r.avaflow 2.2 Background

Potentially catastrophic mass flows occur in many mountain areas all over the world. They may occur as avalanches of snow, rock, debris, and/or ice, debris flows, lahars, or pyroclastic flows, all of which may also represent components of complex cascading mass flows involving more than one type of process. An adequate management of the risk related to these phenomena requires a detailed and reliable analysis of the mechanisms involved in such processes. Even though much work has been done on this subject, and a number of physically-based models with a varying degree of complexity did exist, some problems still remained unsolved. Until that point, no successful attempts had been made to build readily available open source applications of the more complex models, something that would be essential to make the models available to a broader group of users in universities and public services. Multi-phase flow, flow over arbitrary topography, the role of viscous pore fluid, and particle and/or fluid entrainment had not yet been accounted for in an appropriate way in the existing simulation tools.

r.avaflow is an initiative launched for the purpose of designing, evaluating, and promoting a comprehensive and innovative simulation model for the dynamics of various types of geomorphic mass flows.

avaflow is the acronym for a international research project carried out in the period from July 2014 to December 2017. It was supported by the German Research Foundation (DFG), which was the lead agency, and the Austrian Research Fund (FWF) within the D-A-CH programme (between Austria, Germany and Switzerland).

The project has focused on the development of a novel simulation tool (r.avaflow - Mergili et al., 2017) for the motion of debris flows, snow avalanches, rock avalanches, and process chains consisting of more than one component. This tool represents a comprehensive GIS-based open source computational framework which, in contrast to most other related tools, offers a two-phase-flow model and considers entrainment and deposition of material along the path. These features facilitate the simulation of complex mass flows as well as of process chains and interactions.r.avaflow is freely available, its code is publicly accessible (open source), and is available as a command line interface or graphical user interface (GUI). The major innovation, however, builds on the use of a so-called two-phase flow model (Pudasaini, 2012): this means that fluid and solid parts, such as water and debris, are explicitly modelled including their interactions and different material properties. This allows the simulation of the effects of a landslide impacting a reservoir, and the associated overtopping and, possibly, erosion of the dam.

More information about the avaflow project and the prototypes of the code can be obtained from the project website.

Emphasis was put on the back-calculation of well-documented complex landslide process chains already in the final stage of the avaflow project, seamlessly continuing thereafter. The aim of these exercises was to learn about the capabilities of r.avaflow to plausibly reproduce such events, but also about the pitfalls with regard to possible forward simulations. At least one of the resulting publications has quickly evolved to a highly cited paper, highlighting the importance of this topic.

• The 2012 Santa Cruz multi-lake outburst flood in the Cordillera Blanca, Peru (Mergili et al., 2018a). A landslide from an unstable moraine slope entered a lake and caused a chain of glacial lake outburst floods, involving a total of three lakes and the entrainment of a major amount of sediment. This case study served for evaluating the principal ability of r.avaflow to reproduce a complex process chain. The simulation results were plausible, and the spatial patterns of the impact area correspond reasonably well to those mapped in the field and from remotely sensed datasets. However, many uncertainties remain with regard to the temporal scale, as no reference data are available.
• The 1962 and 1970 Huascarán events in the Cordillera Blanca, Peru (Mergili et al., 2018b). These two rock/ice falls evolved into avalanches of ice, debris, and mud of similar characteristics, but different magnitudes, with travel distances of almost 20 km. In contrast to the smaller 1962 event, the larger 1970 event overtopped a more than 100 m high ridge and destroyed the town of Yungay on the other side. This case study was used to evaluate the transferability of parameter sets optimized for one event to another event of the same characteristics, but a different magnitude. Both events are reasonably well reproduced in terms of impact area, involved volumes, and travel times as long as the parameter sets used for the simulation are empirically optimized for each individual event. However, switching the parameter sets clearly reveals that predictions of possible future events remain a major challenge, whereby threshold effects (overtopping of the ridge) play a major role and are highly critical in terms of hazard analysis.
• The 2017 Piz Cengalo-Bondo landslide cascade (Switzerland) (Mergili et al., 2020), where an initial rock slide impacted a glacier and subsequently evolved into a viscous debris flow propagating down a valley all the way to the village of Bondo at a distance of 6.5 km from the release area of the rock slide. This event is well documented in terms of its spatio-temporal evolution, and is considered useful for evaluating the ability of r.avaflow to appropriately reproduce a large number of reference parameters at the same time. Plausible and empirically largely adequate results are obtained with physically reasonable parameter assumptions. However, it remains a major challenge to reproduce all observations in one simulation: in particular, there is a trade-off between the reproduction of the debris flow volume at the front of the rock avalanche deposit, and the volume arriving at the village of Bondo.

Further test cases are continuously analyzed in order to derive sets of guiding parameter values or ranges for specific process types and magnitudes. Such guiding values will help to parameterize forward simulations of possible future events.

Based on the findings of various case studies, the physical concepts used in the prototype of r.avaflow have been questioned, and various larger and smaller adaptations and extensions have been developed and implemented. The most important of them are:

• Multi-phase flow model (Pudasaini and Mergili, 2019). Some types of interactions involved in landslide cascades can be adequately simulated with three-phase models only, taking into account the various material properties and phase interactions. Consideration of three-phase flows, however, is completely new and has not been implemented in operational landslide simulation frameworks before. Therefore, r.avaflow has been extended to a three-phase model, including solid, fine-solid and fluid materials.
• Viscous flows. The prototype of the r.avaflow software came with the option to consider highly viscous flows. However, this was only possible in theory: the numerical treatment of highly viscous flows was inadequate, and simulations failed. r.avaflow 2 offers an improved strategy to deal with high viscosities, so that the simulation of viscous earth flows, lava flows, and even the formation of lava domes becomes possible, opening up a wide field of applications.
• Block sliding. r.avaflow is a mass flow simulation tool, but for certain types of prosses chains it is useful to imitate more compact movements, i.e. block sliding. r.avaflow 2 includes this possibility, controlled through the option to constrain the passive earth pressure coefficients.
• Drag and virtual mass. Enhanced concepts for controlling the phase interactions have been developed. Their implementation have not only increased the flexibility, but also enhanced the numerical stability of simulations with r.avaflow 2.

Starting already in 2018, the potential of r.avaflow for educational purposes was evaluated and partly employed. Simulation results were included in the movie The Andes give, the Andes take and in the contribution on Yungay of the website Discover the Andes. Complementary scripts are provided facilitating the display and 3D animation of simulation results with external software.