundity, larval development rates, vector biting rates, extrinsic incubation periods, vector-host and host-vector transmission probabilities, and vector density-independent mortality rates). Human-related parameters were assumed to be independent of temperature. Here we modify this model by removing the juvenile equations and eliminating explicit competition between mosquito species. The total adult population of each species is instead prescribed, denoted by Nv1 for Ae. aegypti and Nv2 for Ae. albopictus. Susceptible vectors of each species are found by subtracting the exposed and infectious populations from the total population size, which is assumed to remain constant. Dengue transmission can occur when a susceptible vector of either species bites an infectious human or when an infectious vector of either species bites a susceptible human. After a disease-transmitting bite occurs, the exposed individual must survive the incubation period of the virus in order to become infectious. Model equations are shown in (1), with temperature-dependent parameters shown as functions of T. We assume constant temperature throughout our analysis, so model (1) is autonomous. Parameters are defined in Table 1 and baseline parameter values are given in Appendix A. and where Nh=Sh(t)+Ih(t)+Rh(t) is constant. Model (1) has the disease free equilibrium (Sv1∗,Ev1∗,Iv1∗,Sv2∗,Ev2∗,Iv2∗,Sh∗,Eh∗,Ih∗,Rh∗)=(Nv1,0,0,Nv2,0,0,Nh,0,0,0). Table 1Parameter and variable descriptions for the disease transmission model (1). Species 1 is Ae. aegypti and species 2 is Ae. albopictusSviSusceptible vectors of species iEviExposed vectors of species iIviInfected vectors of species iNviTotal vectors of species iShSusceptible hostsEhExposed hostsIhInfected hostsRhRecovered hostsNhTotal host populationbi(T)Biting rate of species i on human hostsρvi(T)Probability of transmission from humans to species iρhi(T)Probability of transmission from species i to humansσi(T)Extrinsic incubation rate for species iμi(T)Death rate of vector species iσhIntrinsic incubation raterhHuman recovery rateμhBirth/death rate of humans Parameter and variable descriptions for the disease transmission model (1). Species 1 is Ae. aegypti and species 2 is Ae. albopictus The basic reproduction number, R0, is the expected number of secondary cases arising from one typical infected individual introduced into a disease-free population. R0 for model (1) is the same as that for the stage-structured model explicitly incorporating competition in Beck et al. (2025): where and RV1H is the expected number of new human infections resulting from an exposed Ae. aegypti, RHV1 is the expected number of new Ae. aegypti infections resulting from an exposed human, RV2H is the expected number of new human infections resulting from an exposed Ae. albopictus, and RHV2 is the expected number of new Ae. albopictus infections resulting from an exposed human. We note that in the presence of only one species (e.g., Nv1=0 or Nv2=0), this expression for R0 reduces to the R0 obtained from a vector-host model with a single vector species. Stochastic Model We next derive a continuous-time Markov chain (CTMC) model from the deterministic transmission model (1). We assume the model is time-homogeneous and satisfies the Markov property, so the future state of the model is only dependent on the current state (Allen 2010; Maliyoni 2020). The state variables are the same as those in model (1), but are now integer-valued random variables. A list of all possible state transitions is created from model (1), and the rates at which those transitions occur are given in Table 2. Table 2Transitions and Rates for CTMC Model. For each of (Sv1, Ev1, Iv1, Sv2, Ev2, Iv2, Sh, Eh, Ih, Rh), a 1 represents an increase by 1, -1 represents a decrease by 1, and 0 represents no change in the variable’s state from time t to time (t+Δt). Temperature dependent parameters are shown as functions of T. Parameter values are given in Appendix A and details are in Beck et al. (2025)EventChange in StateRateExposure of Ae. aegypti(-1,1,0,0,0,0,0,0,0,0)b1(T)ρv1(T)IhNhSv1Infection of Ae. aegypti(0,-1,1,0,0,0,0,0,0,0)σ1(T)Ev1Death of Ev1(0,-1,0,0,0,0,0,0,0,0)μ1(T)Ev1Death of Iv1(0,0,-1,0,0,0,0,0,0,0)μ1(T)Iv1Exposure of Ae. albopictus(0,0,0,-1,1,0,0,0,0,0)b2(T)ρv2(T)IhNhSv2Infection of Ae. albopictus(0,0,0,0,-1,1,0,0,0,0)σ2(T)Ev2Death of Ev2(0,0,0,0,-1,0,0,0,0,0)μ2(T)Ev2Death of Iv2(0,0,0,0,0,-1,0,0,0,0)μ2(T)Iv2Recruitment of Sh(0,0,0,0,0,0,1,0,0,0)μhNhExposure of host(0,0,0,0,0,0,-1,1,0,0)b1(T)ρh1(T)Iv1+b2(T)ρh2(T)Iv2NhShInfection of host(0,0,0,0,0,0,0,-1,1,0)σhEhRecovery of host(0,0,0,0,0,0,0,0,-1,1)rhIhDeath of Sh(0,0,0,0,0,0,-1,0,0,0)μhShDeath of Eh(0,0,0,0,0,0,0,-1,0,0)μhEhDeath of Ih(0,0,0,0,0,0,0,0,-1,0)μhIhDeath of Rh(0,0,0,0,0,0,0,0,0,-1)μhRh Transitions and Rates for CTMC Model. For each of (Sv1, Ev1, Iv1, Sv2, Ev2, Iv2, Sh, Eh, Ih, Rh), a 1 represents an increase by 1, -1 represents a decrease by 1, and 0 represents no change