Hi @benbovy , could you please have a look at fastscape-lem/fastscapelib#62 ? It is a big one, and it's preferable to review and merge it before working on flow routing API. Thanks!
One thing that I'll try is to refactor the
fill_sinks functions in sinks.hpp so that those are grid agnostic.
I'm starting to implement the
flow_graphe API!
do you remember why ? is it related to capability to load a computation from a config file, or from the python side?
We should be able to set those parameters from Python, so I think we need to provide instances of those subclasses when creating an instance of
flow_graph.
:+1:
Added support for log scales as well
although changing from linear to log or from log to linear seems to be broken on the shaders
@benbovy I thought you'd find this cool https://twitter.com/BarrettMeeker/status/1353123863257735168?s=20 https://twitter.com/BarrettMeeker/status/1350208383060643841?s=20
the simulation (and rendering) is made entirely in OpenGL shaders, on a 1000x1000 grid
I'd love to see something like that to simulate mountain formation (with some realistic textures representing vegetation levels, snow, etc.)
That would be super cool indeed
Given the text on the tweet. I understand they have a heightmap for the sand (each "pixel" of the grid has a given sand height). And they compute the height of each pixel given the wind, the previous state, and the neighborpixels. I wonder how accurate their simulation is. But it looks correct at least :)
Regarding the formation of mountains, I've been working with someone (Guillaume Cordonnier) who is developing very fast algorithms and scientifically quasi-accurate for such kind of application (e.g., https://team.inria.fr/imagine/sculpting-mountains-interactive-terrain-modeling-based-on-subsurface-geology/)
(I don't have a link to an animating landscape, unfortunately)
It might be possible to implement this in shaders and get a very fast implementation of this :)
I'm thinking right now about a terrain texture component for ipygany :-). Something like described here: http://unitytipsandtricks.blogspot.com/2014/07/terrain-texture-splatting-and-threejs.html. One could pick a few built-in textures (or load custom textures) and with some parameters tweak texture splatting.
Hi @xlia62, sorry for the late reply here. Yes it is possible, you could replace the
DrainageArea process by the FlowAccumulator process (https://fastscape.readthedocs.io/en/latest/_api_generated/fastscape.processes.FlowAccumulator.html#fastscape.processes.FlowAccumulator) in which you can provide a value for runoff (if you assume that all rainfall on the land surface is participating to surface runoff). You can also check the orographic precipitation extension to see how we can reuse those components (https://github.com/fastscape-lem/orographic-precipitation/blob/master/orographic_precipitation/fastscape_ext.py).
Hi @benbovy, I was looking at using the stratigraphy tracking of reflectors in FastScape (fortranlib) but I noticed that the reflectors are only uplifted and not horizontally advected in the subroutine run_Strati. So, in the case of active margins in a seafloor spreading scenario, this would cause an error in the reflector location over time. Thoughts? Thanks.
Hi @rbrtmch, I've inspected the Fortran source and I'm not sure what's going on, as I see that the
Advect subroutine updates the bedrock elevation b, the Strati subroutine reuses b to update the reflectors, and the FastScape_Execute_Stepsubroutine should call Advect before Strati... I rarely use the Fortran library directly, though. I'm rather using the fastscape Python library which has its own StratigraphicHorizons component.
@benbovy , from what I see the issue is the same in the python version, the horizons are only corrected for vertical motion. As I understand it, the use of b is only to prevent crossovers resulting from vertical and/or horizontal motion. Horizontal advection in both python and fortran version does not advect the horizons horizontally.
Ah yes I guess you are right @rbrtmch, at least regarding the
fastscapelib-fortran (and its Python bindings). All active horizons should be advected (as their elevation corresponds to the elevation of the simulated topographic surface), but this is not the case for inactive horizons, which are indeed only uplifted in the subroutine run_Strati. However, in the fastscape library, unless I'm missing something, the elevation of inactive horizons is updated with the bedrock vertical motion, which should include the vertical effect of advection if the HorizontalAdvection process is included in the model.
Hi @benbovy I'm trying to install the python package from source (I only need the python module fastscapelib-fortran). I had success on my Mac, but not so on Ubuntu 20.04. An error occurred while building with CMake. The install exits with "ERROR: Failed building wheel for fastscapelib-fortran". There is a lot of information to post on here.
4 replies