[bjlittle]

This PR extends the CF dimensionless vertical coordinates coverage of Iris by supporting the ocean sigma over z coordinate auxiliary factory.

At present only NetCDF loading has been included in this PR, which is a re-factor of #509.

[bjlittle]

@rsignell-usgs I hope you've not been holding your breath for this ... 😜

If you get the opportunity, I'd greatly appreciate if you could kick the tyres (once again) to see if the derived coordinate is behaving as expected using some of your datasets. Thanks!

[esc24]

A minor thing, but you could use is_unknown() here.

[bjlittle]

I agree, that's better, thanks.

[esc24]

In my opinion, this change should not take place here. The method is called _check_dependencies(), which suggests it just checks for compatibility. This looks like a side effect to me.

[bjlittle]

Yup, being graceful I thought I could squeeze it in, as it is dependency related. It is the only exception and I want consistency as _check_dependencies is called from both the __init__ and update methods.

[bjlittle]

Raised #1106 in support of this PR, to ensure that NetCDF quantities with missing units default to being dimensionless, as per the stipulated CF convention, see http://cf-convention.github.io/1.6.html#units

[rsignell-usgs]

So I should go ahead and test this PR?

[bjlittle]

Yes please, thanks @rsignell-usgs.

If you need to blend in #1106 you can as there's a check in the code to raise an exception if sigma is not explicitly dimensionless.

[bjlittle]

I tried to re-use the notebook http://nbviewer.ipython.org/urls/raw.github.com/rsignell-usgs/ipython-notebooks/master/ncom_coords.ipynb from #509 ... but it's since been deleted. Understandable really, it was 8 months ago 😉

[rsignell-usgs]

I'm just back from travel, so probably won't be able to test today, but that notebook for testing is now here:
http://nbviewer.ipython.org/github/rsignell-usgs/notebook/blob/master/ncom_coords.ipynb

[bjlittle]

No problem, thanks for posting the notebook again, that's really helpful.

[bjlittle]

Seems to do the job @rsignell-usgs !

Over to you @esc24 ... let's get this in before we get distracted again.

[esc24]

Thanks @bjlittle.

[bjlittle]

Awesome! Nice one @esc24 and thanks @rsignell-usgs

[rsignell-usgs]

Awesome! Are you guys proceeding on to handle other dimensionless vertical coordinates, or was this the specific one you needed for now?

[bjlittle]

I don't know if @esc24 has plans to schedule further development effort in this area in the short-term, but I'm assuming that you'd like to see support for the following dimensionless vertical coordinates in the following order:

1. ocean_s_coordinate_g1 (used by most ROMS models > 2010)
2. ocean_s_coordinate (used by ROMS models < 2010)
3. ocean_s_coordinate_g2 ( used by some ROMS models > 2010)
4. ocean_sigma_coordinate (used by POM and ECOM models)

Is that right @rsignell-usgs ?

[rsignell-usgs]

Yes, that is exactly right! If those additional 4 coordinate systems were handled, we would be able to handle about 90% of the model results in the US Integrated Ocean Observing System, which would be awesome.

[bjlittle]

@rsignell-usgs do you have example OPeNDAP datasets for ocean_s_coordinate_g1 ? I'm not making any promises here ... just teeing up everything I (we) might need next, sometime 😉

[rsignell-usgs]

Oh Yes! Here is the OPeNDAP URL for the US Integrated Ocean Observing Sysstem mid-Atlantic ocean forecast model, which uses ocean_s_coordinate_g1:

url = 'http://tds.marine.rutgers.edu/thredds/dodsC/roms/espresso/2013_da/avg_Best/ESPRESSO_Real-Time_v2_Averages_Best_Available_best.ncd'
s = iris.load_cube(url,'sea_water_salinity')

[bjlittle]

Cheers!