HostDynamics Driver

[pbosler]

Summary

1D host dynamics implemented, with support for writing HOMME prognostic variables and Haero::Atmosphere state data to netCDF file.

To check it out: Run the new host dynamics unit test. It will create 2 netcdf files, one for height-initialized interfaces, the other for pressure -initialized interfaces. Each file contains 2 time steps: the initial time t=0 and t = period/2, to show the state after it's been vertically lifted to the furthest extent. To view the data, run

ncdump host_dynamics_test_zinit.nc

Features

The model is vertically Lagrangian, meaning that the geopotential surfaces (and height surfaces) move with the velocity of the vertical wind. If you need to analyze a fixed pressure level or fixed height level, we'll need to implement an interpolation scheme. Ekat already has a piecewise linear scheme built-in.

As demonstrated in the Host Dynamics unit test:

• The test case has an updraft for the first half of its oscillation period, followed by a down draft.
• For the parameter values in the unit test, the 2 model levels nearest the surface approach saturation at t = period/2. (97% relative humidity).
• Temperature accounts for changes in density due to lifting, and changes in humidity.

Questions / issues that came up

1. In Haero::Atmosphere do we want to keep height on interfaces, or would it be better to have it at level midpoints (so that every variable passed to physics is at a level midpoint)?
2. Intent: Use SI units always.
3. Relative humidity computed as qv/qvsat, where qv is the water vapor mixing ratio and qvsat is the saturation mixing ratio. Saturation mixing ratio computed with the Tetens equation from Soong-Ogura 1973, Appendix A eqn. (A1), which is also used in Klemp-Wilhelmson 1978 eqn. (2.11). Is there a better way to compute relative humidity?

Documentation

Documentation is done, but needs to be integrated to the haero design doc; see #21, #71.

Developments & fixes included in this PR

1. HostDynamics impl'd, creates & updatesAtmosphere state containers for use with params. Doesn't build tendencies yet. Created Define Driver dynamics tendencies #77.
2. Bug fix for read_yaml unit test.
3. Updated NcWriter to new column-based indexing scheme, added support for Species and Atmosphere. Model support is likely something that we'll want for NcWriter: created Add aerosol config support in NcWriter #76 .

Closes #75.
Closes #70.
Closes #14.
Closes #19.

Testing

New test of host dynamics passes on Mac with pack size = 1, 2.

[jeff-cohere]

Any reason you chose not to declarez0 as a const reference? Just curious.

[pbosler]

Yes, I want to force us to think like Homme, with vertical levels starting at model top and going down to the surface as the index k increases. So, in case a user puts z0 values in ascending order, I wanted a non-const copy to run std::reverse on them.

[jeff-cohere]

Strictly speaking (at least conceptually), ${HAERO_DRIVER_LIBRARIES} should precede ${HAERO_LIBRARIES}, since systems with a single-pass linker (like Linux) gather dependencies from earlier items and satisfy them with later ones, and arguably the driver depends on the library. But I think the only libraries we build into the driver and not the library are I/O-related, so it doesn't matter in this case.

[jeff-cohere]

At some point, it might be nice to have a reader for these kinds of tests that can verify properties of the written files. Not urgent, though.

[jeff-cohere]

Nice job, Pete! I left some comments here and there, but overall it looks good to me. I didn't look very closely at the atmospherics part, so maybe others can inspect that more carefully.

[mjs271]

A lot of this appears above the level of my current expertise, but can confirm that it builds/runs on my Mac for pack size 1 and 2, and the output appears to be as expected.

[jeff-cohere]

Does anyone object to merging this PR? The driver in the main branch doesn't use our single-column indexing scheme, so it's out of date.

I'd like to claim that our future PRs will probably be smaller and more incremental, since we've made a lot of important and significant decisions lately, but I know better than to promise that. :-)

[singhbalwinder]

I didn't go through the entire PR but I think it is okay to merge this PR. The changes that I went through looks like a great addition to me. I would also prefer small incremental PRs but I know it is sometimes hard to do that.