NH3.7

Early Warning Systems are one of the most effective tools for reducing disaster risk, however the development of Landslide Early Warning Systems (LEWS) is complicated due to the random nature of landslide occurrence and the uncertainty in mapping the parameters that cause them. Local LEWS have been effective for known landslides, but regional scale LEWS based on rainfall thresholds have not been very effective up to now. In recent years physically-based multi-hazard models have been developed which allow to predict mass movement hazards at a local scale. However, it is still difficult to apply these in LEWS in a local scale due to the coarse resolution of rainfall estimates and the high computational modelling requirements for running such models real-time. On the other hand, machine learning approaches have been used to assess the relationship between the distribution of the landslide hazard and the catchment morphometric features.

This research applies a physically-based multi-hazard model combined with machine learning to forecast the mass movement impact, based on rainfall predictions in an area in Java, Indonesia. The landslide inventory was developed using a combination of local reporting data and machine learning techniques. The integrated physically-based multi hazard model OpenLISEM is used to create a database of hazard intensity maps under various rainfall scenarios. The resulting hazard intensity maps are subsequently used to subdivide the area in homogeneous zones for which warning levels are given. Machine learning is used to query the database and extract the most likely hazard intensity map based on the rainfall prediction. The intensity is then combined with exposure information of people, buildings, transportation infrastructure and agriculture to provide impact forecasts. The output of combining physically-based models with machine learning approaches has the potential to improve the prediction of landslide impact. The method also allows to make more specific local decisions related to the actions for various levels of warning (e.g. increased vigilance, removal of resources, evacuation of people). The method is currently under development as part of an Indonesian-Netherlands collaboration project to develop a blueprint to use tailored rainfall data, in combination with empirical and physically-based hydrological and landslide models, and historical landslide data for the development of thresholds for landslides and debris flows, as the basis for early warning at settlement level, applied to several test sites in Java.

How to cite: Subiyantoro, A., Westen, C. V., Bout, B. V. D., Victor, J., Muntohar, A., Mushthofa, A., Yuniawan, R. A., and Satyaningsih, R.: Development of a local impact-based Landslide Early Warning System using physically-based multi-hazards modelling and machine learning in Java, Indonesia., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6969, https://doi.org/10.5194/egusphere-egu22-6969, 2022.

As observed worldwide during the last decades, landslides are one of the deadliest natural hazards in mainland Portugal and Azores archipelago being responsible for significant direct and indirect societal and economic losses, justifying the implementation of a landslide early warning system at the regional scale.

The BeSafeSlide project aims to develop and implement a soft technology/low-cost prototype for precipitation-triggered landslide early warning system (LEWS) in Portugal. We plan it to allow be adaptable to a changing climate and a changing land use by working with different climate scenarios. Future changes on regional rainfall patterns due to climate change were evaluated in the LEWS for 2071-2100 period, considering two emission scenarios: RCP 4.5 and RCP 8.5. To evaluate future exposure trends and effects in risk analysis, simulations of changes in land use, by the end of the 21th century, will be carried out. The uncertainty of future projections will be addressed by developing a set of different scenarios.

The LEWS prototype for Portugal is sustained on different types of regional rainfall thresholds for landslide occurrence based on daily/hourly rainfall series available for each BeSafeSlide study area. The proposed prototype aims at integrating 3-day rainfall forecasts on rainfall thresholds monitoring and on dynamic physically based susceptibility models, to anticipate changes in hydrological conditions and consequently on the spatio-temporal occurrence of landslides. Special attention is given to two different types of rainfall-triggered landslide events, recognized as responsible for shallow and deep-seated landslides occurrence on natural slopes, which are permanently monitored within the regional early warning system in hotspot risk areas: (i) landslide events associated to intense, short-duration rainfall periods; and (ii) landslide events associated to long-lasting rainfall periods.

The LEWS main goals are to provide information to civil protection services to anticipate and manage people’s evacuation from landslide prone areas and to ensure the maintenance and operability of regional transport, energy and communications networks and the safeguarding of people´s lives. Although the LEWS is being developed within the framework of Portugal we expect to be applicable in different settings. The application of the LEWS will define warning communication procedures, assess response capacity of stakeholders and develop social capacity practices, to reduce vulnerability and mitigate risk, providing a reduction of affected people, economic losses and critical infrastructures/basic services disruptions.

