Hi all,
I’m trying to use SMOKE version 4.7 to process data from CAMS inventory (ECCAD – Emissions of atmospheric Compounds and Compilation of Ancillary Data), which has a resolution of 0.5 x 0.5 degree. For that, I created a new gridmask since the one available in the SMOKE directory (for EDGAR) has a resolution of 0.1 x 0.1 degree. To check if the new gridmask was okay, I used the m3xtract program that successfully read the file.
Noting that we haven’t processed this specific dataset, we notice that the grid information from the m3xtract of the grid mask and the met in grdmat show two grids with different dimensions and different names. The grid mask is called “LATLON_ECCAD0.5” and is 360 x 720, the met is called “RMGV_CROSS” and is 160 x 160. To fully debug the issue, the GRIDDESC file that defines both of these grids would be needed.
However, something you should be able to do is to load your grid mask file into VERDI or another tool that can visualize netCDF files. You can plot the GEOCODE variable with country boundaries to see if it is in the place you expected. I suspect that your coordinates are off and therefore the grids are found not to intersect.
Hi @eyth.alison,
Thanks for your response.
I’ve trying to process this dataset in the same way that I did with EDGAR, since both are global gridded data. So, I basically replaced the input emissions data, and created a new grid mask because of the coarser resolution of ECCAD data. I guess that is why the GRIDDESC names are different.
I plotted the grid masks on VERDI for both, EDGAR (from yours repository), and the one that I created.
These are them:
The output modeling domain through SMOKE do not have to be same as input domain (CAMS 0.5x0.5 grid cells) but GRIDMASK (GEOCODE and TIMEZONE) domain have to be matched with your new CAMS domain. You can define any output domain using SMOKE GRIDDESC input file. I would suggest for you to check the output domain from GRIDDESC and the MCIP domain. They have to be matched for Grdmat to generate the spatial gridding matrix.
After some manipulation, I was able to let both files (grid mask and emissions ECCAD) with same header and readable by M3XTRACT program. I appreciated the help, despite now I’m having a Segmentation fault, which stops during the temporal processing.
Looks like memory issue. It may require more RAM memory? You may set stacksize to unlimited in your run script for Temporal program to see whether that helps or not.
I also already tried to write it in terminal, but the segmentation faults still appeared.
which, the export file is:
module load gcc/7.3.0
export SMK_HOME=/opt/wrf_cmaq/smoke/4.7
export BIN=Linux2_x86_64gfort
export FC=gfortran
export IOAPIDIR=/opt/wrf_cmaq/ioapi
export LD_LIBRARY_PATH=/opt/openmpi/3.1.0/lib:/opt/wrf_cmaq/cmaq/libraries_7.3/netcdf/lib:/opt/wrf_cmaq/cmaq/libraries_7.3/ioapi/ioapi:$LD_LIBRARY_PATH
I would say that the problem is not related to a memory issue, since I was able to run successfully EDGAR inventory (which has a higher resolution than ECCAD) for the same area.
Then, I looked into the inventory stored in the $INVOPD directory, and the TZONES had strange values. However, I dont know from where they came from, since the grid mask did not have these values.
The left is the ECCAD inventory, and in the right the EDGAR.
You are right. Looks like you need to set the default TZONES value (like 0) for non-country area like ocean? To me, some random values got assigned to those cells that do not intersect with county boundaries.
Yes, this inventory that I’m trying to use has only emissions of CH2Br2, CH3I, CHBr3, DMS, assigned to the ocean, exclusively. Thanks for helping BH, Liz and Alison.
Hi. I worked with CAMs dataset and used as an Input in SMOKE. I regridded the CAMS-OCEAN from 0.5 to 0.1 and then, I used some preprocessing steps to get the desired file format as input in SMOKE. If you send me an email (ernesto.pino@pucvl.) I can send you the details. I can´t post here now because it is part of a manuscript not yet published.
