NH3.12

Precipitation is one of the main causes for the initiation or reactivation of deep seated slow moving landslides. Preferential flow paths can have multiple origins, they can be due to changes in soil water content such as desiccation, due to mechanical movement or due to biological activity. The overarching characteristic is that they strongly alter the hydraulic properties of the landslide material. This results in a complex hydrological behaviour of deep-seated slow moving landslides. Research has shown that for instance the porosity of the soil, the fissure distribution and fissure connectivity are very important to predict the behaviour of the hydrological response of precipitation within a landslide body. However, due to large heterogeneity of landslide lithology and spatial and temporal variation of a landslide, it is hard to model water levels in landslides. Cracks and fissures inside the landslide are the cause of preferential flow paths, which can work as infiltration networks to the groundwater, but also as drainage networks lowering the (perched) groundwater levels.

In the last decades, both methodological progress has been made and several case studies have been published. However, most are still somewhat anecdotic examples and a more overarching conceptualisation has not been made yet. In this overview I want to highlight the progress as well as obstacles and challenges ahead of us when assessing and quantifying the impact of preferential flow paths on the mechanisms of a slow moving deep-seated landslide and to improve our understanding and modelling of complex landslides.

How to cite: Bogaard, T.: Conceptualising the effect of preferential flow on slow-moving landslides: from experiments to concepts and models., EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-15777, https://doi.org/10.5194/egusphere-egu21-15777, 2021.

The Hooskanaden landslide is a large (~600 m wide × 1,300 m long), deep (~30 – 45 m) slide located in southwestern Oregon, United States. Since 1958, it has had five moderate/major movements that catastrophically damaged the intersecting U.S. Highway 101, along with persistent slow wet‐season movements and a long‐term accelerating trend due to coastal erosion. Multiple remote sensing approaches, borehole measurements, and hydrological observations were integrated to interpret the motion behaviors of the slide. Pixel offset tracking of both Sentinel‐1 and Sentinel‐2 images was carried out to reconstruct the 3D displacement field of the 2019 major event, and the results agree well with field measurements. A 12‐year displacement history of the landslide from 2007 to 2019 was retrieved by incorporating offsets from LiDAR DEM gradients and InSAR (Interferometric Synthetic Aperture Radar) processing of ALOS and Sentinel‐1 images. Comparisons with daily/hourly ground precipitation reveal that the motion dynamics are predominantly controlled by intensity and temporal pattern of rainfall. A new empirical threefold rainfall threshold was therefore proposed to forecast the dates for the moderate/major movements. This threshold relies upon antecedent water‐year and previous 3‐day and daily precipitation, and was able to represent observed movement periods well. Adaptation of our threshold methodology could prove useful for other large, deep landslides for which temporal forecasting has long been generally intractable. The averaged characteristic hydraulic conductivity and diffusivity were estimated as 6.6 × 10⁻⁶ m/s and 6.6 × 10⁻⁴ m²/s, respectively, based on the time lags between rainfall pulses and slide accelerations. Hydrologic modeling using these parameters helps to explain the ability of the new rainfall threshold.

How to cite: Xu, Y., Lu, Z., and Kim, J.: Dynamics and physics-based rainfall thresholds for a deep-seated landslide, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-3692, https://doi.org/10.5194/egusphere-egu21-3692, 2021.

Many mountainous areas of Campania, southern Italy, are characterized by steep slopes covered with shallow deposits of loose pyroclastic materials, usually in unsaturated conditions, mainly constituted by layers of volcanic ash and pumice lapilli. The total cover thickness is quite variable, between 1.5 m and 2.5 m in the steepest part of the slopes while it reaches several meters at the foot, and it lays upon fractured limestone bedrock. Such pyroclastic materials usually exhibit extremely high porosity (even up to 75%) and saturated hydraulic conductivity (in the order of 10^-4 m/s). The equilibrium of the soil cover is ensured, up to inclination angles of 50°, by the contribution of soil suction to shear strength. Wetting of the soil cover during rainfall infiltration can cause a reduction of suction and, therefore, of the effective shear strength. This action sometimes leads to the triggering of shallow landslides, which often develop in the form of fast and destructive flows.

