In this study we used data from 410 Ae. aegypti
and Ae. albopictus
adult female survival experiments to construct a model of temperature and age-dependent survival. The best-fitting model was an age and temperature dependent GAM, which we combined with field-based mortality to estimate the distribution of longevity of mosquitoes in the field. These models should be interpreted in light of the limitations to the data used and the modelling methods. Of the data compiled, there were relatively few experiments conducted at more extreme temperatures, resulting in higher uncertainty in these temperature ranges. Considering the implications these extreme temperatures have for the geographic range of the mosquito, more data is necessary for better defining that range. Mosquito survival may also vary geographically due to variation in the genetic background of local mosquito populations. Our analysis incorporates experimental data from a wide variety of locations and conditions but, given the paucity of MRR data at climatic extremes, it is possible that estimates of field-based mortality may not be representative, particularly in locations where mosquitoes may be more tolerant of extreme climates. However, observations suggest that adult survival is one of the few bionomics of Ae. aegypti
that remains consistent among geographically disparate populations [67
]. North American samples of Ae. albopictus
were over-represented in our laboratory data (91%) which may overestimate thermal tolerance at the lower temperature limits due to evidence of Ae. albopictus
adaptation in temperate climates [53
]. Perhaps the most important data limitation is that there are very few field mortality experiments upon which to estimate mortality outside of the laboratory, where arbovirus transmission actually occurs. More data would allow better estimation of field mortality and allow relaxation of the assumption that external contributors to mortality are not age-related. An extension should include alternatives to MRR approaches that might be more sensitive for estimating survival of older potentially infectious mosquitoes, for example, using transcript levels of an age-associated gene [70
] or measuring mosquito infection rates [71
It is also worth considering the scope of the model we developed. While it may suggest that adult survival is possible even at low temperatures, factors limiting the development of the immature stages may preclude the establishment of an adult population. For adult Aedes
, there may also be additional temperature-based limits to survival. Below 14-15°C, Ae. aegypti
experiences reduced mobility and struggles to imbibe blood [18
]. As observations (at tropical temperatures) suggest Ae. aegypti
cannot survive longer than 2-3 days without a blood meal [73
], extended periods of time below 14-15°C are likely to result in complete mortality. We extended our predictions below these temperatures because limited time spent at these colder temperatures, such as a few hours in the early morning, may not confer a lasting negative effect on survival, however, these effective mortality limits must be considered when integrating this into a wider Aedes
/DENV transmission model.
Here we have shown that GAMs captured more of the variation in survival between different experiments than conventional, parametric models of mosquito mortality, including the Gompertz and logistic models which previously provided the best fit of the options examined [27
]. In our analysis we did not consider frailty-based mortality models, however non-parametric approaches, such as GAM, offer more flexibility than parametric models, suitable predictive power and straight-forward implementation in a variety of statistical software packages. We have also emphasised the importance of including random effects in analyses of this type as it has allowed us to comment on the effect of specific experimental treatments, such as feeding regime, and thus guide laboratory and field experimental design. Furthermore adding a random effect at the study level allowed us to comment on the generalisability of any one experiment to a wider context.
The relative fit of the GAM indicated the presence of age-dependent survival. This age-related effect may be related to senescence or other factors, but appears to be less important in the wild. On average we found that mortality due to external causes was significantly greater (4.5 times for Ae. aegypti and 3.0 times for Ae. albopictus) than mortality due to temperature and age at mid-ranging temperatures (20-30°C) indicating that few mosquitoes will be alive at older ages by the time senescence measurably impacts mortality. Results from our analysis indicate that across different temperature regimes the assumption of age-independent mortality for Ae. aegypti and Ae. albopictus in the field is likely to be appropriate. It may be possible that senescence acts at different ages in the laboratory than it does in the field, however, quantifying this would require unfeasibly large mosquito cohorts, or innovative new techniques to measure mosquito survival in the field. As a result, the statistical model presented here provides a novel approach for quantifying the effects of age-dependent survival among wild mosquitoes.
The greater tolerance of lower temperatures observed for Ae. aegypti
compared to Ae. albopictus
appears at odds with their observed geographic distribution. This finding can be explained, however, if we consider the adaptations of the egg stages of each mosquito. Aedes albopictus
eggs are able to undergo diapause allowing the species to persist during cold winter temperatures that are unfavourable to adult survival [74
]. Aedes aegypti
shows only limited adaptation to egg stage survival in unfavourable periods [76
] and there may, therefore, be greater selection pressure for thermal tolerance at the adult stage to resist diurnal and inter-seasonal variations in temperature.
The seasonal variation and geographic limits of Aedes
transmitted viruses are intrinsically linked to the seasonality and geographic limits of mosquito populations required for their transmission. It is, therefore, important that the survival of adult Aedes
is well estimated using appropriate data and considering the variety of conditions they may encounter in nature. Quantifying how changes in survival translate to dengue transmission potential is also important and requires integration with models of the DENV extrinsic incubation period, which is also sensitive to temperature [78
]. Characterising the interaction between mortality and extrinsic incubation could be used further to explore entomological components of seasonal forcing among Aedes
-borne viruses, which may help inform strategies for disease prevention and control.
Given that dengue is a disease that affects over half the world’s population [12
] with an estimated 390 million new infections per year [10
], improvements in existing approaches and new strategies to control the disease are needed. Depending on the situation, both the type of vector control used and how it is implemented must be optimised to achieve the biggest reduction in dengue burden. Modelling and mapping are two important tools for addressing both of these issues, identifying risks and opportunities for control [34
]. Considerable efforts have already been undertaken to re-examine many of the fundamental assumptions of early Aedes
and dengue transmission models [6
]. The next generation of models is likely to incorporate significant advances, including new frameworks for modelling survival of adult mosquitoes. The adult Aedes
survival models developed here can be integrated into these new modelling approaches with the aim of refining the global distribution of Aedes
vectors, guiding seasonal vector control efforts, modelling dengue transmission and developing early warning systems to prevent dengue epidemics [83