https://orcid.org/0000-0002-0665-9047
Figures
Abstract
Climate factors such as temperature and rainfall influence vector-borne disease dynamics. Consequently, climate change is affecting the global distribution of diseases such as dengue. Natural climate variability (NCV), which adds noise to the climate signal, alters the future climate trajectory, but is rarely analysed in climate-health analyses. In this study, extending a published model of climate suitability for Ae. aegypti, we consider a climate-sensitive dengue transmission model. Using 100 climate projections up to 2100 from the Community Earth System Model (a climate model including climate change and NCV), each run under the same shared socioeconomic pathway scenario, we generate 100 equally plausible simulations of the future basic reproduction number of dengue globally and the population at risk. We quantify the difference in transmission suitability between the most-suitable and least-suitable projections (i.e., uncertainty due to NCV) and demonstrate that NCV affects future suitability for transmission, dominating epidemiological parameter uncertainty in many locations. While the global population at risk from dengue (and, as we show, other vector-borne diseases) is expected to increase, NCV affects its precise value. Our findings demonstrate that NCV should be routinely incorporated into climate-sensitive epidemiological projections, and a key output of our research is the provision of a methodological framework for this. Accounting for NCV will enable public health policy decisions to be informed by the range of possible future outcomes.
Citation: Kaye AR, Sun L, Hart WS, Hurrell JW, Tildesley MJ, Thompson RN (2026) The effect of natural climate variability on future vector-borne disease suitability: A mathematical modelling study. PLOS Clim 5(8): e0000877. https://doi.org/10.1371/journal.pclm.0000877
Editor: Jamie Males, PLOS Climate, UNITED KINGDOM OF GREAT BRITAIN AND NORTHERN IRELAND
Received: February 23, 2026; Accepted: July 15, 2026; Published: August 18, 2026
Copyright: © 2026 Kaye et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data generated or analysed during this study, including computing code for reproducing our results, are available at: https://github.com/KayeARK/Climate_Change_NCV_Dengue.
Funding: This research was funded by the EPSRC via the Mathematics for Real-World Systems CDT (EP/S022244/1; ARK and MJT) and the Wellcome Trust via a Digital Technology Development Award (226057/Z/22/Z; WSH and RNT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Climate-sensitive infectious diseases are widespread and negatively impact public health. For example, a large proportion of the global population is at risk from dengue [1]. The most successful dengue vector is Aedes aegypti, and the ability of this vector to thrive depends on environmental conditions. Climate factors (e.g., temperature, rainfall and humidity) and socioeconomic factors affect vector characteristics such as the vector lifespan [2,3], egg survival probability [4,5] and development time [6]. These factors in turn affect the vector population size, altering the vector-borne pathogen transmission potential. Additionally, pathogen characteristics, such as viral replication rates, are affected by climate factors. The climate-sensitive nature of both vector ecology and pathogen dynamics means that the locations at risk from vector-borne disease are being altered by climate change [7,8].
Climate-sensitive disease modelling is an emerging field, with studies using climate datasets to generate ecological and epidemiological projections. For example, the spatial distributions of Ae. aegypti and Ae. albopictus were projected under climate change by Kraemer et al. [9,10] using a machine learning model. Other studies have involved using mechanistic models to explore spatial variations in the vector-borne pathogen transmission potential. For example, Mordecai et al.[11] and Ryan et al.[7] considered a formula for the basic reproduction number () derived from a host-vector model and used it in combination with global temperature simulations to project
for a range of pathogens. A similar approach can be taken to extend the dependence of
to other climate variables, such as rainfall [12]. These studies have provided insights into how climate change is affecting the global spatial distribution of vector-borne disease.
In previous climate-sensitive ecological and epidemiological modelling studies, two main sources of uncertainty in the future climate have been considered. Specifically, the impacts of climate model uncertainty (i.e., variation between different models of climate dynamics) and scenario uncertainty (i.e., uncertainty about the shared socioeconomic pathway, or SSP, that will occur) [13] have been analysed. However, a third source of uncertainty exists, and is often neglected by epidemiological modellers [14]. Specifically, natural climate variability (NCV, sometimes referred to as internal climate variability) refers to climate fluctuations arising due to interactions between Earth system components (e.g., atmosphere, oceans and land). NCV is well-studied by climate scientists and adds noise to climate projections [13,15,16]. One way to quantify the extent of NCV is to run multiple climate projections, each initialised from slightly different initial conditions, under a single climate model and SSP. Due to the chaotic nature of the climate, the initial perturbations grow, and the climate conditions at any specific future timepoint can be viewed as a sample from the ensemble of climate projections at that time. Climate projection ensembles (typically 50–100 simulations) from individual climate models have been created specifically for studying NCV.
