Advancing our ability to predict the dynamics of complex natural systems

Опубликовано 17 августа 2016, 21:23

Ecologists face unprecedented demands for accurate and reliable predictive models to address urgent questions such as: how will the Earth's life support systems change into the future? how should we control the spread of emerging infectious diseases? how should we sustainably harvest species? In my career as an ecological modeller I have addressed problems using a wide spectrum of approaches, from low dimensional mathematical representations of populations, to highly parameterised individual based models of communities and, recently, building the Microsoft Dynamic Climate Model to predict the future of life on earth. Repeatedly I have realised that to make more rapid advances we need dramatic improvements in: the way we go about addressing scientific questions, the way we develop and parameterise predictive models and, the computational methods we use for developing and implementing our models. In this talk I will use my insights to give a perspective on why I think the Computational Science group here at Microsoft Research Cambridge is uniquely positioned to transform ecology as a predictive science, and the ways in which we should conduct our research to most effectively implement that transformation.