Size of zarr store vs grib is 10x

[kingkong-cmd]

I'm donwloading wind forecast data from GFS in GRIB2 format. I'm only getting two variables (u and v 10m), and one GRIB file for each forecast time, which means I am downloading apprx. 150 files. Each file is apprx. 1.5 MB, so in total its around 200 MB.

I tried converting into a zarr store by reading each file into xarray and then write to the zarr store like this: ds.to_zarr('test.zarr', append_dim = 'time') , but noticed that the zarr store is 2+ GB on disk, meaning more than 10x size of original files. Am I doing something wrong here or are GRIB files just really well compressed? Is there anything I can do to reduce size?

For the record, this is the xarray representation of a single GRIB files:

<xarray.Dataset> Size: 17MB
Dimensions:    (time: 1, latitude: 721, longitude: 1440)
Coordinates:
  * time       (time) datetime64[ns] 8B 2024-10-22T17:00:00
  * latitude   (latitude) float64 6kB 90.0 89.75 89.5 ... -89.5 -89.75 -90.0
  * longitude  (longitude) float64 12kB 0.0 0.25 0.5 0.75 ... 359.2 359.5 359.8
Data variables:
    ugrd       (time, latitude, longitude) float64 8MB -0.9188 ... -0.9888
    vgrd       (time, latitude, longitude) float64 8MB -3.25 -3.24 ... -5.33
Attributes:
    meta:     Generated with gribberishpy

And this is the xarray representation when loading the whole zarr store into xarray:

<xarray.Dataset> Size: 3GB
Dimensions:    (latitude: 721, longitude: 1440, time: 156)
Coordinates:
  * latitude   (latitude) float64 6kB 90.0 89.75 89.5 ... -89.5 -89.75 -90.0
  * longitude  (longitude) float64 12kB -180.0 -179.8 -179.5 ... 179.5 179.8
  * time       (time) datetime64[ns] 1kB 2024-10-22T17:00:00 ... 2025-09-18T1...
Data variables:
    ugrd       (time, latitude, longitude) float64 1GB ...
    vgrd       (time, latitude, longitude) float64 1GB ...
Attributes:
    meta:     Generated with gribberishpy