in the variable’s state from time t to time (t+Δt). Temperature dependent parameters are shown as functions of T. Parameter values are given in Appendix A and details are in Beck et al. (2025) Theoretical Probability of Disease Extinction When susceptible populations are large enough, we can approximate the dynamics of the CTMC model near the disease-free equilibrium with a Galton-Watson multi-type branching process, a birth-death process for the number of infected individuals. Starting from a small number of infected individuals, the branching process has two possible outcomes, either quickly reaching the absorbing state of zero or growing exponentially. The former is called disease extinction (or a minor outbreak) and the latter is referred to as a major outbreak. Unlike R0 for the deterministic model, the probability of disease extinction, denoted by P0, depends on which of the 6 types of infected individuals initially introduces the disease - exposed human (type 1), infectious human (type 2), exposed Ae. aegypti (type 3), infectious Ae. aegypti (type 4), exposed Ae. albopictus (type 5), or infectious Ae. albopictus (type 6). Offspring probability generating functions (PGFs) for the distribution of new infections can be computed for each of these 6 infected types (Lahodny and Allen 2013). The offspring PGF for a single infected individual of type i (i=1,...,6) has the form: where u→=(u1,u2,u3,u4,u5,u6)∈[0,1]6 and Pi(k1,...,k6) represents the probability that an infected individual of type i creates exactly kj new infections (i.e. offspring) of type j, j=1,...,6, when the susceptible population is at the disease free equilibrium (Allen 2017; Allen and Lahodny 2012). The probabilities in the offspring PGFs are determined from the rates in Table 2 assuming the susceptible population is at the disease free equilibrium. An exposed human (type 1) will either survive the intrinsic incubation period of the disease to become infectious, or else die prior to becoming infectious. Survival results in the creation of one infectious human (type 2), so the probability of surviving the intrinsic incubation period, σhσh+μh, is multiplied by u2 to the power one in the offspring PGF for type 1. Death prior to the end of the intrinsic incubation period results in no new infections, so the probability μhσh+μh is multiplied by all uj raised to the power zero, or 1. Thus the offspring PGF for exposed humans, assuming Eh(0)=1 and Ih(0)=Ev1(0)=Iv1(0)=Ev2(0)=Iv2(0)=0, is An infectious human (type 2) can either transmit the disease to Ae. aegypti or Ae. albopictus, or else recover or die prior to infecting a mosquito. Transmission to Ae. aegypti results in the creation of two offspring - one of type 2, since the infectious human is still infectious after the transmission event occurs, and one of type 3, the newly exposed Ae. aegypti. Therefore the probability that an infectious human generates an Ae. aegypti infection, b1(T)ρv1(T)Nv1Nhb1(T)ρv1(T)Nv1Nh+b2(T)ρv2(T)Nv2Nh+rh+μh, is multiplied by u2 to the power one and u3 to the power one in the offspring PGF. Similarly, transmission to Ae. albopictus results in two infected offspring, one of type 2 and one of type 5, and recovering or dying prior to transmitting the disease results in zero infectious offspring. The offspring PGF for infectious humans, assuming Ih(0)=1 and Eh(0)=Ev1(0)=Iv1(0)=Ev2(0)=Iv2(0)=0, is then An exposed Ae. aegypti (type 3) can either survive the extrinsic incubation period, producing an infectious Ae. aegypti (type 4), or die prior to becoming infectious, resulting in zero infectious offspring. The offspring PGF for exposed Ae. aegypti, assuming Ev1(0)=1 and Eh(0)=Ih(0)=Iv1(0)=Ev2(0)=Iv2(0)=0, is then An infectious Ae. aegypti (type 4) can either transmit the disease to a susceptible human, producing two offspring (one of type 4 and one of type 1) or die before transmitting the disease, producing zero infectious offspring. The offspring PGF for infectious Ae. aegypti, assuming Iv1(0)=1 and Eh(0)=Ih(0)=Ev1(0)=Ev2(0)=Iv2(0)=0, is then An exposed Ae. albopictus (type 5) can either survive the extrinsic incubation period, producing an infectious Ae. albopictus (type 6), or die prior to becoming infectious, resulting in zero infectious offspring. The offspring PGF for exposed Ae. aegypti, assuming Ev2(0)=1 and Eh(0)=Ih(0)=Ev1(0)=Iv1(0)=Iv2(0)=0, is then An infectious Ae. albopictus (type 6) can transmit the disease to a susceptible human, thereby producing two offspring (one of type 6 and one of type 1) or die before transmitting the disease, producing zero infectious offspring. The offspring PGF for infectious Ae. albopictus, assuming Iv2(0)=1 and Eh(0)=Ih(0)=Ev1(0)=Iv1(0)=Ev2(0)=0, is then A 6×6 matrix M can be constructed from the offspring PGFs where the (j, i) entry of M is the expected number of infections of type j produced by an infected individual of type i (Allen and Lahodny 2012) (M is shown in Appendix B). The system (4)-(9) always has