Background

The frequency of severe weather conditions is increasing due to climate changes caused by human-induced greenhouse gases released into the atmosphere. Extreme weather events that could trigger adverse effects, such as landslides, and the failure to mitigate and adapt to climate change are the risks with the highest probability of occurrence and the most significant global impact. The construction industry is a significant contributor, accounting for nearly 40% of the greenhouse gas emissions [27, 47]. Consequently, the durability of the very same built environment (buildings and infrastructure) contributing to these emissions is under global threat from extreme weather events like heavy rainfall and extreme temperatures leading to landslides [13, 16, 20]. On the one hand, there is an unprecedented need for a green transition in the construction sector. On the other hand, managing the risks to the built environment and its users associated with extreme weather events to an acceptable level is long overdue. Considering climate change, the global average annual loss (AAL)Footnote 1 for infrastructures currently lies between $732–$845 billion CDRI [8]. This statistic highlights the pressing need for the construction industry to embrace more eco-friendly practices and for infrastructure to be built with a focus on resilience to climate-related risks.

In India alone, primarily in the Himalayan region, economic losses exceeding INR 1000 crores per year and over 1000 deaths per year have been recorded [10]. This underscores the need for improved tools to mitigate the risks associated with landslides. The landslide risk reduction encompasses both structural and non-structural measures. Effective sense-making and decision-making processes will depend on the reliability of the measures. In this paper, we summarize some latest developments made on understanding and mitigation of water-triggered landslides (Fig. 1).

Fig. 1
figure 1

A depiction of the impact of severe weather on urban areas globally [53]

Full size image

Recent Developments

Given the complexity and breadth of the topic, the authors wish to maintain a specific focus on water-triggered landslides on urbanized slopes. Mitigation of the potential adverse effects of such landslides often involves implementation of challenging and expensive measures. This section addresses some key developments that have been made in terms of the accurate assessment of landslide susceptibility in changing climate, landslide failure mechanism, effectiveness of structural countermeasures and novel techniques to improve landslide risk awareness.

Landslide Susceptibility: Probabilistic Framework for the Assessment of Overall Impact of Climate Change on Landslide Susceptibility for More Accurate Early Warning Assessment

Many studies on the impact of climate change on landslide susceptibility focus on extreme conditions [32], Shou and Yang [41]). The extreme conditions represent intense rainfall events with long return intervals, such as 50 or 100 years.

Oguz [36] developed a novel probabilistic framework proposed to obtain the overall climate change impact on landslide susceptibility by accounting for the probability of occurrence of rainfall events within a given duration. The framework has a potential to provide a more objective assessment of the climate change impacts without bias due to extreme rainfall events. The framework integrates landslide and climate modeling chains within a probabilistic framework to estimate landslide susceptibility as the probability of landslide initiation:

$$P_{f} \left( L \right) = P\left( {F_{S} \le 1.0{|}L} \right) = \smallint P\left( {F_{S} \le 1.0{|}L,I} \right) f\left( {I{|}L} \right) dI$$
(1)

where \(P\left( {F_{S} \le 1.0{|}L} \right)\), i.e., \(P_{f} \left( L \right)\), is the probability of landslide initiation for a given rainfall duration, L, \(F_{S}\) is the factor of safety, \(P\left( {F_{S} \le 1.0{|}L,I} \right)\), i.e., \(P_{f} \left( {L,I} \right)\), is the probability of landslide initiation conditioned on \(L\) and \(I\) obtained from the landslide modeling chain. \(f\left( {I{|}L} \right)\) is the probability distribution function of rainfall event with an intensity of \(I\) conditioned on \(L\), and parametrized by location, \(\kappa_{G}\), and scale, \(\beta_{G} ,\) parameters of a Gumbel distribution, which models the marginal distribution of rainfall intensity for a given duration. The distribution and the parameters are obtained from Intensity–Duration–Frequency (IDF) curves that model the distribution of rainfall intensity and frequency. The effects of climate change are included in the IDF curves by examining the evolution of the IDF parameters based on the climate projections in the climate modeling chain.

