Modelling adult Aedes aegypti and Aedes albopictus survival at different temperatures in laboratory and field settings

Abstract

Background: The survival of adult female Aedes mosquitoes is a critical component of their ability to transmit pathogens such as dengue viruses. One of the principal determinants of Aedes survival is temperature, which has been associated with seasonal changes in Aedes populations and limits their geographical distribution. The effects of temperature and other sources of mortality have been studied in the field, often via mark-release-recapture experiments, and under controlled conditions in the laboratory. Survival results differ and reconciling predictions between the two settings has been hindered by variable measurements from different experimental protocols, lack of precision in measuring survival of free-ranging mosquitoes, and uncertainty about the role of age-dependent mortality in the field.

Methods: Here we apply generalised additive models to data from 351 published adult Ae. aegypti and Ae. albopictus survival experiments in the laboratory to create survival models for each species across their range of viable temperatures. These models are then adjusted to estimate survival at different temperatures in the field using data from 59 Ae. aegypti and Ae. albopictus field survivorship experiments. The uncertainty at each stage of the modelling process is propagated through to provide confidence intervals around our predictions.

Results: Our results indicate that adult Ae. albopictus has higher survival than Ae. aegypti in the laboratory and field, however, Ae. aegypti can tolerate a wider range of temperatures. A full breakdown of survival by age and temperature is given for both species. The differences between laboratory and field models also give insight into the relative contributions to mortality from temperature, other environmental factors, and senescence and over what ranges these factors can be important.

Conclusions: Our results support the importance of producing site-specific mosquito survival estimates. By including fluctuating temperature regimes, our models provide insight into seasonal patterns of Ae. aegypti and Ae. albopictus population dynamics that may be relevant to seasonal changes in dengue virus transmission. Our models can be integrated with Aedes and dengue modelling efforts to guide and evaluate vector control, better map the distribution of disease and produce early warning systems for dengue epidemics.

Figure 1

Schematic overview of the methods. Green parallelograms indicate input data, orange rectangles show processing or modelling steps, blue diamonds show decision steps and red boxes show output analysis and models (dotted shows intermediate, unbroken line shows final outputs). MRR = Mark-release-recapture.

Figure 2

Relative likelihood of four different parametric models for Aedes adult female survival data over a range of constant temperatures. The models included are i) a two parameter Log-logistic model (shape and scale), ii) a two parameter Gompertz model (shape and rate), iii) a one parameter Exponential model (rate) and iv) a two parameter Weibull model (shape and scale).

Figure 3

Examples of parametric and non-parametric model fit. Open circles show Ae. aegypti survival data under controlled laboratory conditions from Joy et al.[64](A) and Yang et al.[21](B), two experiments that show contrasting survival curve shape. Parametric models are shown as dashed lines and the non-parametric GAM is shown as a solid orange line.

Figure 4

The distribution of adult female Aedes aegypti and Aedes albopictus survival across a range of temperatures under laboratory conditions (A and B) and field conditions (C and D). Colours from red to yellow show survival from 100% - 1% of the population remaining. Grey indicates <1% of the population remaining. Dotted blue lines show the limits for 50% and 95% of the original population remaining.

Figure 5

The distribution of uncertainty of the laboratory model prediction. Colours from blue to beige show the interquartile range (IQR) in predictions from 200 bootstrap runs of the laboratory model (A and B). This uncertainty is then combined with the field data uncertainty quantified by 200 bootstrap runs of MRR data to give the IQR predictions for the field survival model (C and D). Red dotted lines of the 50% and 95% of the population remaining are added for reference.