yantosca
commented
4 years ago Thanks for writing. I think this is because the dryrun code in hcoio_read_std_mod.F90 is a little simple-minded in that it lists the file for the given date as missing.
For example, on the AWS c;loud, we have these years of offline biogenics available:
$ s3ls s3://gcgrid/HEMCO/OFFLINE_BIOVOC/v2019-10/0.5x0.625/
PRE 2014/
PRE 2015/
PRE 2016/
PRE 2017/So there are offline BIOVOC files from 2014-2017 available. But the HEMCO code where we look for missing data files is in this permalink:
As you can see, at line 408, we exit the routine without going deeper into the HEMCO code to figure out what the closest date would be.
Perhaps this could be a feature request for the new HEMCO that will go into dev/13.0.0. Right now for some of these edge cases the dry-run might need to be augmented by manual download.
jennyfisher
pkasibhatla
Describe the bug
When running the GEOS-Chem dry run, it will usually correctly identify if (for example) OFFLINE_BIOVOC files are missing. However, if there are files present from before the target date (e.g. if I have downloaded files for Y2007 but not Y2008 or Y2009), then the dry run will reuse the earlier files rather than flagging that the target files are missing. If the cycling flag is changed from "C" to "EF" or "E", this still does not resolve the error - no new BIOVOC files are flagged.
To Reproduce
Include the steps that must be done in order to reproduce the observed behavior:
Preparation
LOCAL_BIOVOCLOCAL_BIOVOCRun commands
input.geosto run from 2008-01-01 to 2008-02-01./geos --dryrun > log.dryrunlog.dryrunYou should see that the old OFFLINE_BIOVOC files are being used, rather than the missing target files being flagged.
Expected behavior
The missing files should be flagged in the log.
Required information
Please include the following:
Input and log files to attach
log.dryrun.2008.txt The attached log file shows the run output - it correctly opens the 2008-07-01 file for BIOVOC (which is present on the system), but then starts looping over earlier years. For a clear example see lines 898-901; HEMCO is reading in data for 2008-07-27, but the lines for the four offline emissions read:
In this case, I'm using directories which are softlinked to all the correct ones on the Harvard Cannon cluster, but my own directories for
OFFLINE_BIOVOC(as the MERRA-2 biogenic VOC emissions aren't all available at Harvard). The key point is that the BIOVOC file is still20080701, whereas the others are all (correctly)2008-07-27.Additional context
This is a slightly difficult issue to resolve because sometimes we specifically want to loop earlier data. Perhaps for the dryrun option we should be expecting that fields with the
Rfield actually have requirements satisfied?Edit: tagging @msulprizio and @jimmielin as I think you'll both have insights on this particular issue!