The probabilistic framework proposed by Oguz [36] is not only accounting for the uncertainties in the rainfall conditions but also for the uncertainties in the climate projections. This has been implemented by quantifying uncertainties in the entire climate modeling chain, as shown in Fig. 2. The probabilistic approach was also extended to the landslide modeling chain by quantifying uncertainties in landslide model parameters (e.g., strength parameters). Given probabilistic approaches in both chains, this finally allowed for seamless integration of the modeling chains in Eq. (1) to obtain probabilistic estimates of landslide susceptibility.

Fig. 2
figure 2

Coupled climate and landslide modeling chains (modified after [37])

Full size image

The climate modeling chain was based on the emission scenario RCP 8.5 (regional climate projection 8.5), which referred to the concentration of carbon that delivers global warming at an average of 8.5 watts per square meter across the planet [36]. The RCP 8.5 pathway delivers a temperature increase of about 4.3˚C by 2100, relative to preindustrial temperatures (e.g., Oguz [36]). RCP 8.5 was selected in [36] because it was considered to be the most impactful scenario. In simple words, RCP 8.5 referred to a scenario with no climate policy, high population growth and low technological development to combat climate change.

The climate-dependent IDF curves are derived by a formulation developed by Benestad et al. [7] as follows:

$$x_{L} = \alpha \cdot \mu \cdot \left( {L/24} \right)^{\zeta } \cdot \ln \left( {f_{w} \cdot \tau } \right)$$
(2)

where \(x_{L}\) is the rainfall amount in mm, \(L\) is duration in hours, \(\tau\) is return interval in years, \(\alpha\) is correction factor accounting for the divergence from the exponential distribution, \(\zeta\) is correction factor for different rainfall durations, \(\mu\) is the wet-day mean precipitation in mm/day and \(f_{w}\) is the wet-day frequency. Wet day in this study is defined as a day with more then 1 mm of rain.

The values of \(\mu\) and \(f_{w}\) are calibrated based on the available historical information to obtain IDF curves for current climate and estimated based on climate projection for future climate.

A case study in central Norway was selected to demonstrate the probabilistic framework [36, 37]. IDF curves were estimated for three time periods: present (1981–2010), future (2021–2050) and far future (2071–2100) using the mean values of wet-day mean precipitation and wet-day frequency applied to the Eq. (2) [36]. For the future and far future periods, parameters of the IDF model were estimated with the delta method, by adding the projected changes in the mean values of the parameters \(\mu\) and \(f_{w}\) to the mean values of the parameters for the present time period [36]. More details on the analysis of climate projections and their downscaling can be found in [36, 37].

The landslide modeling is implemented based on transient rainfall infiltration and grid-based regional slope stability (TRIGRS) [6]. The effects of climate change on landslide susceptibility are analyzed probabilistically by simulating rainfall events with varying durations and intensities based on the IDF curves [36]. For a given the rainfall event, the uncertainties in the TRIGRS model parameters are propagated to the model predictions with the Monte Carlo method. The TRIGRS model predictions were presented in terms of probability of landslide initiation [36]. Different probability threshold levels were examined to express the extent of landslide susceptible zones, specified as \(P_{f} \left( L \right) > P_{{f,{\text{limit}}}}\) [36].

The overall climate change impact on landslide susceptibility was assessed for 6-, 12- and 24-h rainfall events by providing the \(P_{f} \left( L \right)\) maps at present, near future and far future climate conditions [36]. Oguz [36] showed that the projections of landslide susceptibility maps at three time periods indicate a considerable increase in \(P_{f} \left( L \right)\) for all durations of rainfall events and up to 9.9%, 8.4% and 3.7% for 6-, 12- and 24-h rainfall events, respectively, until the end of the twenty-first century.

The results in [36] showed somewhat smaller changes in the values of overall probability of landslide initiation that includes also higher likelihood and lower intensity rainfall events because the high probabilities of extreme rainfall events (e.g., 50-, 100-year return intervals) were weighted down by their relatively low occurrence probabilities.