In a previous study [8], NCV was shown to generate substantial variations in ecological projections in the context of climate suitability for Ae. aegypti. Here, we extend that work to consider the role of NCV in a climate-sensitive epidemiological system (specifically, dengue transmission by Ae. aegypti). We demonstrate that NCV affects epidemiological projections by considering a large ensemble of projections from a single climate model, and use a climate-sensitive epidemiological model to calculate the number of months that are projected to be suitable for any local dengue transmission () or sustained local dengue transmission (
) in locations globally each year in future in each climate projection. We examine the variation in these quantities due to NCV, comparing uncertainty due to NCV against uncertainty resulting from imprecise knowledge of epidemiological parameter values. We project the global population at risk from dengue in each climate projection, again demonstrating the substantial effect of NCV.
The primary goals of this article are to demonstrate the general principle that NCV affects epidemiological projections and to provide a methodological framework for accounting for NCV in climate-sensitive epidemiological projections. To demonstrate the broad applicability of our results, in addition to our main analyses we conduct further analyses using alternative climate-sensitive epidemiological models from the literature, considering dengue and malaria transmission by various vectors and different climate scenarios. In all our analyses, we find that both anthropogenic climate change and NCV play a key role in future global climate suitability for vector-borne disease. It is therefore essential to analyse both these features of the future climate when generating climate-health projections, particularly if the analyses are to be used to inform public health policy.
Methods
Climate projections
Climate data were obtained from the Community Earth System Model (CESM) version 2 [17,18] Large Ensemble Community Project (LENS2) [19]. This dataset consists of 100 equally plausible climate simulations from 1850 to 2100. Differences between projections arise due to small variations in the initial conditions [19]. The impact of the externally forced anthropogenic climate change signal can be obtained by averaging out the “noise” of NCV (i.e., by examining the mean of the climate simulations). Real-world climate trajectories are substantially less smooth than this ensemble mean (see Fig S1 from Kaye et al.[8]), illustrating the need to consider NCV in addition to anthropogenic climate change when projecting future suitability for vector-borne disease.
The 100 CESM projections used in most of our analyses were generated under the SSP3-7.0 scenario. This scenario characterises a future with fragmented international cooperation and limited environmental sustainability efforts, resulting in moderate-to-high challenges for climate change mitigation and relatively high greenhouse gas emissions [20]. The dataset provides the values of projected climate variables across a global grid of 192 latitude and 288 longitude values. We considered the average air temperature (at a reference height of two metres above the Earth’s surface) and total rainfall each month in each location for each CESM projection individually from 2020-2100. The ecological parameters of our model (see below) depend on daily rainfall rather than monthly rainfall; we therefore converted the average monthly rainfall value to its corresponding average daily value.
When testing the robustness of the conclusions derived from our results, we considered an alternative climate change scenario (SSP2-4.5) under which 19 CESM projections are available. This scenario represents a more moderate climate future, characterised by socioeconomic development and emissions trends broadly following historical patterns [20].
Human population density
For 2020, we used human population size estimates from WorldPop [21]. Specifically, we used the unconstrained global mosaics dataset, which provides a global grid of data at a spatial resolution of 30 arcseconds (yielding pixels that are approximately 1km-by-1km at the equator).
We used human population size projections from Wang et al.[22] for 2060 and 2100, again at a spatial resolution of 30 arc. In our main analyses, we used projections generated under SSP3-7.0 for consistency with the climate dataset. When we instead analysed climate projections under SSP2-4.5, we used population projection data based on SSP2-4.5.
When we projected the value of for dengue in each population data pixel (see below), we assumed that the temperature and rainfall are determined by the CESM pixel in which the centre of the population data pixel lies. Only pixels with human population density exceeding 20 inhabitants km-2 (using population data from WorldPop [21]) were included in our analyses, since we assumed that host-vector transmission cannot occur at lower population densities. This leads to a predicted
of zero for locations with population densities below this threshold. Similar thresholds have been implemented previously [23,24].
Epidemiological model
The climate-sensitive epidemiological model that we used is an extended version of an existing ecological model of Ae. aegypti population dynamics [8]. Specifically, we extended the previous model to include vector-borne disease dynamics. In our approach, Ae. aegypti dynamics in each location are represented by an ordinary differential equation model in which vectors are compartmentalised according to their life cycle stage: eggs, aquatic (larvae or pupae) and adults (Fig 1A; adults are then divided according to their infection status as described below). Each model parameter was assumed to depend on either temperature or rainfall; we used the functional forms of the ecological parameters from Kaye et al.[8] (see Figs S2–S6 of that publication).
