Created by on - Updated on 02/13/2020 10:42

Thu, 02/13/2020 - 10:44

Some pressure level RH values are not within 0-100 range and some as as far off as -200%. Should we be setting these values to 0-200 or do the negative values represent supersaturation or something like that?

Hello Cathy,

Supersaturation with respect to ice can occur and values can be greater than 150%. However, the values shouldn't be negative. Given that pressure level RH is stored in spherical harmonics, it could be due to the spectral transformations or maybe a spectral truncation before the transformation. From where did you get the data and how did you process it (if at all)? If you tell me a date, time, level and location, I will check with the native grid.



We checked Jan 1 0z 1979. Values had the range over all levels. 

 r_min = -6.01934528350829 ;

 r_max = 162.923529684544 ;


We did get the files from NCAR but the values from Copernicus were the same. Grid was netCDF


        longitude = 1440 ;

        latitude = 721 ;

        level = 37 ;

        time = 1 ;

        short r(time, level, latitude, longitude) ;

                r:scale_factor = 0.00257798170338687 ;

                r:add_offset = 78.450803209666 ;


from Copernicus. Let us know if you need further info. 



Also, we checked our specific humidity and values were also below 0 for pressure level values.

I know that can be the result of post processing. If it isn't in the native grid, should we truncate either rhum or shum at 0? If it is, should we truncate? Thanks.



Hello Cathy,

Values of relative and specific humidity should not be below zero, so yes, you should set them to zero. Relative humidity is archived as spherical harmonics, so even though the model should not have negative values, the spectral transformations can introduce them because of Gibbs oscillations. Specific humidity is archived in grid point space on a reduced Gaussian grid, but you are using data on a regular latitude/longitude grid, so the interpolation may have introduced the negative values. However, I would have thought the interpolation scheme would be designed not to do that, so I will check that out.

Best wishes,


Add new comment

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.