Detecting causality in complex ecosystems: lesson from tiger mosquitoes and implications for biodiversity research.

There is growing empirical evidence that most ecological and bio-medical processes follow nonlinear dynamics. However, data analysis of nonlinear systems is fundamentally different from “standard statistics”. This is because each state of a nonlinear system depends on previous states in a nonlinear way. Recently, new methods were developed for the analysis of short (ecological) time series, including tests for causality between biotic or abiotic variables.

In the first part of the talk, I will give a brief introduction to nonlinear dynamics with a special emphasis on the problem of mirage correlations and the causality test convergent cross mapping (CCM) (Sugihara et al. 2012, Science).

In the second part, nonlinear methods are applied to the Polynesian tiger mosquito (Aedes polynesiensis), which is a main vector for dengue fever and lymphatic filariasis in the South Pacific. The times series analysis is based on a unique data set of a mosquito meta-population collected in French Polynesia over a period of more than two years.

We tested for causal interactions between climate variables and mosquito abundances using CCM, and found several previously undetected factors. A further key insight from this analysis is that subpopulations in close proximity (200m-5km) differ remarkably with respect to their dynamics and the environmental factors that affect their dynamics. As an outlook, I will discuss how dynamical thinking can lead to new research questions in biodiversity research, and problematize the issue of appropriate data collection, which is critical to make full usage of nonlinear time series analysis (e.g., for testing causal interactions).