Add preprocessing functions

[sjavis]

Adds preprocessing functions for all of the topics in the 'data preprocessing' page of the docs. Intended for use with the batching (#194). Generally the functions are designed to accept either file paths or cf.Field inputs and return single or lists of cf.Field. Each function can also write an output file by providing the output_file argument.

• Preprocessing utility functions in src/tctrack/preprocessing.py
• Tests
• Documentation & add to the 'data preprocessing' page.
• Replace preprocessing in tutorial
• Add checks that esmpy is installed for regridding. Update the docs to reflect that this is now a "proper dependency".

Closes #189

[MarionBWeinzierl]

Just had a quick look and some comments -- @surbhigoel77 , could you have a more thorough look through this PR?

[sjavis]

In fc12ab3 I have also added a correction for what I suspect is an error in cf.curl_xy that causes the result to be the negative of the actual value. I will add an issue to cf-python about this when I get time.

This seems to resolve the issue in #156 when combined with updates to TSTORMS. This will be done in another PR.

[surbhigoel77]

Maybe we can add a check before we build the dict that raises ValueError error if there is a missing variable name

[sjavis]

Perhaps I'm misunderstanding but it is already checking if any output variable names are missing in the loop below.

I don't think it would be easy to do it before building the dict because it is only after reading the input files that we can know which variables are available.

[surbhigoel77]

Not quite clear on why we are negating the curl?

[surbhigoel77]

I found this formula of curl :
curl(u, v) = (1/r) * dv/dλ − (1/r·cosφ) * d(u·cosφ)/dφ

As per your comment, only first term gradient needs negation due to the opposite direction of latitude and theta. Why would we negate the entire curl?

[sjavis]

cf.curl_xy uses the spherical polar form of the equation:

$$\frac{1}{r \sin\theta} \left( \frac{\partial (f_\phi \sin\theta)}{\partial \theta} - \frac{\partial f_\theta}{\partial \phi} \right)$$

So the first term is negative because the derivative is taken in the code wrt latitude rather than $\theta$ as already stated.

For the second term, I mistakenly thought it had been added rather than subtracted. The actual issue is that $f_\theta$ should be the component of $f$ in the southward direction. So $f_\theta = -v$. I've modified the comment to reflect this.

So overall it is calculating the negative of the curl, hence the need for the negation.

I will open an issue on cf-python regarding this.

[MarionBWeinzierl]

I am happy from my part, but leave it to @surbhigoel77 to accept this PR after she has done another review