Acknowledgments: This work was financed by national funds through FCT (Foundation for Science and Technology, I. P.), in the framework of the project BeSafeSlide – Landslide early warning soft technology prototype to improve community resilience and adaptation to environmental change (PTDC/GES-AMB/30052/2017), and the Research Unit UIDB/00295/2020 and UIDP/00295/2020.

How to cite: Oliveira, S. C., Zêzere, J. L., Trigo, R. M., Marques, F., Tavares, A., Marques, R., Ramos, A. M., and Melo, R.: BeSafeSlide – A Landslide early warning soft technology prototype to improve community resilience and adaptation to environmental change, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10685, https://doi.org/10.5194/egusphere-egu22-10685, 2022.

Landslides are one of the most disastrous natural hazards that frequently occur in Indonesia. Since 2017, Balai Sabo has developed an Indonesia Landslide Early Warning System (ILEWS) by utilizing a single rainfall threshold for an entire nation. This condition might lead to inaccuracy of the landslide prediction. Therefore, this study aims to improve the accuracy of the system by updating the rainfall threshold. This study focused on Java Island, where most of the landslides in Indonesia occur. We analyzed 420 landslide events with the one-day and three-day cumulative rainfall for each landslide event. Rainfall data were obtained from the Global Precipitation Measurement (GPM), which is also used in the ILEWS. We propose four methods to derive the thresholds, 1^st is the existing threshold applied in the Balai Sabo-ILEWS, the 2^nd and the 3^rd use the average and minimum of rainfall that trigger landslides, respectively, and the 4^th uses the minimum values of rainfall that induce major landslides. We employed the Receiver Operating Characteristic (ROC) analysis to evaluate the predictability of the rainfall thresholds. The 4^th method shows the best result compared to the others, and this method provides a good prediction of landslide events with a low error value. The chosen threshold will be used as a new threshold in the Balai Sabo-ILEWS.

How to cite: Yuniawan, R. A., Rifai, A., Faris, F., Westen, C. V., Jetten, V., den Bout, B., Subiyantoro, A., Muntohar, A., Musthofa, A., Hidayat, R., Hidayah, A., Ridwan, B., Priangga, E., Satyaningsih, R., and Sutanto, S.: Revisited Rainfall Threshold In the Indonesia Landslide Early Warning System, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7863, https://doi.org/10.5194/egusphere-egu22-7863, 2022.

In this work, we use a probabilistic approach for modelling rainfall thresholds (Caine 1980) triggering shallow landslides with a case study for the Alpes-Maritimes region (France).

In particular, the CTRL-T algorithm (Melillo and al. 2018) is tested to output critical rainfall thresholds, based on the accumulated rainfall – duration parameters (E-D), for different exceedance probabilities from respectively a landslide and two climate datasets. The first climate dataset stores high resolution gridded rainfall data (1km resolution, hourly) and the second climate dataset contains lower resolution gridded rainfall, snow, temperature and evapotranspiration data (8km resolution, daily); the first dataset provides the rainfall records directly used for defining the rainfall events and then for the threshold construction; the second one enables to assess the region’s climate via parameters imported in CTRL-T. The thresholds are then validated using a method designed by Gariano and al. (2015).

Several improvements are made to the method. First, potential evapotranspiration values approximated from temperatures and latitudes in one of the process’ steps are replaced by values from the second climate dataset, the result accounting best for the regional climate. Then, climate-specific duration values, used to split the raw rainfall records in events and sub-events, are computed for each mesh point. This second modification enables considering the heterogeneity of the Alpes-Maritimes climate.

Rainfall thresholds are eventually obtained for different exceedance probabilities, first from a set of probable conditions (MRC), then from a set of highly probable conditions (MPRC). The validation process strengthens the analysis as well as enables to identify best performing thresholds. This work represents novel scientific progress towards landslide reliable warning systems by (a) making a case study of probabilistic rainfall thresholds for Alpes-Maritimes, (b) using for the first time high-resolution rainfall data and (c) adapting the method to climatically heterogeneous zones.

How to cite: Barthelemy, S., Bernardie, S., and Grandjean, G.: Assessing rainfall triggering of shallow landslides with an automatic tool generating thresholds: a case study for the Alpes-Maritimes region, France, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5864, https://doi.org/10.5194/egusphere-egu22-5864, 2022.

