Shallow landslides are a widespread phenomenon in many regions of the world, posing significant risks to lives and economies. Triggered mainly by intense rainfalls, these events are especially susceptible to climate change, although predicting the precise impact of global warming on landslide risks remains challenging despite increasing research. Italy, one of the five most disaster-prone European countries, is particularly vulnerable due to its complex terrain, with 75% of its land being mountainous or hilly, making landslide risk management a priority. This study comprehensively evaluates shallow landslide susceptibility in two areas with distinct geological, geomorphological, and climatic characteristics, assessing the roles of various predisposing factors in slope instability. Findings emphasize the importance of context-specific analyses, as diverse regions may show different landslide triggers. For one study area, an event-based susceptibility model was created to simulate the impact of rainfall events, characterized by different intensities and cumulative antecedent rainfall, on slope stability. The research notes challenges when using data from single events to calibrate models, as unique geological factors may skew results and overemphasize certain factors. To address this, an ensemble of rainfall events, including non-landslide events, is recommended for a balanced understanding of interactions between geological, geomorphological, and meteorological factors. Additionally, a novel classification scheme is introduced, based on misclassification costs, providing a practical tool for early warning systems. By adjusting thresholds according to the costs associated with false positives and negatives, the model becomes more flexible for specific needs and risk management goals. The time-dependent landslide susceptibility model was then used to calculate susceptibility under various combinations of rainfall intensity and cumulative antecedent rainfall. Results, summarized in matrices and paired with future rainfall projections under RCP4.5 and RCP8.5 scenarios, illustrate potential climate impacts on slope stability. The study finds minimal changes in time-dependent susceptibility over time, without clear trends, as extreme rainfall events were not projected in the scenarios considered. Finally, a physically-based model was developed for a single slope in the study area, incorporating downscaled, remotely-sensed soil moisture data with a geomechanical characterization to calculate safety factors during rainfall events. Due to high computational demands, a surrogate model was trained to allow larger-area and longer-duration analyses with significantly reduced computation time. This surrogate model, applied to evaluate climate change impacts, effectively simulated the complex evolution of slope stability under future scenarios, with results indicating no clear upward or downward trends in the safety factor, as seen in previous analyses. In conclusion, the study successfully combines conditioning and triggering factors in shallow landslide susceptibility modelling, offering a versatile approach adaptable to diverse geographic and climatic contexts for improving early warning systems and assessing landslide risks in a changing climate. The complex effects observed from climate change on slope stability, both locally and regionally, highlight the need for further research to fully understand these interactions and their broader implications.
Le frane superficiali sono eventi diffusi in molte regioni del mondo che rappresentano un grave rischio per la vita e con potenziali costi economici elevati. In genere, sono causate da eventi di precipitazione intensi, rendendole strettamente legate ai cambiamenti climatici. Sebbene la ricerca sui rischi idrogeologici legati al cambiamento climatico sia aumentata, prevedere con precisione questa relazione resta complesso. In Europa, l’Italia è particolarmente vulnerabile, con circa il 75% del territorio in aree montuose o collinari, dove la gestione del rischio idrogeologico si rivela essere cruciale. Questo studio esamina due diverse aree con condizioni geologiche, geomorfologiche e climatiche variabili, valutando la suscettibilità alle frane superficiali e analizzando il ruolo di vari fattori predisponenti nell’instabilità dei versanti. I risultati evidenziano come ciascuna regione presenti dinamiche di innesco differenti, rendendo indispensabili valutazioni specifiche. In un’area di studio è stata poi condotta un'analisi della suscettibilità tempo-dipendente per vari eventi di pioggia, producendo un modello che simula l'impatto delle precipitazioni, sia in termini di intensità che di pioggia cumulativa antecedente, sulla stabilità dei versanti. Tuttavia, l’uso di dati relativi a singoli eventi nella calibrazione del modello presenta delle difficoltà, poiché condizioni geologiche particolari possono influenzare i risultati, sovrastimando alcuni fattori. Per ridurre questo effetto, si consiglia di usare un insieme più ampio di eventi di precipitazione, per meglio comprendere le interazioni tra fattori geologici, geomorfologici e meteorologici. Inoltre, lo studio propone un nuovo schema di classificazione basato sui misclassification costs, utile per applicazioni pratiche come i sistemi di early-warning. Regolando le soglie di classificazione a seconda dei costi associati a falsi positivi e falsi negativi, il modello risulta più flessibile e adattabile agli specifici obiettivi di gestione del rischio. Successivamente, il modello di suscettibilità è stato usato per calcolare la probabilità di frana in