Search

The increasing availability of long-term observational data can lead to the development of innovative modelling approaches to determine landslide triggering conditions at regional scale, opening new avenues for landslide prediction and early warning. This research blends the strengths of existing approaches with the capabilities of generalized additive mixed models (GAMMs) to develop an interpretable approach that identifies seasonally dynamic precipitation conditions for shallow landslides. The model builds upon a 21-year record of landslides in South Tyrol (Italy) and separates precipitation that induced landslides from precipitation that did not. The model accounts for effects acting at four temporal scales: short-term “triggering” precipitation, medium-term “preparatory” precipitation, seasonal effects and across-year data variability. It provides relative landslide probability scores that were used to establish seasonally dynamic thresholds with optimal performance in terms of hit and false alarm rates, as well as additional thresholds related to user-defined performance scores. The GAMM shows a high predictive performance and indicates that more precipitation is required to induce a landslide in summer than in winter/spring, which can presumably be attributed mainly to vegetation and temperature effects. The discussion illustrates why the quality of input data, study design and model transparency are crucial for landslide prediction using advanced data-driven techniques.

Italy is frequently hit and damaged by landslides, resulting in substantial and widespread disruptions. In particular, slope failures have a high impact on the population, communication infrastructure and the economic and productive sectors. The hazard posed by landslides requires adequate responses for landslide risk mitigation, with special attention to the risk to the population. In 2006 the Italian Department of Civil Protection, an Office of the Prime Minister, commissioned the Research Institute for Geo-Hydrological Protection (Istituto di Ricerca per la Protezione Idrogeologica), a research institute of the Italian National Research Council, to carry out operational forecasting of rainfall induced-landslides. Collecting landslide information in a catalogue is a preliminary action toward landslide forecasting. The use of spatially and temporally inaccurate landslide catalogues results in an uncertain and unreliable operational landslide forecasting. Consequently, accurate catalogues are needed to reduce the uncertainties, which are to some extent unavoidable. To this end, over the last 15 years many researchers have been involved in compiling a catalogue called ITALICA (ITAlian rainfall-induced LandslIdes CAtalogue), which currently lists 6312 records with information on rainfall-induced landslides that occurred over the Italian territory between January 1996 and December 2021. Overall, more than a third of the catalogue has very high geographic accuracy (less than 1 km2) and hourly temporal resolution. In contrast, less than 2 % of the catalogue have low and very low geographical accuracy and daily temporal resolution. This makes ITALICA the largest catalogue of rainfall-induced landslides accurately located in space and time available in Italy. Without this high level of accuracy, the precipitation responsible for the initiation of landslides cannot be reliably reconstructed, thus making the prediction of landslide occurrence ineffective. ITALICA's information on rainfall-induced landslides in Italy places a special emphasis on their spatial and temporal location, making the catalogue especially suitable for defining the rainfall conditions capable of triggering future landslides on the Italian territory. This information is fundamental for decision-making in landslide risk management.

In India, rainfall-induced landslides cause a high toll in terms of fatalities and damages. Therefore, the adoption of tools to predict the occurrence of such phenomena is urgent. For the purpose, the LANDSLIP project aimed at developing a landslide early warning system (LEWS) to forecast the occurrence of rainfall-induced landslides in two Indian pilot areas: Darjeeling and Nilgiris. Rainfall thresholds are a widely used tool to define critical probability levels for the possible occurrence of landslides in large areas, and are particularly suitable to be implemented in LEWSs.In this work, we exploited two catalogues of 84 and 116 rainfall conditions likely responsible for landslide triggering in Darjeeling and Nilgiris, respectively. Adopting a frequentist statistical method and using an automatic tool, we determined rainfall thresholds at different non-exceedance probabilities for the two pilot areas. Despite the daily temporal resolution of rainfall data and the spatial and temporal distribution of the documented landslides, the thresholds calculated for the two areas have acceptable uncertainties and were implemented in the LANDSLIP LEWS prototype. We expect that the new thresholds and the whole system will contribute to mitigate the landslide risk in the study areas.

Shallow landslides are frequently caused by an interplay of static predisposing factors (e.g., topography), mid-term preparatory factors (e.g., prolonged rainfall, seasonal changes of vegetation, snow melt), and short-term triggers (e.g., heavy rainfall). For large-area assessments, statistical analyses and data-driven approaches are often used to model landslide susceptibility based on spatial environmental variables or to derive critical landslide-triggering rainfall conditions through the definition of empirical rainfall thresholds. Attempts to integrate the spatial and temporal domains in the context of quantitative regional-scale landslide prediction are still rare.This contribution focuses on the landslide-prone area of South Tyrol, northern Italy (7,400 km²) and presents a novel data-driven modelling procedure that integrates spatial predisposing factors and dynamic preparatory and triggering factors to predict the probability of landslide occurrence in space and time. The approach is based on time-stamped landslide inventory data from 2000 to 2021, high-resolution gridded daily precipitation observations for the same period, and a set of relevant static environmental variables (e.g., including topographic indices, lithology). Data preparation included an initial filtering of rainfall-induced landslide presence observations and a rule-based stratified sampling of landslide absence observations at landslide locations and at non-landslide locations. Cross-validation was implemented in the model developing stage to select optimal time windows to represent pre-landslide preparatory and triggering cumulative rainfall conditions. Modelling was based on a binomial Hierarchical Generalized Additive Model (HGAM) that considered the non-linear influence and interactions (i.e., via tensor products) of static and dynamic environmental variables on landslide occurrence (presence vs. absence) while simultaneously accounting for the nested data structure (i.e., multiple considerations of each location) and seasonal effects. The study also considered different biases inherent in the input data, such as the known underrepresentation of landslide data in locations far from infrastructure or potential reporting biases across years, by averaging-out associated random effect variables. The results were validated quantitatively (spatial and temporal cross-validation) and qualitatively (geomorphic plausibility) and visualized as maps and probability surface plots.The assessment and validation confirmed the high generalizability and predictive performance of the model. A closer look at the derived relationships allowed to uncover the effects of predisposing, preparatory and triggering factors on landslide occurrence as well as the associated season-dependent variations. From our perspective, this new approach represents a good compromise between the high model flexibility for landslide prediction purposes and the high interpretability for understanding the underlying relationships.The research leading to these results is related to the PROSLIDE project that received funding from the research program Research Südtirol/Alto Adige 2019 of the Autonomous Province of Bozen/Bolzano – Südtirol/Alto Adige.

