Debris Flow Modelling Using RAMMS Model in the Alpine Environment With Focus on the Model Parameters and Main Characteristics

Debris flows are among the natural hazards that can occur in mountainous areas and endanger people’s lives and cause large economic damage. Debris flow modelling is needed in multiple applications such as design of protection measures or preparation of debris flow risk maps. Many models are available that can be used for debris flow modelling. The Rapid Mass Movement Simulation (RAMMS) model with its debris flow module, (i.e. RAMMS-DF) is one of the most commonly used ones. This review provides a comprehensive overview of past debris flow modelling applications in an alpine environment with their main characteristics, including study location, debris flow magnitude, simulation resolution, and Voellmy-fluid friction model parameter ranges, (i.e. μ and ξ). A short overview of each study is provided. Based on the review conducted, it is clear that RAMMS parameter ranges are relatively wide. Furthermore, model calibration using debris-flow post-event survey field data is the essential step that should be done before applying the model. However, an overview of the parameters can help to limit the parameter ranges. Particularly when considering the similarity between relevant case studies conducted in similar environments. This is especially relevant should the model be applied for estimating debris-flow hazard for potential future events. This model has been used mostly in Europe, (i.e. Alpine region) for modelling small and extremely large debris flows.


INTRODUCTION
According to the updated Varnes classification, debris flows are defined as very to extremely rapid surging flows of saturated debris that occur in steep channels with significant entrainment of material and water (Hungr et al., 2014). Due to these characteristics debris flows can cause large economic damage and endanger human lives (Mikoš et al., 2004(Mikoš et al., , 2007. Especially endangered are the so-called debris flows and torrential fans, (i.e. alluvial fans). These are relatively flat parts of mountainous regions that are often quite heavily populated (Bezak et al., 2019). Reliable debris flow prediction is often not possible due to limited geological information or details about triggering mechanisms such as extreme rainfall event (Takahashi, 2014). Therefore, the so-called back analysis of past debris flow events can be used to design TABLE 1 | A review of debris flow (DF) and Glacial Lake Outburst Flood (GLOF) modelling using RAMMS software and its debris flow module. Studies are sorted by the publication year of the source, and then in alphabetical order. NA indicates that the information was not provided in the cited reference. Multiple parameters are shown when combinations of these parameters were used.

Source (authors, year)
Location ( Short description of the study Cesca and D'Agostino (2008) Fiames DFs, Dolomites, Italy (2006) Six cases: 15,000; 10, 600; 46,800; 11,000; 5,200; 2,100 20; 10; 10; 10; 5; 5 0.18; 0.2; 0.19; 0.37; 0.39; 0.45 500; 40; 15; 40; 100; 1,000 Comparison between RAMMS and FLO-2D. Cell size affected the shape of the inundated area markedly. The deposition area was overestimated, and deposition thickness underestimated, especially for the cell size 20 m. RAMMS had constantly excessive lateral dispersions. Hauser (2011) Arth DF; Goldau DF (2005), Switzerland 50.000-80.000; ∼190.000 2; 2.5 0.20 10 ξ < 50 m/s 2 was used to limit DF velocities. ξ was positively correlated with DF velocity and µ was negatively correlated. Hussin (2011) Barcelonnette, France (1996;2003) 100,000; 83,000-95,000. Both cases included entrainment where this is much larger than initial volume 5 0.06 500 Detailed sensitivity and probability analyses were conducted. The DEM accuracy greatly affected the topography of the area and the geometry of DF channel. This directly affected the DF behavior in terms of velocity and DF height. A rapid decrease of the channel slope caused a decrease in velocity and run-out distance. This led to an increase in the deposit height at the head of deposited DF. The run-out distance and the maximum DF height of the modeled DF was most sensitive to changes in ξ followed by µ. Scheuner et al.  1 | (Continued) A review of debris flow (DF) and Glacial Lake Outburst Flood (GLOF) modelling using RAMMS software and its debris flow module. Studies are sorted by the publication year of the source, and then in alphabetical order. NA indicates that the information was not provided in the cited reference. Multiple parameters are shown when combinations of these parameters were used.

Source (authors, year)
Location ( Creekbest fit for runout distance (15,000-50,000 m 3 , µ 0.03-0.16, ξ 100-700 m/s 2 ), and ii) the Festeticgrabenbest fit for deposition area (10,000-20,000 m 3 , µ 0.01-0.24 (and 0.03-0.32 outside the DF channel), ξ 100-1,400 m/s 2 ). Significant sensitivity was found to the variation in µ and DF volume, and lower sensitivity to variation in ξ. Fischer et al. (2016) Richleren DF (1987) Sensitivity analysis without entrainment showed that max. DF height is not overly sensitive to ξ, whereas it strongly depended on µ (inverse law for µ 0.08-0.14). The DF velocity increased with ξ increasing and µ decreasing. The sensitivity of DF velocity to µ was much higher for high values of ξ. The application of entrainment in the simulation led to a decrease in the best-fit value of ξ, which corresponds to an increase in DF velocity.
(Continued on following page) TABLE 1 | (Continued) A review of debris flow (DF) and Glacial Lake Outburst Flood (GLOF) modelling using RAMMS software and its debris flow module. Studies are sorted by the publication year of the source, and then in alphabetical order. NA indicates that the information was not provided in the cited reference. Multiple parameters are shown when combinations of these parameters were used.

