Bug fix for HRLDAS urban array allocation with SF_URBAN_PHYSICS 2 and 3

[sashanksilwal]

In the HRLDAS driver's NoahmpIOVarInitMod.F90, urban arrays (e.g. TR_URB2D, QC_URB2D, DZR, DZB, DZG, TRL_URB3D) were only allocated and initialized inside the SF_URBAN_PHYSICS == 1 (SLUCM) block. These arrays are passed as non-optional arguments to urban_param_init, urban_var_init, and noahmp_urban, all called when SF_URBAN_PHYSICS > 0.

Issue I faced: Using BEP (SF_URBAN_PHYSICS=2) or BEM (SF_URBAN_PHYSICS=3) causes a segmentation fault due to unallocated arrays

Fix: remove the SF_URBAN_PHYSICS == 1 guards so these arrays are allocated/initialized for all urban modes. The WRF driver is unaffected as its equivalent section is disabled (if (0 == 1))

[tslin2]

will also need to update HRLDAS driver codes

[cenlinhe]

Thank you for identifying this issue. Instead of entirely removing the if-statement for SF_URBAN_PHYSICS = 1, I prefer identifying the variables shared by SF_URBAN_PHYSICS = 1-3 and then keep the if-statement for SF_URBAN_PHYSICS = 1 for those variables only used by SLUCM.

[tslin2]

@sashanksilwal have you tested it? I am still getting error

I suggest we can revise codes in the module_sf_urban.f90 but not here, @cenlinhe how do you think

Because the issue happen in urban_param_init, urban_var_init, and noahmp_urban,

[cenlinhe]

@tslin2 I do not suggest modifying module_sf_urban.F as these files should be as close as the version inside WRF model, so I would avoid additional code changes within module_sf_urban.F if it can be fixed on the noahmp part. However, if it cannot be fixed from the noahmp urban driver part only and it is indeed a bug inside module_sf_urban.F, then we can modify module_sf_urban.F.

[tslin2]

@tslin2 I do not suggest modifying module_sf_urban.F as these files should be as close as the version inside WRF model, so I would avoid additional code changes within module_sf_urban.F if it can be fixed on the noahmp part. However, if it cannot be fixed from the noahmp urban driver part only and it is indeed a bug inside module_sf_urban.F, then we can modify module_sf_urban.F.

For example there is the issue below:
when running urban 2 or 3, QC_URB2D is not allocated,
https://github.com/NCAR/hrldas/blob/0dc57ea6cda02f89ca0eb8634ffcb3c5b55476a0/urban/wrf/module_sf_urban.F#L2822

Either change to IF (sf_urban_physics == 1) QC_URB2D(I,J)=0.01 for this line in module_sf_urban.F

or need to allocate QC_URB2D in NoahmpIOVarInitMod.F90

Another example is https://github.com/NCAR/hrldas/blob/0dc57ea6cda02f89ca0eb8634ffcb3c5b55476a0/urban/wrf/module_sf_urban.F#L2826-L2840

  XXXR_URB2D(I,J)=0.
  XXXB_URB2D(I,J)=0.
  XXXG_URB2D(I,J)=0.
  XXXC_URB2D(I,J)=0.

can move to the loop inside IF ( sf_urban_physics == 1 ) THEN

or allocate these four variables in NoahmpIOVarInitMod.F90

I think both method can fix this issue.

[cenlinhe]

OK, I see. I would recommend allocating those variables in NoahmpIOVarInitMod.F90

[tslin2]

I submit a PR to fix this at #243

[cenlinhe]

OK, I am closing this PR since the new PR #243 is submitted.