A. Schematic illustrating the life cycle of Ae. aegypti and the epidemiological dynamics of dengue; different vector stages and host/vector infection statuses are represented by different compartments in the transmission model. B. The 2.5th percentile estimate as a function of rainfall and temperature, assuming a population density of 15,000 people km-2 (similar to Buenos Aires, which experiences substantial dengue transmission), as derived from the transmission model (see Methods). C. Identical to panel B, but showing the median
estimate. D. Identical to panel B, but showing the 97.5th percentile
estimate.
In the current study, we assigned the adult vectors to compartments according to their infection status. At any time, each adult vector is either susceptible, exposed or infectious (Fig 1A). Once infectious, vectors remain infectious throughout their lives [25]. Hosts are compartmentalised similarly, according to whether they are susceptible, infectious or recovered at any time (Fig 1A). We omitted the exposed compartment for hosts, as this does not affect the instantaneous value of . Host susceptibility, vector susceptibility, the vector bite rate and the extrinsic incubation period are assumed to be temperature-dependent. The infectious period of infected hosts is assumed to be independent of temperature.
To infer the relationship between temperature and the temperature-dependent epidemiological parameters, we adapted the modelling framework of Mordecai et al.[11] We assumed a functional form of the relationship between temperature and each temperature-dependent model parameter and then fit these functions to data from that study using Markov chain Monte Carlo (MCMC). Further details about the model and model fitting are provided in the Supplementary Material. Model fitting results are shown in Figs A–E in S1 Appendix and Table A in S1 Appendix.
We calculated at each temperature-rainfall combination under our epidemiological model using the next generation method (see Supplementary Material). Our MCMC fitting procedure enabled quantification of uncertainty in our model parameter values. We calculated the corresponding uncertainty in
at each temperature-rainfall combination, considering its median, 2.5th and 97.5th percentile values (Fig 1B–1D). In most of our analyses, we used the median estimate, although we considered the full range of possible
values when we compared the effect of epidemiological uncertainty on projected suitability for dengue transmission with uncertainty due to NCV.
Results
Quantification of 
Using our epidemiological model, we first computed the median (Fig 1C), 2.5th (Fig 1B) and 97.5th (Fig 1D) percentile values across a range of temperature and rainfall conditions, assuming a fixed human population density of 15,000 people km-2. The dengue transmission potential is maximised at intermediate temperature and rainfall levels (see Fig 1B–1D). At temperatures deviating from the optimum, conditions are less conducive to transmission due, for example, to vector survival or viral transmission probabilities declining under suboptimal thermal conditions. Similarly, rainfall can limit transmission suitability through different mechanisms: low rainfall may reduce the availability of vector breeding sites, whereas high rainfall can flush vector larvae from their habitats.
Future climate suitability for dengue transmission
For each pixel in the human population dataset, we compared the number of months suitable for sustained dengue transmission () using climate data from the CESM in 2020 and 2100. We used median
estimates and computed the number of months with
for each CESM projection individually. The mean of the resulting values in 2100 is shown in Fig 2A. We also verified that locations in which climate conditions were predicted to be suitable for sustained transmission for at least three months of the year (on average across all climate simulations) in 2020 aligned with locations with recorded Ae. aegypti populations [10] (Fig F in S1 Appendix).
A. The mean number of months suitable for sustained dengue transmission () in each 1km-by-1km pixel in 2100. Magnified versions of this plot focused on Europe (Fig I, panel C, in S1 Appendix), the Americas (Fig J, panel C, in S1 Appendix), Asia (Fig K, panel C, in S1 Appendix) and Africa (Fig L, panel C, in S1 Appendix) are shown in the Supplementary Material. B. Locations in which mean suitability for sustained dengue transmission is projected to increase (red) or decrease (blue) by more than one month between 2020 and 2100. In both panels, white areas indicate regions with a population density below 20 people km-2; we assume that transmission cannot occur at population densities below this threshold due to the limited availability of hosts. Coastline data were obtained from the MATLAB Mapping Toolbox dataset, coastlines.mat (dataset documentation: https://www.mathworks.com/help/map/mapping-toolbox-data-sets.html; licence information: https://www.mathworks.com/license/mll/license.txt).
