Abstract. Despite the large number of recent advances and developments in landslide susceptibility mapping (LSM) there is still a lack of studies focusing on specific aspects of LSM model sensitivity. For example, the influence of factors such as the survey scale of the landslide conditioning variables (LCVs), the resolution of the mapping unit (MUR) and the optimal number and ranking of LCVs have never been investigated analytically, especially on large data sets. In this paper we attempt this experimentation concentrating on the impact of model tuning choice on the final result, rather than on the comparison of methodologies. To this end, we adopt a simple implementation of the random forest (RF), a machine learning technique, to produce an ensemble of landslide susceptibility maps for a set of different model settings, input data types and scales. Random forest is a combination of Bayesian trees that relates a set of predictors to the actual landslide occurrence. Being it a nonparametric model, it is possible to incorporate a range of numerical or categorical data layers and there is no need to select unimodal training data as for example in linear discriminant analysis. Many widely acknowledged landslide predisposing factors are taken into account as mainly related to the lithology, the land use, the geomorphology, the structural and anthropogenic constraints. In addition, for each factor we also include in the predictors set a measure of the standard deviation (for numerical variables) or the variety (for categorical ones) over the map unit. As in other systems, the use of RF enables one to estimate the relative importance of the single input parameters and to select the optimal configuration of the classification model. The model is initially applied using the complete set of input variables, then an iterative process is implemented and progressively smaller subsets of the parameter space are considered. The impact of scale and accuracy of input variables, as well as the effect of the random component of the RF model on the susceptibility results, are also examined. The model is tested in the Arno River basin (central Italy). We find that the dimension of parameter space, the mapping unit (scale) and the training process strongly influence the classification accuracy and the prediction process. This, in turn, implies that a careful sensitivity analysis making use of traditional and new tools should always be performed before producing final susceptibility maps at all levels and scales.
This paper concerns a regional scale warning system for landslides that relies on a decisional algorithm based on the comparison between rainfall recordings and statistically defined thresholds. The latter were based on the total amount of rainfall, which was cumulated considering different time intervals: 1-, 2-and 3-day cumulates took into account the critical rainfall influencing shallow movements, whilst a variable time interval cumulate (up to 240 days) was used to consider the triggering of deep-seated landslides in low permeability terrains. A prototypal version of the model was initially set up to define statistical thresholds. Then, thresholds were calibrated using a database of past georegistered and dated landslides. A validation procedure showed that the calibration highly improves the results and therefore the model was integrated in the regional warning system of Emilia Romagna (Italy) for civil protection purposes. The proposed methodology could be easily implemented in other similar regions and countries where a sufficiently organised meteorological network is present.Keywords Landslide . Optimization . Rainfall . Threshold Introduction In Italy, landsliding is a recurrent phenomenon responsible for casualties, destruction of assets and infrastructures and major economical loss (Guzzetti 2000). Since rainfall represents the most common triggering factor, many Italian civil protection agencies are setting up warning systems based on the interaction between rainfall and landslides. These agencies are responsible for large territories (e.g. regions or large subdivisions such as provinces), therefore they cannot rely on physically based approaches because of the difficulty of defining the exact spatial and temporal variation of the many involved factors (rainfall variation in space and in time, effect of vegetation, mechanic and hydraulic properties of both bedrock and soil layer). As a consequence, physically based approaches can effectively be applied only over small sites (Segoni et al. 2009), while at regional scale the most diffused methodology is the use of black box models based on empirical or statistical rainfall thresholds. The term 'black box' refers to a methodology in which the complex physical processes involved in landslide initiation are ignored (because too difficult to correctly calibrate over large areas) and a more simple and functional empirical correlation is found between the primarily cause (rainfall) and the effect (landslide). Amongst all the factors influencing the triggering of landslides, rainfall is-for instance-one of the most important and the easiest to correctly quantify, e.g. using rain gauges or radar measurements. The majority of the black box approaches are based on an empirical or statistical study of the rainfall characteristics that in the past led to landslides initiation (Caine 1980;Wieczorek 1996;Aleotti 2004;Guzzetti et al. 2008;Brunetti et al. 2010). Such study is aimed at deriving a mathematical equation which represents the threshold beyond whi...
