Methodologically Enhanced Virtual Labs for Early Warning of Significant or Catastrophic Change in Ecosystems: Changepoints for a Changing Planet

点击翻译按钮获取中文摘要

Virtual labs are emerging as a key component in the construction of future digital environments, particularly to abstract over the complexities of the underlying distributed networks of sensors and associated computational infrastructure. We define a virtual lab as a transdisciplinary collaboration space hosted in the cloud (public/private/hybrid) that allows stakeholders to access a range of data, analytical methods and assessment tools (e.g. visualisation tools and/or statistical tools), and to execute these analyses using the elastic capacity of a cloud. In the environmental science community, most existing virtual labs focus on the problem of integrating often complex and heterogeneous data. We seek to significantly advance the state-of-the-art by enhancing virtual labs with sophisticated methodological capability, embracing state-of-the-art data science techniques to assist in the societally-relevant interpretation of these data. This is a bold and broad vision and, to make this feasible in a year, we elect to work with a particular family of data science techniques, that is, changepoint detection methods, designed to identify fundamental changes and anomalous behaviour in data, typically within time-series, but also applicable across space and time and to complex, multivariate problems.This feasibility study will therefore bring together a cross-disciplinary team working on virtual labs, changepoint methods and evidence for impacts of global environmental change on ecosystem structure and function. Our approach will foster a deep, cross-disciplinary dialogue through workshops, enhanced by rapid prototyping of virtual labs to stimulate thinking about what is possible/desirable w.r.t. ecosystem early warning methods.The project will build on the rich, complex, multi-faceted data available from the Environmental Change Network (ECN), that offers detailed multivariate 25-year long data sets for a range of ecosystems in the UK. We seek to understand the role of data science, including, but not limited to changepoint detection, in the construction of environmental early warning alert systems capable of operating at a variety of scales, from catchments to global planetary level systems.