Analyses of daily and sub-daily rainfall events by [36] revealed that shorter duration rainfall events result in higher overall values of \(P_{f} \left( L \right)\). From 3, one can observe that the probabilities of landslide initiation are higher for 6-h rainfall compared to the 12- and 24-h rainfall at all time periods [36]. This was due to the higher likelihood of shorter rainfall events with a sufficient intensity to initiate a landslide, rather than more extreme but less likely 12- and 24-h rainfall events [36].

The results (Fig. 3) show an increase in the probability of landslide initiation under climate change. Statistics of the probabilities of landslide initiations is shown in Table 1, the mean values of the \(P_{f} \left( L \right)\) f [36]. It was interesting to observe that 6-h rainfall led to higher values mean \(P_{f} \left( L \right)\) compared to the 12- and 24-h rainfall at all time periods [36].

Fig. 3
figure 3

\(P_{f} \left( L \right)\) maps for a, d, g 6-h, b, e, h 12-h and c, f, i 24-h rainfall events at a, b, c present, d, e, f near future and g, h, i far future climate conditions from (Oguz et al. 2022)

Full size image
Table 1 Mean values of \(P_{f} \left( L \right)\) for the entire study area for 6-, 12- and 24-h rainfall duration at present, near future and far future climate conditions from Oguz et al. [36]
Full size table

Several probability threshold levels were used by [36] to determine the extent of landslide susceptible zones as shown in Table 2. Table 2 shows that there is a substantial increase in the extent of landslide susceptible zones due to climate change for all rainfall durations. Additionally, shorter duration rainfall events led to a larger relative increase in landslide susceptible extents for each threshold level.

Table 2 Extent of landslide susceptible zones, specified as \(P_{f} \left( L \right) > P_{{f,{\text{limit}}}}\) for the entire study area for 6-, 12- and 24-h rainfall duration at present, near future and far future climate conditions from (Oguz et al. 2022)
Full size table

Landslide Countermeasures: Debris Flow Screens for the Detrainment of Bed Material and Impact Reductions

The idea of using debris flow screens was conceived in Japan with the intention to reduce the energy in debris flows and, thus, contribute to mitigating damage in downstream areas. The main purpose of this screen is to separate water or the fluid from the moving debris. As a result, the pore-water pressure developed inside the debris flow, particularly in the shearing zone, would dissipate. In return, the debris’ solid particles regain their contact friction and thereby increase the shearing resistance of the moving debris, Ref. Figure 4.

Fig. 4
figure 4

The screen: a before a debris flow event giving its schematic representation, and b after a debris-ow event showing the material detrainment  [56]

Full size image

A number of studies have investigated the debris flow screens including a field trial in the Kamikami-Horisawa Valley of Japan have also been reported throughout literature (e.g., [17, 33], and [22, 26], [23], [35, 45] and [59]). Due to their simple construction and cost-effectiveness, screens have also been implemented and used in other countries, including Mainland China (e.g., Lien [7], [31], [54] and [34]) and the Philippines (e.g., ICHARM [20, 21] and [5]). In the Philippians, screens have been used to protect mountain roads by installing them in narrow sections of streams where recurrent debris flows occur.

Recently Yifru et al. [57] and Yifru [56] have made an attempt to advance the use of debris flow screen. The focuses on how to channelize the moving debris safely by using underpasses instead stopping it debris flow completely. Yifru [56] conducted series of flume tests, Ref. Figure 5. In these tests, two screen lengths (0:25 m and 0:50 m long with 2 mm openings) were evaluated along with their respective topographical contributions represented by solid plates. These two lengths were selected in order to evaluate the flow reduction capacity of screen in a continuous debris flow after the 1-m-long screen was seen to accumulate almost the entire volume. The screens were also assessed in combination with guide wall and underpass. The guide wall and underpass (referred here after as underpass) can be considered as a safe passage for debris flows under an elevated road near at the foot of a mountain.

Fig. 5
figure 5

Flume model with screens and underpass placement and instrumentation. All linear dimensions are in meters. [56]

Full size image

The test set was categorized into two: Tests conducted with underpass and tests conducted without underpass. Variations were made on volume, V (30, 40 and 50 L) and solids concentrations by volume, Cs (50, 55, 60%). The average grain size of the tested material is 2.2 mm. The flow impact force was measured by the circular pillar that was placed at the end of the run-out channel (as shown in Fig. 6).

