UROP Project

Incorporating vegetation effects into machine learning based shallow landslide hazard mapping

Machine learning based shallow landslide hazard mapping assumes that future landslides will occur under similar conditions as past ones which makes including extensive information on environmental conditions of past events in the model training essential. We would like to focus in this project on the influence of including vegetation information on the predictive quality of the hazard mapping process. Since vegetation is known to stabilize slopes susceptible to failure both mechanically and hydrologically, a collection of data needs to be investigated to unveil the maximum information which can improve the predictivity. The project comprises the design of a computational experiment to assess the influence of different vegetation parameters on the hazard mapping result, the search for open access vegetation datasets, the conduction of the experiment and its evaluation. An already existing flexible shallow landslide hazard mapping workflow can be used for this purpose.

Task

•Research on open-access (satellite) datasets providing vegetation information •Preparation of (satellite) data for use in hazard mapping workflow •Planning, conduction and evaluation of an experiment for the assessment of the influence of vegetation information on the predictive quality of a Random Forest-based shallow landslide hazard mapping workflow

Requirements

•Programming experience, preferably in Python •Basic geoscientific and/or geodetic knowledge •Preferably basic data science knowledge