Suchergebnisse

Abstract. Coastal hillslopes often host higher concentrations of
earthquake-induced landslides than those further inland, but few studies
have investigated the reasons for this occurrence. As a result, it is
unclear if regional earthquake-induced landslide susceptibility models
trained primarily on inland hillslopes are effective predictors of coastal
susceptibility. The 2016 Mw 7.8 Kaikōura earthquake on the
northeastern South Island of New Zealand resulted in ca. 1600 landslides > 50 m2 on slopes > 15∘ within 1 km of
the coast, contributing to an order of magnitude greater landslide source
area density than inland hillslopes within 1 to 3 km of the coast. In this
study, logistic regression modelling is used to investigate how landslide
susceptibility differs between coastal and inland hillslopes and to determine
the factors that drive the distribution of coastal landslides initiated by
the 2016 Kaikōura earthquake. Strong model performance (area under the
receiver operator characteristic curve or AUC of ca. 0.80 to 0.92) was
observed across eight models, which adopt four simplified geology types. The
same landslide susceptibility factors, primarily geology, steep slopes, and
ground motion, are strong model predictors for both inland and coastal
landslide susceptibility in the Kaikōura region. In three geology types
(which account for more than 90 % of landslide source areas), a 0.03 or
less drop in model AUC is observed when predicting coastal landslides using
inland-trained models. This suggests little difference between the features
driving inland and coastal landslide susceptibility in the Kaikōura
region. Geology is similarly distributed between inland and coastal
hillslopes, and peak
ground acceleration (PGA) is generally lower in coastal hillslopes. Slope angle,
however, is significantly higher in coastal hillslopes and provides the best
explanation for the high density of coastal landslides during the 2016
Kaikōura earthquake. Existing regional earthquake-induced landslide
susceptibility models trained on inland hillslopes using common predictive
features are likely to capture this signal without additional predictive
variables. Interestingly, in the Kaikōura region, most coastal
hillslopes are isolated from the ocean by uplifted shore platforms. Enhanced
coastal landslide susceptibility from this event appears to be a legacy
effect of past erosion from wave action, which preferentially steepened
these coastal hillslopes.