Two important papers on the nudging in WRF :

• Omrani et al. : DOI 10.1007/s00382-014-2453-5
• Spero et al. : DOI: 10.1175/JAMC-D-17-0360.1

 &fdda
 grid_fdda			     = 2, 
 gfdda_inname			     = "wrffdda_d<domain>",
 gfdda_end_h                         = 800000,
 gfdda_interval_m 		     = 360, 
 fgdt                                = 0, 
 if_no_pbl_nudging_uv		     = 1, 
 if_no_pbl_nudging_t		     = 1, 
 if_no_pbl_nudging_q		     = 1,
 guv				     = 0.00005,
 gt				     = 0.00005,
 gq				     = 0.00005,
 fgdtzero                            = 1,
 if_no_pbl_nudging_ph                = 1,
 if_zfac_ph                          = 0,
 gph                                 = 0.00005,
 xwavenum                            = 5,
 ywavenum                            = 3,
 /

• All parameters are nudged using the spectral method with a time constant of 0.5 10⁻4 1/s
• The nudging only applies above the PBL.
• The spatial scales are about 1200km in longitude and 1300km in latitude.

• Why is q nudged ? : In fact the version 3.7 of WRF does not have the code for spectral nudging of q. This was only introduced by Spero in a later version. This can be verified in WRF/phys/module_fdda_spnudging.F. This code can only nudge the following variables

!-- u3d         3d u-velocity staggered on u points
!-- v3d         3d v-velocity staggered on v points
!-- th3d        3d potential temperature (k)
!---ph3d        3d perturbation geopotential

• Tests done to try and remove q nudging have demonstrated that this variable is not nudged. A priori a good thing also Spero et al. find some advantage to it. But these are all big-brother experiments while we are using ERA-I directly.

• Only 0.5 10⁻4 1/s relaxation time : This is a weaker relaxation that the 3.0 10⁻4 which is usually recommended. But as the model behaves well there is perhaps no need to nudge more strongly.