Statistical inference on niche choices based on behavioural time series data
Over the last decade, the advancements made in telemetry, video-tracking and other technology have led to an explosion in the amount of data being collected on animal behaviour over space and time. The corresponding novel types of data offer vast new opportunities for ecologists to study animal behaviour, both in the field and in controlled lab experiments. In particular, behavioural time series data can provide important insights into the influence of internal and external drivers of behaviour. When multiple individuals are observed, then this gives the additional opportunity to study individualised niche choices, as manifested in behavioural traits such as specific movement patterns. In particular, such data sets offer vast opportunities for studying whether individual heterogeneity in realised niches is explained a) by external factors (e.g. habitat), b) by measurable internal factors (e.g. age, size), c) by differences in animal personality (individual variation that cannot be attributed to any measurable influencing factor), or a combination of these.
However, corresponding individual-level behavioural time series data are typically noisy, highly non-linear and highly correlated, rendering them intractable to analysis using conventional statistical methods. Furthermore, past research in behavioural ecology has mostly focused on population-level inference, with individual heterogeneity within populations usually treated as a nuisance rather than an interesting topic.
This project will provide and further develop the necessary statistical tools for the investigation of individualised niches from complex behavioural time series data. The focus hereby will lie on the versatile class of hidden Markov models as a very widely applicable, relatively accessible and extendable statistical tool for the analysis of behavioural data collected over time. More specifically, the key contributions of the proposed project are as follows:
- develop novel statistical methods for drawing inference on niche choice from behavioural time series, in particular (i) widely applicable statistical machinery for investigating different types of individual heterogeneity in niche choice, and (ii) a general modelling framework for analysing multi-stream and multi-scale that allows to study niche choices at different temporal and spatial scales within a joint inferential approach;
- conduct the statistical analyses in particular of the fur seal field data (on movement, habitat use and growth) collected within project A01 (Hoffman), of the turnip sawfly lab data (on movement) collected within project B02 (Müller), of the beetle lab data (on movement) collected within project C01 (Kurtz), and of the fruit fly lab data (on movement and network structures) collected within project C04 (Fricke).
From previous successful research collaborations with international ecologists working with telemetry technology, I have access to data sets collected for various shark and whale species, for harbour, elephant and grey seals, and also for terrestrial species such as muskoxen and moose. Within the project, these data sets can be used as case studies for any novel methodology developed. However, many of the data sets that will need to be analysed within the CRC are of slightly different types than those field data, such that I will likely need to further adapt and refine the existing methodology.