[1] Catchment modeling in areas dominated by active geomorphologic processes, such as soil erosion and landsliding, is often hampered by the lack of reliable methods for the spatial estimation of soil depth. In a catchment, soil thickness h can vary as a function of many different and interplaying factors, such as underlying lithology, climate, gradient, hillslope curvature, upslope contributing area, and vegetation cover, making the distributed estimation of h challenging and often unreliable. In this work we present an alternative methodology which links soil thickness to gradient, horizontal and vertical slope curvature, and relative position within the hillslope profile. While the relationship with gradient and curvature should reflect the kinematic stability of the regolith cover, allowing greater soil thicknesses over planar and concave areas, the distance from the hill crest (or from the valley bottom) accounts for the position within the soil toposequence. This last parameter is fundamental; points having equal gradient and curvature can have significantly different soil thickness due to their dissimilar position along the hillslope profile. The proposed model has been implemented in a geographic information system environment and tested in the Terzona Creek basin in central Italy. Results are in good agreement with field data (mean absolute error is 11 cm with 8.5 cm standard deviation) and average errors are lower than those obtained with other topography-based methods, where mean absolute error ranges from 47 cm for a model based on curvature, position, and slope gradient to 94 cm for a model based solely on slope gradient. As a further test, the predicted soil thickness was used to determine derived quantities, such as the factor of safety for landsliding potential. Our model, when compared to other empirical methods, returns the best results and, therefore, can improve the prediction of soil losses and sediment production when utilized in conjunction with hydrological and landsliding models.Citation: Catani, F., S. Segoni, and G. Falorni (2010), An empirical geomorphology-based approach to the spatial prediction of soil thickness at catchment scale, Water Resour. Res., 46, W05508,
Abstract. HIRESSS (HIgh REsolution Slope StabilitySimulator) is a physically based distributed slope stability simulator for analyzing shallow landslide triggering conditions in real time and on large areas using parallel computational techniques. The physical model proposed is composed of two parts: hydrological and geotechnical. The hydrological model receives the rainfall data as dynamical input and provides the pressure head as perturbation to the geotechnical stability model that computes the factor of safety (FS) in probabilistic terms. The hydrological model is based on an analytical solution of an approximated form of the Richards equation under the wet condition hypothesis and it is introduced as a modeled form of hydraulic diffusivity to improve the hydrological response. The geotechnical stability model is based on an infinite slope model that takes into account the unsaturated soil condition. During the slope stability analysis the proposed model takes into account the increase in strength and cohesion due to matric suction in unsaturated soil, where the pressure head is negative. Moreover, the soil mass variation on partially saturated soil caused by water infiltration is modeled.The model is then inserted into a Monte Carlo simulation, to manage the typical uncertainty in the values of the input geotechnical and hydrological parameters, which is a common weak point of deterministic models. The Monte Carlo simulation manages a probability distribution of input parameters providing results in terms of slope failure probability. The developed software uses the computational power offered by multicore and multiprocessor hardware, from modern workstations to supercomputing facilities (HPC), to achieve the simulation in reasonable runtimes, compatible with civil protection real time monitoring.A first test of HIRESSS in three different areas is presented to evaluate the reliability of the results and the runtime performance on large areas.
Abstract.We propose an original approach to develop rainfall thresholds to be used in civil protection warning systems for the occurrence of landslides at regional scale (i.e. tens of thousands of kilometres), and we apply it to Tuscany, Italy (23 000 km 2 ).Purpose-developed software is used to define statistical intensity-duration rainfall thresholds by means of an automated and standardized analysis of rainfall data. The automation and standardization of the analysis brings several advantages that in turn have a positive impact on the applicability of the thresholds to operational warning systems. Moreover, the possibility of defining a threshold in very short times compared to traditional analyses allowed us to subdivide the study area into several alert zones to be analysed independently, with the aim of setting up a specific threshold for each of them. As a consequence, a mosaic of several local rainfall thresholds is set up in place of a single regional threshold. Even if pertaining to the same region, the local thresholds vary substantially and can have very different equations. We subsequently analysed how the physical features of the test area influence the parameters and the equations of the local thresholds, and found that some threshold parameters can be put in relation with the prevailing lithology. In addition, we investigated the possible relations between effectiveness of the threshold and number of landslides used for the calibration.A validation procedure and a quantitative comparison with some literature thresholds showed that the performance of a threshold can be increased if the areal extent of its test area is reduced, as long as a statistically significant landslide sample is present. In particular, we demonstrated that the effectiveness of a warning system can be significantly enhanced if a mosaic of site-specific thresholds is used instead of a single regional threshold.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.