the fixed point u→=(1,1,1,1,1,1) since fi(1,1,1,1,1,1)=1, i=1,...,6. If the spectral radius of M is less than 1, the branching process is subcritical and this is the only fixed point. In this case the probability of disease extinction is 1. If the spectral radius of M is greater than 1, the branching process is supercritical and the system has an additional fixed point q→=(q1,q2,q3,q4,q5,q6), so fi(q→)=qi, with 01) if and only if ρ(M)<1(=1,>1). In the southern United States, dengue is likely to enter naive areas by travelers returning by plane from areas where the disease is endemic. Therefore type 1 or 2 individuals (exposed or infectious humans) are most likely to introduce the disease. Mosquitoes typically do not travel long distances, but since we are interested in determining risk at the neighborhood scale, it is possible a mosquito can travel between neighborhoods and introduce the disease to a new neighborhood. Therefore we consider results for disease introduction by both humans and mosquitoes. Results We explore how the basic reproduction number R0 for the deterministic model and the theoretical probability of disease extinction P0 for the stochastic CTMC model depend on vector community composition (overall vector abundance as well as relative abundance of each species), temperature, and biting behavior. The equation for R0 is given in (2), and the fixed points of equations (4)-(9) are found numerically in Matlab and used to compute P0 as described in Section 3.1. For every parameter set, the deterministic model yields a single estimate of risk, R0, while the stochastic model gives an estimate of risk, P0, following introduction by each of the 6 types of infected individuals. We note the probability of extinction for introduction by an exposed vector of either species will always be higher than the probability of extinction for introduction by an infectious vector of the same species, since an exposed vector may die prior to becoming infectious. However, the difference between the probabilities of extinction for introduction by exposed and infectious vectors depends on the length of the extrinsic incubation period of the disease and the mortality rate, which both vary with temperature. Since humans have a much longer lifespan than mosquitoes, the length of the intrinsic incubation period of the disease is short in comparison and the chance of exposed humans dying before becoming infectious is negligible. Therefore probabilities of extinction for exposed and infectious humans are essentially identical and we only show P0 for infectious humans in the results to follow. In the remainder of this section, we explore the effects of Aedes species composition, vector-host ratio, temperature, and biting behavior on disease risk. We first investigate the impacts of these variables independently while considering influences of temperature and biting rate assumptions before looking more closely at how combinations of these variables work together to impact disease risk. As in Beck et al. (2025), we consider either a single bite per gonotrophic cycle on humans for each species, referred to as unscaled biting rates, or 50% of bites on humans for Ae. albopictus and 2 bites per gonotrophic cycle on humans for Ae. aegypti, referred to as scaled biting rates. Vector species composition To explore the effect of vector species composition on disease risk, we hold the total vector abundance, Nv1+Nv2, constant at 200 and vary the percentage of the population comprised of Ae. aegypti mosquitoes, Nv1Nv1+Nv2. We set Nh=100 to maintain an overall vector-host ratio of 2, which is similar to ratios included in previous modeling studies (Manore et al. 2014; Robert et al. 2019). Results are shown in Figure 1 for three different temperatures – 25, 30, and 35∘C – as well as two different biting rate scenarios. The effect of Aedes species composition on disease risk. Total vector abundance is held constant (Nv1+Nv2=200) and the percentage of each species in the population is varied between 0 and 100. The basic reproduction number R0 and the probability of disease extinction P0 (for introduction by infectious humans, exposed and infected Ae. aegypti, exposed and infected Ae. albopictus) are shown for three different temperatures (25, 30, and 35∘C) and two biting rate scenarios (unscaled and scaled). Nh=100 for an overall vector-host ratio of 2. All additional parameter values are described in Appendix A (Color figure online) We find that as the percentage of Ae. aegypti mosquitoes in the population increases, the effect on R0 depends on both temperature and vector biting behavior. For unscaled biting rates, R0 decreases with the percentage of Ae. aegypti in the population for lower temperatures, but increases at higher temperatures. When biting rates are scaled, R0 increases with increased Ae. aegypti in the population regardless of temperature. For unscaled biting rates, temperature affects the direction of