THE HYDROLOGIC CONTROL ON SHALLOW LANDSLIDE TRIGGERING: EMPIRICAL AND MONTE CARLO PYSICALLY-BASED APPROACHES

In the dissertation, a Monte Carlo approach, that combines stochastic and deterministic modeling approaches, is used to analyze the hydrological control on shallow landslide triggering. In particular, an integrated stochastic rainfall and deterministic landslide simulator has been developed for the purpose. The simulator is composed by the following components: (i) a seasonal Neyman-Scott Rectangular Pulses (NRSP) model to generate synthetic hourly point rainfall data; (ii) a module for rainfall event identification and separation from dry intervals; (iii) the Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability (TRIGRS) model, version 2 (Baum et al., 2008, 2010) to simulate landslide triggering by rainfall infiltration, combined with a water table recession (WTR) model that computes the initial water table height to consider in simulating rainfall events with TRIGRS. The Monte Carlo simulator has been applied to the Loco catchment in the Peloritani Mountains in northeastern Sicily of Italy, an area with high landslide risk, as recently demonstrated by the regional debris-flow event that occurred on 1 October 2009, which caused 37 casualties and millions of euros of damage. The Monte Carlo approach has been applied for estimation of return periods of shallow landslide triggering and for the evaluation of the most commonly-used types of empirical rainfall threshold. Use of the Monte Carlo approach for estimation of the return period of landsling, represents an advance to approaches based on rainfall Intensity-Duration-Frequency (IDF) curves, applied by several different researchers, for two reasons. Firstly because the response of an hillslope to hyetographs of rectangular (or any other predifined) shape may be significantly different from that to a real-like stochastically variable hyetograph. Secondly, and more importantly, the use of the Monte Carlo approach, in which water table depths at the beginning of each rainfall event are determined in response to antecedent rainfall time history, enables to avoid the drawback of assuming an arbitrary initial water table depth (for instance equal to zero), which has a probability to occur that should be taken into account in estimating the return period. In fact, IDF-based return period estimation is in principle flawed by the fact that in estimating return period the conditional probability of the rainfall event, given the assumed initial water table height, should be considered. Monte Carlo simulations have allowed to map return period of landslide triggering on the case-study catchment. Simulation results have been analyzed to evaluate from a theoretical perspective the Intensity-Duration empirical model paradigm, i.e. to understand if the stochastic nature of rainfall combined with the physical processes of soil-water movement provide a theoretical justification to this most widely used empirical model. In fact, in spite of its consolidated use, no particular theoretical justification for the use of the Intensity-Duration empirical model exists. The paradigm is that a rainfall threshold for landslide triggering assumes a straight line in a bi-logarithmic rainfall (mean) Intensity - Duration plane. The obtained results allow to state that, actually, stochastic structure of real rainfall events combined with the infiltration response reveal in a certain sense a theoretical justification to the I - D relationship. Iso-pore-pressure points, in the bi-logarithmic rainfall (mean) Intensity - Duration plane, lay, with relatively low scattering, around a straight line, in the cases that initial water table height is negligible. This means that the I-D model represents a valid model to interpret data in the case that memory of pore pressures is negligible. In the opposite, most likely, case, the I-D model should be coupled with an antecedent rainfall model.