News

Details

Created: 02 December 2024

Following the thourough analysis done by François Noël and co-authors (published in 2021), pointing out an issue with tree impact detections in Rockyfor3D, we re-analyzed the C-coded version of the program distributed via ecorisQ. We discovered that, due to the integer values used for the indices in the hash table for selecting the set of trees that are potentially impacted by the simulated block at a given location, only one tree per cell could be accounted for. These indices were calculated solely based on the row and column number of the cell. The tree selected for the actual impact calculation was chosen randomly, which could result in selecting the tree with the smallest diameter at breast height (DBH) in the given cell by chance. In the newly released version 6.0 of Rockyfor3D, each tree has a unique index in the hash table and therefore, all trees in the input data are taken into account in the 3D trajectory simulations.

Furthermore, as Noël et al. (2021) correctly noted, tree impacts were only calculated if the center of the block intersected with the tree stem perimeter. In the current version, this detection has been adjusted to align with the original Matlab version of Rockyfor3D (developed in 2007). Mea culpa: someone else translated my original Matlab code to C almost 20 years ago, and in my test simulations, this error did not appear sufficiently. The program now properly calculates the interaction between the tree stem disc and the planimetrically projected perimeter of the falling block. As a result, simulations with forest using v6.0 show less conservative results, and the difference between simulations with and without forest is more pronounced than in previous versions.

Since this revision made us dive deep in to the Rockyfor3D code again, we continued making further corrections. Most importantly, we corrected errors that, in exceptional cases, led to extremely high energies, which were especially noticeable on the energy maps (though less so in the statistics, except for extreme cases like the 98% or 99% quantiles). This issue was caused by trajectories that exited a cell precisely through a corner, and under certain conditions, were transferred instantly to an opposite corner. This created a quasi-infinite velocity loop. With this error resolved, the energy values and trajectories now look more realistic (without inexplicable local energy increases or stains with very high energy values in the raster maps) and therefore more plausible. This also allowed us to eliminate the previously used maximum velocity limit of 70 m/s.

Additionally, we reintroduced the friction and scaling function in the energy loss (rebound) calculation during ground impacts, as originally proposed by Pfeiffer and Bowen (1989), which was included in one of the first versions of Rockyfor, but with updated parameter settings as described by Noël et al. (2021). Another factor influencing energy loss is the correction of the restituted kinetic energy of the impacting block based on the total apparent kinematic coefficient of restitution (CORv), proposed by Noël et al. (2023). This factor adjusts the energy loss during ground impacts that result in extreme deflections from the fall direction. The greater the deviation, the more energy is lost. However, we did not use the relationships proposed by Noël et al. (2023) as they reduced the runout and lateral deviations of trajectories in Rockyfor3D too much, compared to well-documented rockfall events used for calibration. In the new version, we limit the total energy restitution only for total deviation angles (TOTdev) larger than 55°, following the equation: CORv = -0.01 * TOTdev + 1.3. Moreover, the maximum calculated value of the rotation velocity is limited in the new version based on the block's shape and mass, following the relationship published by Caviezel et al. (2021).

All the new integrated aspects described above result in shorter, less extreme trajectories, and generally less kinetic energy compared to previous versions of Rockyfor3D. We believe this represents a significant improvement in the software. Just a reminder: soil type is a sensitive parameter in Rockyfor3D and can be difficult to estimate for different block sizes. For example, a soil type 3 might represent a slope where 0.5 m³ blocks are falling but may not be accurate for a 125 m³ block, as larger blocks interact with deeper soil layers and penetrate more in soft soils. This leads to different compression of the substrate and greater changes in local slope morphology. Therefore, if trajectories seem unrealistic if compared to observed silent witnesses, soil type should be the first parameter to be re-assessed (assuming surface roughness has been mapped accurately).

We also adjusted the algorithm for calculating the effect of the impact height on a tree stem. The base value of our algorithm, previously 1.62, has been lowered to 1.20. This means that for impact heights above 1.3 m, the energy loss value is now less than 1. Consequently, energy dissipation during a tree impact above 1.3 m is now lower than before. The effect of tree impact height on total energy loss is now calculated using the following equation: dE_Height = 1.2*((1.0/(1.0+exp(18.04*(Z/Htree)+0.02*DBH-2.35)))-(1/(1.0+exp(15.69+0.02*DBH)))).

Next, we added the option “Fix” or “Cliff area weighted” to define the number of trajectories simulated from each start cell. “Fix” simulates the defined number of trajectories from all defined start (or release) cells. “Cliff area weighted” calculates the number of simulations per cell based on the cliff area, since steep start cells cover much more rock face area and therefore produce rockfall more frequently than flatter start cells. See the  pdf manual of Rockyfor3D v6.0 (2.66 MB) for further explanations on the underlying calculations.

Finally, we corrected an error where, in rare cases, a block could pass through a net. If the net (or part of it) formed a perfect diagonal line in the raster, the block could pass through a corner point from one no-net cell to another, without touching the neighboring net cells. The situation is semi-graphically explained below, where N represents a net cell and 0 a no-net cell:

000N0000000
0000N000000
00000N00000
000000N0000
0000000N000

In the new Rockyfor3D, we resolved this issue.

Voilà, that was all for now! We hope you appreciate the revised version as much as we do!

Luuk Dorren and Christoph Schaller

Details

Created: 22 July 2024

Recurrent precipitation in June and July 2024 led to flooding in many regions of Switzerland. Heavy rainfall and storms in the second half of June caused rivers and streams to rise rapidly, particularly in Valais and Ticino. Floods and debris flows caused extensive damage to settlements and infrastructure. To date, 10 people have lost their lives and several others are still missing. Dealing with and coming to terms with these events will take a long time. More information here.

In the image shown above, the village of Fontana (Maggia valley) - ©swisstopo 2024 

 Before & After event comparison

Village of Fontana (Maggia valley)

Village of Sorte (Grisons)

Val d'Anniviers (Valais)

For natural hazard and risk analyses, the following ecorisQ tools can be used:

BankforNET - Bank erosion assessment tool

This tool aims to support experts in the quantification of bank erosion hazards and the effects of biological measures.

SlideforMAP - Shallow landslide simulation at regional scale

SlideforMAP is a probabilistic model to assess shallow landslide probability on a regional scale with an explicit focus on vegetation scenarios.

Details

Created: 22 April 2024

On Sunday 14 April 2024, a Bergsturz with more than 1 million m³ occurred on the north face of Piz Scerscen (Switzerland). Although this event occurred spontaneously after several days with high temperatures, nobody got hurt. Since a lot of water has been observed in the failure zone, melting of snow and ice very probably played an important role. The runout distance was 5.4 km, which coressponds to a energy line angle of 15° (following the travel path) and 16° (following a straight line) - typical values for such events. An impressive movie showing the release, transit and deposit area can be found here (commented in the beautiful Rumantsch - the fourth Swiss national language).

With our new rockfall runout simulation tool named RockavELA, we try to predict the propagation of such events.