The occurrence of rainfall-induced shallow landslides is frequently caused by an interplay of predisposing environmental factors and dynamic preparatory and triggering conditions. For large-area assessments and for regional early warning, event-based landslide inventories are often employed to establish critical rainfall thresholds using statistical procedures (e.g., non-exceedance probability curves). These approaches typically put the spotlight on rainfall conditions associated with known landslide occurrences. Not accounting for rainfall conditions that did not induce slope instability comes along with a variety of criticalities, such as the impossibility to discriminate landslide from non-landslide rainfall conditions or the difficulty to validate the results.This contribution proposes a data-driven approach based on Generalized Additive Mixed Models (GAMM) to identify season-dependent shallow landslide rainfall conditions for the province of South Tyrol, Italy. The work builds upon high resolution gridded daily rainfall data and landslide observations for the period from 2000 to 2020. The workflow comprised an initial filtering of rainfall-induced landslides (presence data) and a rule-based stratified random sampling procedure to select non-landslide rainy days at the same locations (absence data). The time periods (time windows in days) to describe preparatory and triggering cumulative rainfall conditions were determined using an optimization procedure based on cross validation. In addition to modelling a yearly effect, a circular day-of-the-year variable was included in the model to consider additional seasonal influences. The underlying nested data structure (i.e., repeated measurements at each landslide location) was accounted for via a location-dependent random intercept. The resulting probability scores for the analysed variables were validated using space-time cross validation, visualized in the form of probability surface plots and complemented with quantitative thresholds (e.g., curves that optimally separate landslide presences and absences).Validation of the model showed a high capability to distinguish the two groups (presence vs. absence observations). The results further indicate that the temporal prediction of shallow landslides in South Tyrol can be improved by accounting for systematic seasonal effects other than triggering and preparatory rainfall variables. This novel approach is flexible and will further be extended to derive space-time predictions. Strengths and limitations for regional landslide early warning will be discussed.The research leading to these results are related to the Proslide project that received funding from the research program Research Südtirol/Alto Adige 2019 of the Autonomous Province of Bozen/Bolzano – Südtirol/Alto Adige.

In Alpine regions changes in seasonal climatic parameters, such as temperature, rainfall, and snow amount have already been observed. Specifically, in the South Tyrol area, meteorological observations indicate that temperatures are increasing and the number of snow days has generally diminished over time with perennial snow line now observed at higher elevations. Changes in rainfall have also been observed with more events associated with higher temperatures in the summer season. Natural hazards - mainly debris and mud flows, landslides, avalanches, rock falls, and (flash) floods - that affect this area every year, damaging population and infrastructures, are either weather or cryosphere-related. While these events have been recorded sporadically since the beginning of the 20th century, a systematic approach of their inventory has been done by local authorities since the 1990s. So far, Earth observation data has not been exploited to complete or complement existing inventories nor have they been used to investigate the influence of climate perturbation on potentially dangerous natural phenomena. The research presented here thus has three objectives: (i) analyse long time series of climate data and hazard occurrence in South Tyrol to examine if these records exhibit a coherent response of hazards to changes in climate; (ii) measure the spatio-temporal evolution of climatic and natural hazard events recorded, and (iii) explore potential relations between meteorological conditions and the hazard occurrence. In this context, in-situ and satellite-based climate data are exploited to study natural hazard triggers while the potential of Earth observation data is evaluated as a complement to the existing historical records of natural hazards. Specifically, Copernicus Sentinel-1 images are used to detect the spatio-temporal distribution of slow earth surface deformations and the results used for checking the completeness of the actual slow-moving landslide inventories. Hazard-related changes in the South Tyrolian landscape have also been analysed in relation to particular meteorological events at a regional scale, assessing trends and anomalies. Results show that: (i) satellite data are very useful to complement the existing natural hazard inventories; (ii) in-situ and satellite-based climate records show similar patterns but differ due to regional versus local variability; (iii) even in a data-rich region such as the analysed area, the overall response of natural hazard occurrence, magnitude, and frequency to change in climate variables is difficult to decipher due to the presence of multiple triggers and locally driven ground responses. However, an increase in the average annual duration of rainfall events and debris flow occurrence can be observed.

Several territorial landslide early warning systems in different parts of the world are based on empirical rainfall thresholds for landslide triggering. The calculation of such thresholds, using rainfall measurements gathered from rain gauges, has been examined frequently, especially considering uncertainties, modeling complexity, spatial assumptions, and analytical tools. Installed rain gauge networks that are spatially clustered in crowded areas have different spatial and attribute settings based on landslide occurrence conditions, such as rainfall record accessibility, processing, and usability, as well as specific locational, morphological, and hydrological settings.In this research work, we introduce an automatic tool called DEWS (Distance, Elevation, Watershed, and Slope unit) for rainfall-induced landslide spatial reference rain gauge selection. DEWS can be considered supplementary and complementary to the CTRL-T tool (Calculation of Thresholds for Rainfall-induced Landslides Tool) developed earlier, and works on a macro-to-micro scale of the spatial components of CTRL-T rain gauge selections. The output information, i.e. the list of selected reference rain gauges, can be used as input for CTRL-T to calculate frequentist rainfall thresholds at different non-exceedance probabilities. The DEWS tool fills the gap of the current literature, where the selection of reference rain gauges is mostly based on the nearest distance location and on statistical or manual procedures, without considering the morphological and hydrological settings of the area in which landslides occurred.The tool allows extracting rain gauges referring to landslide locations by employing four spatial filters: F1 (Distance), F2 (Elevation), F3 (Watershed), F4 (Slope unit), needing only a DEM, the coordinates of landslide and rain gauge locations and the parameters of the filter’s algorithms as inputs. More in detail, F1 selects rain gauges within a specified buffer distance from the landslide locations using the setting parameters and the coordinates of the landslides and rain gauges. Then, F2 uses the DEM to extract the elevation of the rain gauges and the landslides and then calculates the differences within each buffer circle; therefore, the filter keeps only the rain gauge with closest elevation values to each landslide (within F1 results) using the recommended/preferred/ or allowable elevation difference defined by the parameter’s settings. In F3, the rain gauges falling in the watershed that contains the landslide locations are extracted (within F1 and F2 results). F4, which is the smallest and most focused filter, uses a previously developed tool pack (within F1, F2, and F3 results) to extract the slope units associated with each landslide. Consequently, only the rain gauges falling within these slope units are selected.DEWS was implemented in a free tool pack in QGIS software, with default parameter values for non-expert users. The tool pack is divided into three main blocks following the filter structure (F1 and F2 are kept together). The reliability of DEWS was tested at a territorial scale in South Korea, using 223 landslides and 328 rain gauges. As a second step, frequentist rainfall thresholds were calculated in the study area.

