ONEPlanet DTP - Using big data to investigate the predicatbility of landslides (OP2174)

The aim of this PhD project to investigate how complex environmental drivers such as storms affect the occurrence of landslides, using big data and data mining methods to investigate persistently problematic slopes. The research will firstly identify the relationship between environmental variables and landslides through big data, and then generate associations using data mining methods, for example deep learning, random forest and decision tree approaches, and finally it will combine model outputs with recent landslide events, to explore the quantitative prediction of events under known conditions.
The occurrence of landslides that impact key infrastructure such as utilities, buildings and key transport routes remains a costly and potentially deadly hazard. Recent literature has shown that changes in climate lead to instabilities on natural and engineered slopes and cause higher landslide potential. However, the true effects of extreme temperature, rainfall, wind speed events on the occurrence and nature of landslides remain unclear[1]. The uncertainty caused by climatic changes makes the landslide prediction and management more difficult. Many of existing landslide warning systems often ignore short term extremes and simplify environmental drivers into distinct and consistent seasonal variations [2].
The PhD student will analyze and identify the key environmental conditions associated with landslide occurrences in the study area over last decade, and generate data mining models between climate change and landslides and evaluate them. The successful candidate will also integrate the resulting data mining models into a landslide warning system developed for this study side.
[1]Landslides in a changing climate, Earth-Science Reviews. 2016, doi.org/10.1016/j.earscirev.2016.08.011
[2] Geographical landslide early warning systems, Earth-Science Reviews. 2020,
doi.org/10.1016/j.earscirev.2019.102973
This project only focuses on two types of landslides: OSF (Open Slope Failures) and CDF (Channelized Debris Flows). The data has been collecting by NERC (NE/P000010/1, NE/T00567X/1) at the study site: the A83 Rest and be Thankful, Scotland, together with Research England, Transport Scotland and the Scottish Road Research Board. Our industrial partner, Rakwireless, leading in intelligent sensing is willing to support.