The difference in the mean number of months with between 2020 and 2100 was calculated, and the pixels with either an increase (Fig 2B - red) or decrease (Fig 2B - blue) by more than one month were identified. Some equatorial regions are projected to become less suitable for dengue transmission in 2100 compared to 2020 due to temperature values exceeding the optimum for transmission (Fig 2B). However, this neglects the possibility of vector adaptation to gradually changing temperatures permitting survival at higher temperatures. Our main finding in Fig 2B is therefore that a polewards expansion of suitable climate conditions for dengue is projected, aligning with the results of previous studies [7]. The same polewards expansion was observed when we did not account for changing human population densities, instead fixing them at their values in 2020 (Fig G in S1 Appendix).
Effect of natural climate variability
While the results in Fig 2 indicate how suitability for sustained dengue transmission might be expected to change, considering the mean result across all climate simulations does not capture the range of possible outcomes. To consider NCV, we compared the minimum number of months projected to be suitable for sustained dengue transmission in 2100 in each pixel (over all CESM projections; Fig 3A) with the corresponding maximum number of months (Fig 3B). Neither panel represents a single climate projection globally; this figure instead demonstrates the wide variation in the number of months suitable for sustained transmission in each location between CESM projections.
A. The minimum number of months projected to be suitable for sustained dengue transmission in 2100. B. The maximum number of months projected to be suitable for sustained dengue transmission in 2100. In each pixel, the CESM projection corresponding to the smallest (panel A) or largest (panel B) number of suitable months in 2100 is chosen. In both panels, white areas indicate regions with a population density below 20 people km-2. Coastline data were obtained from the MATLAB Mapping Toolbox dataset, coastlines.mat (dataset documentation: https://www.mathworks.com/help/map/mapping-toolbox-data-sets.html; licence information: https://www.mathworks.com/license/mll/license.txt).
The difference between Fig 3A and 3B (Fig H in S1 Appendix) highlights the importance of NCV. For example, the minimum number of months projected to be suitable for sustained dengue transmission by Ae. aegypti in 2100 in London is zero (Fig 3A), compared to a worst-case scenario of four months (Fig 3B). In some locations, this difference is even more substantial, with a difference of 11 months in locations such as Ho Chi Minh City (Vietnam), Nabire (Indonesia) and Colombo (Sri Lanka). Magnified versions of Fig 3A and 3B focused on Europe (Fig I, panels A and B, in S1 Appendix), the Americas (Fig J, panels A and B, in S1 Appendix), Asia (Fig K, panels A and B, in S1 Appendix) and Africa (Fig L, panels A and B, in S1 Appendix) are shown in the Supplementary Material.
NCV is responsible for substantial year-on-year variation in climate conditions, and this translates into substantial uncertainty in transmission suitability in any given year. We therefore expect that some regions in which dengue transmission does not occur currently may experience sporadic years with suitable transmission conditions earlier than suggested by ensemble mean projections. This implies that public health policy advisors in those locations should prepare for outbreaks sooner, since transmission may arise earlier than might be expected based on anthropogenic climate change alone.
To demonstrate the robustness of the results shown in Fig 3, we conducted supplementary analyses. First, we repeated our analysis using CESM simulations generated under SSP2-4.5, which represents a less severe climate change scenario than SSP3-7.0 (Fig M in S1 Appendix). Second, we considered an alternative model [11] of climate suitability for dengue transmission by Ae. aegypti (Fig N, panels A and B, in S1 Appendix and Fig O, panels A and B, in S1 Appendix). Third, we considered models of climate suitability for dengue transmission by Ae. albopictus [11] (Fig N, panels C and D, and Fig O, panels C and D, in S1 Appendix) and malaria transmission by different vectors [26,27] (Figs P and Q in S1 Appendix). Fourth, we considered results generated over a shorter timescale (up to 2060; Fig R in S1 Appendix). In every scenario, NCV generated substantial variation in the number of months projected to be suitable for vector-borne disease transmission in a range of locations.
Effect of epidemiological parameter uncertainty
While NCV generates uncertainty in climate-sensitive epidemiological projections, there is also uncertainty about the ecological and epidemiological parameter values in our model. This generates additional uncertainty in the value of (Fig 1B–1D). To test the relative importance of these uncertainty sources, we projected
for dengue in each month in 2100 using each CESM climate projection. We then calculated the ratio of uncertainty due to NCV (i.e., between projections) compared to uncertainty due to imprecise knowledge of the ecological and epidemiological parameter values (Fig 4; see Supplementary Material for further details). Under the model considered here, we found that in many locations NCV is the dominant uncertainty source (green pixels in Fig 4B). Parameter uncertainty tends to be the dominant source of uncertainty in locations where climatic conditions are already highly suitable for dengue transmission throughout much of the year. In those regions, differences between climate projections may have relatively limited effects on transmission suitability. Conversely, in other areas, variations in temperature or rainfall associated with NCV can substantially alter the duration of the year for which climate conditions permit transmission. While epidemiological parameter uncertainty could theoretically be reduced by improving our understanding of the role of climate in shaping vector ecology and epidemiology, NCV represents an uncertainty source that is exceptionally challenging to reduce due to the chaotic nature of the climate system. We repeated this analysis using the CESM projections under SSP2-4.5, again finding that NCV represents the dominant uncertainty source in most locations (Fig S in S1 Appendix).
