Search

This study aims at evaluating the performance of the Maximum Entropy method in assessing landslide susceptibility, exploiting topographic and multispectral remote sensing predictors. We selected the catchment of the Giampilieri stream, which is located in the north-eastern sector of Sicily (southern Italy), as test site. On 1 October 2009, a storm rainfall triggered in this area hundreds of debris flow/avalanche phenomena causing extensive economical damage and loss of life. Within this area a presence-only-based statistical method was applied to obtain susceptibility models capable of distinguishing future activation sites of debris flow and debris slide, which where the main source of failure mechanisms for flow or avalanche type propagation. The set of predictors used in this experiment comprised primary and secondary topographic attributes, derived by processing a high resolution digital elevation model, CORINE land cover data and a set of vegetation and mineral indices obtained by processing multispectral ASTER images. All the selected data sources are dated before the disaster. A spatially random partition technique was adopted for validation, generating 50 replicates for each of the two considered movement typologies in order to assess accuracy, precision and reliability of the models. The debris slide and debris flow susceptibility models produced high performances with the first type being the best fit. The evaluation of the probability estimates around the mean value for each mapped pixel shows an inverted relation, with the most robust models corresponding to the debris flows. With respect to the role of each predictor within the modelling phase, debris flows appeared to be primarily controlled by topographic attributes whilst the debris slides were better explained by remotely sensed derived indices, particularly by the occurrence of previous wildfires across the slope. The overall excellent performances of the two models suggest promising perspectives for the application of presence-only methods and remote sensing derived predictors. Copyright © 2016 John Wiley & Sons, Ltd.

Landslide susceptibility may be evaluated by defining statistical relationships between the spatial distribution of past slope failures and the variability of landslide triggering factors. In this research, susceptibility to landsliding was assessed by employing multivariate adaptive regression splines (MARS), a statistical model that has been rarely used to this aim. The experiment was carried out in an area of central Sicily (Italy), which is severely affected by shallow landslides mainly occurring during the wet autumn- winter semester. Bedrock lithology and a set of primary and secondary topographic attributes were exploited as proxies of main landslide driving factors. The robustness of the procedure and the predictive skill of the susceptibility models were evaluated by calibrating and validating the models on different samples of raster cells. Stable cells (absence of landslides) of these samples were selected using two different methods: i) a random selection of individual stable cells; ii) a selection of cells intersecting circles randomly distributed within the stable sectors of the study area. The fit of the models to calibration and validation samples was quantitatively assessed by drawing receiver operating characteristic (ROC) curves and by calculating the area under the ROC curve (AUC). The validation results indicate high accuracy and stability of the models, demonstrating the ability of MARS to effectively predict the spatial distribution of landslides. Moreover, a better performance was observed for models trained and tested using circular groups of stable cells.

The determination of soil properties is considerable challenge when it is aimed to evaluate the spatial distribution of one or more parameters across significant surfaces. In fact, terrain sampling and field data are punctual measurement; therefore, quantitative models are needed to predict the spatial distribution of soil attributes. The spatialization of field and laboratory data is a very important information in landslide studies. Raster layers displaying soil properties can be used both for statistical models and for the parameterization of physically based models. The purpose of this work is to produce a detailed hydrological and mechanical characterization of the soil affected by shallow landsliding processes. The study area is located in Messina (Southern Italy) where a debris flow event occurred on the 1st October 2009. In particular, two small and independent (2 km2) hydrological units were chosen: Racinazzi and Saponarà catchments. The sample sites were selected using a provisional predictive pedologic model based on the topographic attributes Topographic Wetness Index and Steepness of slope. The field analyses were aimed to determine soil thickness, hydraulic conductivity and other soil mechanical properties. The fieldwork was carried out using: (1) Dynamic Cone Penetrometer (2) Amoozemeter (3) Auger sampler. The laboratory analysis of the collected samples have performed in order to characterize granulometry and the Atterberg limits. Stochastic approaches have been then adopted to regionalize the punctual information for each of the collected properties resulting in robust spatial distributions used to characterize landslide prone conditions. The results show the slope instability mainly affect the terraced areas characterized by the presence of a thin layer of soil which, according to the laboratory analysis, testifies an incomplete pedogenesis.