To capture the main effects of precipitations on the equilibrium of these slopes, hydrological monitoring activities have been carried out at the slope of Cervinara, located around 40 km northeast of Naples, where a destructive flowslide occurred in December 1999. An automatic hydro-meteorological station was installed at the elevation of 585m a.s.l., immediately near the scarp of the major landslide occurred in 1999. The meteorological equipment includes a rain gauge, a thermo-hygrometer, a thermocouple for soil temperature, an anemometer, a pyranometer, and a barometric sensor. The hydrological equipment consists of six tensiometers (located at depths between -0.2 m and -3.0 m below the ground surface) and six metallic time domain reflectometry probes (buried at depths between -0.3 m and -2.0 m) for the measurements of soil suction and water content, respectively. Furthermore, the water level in two streams located at the foot of the slope has been first manually monitored every month, and then, since March 2019, one of the two stream sections was instrumented with a probe, measuring water pressure, temperature, and electrical conductivity with hourly resolution.

The measurements allowed quantifying the major hydrological processes draining the soil cover after rainwater infiltration (i.e. evapotranspiration, overland and sub-surface runoff, leakage through the soil-bedrock interface), eventually assessing the water balance of the slope for three hydrological years (2017-2018, 2018-2019, 2019-2020).  The field monitoring data allowed the identification of the complex hydrological processes involving the unsaturated pyroclastic soil and the shallow groundwater system developing in the limestone bedrock, which control the conditions that potentially predispose the slope to landslide triggering. Specifically, late autumn has been identified as the potentially most critical period, when drainage through the soil-bedrock interface is not yet effective, owing to the still dry conditions at the base of the soil cover, but the slope already receives large amounts of precipitation.

How to cite: Greco, R., Comegna, L., Damiano, E., Marino, P., and Olivares, L.: Water balance based on field monitoring for the assessment of landslide predisposing conditions in a slope covered with pyroclastic deposits over fractured limestone bedrock, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-12447, https://doi.org/10.5194/egusphere-egu21-12447, 2021.

In this work we have tackled as a ‘cold case’ the prolonged rainfall and flood event occurred in the Piedmont region (Northern Italy) in November 1994, when several hundreds of shallow landslides occurred. The aim is to investigate the key trigger factors of the landslides and to put some focus on the possibility to obtain calibrated parameters thanks to the use of a regional geotechnical database.

This research has been motivated by the effort to close the methodological and conceptual gap between the use of low-detail approaches, proposed to explore wide investigation domains and that of complex ones, applied to single hillslope scale, typically relying on finite elements solutions.

To achieve the above-mentioned goals, a simple model was preferred (i.e. that of Rosso, Rulli, Vannucchi, 2006), since it allowed a better check on the sensitivity of soil parameter values to the instability condition, under the assumption that these were the main sources of uncertainty.

With reference to the 1994 event, a database of 238 observed landslide has been used, for which well-documented geometries and geotechnical parameters are available.

To address the specific aim of cohesion and permeability validation, the safety factor expression from Limit Equilibrium Analysis has been targeted to assume the value 1 for all the considered slopes subjected to the actual (measured) rainfall.

The comparison between locally calibrated cohesion and permeability and the reference ones found in the database shows some differences; in particular, in several cases, safety factors quite lower than 1 have been derived, compared to those obtained using the published parameter values. The overall uncertainty resulting from this gap has been analysed for a limited (5%) number of carefully examined landslides and it will lay the foundations for subsequent, more geometrically accurate, investigations.

How to cite: Evangelista, G., Barbero, M., Butera, I., Castelli, M., Claps, P., and Tamea, S.: Parameters calibration in rainfall induced landslides in the Langhe area (1994), EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-10999, https://doi.org/10.5194/egusphere-egu21-10999, 2021.