Rainfall intensity-duration landslide-triggering thresholds have been proposed as a possible component for the implementation of territorial landslide early warning systems. Given a set of rainfall and landslide data, three approaches can be distinguished to determine thresholds: (i) methods based on triggering events only, (ii) methods based on the non-triggering events only, and (iii) methods based on both type of rainfall events. The aim of the present research is to compare these three possible approaches based on statistical criteria: robustness, sampling variation, and performance. This comparison can provide an insight on which of the three approaches is more appropriate based on the dataset that happens to be available for the area of interest.

We address these aspects by setting up a virtual simulation framework combining a stochastic rainfall model with a hydrological and slope stability model, which allows to make repeated experiments and to simulate different uncertainty conditions.

Our analysis shows that methods based on triggering rainfall only can be the worst with respect to the three investigated statistical properties. Methods based on both triggering and non-triggering rainfall have the highest performances in terms of the ROC true skill statistic; they are also robust, but still require a quite large sample to sufficiently limit the sampling variation of the threshold parameters. On the other side, methods based on non-triggering rainfall only, which are mostly overlooked up, are characterized by good robustness and low sampling variation. It can also be shown that in realistic scenarios their performances can be acceptable and even higher than thresholds derived from triggering events only. Indeed, the use of triggering rainfall only, a common practice in the past literature, yields to thresholds with the worse statistical properties, except when there is a clear separation between triggering and non-triggering events.

Based on these results, it can be stated that methods based on non-triggering rainfall only deserve wider attention, as they have also the practical advantage that can be in principle used where limited information on landslide occurrence is available. The fact that relatively large samples (about 200 landslides events) are needed for a sufficiently precise estimation of threshold parameters when using triggering rainfall, provides a possible insight on the level of uncertainty of thresholds proposed in the past literature.

How to cite: Peres, D. J. and Cancelliere, A.: An analysis of robustness, sampling variation and performances of landslide triggering thresholds determined by different approaches, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-3630, https://doi.org/10.5194/egusphere-egu22-3630, 2022.

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.

How to cite: Steger, S., Kohrs, R., Crespi, A., Moreno, M., Zellner, P. J., Goetz, J., Mair, V., Gariano, S. L., Brunetti, M. T., Melillo, M., Peruccacci, S., and Pittore, M.: A data-driven approach to establish prediction surfaces for rainfall-induced shallow landslides in South Tyrol, Italy, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5902, https://doi.org/10.5194/egusphere-egu22-5902, 2022.

In this research the occurrence of landslides triggered by rainfall is investigated through the assessment of rainfall thresholds for three regions in Portugal: Lisbon, Oporto, and Coimbra. An historical landslide inventory is used, based on newspapers published between 1865 and 2010, associated with daily precipitation databases from three long-term meteorological stations. An empirical approach based on antecedent rainfall is applied to define the rainfall thresholds for landslides occurrence. The analysis is focused on each single rain gauge, for which the spatial representativeness is evaluated. The daily rainfall data is analysed using the Gumbel probability distribution for different durations. The critical rainfall combinations (cumulated rainfall duration) with the highest return period are associated with the landslide occurrence.

For the three regions, rainfall thresholds are defined from regression (linear and potential), extreme values (upper and lower) and probability (probability of a rainfall event resulting in a landslide event when the threshold is exceeded), and the results are assessed and calibrated using the receiver operating characteristic (ROC) metrics. The thresholds comparation reveal regional patterns in rainfall thresholds. The differences in regional critical rainfall conditions for landslide occurrence between regions are associated with geological, geomorphological, and climatic features.

This work is part of the project BeSafeSlide (BSS) - Landslide Early Warning soft technology prototype to improve community resilience and adaptation to environmental change [PTDC/GES-AMB/30052/2017]. JL Zêzere was supported by the RiskCoast project - Development of tools to prevent and manage geological risks on the coast linked to climate change, Interreg SUDOE [SOE3/P4/EO868]

How to cite: Vaz, T. and Zêzere, J. L.: Empirical rainfall thresholds for landslide activity based on long-term Portuguese meteorological stations, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6389, https://doi.org/10.5194/egusphere-egu22-6389, 2022.

The prediction of debris flows is relevant because this type of natural hazard can pose a threat to humans and infrastructure. Debris-flow (and landslide) early warning systems often rely on rainfall intensity–duration (ID) thresholds. Multiple competing methods exist for the determination of such ID thresholds but have not been objectively and thoroughly compared at multiple scales, and a validation and uncertainty assessment is often missing in their formulation. As a consequence, updating, interpreting, generalizing and comparing rainfall thresholds is challenging. Here, we present the findings of Hirschberg et al. (2021), which focused on (i) uncertainties related to ID thresholds, (ii) differences in local compared to regional ID thresholds, and (iii) how prediction can potentially be improved using statistical learning algorithms. The findings are of interest for debris-flow (and landslide) early-warning developers.

