Landslide susceptibility in the Turkish Northwesternmost sector: Distinctive patterns of inactive and active landslides

The Western Pontides, an active neo-tectonic region in Northwestern Turkey, are particularly rich in slope movements. Here, data-driven modelling is challenging owing to a prominent variability in geological and geomorphological features, which hinders a straightforward comprehension of ongoing and past processes. Statistical models can indeed be used to recognise portions of the landscape that are more susceptible to landslides, whereas inferring possible causation from spatial correlations with a number of other spatial characteristics typically requires knowledge of triggers and mechanisms. Here, we present results that we obtained using a dataset of active and relict landslides in the Western Pontides, for which we produced two distinct landslide susceptibility models. We subdivided the area in slope units, which are geomorphologically consistent, and worked in R-INLA with a Bayesian version of a binomial Generalised Additive Model. By doing so, we were able to evaluate the effect of the variables in the two models and identify effects of geomorphological factors that depend on the state of activity. Furthermore, we could describe and compare linearities and nonlinearities to better capture differences in spatial patterns possibly related to distinct triggers. We argue that our approach may also serve to assess the reliability of an inventory classification, as the presence of biases would make the emergence of distinctive patterns less likely.