Assessing shallow landslide susceptibility by using the Generalized Additive Model: a case study

Carlotta Bartelletti (1), Yuri Galanti (1), Michele Barsanti (2), Roberto Giannecchini (1), Giacomo D’Amato Avanzi (1), Maria Giuseppina Persichillo (3), Massimiliano Bordoni (3), Claudia Meisina (3), Andrea Cevasco (4) & Jorge Pedro Galve (5)
(1) Dipartimento di Scienze della Terra, Università di Pisa, Via S. Maria, 53, 56126, Pisa. (2) Dipartimento di Ingegneria Civile e Industriale, Università di Pisa, Largo L. Lazzarino, 56122, Pisa. (3) Dipartimento di Scienza della Terra e dell'Ambiente, Università degli studi di Pavia, Via Ferrata, 1, 27100, Pavia. (4) Dipartimento di Scienze della Terra dell'Ambiente e della Vita, Università degli studi di Genova, Corso Europa, 26, 16132, Genova. (5) Departamento de Geodinámica, Universidad de Granada, Campus Universitario Fuentenueva, 18071, Granada. Corresponding author e-mail: galanti@dst.unipi.it.

In this work, the Generalized Additive Model (GAM) technique was implemented for the landslide susceptibility assessment in the Gravegnola T. basin (Eastern Liguria, Italy), affected by many shallow landslides caused by the 25 October 2011 rainstorm. Nine morphological variables, river network, land use and geological settings were considered in GAM. The predictive performance of different combinations of these variables (chosen using a stepwise optimization of the Akaike information criterion) was evaluated through the cross-validation technique and AUROC computation. A susceptibility map using all the shallow landslide types was produced and compared with those obtained by Bartelletti et al. (2017b) for each different landslide type. The results strengthen the ability of this methodology to select the most influent predisposing factors. The bootstrap procedure allowed to compute the 95% probability confidence intervals of the landslide probability. Their amplitude can be interpreted as a measure of the spatial model reliability.

shallow landslide, Generalized Additive Model, predisposing factors, model reliability, Liguria, Italy.