This study was conducted using data collected from 3 catchments in North-Western region of Rwanda; Kivu, upper Nyabarongo and Mukungwa. We used two parsimonious  models, a transfer function noise time series model and a linear reservoir conceptual model, to simulate groundwater levels using rainfall and potential evapotranspiration as model inputs. The transfer function noise model was identified as the model with great explanatory predictive power to simulate groundwater levels as compared to the linear reservoir model. Hereafter, the modelled groundwater levels were used together with precipitation to explain the landslide occurrence in the studied catchments. These variables were categorized into landslide predisposing conditions which include the standardized groundwater level on the landslide day h_t and prior to landslide triggering event h_t-1 and landslide triggering conditions which include the rainfall event, event intensity and duration.  Receiver operating characteristics curve and area under the curve metrics were used to test the discriminatory power of each landslide explanatory variable. The maximum true skill statistics and the minimum radial distance were used to highlight the most informative hydrological and meteorological threshold levels above which landslide are high likely to occur in each catchment. We will discuss our results of incorporation of groundwater information in the landslide predictions and compare these results with landslide prediction capacity which solely use of precipitation thresholds.Here we focus on at the same time on the practicalities of data availability for day-to-day landslide hazard management, both in terms of missed and false alarms

How to cite: Uwihirwe, J., Hrachowitz, M., and Bogaard, T.: Integration of multiple observed and model-derived hydrological variables in landslide initiation threshold models in Rwanda, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-14577, https://doi.org/10.5194/egusphere-egu21-14577, 2021.

The city of Rio de Janeiro is situated within a coastal region with steep slopes, intense seasonal rainfall, and vulnerable populations located on marginal slopes. Landslides are a seasonal challenge within the city and proximate regions and increasing real-time awareness of the hazard and exposure is paramount to saving lives and mitigating damage. A local alerting system has been developed for the city that leverages a global landslide hazard assessment for situational awareness (LHASA) framework, developed by NASA, with local rainfall thresholds and landslide susceptibility information. The LHASA-Rio system uses a decision tree approach to first identify extreme rainfall based on a series of rainfall thresholds established by Geo-Rio (the City’s agency responsible for landslide hazards) for 1 hour, 1 day or 1 hour and 4 day thresholds. This is then coupled with information on landslide susceptibility also developed by the Geo-Rio team. The LHASA-Rio system has been running operationally since 2017 within the city to provide real-time, high resolution estimates of areas within the city at higher hazard at 15-minute intervals consistent with the rainfall gauge network distributed throughout the city. Results of the LHASA-Rio system indicate excellent performance for several case studies where extreme rainfall triggered landslides within the city over areas identified as high hazard zones by LHASA-Rio. The model has recently been updated to accommodate additional rainfall thresholds to differentiate moderate to very high and critical intensities. The modeling effort is also incorporating information on landslide exposure by connecting the hazard estimates to city-wide data on population, road networks and other infrastructure. The goal of this system is ultimately to provide key tools to emergency response teams, civil protection and other hazard monitoring organizations within Rio’s City Government in real-time and provide  actionable information for key communities, city management and planning. Future work of this system is the application of a regional precipitation forecast to improve the lead time.

This work has been done in partnership through an agreement established between NASA and the City of Rio de Janeiro in 2015 that was recently extended in 2020. This agreement seeks to support innovative efforts to better understand, anticipate, and monitor hazards and environmental issues, including heavy rainfall and landslides, urban flooding, air quality and water quality in and around the city. This collaboration leverages the unique attributes of NASA's satellite data and modeling frameworks and Rio de Janeiro's management and monitoring capabilities to improve awareness of how the city of Rio may be impacted by hazards and affected by climate change. If the success of this technology is demonstrated, other cities in the world with physiographic and socioeconomic characteristics similar to Rio de Janeiro may benefit by implementing, or strengthening, their own Early Warning Systems for landslides triggered by heavy rains using LHASA's open source algorithms and the experience gathered by the use of LHASA-Rio. This presentation highlights the achievements and advancements of the LHASA-Rio system and discusses lessons learned regarding the applications of the landslide modeling systems to advance decision-relevant science at the city level.

How to cite: Kirschbaum, D., Mandarino, F., Fonseca, R., D'Orsi, R., Emberson, R., Stanley, T., and Khan, S.: Landslide Hazard and Exposure Modeling for Situational Awareness and Response in Rio de Janeiro, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-8914, https://doi.org/10.5194/egusphere-egu21-8914, 2021.

