Earthquakes have been recognized as a major cause of landsliding (Keefer, 1984), and landslides triggered by earthquakes have been documented since the IV century (Seed, 1968). The spatial distribution of earthquake-induced landslides around the seismogenetic source has been analysed to better understand the triggering of landslides in seismic areas and to forecast the maximum distance at which an earthquake, with a certain magnitude, can trigger landslides. However, when applying such approaches to old earthquakes one should be concerned about the undersampling of smaller landslides, which can be cancelled, by erosion and landscape evolution. For this reason, it is important to characterize carefully the size distribution of landslides as a function of distance from the earthquake source. I analysed six earthquakes in the world that triggered significant amount of landslides (Finisterre 1993, Northridge 1994, Niigata 2004, Wenchuan 2008, Iwate 2008 and Tohoku 2011) to better understand the relation between the spatial distribution of the landslides, the peak ground acceleration (PGA), the distance from the sources, the relief and the lithologies of the area. I observed a strong relationship between landslides size and PGA, while the relationship between the distance from the source and the landslide size distribution is not clear, due to the interaction of different factors such as relief and lithology. I also developed magnitude frequency curves (MFC) for different distances from the source area by using different methods, such as: the maximum likelihood estimator of cumulative power-law distribution (Clauset et al, 2009); the maximum likelihood estimator of non-cumulative power-law function; the least square regression of non-cumulative log power-law function and the maximum likelihood estimator of Double Pareto distribution. I observed a decrease of the spatial density of landslides with distance, with a small effect of the size of these landslides. I also identify the Double Pareto function as the best tool for the fitting of the data (Valagussa et al., 2014a). In order to define the hazard due to earthquake-induced landslides, I developed a methodology for quantitative probabilistic hazard zonation for rockfalls (Valagussa et al., 2014b). I applied and demonstrated the method in the area of Friuli (Eastern Italian Alps) that was affected by the 1976 Mw 6.5 earthquake. Four rockfall datasets have been prepared from both historical data and field surveys. The methodology relies on a three-dimensional hazard vector (RHVmod), whose components include the rockfall kinetic energy, the fly height, and the annual frequency. The values of the first two components are calculated for each location along the slope using the 3D rockfall runout simulator Hy-STONE. The rockfall annual frequency is assessed by multiplying the annual onset frequency by the simulated transit frequency. The annual onset frequency is calculated 2 through a procedure that combines the extent of unstable areas, calculated for 10 different seismichazard scenarios with different annual frequencies of occurrence, and the magnitude relativefrequency relationship of blocks as derived from the collected field data. For each annual frequency of occurrence, the unstable area is calculated as a function of morphometric and earthquake characteristics. A series of discriminant-analysis models, using the rockfall datasets and DEMs of different resolution (1 and 10 m), identified the controlling variables and verified the model robustness. In contrast with previously published research, I show that the slope curvature plays a relevant role in the computation of the unstable area. To ensure the validity of the peak ground acceleration used as seismic parameter in the discriminant function, I also try to define a map of PGA based on the precarious balanced rocks surveyed on the field.

Gli eventi sismici sono riconosciuti come una delle maggiori cause per l’innesco di frane (Keefer, 1984). Le frane sismo-indotte sono documentate sin dal IV secolo (Seed, 1968). È stata condotta un’analisi sulla distribuzione spaziale delle frane sismo-indotte nell’area circostante la sorgente sismogenetica per meglio comprendere il loro innesco in aree sismiche e per delimitare la massima distanza alla quale un sisma con data magnitudo possa indurre frane. Tuttavia, quando si applicano tali approcci a eventi storici si pone un problema legato al sottocampionamento delle frane più piccole, che possono essere obliterate dall'erosione e dall'evoluzione del paesaggio. Per questo motivo è importante caratterizzare accuratamente la distribuzione delle frane, in termini di dimensione, in funzione della distanza dalla sorgente sismica. Sono stati analizzati sei terremoti in tutto il mondo che hanno innescato un significativo numero di frane (Finisterre 1993, Northridge 1994, Niigata 2004, Wenchuan 2008, Iwate 2008 and Tohoku 2011) per meglio comprendere le relazioni esistenti tra la distribuzione spaziale delle frane, l’accelerazione di picco al suolo (PGA), la distanza dalla sorgente, il relief e le litologie presenti nell’area. Si è osservata una forte relazione tra la PGA e la dimensione delle frane, mentre una la relazione tra la loro dimensione e la distanza dalla sorgente non è altrettanto chiara, ciò legato all’interazione tra diversi fattori quali ad esempio il relief e la litologia. Sono state realizzate e analizzate le curve magnitudo-frequenza (MFC) per differenti distanze dall’area sorgente attraverso varie metodologie: stimatore di massima verosimiglianza per distribuzioni di tipo potenza cumulate (Clauset et al, 2009), stimatore di massima verosimiglianza per distribuzioni di tipo potenza non cumulate, regressione ai minimi quadrati per funzioni di tipo potenza non cumulate in scala logaritmica e stimatore di massima verosimiglianza per la distribuzione Double Pareto. Dalle analisi si è potuto osservare un decrescere della densità spaziale delle frane con la distanza, ma un basso impatto della dimensione delle frane. Inoltre la funzione Double Pareto è stata scelta come miglior strumento per il fittaggio dei dati (Valagussa et al, 2014). Allo scopo di definire il rischio legato alle frane sismo-indotte è stata sviluppata una metodologia per la zonazione probabilistica quantitativa del rischio da frane da crollo (Valagussa et al, 2014). Il metodo è stato applicato e dimostrato nell’area del Friuli (Apli orientali) colpita da un terremoto di magnitudo 6.4 nel 1976. Quattro inventari sono stati realizzati sia tramite attività di terreno che da dati storici. La metodologia si basa sul vettore di rischio tridimensionale (RHVmod) le cui componenti includo l’energia cinetica, l’altezza di volo e la frequenza annua. I primi due valori sono calcolati per ogni cella del versante per mezzo del programma Hy-STONE. La frequenza annua è invece determinata moltiplicando la frequenza d’innesco annua per il numero di transiti simulati in ogni cella. La frequenza d’innesco annua è calcolata combinando l’area instabile, calcolata per 10 differenti scenari con differente frequenza annua di occorrenza sulla base di caratteristiche morfometriche e sismiche, e la curva magnitudo-frequenza relativa dei blocchi identificati da attività di terreno. Una serie di analisi discriminanti sono state condotte per determinare le variabili che controllano l’area in frana, sulla base degli inventari redatti e di DEMs a differenti risoluzioni (1 e 10m). L’analisi ha dimostrato il ruolo rilevante della curvatura nella definizione dell’area instabile. Per verificare la validità della mappa di PGA utilizzata nelle analisi, una nuova mappa è stata redatta sulla base delle Precarious Balanced Rocks identificate sul terreno.

