Custom Query (986 matches)

I forgot to add the logic that avoids replacing grid arrays with the _FillValue strategy introduced in ​src:ticket:222. The tindex argument is used to avoid overwriting such variables in masked areas.

In mod_scaras.F line 1570, replace

        SCALARS(ng) % IJwater((1:Lm(ng)+2)*(Mm(ng)+2),1:4) = 0

by

        SCALARS(ng) % IJwater(1:(Lm(ng)+2)*(Mm(ng)+2),1:4) = 0

The syntax of mp_scatter is modified by the READ_WATER option. Two calls to it in white_noise.F at lines 204 and 307 need to be corrected.

In wpoints,F on line 164, replace

# if defined READ_WATER || defined PROPAGATOR

by

# if defined READ_WATER || defined WRITE_WATER || defined PROPAGATOR

otherwise the mask is not defined.

In mod_param.F on line 159 replace

#if defined WRITE_WATER && defined MASKING

by

#if (defined WRITE_WATER || defined READ_WATER) && defined MASKING

The _FillValue logic introduced in ​src:ticket:222 broke when wetting and drying is activated. In this case, the bathymetry array becomes time-dependent and the tindex>0 and nVdim>2 logic to avoid masking no longer works. Recall that we cannot mask the bathymetry array in ROMS because h=0 in such areas and will give NaNs when dividing by h.

This implies that we need a more sophisticate logic to replace masked areas with _FillValue during IO. The less intrusive solution is put an optional argument (..., SetFillVal) to the nf_fwrite*d and def_var routines. Since optional arguments require an explicit interface, these routines were converted to modules. Therefore, several changes were needed in several of the IO routines.

This new logic is more elegant and allows for exception to land fill values in the future.