WRF real.exe "grid%tsk unreasonable" error after Landuse preprocessing

[aba0327]

Hi, I'm working on customizing Landuse data for WRF and encountered an issue when running real.exe. The model stops with the error message:

grid%tsk unreasonable

I suspect the issue is related to Landuse preprocessing. Here are the steps I took to integrate the land cover data:

1. Custom Landuse Data Source:
   • I used EGIS land cover data, which I converted to binary (.bil) format.
   • And in the bin file, I confirmed that the land cover code was properly mapped.
2. Resolution & Projection: The dataset was projected to EPSG:4326 with a resolution of 0.0041666667.
3. Preprocessing Workflow:
   • Used rasterio and rasterize() to convert shapefiles to a raster grid.
   • Assigned 1 to all no-data areas.
   • Processed GEOGRID.TBL with the following settings:
     

     name=LANDUSEF
     priority = 1
     dest_type = categorical
     z_dim_name = land_cat
     dominant = LU_INDEX
     interp_option = egis_landuse:nearest_neighbor
     rel_path = egis_landuse:egis_landuse/

4. Checking geo_em Files
   • Used ncdump -v LU_INDEX geo_em.d01.nc | less to check LU_INDEX values.
   • Found that all values were 17, which seems incorrect.
5. Running real.exe
   • real.exe runs but fails with "grid%tsk unreasonable" error.
   • ncdump -v TSK met_em.d01.2023-06-01_12:00:00.nc shows no TSK variable.

Questions:

1. What causes geo_em to be created with only one code of 17?
2. Could this issue be caused by incorrect Landuse classification mapping?
3. Any recommended debugging steps for Landuse-related real.exe errors?

Any insights would be greatly appreciated. Thanks!

 

[dbpartha]

Hi, I faced the same issue of grid%tsk unreasonable and here's how I solved it,

# it's possible that TSK variable is not present in your met_em* (hence metoa*) files as MET data (NARR or GFS) uses SKINTEMP instead of TSK (I might be completely wrong). Check if you have TSK variable in either metoa* or met* files

>> ncdump -v TSK metoa_em.d01.2023-06-21_00:00:00.nc
ncdump: TSK: No such variable

>> ncdump -v TSK met_em.d01.2023-06-21_00:00:00.nc
ncdump: TSK: No such variable

TRIAL: Try to find the variable that represents skin temperature of earth in your met or metoa* (if using nudging) files and check if the range of the variable values are reasonable (e.g., 270–320 K). There’s SKINTEMP variable in the NARR metoa* files, renaming it to TSK in all files

>> module purge all
>> module load nco
>> module load hdf5            #(you may need other modules as well, load as required)

#rename SKINTEMP variable in metoa* (in my case) netcdf files, directly copy paste these lines on command line

>> for file in metoa_em.d0*.nc; do
>> ncrename -O -v SKINTEMP,TSK "$file"
>> done

#check the last file to ensure SKINTEMP is renamed as TSK

>> ncdump -v TSK metoa_em.d03.2023-09-02_00:00:00.nc

# will print out TSK values

#Run REAL.EXE

>> cd ~/WRF-4.4.2/run

>> rm metoa*                        # will only remove the links but doesn’t hurt to start fresh

>> ln -s /projects/b1045/wrf-cmaq/OBSGRID/20230708/metoa* .

>> ./real.exe

 

[aba0327]

Thank you, and I'll try it.

 

[jchiquetto]

Hello, thank you for this thread, I am having a simillar issue!

I am currently trying to run WRF in AWS (Amazon Web Services) and getting the same error message ("grid%tsk unreasonable").

I realize this could be due to landuse data errors. I am using ERA5 input data (both surface and pressure levels), I have already downloaded the ERA5-Invariant file and placed it in the WPS directory, and linked to the ECMWF vtable before running ungrib. I have run ungrib and metgrid successfully, although in metgrid it shows this warning ("Note: The following floating-point exceptions are signalling: IEEE_OVERFLOW_FLAG IEEE_UNDERFLOW_FLAG IEEE_DENORMAL") along with the successful completion message (possibly not an issue).

The met_em files seem to be ok ,but there is no variable named TSK, only SKINTEMP. I tried following your steps for changing the variable name in the met_em files. But I got the following:
>> module purge all - no message (presumably worked)
>> module load nco and >> module load hdf5 (didn't recognize any of these modules)

Then when I tried
>> for file in metoa_em.d0*.nc; do
>> ncrename -O -v SKINTEMP,TSK "$file"
>> done

There was a message saying it didn't recognize the command ncrename - I assume because the nco and hdf5 modules above didn't load correctly.
Just for the record, I am running WRF v. 4.3.3, WPS v. 4.3.1.

If you you have any ideas on how to work around this issue, or where I can get some help (for example, how do I obtain the nco and hdf5 modules), I'd greatly appreciate it!

Thanks a lot and kind regards,
Julio

 

[dbpartha]

Hi Julio, you may not want to change the variable name “SKINTEMP” to “TSK” as large number of newer WRF model variables may depend on it. My WRF4.4.2 crashed with a RIBX never exceeds RIC error (see here: RIBX never exceeds RIC error in WRF4.4.2). I was using NARR data and found that grid%tsk unreasonable is an issue with the recent NARR data (and probably for ERA5 too, seen some posts on it). I am still figuring this out and switched to NAM instead of NARR. Let me know if you can figure this issue out without changing the datasets.

 

[jchiquetto]

Hi Dbpartha, Thank you for the reply. I will keep this in mind, maybe change the datasets.
Well, the first one of us whom finds a solution, make sure to come back here and let the others know, haha.
Kind regards,
Julio