Fig. 6
figure 6

Plan view photos showing details of placement of a the screen and b the sidewall and underpass. A road and a car representation over the underpass is also given at the top right. [56]

Full size image

This study investigated the potential and performance of screens in reducing the impact force of the debris flow by detrainment of debris on the screens. The presence or absence of the underpass has no or minimal effect on the measured force as the pillar was situated before the underpass. The magnitude of the recorded force was seen to be affected by volume (V) and solid concentration (C) variation as well. The case where large V with high Cs results in the highest impact force in each scenario and vice versa. This indicates that both increase in V and decrease in water content, with more viscous type ow characteristic, may result in higher impact force. The measured run-up height, run-out distances were significantly reduced with increasing length of debris flow screen. On each case, this maximum impact force was observed to decrease between 20 and 50% progressively with the provision of plate, 0:25 m long screen and 0:5 m long screen. The total run-out distance was reduced between 30 and 60% accordingly. Readers are encouraged to refer Yifru [56] to obtain details of test results and analysis in its entirety. Based on exhaustive number of tests conducted on the screen countermeasures, Yifru [60] suggests that this mitigation measure can deter the mobility of debris flows. The study the optimal opening size of screen grid along is the mean grain size diameter. The work demonstrated that the screen performance can vary based on the debris flow composition. The detrainment potential decreased when the fines and water content of the debris flow increased. Although the resulting run-out distance decreased due to the provision of the screen, it is seen to slightly increase with increase of fines and water contents as well as debris flow volume.

Practical application of debris flow screens on steep terrains can be limited; therefore, the idea of upstream flexible barrier was launched and tested by Vicari et al. [51] as presented in the following section.

Use of Upstream Flexible Barrier for the Entrainment and Impact Reductions

Vicari et al. [51] advanced the understanding and applicability of flexible barriers as they examined the use of multiple flexible barriers in conjunction with an erodible bed. They focused on understanding the influence of an upstream flexible barrier on debris flow entrainment and its impact on a terminal barrier. The study conducted large-scale experiments using a 28-m-long flume, ref. Figure 7, to investigate the entrainment of a wet soil bed by a debris flow. The debris, initially in volumes of 2.5 and 6 m3, was stored in a container and could be triggered by a dam break. The debris then flowed along a 15-m-long, 2-m-wide channels inclined at 20 °. The last 6 m of the inclined channel was covered with a 120-mm-thick layer of wet soil. The channel ended in a horizontal 4.4-m-long run-out section, where a terminal flexible barrier was placed to stop the flow. This paper presents results from with and without any upstream barrier, used initial volumes of 6 m3. A 0.6-m-tall flexible barrier placed 4.3 m from the gate to study the effect of an upstream flexible barrier on entrainment reduction. The debris flow material used in the study was a sand–gravel–clay mixture [34], consisting of 36% gravel (2–20 mm), 61% sand (0.075–2 mm) and 3% fines (< 0.075 mm) by mass, with a solid concentration (Cs) of 70%. The initial density of the debris mixture was 2155 kg/m3.

Fig. 7
figure 7

28-m-long flume model in Hong Kong, aerial view (top) and cross section (lower) [51]

Full size image

The study reveals that a compact upstream flexible barrier, designed to retain only 20% of the initial debris volume, can significantly mitigate peak discharge by creating multiple smaller overflowing surges when placed before an erodible bed. The average erosion was recorded to be 110 mm for test T1 and 33 mm for T2. The interaction with the upstream barrier results in a 70% decrease in entrainment and a 94% reduction in the impact force on the terminal barrier, compared to test without an upstream barrier.

Engineers typically use the total peak impact force (F), calculated using the hydrodynamic equation, to design flexible barriers that can withstand debris flow impacts.

$$F = a \rho v^{2} h W$$
(3)

where α is a dynamic pressure coefficient, ρ is the mass density, v and h are the flow velocity and depth before impact, respectively, and W is the width of the flow before impact (2 m). The dynamic pressure coefficient ‘α’ can vary from 1.0 to 5.0 according to different sources, but [52] reported a value of 1.0 for debris flows impacting a flexible barrier in the field.