relationship between P0 and the percentage of Ae. aegypti in the population. As the percent of Ae. aegypti in the population increases, P0 increases for all types of introductions we considered at 25∘C but decreases at 35∘C. At 30∘C, increasing the percent Ae. aegypti in the population has little impact on P0 associated with introductions by mosquitoes despite leading to substantial decreases in R0, and only leads to a subtle increase in P0 following introductions by infectious humans. For scaled biting rates, the temperatures we consider here did not impact the directional relationship between the percent Ae. aegypti and P0. Notably, temperature has a strong influence on the relative values of P0 for introduction by each type of infected individual as the percent of Ae. aegypti in the population increases. Overall, the probability of extinction for introduction by humans varies more with percent Ae. aegypti than for introduction by any other type of infectious individual, especially at higher temperatures. At 35∘C, P0 for introduction by humans varies from 0.9 (100% Ae. albopictus) to 0.5 (100% Ae. aegypti) for unscaled biting rates and 1 (100% Ae. albopictus) to 0.25 (100% Ae. aegypti) for scaled biting rates. For the 2:1 vector-host ratio we consider, the probability of extinction is always lowest for introduction by a vector — Ae. albopictus for unscaled biting rates at 25∘C and 30∘C, and Ae. aegypti for unscaled biting rates at 30∘C and scaled biting rates at 25, 30, and 35∘C. Total Vector Abundance Next we explore the effect of increased overall vector abundance, Nv1+Nv2, and thus the vector-host ratio, on R0 and P0. We again consider three temperatures (25, 30, and 35∘C) and both scaled and unscaled biting rates, and we assume the population is always composed of 50% Ae. aegypti and 50% Ae. albopictus. R0 increases with vector-host ratio for all temperatures for both biting rate scenarios (Figure 2). For both scaled and unscaled biting rates, R0 is greatest at 30∘C, but for unscaled biting rates R0 is lowest at 35∘C while for scaled biting rates R0 is lowest at 25∘C. The effect of vector-host ratio on disease risk. Here Nh=100 and we increase the overall vector abundance Nv1+Nv2 from 100 to 1000 (so the vector/host ratio varies from 1 to 10) while maintaining an equal abundance of each vector species (50% Ae. aegypti and 50% Ae. albopictus). The basic reproduction number R0 and the probability of disease extinction P0 (for introduction by infectious humans, exposed and infected Ae. aegypti, exposed and infected Ae. albopictus) are shown for three different temperatures (25, 30, and 35∘C) and two biting rate scenarios (unscaled and scaled). All additional parameter values are described in Appendix A (Color figure online) At 35∘C, the highest probability of extinction is for introduction by Ae. albopictus, for both unscaled and scaled biting rates. The lowest probability of extinction is for introduction by Ae. aegypti at low vector-host ratios, but once the vector-host ratio becomes large enough, the lowest probability of extinction is for introduction by humans. Across all temperatures, we again see that the probability of extinction for introduction by infected humans varies more with vector-host ratio than for introduction by any other type of infected individual. Temperature Here we explore the effects of temperature on R0 and P0. We consider a vector-host ratio of 2, composed of either 1%, 50%, or 99% Ae. aegypti (with the remainder of the vector population in each case being composed of Ae. albopictus), for both unscaled and scaled biting rates (Figure 3). When temperatures are too low or too high, R0 falls below 1 and P0 is equal to 1 for introduction by all types of infectious individuals. The temperature for which R0 first reaches 1 increases with the percentage of Ae. aegypti in the population for unscaled biting rates and decreases with the percentage of Ae. aegypti in the population for scaled biting rates. At higher temperatures, the temperature for which R0 drops below 1 increases with the percentage of Ae. aegypti. The effect of temperature on disease risk. We set Nv1+Nv2=200 and Nh=100 for a baseline vector-host ratio of 2 with either a 1:99, 50:50, or 99:1 ratio of Ae. aegypti to Ae. albopictus and vary temperature from 15 to 40∘C. R0 and the probability of disease extinction P0 (for introduction by infectious humans, exposed and infected Ae. aegypti, exposed and infected Ae. albopictus) are shown for three different vector compositions (1%, 50%, and 99% Ae. aegypti) and two biting rate scenarios (scaled and unscaled). All additional parameter values are described in Appendix A (Color figure online) For intermediate temperatures, the lowest probability of extinction occurs for Ae. albopictus when biting rates are unscaled and for Ae. aegypti when biting rates are scaled. At higher temperatures the lowest probability of extinction occurs for introduction by Ae. aegypti for