Road influence on landslides is an often-used variable in landslide susceptibility models. However, in many studies, there is limited detail explaining the processes and interactions between roads and landslides; instead a constant road buffer or standard function is used. Here, we present a spatial statistical analysis of landslide proximity to roads across a range of geographic settings and landslide inventory types. We examine the proximity of landslide centroids to roads at regional to national scales using twelve landslide inventories; with a variety of inventory types (6 triggered event, 6 multi-temporal), mapping methods (2 field based, 6 remote sensing, and 4 a combination of the two), and country of origin (6 high and 6 low human development index). Each inventory contains between ≈270 to 81,000 landslides (n_Landslides) and covers areas of ≈80 km² to 385,000 km².

We have developed a pyQGIS tool that calculates the distance between each landslide centroid and road vectors within the same drainage basin; this make sure no distances are calculated between landslides and roads that are on opposite sides of ridges and therefore do not influence each other. For each landslide inventory, we calculate the distance to the closest road for each landslide. We then compare this distribution that of a set of randomly generated points (number of random points is calculated for each landslide inventory using the equation n_Landslides*100) to roads, to test whether landslide occurrence is influenced by road presence.

For ten of the twelve inventories, the results show no strong preference of landslides to occur closer to roads than the random points; the exceptions being landslide inventories that we believe have a bias towards roads due to the mapping remit (e.g. highway agencies). For some of the ten inventories showing no robust relationship with roads, we believe this is related to the location of the roads on the slope (e.g. at the toe, mid-slope or on the ridge), but it is not readily explainable in others. Based on our results, we suggest that a more nuanced use of road proximity within landslide susceptibility models should be adopted, and further research to understand the interactions between landslides and proximity to roads at the regional to national scale.

How to cite: Heijenk, R. A., Malamud, B. D., Taylor, F. E., and Wood, J. L.: Comparing landslide proximity to roads in national and event landslide inventories, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-12288, https://doi.org/10.5194/egusphere-egu21-12288, 2021.

Landslides are among the phenomena that can cause casualties among the population and provoke considerable damage to infrastructures, and public and private assets. Precipitation must be mentioned as the main natural driver of these phenomena.  Over the years, landslide early warning systems (LEWS), aiming at reducing the exposure of the population, have been developed based on rainfall thresholds derived mostly by empirical methods, i.e. on finding a statistical link between the characteristics of precipitation (e.g. intensity and duration) and the occurrence of landslide movements. Commonly, landslide early warning thresholds are derived in the form of a power-law between rainfall intensity (or event rainfall) and duration.

One of the limitations of this approach is that using a predetermined form of the law may reduce its performances. In this work we investigate the advantages of removing such constraint by using artificial neural networks, which impose very low restrictions on the functional form of the threshold. We investigate this issue with reference to Sicily, Italy, where several landslide events have been documented in the past decades. In particular, we use rainfall data from almost 300 rain gauges from different monitoring network as the Sicilian hydrological observatory (Osservatorio delle Acque), the SIAS (Sicilian Agro-meteorological Information Service), the Department of Civil Protection (DPC) and almost 250 landslide events from FraneItalia (Calvello and Pecoraro, 2018). We then apply the CTRL-T code (Melillo et al., 2018) to automatically reconstruct rainfall events, identify the most probably rainfall condition that leads to slope failures and derive the traditional power-law threshold.  Then, the pattern recognition skills of artificial neural networks (ANN) are exploited to search the possible empirical relationship between rainfall characteristics and slope failure. Several options for the ANN structure are investigated. Finally, we show a comparison between the results of the two approaches, based on Receiver Operating Characteristic (ROC) analysis. Results show some potential of the ANN-based approach in improving landslide forecasting, although some limitations may exist due to possible quality issues of landslide and rainfall data.

References

Calvello, M., & Pecoraro, G. (2018). FraneItalia: a catalog of recent Italian landslides. Geoenvironmental Disasters, 5(1), 1-16.

Melillo, M., Brunetti, M. T., Peruccacci, S., Gariano, S. L., Roccati, A., & Guzzetti, F. (2018). A tool for the automatic calculation of rainfall thresholds for landslide occurrence. Environmental Modelling & Software, 105, 230-243.