In test T2, the dynamic pressure coefficient ‘α’ was back-calculated as 2.2 for the upstream barrier and 1.2 and 0.2 for the terminal barrier in tests T1 and T2, respectively, ref Table 3. This suggests that the installation of a compact upstream flexible barrier can optimize the design of the terminal barrier, as the design dynamic pressure coefficient ‘α’ for the terminal barrier decreased from 1.2 to 0.2 when an additional upstream barrier is present.

Table 3 Test details and results
Full size table

The value of ‘α’ is higher for the upstream barrier (2.2) in test T2 compared to that of the terminal barrier (1.2) for the test with only the terminal barrier (test T1). This higher value at the upstream barrier could be due to the centrifugal component of the force from the curved layer flowing over the dead zone and drag from overflow, as well as the small barrier to flow depth ratio (1.5) which may have limited energy dissipation internally in the flow, resulting in higher force transferred to the upstream barrier.

However, the value of ‘α’ on the terminal barrier decreases significantly when the upstream flexible barrier is used. This reduction can be attributed to the impact process on the terminal flexible barrier, where a higher volume impacted the barrier with a higher flow velocity in test T1, leading to a run-up impact mechanism. In contrast, in test T2, the volume and velocity impacting the terminal barrier were much smaller due to the upstream barrier, resulting in a pileup mechanism.

Despite the upstream barrier’s effectiveness in reducing entrainment and the impact force on the terminal barrier, its deformation upon impact forms a curved ramp-like deposit that induces a centrifugal force on the barrier for subsequent flow. This configuration results in a back-calculated dynamic pressure coefficient of 2.2, significantly higher than the recommended 1.5 in design guidelines, indicating the need for further consideration of the maximum deformation and orientation of a flexible barrier relative to the channel.

Landslide Failure Mechanism: Stability of Reservoir Rim Slopes Subjected to Water Level Fluctuations

Rim slope failure is a major hurdle in the smooth functioning of the hydropower projects across globe. Many rim slope failures have been witnessed amidst monsoons and due to water level fluctuations in the reservoir during past several decades causing severe destruction. The tragedy of Mt. Vajont Dam, Italy, in 1963, where a massive upstream landslide along rim slope caused havoc in the nearby region claiming more than 2000 lives, cannot go unnoticed [37]. Rainfall and water level fluctuations causing internal seepage in the rim slopes have been regarded as the main triggering factors of reservoir-induced landslides [43, 44]. Reservoir rim slope failures in the vicinity of hydropower dams are increasing challenge in India, but, as of now a very limited attention is given. Figure 8 presents a glimpse of few of the slope instabilities observed along the reservoir rims in lower Himalayas, India. It becomes necessary to investigate the failure mechanism of such rim slope instabilities.

Fig. 8
figure 8

Slope instabilities observed along the reservoir rims in lower Himalayas, India

Full size image

Establishment of internal seepage flow within rim slopes is a predominant phenomenon as a consequence of rainfall and water level fluctuations. The infiltrated water within rim slopes during rainfall and water level fluctuations seeps toward the drainage basin and may induce instability in the slope. Internal seepage flow can be steady or unsteady in nature. Apart from the seepage flow, several other factors contribute to rim slopes instabilities such as the slope geometry and slope material. In context to Indian Himalayas, the slopes forming the reservoir rims are mostly debris or riverbed material. The present study aims at exploring the failure mechanism of rim slopes with riverbed material as the slope material under the steady-state seepage flow. A physical slope model test (PSMT) apparatus was developed indigenously as shown in Fig. 9a.

Fig. 9
figure 9

Physical slope model test (PSMT): a test apparatus, b prepared 30° model slope with 12 piezometer nodes

Full size image

The setup has the facility to test model slopes of height up to 28 cm, width 20 cm and length of up to 50 cm. Model tests with slope inclination between 30° and 60° can be performed. Pore-water pressure at 12 different locations inside the slope mass can be measured using piezometers provided in the setup (Fig. 9b). Apparatus has the provision to measure deformation of slope profile with the help of vertical and horizontal scales attached to the sidewalls of the test setup and with a deformation gauge. Physical slope model test (PSMT) under steady-state condition was performed on a model slope made with riverbed material. The properties of the riverbed material are tabulated in Table 4.

Table 4 Properties of the riverbed material
Full size table

The PSMT was carried out by first preparing the 30° model slope of height—25 cm as shown in Fig. 9b. The model slope geometry was recorded with the help of deformation gauge, and horizontal and vertical scales. The slope was then subjected to water total heads at either end of the slopes to induce internal seepage. The water total head was kept as 20 cm and 5 cm at the crest end and toe end of the slope to induce seepage toward the toe region of the slope. Soon, the model slope achieved steady-state condition. The steady-state condition was ensured by monitoring the constant discharge at the toe end of the slope.

A retrogressive multiple rotational slide type of slope failure occurred during PSMT. It was observed that the failure did not occur in one go; instead, it was progressive with time (Fig. 10). The slope failure started near the toe, where the computed exit hydraulic gradient was found maximum (Fig. 10a).

Fig. 10
figure 10

Images of the slope failure in PSMT at different time instants: a Failure 1 at time T–16 min, b Failure 2 at time T–20 min, c Failure 3 at time T–26 min, d Failure 4 at time T–31 min, e Failure 5 at time T–58 min

Full size image

The initial instability was mainly due to the seepage forces acting toward toe of the slope. Post the initial instability, the rest of the failures were induced mainly due to the geometric instability. It was observed that the failure surface retrogressed backward with larger block slides causing a retrogressive slide. The retrogressive phenomenon observed during PSMT was numerically simulated and validated using limit equilibrium approach [9]. Stagewise 2D schematic diagrams of the PSMT were drawn to showcase the retrogressive failure phenomenon as shown in Fig. 11.

Fig. 11
figure 11

Slope geometry with potential failure surface depicting retrogression. a Failure 1, b Failure 2, c Failure 3, d Failure 4, e Failure 5, f All failure surfaces with final slope geometry

Full size image

The failure surfaces retrogressed backward by a distance of more than 30 cm until the slope inclination reduced to nearly 18°. Each successive failure surface was found larger than before where the diameter of the failure surfaces appears to be the multiple of the diameter of initial failure surface caused due to geometric instability (Fig. 11f). Also, with the increase in diameter of the failure surfaces, the time period between two successive collapses increased (Fig. 11). Further, the spread near toe was observed to be more than 22 cm nearly equal to the height of the slope. The retrogressive failure phenomenon was also observed by Jia et al. [25] while subjecting a model slope of sandy silt to large water level fluctuations. In the study by Jia et al. [25], the slope collapse began after the drawdown was initiated in the form of large blocks retrogressing backward.

Thus, the rim slopes in the vicinity of hydropower dams are susceptible to failure even at a gentle slope inclination. Retrogressive failure can easily occur in cohesionless soils along the rim slopes under internal seepage flow. The cause of the seepage flow is mostly rainfall and water level fluctuations. The rim slope stability should be given due consideration and constant monitoring of the rim slopes is necessary. Also, in usual practice by engineers, while performing numerical simulations for a field problem, the focus is on the formation of failure surface and deformations observed. The limitation with such studies is that the movement of the slope surface over a period of time cannot be investigated, and the possibility of retrogression cannot be studied. The past and future of such rim slope instabilities are mostly unknown or are disregarded. It is, therefore, suggested that emphasis should be laid upon the time-dependent analysis as well. Further, new approaches should be developed to predict the time-dependent behavior of rim slopes under the influence of rainfall and reservoir water level fluctuations for the smooth functioning of hydropower projects.

Landslide Risk Awareness: Use of Virtual Reality as a Tool for Landslide Risk Awareness and Preparedness

In recent years, advancements in virtual reality (VR) technology, alongside the accessibility and affordability of VR hardware and software tools, have resulted in its widespread adoption and diverse applications across various domains ranging from entertainment to military, mental health, disaster risk management and daily life. In the context of disaster risk management, a comprehensive examination conducted by Zhu and Li [60] revealed that immersive technologies, such as VR, have been employed across various phases: from pre-emergency preparedness, encompassing hazard identification and prevention, to the emergency response stage, including evacuation and rescue operations, and finally to the postdisaster phase, involving rebuilding, recovery and restoration efforts. VR is widely applied in construction safety and hazard recognition training [24, 31, 39, 55], earthquake disaster evacuation training [18, 28, 29] and flood hazard awareness training and education ([14, 40, 42]).

VR offers a significant advantage in disaster risk management by enabling the creation of disaster scenes which their construction in real life is dangerous and expensive. However, even in the most studied disasters, there is little research that shows how and what elements should be included in the VR disaster scenes [15]. Alene et al. [1] undertook a co-design approach involving experts from diverse disciplines such as geotechnical engineering, hydraulic engineering, psychology and VR technology. Their objective was to gather insights into the crucial components of VR disaster scenes and how these elements should be integrated to effectively represent quick clay landslide disaster scenarios. Based on the insights, the authors developed a VR tool, named QuickAware, which aimed to enhance awareness and understanding of quick clay landslide hazards through immersive VR experiences. The VR tool the underwent testing with a small group of participants to assess the realism of its elements for an immersive experience and to evaluate whether the key objects within the virtual environment are compelling enough to capture users’ attention, ensuring they focus on the intended issues addressed by the tool.

Alene et al. [4], [1] further emphasized that especially for flow-type landslide, the dynamics of the flow, often represented by numerical model simulation results, should be an integral part of the disaster scene. However, unlike flow landslide models, the game engines used to create virtual environments in VR applications lack inherent compatibility with georeferenced systems, posing challenges for integrating simulation results into VR environments. Alene et al. [3] developed a novel framework to incorporate the results from a debris flow numerical model [12] into VR. This framework was further upgraded to include numerical simulation results from various geophysical flows such as snow avalanche, landslides and flood [2]. The framework is based on Unity game engine,Footnote 2 one of the widely used game engine in developing immersive virtual environments [11].

In our proposed framework for flow landslide visualization, the initial step involves selecting an appropriate numerical simulation model. Key input parameters include defining the computational domain (terrain), specifying initial and boundary conditions, determining the time step for output data storage and accounting for additional model-specific parameters. These parameters must be prepared in advance before commencing the numerical simulation.

The framework is specifically tailored to mesh-based numerical models operating within the Eulerian frame of reference. This choice aligns with the prevailing practice, as most existing flow landslide models utilize the Eulerian reference frame [48]. Consequently, the expected output results manifest as point clouds, where flow variables are provided at each mesh grid point or cell. However, not all numerical models directly yield the variables necessary for VR visualization. In such cases, postprocessing tasks—such as filtering and interpolation—may be required to export the relevant variable values. These exported variables can be saved in various formats, including structured ASCII raster or basic XYZ coordinates in CSV files.

The immediate input data for our framework comprise flow depth, flow velocity and terrain, all represented in a structured ASCII raster format (Fig. 12a). Importing these data into Unity is straightforward using the preprocessing module (Fig. 12b). However, if results are provided as generic XYZ coordinates in CSV format, additional steps involving Python scripts are necessary to convert the data into a structured ESRI ASCII raster format (as detailed in [2]). During preprocessing, the terrain data are transformed into Unity’s native terrain format, while flow depth values are integrated with terrain elevation to create flow surface elevation coordinates—a representation of the flow landslide. These flow surface elevation coordinates are organized into 2D arrays, capturing both spatial location and corresponding velocity information. To optimize runtime performance in Unity, we store these 2D arrays as binary files rather than the original ASCII format.

Fig. 12
figure 12

Flowchart of the VR visualization (Alene et al. [1])

Full size image

The simulation module (Fig. 12d) serves as the core component responsible for processing binary files and generating 3D models representing flow data. Its workflow begins by extracting vertices from flow surface elevation coordinates. These vertices are then transformed into 3D models through mesh creation. To render the scene, we utilize the Universal Render Pipeline (Unity [49]), which handles the visualization of these 3D models and other elements. Within the framework, two shaders play distinct roles:

Terrain Shader: This Unity-built tool allows users to apply predefined textures to the terrain, mimicking its real-world appearance. It enhances the visual fidelity of the terrain within the VR environment.