Shallow landslide and debris flows are among the most dangerous natural hazards triggered by extreme meteorological events. These phenomena have recently caused catastrophic scenarios in Italy (e.g. in Sarno-Quindici and Giampilieri) and, according to expected changes in the climate pattern, an increasing frequency of these phenomena is expected. The aim of this research is to assess the debris flow susceptibility in the Giampilieri area (Sicily) using a spatially-distributed debris flow runout model based on topographic information. The application of the model starts with the identification of the source areas from which debris flows are propagated on the basis of frictional laws and flow direction algorithms. The area selected for this study is located in the Ionian sector of the Peloritanian area in Sicily, in the South part of Messina (Sicily) and includes the villages of Giampilieri, Briga Itala and Scaletta Zanclea. There, the 1st October 2009 thousands of debris and mud flows were activated by a cumulative rainfall of about 160 mm in 6 hours, which followed two previous rainfalls events occurred on16th September (76 mm in six hours) and 23rd – 24th September (190 mm in 10 hours). Among the catchments hit by the 2009 event, the Giampilieri basin (10 km2) has been chosen as sub area in order to set the algorithms for the spreading assessment and the friction parameters of the model. In this catchment, a complete inventory of the source areas and accumulation zone was created by photointerpretation of post event images. Moreover, volume and velocity estimations of the mobilized material have been carried out. The susceptibility was evaluated using the source areas of the 2009 event and its accuracy was estimated by the comparison of the results with the accumulation areas and the velocity and volume estimated. In a second step we performed the analysis at the medium scale on the whole area hit by the 2009 event using the parameters calibrated on the Giampilieri basin. The presented approach of debris flow susceptibility analysis demonstrates that a simple assessment of the debris flow spreading calculated using defined source areas and calibrated on past events, provided good results for consequent hazard and risk studies.

Debris flows are among the most hazardous phenomena in nature, which typically take the form of multiple-occurrence regional landslide events triggered by intense driving inputs such as storms or earthquakes. The main tasks of this study were to verify whether cell-based susceptibility models is capable of predicting debris flow initiations in the Giampilieri catchment (southern Italy) and to explore the relationships between the pixel size of the adopted mapping units in terms of predictive performances of the derived models. The Giampilieri catchment is a small area (10km 2 ) hit by a storm on the 1 st October 2009 which resulted in the triggering of more than one thousand landslides and caused severe damages and 37 fatalities.

The SUFRA project is based on a three level susceptibility mapping. According to the availability of more detailed data, the three scale for susceptibility mapping are increased respect to the ones suggested by the TIER group to 1:100,000, 1:50,000 and 1:25,000/1:10,000. The mapping levels exploit climatic, soil use (CORINE2009) and seismic informative layers, differentiating in the details of the core data (geology and topography), in the quality and resolution of the landslide inventory and in the modelling approach (Tab. 1). SUFRA_100 is based on a heuristic approach which is applied by processing a geologic layer (produced by ARTA integrating pre-CARG 1:100,000 geologic maps); the DEM exploited are IGMI 250m and the mapping units are 1km side square cells. Models are validated with respect to the PAI LIPs (Landslide Identification Points) which are reclassified adopting a simplified scheme. Output cuts of SUFRA100 will be referred to administrative boundaries (provinces). SUFRA50 is based on statistical analysis of new CARG geologic maps and 20m (ITA2000) - 2m (ATA2007) DEM. The mapping units are 500m and 50m cells, hydrographic and hydro-morphometric units. The landslide inventory is the IFFI2012_LIPs (first level) which is the result of the conversion in IFFI format of the PAI archive, which will be supported by remote landslide mapping (exploiting the ATA2007 aerial photos), according to the IFFI first level approach. Validation of the models will be performed exploiting both random spatial partition and temporal partition methods. Output cuts of SUFRA50 will be based on physiographic (basin) and administrative (municipalities) boundaries. SUFRA10/25 is based on statistical analysis of new CARG geologic maps (remotely and field adapted) and 2m (ATA2007) DEM. The mapping units are the slope units (SLUs) which are derived by further partitioning the hydro-morphometric units so to obtain closed morphodynamic units. The landslide inventories is the IFFI2012 which is the results of a field supported (on focus) landslide remote systematic mapping, according to the IFFI full level approach. Examples of SUFRA_100, SUFRA_50 and SUFRA_10 are presented for some representative key sector of Sicily. First results attest for the feasibility and goodness of the proposed protocol. The SUFRA program aims at enabling the regional governmental administration to cope with landslide prevision, which is the required operational concept in land management and planning. PAI has been a great advance with respect to the “pre-SARNO” conditions, but it is very exposed to fail: it is a blind approach for new activations; it is critically dependent on the quality of the landslide inventories; it cannot project the susceptibility outside the landslide areas