Earth observation data are useful to analyze the impact of climate-related variables on geomorphological processes. This work aims at evaluating the impact of rainfall on slow-moving landslides, by means of a quantitative procedure for identifying satellite-based displacement clusters, comparing them with rainfall series, and applying statistical tests to evaluate their relationships at the regional scale. The chosen study area is the Basento catchment in the Basilicata region (southern Italy). Rainfall series are gathered from rain gauges and are analyzed to evaluate the presence of temporal trends. Ground displacements are obtained by applying the P-SBAS (Parallel Small BAseline Subset) to three datasets of Sentinel-1 images: T146 ascending orbit, and T51 and T124 descending orbits, for the period 2015–2020. The displacement series of the pixels located in areas mapped as landslides by the Italian Landslide Inventory and sited within rain gauge influence regions (defined as 10 km circular buffers) are studied. Those displacement series are analyzed and compared to the rainfall series to search for correlations, by employing statistical and non-parametric tests. In particular, two landslides are selected and investigated in detail. Significant results were obtained for the T124 descending orbit for both landslides, for a 3-day cumulative rainfall and a 7-day delay of the slope response. Challenges in the whole procedure are highlighted and possible solutions to overcome the raised problems are proposed. Given the replicability of the proposed quantitative procedure it might be applied to any study area.

Rainfall-triggered shallow landslides represent a major threat to people and infrastructure worldwide. Predicting the possibility of a landslide occurrence accurately means understanding the trigger mechanisms adequately. Rainfall is the main cause of slope failures in Slovenia, and rainfall thresholds are among the most-used tools to predict the possible occurrence of rainfall-triggered landslides. The recent validation of the prototype landslide early system in Slovenia highlighted the need to define new reliable rainfall thresholds. In this study, several empirical thresholds are determined using an automatic tool. The thresholds are represented by a power law curve that links the cumulated event rainfall (E, in mm) with the duration of the rainfall event (D, in h). By eliminating all subjective criteria thanks to the automated calculation, thresholds at diverse non-exceedance probabilities are defined and validated, and the uncertainties associated with their parameters are estimated. Additional thresholds are also calculated for two different environmental classifications. The first classification is based on mean annual rainfall (MAR) with the national territory divided into three classes. The area with the highest MAR has the highest thresholds, which indicates a likely adaptation of the landscape to higher amounts of rainfall. The second classification is based on four lithological units. Two-thirds of the considered landslides occur in the unit of any type of clastic sedimentary rocks, which proves an influence of the lithology on the occurrence of shallow landslides. Sedimentary rocks that are prone to weathering have the lowest thresholds, while magmatic and metamorphic rocks have the highest thresholds. Thresholds obtained for both classifications are far less reliable due to the low number of empirical points and can only be used as indicators of rainfall conditions for each of the classes. Finally, the new national thresholds for Slovenia are also compared with other regional, national, and global thresholds. The thresholds can be used to define probabilistic schemes aiming at the operative prediction of rainfall-induced shallow landslides in Slovenia, in the framework of the Slovenian prototype early warning system.

Landslides are among the most dangerous natural hazards, particularly in developing countries where ground observations for operative early warning systems are lacking. In these areas, remote sensing can represent an important tool to forecast landslide occurrence in space and time, particularly satellite rainfall products that have improved in terms of accuracy and resolution in recent times. Surprisingly, only a few studies have investigated the capability and effectiveness of these products in landslide forecasting, to reduce the impact of this hazard on the population. We have performed a comparative study of ground- and satellite-based rainfall products for landslide forecasting in India by using empirical rainfall thresholds derived from the analysis of historical landslide events. Specifically, we have tested Global Precipitation Measurement (GPM) and SM2RAIN-ASCAT satellite rainfall products, and their merging, at daily and hourly temporal resolution, and Indian Meteorological Department (IMD) daily rain gauge observations. A catalogue of 197 rainfall-induced landslides occurred throughout India in the 13-year period between April 2007 and October 2019 has been used. Results indicate that satellite rainfall products outperform ground observations thanks to their better spatial (10 km vs 25 km) and temporal (hourly vs daily) resolution. The better performance is obtained through the merged GPM and SM2RAIN-ASCAT products, even though improvements in reproducing the daily rainfall (e.g., overestimation of the number of rainy days) are likely needed. These findings open a new avenue for using such satellite products in landslide early warning systems, particularly in poorly gauged areas.

Landslides are frequent and widespread destructive processes causing casualties and damage worldwide [...]

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

In January 2020, the Istituto di Ricerca per la Protezione Idrogeologica of the Italian National Research Council, the British Geological Survey, the Norwegian Water Resources and Energy Directorate, the Swiss Federal Institute for Forest, Snow and Landscape Research, and the University of Salerno - Italy have organised a 3-day workshop on regional Landslide Early Warning Systems (LEWS). The workshop, held in Perugia, Italy, follows a previous meeting held in Oslo, Norway, in October 2016. The main aims of the initiative are: to collect experiences from worldwide invited experts involved in the design, the development, the operation or the analysis of LEWS, and to exchange knowledge, experiences, challenges and best practices. The first day of the workshop is dedicated to presentations from identified participants on specific topics relevant for the optimal design, implementation, and operation of global, national and regional LEWS. This is followed by a long discussion session, aimed at addressing many of the issues that are relevant for regional LEWS, including system performance, warning communication and involvement of the stakeholders. The second day is organized around four round tables on the following four topics: (i) data; (ii) landslide forecast models; (iii) warning models; (iv) scope, management structure, stakeholder involvement, and communication. The third day is focused on summarizing and formalizing the main issues discussed in an open document to be later shared with colleagues interested in LEWS. The final purpose of the workshop is to establish and consolidate a community of experts in LEWS and to build relationships with other communities (e.g., meteorologists, climate scientists, communications scientists). This will help to level up the quality of both theory and practice, and to define standards in early warnings in order to provide timely advisories and to initiate emergency responses to landslides (particularly rainfall-induced) avoiding or reducing life and economic losses. The main outcomes of the workshop, the most debated issues, and the key recommendations included in the open document will be presented and shared.

Rockfalls are frequent and harmful phenomena occurring in mountain ranges, coastal cliffs, and slope cuts. Although several natural processes occur in their formation and triggering, rainfall is one of the most common causes. The prediction of rock failures is of social significance for civil protection purposes and can rely on the statistical analysis of past rainfall conditions that caused the failures. The paper describes the analysis of information on rainfall-induced rockfalls in Gran Canaria and Tenerife, Canary Islands (Spain). An analysis of the monthly rainfall versus the monthly distribution of rockfalls reveals that they are correlated for most of the year, except in summer, when other triggers act to induce collapses. National and regional catalogs with hourly and daily rainfall measurements are used to reconstruct the cumulated amount (E) and the duration (D) of the rainfall responsible for the rock failures. Adopting a consolidated statistical approach, new ED rainfall thresholds for possible rockfall occurrence and the associated uncertainties are calculated for the two test sites. As far as is known, this is the first attempt to predict this type of failure using the threshold approach. Using the rainfall information, a map of the mean annual rainfall is obtained for Gran Canaria and Tenerife, and it is used to assess the differences between the thresholds. The results of this study are expected to improve the ability to forecast rockfalls in the Canary Islands in view of implementing an early-warning system to mitigate the rockfall hazard and reduce the associated risk.