Flow Surface Shaders: Specifically designed for 3D objects derived from flow surface elevations, these shaders create the illusion of flowing fluid. They contribute to the realistic representation of flow landslides in the VR scene.

To simulate flow dynamics, velocity vectors obtained from numerical simulations are integrated into shaders. These vectors control the flow’s direction and speed, adding fluid-like behavior to the visual representation. Temporal evolution is captured using 3D models as frames in an animation, displayed sequentially on the screen according to the simulation’s time step. Within the Unity project development environment, a rendered VR scene—including the flow and terrain—can be viewed on the computer screen.

The interaction module plays a crucial role in connecting the virtual environment with XR devices like Meta Quest5. These devices allow users to immerse themselves in the virtual world and interact with it. By incorporating the XR Interaction Toolkit plugin (Unity [50]) into the scene, the system supports various user interactions, including movement, manipulation and selection via different input devices and controllers. Custom scripts further facilitate specific interactions, such as accessing flow variables, taking field measurements and controlling simulation playback.

Real debris flow events that occurred in Hunnedalen, Norway, in 2016 and flow slide in Gjerdrum in 2020 were studied and simulated using the proposed methodology. The details can be found in Alene et al. [1] and [2]. For Hunnedalen debris flow, a depth-averaged numerical model was used to back-calculate the event, whereas for the Gjerdrum landslide, a full 3D resolution numerical model was employed. In both cases, the simulation outcomes are imported into the Unity game engine to build a digital model of the debris flow event, see Fig. 13. The study focused on visualizing the time and space progression of the flow landslide in a virtual reality setting to assess the extent of landslides and damages to the built environment.

Fig. 13
figure 13

VR visualization of postfailure events of a a real flow landslide scarp including destroyed houses in Gjerdrum landslides in 2020 and b a destroyed road in Hunnedalen

Full size image

Alene et al. [2] investigate the use of dynamic 3D visualization in VR to immerse users in flow landslide hazards. Unlike traditional 2D screens, this approach provides a better sense of scale and depth perception. The intuitive understanding of hazard power and speed through VR may assist politicians and municipal authorities in making informed decisions. The suggested a framework that integrates VR into disaster scenes, offering realistic simulations for first responders during training. For early warning systems, the proposed VR tool will complement field visits by testing camera locations to detect flow landslides promptly. In education, the VR tool will enhance student engagement and understanding of complex numerical methods related to flow-type landslides. Overall, the proposed approach aims to enhance hazard communication and management by leveraging VR technology.

Overall conclusions

Selected recent developments in structural and non-structural solutions, early warning and preparedness to manage water-triggered landslides are discussed. This integration considers multiple rainfall events and their likelihood of occurrence within a probabilistic framework. The model framework is tested in a real case study, revealing that the proposed integration framework enhances the detection of critical zones with a higher probability of landslide initiation under climate change conditions. Consequently, this approach enables the development of more effective strategies to address upcoming risks and achieve better preparedness. The paper summarized the latest advancements in landslide risk communication through the creation of immersive and participatory visualizations. These visualizations leverage new digital media, particularly augmented/virtual/mixed realities, to facilitate a deeper understanding of landslides. The advancement and applicability of usefulness of two new structural solutions, debris flow screens and upstream flexible flow barriers, for water-triggered landslides in steep terrains are discussed. Experimental results from 10- and 28-m-long flume tests were used to illustrate the effectiveness of these countermeasures in partially or fully reducing the mobility of water-triggered landslides. The flume test results indicate that upstream flexible barriers could reduce the entrainment of bed materials in the landslide mass by 70% and the impact force on the terminal barrier by 94%, compared to cases without an upstream barrier along run-out channels. The study also addressed the seepage-induced stability and failure mechanisms of reservoir rim slopes. Pumping in and out of water from the reservoir can endanger the safety of rim slopes. Regarding non-structural countermeasures, this paper presented a pioneering effort to integrate climate modeling into landslide susceptibility assessment.