Abstract
A map of susceptibility to submarine landslides in European seas has been developed using the maximum entropy model (MaxEnt) and landslide cartography acquired from European Marine Observation and Data Network (EMODnet) Geology (https://www.emodnet-geology.eu/). A total of 15% landslide points were set aside and used as test data, while the remaining 85% were used for model training. Based on EMODnet bathymetry, 33 environmental maps including bathymetric position index (BPI), slope, features, flat curvature and profile curvature, all calculated at different spatial scales, were processed. In an iterative process, all the variables that were highly collinear and those that did not contribute significantly to the model were eliminated. In the end, the model was constructed using only bathymetry and the slope of the seabed. The elaborated model has an area under the ROC (receiver operating characteristic) curve (AUC) of 0.910 calculated on the training set and 0.911 on the test set, and the ROC curves are coincident. The slope is the variable that gives the greatest contribution to the model (82%), but without the contribution of depth (18%) the results degrade significantly. The susceptibility map correctly identifies all the main structures where the presence of landslides is known. The model shows a certain sensitivity to bathymetric acquisition noise.
Thematic collection: This article is part of the Mapping the Geology and Topography of the European Seas (EMODnet) collection available at: https://www.lyellcollection.org/cc/EMODnet
- © 2020 The Author(s). Published by The Geological Society of London. All rights reserved
Please note that if you are logged into the Lyell Collection and attempt to access content that is outside of your subscription entitlement you will be presented with a new login screen. You have the option to pay to view this content if you choose. Please see the relevant links below for further assistance.