We use a 17-year record of rainfall and 67 debris flows in a Swiss Alpine catchment (Illgraben) to determine ID thresholds and associated uncertainties as a function of record du- ration. This included comparing two methods for rainfall threshold definition based on linear regression and/or true-skill-statistic maximization. The main difference between these approaches and the well-known frequentist method is that non-triggering rainfall events were additionally considered for obtaining ID-threshold parameters. Depending on the method applied, the ID-threshold parameters and their uncertainties differed significantly. We found that 25 debris flows are sufficient to constrain uncertainties in ID-threshold parameters to ±30% for our study site. We further demonstrated the change in predictive performance of the two methods if a regional landslide data set with a regional rainfall product was used instead of a local one with local rainfall measurements. Hence, an important finding is that the ideal method for ID- threshold determination depends on the available landslide and rainfall data sets. Furthermore, for the local data set we tested if the ID-threshold performance can be increased by considering other rainfall properties (e.g. antecedent rainfall, maximum intensity) in a multivariate statistical learning algorithm based on decision trees (random forest). The highest predictive power was reached when the peak 30 min rainfall intensity was added to the ID variables, while no improvement was achieved by considering antecedent rainfall for debris-flow predictions in Illgraben. Although the increase in predictive performance with the random forest model over the classical ID threshold was small, such a framework could be valuable for future studies if more predictors are available from measured or modelled data.

How to cite: McArdell, B., Hirschberg, J., Badoux, A., Leonarduzzi, E., and Molnar, P.: Uncertainties in local and regional mass movement prediction using rainfall, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-990, https://doi.org/10.5194/egusphere-egu22-990, 2022.

After the 2008 Wenchuan earthquake (Mw 7.9), increased occurrence of rainfall-induced debris flows was initially observed in the earthquake-hit region (Sichuan, China). In the following years, the frequency of debris flows gradually reduced, indicating a progressive recovery of stability of debris deposits accumulated along slopes and in gullies after the earthquake. To assess these dynamically changing conditions, empirical thresholds have been identified to predict post-seismic debris flow occurrence with two approaches: a meteorological approach based only on precipitation characteristics, and a hydrometeorological approach that also considers the hydrologic conditions before the onset of rainfall. Both used the available record of precipitations and debris flows that occurred between 2008 and 2015 in several gullies, tributary of the upper Minjiang river course, in Wenchuan county. Hydrometeorological thresholds for debris flows were identified at the gully catchment scale, by assessing the water balance with a simplified lumped hydrological model, based on the Budyko framework. The parameters of the model were estimated based on the scarce available information about the water balance of the entire watershed of the upper Minjiang. Simulated catchment water storage was used as a proxy of the moisture state of the slopes. The results indicate that both meteorological and hydrometeorological thresholds allow catching the progressive recovery of stability of the debris deposits. Specifically, the assessment of water balance at the catchment scale highlights the role played by the hydrological processes affecting the slopes, leading to the definition of reliable thresholds, that resulted robust despite the uncertainty of the estimated parameters of the hydrological model. Therefore, the hydrometeorological approach appears suitable to define thresholds for early warning of debris flows at the catchment scale.

How to cite: Greco, R., Marino, P., and Fan, X.: Hydrometeorological thresholds based on catchment storage to predict changes in debris-flow susceptibility after the Wenchuan earthquake (Sichuan, China), EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8081, https://doi.org/10.5194/egusphere-egu22-8081, 2022.

Norway's high-relief landscape is susceptible to gravity-driven natural hazards including snow avalanches, landslides, debris flows, and rockfalls. Rockfalls are the most numerous geohazard in Norway. There are currently over 35 000 rockfall events registered in Norway's national hazard database, accounting for nearly 50% of the total number of events for all hazard types. Rockfalls commonly impact the functioning of infrastructure assets such as roads and railways, and occasionally damage buildings and result in death.

The relationship between rockfall events and weather conditions is recognised but not straightforward. Several hydrometeorological variables are significant for rockfall triggering including precipitation, snow melt, freezing and thawing, temperature, insolation, and soil or rock moisture. The highest frequency of rockfall activity in Norway is observed in spring, a period of snowmelt and freeze-thaw cycling. Given the links to meteorological variables, rockfall frequency is expected to change with climate, altering the exposure of population and infrastructures to rockfalls.