A. The relative importance of uncertainty in the value of each month in 2100 arising from NCV as opposed to epidemiological parameter uncertainty. This is computed as the logarithm of the ratio between the length of the 95% credible interval due to NCV and the length of the 95% credible interval due to epidemiological parameter uncertainty (see Supplementary Material for further details). B. The dominant uncertainty source in the projected value of
in 2100. Green and pink pixels indicate locations where more uncertainty is attributable to NCV and epidemiological parameter uncertainty, respectively. In both panels, white areas indicate regions with a population density below 20 people km-2. Coastline data were obtained from the MATLAB Mapping Toolbox dataset, coastlines.mat (dataset documentation: https://www.mathworks.com/help/map/mapping-toolbox-data-sets.html; licence information: https://www.mathworks.com/license/mll/license.txt).
Population at risk
In addition to quantifying climate suitability for dengue transmission, we considered the future human population at risk. We calculated the total number of individuals globally each month living in pixels with (potential for any local transmission; Fig 5A) and
(potential for sustained local transmission; Fig 5B) in 2020, 2060 and 2100.
A. The total population in pixels where . B. The total population in pixels where
. In both panels, the population at risk is computed separately for each CESM projection. Points represent median estimates and error bars show 95% credible intervals (over all 100 climate projections).
Generally, climate change and shifts in human population densities are expected to lead to a larger population at risk from dengue in future than currently (for example, in Fig 5B, averaging across the year, the median population at risk is 2.67 billion in 2020, 3.93 billion in 2060 and 4.60 billion in 2100). The estimate of 2.67 billion people at risk from dengue in 2020 is in broad agreement with previous estimates indicating that 2.5-3 billion people live in dengue at-risk regions [28]. Large error bars in Fig 5 indicate that NCV again generates substantial uncertainty. To isolate the effect of climate differences alone, we repeated the analysis shown in Fig 5 but fixing human population densities at their 2020 values, again finding that NCV affects the projected population at risk (Fig T in S1 Appendix).
To verify the wider applicability of this result, we ran additional analyses. We repeated the results in Fig 5 for an alternative model [11] of climate suitability for Ae. aegypti-borne dengue transmission, for alternative models of dengue transmission by Ae. albopictus [11] and for malaria transmission by different vectors [26,27] (Fig U in S1 Appendix). We also considered CESM simulations generated under SSP2-4.5 (Figs V and W in S1 Appendix). In every case, we found that climate change is expected to increase the global population at risk compared to 2020, and NCV generates substantial uncertainty in projections.
Discussion
Vector-borne diseases are responsible for substantial morbidity and mortality worldwide, particularly in low- and middle-income countries. An improved understanding of the interplay between climate and epidemiological projections can inform monitoring and intervention strategies. We constructed a climate-sensitive model of dengue transmission by Ae. aegypti and computed (Fig 1B–1D) under different climate conditions. Using climate data from the CESM [17,18] from 2020-2100 and human population density projections, we generated global
projections.
Similar projections of transmission suitability have been generated in previous studies [7] using a re-scaled version of where the possibility of transmission corresponds to
. By analysing human population density data, we evaluated where and when
, corresponding to the potential for sustained transmission. Like in previous studies [7], we identified a polewards expansion in suitable conditions for any dengue transmission (
), and in addition we found that the number of months suitable for sustained transmission (
) is expected to increase at northerly and southerly latitudes under climate change (Fig 2).
The key feature of the research presented here is the use of a large ensemble of climate projections from a single model to highlight the effect of NCV on epidemiological projections. In the context of climate-sensitive epidemiological modelling, this is a methodological advance that we hope will be adopted by other researchers to account for NCV in future studies. We found that NCV generates substantial variation in the number of months with suitable climate conditions for sustained dengue transmission () in future years (Fig 3). In particular, the minimum and maximum number of months with suitable transmission conditions in 2100 varied substantially (by up to 11 months in some locations) between different, but equally plausible, climate projections. We demonstrated the general nature of this finding by considering an alternative model of Ae. aegypti-borne dengue transmission (Fig N, panels A and B, in S1 Appendix), as well as models of dengue transmission by Ae. albopictus (Fig N, panels C and D, in S1 Appendix) and malaria transmission (Fig P in S1 Appendix). We also analysed all models under SSP2-4.5 (Figs M, O and Q in S1 Appendix).
