esds/posts/2021/map_blocks_example/index

[utterances-bot]

How to Use xarray.map_blocks for Vertical Interpolation of a 3D Field — NCAR-ESDS 0.1 documentation

https://ncar.github.io/esds/posts/2021/map_blocks_example/index.html

[tlsw231]

"Currently, this implementation takes 18 minutes to compute"

I just wanted a clarification what is meant here by the "entire spatial domain". Is it taking 18 minutes for the subset defined by:

ds.isel(member_id=slice(0, 2), nlat=slice(2, 5), nlon=slice(1, 5))

Or is it for the full dataset defined by the full range of nlat and nlon values?

[tlsw231]

@matt-long @mgrover1 Forgot to tag you in the previous reply. I wanted to follow up on this. I am working with 0.1 degree POP output and am finding that with 96 cores, it takes around 1 hour to process 12 monthly files (i.e., 1 year). Each high resolution file has 2400 x 3600 x 62 points.

[mgrover1]

"entire spatial domain" here refers to the entire dataset, across all nlat/nlon values

[tlsw231]

What is the size of the dataset? Can you give me all the dimensions? I am trying to see how fast of slow my runtimes are. My dataset size is (12,62,2400,3600).

[tlsw231]

I am sorry I meant to ask what is the shape (not size) of the dataset.

[tlsw231]

Okay, the docs mention "a single ensemble member". So am I right that your dataset is 5D? The fifth dimension being the ensemble member, the other four being the ones in mine as well (time,depth,nlat,nlon).