Closed FariborzDaneshvar-NOAA closed 5 months ago
Here some example results for Sandy | Sandy, 12hr leadtime | `Sandy, 72hr leadtime |
---|---|---|
Just for the sake of documentation: As we discussed this is due to an issue with setup (there's no track!). It stems from not having internet access. I'm looking into modifying the prep code to avoid the internet requirement.
As @SorooshMani-NOAA noted, the workflow was not using atmospheric forcing. Thanks to Soroosh for fixing this bug, I re-ran the workflow with updated prep.sif
image.
But still getting one eigenvalue!
Here are comparison of old and new tracks for some scenarios: | old tracks | new tracks |
---|---|---|
12-hr | 12-hr | |
24-hr | 24-hr | |
@SorooshMani-NOAA Here are pressure fields for some timesteps of the original track of Sandy with 12hr leadtime, suggesting that there is an issue with the hurricane model!
October 24, 2012, at 19:00
October 23, 2012, at 18:00
October 22, 2012, at 23:00
Path to the output on Hercules: /work2/noaa/nos-surge/shared/nhc_hurricanes/sandy_2012_OFCL_12hr_korobov_30/setup/ensemble.dir/runs/original/outputs/out2d_1.nc
@FariborzDaneshvar-NOAA I noticed I get a bunch of these warnings:
No nodal points found inside the radius of the last closed isobar (rrp = 0.000 m)
in a test run for Florence. Do you see the same/similar thing in your Sandy runs? It should be in outputs/mirror.out
. I'm not sure if that means we have to modify the track or there's something in the PaHM settings we need to fix.
Update
I checked your mirror.out
I see the same warning, but not for all timesteps. But still I see something that looks strange:
TIME STEP= 6623; TIME= 993450.000000
before GetGAHMFields
Working on time frame: 993600.000000000
Requested output time: 993600.000000000
, skipping generating data for this time; 0.000000000000000E+000
21600.0000000000 43200.0000000000 64800.0000000000
86400.0000000000 108000.000000000 129600.000000000
151200.000000000 172800.000000000 194400.000000000
216000.000000000 237600.000000000 259200.000000000
280800.000000000 302400.000000000 324000.000000000
345600.000000000 367200.000000000 388800.000000000
410400.000000000 432000.000000000 453600.000000000
475200.000000000 496800.000000000 518400.000000000
540000.000000000 561600.000000000 583200.000000000
604800.000000000 626400.000000000 637200.000000000
669600.000000000 712800.000000000 756000.000000000
799200.000000000 885600.000000000 972000.000000000
after GetGAHMFields
Is it skipping generating wind?! Is the issue the timestep settings? We need to ask @pvelissariou1
@FariborzDaneshvar-NOAA I think we have discussed this issue before. I sent the workaround for missing RRP fields by commenting out a couple of if blocks in the PaHM code contained in SCHISM. I think Soroosh implemented it. I'll push the updated modifications to SCHISM. If the hurricane is way outside the domain at the specific time you will get these warnings as wind fields are not calculated in these cases.
@pvelissariou1 I see, I just remembered you had it in a dev branch. I didn't realized it's still not in the main SCHISM code? When do you think you'll be able to merge the code?
In any case, I recompiled using the change I got from Takis into my schism fork. I'm now retesting the binaries on Hercules to make sure we get the wind correctly.
@SorooshMani-NOAA @pvelissariou1 Thanks for your efforts. The issue is been resolved! PCA is working as expected! Here are example results for Sandy with 48hr leadtime:
I will close this ticket then.
@FariborzDaneshvar-NOAA , @SorooshMani-NOAA I have updated PaHM in SCHISM and pushed the changes in the cmmb branch:
git clone https://github.com/schism-dev/schism.git -b cmmb
This branch is already synchronized with master and it is up to date.
Could you please try it and see if it compiles and runs as expected?
In my side I saw no issues.
After confirmation of successful runs, I will create a PR.
@pvelissariou1 thank you. I will recompile binaries using the cmmb branch and run to see what happens. If you don't have a separate ticket I'll close this again after testing your branch.
@pvelissariou1 I have issues compiling the cmmb
branch. I'll follow up with you on chat and then summarize findings here
The cmmb
branch can be compiled using BLD_STANDALONE=ON
option when running cmake.
The K-L expansion is returning only one eigenvalue for nodes from 30 ensembles. This is the case for all 6 leadtimes of three storms (Sandy, Michael, and Harvey) that I ran recently on Hercules! so there should be something with the workflow on Hercules. @SorooshMani-NOAA is there any part of the workflow on Hercules different from what we had on PW? what's your take on this?