The Emilia Romagna region (22,446 km2, Northern Italy) is widely affected by landslides. The Civil protection Agency of the Emilia Romagna Region uses a regional scale warning system (WS) for the management of the risk related to rainfall induced landslides. The WS is used to perform a temporal forecasting of landslides, as it provides an alert level for each of the eight subdivisions (called alert zones—AZ) of the regional territory. To improve the spatial information of the WS, we developed a susceptibility map and we tested the feasibility of coupling the temporal forecasting of the WS with the spatial forecasting of the susceptibility map. To map the landslide susceptibility at regional scale, we adopted a simple implementation of the Random Forest (RF) classification family. Random forest is a combination of tree (usually binary) Bayesian predictors that allows relating a set of contributing factors with the actual landslides occurrence. As a non-parametric model, it is possible to incorporate a wide range of numeric or categorical data layers and there is no need to select unimodal training data. Many classical and widely acknowledged landslide predisposing factors have been taken into account as mainly related to lithology, land use and morphometry (primary and secondary attributes derived from the DTM). The use of random forest enabled us to estimate the relative importance of the single input parameters and to select the optimal configuration of the regression model: an automated procedure selected the optimal configuration of parameters discarding the uninfluential and the pejorative ones. To develop the susceptibility map we considered the parameter set characterized by the lowest misclassification probability. To calibrate the model we used a training set and a test set considering the 10 % of the study area with random sampling. Following this approach we obtained a regional scale susceptibility map with 100 m resolution. We verified that the majority of the landslides forecasted by the WS in the past 7 years (this dataset is completely independent from the one used for the susceptibility assessment) occurred in areas mapped as highly (53 %), very highly (26 %) or moderately (19 %) susceptible to landsliding. Therefore, we concluded that the regional scale susceptibility map can be fruitfully exploited by Civil Protection Authorities to better focus their efforts in case of warnings issued by the regional warning systems.

Regional scale landslide susceptibility mapping in Emilia Romagna (Italy) as a tool for early warning / Lagomarsino D.; Segoni S.; Fanti R.; Catani F.; Casagli N.. - STAMPA. - (2014), pp. 443-449. [10.1007/978-3-319-05050-8_69]

Regional scale landslide susceptibility mapping in Emilia Romagna (Italy) as a tool for early warning

LAGOMARSINO, DANIELA;SEGONI, SAMUELE;FANTI, RICCARDO;CATANI, FILIPPO;CASAGLI, NICOLA
2014

Abstract

The Emilia Romagna region (22,446 km2, Northern Italy) is widely affected by landslides. The Civil protection Agency of the Emilia Romagna Region uses a regional scale warning system (WS) for the management of the risk related to rainfall induced landslides. The WS is used to perform a temporal forecasting of landslides, as it provides an alert level for each of the eight subdivisions (called alert zones—AZ) of the regional territory. To improve the spatial information of the WS, we developed a susceptibility map and we tested the feasibility of coupling the temporal forecasting of the WS with the spatial forecasting of the susceptibility map. To map the landslide susceptibility at regional scale, we adopted a simple implementation of the Random Forest (RF) classification family. Random forest is a combination of tree (usually binary) Bayesian predictors that allows relating a set of contributing factors with the actual landslides occurrence. As a non-parametric model, it is possible to incorporate a wide range of numeric or categorical data layers and there is no need to select unimodal training data. Many classical and widely acknowledged landslide predisposing factors have been taken into account as mainly related to lithology, land use and morphometry (primary and secondary attributes derived from the DTM). The use of random forest enabled us to estimate the relative importance of the single input parameters and to select the optimal configuration of the regression model: an automated procedure selected the optimal configuration of parameters discarding the uninfluential and the pejorative ones. To develop the susceptibility map we considered the parameter set characterized by the lowest misclassification probability. To calibrate the model we used a training set and a test set considering the 10 % of the study area with random sampling. Following this approach we obtained a regional scale susceptibility map with 100 m resolution. We verified that the majority of the landslides forecasted by the WS in the past 7 years (this dataset is completely independent from the one used for the susceptibility assessment) occurred in areas mapped as highly (53 %), very highly (26 %) or moderately (19 %) susceptible to landsliding. Therefore, we concluded that the regional scale susceptibility map can be fruitfully exploited by Civil Protection Authorities to better focus their efforts in case of warnings issued by the regional warning systems.
2014
9783319050492
9783319050508
Landslide Science for a Safer Geoenvironment - Volume 2: Methods of Landslide Studies
443
449
Lagomarsino D.; Segoni S.; Fanti R.; Catani F.; Casagli N.
File in questo prodotto:
File Dimensione Formato  
Lagomarsino et al SPRINGER 2014.pdf

Accesso chiuso

Descrizione: Articolo principale
Tipologia: Pdf editoriale (Version of record)
Licenza: DRM non definito
Dimensione 2.51 MB
Formato Adobe PDF
2.51 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/903332
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact