Empirical Dynamic Modelling Identifies different Responses of Aedes Polynesiensis Subpopulations to Natural Environmental Variables

Florian Grziwotz,Jakob Friedrich Strauß,Chih-Hao Hsieh,Arndt Telschow

doi:10.1038/s41598-018-34972-w

Abstract

To control mosquito populations for managing vector-borne diseases, a critical need is to identify and predict their response to causal environmental variables. However, most existing attempts rely on linear approaches based on correlation, which cannot apply in complex, nonlinear natural systems, because correlation is neither a necessary nor sufficient condition for causation. Applying empirical dynamic modelling that acknowledges nonlinear dynamics on nine subpopulations of tiger mosquitos from three neighbouring reef islets of the Raiatea atoll, we identified temperature, precipitation, dew point, air pressure, and mean tide level as causal environmental variables. Interestingly, responses of subpopulations in close proximity (100–500 m) differed with respect to their causal environmental variables and the time delay of effect, highlighting complexity in mosquito-environment causality network. Moreover, we demonstrated how to explore the effects of changing environmental variables on number and strength of mosquito outbreaks, providing a new framework for pest control and disease vector ecology.

Highlights

There are two main strains of research aiming to examine how mosquito populations respond to climate changes
When analysing the mosquito dynamics for each of the nine sampling sites separately using convergent cross mapping (CCM), we found that temperature, precipitation, dew point, air pressure, and mean tide level are causal variables for the mosquito populations (Fig. 1; Supplementary Table S2)
We found that sampling site S7 at Toamaro island is affected by four climate variables, site S8 by only one, and site S9 by none

Summary

Introduction

There are two main strains of research aiming to examine how mosquito populations respond to climate changes. Time series data of mosquito abundance in natural systems are analysed for statistical associations with climate variables. Majority of the studies linking natural mosquito abundances with climate variables employed linear correlation analyses. Abundance data of lab and field populations are best described by nonlinear mathematical models[19,20,21] All these studies suggest that, in general, mosquito-environment interactions are complex, nonlinear, and inter-dependent; as such, linear methods are not adequate for these questions. The motus have similar climate, but differ in vegetation and human activity as well as in larval competition and average body size of adult mosquitoes This exceptionally high temporal and spatial resolution gives unique opportunities for studying population dynamics of mosquitos. Our objectives are (1) to identify the critical environmental variables that causally affect mosquito populations using convergent cross mapping (CCM) (Methods), (2) to test for causal interactions between mosquito subpopulations, and (3) to explore scenarios how change in a causal environmental variable may affect mosquito abundance as well as number and strength of mosquito outbreaks

Objectives

Methods

Results

Conclusion