Bhutan is highly prone to landslides, particularly during the monsoon season. Several landslides often occur along the Phuentsholing-Thimphu highway, a very important infrastructure for the country. Worldwide, empirical rainfall thresholds represent a widely used tool to predict the occurrence of rainfall-induced landslides. Nevertheless, no thresholds are currently designed and proposed for any region in Bhutan. In this work, we define empirical cumulated event rainfall-rainfall duration thresholds for the possible initiation of landslides using information on 269 landslides that occurred between 1998 and 2015 along the 90-km highway stretch between the towns of Phuentsholing and Chukha, in southwestern Bhutan, and daily rainfall measurements obtained from three rain gauges. For this purpose, we apply a consolidated frequentist method and use an automatic tool that identifies the rainfall conditions responsible for the failures and calculates thresholds at different exceedance probabilities and the uncertainties associated with them. Analyzing rainfall and landslide data, we exclude from the analysis all the landslides for which the rainfall trigger is not certain, so we reduce the number of landslides from 269 to 43. The calculated thresholds are useful to identify the triggering conditions of rainfall-induced landslides and to predict the occurrence of the failures in the study area, which is, to date, poorly studied. These rainfall thresholds might be implemented in an early warning system, in order to reduce the risk posed by these phenomena to the population living and traveling along the investigated road stretch.

In many areas of the world, the prediction of rainfall-induced landslides is usually carried out by means of empirical rainfall thresholds. Their definition is complicated by several issues, among which are the evaluation and quantification of diverse uncertainties resulting from data and methods. Threshold effectiveness and reliability strongly depend on the quality and quantity of rainfall measurements and landslide information used as input. In this work, the influence of the temporal resolution of rainfall measurements on the calculation of landslide-triggering rainfall thresholds is evaluated and discussed. For the purpose, hourly rainfall measurements collected by 172 rain gauges and geographical and temporal information on the occurrence of 561 rainfall-induced landslides in Liguria region (northern Italy) in the period 2004–2014 are used. To assess the impact of different temporal resolutions on the thresholds, rainfall measurements are clustered in increasing bins of 1, 3, 6, 12 and 24 h. A comprehensive tool is applied to each dataset to automatically reconstruct the rainfall conditions responsible for the failures and to calculate frequentist cumulated event rainfall–rainfall duration (ED) thresholds. Then, using a quantitative procedure, the calculated ED thresholds are validated. The main finding of the work is that the use of rainfall measurements with different temporal resolutions results in considerable variations of the shape and the validity range of the thresholds. Decreasing the rainfall temporal resolution, thresholds with smaller intercepts, higher slopes, shorter ranges of validity and higher uncertainties are obtained. On the other hand, it seems that the rainfall temporal resolution does not influence the validation procedure and the threshold performance indicators. Overall, the use of rainfall data with coarse temporal resolution causes a systematic underestimation of thresholds at short durations, resulting in relevant drawbacks (e.g. false alarms) if the thresholds are implemented in operational systems for landslide prediction.

The design, implementation, management, and verification of landslide early warning systems (LEWSs) are gaining increasing attention in the literature and among government officials, decision makers, and the public. Based on a critical analysis of nine main assumptions that form the rationale for landslide forecasting and early warning, we examine 26 regional, national, and global LEWSs worldwide from 1977 to August 2019. We find that currently only five nations, 13 regions, and four metropolitan areas benefit from LEWSs, while many areas with numerous fatal landslides, where landslide risk to the population is high, lack LEWSs. Operational LEWSs use information from rain gauge networks, meteorological models, weather radars, and satellite estimates; and most systems use two sources of rainfall information. LEWSs use one or more types of landslide forecast models, including rainfall thresholds, distributed slope stability models, and soil water balance models; and most systems use landslide susceptibility zonations. Most LEWSs have undergone some form of verification, but there is no accepted standard to check the performance and forecasting skills of a LEWS. Based on our review, and our experience in the design, implementation, management, and verification of geographical LEWSs in Italy, we conclude that operational forecast of weather-induced landslides is feasible, and it can help reduce landslide risk. We propose 30 recommendations to further develop and improve geographical LEWSs, and to increase their reliability and credibility. We encourage landslide forecasters and LEWSs managers to propose open standards for geographical LEWSs, and we expect our work to contribute to this endeavour.

Empirical rainfall thresholds are a tool widely used to predict the possible landslide occurrence in large areas. Review of the literature reveals that most of the thresholds are affected by several uncertainties related to: the rainfall and the landslide data accuracy, the reconstruction of the rainfall amount responsible for the failure, and the method adopted to calculate the thresholds. These uncertainties limit the use of thresholds in operational landslide early warning systems. To overcome the problem, we developed a comprehensive algorithm (named CTRL-A, Calculation of Thresholds for Rainfall-induced Landslides-Algorithm) that automatically and objectively reconstructs rainfall events and rainfall conditions responsible for the failures, and calculates rainfall thresholds at different exceedance probabilities and the associated uncertainties. The algorithm consists of three distinct blocks. The first block operates the reconstruction of rainfall events in terms of duration (D, in h) and cumulated event rainfall (E, in mm) using rainfall series and several parameters to account for different seasonal and climatic settings. Rainfall event parameters were calibrated adopting the monthly soil water balance model and evapotranspiration analysis. The second block performs the selection of the rain gauge representative for the landslide, the modeling of the antecedent rainfall, and the reconstruction of single or multiple rainfall conditions (MRC) likely responsible for the failures (associating a probability to each condition). The third block calculates the rainfall thresholds using the multiple rainfall conditions or the condition with the maximum probability (MPRC) for each failure. CTRL-A uses a set of adjustable parameters to account for different physical and climatic environments. We tested CTRL-A in Liguria, a northern Italian region highly prone to landslides, exploiting geographical and temporal information on 561 rainfall-induced landslides, occurred between October 2004 and November 2014, and hourly rainfall data measured by 172 rain gauges. The algorithm reconstructed 801 MRC and 440 MPRC associated to the observed landslides and determined ED frequentist thresholds, which may be implemented in an operational regional early warning system for shallow landslide forecasting in Liguria.