NCV was not the only uncertainty source in our model. We compared the extent of uncertainty in projections arising from NCV and imprecise knowledge of ecological and epidemiological parameter values (Fig 4 and Fig S in S1 Appendix). Under the model considered here, we found that NCV is the dominant uncertainty source in most locations.
As well as projecting , we quantified the global population at risk, and verified that our estimate for 2020 is in agreement with estimates in prior literature. This quantity is expected to increase due to a combination of climate change and population growth. In 2100, a dip is projected during the northern hemisphere’s summer months (Fig 5). This arose because temperatures are projected to exceed the optimal conditions for vectors in some locations (see Fig 1B–1D). Since the majority of the global population lives in the northern hemisphere, this leads to a projected drop in the population at risk from June to September. We note, however, that this does not account for the possibility that vectors adapt to rising temperatures [29], which may occur over the timescale studied here. Adaptation would likely act to increase the population at risk by widening the geographical range of climate suitability for vectors. As in our
projections, NCV drives substantial variation in projections of the population at risk in specific years (see error bars in Fig 5). This conclusion again held over a range of systems and under the alternative SSP2-4.5 scenario (Figs U–W in S1 Appendix).
Our findings have important implications for vector-borne disease outbreak preparedness, as well as vector and pathogen surveillance strategies. In regions where vector-borne disease outbreaks do not currently happen but may in future, NCV could advance the timing of outbreak occurrence. In particular, NCV can produce some years with substantially higher transmission suitability than would be inferred from ensemble mean climate projections alone, increasing the likelihood that outbreaks occur earlier than anticipated under long-term climate change trends. Public health planning should therefore consider not only gradual climatic shifts, but also the potential for episodic high-risk years driven by NCV. Preparedness measures that should be considered as a result include strengthened vector surveillance, improved diagnostic and clinical capacity, public awareness campaigns focused on vectors and vector-borne diseases, and the development of outbreak response plans in regions where competent vectors may emerge and establish. In areas where vector populations are not yet present but may establish in future, vector surveillance at likely points of entry is especially important. When vectors invade, interventions such as eliminating breeding sites will be critical for delaying vector establishment and limiting pathogen transmission risks. Monitoring and response capacity may be strengthened through coordination across sectors, including climate expertise, vector control programmes and public health systems.
Our work represents a detailed study of the effect of NCV on climate-sensitive epidemiological projections. Previous studies have projected both vector suitability (for Ae. aegypti and Ae. albopictus [9,10,30]) and the transmission potential of various pathogens [7,31]. While temperature has often been the only climate variable considered, some studies have considered rainfall and humidity [9,30], alongside socioeconomic factors such as urbanicity [9] and gross domestic product [30]. While other climate uncertainty sources have been considered (specifically, climate model uncertainty and scenario uncertainty, as described in the Introduction), NCV has often been neglected in climate-sensitive disease modelling analyses. It has, however, been analysed in a limited number of ecological studies. Notably, Monaghan et al.[32] considered climate suitability for Ae. aegypti, albeit only analysing a small number of climate projections (compared to the 100 projections that we used for SSP3-7.0). Kaye et al.[8] projected climate suitability for Ae. aegypti using the same climate simulations as considered here. While that previous research showed that NCV affects ecological projections, in the current study we have demonstrated further that NCV must be considered when generating epidemiological projections.
Like any modelling study, our research involved assumptions. Our aim was to demonstrate that NCV introduces uncertainty into epidemiological projections rather than to make precise predictions about transmission suitability. We used threshold values to characterise locations and time periods in which there is either a risk of any transmission (
) or sustained transmission (
). However, more complex metrics could be considered, and in principle simulations of dynamic models could be used to assess transmission risks (rather than relying on “steady state” summary statistics such as
). However, the development of complex dynamic models with the capability of generating accurate projections would likely require both enhanced data and an increased understanding of the impacts of a range of factors on ecological and epidemiological dynamics. For example, climate variables such as humidity or wind speed (which affects the vector biting rate) could be analysed, and human behavioural factors (e.g., water storage methods) that affect Ae. aegypti population dynamics could be accounted for. The climate data used here have a spatial resolution of 100km-by-100km. Future work could use finer-grained spatial data to examine the effects of microclimates on vector and disease dynamics.
