troubles, could not find trapping x locations - NOT solved by sfcp_to_sfcp = .true.

[alexjt]

I am running a 4km WRF4.6.1 simulation over the CONUS and am inconsistently running into the error "troubles, could not find trapping x locations" when running real.exe. I am running several days of the model and some days encounter this error, while other days do not. The error popped up when I specified sst_update and sst_skin equal to 1 in namelist.input.

I have run two distinct days of the simulation: 1) 2036-09-16 ; this day runs real.exe and wrf.exe successfully with no errors, and 2) 2036-09-17 ; the next day encounters the "troubles" error when running real.exe. Both days use the exact same namelist and input data source, so the only difference is the actual values of the input data.

These simulations are downscaled from CESM2 Large Ensemble model output that is converted to WRF intermediate files. I have confirmed that there are no NaN values anywhere within the met_em*.nc files generated from WPS.

The error message I get when running real.exe for 2036-09-17 is as follows:

rsl.out.0038:

Using sfcprs3 to compute psfc
i,j = 1131 453
target pressure and value = NaN 6061.74316
column of pressure and value = 11.5168056 0.00000000
column of pressure and value = 11.5080614 295.771973
column of pressure and value = 11.4826183 295.789612
column of pressure and value = 11.4569368 521.401306
column of pressure and value = 11.4305544 752.369751
column of pressure and value = 11.4034300 988.765503
column of pressure and value = 11.3467875 1477.94885
column of pressure and value = 11.2866344 1992.12805
column of pressure and value = 11.2225304 2532.39453
column of pressure and value = 11.1539345 3106.60010
column of pressure and value = 11.0801859 3715.09521
column of pressure and value = 11.0004826 4362.87549
column of pressure and value = 10.9138222 5067.46338
column of pressure and value = 10.8188562 5820.38672
column of pressure and value = 10.7137947 6642.01660
column of pressure and value = 10.5962372 7536.76270
column of pressure and value = 10.4628706 8524.86914
column of pressure and value = 10.3088398 9634.39355
column of pressure and value = 10.1265898 10900.2695
column of pressure and value = 9.90347576 12375.1445
column of pressure and value = 9.61580086 14178.9180
column of pressure and value = 9.21033764 16649.3945
column of pressure and value = 8.85366249 18838.6953
column of pressure and value = 8.51718998 20941.1250
column of pressure and value = 8.00636482 24169.8359
column of pressure and value = 7.60090065 26782.6387
column of pressure and value = 6.90775537 31343.9180
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 6302
troubles, could not find trapping x locations
-------------------------------------------

I tried adding sfcp_to_sfcp = .true. to my namelist.input and this fixed the "troubles" error in real.exe and caused real.exe to use "Using sfcprs2 to compute psfc". However, when I subsequently ran wrf.exe, the simulation failed with this error:

rsl.out.0770:

emitted longwave <0; skin T may be wrong due to inconsistent
input of SHDFAC with LAI
375 317 SHDFAC= 0.356190503 VAI= 1.38770485 TV= 127.578300 TG= -495.974091
LWDN= 1118.19128 FIRA= -14036.8301 SNOWH= 0.00000000
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 2327
STOP in Noah-MP
-------------------------------------------

I am unsure how to solve this set of errors. It appears to me that the fortran module sfcprs3 is producing a NaN value from my input data, even though there are no NaNs in my input data to begin with.

Any insight is greatly appreciated!

 

[kwerner]

Apologies for the long delay in response while our team tended to time-sensitive obligations. Thank you for your patience. Since it's been a while, I'd like to first ask if you are still experiencing this issue. If so, is there any updated information you want to share? Thanks.

 

[alexjt]

Thanks for the response. I am pretty sure I solved this last week but have some more testing to do before I confirm that the issue is resolved.

For reference, I tried many different tests to try to understand this error. It turns out that it was likely an issue in the pressure profile of geopotential height and below surface values and how real.exe interpolated these values at high spatial resolution. I tried many tests to solve the problem including specifying various namelist options, using bias corrected and non-bias corrected CESM2 data, specifying different nodes/processors for real.exe, but none of this resolved the issue. I did find out that coarsening the domain from 4km to 25km resolved the issue, but was not a suitable option for my simulations.

What ended up resolving the problem (I think) was modifying METGRID.TBL at name=GHT and I changed "interp_option=sixteen_pt+four_pt+average_4pt" to "interp_option=four_pt+average_4pt". I believe that the interpolation with sixteen_pt was producing NaN values in the GHT interpolation within real.exe.

I will provide any further updates if they arise from future testing.