How to cite: Distefano, P., Peres, D. J., Scandura, P., and Cancelliere, A.: Derivation of landslide triggering thresholds in Sicily through artificial neural networks, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-2674, https://doi.org/10.5194/egusphere-egu21-2674, 2021.

In snow-covered regions, a large number of landslides are induced by infiltration of snowmelt water. Although it is very important to early find signs of increase in landslide activity such as cracks or bulges for preventing or mitigating snowmelt-induced landslide disasters, thick snow cover often makes it difficult to find them. In such cases, frequent patrols of slopes prone to landslides during periods with high risk can be effective. In Japan, snowmelt advisories are issued by the Japan Meteorological Agency while snowmelt-induced disasters (e.g., flood and landslides) are predicted based on meteorological conditions. Although it seems that snowmelt advisories can be used for judging whether patrols are required, it has been reported that snowmelt advisories are not issued for some days with high risk of snowmelt-induced landslides (Irasawa et al, 2011). Focused exclusively on landslides, Nakaya et al (2008) and Touhei et al (2016) proposed methods for capturing 70% of landslides by setting a critical level using reservoir inflow and river water level and flow rate as hydrological indices. These methods, however, are difficult to apply for areas affected by human impacts including irrigation and water intake and drainage of power stations. In this study, based on the antecedent precipitation index, reported as a hydrological index showing a good correlation with slow-moving landslide velocity (e.g., Enokida et al, 2002), we propose an extensively applicable method for setting snowmelt-induced landslides warning periods. The target areas are three 5-km meshes in Joetsu and Myoko Cities, Niigata Prefecture, central Japan, where heavy snowfall in winter and the underlying Tertiary sedimentary rocks cause many snowmelt-induced landslides every year. We used for analyses 285 landslide cases that occurred from December to May in 1979 to 2020 reported in data set on landslides compiled by the Niigata Prefectural government. We used  (meltwater and/or rainwater), which is the total amount of water reaching the ground surface, instead of precipitation, for calculating the antecedent precipitation index. The amount of snowmelt was estimated based on the heat balance method using the Japan Meteorological Agency observation data alone (Matsunaga, 2019) for the center of each mesh with an average elevation within the mesh.  and the antecedent  index with a various half-life were calculated hourly. Using the standard score, calculated by normalizing the antecedent  index, we determined the critical standard score capturing 70% of the target landslides in each mesh and the half-life minimizing the landslides warning periods (i.e., periods during which the standard score exceeds the critical standard score). These procedures resulted in the average landslides warning periods per year of 36 to 50 days with 36 to 318 hours of the half-life for all meshes. On the other hand, snowmelt advisories were issued for 30 days per year in average from 2013 to 2020, capturing only 36% of the target landslides. Thus, the method proposed in this study shows more than 30% higher landslide capture ratio and therefore is better than snowmelt advisories for setting snowmelt-induced landslides warning periods.

How to cite: Matsunaga, T. and Katsura, S.: A study on an extensively applicable method for determining snowmelt-induced landslides warning periods based on a hydrological index, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-5454, https://doi.org/10.5194/egusphere-egu21-5454, 2021.

SIGMA (Sistema Integrato Gestione Monitoraggio Allerta – integrated system for management, monitoring and alerting) is a landslide forecasting model at regional scale which is operational in Emilia Romagna (Italy) for more than 20 years. It was conceived to be operated with a sparse rain gauge network with coarse (daily) temporal resolution and to account for both shallow landslides (typically triggered by short and intense rainstorms) and deep seated landslides (typically triggered by long and less intense rainfalls). SIGMA model is based on the statistical distribution of cumulative rainfall values (calculated over varying time windows), and rainfall thresholds are defined as the multiples of standard deviation of the same, to identify anomalous rainfalls with the potential of triggering landslides.