Oggigiorno, la valutazione della diversa importanza delle variabili geoambientali nel determinare le condizioni di suscettibilità da frana di un’area è uno dei problemi più attuali della geologia. L’uso ed il confronto di due differenti approcci statistici, ha consentito di stimare le condizioni di predisposizione all’instabilità gravitativa dei versanti, per un esteso settore settentrionale della catena appenninica siciliana, ricadente all’interno delle tavolette I.G.M.I. nn. 259 I SE “Scillato” e 259 II NE “Caltavuturo”. L’area oggetto della sperimentazione, estesa circa 200 Km2, è stata suddivisa in maniera semi-automatica in 1827 unità idro-morfologiche o unità di versante. Per ciascun’unità di versante, è stato definito un set di 15 proprietà fisico-ambientali (caratteristiche geologiche, di uso del suolo, attributi topografici primari e secondari), che rappresentano le variabili indipendenti capaci di descrivere i fattori di controllo della stabilità dei versanti. Nell’area era disponibile un archivio di 233 frane di colamento che erano state riconosciute sulla base di fotointerpretazione e controlli di campo condotti nel 1999. In occasione della presente ricerca, utilizzando immagini GoogleTM relative al 2008, è stato realizzato un nuovo archivio eventi contenente 55 attivazioni successive. Per caratterizzare le condizioni di stabilità di ciascuna unità di mappatura, si sono utilizzate due differenti aree diagnostiche: il centroide di ogni area in frana (LC) e l’intera sua area di rottura (AR). Per definire il modello di suscettibilità sono state utilizzate due differenti tecniche di regressione: la Forward Logistic Regression (FLR) e Multivariate Adaptive Regression Splines (MARS). In particolare, si sono utilizzate le frane mappate al 1999 per allenare i modelli e le successive 55 attivazione per la validazione degli stessi. I due metodi sono stati applicati per ciascuna delle due aree diagnostiche, al fine di confrontare anche, per una stessa metodologia, la capacità previsionale offerta. I modelli di suscettibilità ricavati mostrano caratteristiche comuni, selezionando quali variabili indipendenti di maggiore impatto sulle condizioni di stabilità delle unità di mappatura: la pendenza e la lunghezza dell’unità di versante, l’uso del suolo, il comportamento litotecnico atteso dei terreni, la litologia affiorante, le condizioni geomorfologiche, la curvatura longitudinale al versante e l’indice di erosione potenziale. I modelli di regressione costruiti utilizzando quale area diagnostica il punto centroide del movimento franoso hanno performance previsionali (AUC = 0.693) inferiori ai modelli di regressione logistica implementati utilizzando l’intera area di rottura (AUC = 0.722). Il metodo MARS, pur selezionando le stesse variabili predittive, ha raggiunto maggiori prestazioni sia con il modello di previsione costruito sui punti centroidi (AUC > 0.90), sia con quello che utilizza l’intera area di rottura (ROC = 0.84), con un errore GCV sempre minore di 0.09. Entrambi gli approcci statistici hanno prodotto complessi algoritmi che riflettono la complessità dell’area oggetto dello studio e l’interdipendenza tra il set di variabili geoambientali e la franosità. Dai risultati ottenuti con entrambi gli approcci, si evidenzia che l’uso delle unità di versante come unità di mappatura di base, unitamente al punto centroide quale area di attivazione dell’unità morfologica, si presta alla costruzione di un modello di suscettibilità con buone prestazioni previsionali, consentendo un indubbio risparmio in termini di risorse economiche e temporali e, anche e soprattutto, di limitare fortemente gli eventuali errori di mappatura delle forme franose derivate dall’intervento di un operatore.

