Inconsistent XLAND and ISLTYP

[lkugler]

Hi,

During the cold air outbreak in January, we had a crash in RUC LSM due to ISLTYP/XLAND being at inconsistent values (one indicating LAND, the other WATER).

According to ncks, the ISLTYP in the wrfinput_d01 file is 3 and xland=1, which is fine.
In SURFACE_DRIVER() we had xland=1 (land) and ISLTYP=14 (water) such that in LSMRUC() the subroutine SFCTMP() is called, which should never be called for ISLTYP=14. Finally there is a division by zero in SOILPROP(), such that the heat capacity becomes NaN.

To see where this comes from, we printed XLAND and ISLTYP for this gridpoint after reading wrfinput_d01 and wrfbdy_d01:

After reading wrfinput:

   wrf main: calling open_r_dataset for wrfinput
 after reading wrfinput
 xland(144,701), grid%isltyp(144,701)  0.0000000E+00           0

After reading wrfbdy_d01:

 d01 2019-01-19_12:00:00 in med_latbound_in preparing to read
 d01 2019-01-19_12:00:00  Opening: wrfbdy_d01
 after reading wrfbdy_d01
 xland(144,701), grid%isltyp(144,701)   1.000000              14

I assume that after reading wrfinput, the xland and isltyp are not already written to these variables.
After reading wrfbdy the variables are inconsistent.

Questions:

• Is it possible that the error is related to reading the wrfbdy file?

• Is it possible that it is an issue with indexing, i.e. MPI / OpenMP reading the wrong (i,j)? Or that the grid is shifted?

• How can we solve this? An additional IF to find inconsistencies in xland/isltyp at initialization time?

Best regards,
Lukas

 

[lkugler]

It seems that the issue is related to a shifted grid / wrong indices, as all the inconsistent gridpoints (land-points with xland=1) are one gridpoint to the north-east of a water gridpoint that is changed to sea-ice (due to the cold air outbreak), but it is still not clear why and what one can do to avoid that.

 

[Ming Chen]

We will take a look at this issue probably after Feb. 19. This is because we are on a business trip the following two weeks. Thanks for your patience.
Please keep us updated if there is more information related to this.