Detection of abrupt changes in land and ocean ecosystems

Dr Claudie Beaulieu, Ocean and Earth Science, University of Southampton, PI

Dr Rebecca Killick, Mathematics and Statistics, Lancaster University, Co-PI

 Man-made pressure on the Earth’s climate and ecosystems is increasing vulnerability to abrupt changes which, due to the limited time to adapt, can have severe socio-economic impacts. The ecosystem shift can be a response to change in external forcing (e.g. climate shift) or a random reorganization of the system, which can often be characterized by a simple autoregressive process. Distinguishing between random and deterministic regime shifts is fundamental to gain a better understanding of the underlying mechanisms and to predict their behaviour, but challenged by the limitations of statistical techniques available to address this problem.

Complex behaviour in climate and environmental time series, including the potential presence of autocorrelation and long-term climate change trend challenges the detection of abrupt changes. These different components can easily be confused, especially in the short and noisy time series characteristic of climate and environmental time series. Changepoint detection techniques offer great potential for the detection of abrupt changes in the Earth’s climate system and ecosystems. However, to the best of our knowledge, there is no available software that can distinguish between the different types of changes, which are often embedded in environmental time series.

In this project, we are developing software for use by environmental scientists and statisticians that can automatically discern between these different types of change and whether none, single or multiple changepoints are present without the need to make potentially arbitrary a priori decisions as to what type of change may be present. We expect the outcomes of this project to improve robustness and reliability of regime shift detection through our methodology that is especially tailored for the problem, thus leading to a better evidence and understanding of the underlying mechanism.

The talk given by Rebecca Killick can be viewed below: