HDF error with ERA5 forcing data in CLM5

[Xiaoyu yu]

Hello everyone,
I have replaced the default GSWP3 forcing data with other 3-hour forcing (CMFD) and run the compset I2000Clm50Sp sucessfully by using cesm2.2.0. However, some errors came up when I tried to replaced it with hourly ERA5 data.

It works with the following three steps:
(1) Setup the case using GSWP3 forcing and modify files;
(2) create a new domain file which is matched with atm forcing;
(3) modify the three stream files (attached below);
(4) make a regional domain (with 'frac' variable) and change the "LND_DOMAIN_FILE" and "ATM_DOMAIN_FILE" to run at reginal scale.

I only run two months (Jan and Feb in 1980) to test the forcing. After submitting the case, the errors in cesm.log file showed:

NetCDF: HDF error
 pio_support::pio_die:: myrank=          -1 : ERROR: ionf_mod.F90:         272 :
  NetCDF: HDF error

The processes in atm.log file showed:

(datm_comp_run) atm: model date 19800131   79200s
(shr_dmodel_readstrm) file ub: /scratch01/yuxy/inputdata/ERA5/Solar_test/clmforc.GSWP3.c2011.0.5x0.5.Solr.1980-01.nc     744
(shr_dmodel_readstrm) file ub: /scratch01/yuxy/inputdata/ERA5/Precip_test/clmforc.GSWP3.c2011.0.5x0.5.Prec.1980-01.nc     744
(shr_dmodel_readstrm) file ub: /scratch01/yuxy/inputdata/ERA5/TPHWL_test/clmforc.GSWP3.c2011.0.5x0.5.TPQWL.1980-01.nc     744
(datm_comp_run) atm: model date 19800131   81000s
(datm_comp_run) atm: model date 19800131   82800s
(shr_dmodel_readstrm) close  : /scratch01/yuxy/inputdata/ERA5/Solar_test/clmforc.GSWP3.c2011.0.5x0.5.Solr.1980-01.nc

It seemed that there was some errors in Solar stream of ERA5 forcing, but I have checked the data, and didnt find any errors in the value or spatial variation of Solar data. Is it because of reading data in hour step or the large amount of data? I don't know how to find the exact problems in the data. Does anyone know how to fix it?

Any suggestions will be appreciated! Thanks in advance.

 

[oleson]

I guess I would start by doing a ncdump -k on both of those forcing files (1980-01 and 1980-02) to make sure they are not NETCDF-4. Then make sure you get output when you do a ncdump -h on both files to make sure they are not corrupted.
I'd repeat that on your other stream files.

 

[Xiaoyu yu]

I guess I would start by doing a ncdump -k on both of those forcing files (1980-01 and 1980-02) to make sure they are not NETCDF-4. Then make sure you get output when you do a ncdump -h on both files to make sure they are not corrupted.
I'd repeat that on your other stream files.

Thanks a lot! I checked the data format. The format was NETCDF-4 classic model. I will change the format and make the forcing files again.

 

[Yuan Sun]

Great! I also made this mistake! Thank you for suggestions~