Only a few studies have investigated the geographical and temporal variations in the frequency and distribution of rainfall-induced landslides, and the consequences of the variations on landslide risk. Lack of information limits the possibility to evaluate the impact of environmental and climate changes on landslide frequency and risk. Here, we exploit detailed historical information on landslides and rainfall in Calabria, southern Italy, between 1921 and 2010 to study the temporal and the geographical variation in the occurrence of rainfall-induced landslides and in their impact on the population. We exploit a catalogue with information on historical landslides from June 1920 to December 2010, and daily rainfall records obtained by a network of 318 rain gauges in the same period, to reconstruct 448 493 rainfall events (RE). Combining the rainfall and the landslide information, we obtain a catalogue of 1466 rainfall events with landslides (REL), where an REL is the occurrence of one or more landslide during or immediately after a rainfall event. We find that (i) the geographical and the temporal distributions of the rainfall-induced landslides have changed in the observation period, (ii) the monthly distribution of the REL has changed in the observation period, and (iii) the average and maximum cumulated event rainfall that have resulted in landslides in the recent 30-year period 1981–2010 are lower than the rainfall necessary to trigger landslides in previous periods, whereas the duration of the RE that triggered landslides has remained the same. We attribute the changes to variations in the rainfall conditions and to an increased vulnerability of the territory. To investigate the variations in the impact of REL on the population, we compared the number of REL in each of the 409 municipalities in Calabria with the size of the population in the municipalities measured by national Censuses conducted in 1951, 1981, and 2011. We adopted two strategies; the first strategy considered impact as IREL = #REL / P, and the second strategy measured impact as RREL = #REL × P, where #REL is the total number of REL in a period, and P is the size of the population in the same period and geographical area. The analysis has revealed a complex pattern of changes in the impact of rainfall-induced landslides in Calabria in the recent past, with areas where IREL and RREL have increased, and other areas where they have decreased. Municipalities where IREL has increased are mainly in the mountains, and municipalities where RREL has increased are mainly along the coasts. The complexity of the changes in the frequency and impact of rainfall-induced landslides observed in Calabria suggests that it remains difficult and uncertain to predict the possible variations in the frequency and impact of landslide in response to future climatic and environmental changes.

GASAKe is a new hydrological model aimed at forecasting the triggering of landslides. The model is based on genetic algorithms and allows one to obtain thresholds for the prediction of slope failures using dates of landslide activations and rainfall series. It can be applied to either single landslides or a set of similar slope movements in a homogeneous environment. Calibration of the model provides families of optimal, discretized solutions (kernels) that maximize the fitness function. Starting from the kernels, the corresponding mobility functions (i.e., the predictive tools) can be obtained through convolution with the rain series. The base time of the kernel is related to the magnitude of the considered slope movement, as well as to the hydro-geological complexity of the site. Generally, shorter base times are expected for shallow slope instabilities compared to larger-scale phenomena. Once validated, the model can be applied to estimate the timing of future landslide activations in the same study area, by employing measured or forecasted rainfall series. Examples of application of GASAKe to a medium-size slope movement (the Uncino landslide at San Fili, in Calabria, southern Italy) and to a set of shallow landslides (in the Sorrento Peninsula, Campania, southern Italy) are discussed. In both cases, a successful calibration of the model has been achieved, despite unavoidable uncertainties concerning the dates of occurrence of the slope movements. In particular, for the Sorrento Peninsula case, a fitness of 0.81 has been obtained by calibrating the model against 10 dates of landslide activation; in the Uncino case, a fitness of 1 (i.e., neither missing nor false alarms) has been achieved using five activations. As for temporal validation, the experiments performed by considering further dates of activation have also proved satisfactory. In view of early-warning applications for civil protection, the capability of the model to simulate the occurrences of the Uncino landslide has been tested by means of a progressive, self-adaptive procedure. Finally, a sensitivity analysis has been performed by taking into account the main parameters of the model. The obtained results are quite promising, given the high performance of the model against different types of slope instabilities characterized by several historical activations. Nevertheless, further refinements are still needed for application to landslide risk mitigation within early-warning and decision-support systems.

Empirical rainfall thresholds are tools to forecast the possible occurrence of rainfall-induced shallow landslides. Accurate prediction of landslide occurrence requires reliable thresholds, which need to be properly validated before their use in operational warning systems. We exploited a catalogue of 200 rainfall conditions that have resulted in at least 223 shallow landslides in Sicily, southern Italy, in the 11-year period 2002–2011, to determine regional event duration–cumulated event rainfall (ED) thresholds for shallow landslide occurrence. We computed ED thresholds for different exceedance probability levels and determined the uncertainty associated to the thresholds using a consolidated bootstrap nonparametric technique. We further determined subregional thresholds, and we studied the role of lithology and seasonal periods in the initiation of shallow landslides in Sicily. Next, we validated the regional rainfall thresholds using 29 rainfall conditions that have resulted in 42 shallow landslides in Sicily in 2012. We based the validation on contingency tables, skill scores, and a receiver operating characteristic (ROC) analysis for thresholds at different exceedance probability levels, from 1% to 50%. Validation of rainfall thresholds is hampered by lack of information on landslide occurrence. Therefore, we considered the effects of variations in the contingencies and the skill scores caused by lack of information. Based on the results obtained, we propose a general methodology for the objective identification of a threshold that provides an optimal balance between maximization of correct predictions and minimization of incorrect predictions, including missed and false alarms. We expect that the methodology will increase the reliability of rainfall thresholds, fostering the operational use of validated rainfall thresholds in operational early warning system for regional shallow landslide forecasting.

The primary trigger of damaging landslides in Italy is intense or prolonged rainfall. Definition of the rainfall conditions responsible for landslides is a crucial issue and may contribute to reducing landslide risk. Criteria for identifying the rainfall conditions that could initiate slope failures are still lacking or uncertain. Expert investigators usually reconstruct rainfall events manually. In this paper, we propose an algorithm for the objective and reproducible definition of rainfall conditions responsible for landslides, from a series of hourly rainfall data. The algorithm, which is implemented in R (http://www.r-project.org), performs a series of actions: (i) removes isolated events with negligible amount of rainfall and random noise generated by the rain gauge; (ii) aggregates rainfall measurements in order to obtain a sequence of distinct rainfall events; (iii) identifies single or multiple rainfall conditions responsible for the slope failures. The result is the objective reconstruction of the duration, D, and the cumulated rainfall, E, for rainfall events, and for rainfall conditions that have resulted in landslides. We tested the algorithm using rainfall and landslide information for the period between January 2002 and December 2012 in Sicily, Southern Italy. The algorithm reconstructed 13,537 rainfall events and 343 rainfall conditions as possible triggers using the information on 163 documented landslides. The comparison between automatic and manually method highlights that most (87.7%) of the rainfall conditions obtained manually were reconstructed accurately. Use of the algorithm is the objective reconstruction of the duration, D, and the cumulated rainfall, E, for rainfall events, and for rainfall conditions that have resulted in landslides. We tested the algorithm using rainfall and landslide should contribute to reducing the current subjectivity inherent in the manual treatment of the rainfall and landslide data.

