Past and future spread of the arbovirus vectors Aedes aegypti and Aedes albopictus

The global population at risk from mosquito-borne diseases—including dengue, yellow fever, chikungunya and Zika—is expanding in concert with changes in the distribution of two key vectors: Aedes aegypti and Aedes albopictus. The distribution of these species is largely driven by both human movement and the presence of suitable climate. Using statistical mapping techniques, we show that human movement patterns explain the spread of both species in Europe and the United States following their introduction. We find that the spread of Ae. aegypti is characterized by long distance importations, while Ae. albopictus has expanded more along the fringes of its distribution. We describe these processes and predict the future distributions of both species in response to accelerating urbanization, connectivity and climate change. Global surveillance and control efforts that aim to mitigate the spread of chikungunya, dengue, yellow fever and Zika viruses must consider the so far unabated spread of these mosquitos. Our maps and predictions offer an opportunity to strategically target surveillance and control programmes and thereby augment efforts to reduce arbovirus burden in human populations globally.


Model validation results
In areas where there is time-series data and empirical human movement data available such as the United States we find that great-circle distance is capturing the spread process of the vector with a univariate correlation between origin of continental spread in the southern Using a comprehensive probabilistic geographic spread model we reconstructed the spatial spread process, specifically the timing and likelihood of the vectors persistence. For Ae.
aegypti, in addition to great-circle distance, county-to-county commuting trips, and mobility  Figure   6). Covariates best explaining the spatial spread are direct neighbour adjacency, great-circle distance, radiation, and gravity models. We repeated the analysis for Ae. albopictus in Europe and also show robust predictive ability (mean AUC = 0.90, 95% CI: 0.88 -0.93, Supplementary Figure 6). Covariates retained in the model were the same for both models except the greater risk of infestation when a district was two degrees away from an infested district.
The environmental niche modelling was performed for both species. The 2015 baseline model using environmental variables and species occurrence data resulted in high predictive power for Ae. aegypti (AUC = 0.865, 95% CI = 0.85 -0.87) and Ae. albopictus (AUC = 0.90, 95% CI = 0.88 -0.91). Variables that explained most of the variation of the species environmental niche was the species-specific temperature suitability indices (58% for Ae. aegypti and 49% for Ae. albopictus) followed by maximum precipitation (16% for Ae. aegypti and 19% for Ae. albopictus). A full list is available in Table 1 and 2.

Comparison to previously published work and CDC and ECDC records
Aedes albopictus in Europe: Overall, there is a high concordance between the current reported distribution of Ae. albopictus and the present day predicted distribution as determined by our model. Both suggest that the primary range of Ae. albopictus is concentrated around the Mediterranean coastline with extensions across southern France.
Our model-based predictions also suggest a high probability of presence in Bulgaria, Romania and Hungary, regions for which ECDC data is patchy or missing entirely.
Furthermore, we predict relatively low levels of suitability for areas for which ECDC has observed introduction but no subsequent establishment, such as central France, southern Germany and southern England. One noticeable discrepancy is the reported absence of Ae. albopictus in Portugal, a country we predict to be both highly suitable and highly connected to the species' established range. This may be due to the biogeographical barrier of the arid interior of the Iberian peninsula which may desiccate eggs and juveniles in transit.  Table 9 based on the inclusion criteria: i) gave projections of the future distribution of either species and ii) made predictions at the scale of whole country or US state or higher.
Methodologically, past approaches to project the future distribution of Aedes have largely fallen into two camps: i) agent based, mechanistic dynamic models or ii) species distribution models using the Maximum Entropy (MAXENT) algorithm. These methods, in isolation, are only suitable for estimating suitability for Aedes populations and do not attempt to estimate whether it would be possible for any Aedes species to spread into this newfound niche. In addition, the Boosted Regression Tree species distribution model used here presents a considerable advantage over the MAXENT algorithm for this application due to its more explicit handling of biases in reporting 2 , a key feature of spatially variable Aedes surveillance.
All previous modelling attempts also only consider changes in climatological factors, principally temperature and precipitation (Supplementary Table 9). Previous mapping studies have shown that the current global distribution of Ae. aegypti and Ae. albopictus is also determined by socioeconomic factors such as urbanization 3 . Given the importance of these variables, we have included them in our assessment of the future distribution of these mosquito species.
In Europe, the majority of these climatological projection models suggest Ae. albopictus will spread through France to cover much of western Europe and southeast England with some suggesting extensions as far north as southern Sweden. In contrast, our own mapping efforts suggest much more limited spread over this time period with modest increases in range along the fringes of the European distribution of Ae. albopictus.
In the USA, previous predictions for Ae. albopictus suggest the species will spread throughout the country by 2080 with the exception of the arid southern state of Arizona and the northern states of north Dakota and Wisconsin 4,5 , while much of the south east of the USA from Texas to Virginia is predicted to be suitable for Ae. aegypti at least for some times of the year 6 .
For Ae. aegypti in the USA our predictions are similar to previous efforts, however we do predict further spread to some major cities in the northeast and inland California that previous models have not. For Ae. albopictus while our estimates agree on the northern limit of the species with those from previous approaches, we predict that the species' distribution will be largely restricted East of the Rocky Mountains.
Elsewhere, all past approaches agree that there will be considerable change in suitability in Asia (in particular India, Thailand and Vietnam) for both species with both increases and decreases in suitability, but there is a lack of consensus on which areas will see expansion Supplementary Figure 13: Histograms of predicted suitability for locations of observations. This data was used to derive cut-offs to build binary presence and absence maps for Ae. aegypti (a) and Ae. albopictus (b) species so the relevant population at risk and area expansion could be calculated.