Rockfall risk mitigation at regional scale is challenging. Early warning systems are a helpful tool to depict the time and location of future rockfall events so that emergency managers can act in advance. At present, most existing rockfall early warning systems (REWS) are based on the monitoring and analysis of seismic signals to determine the movement of boulders or the cracking of joints. Little previous research has been conducted to analyse the meteorological conditions that could trigger rockfalls. There is currently no REWS in Norway.

The main objective of this work is to investigate the feasibility of using hydrometeorological thresholds for regional scale rockfall warning. To do so rainfall, temperature, and soil moisture data from SeNorge.no, and the rockfall inventory contained in the Norwegian national hazard database have been analysed to find relations between the hydrometeorological conditions and the triggering of rockfalls in Norway.

How to cite: Palau, R. M., Gilbert, G. L., Solheim, A., Capobianco, V., and Gisnås, K.: Are hydrometeorological thresholds useful for regional-scale rockfall early warning systems? A preliminary analysis of the hydrometeorological conditions leading to rockfalls in Norway, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8464, https://doi.org/10.5194/egusphere-egu22-8464, 2022.

A combination of extreme environmental conditions such as high soil moisture content and heavy or prolonged precipitation contribute to landslide initiation in mountainous areas worldwide. On-site soil moisture monitoring equipment and rain gauge have been widely used to record these variables despite the sparse spatial coverage. Satellite‐based technologies provide estimates of rainfall and soil moisture over large spatial areas sufficient to be explored for landslide hazard assessment in data scarce regions. In this study, we used statistical metrics to compare the gauge based to the satellite precipitation products: TRMM42, CHIRPS, PERSIANN-CDR, GLDAS-2.1, CFSV2, GPM-IMERG, and ERA-5 and assess their performance. Similarly, high resolution satellite and hydrological model derived soil moisture was compared to the automated soil moisture observations at Rwanda weather station sites to assess the usefulness in empirical landslide hazard assessment thresholds in Rwanda. Based on statistical indicators, the NASA GPM based IMERG showed the highest skill to reproduce the main spatiotemporal precipitation patterns. Similarly, the satellite and hydrological model derived soil moisture broadly reproduce the in situ measured soil moisture. The landslide explanatory variables from IMERG satellite precipitation; event rainfall volume E and Duration D in bilinear thresholds framework reveal promising results with improved landslide prediction capabilities in terms of true positive alarms ~80-90% and low rate of false alarms ~14-16%. However, the incorporation of satellite and model derived antecedent soil moisture to the empirical landslide hydro-meteorological thresholds showed no significant improvement. This may be attributed to the probable long and no constant timescale of the defined landslide triggering events that could be shortened to further improve the landslide prediction and support the early warning system development in Rwanda.

How to cite: Uwihirwe, J., Riveros, A., Sadeghi Tehrani, F., Sperna Weiland, F., Hrachowitz, M., and A. Bogaard, T. A. B.: The potential of Satellite and model derived precipitation and soil moisture for estimation of landslide hazard thresholds in Rwanda, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9375, https://doi.org/10.5194/egusphere-egu22-9375, 2022.

Landslide susceptibility maps (LSMs) are inevitable parts of regional scale landslide forecasting models. The susceptibility maps can provide the spatial probability of occurrence of landslides and have crucial role in the development and planning activities of any region. With the wide availability of satellite-based data and advanced computational facilities, data driven LSMs are being developed for different regions across the world. Since a decade, machine learning (ML) algorithms have gained wide acceptance for developing LSMs and the performance of such maps depends highly on the quality of input data and the choice of ML algorithm. This study employs a k fold cross validation technique for evaluating the performance of five different ML models, viz., Naïve Bayes (NB), Logistic Regression (LR), Random Forest (RF), K Nearest Neighbors (KNN) and Support Vector Machines (SVM), to develop LSMs, by varying the train to test ratio. The ratio is varied by changing the number folds used for k fold cross validation from 2 to 10, and its effect on each algorithm is assessed using Receiver Operating Characteristic (ROC) curves and accuracy values. The method is tested for Wayanad district, Kerala, India, which is highly affected by landslides during monsoon. The results show that RF algorithm performs better among all the five algorithms considered, and the maximum accuracy values were obtained with the value of k as 8, for all cases. The variation between the minimum and maximum accuracy values were found to be 0.6 %, 0.74 %, 1.71 %, 1.92 % and 1.83 % for NB, LR, KNN, RF and SVM respectively.

