Segmentation Fault using UCM

[einaraz]

Hi everyone,

I recently started running WRF on Cheyenne. To get started, I'm comparing the different urban physics models. I have a nested domain (9, 3, 1) km and I am using NARR and NLCD2011 for land cover. My three-day simulation finishes when I use Noah LSM + BEP or BEM. However, I keep running into errors when I try UCM (sf_urban_physics=1). The first issue was regarding the height of the first grid level, which seems to be solved after I increase dzbot. However, now my simulation crashes after a few hours (the time it crashes varies for 3-nested domain or single domain). I'm using almost all default settings, but I did try changing lw/sw schemes to no avail (I think the problem is related to the surface energy budget).
The details of my simulation and the rsl files are located here /glade/scratch/einaraz/WRFcomp/Simuls/NLCD2011-UCM
Do you have any suggestions to fix this problem?

Any help would be much appreciated!
Thank you!

 

[Ming Chen]

cheyenne is down this week. Can you clarify what errors you saw in your rsl files?

 

[einaraz]

Hi Ming,
Thank you for your reply. I attached here my namelist.input and wps, as well as the rsl file with the error message (rsl.error.0021).

 

[Ming Chen]

Something in the physics went wrong in this case. In the rsl.error.0021, we can find some unreasonable value like:

f_qv=4294967295, f_qc=4294967295, f_qr=4294967295,

In your namelist.input, everything is fine except the option

use_wudapt_lcz

If you run with this option, then I suppose you should also use the wudapt urban land use type. I don't think this option works with NLCD.

 

[einaraz]

Hi Ming,

I tried running the simulation without the option use_wudapt_lcz=1, and apparently it worked, thank you!

At first I was running the pre-compiled version available on Cheyenne (the most recent). I was first getting an error related to the VEGPARM.TBL, which I replaced after reading a similar question in the forum. After fixing that, I got the following error message

USING 10 WUDAPT LCZ WITHOUT URBPARM_LCZ.TBL. SET USE_WUDAPT_LCZ=1

That is why I changed that option. I would then get segmentation fault after a few hours of simulation.

When I run the same simulation using the code I compiled, I see no error messages and I am able to finish a three-day simulation using the UCM option. However, now I'm having issues when I try to activate the mosaic option (simulation here /glade/scratch/einaraz/WRFmycomp/WRF/run). I tried both NLCD 2006 and 2011 and the same problem happens, with the simulation stopping at minute 9 (same value as radt) without a specific error message. I was not able to find more information about the mosaic option in the user's guide, so I was wondering if the problem could be the combination of physics schemes or the data that I'm using.

Thank you!

 

[Ming Chen]

The mosaic option is only used for diagnostic and it only works for Noah LSM. It is well tested with MODIS/USGS as input, and I am not sure whether there is any issues related to NLCD data.

 

[Benjamin_Obe]

Hello Ming and einaraz,

I got the same error with my SLUCM (sf_urban_physics=1) with LCZ and wudapt option set to 1., my dzbot is set at 50

01 2020-03-01_06:00:00 Input data is acceptable to use:
d01 2020-03-01_07:00:00 Input data is acceptable to use:
d01 2020-03-01_08:00:00 Input data is acceptable to use:
[sonic49:23280:0:23280] Caught signal 11 (Segmentation fault: address not mapped to object at address 0xfffffffe080755e0)

I have been on this without success for 2 months. However, I have got no issues with BEP and BEP+BEM. any clue will be greatly appreciated... my namelist.input is attached

 

[Ming Chen]

I am suspicious that the mosaic option may have some issues related to LANDUSEF, although I am not 100% sure. Please avoid using this option at present.

 

[einaraz]

Hi everyone,

I just wanted to give an update on this issue.
I was able to finish the simulation using the same setup (NLCD data, UCM + mosaic) using an older version of WRF (v3.9.1.1). I haven't tried other versions, but it seems that the mosaic option has issues at least in the most recent version (v4.4).

Best,