[com4FlowPy] Use CAseStrange runout pattern

[saudeoud]

Hello all,

Trying to apply flow py on an ATES study case, I obtainend some strange outputs, the flow lines seem to follow geomtric rule, using 0° or 45° north preferential paths. The lateral spreading seem kind of blocked or altered.
Perhaps this has to do to some parametrization ?
I assume the DEM is of good quality
This doesn't seem normal, any ideas ?

Alpha Angle: 24.0
Exponent: 8.0
Flux Threshold: 0.003
Max Z_delta: 270.0

Zdelta :

Detail :

[fso42]

Hi @saudeoud,
thanks for your feedback. @PaulaSp3 or @ahuber-bfw will get back to you. (They are the experts for com4FlowPy)

[ahuber-bfw]

Hi @saudeoud,

Thx for the feedback :)

You could try setting the flux threshold ($R_{\text{stop}}$ in D'Amboise et al., 2022) to 0.0003 ($3\times10^{-4}$) instead of 0.003 ($3\times10^{-3}$) --> this is our recommended setting and should produce smoother results with less geometric effects in the runout.

Btw, which DEM resolution are you using - 10m?

[saudeoud]

Thanks @fso42 and @ahuber-bfw for your answers.

I use a 5m DEM (France IGN RGE ALTI).

I try with your suggestion.

[fso42]

Hi @saudeoud ,
Did the suggestion work?

[saudeoud]

Hi all,

Thanks for your interest.

Results are far better with the suggestion of @ahuber-bfw .

Still some marginal geometric effects

24° alpha angle seem too low for my case, I try to capture max. size 3 avalanches, typical for skiers
I will try with increased values of alpha angle, starting with 26°

Next step is to include forest interaction.
Do you know if the forest implementation is operational ?

#++++++++++++ Use Forest Information

NOTE AH 20240408: Dummy settings for Forest Interaction

The forest implementation should mimick/reproduce the

one from the "foreste_detrainment" branch in the avaframe/FlowPy repo

which is partly described in D'Amboise et al. (2021)

Forest-Interaction is only used if 'forest' is set to 'True'

#++++++++++++

[ahuber-bfw]

Hi @saudeoud,

Good to see that your results improved a bit.
I would also suggest, that you down-sample your DTM to 10m Resolution - if you have no specific reason to stick with 5m.

This will reduce the computation time and should also further reduce geometrical effects/artefacts - without dramatically impacting the model outputs.

For max. size 3 avalanches you could try using this parameter set as a first guess - we used it in this study:

Besides, the different forest interaction modules are operational in the current master branch - the 'forestFriction' option is described in these publications (1, 2). Information on how to use the forest modules can be found in the documentation under point iii).

Hope this is helpful as a starting point!

Btw, thx for pointing out the comment in the .ini file - we will fix it shortly ;)

[saudeoud]

@ahuber-bfw
Thanks a lot. I will try what you propose and will be back soon.