How to cite: Abraham, M. T., Satyam, N., and Pradhan, B.: Effect of data splitting and selection of machine learning algorithms for landslide susceptibility mapping, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6795, https://doi.org/10.5194/egusphere-egu22-6795, 2022.

Rainfall-induced shallow landslides and debris flows often cause casualties and significant damage to property. Territorial landslide early warning systems are recognized as an important countermeasure to avoid or reduce fatalities during rainfall events. A reliable warning model is a key component of these systems. Warning models operating over large areas usually relate the occurrence of landslides to rainfall monitoring data adopting appropriate thresholds (e.g., intensity-duration, cumulated rainfall-duration, hourly/daily rainfall indicators). The increasing availability of large sets of atmospheric and land monitoring data represents an opportunity to upgrade and improve existing landslide warning models. At the same time, appropriately treating such data may pose a significant challenge to analysts that are used to deal with much smaller amounts of data.

The objective of this preliminary study is to demonstrate that machine learning techniques can be effectively used to process monitoring data over large areas at regional scale, with the aim of defining and selecting the variables that best correlate with the initiation of shallow landslides and debris flows. The machine learning models have been tested in one of the warning zones defined by the regional civil protection agency for hydrogeological risk management in Campania (Italy). Two categories of data are used for the analyses: distributed monitoring data, and a landslide inventory. The monitoring variables are derived from the fifth generation of ECMWF atmospheric reanalysis (ERA5), available with a spatial resolution of about 31 km and a temporal resolution of 1 h (http://dx.doi.org/10.24381/cds.adbb2d47). Data on landslide events come from “FraneItalia”, a geo-referenced openly available catalogue of Italian landslides created consulting online news from 2010 onwards (http://dx.doi.org/10.17632/zygb8jygrw.2). Different machine learning models have been defined, trained, and tested to relate the occurrence of landslides in the case study area to multiple variables arising from different combinations of the adopted monitoring data, mainly rainfall and soil water content. The performance of these models is evaluated by means of standard contingencies and skill scores. The best performing variables are used to define an optimal multivariate threshold to be adopted in the landslide warning model. The results of the optimal model are also compared with the outcomes of an application of a more classical exceedance probability statistical methodology based on cumulated rainfall-duration thresholds.

How to cite: Calvello, M., Pecoraro, G., Esposito, M., Pota, M., Rianna, G., and Reder, A.: Using machine learning for defining distributed monitoring variables correlated to the occurrence of rainfall-induced shallow landslides and debris flows: a case study in Campania region, Italy, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5272, https://doi.org/10.5194/egusphere-egu22-5272, 2022.

On August 8, 2017, a magnitude 7 earthquake struck Jiuzhaigou County, Aba Prefecture, Sichuan Province, inducing a large number of landslides. Evaluating the susceptibility to landslides induced by strong earthquakes can provide a scientific basis for disaster risk management and monitoring. However, different evaluation models can obtain different spatial distributions of landslide susceptibility, and thus, selecting the optimal model is the most effective way to improve the susceptibility evaluation. To select the most suitable evaluation model for a strong earthquake area (Jiuzhaigou), 12 influencing factors affecting the landslide occurrence, including slope, elevation, and aspect, were extracted, and different statistical analysis methods and machine learning models were used to calculate the susceptibility index. The results show that the deep neural network model had the highest accuracy (85.4%), followed by the random forest and support vector machine models (84.2% and 82.3%, respectively), while the logistic regression model and certainty factor models achieved accuracies of 80.8% and 76.2%, respectively. Accordingly, the deep neural network model can be considered a new tool to achieve the more accurate zonation of landslide susceptibility in meizoseismal regions.

How to cite: Wang, X., Wang, D., and Li, S.: Multi-model-based evaluation of landslide susceptibility in a meizoseismal area, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6895, https://doi.org/10.5194/egusphere-egu22-6895, 2022.

An Idealized curve surface (ICS) with two constant curvatures was proposed in Tai et al. (2020) for mimicking the plausible landslide failure surface in numerical simulation.  For ease of illustrating the ICS, Ko et al. (2021) suggested the concept of a reference ellipse for constructing the associated ICS, i.e. the ellipse-ICS method. Hence, with respect to a landslide-prone area, the most appropriate ICS can be figured out by translating, rotating and side-tilting the reference ellipse.

