isobar | Python library | Audio Utils library

This process can get a little cleaner by making use of MetPy's XArray accessor to simplify selection based on generic dimensions like 'time' and 'vertical'. We can also get the values on common levels formed using np.intersect1d:

Notice that the metpy_crs identified in your dataset from .metpy.parse_cf() is given as Projection: latitude_longitude, which is not correct, as your dataset has 2D latitude/longitude coordinates with 1D y/x coordinates in a projected grid space. This is likely triggered by your dataset's metadata not being CF-compliant. At it is now, because of this, MetPy is treating your x coordinate as if it were longitude and y as if it were latitude (notice the ranges in those coordinates), even though they are not. The outcome here is definitely far from ideal (no error message or anything), so I'd recommend submitting an issue on MetPy's issue tracker.

To directly fix your issue, you will need to manually specify the CRS/projection of your data. To do so (in MetPy v1.0 and later), use the assign_crs method on MetPy's Dataset accessor instead of using parse_cf (which only works for CF-compliant datasets).

Looks like I have found a bug within the MetPy code, which appear when plotting the data which cover both sides of the zero meridian: https://github.com/Unidata/MetPy/pull/1595

A temporal solution is to edit the file src/metpy/plots/declarative.py and replace the line:

The problem with xarray 0.15.1 is a known issue that should be fixed in the next release.

The error you get with 0.15.0 is certainly confusing. The problem is that the example code you're using is from the docs, which were pointing to the docs for 1.0, which is currently in release candidate status--which means you're most likely not running it, but MetPy 0.12. The problem is that MetPy 1.0 changes the order of arguments to the relative_humidity_from_specific_humidity. So the example code you downloaded wasn't appropriate for the version of MetPy you're running.

I have updated the MetPy docs to point the latest docs to the 0.12 release docs. You should now be able to download the correct example code for your version.

The problem you're running into is that multiplying units works with numpy arrays, but not with xarray.DataArray instances--which is what you get when you get vertical. It'd be nice if this weren't so, but currently that's the world we live in. To get around this MetPy has some helpful tools for xarray and units.

In this case, though, there's no need to manually convert your units from Pa to hPa--the beauty of the unit support is as long as the units are known and of the proper dimensionality, potential_temperature does not care what they are. This is how I would write your code:

So while it seems like your data are 2D since you only have a single time, based on this output:

I think your issue is that stat_contour does not work because it needs a complete grid. I found this blog's article that explain how to deal with this issue: https://www.r-statistics.com/2016/07/using-2d-contour-plots-within-ggplot2-to-visualize-relationships-between-three-variables/

I used this blog's answer to build the following answer adapted to your question and the minimal example you provided.

First, you need to create a predicted model based on your restricted dataset "datgeo".