The large physiographic variability and the abundance of landslide and rainfall data make Italy an ideal site to investigate variations in the rainfall conditions that can result in rainfall-induced landslides. We used landslide information obtained from multiple sources and rainfall data captured by 2228 rain gauges to build a catalogue of 2309 rainfall events with – mostly shallow – landslides in Italy between January 1996 and February 2014. For each rainfall event with landslides, we reconstructed the rainfall history that presumably caused the slope failure, and we determined the corresponding rainfall duration D (in hours) and cumulated event rainfall E (in mm). Adopting a power law threshold model, we determined cumulated event rainfall–rainfall duration (ED) thresholds, at 5% exceedance probability, and their uncertainty. We defined a new national threshold for Italy, and 26 regional thresholds for environmental subdivisions based on topography, lithology, land-use, land cover, climate, and meteorology, and we used the thresholds to study the variations of the rainfall conditions that can result in landslides in different environments, in Italy. We found that the national and the environmental thresholds cover a small part of the possible DE domain. The finding supports the use of empirical rainfall thresholds for landslide forecasting in Italy, but poses an empirical limitation to the possibility of defining thresholds for small geographical areas. We observed differences between some of the thresholds. With increasing mean annual precipitation (MAP), the thresholds become higher and steeper, indicating that more rainfall is needed to trigger landslides where the MAP is high than where it is low. This suggests that the landscape adjusts to the regional meteorological conditions. We also observed that the thresholds are higher for stronger rocks, and that forested areas require more rainfall than agricultural areas to initiate landslides. Finally, we observed that a 20% exceedance probability national threshold was capable of predicting all the rainfall-induced landslides with casualties between 1996 and 2014, and we suggest that this threshold can be used to forecast fatal rainfall-induced landslides in Italy. We expect the method proposed in this work to define and compare the thresholds to have an impact on the definition of new rainfall thresholds for possible landslide occurrence in Italy, and elsewhere.

According to the fifth report of the Intergovernmental Panel on Climate Change, "warming of the climate system is unequivocal". The influence of climate changes on slope stability and landslides is also undisputable. Nevertheless, the quantitative evaluation of the impact of global warming, and the related changes in climate, on landslides remains a complex question to be solved. The evidence that climate and landslides act at only partially overlapping spatial and temporal scales complicates the evaluation. Different research fields, including e.g., climatology, physics, hydrology, geology, hydrogeology, geotechnics, soil science, environmental science, and social science, must be considered. Climatic, environmental, demographic, and economic changes are strictly correlated, with complex feedbacks, to landslide occurrence and variation. Thus, a holistic, multidisciplinary approach is necessary. We reviewed the literature on landslide-climate studies, and found a bias in their geographical distribution, with several studies centered in Europe and North America, and large parts of the world not investigated. We examined advantages and drawbacks of the approaches adopted to evaluate the effects of climate variations on landslides, including prospective modelling and retrospective methods that use landslide and climate records, and paleoenvironmental information. We found that the results of landslide-climate studies depend more on the emission scenarios, the global circulation models, the regional climate models, and the methods to downscale the climate variables, than on the description of the variables controlling slope processes. Using ensembles of projections based on a range of emissions scenarios would reduce (or at least quantify) the uncertainties in the obtained results. We performed a preliminary global assessment of the future landslide impact, presenting a global distribution of the projected impact of climate change on landslide activity and abundance. Where global warming is expected to increase, the frequency and intensity of severe rainfall events, a primary trigger of shallow, rapid-moving landslides that cause many landslide fatalities, an increase in the number of people exposed to landslide risk is to be expected. Furthermore, we defined a group of objective and reproducible methods for the quantitative evaluation of the past and future (expected) variations in landslide occurrence and distribution, and in the impact and risk to the population, as a result of changes in climatic and environmental factors (particularly, land use changes), at regional scale. The methods were tested in a southern Italian region, but they can easily applied in other physiographic and climatic regions, where adequate information is available.

Seismic techniques are increasingly adopted to detect signals induced by mass movements and to quantitatively evaluate geo-hydrological hazards at different spatial and temporal scales. By analyzing landslide-induced seismicity, it is possible obtaining significant information on the source of the mass wasting, as well as on its dynamics. However, currently only few studies have performed a systematic back analysis on comprehensive catalogues of events to evaluate the performance of proposed algorithms. In this work, we analyze a catalogue of 1058 landslides induced by rainfall in Italy. Among these phenomena, there are 234 rock falls, 55 debris flows, 54 mud flows, and 715 unspecified shallow landslides. This is a subset of a larger catalogue collected by the Italian research institute for geo-hydrological protection (CNR IRPI) during the period 2000-2014 (Brunetti et al., 2015). For each record, the following information are available: the type of landslide; the geographical location of the landslide (coordinates, site, municipality, province, and 3 classes of geographic accuracy); the temporal information on the landslide occurrence (day, month, year, time, date, and 3 classes of temporal accuracy); the rainfall conditions (rainfall duration and cumulated event rainfall) that have resulted in the landslide. We consider here only rainfall-induced landslides for which exact date and time were known from chronicle information. The analysis of coeval seismic data acquired by regional seismic networks show clear signals in at least 3 stations for 64 events (6% of the total dataset). Among them, 20 are associated to local earthquakes and 2 to teleseisms; 10 are anomalous signals characterized by irregular and impulsive waveforms in both time and frequency domains; 33 signals are likely associated to the landslide occurrence, as they have a cigar-shaped waveform characterized by emerging onsets, duration of several tens of seconds, and low frequencies (1-10 Hz). For this last category of events, we have applied the approach proposed in Manconi et al. (2016), in order to evaluate the performance of automatic identification, location and first order classification of landslide events trough seismic data only. Our analysis may provide important insights for the development and calibration of landslide identification algorithms, which might be used to enhance the completeness of landslide catalogues by relying on quantitative data.

The study area of the Costa Viola mountain ridge is strongly exposed to shallow landslides triggered by rainfall. Starting from a susceptibility map, recently published by the same team, the related risk has been assessed, limitedly to landslide sources. The hazard has first been evaluated by considering the recurrence periods of triggering events. Economic value and physical vulnerability of the types of elements at risk have been assumed by considering land-use and the effects of similar events in Southern Italy, respectively. The 1:30,000-scale risk map (Main Map) has been produced by combining the above-mentioned layers. Low risk prevails in the study area, whilst the highest values are concentrated along the coast, where villages and the main infrastructure are located. The expected annual damage exceeds 570 million EUR (about 68% pertaining to low-risk cells). Risk maps produced through similar simplified geographic information system-based procedures may play a crucial role in accurate land-use planning and effective decision-making.