In the present study, the criteria for the searching the most appropriate ICS consist of the terrain characteristics (cracks, scarps, erosion gullies) and the data of the gauging station (inclinometer and groundwater indicators), where the terrain characteristics indicate the plausible boundary of the failure area, the records of inclinometer help to identify the (local) depth of sliding surface. Since the inclinometer and groundwater indicators provide the local data only, the proposed ellipse-ICS method is employed as an efficient tool to construct the plausible ICS and to investigate the impacts of the groundwater distribution on the slope stability.

The ellipse-ICS method is therefore applied to two potential large-scale landslide areas in Taiwan, i.e., the T003 at Yanping Township in eastern Taiwan and the T002 at Fuxing District in northern Taiwan. The ICSs are identified with respect to the failure depths measured by inclinometer, where the safety factors are estimated. Together with the numerical approach given in Tai et al. (2019), the subsequent flow paths of post-failure can be estimated and may serve as useful information for hazard assessment.

Keywords:

ellipse-ICS, inclinometer, groundwater level, safety factors, flow paths

References

• Tai, Y. C., Heß, J., & Wang, Y. (2019). Modeling Two‐Phase Debris Flows with Grain‐Fluid Separation over Rugged Topography: Application to the 2009 Hsiaolin Event, Taiwan. Journal of Geophysical Research: Earth Surface, 124(2), 305-333.
• Tai, Y. C., Ko, C. J., Li, K. D., Wu, Y. C., Kuo, C. Y., Chen, R. F., & Lin, C. W. (2020). An idealized landslide failure surface and its impacts on the traveling paths. Frontiers in Earth Science, 8, 313.
• Ko, C. J., Wang, C. L., Wong, H. K., Lai, W. C., Kuo, C. Y. & Tai, Y. C. (2021). Landslide Scarp Assessments by Means of an Ellipse-Referenced Idealized Curved Surface. Frontiers in Earth Science, 9,862.

How to cite: Wong, H.-K., Wang, C.-L., Ma, C.-Y., and Tai, Y.-C.: Integration of the Terrain Characteristics and Data of Gaging Station for Mimicking the Plausible Surface of Slope Failures., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10918, https://doi.org/10.5194/egusphere-egu22-10918, 2022.

Geohazard monitoring is a key component in an early warning system (EWS). The implementation of monitoring actions provides data for the acquisition of variables related to the landslide, its triggering and kinematic; additionally, it can provide insights of its evolution in time in order to plan mitigation actions, including alarms and warnings. The use of high-frequency systems can also provide such data in the shortest time thus optimizing the aforementioned actions. In the last decades, numerous surface monitoring systems were developed, with various features, providing punctual information, like GNSS or Robotized Total Stations, at high frequency or large-scale data, i.e. Remote Sensing, at lower temporal resolution. The choice of the best one is related to the goals to be fulfilled but, independently from the selected method, surface techniques monitor only a displacement resulting from the sum of all the deep-seated ground deformations. To properly detect the subsoil behavior of a landslide, the use of subsurface sensors is necessary. To couple the subsoil survey with high frequency monitoring a robotic inclinometric system was developed, and patented, by the Geohazard Monitoring Group (GMG) of CNR-IRPI. The instrumentation features the operational characteristics of the manual inclinometric measures (reliability, double readings 0/180˚…) but integrates the advantages of the robotization (accuracy, measurement frequency…), too. The robotized instrumentation also called “Automated Inclinometer System” (AIS) allows the automatic exploration of all the borehole length (up to 120 meters in the standard configuration) with a single probe. The AIS is remotely connected by a 4G modem so it is possible to define the acquisition parameters, download measured data and check the device functioning parameters. The instrumentation was deployed, at the beginning of December 2021, in a borehole located in Passiria Valley (northeastern Italy) to monitor a large and slow-moving landslide involving the whole mountain face; thanks to instrumentation modularity, the AIS is ready to measure after only 4÷5 hours of installation time. Concurrently with the main installation, a GNSS benchmark was positioned and surveyed to provide, with the next measurement campaigns, a crosscheck with the AIS results. After 10÷15 days of monitoring at 1 measurement/day the landslide’s sliding surface, its depth and deformation rate, were clearly identified, thus confirming the capability of the AIS to perform early detection of the landslide kinematic. This result is key information in the risk reduction chain as it shortens the time necessary to achieve the numerical parameters describing the landslide and, consequently, plain the following, mitigating, actions.