On the 1st of October 2009, the area centred on the village of Giampilieri (Messina), on the Ionian side of the Peloritan belt, suffered thousands of landslides activated in the time lapse of about five hours, which caused 36 victims, more than 100 injured and more than 0.5M€ of damage to structures. This unprecedented phenomenon was triggered by an exceptional meteorological event, recorded at the foothills with 250mm of rain in just 8 hours; this amount of rainfall was cumulated to two previous events (16/IX: 75mm; 23/IX: 190mm) for a total amount of more than 500mm in less than two weeks. Due to the peculiar triggering conditions a huge number of debris flows involved the shallow weathered layer of the outcropping lithologies, consisting of phyllites of the Mandanici Units, mica schists of the Mela Units and medium to high grade Varisic metamorphic rocks. The purpose of this study is to evaluate the exportability of a landslide susceptibility model, obtained by using logistic regression method, within a training hydrographic unit (the “Torrente Briga” catchment) to predict the landslide spatial distribution in a test hydrographic units (the “Torrente Giampilieri” catchment). Both the basins extend for about 10km2. Exporting procedures for susceptibility model are in fact of great importance to optimize the survey costs or when facing phenomena which are locally triggered, such as the ones activated under extreme rainfall events; in fact, in this cases the landslide scenario used to train the statistical model is local and spatially more limited than the extension of while investigated area. In this research we prepared a susceptibility model by means of forward logistic regression in the “Torrente Briga” catchment (871 landslides, regressing an optimized set of computed physicalenvironmental predictors and obtaining the log-function which was then applied to the “Torrente Giampilieri” catchment (1121 landslides). By using a 2m cell DEM (from which we calculated and tested a large set of primary and secondary topographic attributes) and some thematic maps (geology and land use), the following predictors have been selected: height, slope, stream power index, topographic wetness index, profile and plan curvatures, landform classification, outcropping geology, landuse. Unstable slope conditions were assigned to the cells within a 2.9m neighbourhood of the landslide identification points (located on the highest point of the landslide areas). Models were built by merging the unstable cells with an equal number of randomly selected stable cells. To assess the sensitivity of the models with respect to the selection, 8 extractions were performed for each of the two basins, obtaining 8 models for the “Torrente Briga” area and 8x8 exporting combination, for the “Torrente Giampilieri”. The results attest for a high performance of the models, as we obtained excellent AUC both for the the 8 Briga models (>0.84) and the 64 exported Giampilieri models (>0.8). High steepness, low height, plan concave and profile convex curvatures, together with south and south-west verging slopes, are the main controlling factors of debris flow initiations in the two areas.

