<p>When planning for the implantation of transport infrastructures or buildings, it is necessary to identify the land zones that can be reached by rockfalls. These zones should then be avoided if possible, or stabilisation and risk mitigation measures must be considered. 3D preliminary rockfall simulations can be used to help finding the areas where inspections should be prioritised. Using orthophotos, a detailed shaded representation of the terrain and field work, geologists can then note the position of the deposited blocks and sources from past events, among other things. Collecting this information can however be complex, and the blocks can sometimes be mistaken for glacial deposits.</p><p>To increase the accuracy of this inspection task, the land can be analysed using a 3D detailed terrain model with artificial colors based on its aspect orientation and slope steepness and artificial shadows based on the ambient occlusion and eye dome lighting methods. Scars left by past rockfall events are then highlighted and some trajectories can be reconstituted. This method can help isolating identified rockfall deposited boulders from erratic blocks and help finding where is the source from. It can also draw attention to the location where a block has settled by showing parts of its trajectory. A relative aging can also be attributed based on the sharpness of the scar edges, with older events appearing smoother or partly erased. This can help estimating the activity of the site when no other information is known.</p><p>We applied this method to the Mel de la Niva site in Switzerland while analysing the two main rockfalls from the 2015 event. The 3D model used was created from SfM photogrammetry using pictures acquired on the field by manually flying a DJI Phantom 4 drone over the terrain. The method allowed to identify 1 rockfall that followed the main 2015 event and 7 rockfalls that preceded it, which is quite interesting. Indeed, if activity is observed on a site, inspection of the source cliff should be done to try to identify if a larger event is about to occur.</p><p>These identified rockfalls trajectories were validated using a time series of available orthophotos from SWISSIMAGE. Two paths were present before the oldest photo from 1983. Three appear on the 1999 photo. They then happened in between the previous photo from 1995 and the 1999 one. One happened in between the 1999 and 2005 photos. One happened in between the 2010 and 2013 photos and one in between the 2016 and 2017 photos.</p><p>The 8 identified trajectories combined with the 2 from 2015 also have an interesting shape. They tend to not directly follow the steepest path of the terrain. This behavior seems to be frequent, especially when the blocks are disk-shaped, and it has also been observed and partly quantified from the rockfall experiment we presented here last year (2019). Data from the Mel de la Niva site has been added to our rockfall database and it will used for the calibration and further developments of our rockfall simulation model.</p>

<p>Working with 3D point clouds offers many benefits for reducing the subjectivity of rockfall simulations at a local scale. Indeed, many “dynamic” rockfall rebound models are strongly affected by the topography and the perceived surface roughness, which can be objectively represented with detailed terrain models. This reduces the need for complex time intensive back analyses and associated sensitive adjustments of parameters used for subjectively adjusting the simulations to the desired runout distances.</p><p>Predictable and reproductible simulations from a constrained set of parameters while still managing to reproduce observed runouts on a wide range of sites could be time saving for practitioners and their clients, ultimately improving quality at lower costs to the society. This could speed up the process for practitioners to deliver concise reports easier to interpret and quality-check by a wider range of employees on the client side.</p><p>However, working with 3D point clouds can have a steep learning curve and quickly becomes impractical at a larger scale for regional analysis, partially obscuring some of the previously mention advantages. To explore potential ways to circumvent these issues, a prototype of an algorithm that runs the stnParabel rockfall simulation freeware in batch was quickly implemented in 2020. It was developed to expand such dynamic simulation capabilities to larger regions and up to potentially national-covering capabilities.</p><p>Slight modifications were done on the impact detection algorithm to also work with high resolution gridded terrain models (DTMs) with a focus at not sacrificing the benefits of working on 3D point clouds. The sources biases due to the stretched grided cells underrepresenting the steep cliffs are worked around by randomly distributing the sources based on the 3D stretched surface occupied by the cells.</p><p>Preliminary results were produced regionally over 6000 km<sup>2</sup>, involving 115 000 000 simulated rockfalls with 10 m<sup>3</sup> blocks of dimensions 3.8x3.2x1.8 m. The simulations were performed on the Norwegian national 1 m DTM from airborne LiDAR, up sampled to 50 cm cells for future proofing the approach. They were produced at a rate of about 15 000 000 simulated 3D trajectories per hour when ran on a small Ultrabook laptop with fast SSD.</p><p>The preliminary results from the dynamic rockfall model were then combined with databases of observed deposited blocks from previous rockfall events to act as a calibration guide for FlowR model. This simpler model based on gridded topographic-hydrologic spreading and sliding block approaches can be adjusted to produce a wide range of desired runouts envelopes from numerous processes, like rockfalls. The simpler simulations on 10 m DTM were used as a candidate for the revision of the national rockfall susceptibility mapping methodology.</p><p>The prototype approach to run detailed dynamic rockfall simulations regionally would require validations. Such potentially useful approach with objective dynamic simulations for hazard mapping as well as for the design of mitigation measures could then be shared through publications and be implemented in the distributed rockfall simulation freeware stnParabel. </p>

Rockfall simulations are often performed at various levels of detail depending on the required safety margins of rockfall-hazard-related assessments. As a pseudo benchmark, the simulation results from different models can be put side-by-side and compared with reconstructed rockfall trajectories, and mapped deposited block fragments from real events. This allows for assessing the objectivity, predictability, and sensitivity of the models. For this exercise, mapped data of past events from the Mel de la Niva site are used in this paper for a qualitative comparison with simulation results obtained from early calibration stages of the Flow-R 2.0.9, Rockyfor3D 5.2.15 and RAMMS::ROCKFALL 1.6.70 software. The large block fragments, reaching hundreds of megajoules during their fall, greatly exceed the rockfall energies of the empirical databases used for the development of most rockfall models. The comparison for this challenging site shows that the models could be improved and that combining the use of software programs with different behaviors could be a workaround in the interim. The findings also highlight the inconvenient importance of calibrating the simulations on a per-site basis from onsite observations. To complement this process, a back calculation tool is briefly described and provided in the supplementary material. This work also emphasizes the need to better understand rockfall dynamics to help improve rebound models.

No abstract

Process-based rockfall simulation models attempt to better emulate rockfall dynamics to different degrees. As no model is perfect, their development is often accompanied and validated by the valuable collection of rockfall databases covering a range of site geometries, rock masses, velocities, and related energies that the models are designed for. Additionally, such rockfall data can serve as a base for assessing the model’s sensitivity to different parameters, evaluating their predictability and helping calibrate the model’s parameters from back calculation and analyses. As the involved rock volumes/masses increase, the complexity of conducting field-test experiments to build up rockfall databases increases to a point where such experiments become impracticable. To the author’s knowledge, none have reconstructed rockfall data in 3D from real events involving block fragments of approximately 500 metric tons. A back analysis of the 2015 Mel de la Niva rockfall event is performed in this paper, contributing to a novel documentation in terms of kinetic energy values, bounce heights, velocities, and 3D lateral deviations of these rare events involving block fragments of approximately 200 m3. Rockfall simulations are then performed on a “per-impact” basis to illustrate how the reconstructed data from the site can be used to validate results from simulation models.