Forest Canopy Model

[hgopalan]

This PR provides a forest canopy model in AMR-Wind. Two kinds of forests are supported: (i) homogeneous forest and (ii) heterogeneous forest using an empirical model for leaf area density.

Forest overlap is allowed and is not checked. It is expected that the users provide accurate inputs and currently no checking is included. The model requires following inputs from the user: (i) tree type (ii) tree center x (iii) tree center y (iv) tree height (v) tree girth preferably at the widest point (vi) tree drag coefficient (vii) tree leaf area index (viii) location of maximum leaf area density for heterogeneous forests.

The model will work with any LES model as the drag forcing due to the turbulence is required only in momentum equation. It is also possible to include a turbulence forcing term in one equation kegs model.

The model will be extended to the one-equation RANS model in a future PR by including the turbulence contribution.

Please check the type of change introduced:

• Bugfix
• [X ] Feature
• Code style update (formatting, renaming)
• Refactoring (no functional changes, no api changes)
• Build related changes
• Documentation content changes
• Other (please describe):

Checklist

The following is included:

• new unit-test(s)
• [X ] new regression test(s)
• documentation for new capability

This PR was tested by running:

• the unit tests
  • on GPU
  • on CPU
• the regression tests
  • on GPU
  • [X ] on CPU

Issue Number: #1322

[hgopalan]

Sample result. Top is a heterogeneous tree model while other three are homogeneous.

[marchdf]

out of date comment

[marchdf]

@hgopalan sorry this slipped off my radar. Can you remind me where we are at?

[hgopalan]

@hgopalan sorry this slipped off my radar. Can you remind me where we are at?

You wanted to discuss with me about the multiple for loops and may be come up with a better way to do it. Rest of issues are sorted.

[marchdf]

After chatting with @hgopalan: he will remove the n_cells x n_forest loop with a bounding box for the forest.

The other thing I would recommend is to create a Forest struct that contains all the data for a forest and that can be stored in a vector (device or host). That will simplify the code quite a bit.

struct Forest{
real rad
int type
Box bounding_box
etc
}

// later:
amrex::Vector<Forest> m_forests;

// read the data from file
loop on file lines{
  Forest f;
  f.rad.= value_from_file
  f.type = value_from_file
  f.bouding_box = bx
  forests.push_back(f)
}

// and then a single copy to device
amrex::DeviceVector<Forest> m_d_forests
DeviceCopy blah

[hgopalan]

Code has been updated with rearrangement. @marchdf I have not added bounding box since I wanted a circular search. Will it be possible to define a cylinder for contains?

[marchdf]

Just make a bounding box wherein the cylinder fits. It's fine if the bounding box technically has more volume than the cylinder it contains.

[marchdf]

@hgopalan confirmed I didn't break anything. While I tidy up a couple other things, could you add a unit test, @hgopalan ? Just add something like you did with test_abl_terrain.cpp. Maybe instead of point probes, just do a field norm on the drag and blank fields?

[marchdf]

Ok I am done with all my changes. There's a last bit that's bothering me (the if condtion on the type in the lm and area_fraction member functions). But I need to move on to other things.

Tagging some others (@moprak-nrel and @mbkuhn) on the PR review since I am now a major author ;)

Tests pass on CPU and GPU.

[marchdf]

Blocking with a change request until others have been able to look.