Christine Smit

This is the zarr store I generated with xarray: [generated.tar.gz](https://github.com/Unidata/netcdf-c/files/9064372/generated.tar.gz)

Interestingly, when I open and run ncdump on your zarr store, however, I see the same issue: ``` % ncdump -v lat "file://`pwd`/gldas.zarr/#mode=nczarr,zarr" netcdf \#mode\=nczarr\,zarr { dimensions: time = 1...

Strangely, when I open your dataset with xarray, the latitude values look fine: 

When I open your zarr store directly with the zarr library, the values also look fine: 

Awesome! The tricky thing is that I didn't install any NetCDF or HDF stuff directly. I'm using anaconda python. So I used anaconda's package manager, conda, to install the NetCDF...

% uname -a Darwin gs6102dscmc1csm 21.5.0 Darwin Kernel Version 21.5.0: Tue Apr 26 21:08:22 PDT 2022; root:xnu-8020.121.3~4/RELEASE_X86_64 x86_64

Am I understanding correctly that filters are somehow related to the compression? Can I somehow pick a different compression scheme that would work? Or is it that just reading the...

I was more thinking about the fact that I am currently the one creating these zarr stores. So I can pick which compression algorithm is used. Can I pick one...

Okay! I created a zarr store where none of the variables have compression and I can run ncdump on everything just fine. Note: This is different data. I didn't use...

I ran the above script last night. It appears to have stopped before doing 500 iterations, but you can at least see the trend. [loop_zarr.txt](https://github.com/zarr-developers/zarr-python/files/7871287/loop_zarr.txt)