In this study, SIGMA model is applied for the first time in a geographical location outside of Italy, i.e. Kalimpong town in India. The SIGMA algorithm is customized using the historical rainfall and landslide data of Kalimpong from 2010 to 2015 and has been validated using the data from 2016 to 2017. The model was validated by building a confusion matrix and calculating statistical skill scores, which were compared with those of the state-of-the-art intensity-duration rainfall thresholds derived for the region.

Results of the comparison clearly show that SIGMA performs much better than the other models in forecasting landslides: all instances of the validation confusion matrix are improved, and all skill scores are higher than I-D thresholds, with an efficiency of 92% and a likelihood ratio of 11.28. We explain this outcome mainly with technical characteristics of the site: when only daily rainfall measurements from a spare gauge network are available, SIGMA outperforms other approaches based on peak measurements, like intensity – duration thresholds, which cannot be captured adequately by daily measurements. SIGMA model thus showed a good potential to be used as a part of the local Landslide Early Warning System (LEWS).

How to cite: Segoni, S., Abraham, M. T., Satyam, N., Rosi, A., and Pradhan, B.: Application of SIGMA model for landslide forecasting in Darjeeling Himalayas, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-669, https://doi.org/10.5194/egusphere-egu21-669, 2021.

Rainfall induced landslide is one of the most common hazards worldwide and it is responsible every year of huge losses, both economic and social. 
Because of the high impact of this kind of natural hazard, the forecasting of the meteorological condition associated with the initiation of landslide has become paramount in the recent years and several papers addressing this issue have been published.
When working over large areas, the definition of rainfall thresholds is the most used approach, since it requires few data that can be easily retrieved: landslide triggering date and location and rainfall recording associated to landslide events.
The intensity-duration threshold is the most used approach and it showed over the time its potential to be implemented in an operative landslide early warning system (LEWS), but literature papers showed that this approach is affected by a main drawback, i.e., the high number of false positives (events that are not capable of triggering landslides are classified as landslide triggering events).
To overcome this problem several authors tried to combine these thresholds with other parameters and recently one of the most promising approach is the use of the antecedent soil moisture condition, but this parameter is note very easily available for large areas and it is difficult to retrieve it in real time, so as it can be used in a LEWS.
In our work we used antecedent rainfall to simulate the progressive saturation of the soil and then the soil moisture condition associated with the initiation of landslides.
In a given area the total rainfall recorded by each rain gauge over a defined period of time prior the landslide is considered and used to define a parameter named MeAR (Mean Antecedent Rainfall), which represent the mean rainfall of the area over a given time interval, as recorded by all the active rain gauges.
The MeAR parameter has been coupled with classical I-D thresholds to define 3D thresholds, where the conditions associated with the initiation of a landslide are defined by a portion of a 3D space, instead of a portion of a 2D plane. This approach has been tested in Emilia-Romagna region (Italy) and it resulted the possibility of reducing false positives from 30% up to 80% on different areas.

How to cite: Rosi, A., Monni, A., Gallucci, A., and Casagli, N.: Definition of 3D rainfall thresholds for operative LEWS, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-2271, https://doi.org/10.5194/egusphere-egu21-2271, 2021.

Landslide Early Warning Systems (LEWS) can provide enough time to take necessary precautions before the occurrence of landslides and can reduce the risk associated with it. Deriving empirical rainfall thresholds is the conventional approach in developing regional scale LEWS, but the major drawback of this approach is the relatively high number of false alarms. In this study, a prototype method for LEWS is proposed by combining rainfall thresholds and field monitoring data from MicroElectroMechanical Systems (MEMS) units that integrate a tilt sensor, a soil moisture meter and a real-time wireless transmitter. The study was conducted in the Kalimpong district of West Bengal, India. Tilt sensors were installed at different locations on unstable slopes of Kalimpong since July 2017 and the observations from July 2017 to August 2020 were used to enhance the performance of the existing rainfall thresholds.

During this period, both rainfall thresholds and tilt meters, when used separately, systematically overestimated landslide hazard, producing high false alarm rates. However, it was found that using a decisional algorithm that combines both approaches can reduce the false alarms and improve the overall efficiency of the system from 84 % (based on rainfall thresholds) to 92 % (combined method). The prototype LEWS is found to be promising to be developed as an operational LEWS capable to issue alerts with a lead time of 24 h. 