both scenarios. When Ae. aegypti only make up a small percentage of the population, the highest P0 values occur when the disease is introduced by humans. At higher temperatures (above 35∘C), P0 decreases with the percentage of Ae. aegypti in the population, as long as the disease is not introduced by Ae. albopictus, as Ae. albopictus cannot survive long at that temperature. Temperature and Species Composition In Figure 4, we further explore the combined impact of temperature and species composition on R0 and P0. Horizontal cross-sections at temperatures of 25, 30, and 35∘C recover Figure 1 while vertical cross sections at 1, 50, and 99% Ae. aegypti recover Figure 3. Contours where R0=P0=1 are shown in black. To simplify presentation of our results, we focus only on disease introduction by infectious individuals rather than exposed. The effect of temperature and species composition on disease risk. The basic reproduction number R0 and the probability of disease extinction P0 (for introduction by infectious humans, infectious Ae. aegypti, and infectious Ae. albopictus) are shown for temperatures from 15 to 40 C) and for 1% to 99% Ae. aegypti, and for two biting rate scenarios (unscaled and scaled). Black contours denote R0=1 or P0=1. Nv1+Nv2=200 and Nh=100 for a vector-host ratio of 2. All additional parameter values are described in Appendix A (Color figure online) When considering the impacts of both species composition and temperature, we find that R0 values were highest with scaled biting rates and 100% of the mosquito population composed of Ae. aegypti (Figure 4). These values of R0 occur between 30-35∘C. When biting rates are unscaled, R0 values are highest when Ae. albopictus make up 100% of the population and temperatures are between 25-31∘C, although the highest R0 values in these scenarios are still lower than those for scaled biting rates. In considering P0 values, the lowest probability of extinction occurs either when Ae. aegypti introduces infection into a population composed mostly of Ae. aegypti and biting rates are scaled or when Ae. albopictus introduces infection into a population mostly composed of Ae. albopictus and biting rates are unscaled. For any given temperature, the percent composition does not have a substantial impact on the probabilities of extinction following introductions by mosquitoes; however, more variation is observed following introductions by humans. For example, at around 25∘C, when biting rates are scaled, P0 is higher than 0.8 when the population is primarily composed of Ae. albopictus, but decreases to below 0.3 when the population is composed primarily of Ae. aegypti. We observe similar changes in the probability of extinction when temperatures are higher (around 35∘C) in the unscaled biting rates scenario. Vector biting behavior The actual percentage of Ae. albopictus bites on humans may vary depending on the availability of human hosts compared to alternative hosts independently of whether Ae. aegypti takes multiple bites per gonotrophic cycles. In Figure 5, we compare the effect of species composition (percent Ae. aegypti) and the percentage of Ae. albopictus bites on humans on R0 for different assumptions about the biting rate of Ae. aegypti (a1 either 1 (unscaled) or 2 (scaled) bites per gonotrophic cycle; see Appendix A.2) and for different temperatures (25, 30, or 35∘C). When a1=1, a horizontal cross section at 100% Ae. albopictus bites on humans corresponds to the previously explored unscaled biting rate scenario, and when a1=2, a horizontal cross section at 50% Ae. albopictus bites on humans corresponds to the scaled biting rate scenario. Recall that for these scenarios, R0 decreases with percent Ae. aegypti at 25 and 30∘C with unscaled biting rates, but increases with percent Ae. aegypti at 35∘C with unscaled biting rates and for 25, 30, and 30∘C with scaled biting rates. The effect of biting behavior on disease risk. R0 is plotted as a function of vector composition (percent Ae. aegypti in the population) and the percentage of Ae. albopictus bites on humans. We show results for three temperatures (25, 30, and 35∘C) for both unscaled (a1=1) and scaled (a1=2) assumptions for the biting rate of Ae. aegypti. We set Nv1+Nv2=200 and Nh=100 for a vector-host ratio of 2. All additional parameter values are described in Appendix A (Color figure online) In Figure 5 we again see that R0 increases with percent Ae. aegypti at 35∘C with unscaled Ae. aegypti biting rates and for 25, 30, and 30∘C with scaled Ae. aegypti biting rates regardless of the biting behavior of Ae. albopictus. However, when Ae. aegypti biting rates are unscaled, at the lower temperatures of 25 and 30∘C we see R0 decrease with percent Ae. aegypti if Ae. albopictus has a high enough biting rate on humans, but R0 will increase with percent Ae. aegypti if the percentage of Ae. albopictus bites on humans is small. At 30 C, when half of Ae. albopictus bites are on humans, we see very