How to cite: Godone, D., Allasia, P., Baldo, M., Guenzi, D., and De Polo, F.: The use of robotized inclinometric system in Early Warning System. The case study of a large landslide monitoring., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7388, https://doi.org/10.5194/egusphere-egu22-7388, 2022.

Waste materials produced by mining activities (tailings) can be collected in artificial ponds delimited by earth embankments (tailings dams). In case of tailings dam failure, the consequences are often catastrophic for the surrounding communities and livelihoods as this rupture may release large amounts of tailings and mining wastewater that moves downstream. Furthermore, the mining by-products cause, in many cases, a devastating impact on the surrounding environments and ecosystem. As an increased trend of tailings dam failure has been observed in the last decade, there is an urgent demand from the industry as well as the civil society and the investor community to gain a broader understanding of the risks posed by tailings facilities. Furthermore, efficient techniques to monitor and predict the failure of tailings dams are also crucial.
 
This study investigates how the satellite remote sensing interferometric synthetic aperture radar (InSAR) technique can be used to monitor tailings dams and the applicability of the inverse velocity method to predict failures. InSAR data have been used to map surface displacement prior to dam failures in two case studies: the Feijao tailings dam in Brazil and the Cadia tailings dam in Australia. In the case of the Feijao dam, both the SBAS and PS techniques were applied to process displacement time-series from the satellite data. For the Cadia dam, data processing was carried out using the SqueeSAR algorithm.

The inverse velocity method uses surface displacement measurement points to predict a time of failure. For the Feijao dam InSAR dataset, the inverse velocity method was applicable to different periods presenting an evident increase in the displacement rate. However, it was difficult to retrieve any reliable indication of failure. Contrary to the Feijao dam, the results from the Cadia dam shows a significantly accelerating deformation with time, and by applying the inverse velocity method a predicted time of failure can be retrieved in good agreement with the actual failure.  

How to cite: Svendsen, I., Piciullo, L., Vöge, M., Montalti, R., and Intrieri, E.: Tailings dam monitoring and early warning with InSAR technique, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11011, https://doi.org/10.5194/egusphere-egu22-11011, 2022.

Abstract:

Karst aquifers belong to the fractured aquifer family. The Zaghouan region located in NE of Tunisia (North Africa) is characterized by a high degree of karstification due to the climate impact and the development of fracture network. Survey using electrical resistivity tomography (ERT) is deployed to provide a cost-effective characterization of the subsurface karst environments. A total of three ERT profiles with a length of 300 meters were evaluated in Zaghouan region.

The area represents an anticline of Jurassic limestone rocks, which is overlain by a thin clay layer. In this study, an ERT survey was conducted to examine the spatial distribution and shape of underground cavities in the karst area of Jebel Bent Saidan. In this study, geological, hydrogeological and electrical resistivity tomography (ERT) methods were applied to determine the geometry of the karst aquifer in the Zaghouan area (NE Tunisia). The area is characterized by fractured and karstic limestone aquifer of Jurassic. Three resistivity profiles were carried out along the study area (Jebel Bent Saidan). The correct resistivity data was interpreted using ZONDRes 2D software.  The results of the interpreted geo-electrical sections showed that the resistivity of the carbonate aquifer ranges from 350 to over 4000 Ωm. The thickness of the aquifer varies between 15 and 30 meters, while its depth from the surface is between 10 and 40 meters. The ERT not only provided accurate near-surface information, but was also very useful in establishing the geometry of the aquifer. It was also very useful in establishing the 3D geometry and position of several potential karst cavities and conduits. The results show the presence of two large isolated cavities at different depths. The low resistivity of karst cavities in the Jurassic carbonate of Jebel Bent Saidane was explained by the saturation of groundwater. The ERT imaging technique using to identify and characterize the discontinuities, faults and water investigation of the fractured and karstified limestone aquifers in the Bent Saidan Mounts. The conducted research demonstrated that the ERT method was an effective tool for imaging the subsurface in the karst terrain.

Keywords: Bent Saidan (NE Tunisia), karst aquifers, electrical resistivity tomography (ERT), cavities.

How to cite: Mhimdi, A., Gabtni, H., Ezzine, I., Hamzaoui, F., Ghanmi, M., and Bouhlila, R.: Electrical Resistivity Tomography (ERT) Applied to the assessment of Karst Carbonate Aquifers structure: Case Study from Zaghouan-Bent Saidan (NE Tunisia), EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10251, https://doi.org/10.5194/egusphere-egu22-10251, 2022.