funzione dell'intensità e della pioggia cumulata antecedente. I risultati sono stati integrati con proiezioni di precipitazione future (scenari RCP4.5 e RCP8.5), per comprendere l'impatto potenziale del cambiamento climatico sulla stabilità dei versanti. Sebbene i modelli climatici non prevedano eventi di pioggia estremi, si osservano leggere variazioni nella suscettibilità tempo-dipendente, seppur difficili da definire in modo preciso. Infine, è stato sviluppato un modello fisico per l'analisi di stabilità di un versante specifico nell'area di studio, integrando dati idrici del suolo da rilevamenti remoti con una caratterizzazione geomeccanica dei materiali, per calcolare il fattore di sicurezza durante le precipitazioni. A causa dell’elevato costo computazionale, è stato creato un modello surrogato che, basandosi sui risultati delle simulazioni numeriche, consente di analizzare aree estese e periodi di pioggia più lunghi con tempi di calcolo ridotti. I risultati confermano un'evoluzione complessa del fattore di sicurezza in condizioni di cambiamento climatico, senza un chiaro trend di aumento o diminuzione della stabilità. In conclusione, questo studio integra in modo efficace i fattori predisponenti e di innesco nella modellazione della suscettibilità alle frane superficiali, offrendo una metodologia versatile applicabile in diversi contesti geografici e climatici. Questo approccio può supportare lo sviluppo di sistemi di early-warning e la definizione di scenari di rischio in un contesto di cambiamento climatico. La complessità osservata negli effetti del cambiamento climatico sulla stabilità dei versanti sottolinea inoltre la necessità di ulteriori studi per una comprensione più approfondita delle implicazioni di tale fenomeno.
Rainfall-dependent susceptibility mapping for shallow landslides: data-driven and physically-based modelling under climate change
FUMAGALLI, MICOL
2025
Abstract
Shallow landslides are a widespread phenomenon in many regions of the world, posing significant risks to lives and economies. Triggered mainly by intense rainfalls, these events are especially susceptible to climate change, although predicting the precise impact of global warming on landslide risks remains challenging despite increasing research. Italy, one of the five most disaster-prone European countries, is particularly vulnerable due to its complex terrain, with 75% of its land being mountainous or hilly, making landslide risk management a priority. This study comprehensively evaluates shallow landslide susceptibility in two areas with distinct geological, geomorphological, and climatic characteristics, assessing the roles of various predisposing factors in slope instability. Findings emphasize the importance of context-specific analyses, as diverse regions may show different landslide triggers. For one study area, an event-based susceptibility model was created to simulate the impact of rainfall events, characterized by different intensities and cumulative antecedent rainfall, on slope stability. The research notes challenges when using data from single events to calibrate models, as unique geological factors may skew results and overemphasize certain factors. To address this, an ensemble of rainfall events, including non-landslide events, is recommended for a balanced understanding of interactions between geological, geomorphological, and meteorological factors. Additionally, a novel classification scheme is introduced, based on misclassification costs, providing a practical tool for early warning systems. By adjusting thresholds according to the costs associated with false positives and negatives, the model becomes more flexible for specific needs and risk management goals. The time-dependent landslide susceptibility model was then used to calculate susceptibility under various combinations of rainfall intensity and cumulative antecedent rainfall. Results, summarized in matrices and paired with future rainfall projections under RCP4.5 and RCP8.5 scenarios, illustrate potential climate impacts on slope stability. The study finds minimal changes in time-dependent susceptibility over time, without clear trends, as extreme rainfall events were not projected in the scenarios considered. Finally, a physically-based model was developed for a single slope in the study area, incorporating downscaled, remotely-sensed soil moisture data with a geomechanical characterization to calculate safety factors during rainfall events. Due to high computational demands, a surrogate model was trained to allow larger-area and longer-duration analyses with significantly reduced computation time. This surrogate model, applied to evaluate climate change impacts, effectively simulated the complex evolution of slope stability under future scenarios, with results indicating no clear upward or downward trends in the safety factor, as seen in previous analyses. In conclusion, the study successfully combines conditioning and triggering factors in shallow landslide susceptibility modelling, offering a versatile approach adaptable to diverse geographic and climatic contexts for improving early warning systems and assessing landslide risks in a changing climate. The complex effects observed from climate change on slope stability, both locally and regionally, highlight the need for further research to fully understand these interactions and their broader implications.File | Dimensione | Formato | |
---|---|---|---|
phd_unimib_790564.pdf
accesso aperto
Dimensione
12.49 MB
Formato
Adobe PDF
|
12.49 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/193761
URN:NBN:IT:UNIMIB-193761