Tracer density #434
Tracer density #434
Conversation
gusto/diagnostics.py
Outdated
|
|
||
| name = "TracerDensity" | ||
|
|
||
| def __init__(self, m_X, rho_d): |
There was a problem hiding this comment.
I think this __init__ method should still have the arguments of the parent class, because the user might still want to specify the function space and the computation method.
We also need to make clear whether the m_X and rho_d arguments should be strings (the names of the fields) or the Function variables. I think if we change the name of the arguments to mixing_ratio_name and density_name then it is clearer?
So I suggest something like:
def __init__(self, mixing_ratio, density_name, space=None, method='interpolate', required_fields=()):
There was a problem hiding this comment.
I agree with having more descriptive variable names.
Is the required_fields=() just to initialise this variable before using it in:
super().init(method='interpolate', required_fields=(mixing_ratio_name, density_name))
?
I've tested the four different methods with this diagnostic. It works when specifying 'interpolate' or 'project'. However, there are issues when specifying 'assign':
ValueError: All functions in the expression must have the same element as the assignee
and 'solve':
NotImplementedError: Solve method has not been implemented for diagnostic TracerDensity
Should this diagnostic be able to be used with these methods, or is it to be expected that these aren't compatible?
There was a problem hiding this comment.
Ah I think you're right, we don't need required_fields in the __init__, as this should just be set in the call to super().__init__.
For methods, 'assign' only makes sense when the density and the mixing ratio field in the same space (what assign does is evaluate the values at the degrees of freedom). The solve method is there if we need to solve a more complicated equation to get a diagnostic, rather than interpolating/projecting/assigning. In this case I don't think we need a solve method so it should throw an error.
There was a problem hiding this comment.
Ok, thanks.
I'll put in the function header that:
Valid options are 'interpolate', 'project' and 'assign', so that one doesn't try to use 'solve' with this
gusto/diagnostics.py
Outdated
| def __init__(self, m_X, rho_d): | ||
| """ | ||
| Args: | ||
| m_X: the mixing ratio of the tracer |
There was a problem hiding this comment.
We need to include the type of the argument in the docstrings, so this should be something like:
mixing_ratio_name (str), the name of the tracer mixing ratio variable.
and so on for the other variables
gusto/diagnostics.py
Outdated
| rho_d = state_fields(self.rho_d) | ||
| space = m_X.function_space() | ||
| self.expr = m_X*rho_d | ||
| super().setup(domain, state_fields, space=space) |
There was a problem hiding this comment.
If we include space as an argument to the __init__ then we won't need to include it here
Added a tracer density diagnostic. This is useful when examining conservation of mass when transporting species via the dry density and mixing ratio.
Here are examples of plotting the normalised difference in the tracer density diagnostic over time, using the planar_nair_gaussian case as examined the Bendall, 'Trilemma' paper.