Timescale: Oct 2021-Sept 2024
Prof. Oliver Korup, University of Potsdam
Prof. Jürgen Kurths, PIK Potsdam & Hu Berlin
PD Dr. Norbert Marwan, PIK Potsdam & University of Potsdam
Summary of the project
Numerous studies devoted to landslide prediction rely on intersecting data of past slope failures and the associated geological, topographic, and meteorological conditions. One common notion is that contemporary atmospheric warming is increasing the magnitude and frequency of landslides. We test this notion by making use of detailed time series of meteorological and seismological data that have become available at high spatial and temporal resolution and encourage means to predict landslides from patterns of weather and seismic activity. In this project we aim to learn whether and how well it is feasible to predict, large—and often destructive—landslides from precursory weather conditions. We also aim to test how to disentangle such predictions from the influence of historic seismic activity. This approach is novel in that it focuses on climatic and seismic precursors for each landslide, and in that it focuses on the larger to largest slope failures that have eluded most comparable studies.
Objectives and Methods
Dedicated Regional Cluster(s): Japan, United States, (and worldwide given some datasets)
Related PhD-projects: P1 (first cohort, Ugur Öztürk) and I8 (second cohort, Katharina Müller, Lisa Luna): this project builds on P1 and I8. With our Japanese collaborators we also have access to landslide damage data, so that we hope to gain first insights into the damage portfolio (and possibly future exposure) of rainfall- and earthquake-triggered landslides at the scale of an entire nation.
Planned external collaborations:
Responsibilities: The PhD-project “Predicting large landslides in a changing climate” is based at the research team “Geohazards” of the University of Potsdam. The PhD-project aims at developing methods to identify from existing landslide catalogues those landslides that have the highest likelihood of having had a rainfall trigger, and thus separate them from co-seismic landslides. Methods for this project will involve the analysis of large and partly inhomogeneous landslide inventories from selected mountain belts using complex spatially embedded networks, and probabilistic classification methods using Bayesian Reasoning and Machine Learning.
Requirements: We are seeking applications from highly motivated individuals with a strong background in quanti-tative geosciences, remote sensing, engineering geology, or landslide research. Fluency in the English language (speaking and writing) as well as the willingness to work in an interdisciplinary team are essential. Experience with statistical software and learning, processing of large and in-homogeneous geodata is desirable. We expect a solid background in mathematics, programming skills, and interest in the quantitative assessment of geohazards and -risks. Basic knowledge of time-series analysis, data mining or machine learning, and modern risk concepts will be of advantage for this post.