The method is simple and can be easy exported to other sites with historical rainfall and landslide data and a network of slope monitoring sensors. Cost of installation of a large number of sensors is a major concern for developing countries like India, hence a cost-effective approach is used in this study: the use of MEMS sensors along with empirical rainfall thresholds is thus a simple and economical approach for the prediction of landslide events.

How to cite: Bulzinetti, M. A., Abraham, M. T., Satyam, N., Pradhan, B., and Segoni, S.: Combining rainfall thresholds and field monitoring data for development of LEWS., EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-2072, https://doi.org/10.5194/egusphere-egu21-2072, 2021.

Landslide early warning systems (LEWS) can be classified in either territorial or local systems (Piciullo et al., 2018). Systems addressing single landslides, at slope scale, can be named local LEWS (Lo-LEWS), systems operating over wide areas, at regional scale, can be referred to as territorial systems (Te-LEWS). Te-LEWS deal with the occurrence of several landslides within wide warning zones at municipal/regional/national scale. Nowadays, there are around 30 Te-LEWS operational worldwide (Piciullo et al., 2018; Guzzetti et al., 2020). The performance evaluation of such systems is often overlooked, and a standardized procedure is still missing. Often the performance evaluation is based on 2 by 2 contingency tables computed for the joint frequency distribution of landslides and alerts, both considered as dichotomous variables. This approach can lead to an imprecise assessment of the warning model, because it cannot differentiate among different levels of warning and the variable number of landslides that may occur in a time interval.

To overcome this issue Calvello and Piciullo (2016) proposed an original method for the performance analysis of a warning model, named EDuMaP, acronym of the method’s three main phases: Event analysis, Duration Matrix computation, Performance assessment. The method is centered around the computation of a n by m duration matrix that quantifies the time associated with the occurrence (and non-occurrence) of a given landslide event in relation to the different warning levels adopted by a Te-LEWS. Different performance criteria and indicators can be applied to evaluate the computed duration matrix.

Since 2016, the EDuMaP method has been applied to evaluate the performance of several Te-LEWS operational worldwide: Rio de Janeiro, Brazil (Calvello and Piciullo, 2016); Norway, Vestlandet (Piciullo et al., 2017a); Piemonte region, Italy (Piciullo et al., 2020), Amalfi coast, Italy (Piciullo et al., 2017b). These systems have different structures and warning models with either fixed or variable warning zones. In all cases, the EDuMaP method has proved to be flexible enough to successfully perform the evaluation of the warning models, highlighting critical and positive aspects of such systems, as well as proving that simpler evaluation methods do not allow a detailed assessment of the seriousness of the errors and of the correctness of the predictions of Te-LEWS (Piciullo et al., 2020).

Calvello M, Piciullo L (2016) Assessing the performance of regional landslide early warning models: the EDuMaP method. Nat Hazards Earth Syst Sc 16:103–122. https://doi.org/10.5194/nhess-16-103-2016

Guzzetti et al (2020) Geographical landslide early warning systems. Earth Sci Rev 200:102973. https://doi.org/10.1016/j.earsc irev.2019.102973

Piciullo et al (2018) Territorial early warning systems for rainfall-induced landslides. Earth Sci Rev 179:228–247. https://doi.org/10.1016/j.earscirev.2018.02.013

Piciullo et al (2017a) Adaptation of the EDuMaP method for the performance evaluation of the alerts issued on variable warning zones. Nat Hazards Earth Sys Sc 17:817–831. https://doi.org/10.5194/nhess-17-817-2017

Piciullo et al (2017b) Definition and performance of a threshold-based regional early warning model for rainfall-induced landslides. Landslides 14:995–1008. https://doi.org/10.1007/s10346-016-0750-2

Piciullo et al (2020). Standards for the performance assessment of territorial landslide early warning systems. Landslides 17:2533–2546. https://doi.org/10.1007/s10346-020-01486-4

How to cite: Piciullo, L. and Calvello, M.: Five years of EDuMaP for the performance analysis of territorial landslide early warning systems, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-5618, https://doi.org/10.5194/egusphere-egu21-5618, 2021.