Detecting wildfire dynamics with multi-dimensional time series of remotely sensed data

Ecological processes like wildfire are important in determining how terrestrial ecosystems like forests influence and are influenced by climate. For example, post-fire regeneration can follow different pathways, some of which may serve to exacerbate climate change, while others may mitigate it. Long time series of remotely sensed imagery are used to detect disturbances and trends such as insect outbreak and increased productivity in response to warming, and methods for working with these data are still in the initial phases of development. This studentship focuses on using remotely sensed data to observe ecological processes relevant to climate, and testing and developing new methods specific to ecosystem recovery from wildfire.

The proliferation of freely available long time series of remotely sensed data has shifted our ability to study ecosystems before and after a disturbance, to studying dynamics over decadal timescales. Methods for analysing long time series of data have focused on variability of a single value, such as a vegetation index, over time. However, we know that the interaction of different variables, such as vegetation cover and moisture, is important in determining how ecosystems influence climate. This studentship will develop methods to incorporate multiple variables to describe ecological change over time.

Incorporating understanding of ecosystem processes with models of vegetation dynamics and climate will improve the sophistication of feedbacks between the two. Long time series of remotely sensed imagery allow us to elaborate our understanding of such feedbacks at landscape to global scales. Models that include ecosystem disturbances generally assume self-replacement of ecosystems over some recovery period, instead of modelling the range of documented post-fire pathways. There is a broad opportunity to contribute to understanding of how ecosystems influence and respond to climate change in developing methods to work with long time series of remotely sensed data.

We will develop a statistical methodology for multivariate time series data, to detect change points indicating potential phase shifts in the ecological process such as recovery from wildfire. The methodological investigation focuses upon the development of time series decomposition approaches utilising nonparametric smoothing techniques, which offer a flexible modelling framework accommodating non-stationarity and varying cyclical patterns that are often observed in the remotely sensed data. We will then expand the modelling framework into a machine learning context to deal with on-line remotely sensed data. This development allows us to process such big data in a semi-automatic manner, i.e., monitoring changes in ecological processes. Processing of long, dense time series of large, multivariate datasets will be supported by the high-performance computing facility at the University of Leicester.

Dr Kirsten Barrett is an expert in using remotely sensed data to study disturbance recovery cycles over decadal timescales, primarily in boreal forest ecosystems as well as tundra and tropical forests. She currently leads a NERC funded project on persistent forest loss following wildfires in Siberia.

Dr Hideyasu Shimadzu is statistician specialised in time series analysis working in the assessment of the change of biological diversity. He has worked in both academia and governmental institutions, and his work involves the estimation of biodiversity around Australian ocean for setting up marine protected areas and the analysis of worldwide biodiversity records to assess the temporal change in biodiversity.

Entry requirement
Applicants are required to hold/or expect to obtain a UK Bachelor Degree 2:1 or better in a relevant subject. The University of Leicester English language requirements apply where applicable.

How to apply
Please refer to the CENTA Studentship application information on our website for details of how to apply.

As part of the application process you will need to:
• Complete a CENTA Funding form – to be uploaded to your PhD application
• Complete and submit your PhD application online. Indicate project CENTA2-GGE9-BAR1 in the funding section.
• Complete an online project selection form Apply for CENTA2-GGE9-BAR1