Dear all,
I experienced a bug in TSLB_MOSAIC, SMOIS_MOSAIC and SH2O_MOSAIC output when using the option sf_surface = 1 with Noah LSM and mosaic_cat=3.
The figure "TSLB_MOSAIC_t0_mosaic0" shows a 2D color plot of TSLB_MOSAIC 45 minutes after the simulation start time, in the first mosaic layer int front of Longitude and Latitude. We can see that the domain is divided in 3 bands of realistic values separated by dark bands. The dark band has a temperature of 0 K (I forced the minimum color at 272 K to show that other values seem realistic).
Here are some notes about the bug :
- I observed the same bug in all mosaic layers of TSLB_MOSAIC, SMOIS_MOSAIC and SH2O_MOSAIC, at all output time, and for my 2 domains (parent + nest).
- surface variables like TSK_MOSAIC, ALBEDO_MOSAIC, EMISS_MOSAIC are not affected by this bug
- I ran here with 9 CPUS. When I tested a run with 10 CPUS, I had 5 bands instead of 3, indicating that the bug is linked to mpi.
- It is quite hard to notice but it seems that the "realistic layers" is flatten in the north-south direction. The missing values of dark bands are not really missing but we should stretch the "realistic layers" to get them back to normal.
- I added the MOSAIC variables to wrfout files just by adding "h" in Registry.EM_COMMON file
- The TSLB output value is not affected by the bug (shown in figure "TSLB_t0_soil0")
- I also tried with mosaic_cat=3 and had the same error
- I used the last version of the code (downloaded 04/07/23) and I compiled with configuration option 34 (parallel distribute memory GNU), I already made simulations with that option without facing any issue on other variables
I didn't find another related issue, is that already known ? I join my namelist.input and namelist.wps files.
Mathieu
I experienced a bug in TSLB_MOSAIC, SMOIS_MOSAIC and SH2O_MOSAIC output when using the option sf_surface = 1 with Noah LSM and mosaic_cat=3.
The figure "TSLB_MOSAIC_t0_mosaic0" shows a 2D color plot of TSLB_MOSAIC 45 minutes after the simulation start time, in the first mosaic layer int front of Longitude and Latitude. We can see that the domain is divided in 3 bands of realistic values separated by dark bands. The dark band has a temperature of 0 K (I forced the minimum color at 272 K to show that other values seem realistic).
Here are some notes about the bug :
- I observed the same bug in all mosaic layers of TSLB_MOSAIC, SMOIS_MOSAIC and SH2O_MOSAIC, at all output time, and for my 2 domains (parent + nest).
- surface variables like TSK_MOSAIC, ALBEDO_MOSAIC, EMISS_MOSAIC are not affected by this bug
- I ran here with 9 CPUS. When I tested a run with 10 CPUS, I had 5 bands instead of 3, indicating that the bug is linked to mpi.
- It is quite hard to notice but it seems that the "realistic layers" is flatten in the north-south direction. The missing values of dark bands are not really missing but we should stretch the "realistic layers" to get them back to normal.
- I added the MOSAIC variables to wrfout files just by adding "h" in Registry.EM_COMMON file
- The TSLB output value is not affected by the bug (shown in figure "TSLB_t0_soil0")
- I also tried with mosaic_cat=3 and had the same error
- I used the last version of the code (downloaded 04/07/23) and I compiled with configuration option 34 (parallel distribute memory GNU), I already made simulations with that option without facing any issue on other variables
I didn't find another related issue, is that already known ? I join my namelist.input and namelist.wps files.
Mathieu