5.2 Derivatives on model levels

If you try to postprocess derivatives on eta levels (like the potential vorticity on the model levels) and you do interpolate on the horizontal (for instance : from a file ARPEGE to a file ALADIN), the software will fail because derivatives will be missing : this is because the horizontal derivatives are available only in the model geometry.

      A way to overcome this limitation is to first change the geometry of your input file to the geometry of your output file (using the configurations 927), then to post-process on the new "model" grid (CFPFMT='MODEL' and LFITS=.FALSE.).

  Unfortunately this does not work if the target geometry is LAT X LON ! In this case you have to trick the software so that the field you wish to interpolate will be considered as a passive scalar field ; this can be achieved in two steps :

 

1.

You should create a history file with the supplementary fields you wish to interpolate ; this can be achieved either by running a configuration of the kind "927" in which namelist you will request the supplementary fields, or by running a normal post-processing job in the model geometry (CFPFMT='MODEL') and using the "appending fields" trick (refer to the previous section).      

2.

If they are spectral you can post-process your supplementary fields as model passive scalar fields (setting NFPASS and the field descriptors TFP_SCVA()). Else you can still trick the software by activating the pronostic field for gridpoint cloud fraction (setting (LGPA=.TRUE.)) and feeding the cloud fraction with one of your supplementary field through a proper setting of TFP_CLF%CLNAME. Notice : this is possible only because there is -- by "chance" ! -- no control of the interpolations overshoot for cloud fraction (In principle the interpolated cloud fraction should be controlled in order to remain between 0. and 1.).