Quantitative ecosystem management typically begins with ecosystem models, which require us to model species interactions. Quantitative models of this sort are fundamental to modern environmental management and population forecasting; however, this approach has the inherent limitation that the most important processes governing dynamics must be known a priori and incorporated correctly into the model. The inverse of this approach – inferring which processes and ecological interactions drive observed dynamics – remains an open problem in fisheries science and ecology more generally. We will lay the groundwork for robust management of marine systems by developing tools to quantify species interactions from time series. To do so, we will extend state-of-the-art machine learning algorithms based on symbolic regression and multivariate Gaussian processes. We will apply these algorithms to studying species interactions on George’s Bank and the coastal California ecosystem. These approaches will provide us with both a better understanding of species interactions in the ecosystems we manage as well as an increased capacity for forecasting in these systems. Moreover, the algorithms we propose are quite generic – they should be of use in many other venues outside marine ecosystem management.

To carry out this work, we seek a postdoctoral fellow with experience in one or more of the following areas: machine learning, symbolic regression, Gaussian process regression, time-delay embedding, high-performance computing. Experience coding in Matlab or R is required, experience programming for parallel architectures in C++ or Python would ideal. The ideal Fellow would also have a demonstrated interest in marine ecology and/or fisheries.

key words

Multi-species forecasting; Ecosystem management; Nonlinear dynamics; Gaussian process; Symbolic regression;

Citizenship:  Open to U.S. citizens, permanent residents and non-U.S. citizens

Level:  Open to Postdoctoral and Senior applicants