(2015). Relationships between landslides size distribution and earthquake source area in a perspective of seismic hazard zoning. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2015).

Relationships between landslides size distribution and earthquake source area in a perspective of seismic hazard zoning

VALAGUSSA, ANDREA
2015

Abstract

Earthquakes have been recognized as a major cause of landsliding (Keefer, 1984), and landslides triggered by earthquakes have been documented since the IV century (Seed, 1968). The spatial distribution of earthquake-induced landslides around the seismogenetic source has been analysed to better understand the triggering of landslides in seismic areas and to forecast the maximum distance at which an earthquake, with a certain magnitude, can trigger landslides. However, when applying such approaches to old earthquakes one should be concerned about the undersampling of smaller landslides, which can be cancelled, by erosion and landscape evolution. For this reason, it is important to characterize carefully the size distribution of landslides as a function of distance from the earthquake source. I analysed six earthquakes in the world that triggered significant amount of landslides (Finisterre 1993, Northridge 1994, Niigata 2004, Wenchuan 2008, Iwate 2008 and Tohoku 2011) to better understand the relation between the spatial distribution of the landslides, the peak ground acceleration (PGA), the distance from the sources, the relief and the lithologies of the area. I observed a strong relationship between landslides size and PGA, while the relationship between the distance from the source and the landslide size distribution is not clear, due to the interaction of different factors such as relief and lithology. I also developed magnitude frequency curves (MFC) for different distances from the source area by using different methods, such as: the maximum likelihood estimator of cumulative power-law distribution (Clauset et al, 2009); the maximum likelihood estimator of non-cumulative power-law function; the least square regression of non-cumulative log power-law function and the maximum likelihood estimator of Double Pareto distribution. I observed a decrease of the spatial density of landslides with distance, with a small effect of the size of these landslides. I also identify the Double Pareto function as the best tool for the fitting of the data (Valagussa et al., 2014a). In order to define the hazard due to earthquake-induced landslides, I developed a methodology for quantitative probabilistic hazard zonation for rockfalls (Valagussa et al., 2014b). I applied and demonstrated the method in the area of Friuli (Eastern Italian Alps) that was affected by the 1976 Mw 6.5 earthquake. Four rockfall datasets have been prepared from both historical data and field surveys. The methodology relies on a three-dimensional hazard vector (RHVmod), whose components include the rockfall kinetic energy, the fly height, and the annual frequency. The values of the first two components are calculated for each location along the slope using the 3D rockfall runout simulator Hy-STONE. The rockfall annual frequency is assessed by multiplying the annual onset frequency by the simulated transit frequency. The annual onset frequency is calculated 2 through a procedure that combines the extent of unstable areas, calculated for 10 different seismichazard scenarios with different annual frequencies of occurrence, and the magnitude relativefrequency relationship of blocks as derived from the collected field data. For each annual frequency of occurrence, the unstable area is calculated as a function of morphometric and earthquake characteristics. A series of discriminant-analysis models, using the rockfall datasets and DEMs of different resolution (1 and 10 m), identified the controlling variables and verified the model robustness. In contrast with previously published research, I show that the slope curvature plays a relevant role in the computation of the unstable area. To ensure the validity of the peak ground acceleration used as seismic parameter in the discriminant function, I also try to define a map of PGA based on the precarious balanced rocks surveyed on the field.
FRATTINI, PAOLO
Earthquake, Landslide Size Distribution, Magnitude-Frequency Curve, Rockfall, Hazard, Discriminant Analysis, 3D Runout Modeling
GEO/05 - GEOLOGIA APPLICATA
English
9-feb-2015
Scuola di dottorato di Scienze
SCIENZE DELLA TERRA - 61R
27
2013/2014
open
(2015). Relationships between landslides size distribution and earthquake source area in a perspective of seismic hazard zoning. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2015).
File in questo prodotto:
File Dimensione Formato  
PhD_unimib_072715.pdf

accesso aperto

Descrizione: Tesi di Dottorato
Tipologia di allegato: Doctoral thesis
Dimensione 26.22 MB
Formato Adobe PDF
26.22 MB Adobe PDF Visualizza/Apri

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

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/68458
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact