sub dataset and interpolation

[galleon]

I am discovering the power of xarray but I am still unsure if I can do what I want.
From a GRIB file, I am reading wind profile for the whole globe at different isobaric & for a 24-hour period.
I have therefore the following:

My application being aeronautics, I want to compute u, v through interpolation at any given (t, alt, lat, lon) in a subset of the original data.

I know that I can do uwind = ds.u.interp(time=t, isobaricInHPa=i, latitude=lat, longitude=lon)

But I wanted to try to optimize space and time so I decided to explore the following steps:

1. Get a sub dataset with all timesteps, all altitudes but a [lat_min, lat_max], [lon_min, lon_max] window
2. Move isobaric to altitude (which is more natural for my use-case)
3. Interpolate my data

I faced newbie's issues in all my steps :-)

1. I tried this uwind4 = uwind.where(uwind.latitude.isin([90, 89.75])).where(uwind_.longitude.isin([359.5, 359.8])) which turned out to give me the same dataset size full of nan while I was expecting a smaller one. So I am really wondering if that is possible.

2. I tried the following

altitude = aero.h_isa(uwind.isobaricInhPa.data * 100) / aero.ft
uwind2 = uwind.assign_coords(altitude=("isobaricInhPa", altitude))

which works well but altitude is not an index apparently so I cannot sel using it. Did I miss something there ?

3. is easy (and a bit mysterious to me so blazing fast)

Thanks for your help

[TomNicholas]

Welcome to xarray!

For 1. Did you try using .sel(latitude=[min, max], longitude=[min, max], meathod="nearest")?

For 2. When you do that altitude becomes a 1d coordinate but is not backed by an index (not bold in the repr). Selection along these will become possible very soon #2028

For 3. Is there there an issue? :)

[galleon]

Thank you !

1. No I did not, how can I have missed that one. It does not do exactly what I want since it will only pick min and max values while I want all values in the intervals. I guess I will need to create from the coordinates a list of all values that are in the interval. Since all indexes seem to have comparison operators (<, ...), it would be a nice addition to the API to enable such selection ... a bit like what where does but just keeping the meaningful data !

2. Cool ! In the meantime is there a way to either:

• create a new index
  or
• transform an index and rename it - I dont mind if I loose isobaric since altitude is just a one to one transformation

3. No interp is cristal clear :-)

And thanks for the great software.

[abkfenris]

Try .sel(latitude=slice(min, max), longitude=slice(min, max)) to get anything in the interval.

Indexes I'm less sure about.

[benbovy]

For 2, with the last Xarray release v2022.09.0 you can now do this:

uwind2 = uwind.assign_coords(altitude=("isobaricInhPa", altitude)).set_xindex("altitude")

uwind2.sel(altitude=...)