CDO

Hello-

I have just started using cdo. Thanks for a wonderful set of tools. I hope this is the right place to report these issues.

I am trying to do a time average on a 500 mb height field and saving the results in a netCDF file so I can read it into Unidata's Integrated Data Viewer (IDV). The original file (ftp://ftp.cdc.noaa.gov/Datasets/ncep.reanalysis/pressure/hgt.2010.nc) defines the height field as:

short hgt(time, level, lat, lon) ;
                hgt:long_name = "mean Daily Geopotential height" ;
                hgt:actual_range = -559.25f, 32277.5f ;
                hgt:valid_range = -700.f, 35000.f ;
                hgt:units = "m" ;
                hgt:add_offset = 32066.f ;
                hgt:scale_factor = 1.f ;
                hgt:missing_value = 32766s ;
                hgt:precision = 0s ;
                hgt:least_significant_digit = 0s ;
                hgt:GRIB_id = 7s ;
                hgt:GRIB_name = "HGT" ;
                hgt:var_desc = "Geopotential height\n",
                        "H" ;
                hgt:dataset = "NCEP Reanalysis Daily Averages" ;
                hgt:level_desc = "Multiple levels\n",
                        "F" ;
                hgt:statistic = "Mean\n",
                        "M" ;
                hgt:parent_stat = "Individual Obs\n",
                        "I" ;

There is an error in the definition in that the valid_range should be the same data type as the packed data (short) instead of float as it is defined in the file. However, the IDV (which uses Unidata's netCDF-Java library) can read this.

When I issue the following command:

cdo timavg -sellevel,500 -seldate,20100710,20100808 hgt.2010.nc hgt.20100710_0808.nc

The variable is now defined as:

short hgt(time, lev, lat, lon) ;
                hgt:long_name = "mean Daily Geopotential height" ;
                hgt:units = "m" ;
                hgt:add_offset = 32066.f ;
                hgt:scale_factor = 1.f ;
                hgt:_FillValue = 32766s ;
                hgt:actual_range = -559.25, 32277.5 ;
                hgt:valid_range = -700., 35000. ;
                hgt:precision = 0 ;
                hgt:least_significant_digit = 0 ;
                hgt:GRIB_id = 7 ;
                hgt:GRIB_name = "HGT" ;
                hgt:var_desc = "Geopotential height\n",
                        "H" ;
                hgt:dataset = "NCEP Reanalysis Daily Averages" ;
                hgt:level_desc = "Multiple levels\n",
                        "F" ;
                hgt:statistic = "Mean\n",
                        "M" ;
                hgt:parent_stat = "Individual Obs\n",
                        "I" ;

The valid_range attribute values have been changed to doubles (vs. floats in the old file). This causes the scale_factor and add_offset to be applied to the valid_range so the grid ends up with all invalid values. I think cdo should maintain the data
type of the attributes when creating the new file. I saw the note in the documenation about scale/offset, but in this case the range of the data are not changing, the scope of the valid_range is changing. I can use the -b 32 to unpack the data, but that creates larger files.

Another issue is that the units of the level have been changed from millibar to Pa:

original:

float level(level) ;
                level:units = "millibar" ;
                level:actual_range = 1000.f, 10.f ;
                level:long_name = "Level" ;
                level:positive = "down" ;
                level:GRIB_id = 100s ;
                level:GRIB_name = "hPa" ;

cdo generated variable:

double lev(lev) ;
                lev:long_name = "pressure" ;
                lev:units = "Pa" ;
                lev:axis = "z" ;

For a 3-D depiction of the data with the IDV, this is problematic since it plots the data at 5 hPa instead of 500 hPa (millibars).

I'm using cdo 1.4.4, but I did not see any indication that these issues were addressed in subsequent releases.

Thank you for your time.

Don Murray