little effect on R0 as we change the composition of the population if Ae. aegypti are assumed to have 1 bite per gonotrophic cycle. If we assume a larger Ae. aegypti biting rate of 2 bites per gonotrophic cycle, then we see a much larger effect on R0 when we increase the percentage of Ae. aegypti in the population. R0 is always greater for a1=2 compared to a1=1, all else equal. For a given value of a1 and set species composition, we generally see R0 increase as the percent of Ae. albopictus bites on humans increases. The sensitivity of R0 to the percentage of Ae. albopictus bites on humans decreases as the percent Ae. aegypti increases, meaning the percent of Ae. albopictus in the population decreases. The exception to this is that at higher temperatures (35∘C) where the lifespan of Ae. albopictus is short, we see little change in R0 with the biting rate of Ae. albopictus regardless of the species composition of biting rate of Ae. aegypti. Discussion In this paper, we developed a stochastic continuous time Markov chain model to extend our previous work exploring the effects of temperature and food quality on the outcome of competition between Ae. aegypti and Ae. albopictus, and the resulting risk of dengue using a deterministic model (Beck et al. 2025). In that work, we showed that at a given temperature, the abundance of each species present was seen to vary with food quality, habitat availability, and potentially initial conditions, and that these variables also impacted disease risk as measured by the basic reproduction number, R0. Here, we further investigate the impacts of temperature, species composition, vector host ratio, biting behavior of both mosquito species, and source of initial introduction on the potential for disease transmission as measured by R0 and the probability of disease extinction, P0. Every set of parameter values considered gives a single estimate of R0, based on disease introduction by a typical infected individual, and an estimate of P0 for introduction by each of the 6 types of infected individuals. We find that R0 and P0 are generally inversely related for each type of infectious individual, with P0 decreasing as R0 increases or visa versa. However, the relative magnitudes of P0 for disease introduction by the different types of infected individuals vary greatly with model parameters, even when estimates of R0 are similar. We show that overall, disease risk following introduction by humans is much more sensitive to vector composition than for introduction by vectors. This occurs because the model assumes an infectious vector will always find a host to bite, and the presence of other vectors of either species does not affect that process. However, if a human introduces the disease, then the vector composition—what vectors are present, what their biting rates are, and how many other potential humans to bite—are all important for determining whether that human will be bitten by a mosquito, and what the probability of passing the disease on to a vector would be. These results have particular implications for interconnected regions with heterogeneity in temperature and mosquito species distributions. For example, in the city of New Orleans and surrounding suburbs, both Ae. aegypti and Ae. albopictus are present, but species abundance and composition varies by neighborhood (de Jesús Crespo R, Rogers RE 2021). This region, too, has many different microclimates, with a mixture of urban heat islands interspersed with cooler areas with more vegetation. Seasonal and long-term changes in temperature lead to shifts in species distributions, which, as we show here, could have important impacts for disease risk, with those impacts dependent on a number of other factors. For example, in a population with an average temperature of 25∘C where Ae. aegypti and Ae. albopictus coexist in relatively similar ratios, a shift in the species composition that favors Ae. albopictus may lead to either larger risk of transmission and outbreaks (assuming unscaled biting rates) or a lower risk (assuming scaled biting rates) (Figure 4). If temperature also increases, this change in risk could be more pronounced. It is likely that changes in temperature will also be accompanied by changes in species composition. For example, Ae. albopictus has higher fecundity and lower mortality rates than Ae. aegypti across most temperatures, but Ae. aegypti is more tolerant of higher temperatures, and previous work has shown that temperature-dependent risks associated with Ae. aegypti are higher than those of Ae. albopictus (Mordecai et al. 2017; Beck et al. 2025; Reinhold et al. 2018). With increased temperatures, areas currently dominated by Ae. albopictus could be replaced by Ae. aegypti, so that increases in temperature are accompanied by increases in the relative abundance of Ae. aegypti which may lead to higher levels of transmission (scaled biting rates) or potentially lower rates of transmission (unscaled