A process chain for the definition and the performance assessment of an operational regional warning model for rainfall-induced landslides, based on rainfall thresholds, is proposed and tested in a landslide-prone area in the Campania region, southern Italy. A database of 96 shallow landslides triggered by rainfall in the period 2003–2010 and rainfall data gathered from 58 rain gauges are used. First, a set of rainfall threshold equations are defined applying a well-known frequentist method to all the reconstructed rainfall conditions responsible for the documented landslides in the area of analysis. Several thresholds at different exceedance probabilities (percentiles) are evaluated, and nine different percentile combinations are selected for the activation of three warning levels. Subsequently, for each combination, the issuing of warning levels is computed by comparing, over time, the measured rainfall with the pre-defined warning level thresholds. Finally, the optimal percentile combination to be employed in the regional early warning system, i.e. the one providing the best model performance in terms of success and error indicators, is selected employing the “event, duration matrix, performance” (EDuMaP) method.

On 28 January 2009, a large debris slide was triggered by prolonged rainfalls at the southern suburbs of San Benedetto Ullano (Northern Calabria). The slope movement affected fractured and weathered migmatitic gneiss and biotitic schist, and included a pre-existing landslide. A detailed geomorphologic field survey, carried out during the whole phase of mobilization, allowed to recognize the evolution of the phenomenon. A set of datum points was located along the borders of the landslide and frequent hand-made measurements of surface displacements were performed. Since 11 February, a basic real-time monitoring system of meteoric parameters and of surface displacements, measured by means of high-precision extensometers, was also implemented. Based on the data gained through the monitoring system, and on field surveying, a basic support system for emergency management could be defined since the first phases of activation of the phenomenon. The evolution of the landslide was monitored during the following months: as a consequence, evidence of retrogressive distribution could be recognized, with initial activation in the middle sector of the slope, where new temporary springs were observed. During early May, the activity reduced to displacements of a few millimetres per month and the geo-hydrological crisis seemed to be concluded. Afterwards, the geological scheme of the slope was refined based on the data collected through a set of explorative boreholes, equipped with inclinometers and piezometers: according to the stratigraphic and inclinometric data, the depth of the mobilized body resulted in varying between 15 and 35 m along a longitudinal section. A parametric limit equilibrium analysis was carried out to explore the stability conditions of the slope affected by the landslide as well as to quantify the role of the water table in destabilizing the slope. The interpretation of the process based on field observations was confirmed by the limit equilibrium analysis: the first activation of the landslide was, in fact, to be expected in the middle portion of the slope, provided that the groundwater levels approximate the ground surface in the same sector. On 1 February 2010, another remarkable phase of landslide mobilization began, following a new period of exceptional and prolonged rainfalls. On 11 February, an abrupt stage of slope acceleration was observed, after further extraordinary rainfalls. The slope movement essentially replicated the phases of mobilization observed on 28 January 2009, thus confirming the results of the limit equilibrium analysis. Based on the outcomes of the parametric analysis, the support system for emergency management could then be tentatively refined on a more physical basis.

Prediction of rainfall-induced landslides can rely on empirical rainfall thresholds. These are obtained from the analysis of past rainfall events that have (or have not) resulted in slope failures. Accurate prediction requires reliable thresholds, which need to be validated before their use in operational landslide warning systems. Despite the clear relevance of validation, only a few studies have addressed the problem, and have proposed and tested robust validation procedures. We propose a validation procedure that allows for the definition of optimal thresholds for early warning purposes. The validation is based on contingency table, skill scores, and receiver operating characteristic (ROC) analysis. To establish the optimal threshold, which maximizes the correct landslide predictions and minimizes the incorrect predictions, we propose an index that results from the linear combination of three weighted skill scores. Selection of the optimal threshold depends on the scope and the operational characteristics of the early warning system. The choice is made by selecting appropriately the weights, and by searching for the optimal (maximum) value of the index. We discuss weakness in the validation procedure caused by the inherent lack of information (epistemic uncertainty) on landslide occurrence typical of large study areas. When working at the regional scale, landslides may have occurred and may have not been reported. This results in biases and variations in the contingencies and the skill scores. We introduce two parameters to represent the unknown proportion of rainfall events (above and below the threshold) for which landslides occurred and went unreported. We show that even a very small underestimation in the number of landslides can result in a significant decrease in the performance of a threshold measured by the skill scores. We show that the variations in the skill scores are different for different uncertainty of events above or below the threshold. This has consequences in the ROC analysis. We applied the proposed procedure to a catalogue of rainfall conditions that have resulted in landslides, and to a set of rainfall events that - presumably - have not resulted in landslides, in Sicily, in the period 2002-2012. First, we determined regional event duration-cumulated event (ED) rainfall thresholds for shallow landslide occurrence using 200 rainfall conditions that have resulted in 223 shallow landslides in Sicily in the period 2002-2011. Next, we validated the thresholds using 29 rainfall conditions that have triggered 42 shallow landslides in Sicily in 2012, and 1250 rainfall events that presumably have not resulted in landslides in the same year. We performed a back analysis simulating the use of the thresholds in a hypothetical landslide warning system operating in 2012.

GASAKe is a new hydrological model aimed at forecasting the triggering of landslides. The model is based on genetic algorithms and allows one to obtain thresholds for the prediction of slope failures using dates of landslide activations and rainfall series. It can be applied to either single landslides or a set of similar slope movements in a homogeneous environment. Calibration of the model provides families of optimal, discretized solutions (kernels) that maximize the fitness function. Starting from the kernels, the corresponding mobility functions (i.e., the predictive tools) can be obtained through convolution with the rain series. The base time of the kernel is related to the magnitude of the considered slope movement, as well as to the hydro-geological complexity of the site. Generally, shorter base times are expected for shallow slope instabilities compared to larger-scale phenomena. Once validated, the model can be applied to estimate the timing of future landslide activations in the same study area, by employing measured or forecasted rainfall series. Examples of application of GASAKe to a medium-size slope movement (the Uncino landslide at San Fili, in Calabria, southern Italy) and to a set of shallow landslides (in the Sorrento Peninsula, Campania, southern Italy) are discussed. In both cases, a successful calibration of the model has been achieved, despite unavoidable uncertainties concerning the dates of occurrence of the slope movements. In particular, for the Sorrento Peninsula case, a fitness of 0.81 has been obtained by calibrating the model against 10 dates of landslide activation; in the Uncino case, a fitness of 1 (i.e., neither missing nor false alarms) has been achieved using five activations. As for temporal validation, the experiments performed by considering further dates of activation have also proved satisfactory. In view of early-warning applications for civil protection, the capability of the model to simulate the occurrences of the Uncino landslide has been tested by means of a progressive, self-adaptive procedure. Finally, a sensitivity analysis has been performed by taking into account the main parameters of the model. The obtained results are quite promising, given the high performance of the model against different types of slope instabilities characterized by several historical activations. Nevertheless, further refinements are still needed for application to landslide risk mitigation within early-warning and decision-support systems.

