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.

No abstract

Regional-scale forecasting of landslides is not a straightforward task. In this work, the spatiotemporal forecasting capability of a regional-scale landslide warning system was enhanced by integrating two different approaches. The temporal forecasting (i.e. when a landslide will occur) was accomplished by means of a system of statistical rainfall thresholds, while the spatial forecasting (i.e. where a landslide should be expected) was assessed using a susceptibility map. The test site was the Emilia Romagna region (Italy): the rainfall thresholds used were based on the rainfall amount accumulated over variable time windows, while the methodology used for the susceptibility mapping was the Bayesian tree random forest in the tree-bagger implementation.

Abstract.We set up an early warning system for rainfallinduced landslides in Tuscany (23 000 km 2 ). The system is based on a set of state-of-the-art intensity-duration rainfall thresholds (Segoni et al., 2014b) and makes use of LAMI (Limited Area Model Italy) rainfall forecasts and real-time rainfall data provided by an automated network of more than 300 rain gauges.The system was implemented in a WebGIS to ease the operational use in civil protection procedures: it is simple and intuitive to consult, and it provides different outputs. When switching among different views, the system is able to focus both on monitoring of real-time data and on forecasting at different lead times up to 48 h. Moreover, the system can switch between a basic data view where a synoptic scenario of the hazard can be shown all over the region and a more in-depth view were the rainfall path of rain gauges can be displayed and constantly compared with rainfall thresholds.To better account for the variability of the geomorphological and meteorological settings encountered in Tuscany, the region is subdivided into 25 alert zones, each provided with a specific threshold. The warning system reflects this subdivision: using a network of more than 300 rain gauges, it allows for the monitoring of each alert zone separately so that warnings can be issued independently.An important feature of the warning system is that the visualization of the thresholds in the WebGIS interface may vary in time depending on when the starting time of the rainfall event is set. The starting time of the rainfall event is considered as a variable by the early warning system: whenever new rainfall data are available, a recursive algorithm identifies the starting time for which the rainfall path is closest to or overcomes the threshold. This is considered the most hazardous condition, and it is displayed by the WebGIS interface.The early warning system is used to forecast and monitor the landslide hazard in the whole region, providing specific alert levels for 25 distinct alert zones. In addition, the system can be used to gather, analyze, display, explore, interpret and store rainfall data, thus representing a potential support to both decision makers and scientists.

No abstract