While such extensions represent future research targets, we contend that our main conclusion is likely to apply irrespective of the complexity of the epidemiological model being analysed: NCV merits consideration when making climate-sensitive epidemiological projections. As we have shown, NCV can be analysed in climate-health projections by using an ensemble of climate projections from a single climate model generated under a single SSP.
In summary, NCV generates substantial uncertainty in climate-sensitive epidemiological projections. We hope that our research will inspire modellers to account for the full range of uncertainty present in climate-health projections, including NCV. Variations in outbreak risks due to NCV should be considered by policy advisors when preparing for future vector-borne disease outbreaks.
Acknowledgments
The CESM project is supported primarily by the U.S. National Science Foundation (NSF). Computing and data storage resources, including the Cheyenne supercomputer (https://doi.org/10.5065/D6RX99HX), were provided by the Computational and Information Systems Laboratory (CISL) at the NSF National Center for Atmospheric Research (NCAR). NSF NCAR is sponsored by the U.S. National Science Foundation. Thanks to members of the Zeeman Institute for Systems Biology and Infectious Disease Epidemiology Research at the University of Warwick, and the Wolfson Centre for Mathematical Biology at the University of Oxford (particularly the Infectious Disease Modelling group), for useful discussions about this research.
References
- 1. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, et al. The global distribution and burden of dengue. Nature. 2013;496(7446):504–7. pmid:23563266
- 2. Yang HM, Macoris MLG, Galvani KC, Andrighetti MTM, Wanderley DMV. Assessing the effects of temperature on the population of Aedes aegypti, the vector of dengue. Epidemiol Infect. 2009;137(8):1188–202. pmid:19192322
- 3. Brady OJ, Johansson MA, Guerra CA, Bhatt S, Golding N, Pigott DM, et al. Modelling adult Aedes aegypti and Aedes albopictus survival at different temperatures in laboratory and field settings. Parasit Vectors. 2013;6:351. pmid:24330720
- 4. Tesla B, Demakovsky LR, Mordecai EA, Ryan SJ, Bonds MH, Ngonghala CN, et al. Temperature drives Zika virus transmission: evidence from empirical and mathematical models. Proc R Soc Ser B Biol Sci. 2018;285(1884):20180795. pmid:30111605
- 5. Juliano SA, O’Meara GF, Morrill JR, Cutwa MM. Desiccation and thermal tolerance of eggs and the coexistence of competing mosquitoes. Oecologia. 2002;130(3):458–69. pmid:20871747
- 6. Alto BW, Juliano SA. Precipitation and temperature effects on populations of Aedes albopictus (Diptera: Culicidae): implications for range expansion. J Med Entomol. 2001;38(5):646–56. pmid:11580037
- 7. Ryan SJ, Carlson CJ, Mordecai EA, Johnson LR. Global expansion and redistribution of Aedes-borne virus transmission risk with climate change. PLoS Negl Trop Dis. 2019;13(3):e0007213. pmid:30921321
- 8. Kaye AR, Obolski U, Sun L, Hart WS, Hurrell JW, Tildesley MJ, et al. The impact of natural climate variability on the global distribution of Aedes aegypti: a mathematical modelling study. Lancet Planet Health. 2024;8(12):e1079–87. pmid:39674197
- 9. Kraemer MUG, Reiner RC Jr, Brady OJ, Messina JP, Gilbert M, Pigott DM, et al. Past and future spread of the arbovirus vectors Aedes aegypti and Aedes albopictus. Nat Microbiol. 2019;4(5):854–63. pmid:30833735
- 10. Kraemer MUG, Sinka ME, Duda KA, Mylne AQN, Shearer FM, Barker CM, et al. The global distribution of the arbovirus vectors Aedes aegypti and Ae. albopictus. Elife. 2015;4:e08347. pmid:26126267
- 11. Mordecai EA, Cohen JM, Evans MV, Gudapati P, Johnson LR, Lippi CA, et al. Detecting the impact of temperature on transmission of Zika, dengue, and chikungunya using mechanistic models. PLoS Negl Trop Dis. 2017;11(4):e0005568. pmid:28448507
- 12. Parham PE, Michael E. Modeling the effects of weather and climate change on malaria transmission. Environ Health Perspect. 2010;118(5):620–6. pmid:20435552
- 13. Hawkins E, Sutton R. The potential to narrow uncertainty in regional climate predictions. Bull Am Meteorol Soc. 2009;90:1095–108.