biting rates) (Laporta et al. 2023; Ryan et al. 2019). While human movement is the primary driver of both short and long range mosquito-borne disease transmission (Stoddard et al. 2013; Horton and Robertson 2024), spread among interconnected neighborhoods in urban and/or metropolitan areas may be driven in part by movement of mosquitoes (Reiner et al. 2014; Chao et al. 2013). We show that the probability of outbreak is influenced by the species of the infected mosquito and that this probability is influenced by numerous factors including temperature and species composition. This could be of particular importance for informing mitigation strategies in urban areas where neighboring regions support different species. For example, if a cooler area with more vegetation borders a warmer, urban heat island, migration of infected Ae. aegypti mosquitoes from the urban heat island to the neighboring region could spark local outbreaks even if the Ae. albopictus population in that region cannot support sustained transmission of a pathogen. Throughout this work, we emphasize the impact that vector biting behavior has on our results. In many cases, the assumption of whether biting rates are unscaled or scaled drastically changes disease risk. In Figure 5, we explore this further to show just how influential biting rate assumptions are on the results in combination with species composition and temperature. While Ae. aegypti may take multiple bites per gonotrophic cycle, the biting rate of the more opportunistic species Ae. albopictus on humans is more likely to vary spatially with the availability of human hosts compared to alternative hosts (i.e. in rural versus urban areas) (Niebylski et al. 1994; Fikrig et al. 2021). R0 is particularly sensitive to the biting rate of Ae. albopictus at 30∘C, where varying the percentage of bites that Ae. albopictus take on humans drives changes in R0 from less than 1, with fewer bites on humans, to greater than 4, when most of their bites occurred on humans. The effect on R0 is greatest when Ae. albopictus is the dominant species in the population, and even more pronounced when Ae. aegypti is assumed to take multiple bites per gonotrophic cycle. These results emphasize that better understanding of biting behaviors of Ae. aegypti and Ae. albopictus is paramount to studying the impacts of temperature and competition on mosquito-borne disease spread. Our results have important implications for dengue control, particularly in regions where both Ae. aegypti and Ae. albopictus are implicated in transmission. We show that species composition influences R0 as well as P0, with the specific effects on both dependent on biting rate assumptions (Figure 4). With the increasing promise of genetic and biological control strategies (Alphey et al. 2013; Benedict 2021; Oliva et al. 2021), species-specific control could result in changes to species composition in areas where both Aedes subspecies coexist. This could ultimately have the impact, for example, of reducing transmission at temperatures where Ae. albopictus are more dominant if they are replaced by Ae. aegypti (Figure 4, unscaled biting rates), or increasing transmission if Ae. aegypti replace Ae. albopictus populations in other scenarios (Figure 4, scaled biting rates). Our work shows that careful consideration is necessary when considering species-targeted control measures in regions where both Ae. aegypti and Ae. albopictus are part of the dengue transmission cycle. Importantly, our results highlight that introductions of dengue by humans are most affected by species composition (Figure 4) and changes in vector-host ratio (Figure 2) when compared to introductions by either mosquito species. This result agrees with previous work suggesting that human movement plays a larger role in dengue spread than mosquito movement (Harish et al. 2024; Reiner et al. 2014; Stoddard et al. 2013), and suggests that human-focused control measures such as testing, quarantining, and vaccines may have the largest impact on reduction of transmission and the potential for outbreaks. Our work makes some simplifying assumptions that should be further explored. Namely, we calculated R0 and the probability of extinction P0 assuming a disease-free equilibrium at constant temperatures. A next step would be to incorporate seasonally varying or diurnal variations in temperature. We also acknowledge that some of the vector abundance and temperature scenarios we consider may be somewhat unrealistic, such as populations comprised of mostly Ae. albopictus at temperatures where the lifespan is very short and the species would likely not persist. However, since there are many factors affecting the outcome of competition between Ae. aegypti and Ae. albopictus that are not all well understood, this approach lets us focus on the implications of vector composition on disease risk, with computations for R0 and P0 retaining the temperature dependence for vector