This page documents the routine PDAF_gather_obs_f2_flex of PDAF, which was introduced with PDAF V1.14.

In the local filters (LESKTF, LETKF, LSEIK, LNETF) this routine can be used to gather the process-local observation coordinates arrays and returns the cooridates for the full observation vector.

The routine is typically used in the routines init_dim_obs_f_pdaf if the analysis step of the local filters is parallelized.

The interface is the following:

  SUBROUTINE PDAF_gather_obs_f2_flex(dim_obs_p, dim_obs_f, coords_obs_p, coords_obs_f, nrows, status)

with the following arguments:

  • dim_obs_p : integer, intent(in)
    Process-local dimension of observation vector
  • dim_obs_f : integer, intent(in)
    Full dimension of observation vector
  • coords_obs_p : real, intent(in), dimension(nrows, dim_obs_p)
    Process-local observation coordinates array
  • coords_obs_f : real, intent(out), dimension(nrows, dim_obs_f)
    Full observation coordinates array
  • nrows : integer, intent(in)
    Number of rows in observation coordinates array
  • status : integer, intent(out)
    Status flag (0 for no error)


  • The routine [Wiki:PDAF_gather_dim_obs_f] can be used to determine dim_obs_f.
  • This is more flexible variant to the routine PDAF_gather_obs_f2. It's functionality is the same, but it does not rely on a previous call to the routine PDAF_gather_dim_obs_f.
  • The routine is used for a coordinate array which stores in each column the coordinates of one observation. Thus for a two-dimensional case, the size of coords_obs_p is (2, dim_obs_p) and nrows=2 is used in the call to PDAF_gather_obs_f2.
  • If the coordindates are stored so that each row of coords_obs_p holds the coordinates of one observation the routine PDAF_gather_obs_f should be applied separate for each column to gather the full coordinates array.
Last modified 5 years ago Last modified on Jul 5, 2019, 10:14:56 AM