Fix runtime error when using sf_surface_mosaic = 1 with use_wudapt_lcz = 0

[epn09]

TYPE: bug fix

KEYWORDS: noah mosaic, wudapt

SOURCE: Do Ngoc Khanh (Tokyo Institute of Technology)

DESCRIPTION OF CHANGES:
Fix runtime error when using sf_surface_mosaic = 1 with use_wudapt_lcz = 0.

Problem:
Segmentation fault occurs when running the model using sf_surface_mosaic = 1 with use_wudapt_lcz = 0 as described in detail in #1633.

Solution:
The code now checks for use_wudapt_lcz and uses different code to define urban categories in lsm_mosaic routine in module_sf_noahdrv.F.

ISSUE:
Fixes #1633

LIST OF MODIFIED FILES:
M dyn_em/module_first_rk_step_part1.F
M phys/module_sf_noahdrv.F
M phys/module_surface_driver.F

TESTS CONDUCTED:

1. When use_wudapt_lcz = 1: bit-by-bit identical output before and after modification.
2. When use_wudapt_lcz = 0: Segmentation fault is fixed.
3. The Jenkins tests are all passing.

RELEASE NOTE: Fix runtime error when using sf_surface_mosaic = 1 with use_wudapt_lcz = 0.

[weiwangncar]

@epn09 Thank you for making this PR. Have you looked how the non-mosaic version of Noah LSM handles the urban categories? That code did not need to check the namelist option use_wudapt_lcz and it appears to work fine.

[epn09]

@weiwangncar In the non-mosaic version of Noah LSM, IVGTYP is constant and assignment of FRC_URB2D is done in urban_var_init in module_sf_urban.F (https://github.com/wrf-model/WRF/blob/master/phys/module_sf_urban.F#L2653-L2686) and use_wudapt_lcz is checked there.

[weiwangncar]

@epn09 This probably makes sense. Thanks. Would you be interested in testing an urban option using mosaic Noah LSM?

[epn09]

@weiwangncar I have run the mosaic Noah LSM with single layer UCM urban option (using the bug-fixed code) for a 3-day run. Comparing with the run without mosaic option, the result looks reasonable to me.

[weiwangncar]

@epn09 Thank you for doing that test.

The base branch was changed.

[davegill]

@weiwangncar
Wei,
I swapped the base branch from master to develop, and that means github tossed out your review. Would you do your review again please?

[pvahmanilbl]

@davegill and @weiwangncar I hope this is the right place to raise this issue: I have been getting segmentation fault error:
Program received signal SIGSEGV: Segmentation fault - invalid memory reference
when I use
SF_SURFACE_MOSAIC = 1
This happens when the radiation scheme is called for the first time (6 min in the simulation). I checked and the fix above is already in the WRFv44 I am using.

I am using
use_wudapt_lcz=1
num_land_cat=61
sf_urban_physics=1
When I remove SF_SURFACE_MOSAIC=1 there are no issues. I would really appreciate it if you can help me with this!

[kkeene44]

@pvahmanilbl
The subject for this PR is specific to when use_wudapt_lcz=0. I just want to check that you're actually using =1.

[pvahmanilbl]

@kkeene44 I am using use_wudapt_lcz=1. I have not tried it use_wudapt_lcz=0. My configuration works with non-mosaic (SF_SURFACE_MOSAIC=0) option but when I use mosaic (SF_SURFACE_MOSAIC=1) it crashes with seg fault. I posted here since this is the closest I found to the issue I am facing.

[kkeene44]

@pvahmanilbl
I ran a test case with those settings and I'm able to run without problems for a case using sf_surface_mosaic=1 and use_wudapt_lcz=1. This must be related to your specific case. Can you post your issue to the WRF & MPAS-A Support Forum so we can look into the problem there? When you do, please attach your namelist.input file, and if you'd like to share larger files (e.g., wrfinput, wrfbdy), take a look at the home page of the forum for instructions on sharing large files. Thanks!

[pvahmanilbl]

@kkeene44 thank you so much for your response! (sorry I am being slow; I was on paternity leave)
I will try the Forum one more time but can I ask a quick question? when you run real.exe with the mosaic option do you get the new variables (e.g., LANDUSEF2, MOSAIC_CAT_INDEX, etc) in wrfinput with values in them? for me, all of these new variables in wrfinput have only zeros.

[kkeene44]

@pvahmanilbl
Mine are also all 0, but if I'm reading the code correctly, the new variables are just initialized during the real.exe process. They are later used during WRF to update other variables (for e.g., the values for mosaic_cat_index become IVGTYP.

[pvahmanilbl]

@kkeene44 thanks so much for your response! I was trying to use CONUS physics and it didn't work with MOSAIC until I used:

sf_sfclay_physics = 1, 1, 1, 1,
bl_pbl_physics = 1, 1, 1, 1,
and it worked!