CDO

I have attached a netcdf file I am converting to a grib using the CDO.
When I do this I lose precision in the depth variable. I am trying to understand why this is happening.
I am running this command.

cdo -f grb -setparam,151.130 -sellonlatbox,0,360,-80,80 test.nc test.grb

Checking the original netcdf file I see the following info

(venv) jlambert@odin phy % cdo -f grb -setparam,151.130 -sellonlatbox,0,360,-80,80 test.nc test.grb
cdo(1) sellonlatbox: Process started
cdo    setparam: Processed 215766720 values from 1 variable over 1 timestep [0.86s 132MB]
(venv) jlambert@odin phy % cdo sinfo test.nc
   File format : NetCDF4
    -1 : Institut Source   T Steptype Levels Num    Points Num Dtype : Parameter ID
     1 : unknown  MOI GLO12 v instant      26   1   8817120   1  F32  : -1            
   Grid coordinates :
     1 : lonlat                   : points=8817120 (4320x2041)
                        longitude : -180 to 179.9167 by 0.08333334 degrees_east  circular
                         latitude : -80 to 90 by 0.08333333 degrees_north
   Vertical coordinates :
     1 : depth_below_sea          : levels=26
                            depth : 0.494025 to 186.1256 m
   Time coordinate :
                             time : 1 step
     RefTime =  1950-01-01 00:00:00  Units = hours  Calendar = gregorian
  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss
  2025-03-12 00:00:00
cdo    sinfo: Processed 1 variable over 1 timestep [0.01s 23MB]

(venv) jlambert@odin phy % cdo sinfo test.grb
   File format : GRIB
    -1 : Institut Source   T Steptype Levels Num    Points Num Dtype : Parameter ID
     1 : unknown  unknown  v instant      26   1   8298720   1  P16  : 151.130       
   Grid coordinates :
     1 : lonlat                   : points=8298720 (4320x1921)
                              lon : 0 to 359.9167 by 0.08333333 degrees_east  circular
                              lat : -80 to 80 by 0.08333333 degrees_north
   Vertical coordinates :
     1 : depth_below_sea          : levels=26
                            depth : 0 to 186 m
   Time coordinate :
                             time : 1 step
  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss
  2025-03-12 00:00:00
cdo    sinfo: Processed 1 variable over 1 timestep [0.10s 21MB]

As you can see below the precision of the depth is reduced for the grib and appears to rounding to an integer.

(venv) jlambert@odin phy % cdo showlevel test.nc 
 0.494024992 1.54137504 2.64566898 3.81949496 5.07822418 6.44061422 7.92956018 9.57299709 11.4049997 13.4671402 15.81007 18.4955597 21.5988197 25.2114105 29.4447308 34.4341507 40.3440514 47.3736916 55.7642899 65.8072662 77.8538513 92.3260727 109.729301 130.666 155.850693 186.125595
cdo    showlevel: Processed 1 variable [0.00s 23MB]
(venv) jlambert@odin phy % cdo showlevel test.grb
 0 2 3 4 5 6 8 10 11 13 16 18 22 25 29 34 40 47 56 66 78 92 110 131 156 186
cdo    showlevel: Processed 1 variable [0.04s 21MB]

This is the output of my cdo -V

(venv) jlambert@odin phy % cdo -V
Climate Data Operators version 2.5.0 (https://mpimet.mpg.de/cdo)
System: aarch64-apple-darwin23.6.0
CXX Compiler: clang++ -std=gnu++11 -std=gnu++20 -g -O2  -pthread
CXX version : Apple clang version 16.0.0 (clang-1600.0.26.4)
CXX library :
C Compiler: clang -g -O2  -pthread -pthread
C version : Apple clang version 16.0.0 (clang-1600.0.26.4)
F77 Compiler: gfortran -g -O2
F77 version : GNU Fortran (Homebrew GCC 14.2.0_1) 14.2.0
Features: 16GB 12threads c++20 Fortran pthreads HDF5 NC4/HDF5 OPeNDAP sz
Libraries: yac/3.1.0 NetCDF/4.9.2 HDF5/1.14.5
CDI data types: SizeType=size_t
CDI file types: srv ext ieg grb1 grb2 nc1 nc2 nc4 nc4c nc5 nczarr 
     CDI library version : 2.5.0
 cgribex library version : 2.3.1
 ecCodes library version : 2.39.0
  NetCDF library version : 4.9.2 of Mar 14 2023 20:34:25 $
    exse library version : 2.0.0
    FILE library version : 1.9.1