Heavy rainstorms often induce flash floods, shallow landslides and debris flows, which cause several damage to manmade infrastructures and loss of lives. The analysis of spatial distribution and temporal features of intense rainfall events is a fundamental step for a better understanding of the phenomena and for its possible prediction. The present study is an attempt to improve, from a statistical point of view, the understanding at sub-hourly scale of the temporal and spatial structure of intense rainfall events, by examining those that have hit Calabria (Southern Italy) in the years 1998-2008. More in detail, a considerable amount of series with high temporal detail (5 min) related to 155 sites (one rain gauge per less than 100 sq km), were analysed. First, more than 152 thousands rainfall events, separated by at least 6 hours of dry weather, were recognized. Then, less than a third (45,533) were selected, since denoted as erosive. Finally, several heavy rainstorm events (HREs) were chosen by considering the rainfall events recorded simultaneously at different rain gauges, even non-contiguous, within the region. In particular, this further selection was conducted, based on heuristic threshold values of cumulated rainfall ( 100 mm), maximum intensity ( 50 mm/h), and kinetic energy ( 29 MJ/ha). Therefore, 25 distinct HREs, including all the well-known catastrophic geo-hydrological events, were subjected to thorough investigation. The obtained HREs, automatically classified according to their structure in time, were analysed as regards both spatial and temporal evolution. At this end, the 25 HREs were distinguished as widespread (17) or localized (8), if the affected area is 500 sq km or < 500 sq km, respectively. In particular, the temporal storm structure was described by means of the standardized rainfall profile (rainfall amount vs. duration, in terms on cumulative percentages). Then, a 4-digit binary shape code was adopted to automatically identify the shape of the profile (Terranova and Iaquinta, 2011; Terranova and Gariano, 2014). HREs have different spatial extents and temporal patterns. A wide spatial extent of the events does not imply damage proportionally high. Generally, a peak at the beginning of the event (thunderstorm-type) characterizes localized events. On the contrary, widespread events present mixed temporal structures with peaks localized in the last half of their duration. The proposed method improves the knowledge regarding the input of rainfall-runoff watershed models. These models can benefit from design storms, based on the synthesis of recorded rainstorms, having a time structure integrated with the results of the spatial analysis. The notable size of the employed sample, including data with a very detailed time resolution that relate to several rain gauges well distributed throughout the region, gives robustness to the obtained results. References O.G. Terranova, and P. Iaquinta.: Temporal properties of rainfall events in Calabria (southern Italy). Nat. Hazards Earth Syst. Sci., 11, 751-757, 2011. O.G. Terranova, and S.L. Gariano.: Rainstorms able to induce flash floods in a Mediterranean-climate region (Calabria, southern Italy). Nat. Hazards Earth Syst. Sci., 14, 2423-2434, 2014.

In Italy, rainfall-induced shallow landslides are frequent and harmful phenomena. The prediction of their occurrence is of social significance for civil protection purposes. For the operational prediction of rainfall-induced shallow landslides empirical rainfall thresholds based on the statistical analysis of past rainfall conditions that triggered slope failures are commonly used. The paper describes a catalogue of 1981 rainfall events, which caused 2408 shallow landslides in Italy in the period 1996–2012. Information on rainfall-induced landslides was collected searching chiefly online newspaper archives, blogs, and fire brigade reports. For each documented failure, we reconstructed the triggering rainfall conditions (rainfall duration D and cumulated rainfall E) using national and regional rain gauge networks. We analysed the rainfall conditions to determine new ED rainfall thresholds for Italy. The calculated thresholds can be implemented in a landslide forecasting system to mitigate landslide hazard and risk.

Heavy rainstorms often induce flash flooding, one of the natural disasters most responsible for damage to man-made infrastructure and loss of lives, adversely affecting also the opportunities for socio-economic development of Mediterranean Countries. The frequently dramatic damage of flash floods are often detected with sufficient accuracy by post-event surveys, but rainfall causing them are still only roughly characterized. With the aim of improving the understanding of the temporal structure and spatial distribution of heavy rainstorms in the Mediterranean context, a statistical analysis was carried out in Calabria (southern Italy) concerning rainstorms that mainly induced flash floods, but also shallow landslides and debris-flows. Thus a method is proposed – based on the overcoming of heuristically predetermined threshold values of cumulated rainfall, maximum intensity, and kinetic energy of the rainfall event – to select and characterize the rainstorms able to induce flash floods in the Mediterranean-climate Countries. Therefore the obtained (heavy) rainstorms were automatically classified and studied according to their structure in time, localization and extension. Rainfall-runoff watershed models can consequently benefit from the enhanced identification of design storms, with a realistic time structure integrated with the results of the spatial analysis. A survey of flash flood events recorded in the last decades provides a preliminary validation of the method proposed to identify the heavy rainstorms and synthetically describe their characteristics. The notable size of the employed sample, including data with a very detailed resolution in time, that relate to several rain gauges well-distributed throughout the region, give robustness to the obtained results.

In Italy, intense or prolonged rainfall is the primary trigger of damaging landslides. The identification of the rainfall conditions responsible for the initiation of landslides is a crucial issue and may contribute to reduce landslide risk. Objective criteria for the identification of rainfall conditions that could initiate slope failures are still lacking or ambiguous. The reconstruction of rainfall events able to trigger past landslides is usually performed manually by expert investigators. Here, we propose an algorithm that reconstructs automatically rainfall events from a series of hourly rainfall data. The automatic reconstruction reproduces the actions performed by an expert investigator that adopts empirical rules to define rainfall conditions that presumably initiated the documented landslides. The algorithm, which is implemented in R (http://www.r-project.org), performs three actions on the data series: (i) removes isolated events with negligible amount of rainfall and random noise generated by the rain gauge; (ii) aggregates rainfall measurements in order to obtain a sequence of distinct rainfall events; (iii) identifies single or multiple rainfall conditions responsible for the slope failures. In particular, the algorithm calculates the duration, D, and the cumulated rainfall, E, for rainfall events, and for rainfall conditions that have resulted in landslides. A set of input parameters allows the automatic reconstruction of rainfall events in different physical settings and climatic conditions. We tested the algorithm using rainfall and landslide information available to us for Sicily, Southern Italy, in the period between January 2002 and December 2012. The algorithm reconstructed 13,537 rainfall events and 343 rainfall conditions as possible triggers of the 163 documented landslides. Most (87.7%) of the rainfall conditions obtained manually were reconstructed accurately. Use of the algorithm shall contribute to an objective and reproducible definition of rainfall conditions responsible for landslides in different geographic areas, reducing the current subjectivity inherent in the manual treatment of the rainfall and landslide data.