L’attuale versione del PAI (Piano Stralcio per l'Assetto Idrogeologico) disponibile per il territorio siciliano è fortemente dipendente dallo scenario di dissesti passati censiti e catalogati, sulla base dei quali, utilizzando un sistema di matrici di valutazione, è possibile ricavare le condizioni di rischio geomorfologico associato. Questo stadio costituisce un primo grande avanzamento delle conoscenze a partire dal quale è ora necessario procedere alla valutazione della suscettibilità da frana e all’adozione dunque di uno strumento di analisi territoriale con carattere previsionale. La realizzazione di una cartografia della suscettibilità da frana a scala regionale pone d’altra parte una serie di problemi di approccio e di tecnica realizzativa, che derivano dalla necessità di trovare un compromesso tra estensione territoriale e risoluzione dei modelli previsionali e, dunque, di risoluzione dei dati necessari in ingresso. Il progetto SUFRA muove dalla stessa analisi metodologica del progetto TIER ma, alla luce della disponibilità per il territorio siciliano di tematismi geologici con maggiore dettaglio e inventari delle forme franose densamente popolati, diverge da questo sia nei dati in ingresso sia nei metodi di costruzione dei modelli, che vengono implementati per la valutazione della suscettibilità nei tre livelli in Sicilia. Tutti i livelli di mappatura sfruttano strati informativi con informazioni riguardo alla geologia, al clima, all’uso del suolo e ai dati sismici. Il progetto SUFRA mira dunque a realizzare una cartografia a varia scala della suscettibilità, articolata secondo i tre livelli: SUFRA100 (scala 1:100,000), SUFRA50 (scala 1:50,000) e SUFRA25/10 (scala 1:25,000 o 1:10,000). SUFRA100 è basato su un approccio euristico utilizzando un DEM con cella di 250 metri derivato dall’IGMI. Le unità di mappatura sono celle quadrate di 1 km di estensione. I modelli sono validati rispetto ai LIP (punti identificazione della frana) che vengono riclassificati con adozione di una semplificazione in due sole principali tipologie (frane di scarpata e frane di versante) dell’archivio PAI, disponibile per tutto il territorio regionale. La rappresentazione di uscita, a questo livello saranno limiti amministrativi come quelli comunali o provinciali. SUFRA50 si basa sull'analisi statistica sfruttando la disponibilità delle nuove carte geologiche (CARG) e un DEM ottenuto ricampionando a celle di 20 metri di lato il DEM ATA2007/08. L’inventario delle frane, utilizzato per questo livello è l’IFFI2012_LIP (solamente il primo livello) che è il risultato della conversione in formato IFFI dell'archivio PAI, che sarà supportato e implementato da un controllo remoto delle aree individuate (sfruttando le foto aeree ATA2007). La rappresentazione di uscita del SUFRA50 sarà basata su confini fisiografici (bacini idrografici) e/o amministrative (comuni). SUFRA10/25 si basa sull'analisi statistica delle carte CARG (verificate con un lavoro di campagna) e sfruttando l’alta risoluzione del DEM ATA2007. Le unità di mappatura di base corrispondono alle unità di versante (SLU), derivate da una nuova ripartizione delle unità idromorfometriche in modo da ottenere unità con coerenza morfodinamica. Gli inventari di frana sono il risultato di una successiva implementazione dell’archivio IFFI2012, controllato e verificato da remoto e con indagini di campo. Esempi di SUFRA_100, SUFRA_50 e SUFRA_10 sono stati realizzati per qualche settore rappresentativo di diversi contesti geologici e geomorfologici Siciliani. I primi risultati attestano per la fattibilità e la bontà del protocollo proposto.

In this work, we assessed gully erosion susceptibility in two adjacent cultivated catchments of Sicily (Italy) by employing multivariate adaptive regression splines and a set of geo-environmental variables. To explore the influence of hydrological connectivity on gully occurrence, we measured the changes of performance occurred when adding one by one nine predictors reflecting terrain connectivity to a base model that included contributing area and slope gradient. Receiver operating characteristic (ROC) curves and the area under the ROC curve were used to evaluate model performance. Gully predictive models were trained in both the catchments and submitted to internal (in the calibration catchment) and external (in the adjacent one) validation, using samples extracted both from all cells of the catchments and only from cells located along flow concentration axes. Model evaluation on the entire catchments shows outstanding predictive performance of models that either include or do not include the predictors selected to reflect potential hydrological connectivity. Conversely, area under the ROC curve values measured on flow concentration axes reveals that almost all the additional predictors improve the performance of the base model, but the most enhanced increase of accuracy occurs when upstream drainage density of each landscape position is included as predictor of gully occurrence.