life-history and disease-related parameters. Taken together, our results emphasize the importance of temperature, species composition, vector-host ratio, mosquito biting behaviors, and source of initial introduction in the emergence and spread of dengue in a population where its two primary vectors share habitat. These results are critical in planning short- and long-term mitigation strategies in interconnected populations with high heterogeneity in human population, temperatures, and mosquito habitat availability. As we show here, small shifts in key variables may lead to important differences in species composition and disease risk, but our model and further investigation of this topic through similar models will be useful for better preparing for potential scenarios resulting from changes in these variables. Although this work has focused primarily dengue and Aedes subspecies, the framework presented here can be adapted to other mosquito-borne diseases where multiple mosquito species transmit pathogens and/or compete for resources. Parameters Baseline model parameters are given below, see Beck et al. (2025) for additional details. Vector mortality rates As in Beck et al. (2025), the mortality rates, measured in days-1, for Ae. aegypti and Ae. albopictus, respectively, are taken to be the inverse of the average of curves fit to the 50% laboratory and field data for adult lifespan in Brady et al. (2013), with a minimum lifespan of 1 day assumed: and where Vector biting rates As in Beck et al. (2025), we define the vector biting rates, measured in bites per mosquito per day, using Brière functions from Mordecai et al. (2017). We multiply the Brière functions by constant scaling parameters a1 and a2 to obtain the biting rates of Ae. aegypti and Ae. albopictus on humans, given by b1(T) and b2(T), respectively: A value of ai less than 1 indicates that species i takes bites on non-human hosts, and a value of ai greater than 1 indicates that species i takes multiple bites per gonotrophic cycle. For the unscaled biting rate scenario, a1=a2=1, and for the scaled biting rate scenario, a1=2 and a2=0.5. Host to vector transmission probability per bite As in Beck et al. (2025), we define the per-bite probability of transmission from host to vector using Brière functions from Mordecai et al. (2017) with ρv1 modified above 31.04∘C to take on the value of the Brière function when T=31.04: Vector to host transmission probability per bite As in Beck et al. (2025), we define the per-bite probability of transmission from vector to host using Brière functions from Mordecai et al. (2017) with ρh1 modified above 31.22∘C to take on the value of the Brière function when T=31.22: Vector extrinsic incubation rate As in Beck et al. (2025), we define the extrinsic incubation rate, measured in days-1, for dengue using Brière functions from Mordecai et al. (2017): Host parameters As in Beck et al. (2025), the host parameters are assumed to be temperature independent with values taken from Huber et al. (2018). Host intrinsic incubation period: σh=1/5.9 (days-1) Human recovery rate: rh=1/5 (days-1) Birth/Death rate of humans: μh=1/(70∗365) (days-1) Relating the Deterministic and Stochastic Thresholds From the deterministic model (1), we can define the vector of infectious classes X=(Eh,Ih,Ev1,Iv1,Ev2,Iv2). The linearization at the disease free equilibrium is where and The basic reproduction number R0 of the deterministic model is the spectral radius of the second generation matrix FV-1. From the offspring PGFs for the 6 types of infectious individuals (4)-(9), we can define a 6-by-6 expectation matrix, M, where the (j, i) entry ∂fi∂uj evaluated at the fixed point (u1,u2,u3,u4,u5,u6)=(1,1,1,1,1,1) gives the expected number of new infections of type j produced by an infected individual of type i (Allen and Lahodny 2012): where and M is non-negative and irreducible, and satisfies the relationship (M-I)W=F-V, where I is the 6×6 identity matrix, W=diag(σh+μh,b1(T)ρv1(T)Nv1Nh+b2(T)ρv2(T)Nv2Nh+rh+μh,σ1(T)+μ1(T),b1(T)ρh1(t)+μ1(T),σ2(T)+μ2(T),b2(T)ρh2(t)+μ2(T)), and F and V are the matrices above in equations (14) and (15). Therefore the conditions for the Threshold Theorem from Allen and van den Driessche (2013) to apply are met, guaranteeing R0<1(=1,>1) if and only if ρ(M)<1(=1,>1), as well as the existence of a unique fixed point of the offspring PGFs (4)-(9), (q1,q2,q3,q4,q5,q6)∈(0,1)6 when the branching process is supercritical. Funding This work was funded by the National Science Foundation (Grant no. DMS1950015). MAR was supported by a Burroughs Wellcome Fund Climate and Health Interdisciplinary Award (BWF 1022621). Data Availibility Statement No datasets were generated or analyzed in this work. MATLAB code used to generate figures is available upon request. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement No datasets were generated or analyzed in this work. MATLAB code used to generate figures is available upon request.