Source (authors, year)
Location (  1 | (Continued) A review of debris flow (DF) and Glacial Lake Outburst Flood (GLOF) modelling using RAMMS software and its debris flow module. Studies are sorted by the publication year of the source, and then in alphabetical order. NA indicates that the information was not provided in the cited reference. Multiple parameters are shown when combinations of these parameters were used.

Source (authors, year)
Location (  1 | (Continued) A review of debris flow (DF) and Glacial Lake Outburst Flood (GLOF) modelling using RAMMS software and its debris flow module. Studies are sorted by the publication year of the source, and then in alphabetical order. NA indicates that the information was not provided in the cited reference. Multiple parameters are shown when combinations of these parameters were used.

Source (authors, year)
Location ( Back analysis of a devastating DF using geotechnical investigation. Runout modeling using block release option in RAMMS. Calibration of friction parameters using digital image processing to compare the shape of the actual DF and the simulated one. Bezak et al. (2020) Brezovški and Lukenjski graben, Slovenia (2018) Debris flood: 48,000, out of that 7,000-10,000 of coarse deposits 1 0.13 for Brezovški graben, 0.2 for Lukenjski graben 400 for Brezovški graben, 900 for Lukenjski graben Best fit parameters were determined using the inundation area. RAMMS was successfully applied for a debris flood modelling.
(Continued on following page) TABLE 1 | (Continued) A review of debris flow (DF) and Glacial Lake Outburst Flood (GLOF) modelling using RAMMS software and its debris flow module. Studies are sorted by the publication year of the source, and then in alphabetical order. NA indicates that the information was not provided in the cited reference. Multiple parameters are shown when combinations of these parameters were used.

RAMMS AND DEBRIS FLOW MODELLING
The RAMMS model uses depth-averaged shallow water equations for granular flow in the single-phase model for debris flow modelling (RAMMS, 2017). The model employs the Voellmy-fluid friction model that includes two parameters, (i.e. the dry-Coulomb type friction μ (Mu) and the viscous-turbulent friction ξ (Xi)). These two parameters are usually calibrated, although other parameters such as stop parameter or simulation resolution also have an effect on the modelling results (Bezak et al., 2019). However, some of these are limited by data availability. A detailed description of the model's theoretical background and key equations are provided in the user's manual. Table 1 provides a review of more than 30 past studies that used RAMMS software for debris flow modelling. It can be seen that RAMMS model has been frequently applied in Europe, (i.e. for the Alpine region) while applications in South America and Asia were also included in the review (Table 1). Furthermore, it can be also seen that RAMMS was used for modelling relatively small debris flows, (i.e. 1,000 m 3 or less) to extreme ones where their magnitude exceeds a couple of million m 3 ( Table 1). The simulation resolution was in most cases very high, especially considering large debris flow magnitudes with resolution ranging from less than 0.5 m to 20 or 30 m ( Table 1). In most cases, the resolution was between 2 and 5 m ( Table 1). Moreover, the Voellmy-fluid friction parameters covered wide ranges (Figure 1). Low values for the both parameters are prevailing, and only a few case studies used the parameters above the line connecting the end points: (μ 0, ξ 1,400 m/s 2 ) (μ 0.65, ξ 0 m/s 2 ). Nevertheless, they mostly stayed within the ranges indicated by Scheidl et al. (2013) as typical for debris flows (Table 1). More specifically, Dry-Coulomb type friction parameter μ (Mu) ranged from less than 0.001 to 0.7. Most often, the value of this parameter was around 0.1 or 0.2 ( Table 1). The Viscous-turbulent friction parameter ξ (Xi) ranged from 10 m/s 2 to 2,000 m/s 2 . Its value was most often between 200 and 500 m/s 2 ( Table 1). The debris flow magnitude slightly decreases and increases with increasing μ and ξ, respectively. Nevertheless, no significant correlation could be detected ( Table 1). As illustrated, the RAMMS model was used for a variety of different applications, including modelling of the glacial lake outburst flood (Table 1). Figure 2 shows a result of a typical application of the RAMMS model in an alpine environment.

CONCLUSION
No clear pattern can be observed in the reviewed studies regarding the frequency of the most suited friction parameters μ and ξ.
Evidently, the RAMMS model parameters clearly depend on local debris flow characteristics such as topography, rheological properties, and hydro-meteorological conditions. Therefore, as already suggested in the RAMMS manual (RAMMS, 2017), model calibration should be the optimal way to determine the FIGURE 1 | Range of the Voellmy-fluid friction parameters (μ and ξ) used in the analyzed studies shown in Table 1, excluding one study with very large ξ parameter, (i.e. Chung et al., 2018).
Frontiers in Earth Science | www.frontiersin.org January 2021 | Volume 8 | Article 605061 friction parameters that clearly have a significant impact on the modelling results (Table 1). Moreover, further research could focus on a better connection of the RAMMS model parameters with the physical features of an area or debris-flow material.

AUTHOR CONTRIBUTIONS
MM prepared the first version of Table 1. NB made an update and verification of the table. Both authors contributed to the writing and editing of the article and approved the submitted version.