- 14. Hart WS, Hurrell JW, Kaye AR, Chand M, Keeling MJ, Thompson RN. Climate variability amplifies the need for vector-borne disease outbreak preparedness. Proc Natl Acad Sci U S A. 2025;122(34):e2507311122. pmid:40825121
- 15. Deser C, Hurrell JW, Phillips AS. The role of the North Atlantic Oscillation in European climate projections. Clim Dyn. 2017;49:3141–57.
- 16. Ricke KL, Caldeira K. Natural climate variability and future climate policy. Nat Clim Chang. 2014;4:333–8.
- 17. Danabasoglu G, Lamarque JF, Bacmeister J. The community earth system model version 2 (CESM2). J Adv Model Earth Syst. 2020;12:e2019MS001916.
- 18. Hurrell JW, Holland MM, Gent PR, et al. The community earth system model: a framework for collaborative research. Bull Am Meteorol Soc. 2013;94:1339–60.
- 19. Rodgers KB, Lee S-S, Rosenbloom N, Timmermann A, Danabasoglu G, Deser C, et al. Ubiquity of human-induced changes in climate variability. Earth Syst Dyn. 2021;12(4):1393–411.
- 20. Meinshausen M, Nicholls ZRJ, Lewis J, Gidden MJ, Vogel E, Freund M, et al. The shared socio-economic pathway (SSP) greenhouse gas concentrations and their extensions to 2500. Geosci Model Dev. 2020;13(8):3571–605.
- 21. WorldPop. Global High Resolution Population Denominators Project - Funded by The Bill and Melinda Gates Foundation (OPP1134076). School of Geography and Environmental Science, University of Southampton; Department of Geography and Geosciences, University of Louisville; Departement de Geographie, Universite de Namur and Center for International Earth Science Information Network (CIESIN), Columbia University; 2018.
- 22. Wang X, Meng X, Long Y. Projecting 1 km-grid population distributions from 2020 to 2100 globally under shared socioeconomic pathways. Sci Data. 2022;9(1):563. pmid:36097271
- 23. Zardini A, Menegale F, Gobbi A, Manica M, Guzzetta G, d’Andrea V, et al. Estimating the potential risk of transmission of arboviruses in the Americas and Europe: a modelling study. Lancet Planet Health. 2024;8(1):e30–40. pmid:38199719
- 24. Guzzetta G, Montarsi F, Baldacchino FA, Metz M, Capelli G, Rizzoli A, et al. Potential risk of dengue and chikungunya outbreaks in Northern Italy based on a population model of Aedes albopictus (Diptera: Culicidae). PLoS Negl Trop Dis. 2016;10(6):e0004762. pmid:27304211
- 25. Guzman MG, Gubler DJ, Izquierdo A, Martinez E, Halstead SB. Dengue infection. Nat Rev Dis Primers. 2016;2:16055. pmid:27534439
- 26. Villena OC, Ryan SJ, Murdock CC, Johnson LR. Temperature impacts the environmental suitability for malaria transmission by Anopheles gambiae and Anopheles stephensi. Ecology. 2022;103(8):e3685. pmid:35315521
- 27. Mordecai EA, Paaijmans KP, Johnson LR, Balzer C, Ben-Horin T, de Moor E, et al. Optimal temperature for malaria transmission is dramatically lower than previously predicted. Ecol Lett. 2013;16(1):22–30. pmid:23050931
- 28. Khan MB, Yang Z-S, Lin C-Y, Hsu M-C, Urbina AN, Assavalapsakul W, et al. Dengue overview: an updated systemic review. J Infect Public Health. 2023;16(10):1625–42. pmid:37595484
- 29. Couper LI, Farner JE, Caldwell JM, Childs ML, Harris MJ, Kirk DG, et al. How will mosquitoes adapt to climate warming? Elife. 2021;10:e69630. pmid:34402424
- 30. Liu-Helmersson J, Brännström Å, Sewe MO, Semenza JC, Rocklöv J. Estimating past, present, and future trends in the global distribution and abundance of the arbovirus vector Aedes aegypti under climate change scenarios. Front Public Health. 2019;7:148. pmid:31249824
- 31. Messina JP, Brady OJ, Golding N, Kraemer MUG, Wint GRW, Ray SE, et al. The current and future global distribution and population at risk of dengue. Nat Microbiol. 2019;4(9):1508–15. pmid:31182801
- 32. Monaghan AJ, Sampson KM, Steinhoff DF, Ernst KC, Ebi KL, Jones B, et al. The potential impacts of 21st century climatic and population changes on human exposure to the virus vector mosquito Aedes aegypti. Clim Change. 2018;146(3–4):487–500. pmid:29610543