Fix unit conversion logic #579
Conversation
There was a problem hiding this comment.
These changes match #521 . I approve these changes with successful testing of the impacted plotting components.
jevs_cam_grid2obs_plotsLog File: /lfs/h2/emc/vpppg/noscrub/mallory.row/verification/EVS_PRs/pr579/EVS/dev/drivers/scripts/plots/cam/jevs_cam_grid2obs_plots.o157072007 |
jevs_mesoscale_grid2obs_plotsLog File: /lfs/h2/emc/vpppg/noscrub/mallory.row/verification/EVS_PRs/pr579/EVS/dev/drivers/scripts/plots/mesoscale/jevs_mesoscale_grid2obs_plots.o157073250 Tagging @PerryShafran-NOAA for awareness |
jevs_global_ens_atmos_gefs_grid2obs_past31days_plotsLog File: /lfs/h2/emc/vpppg/noscrub/mallory.row/verification/EVS_PRs/pr579/EVS/dev/drivers/scripts/plots/global_ens/jevs_global_ens_atmos_gefs_grid2obs_past31days_plots.o157073480 Tagging @GwenChen-NOAA for awareness |
|
jevs_aqm_grid2obs_plotsLog File: /lfs/h2/emc/vpppg/noscrub/mallory.row/verification/EVS_PRs/pr579/EVS/dev/drivers/scripts/plots/aqm/jevs_aqm_grid2obs_plots.o157074109 Tagging @Ho-ChunHuang-NOAA for awareness |
|
The following completed cleanly and output looks good: The following failed but seems to just need COMIN corrected:
Still running: |
Rerun with correct COMIN jevs_analyses_grid2obs_plotsLog File: /lfs/h2/emc/vpppg/noscrub/mallory.row/verification/EVS_PRs/pr579/EVS/dev/drivers/scripts/plots/analyses/jevs_analyses_grid2obs_plots.o157075692 Tagging @PerryShafran-NOAA for awareness |
Plots look good to me. |
|
The AQM plots looks good to me https://www.emc.ncep.noaa.gov/users/verification/air_quality/aqm/pr/grid2obs/pm25/ |
|
Thanks @Ho-ChunHuang-NOAA! ⏳ jevs_cam_grid2obs_plots |
|
@MarcelCaron-NOAA @Ho-ChunHuang-NOAA Thanks! Can you please remind us how this particular PR should impact AQM? I was thinking that this PR primarily fixed temperature and dewpoint units on plots. Is there an AQM plot that should have changed? |
@AliciaBentley-NOAA That's right, no aqm plot should change. The changes in this PR affect unit conversion, so plots with no unit conversion will look the same. It also depends on the metric shown; metrics computed with ffbar, oobar, or fobar are affected (e.g., rmse), but metrics computed with only fbar and obar are not (e.g., bias). Even though unit conversion is not currently used in aqm plots, the same faulty unit conversion logic exists and is corrected here. |
|
It should be the exact same thing as ops, but a script was edited so good to test to make sure it didn't introduced anything unwanted. |
|
That makes sense. I found that the test plots and the prod plots for AQM looked exactly the same and was I worried that I was missing something (since the plots have been changing in most of the other unit PRs). Good to know that this component needed to look the same to be correct. Thank you! |
|
Yes! Those plots should not change. I updated the PR instructions to include that info. Thanks! |
|
The following completed cleanly and output look normal: ⏳ jevs_cam_grid2obs_plots |
|
The following completed cleanly and output look normal: ⏳ jevs_cam_grid2obs_plots |
|
@AliciaBentley-NOAA @MarcelCaron-NOAA AQM plots use Marcel's code as the core modules. AQM uses only fbar and obar. But it would be good to sync the updated code with Marcel 's, and thank you Marcel for doing the lift. |
|
@malloryprow The following completed cleanly and output look normal: 👍 All of the test jobs have completed successfully |
|
Thanks for your work on this @MarcelCaron-NOAA! |
Note to developers: You must use this PR template!
Description of Changes
This fix matches the fix made in PR #521. It addresses a logic error affecting plotting output for five EVS components. The details are reprinted here.
The logic error was found in several instances of code used for converting the units of certain METplus summary statistics. Values of these stats were already converted before being used in follow-up conversion equations (incorrect). As a result, some output plots were displaying incorrect "doubly converted" values of verification metrics.
In this fix, METplus summary statistics are first stored as new variables prior to conversion, and these new variables are used when performing unit conversion and are never overwritten. As a result, all EVS plots affected by the issue should display correct converted values of verification metrics.
Developer Questions and Checklist
Yes. Currently, some EVS plots are inaccurate if they display SL1L2- or VL1L2-based metrics and apply unit conversion. The issue affects five EVS components.
No. My work schedule may be impacted by the Hurricane Helene recovery timeline (will edit when this is no longer the case).
No.
${USER}where necessary throughout the code.HOMEevsare removed from the code.dev/drivers/scriptsordev/modulefileshave been made in the correspondingecf/scriptsandecf/defs/evs-nco.def?Testing Instructions
(1) Set up jobs
HOMEevs - set to your test EVS directory
COMIN - set to /lfs/h1/ops/prod/com/$NET/$evs_ver_2d
COMOUT - set to your test output directory
SENDMAIL - (optional) set to "NO"
(2) Running jobs
[Total: 5 plots jobs]
qsub -v vhr=00 $driver(3) Checking jobs
Log files should be checked by the developer for the following keywords:
Additionally, a sample of test output plots should be compared to EVS prod output.
Note: The changes in this PR affect unit conversion, so plots with no unit conversion will look the same. Whether or not a plot is affected also depends on the metric shown; metrics computed with ffbar, oobar, or fobar are affected (e.g., rmse), but metrics computed with only fbar and obar are not (e.g., bias).