Add script to plot online MOC #415

milenaveneziani · 2018-06-01T21:25:45Z

This is very 'work in progress' for now. I just wanted to open this for discussion of some points.

milenaveneziani · 2018-06-01T21:27:42Z

mpas_analysis/ocean/streamfunction_moc.py

-                    ['timeMonthly_avg_normalGMBolusVelocity',
-                     'timeMonthly_avg_vertGMBolusVelocityTop'])
+        if self.mocAnalysisMemberEnabled:
+            # I think the bolus velocity should always be included if GM is on


I think the following should be changed so that it's no longer an option. If GM is on, we need to include the bolus velocity part.

milenaveneziani · 2018-06-01T21:29:38Z

mpas_analysis/ocean/streamfunction_moc.py

+        config = self.config
+
+#*********************
+# Needs to be edited from here on


The following needs to be edited substantially, but we need to keep the option to read from an already computed climatology. I am still figuring out how I do that, hence no much change can be found below :|

I can help here if you can describe more specifically what is needed.

milenaveneziani · 2018-06-01T21:31:29Z

mpas_analysis/ocean/streamfunction_moc.py

+                                        np.amax(latBins)+latBinSize,
+                                        latBinSize)
+                mocTop = self._compute_moc(latBins, nVertLevels, latCell,
+                                           regionCellMask, transportZ, velArea)


The following 3 lines is the bit to compute the MOC at 26.5. I need to put this into a region loop, keeping in mind that MOC online has two separate arrays: one for the Global bit and one for the regional bit. At the moment, the latter is for the Atlantic, if indRegion=0.

milenaveneziani · 2018-06-01T21:32:53Z

mpas_analysis/ocean/streamfunction_moc.py

+        # }}}
+
+    def _compute_moc_time_series_analysismember(self):  # {{{
+


This is mostly done, but we probably do not need to save the timeseries to file this time, since it's a super simple calculation once we have the online MOC.
Also, there are lots of repetitions with the _compute_time_series_postprocess function.

Please do make sure to save the time series data. The main reason is that it makes it possible to use it as a reference data set when doing a main vs. ref comparison. In the not-too-distant future, we would want to do the same with the MOC lat vs. depth plots, too, so we will also need to store a copy of them.

milenaveneziani · 2018-06-01T21:33:56Z

mpas_analysis/ocean/streamfunction_moc.py

+            self.logger.info('     date: {:04d}-{:02d}'.format(date.year,
+                                                               date.month))
+
+            # hard-wire region=0 (Atlantic) for now


This whole region handling will change substantially once we use the regional capability fully.

Yes, see #246

milenaveneziani · 2018-06-04T20:11:52Z

mpas_analysis/ocean/streamfunction_moc.py

+            depth = ncFile.createVariable('depth', 'f4', ('nz',))
+            depth.description = 'depth'
+            depth.units = 'meters'
+            depth[:] = self.depth


@xylar: I think I took care of most of the changes needed here.
Note the change on the line above depth[:] = self.depth, with respect to the _compute_moc_climo_postprocess respective line. I think we should change it in there too, but I'd like to double check with you first.

I went ahead and made this change in _compute_moc_climo_postprocess as well.

Yep, seems fine to me. We might want to think about how this would change for a non-z-level grid in the future but that's a separate issue and there's already a GitHub issue for it (#120)

milenaveneziani · 2018-06-04T20:12:11Z

I will do some testing now.

milenaveneziani · 2018-06-04T21:01:41Z

This is not working. It hangs, I think at the climatology task, and no log file is even created.
Looking into it.

milenaveneziani · 2018-06-04T21:08:47Z

ok, the problem seems to be NERSC. It's having trouble accessing the scratch directory right now. Will try again when things are back to normal.

milenaveneziani · 2018-06-05T06:53:23Z

ok, after fixing a few things, a test finally gives some reasonable results.
This is the result of running this branch:
http://portal.nersc.gov/project/acme/milena/20180129.DECKv1b_piControl.ne30_oEC.edison_MOConline/ocean/index.html

And this a result I obtained previously using the postprocessing script and accounting for the bolus velocity:
http://portal.nersc.gov/project/acme/milena/20180129.DECKv1b_piControl.ne30_oEC.edison_transects_new/ocean/index.html#moc

Aside from some plot differences (we need to plot better latitude extent), I suppose the differences between these two groups of plots are due to the fact that the online MOC is not accounting for the bolus. I need to test this by looking at a high-res run (no GM).

milenaveneziani · 2018-06-05T06:56:04Z

@xylar: you will probably hate streamfunctionMOC.py even more now than before :) Lots of repetitions in the code.. But I think it is good to first make sure that things work, and then we can re-structure as much as we want.

milenaveneziani · 2018-06-05T06:58:27Z

mpas_analysis/ocean/streamfunction_moc.py

        latAtlantic = self.lat['Atlantic']
        dLat = latAtlantic - 26.5
-        indlat26 = np.where(dLat == np.amin(np.abs(dLat)))
+        indlat26 = np.where(np.abs(dLat) == np.amin(np.abs(dLat)))


I believe this was a bug, and we were just lucky that it worked before.

milenaveneziani · 2018-06-05T07:03:02Z

One TBD thing is to add a temporary config option to force the use of the postprocessing script, even if the AM is on, until we fix the online MOC to take into account the GM bolus velocity, when GM is on.
I will do this tomorrow.

xylar · 2018-06-05T07:47:04Z

From the results, it looks like this is really coming along.

milenaveneziani · 2018-06-05T16:01:20Z

Trying to find a high-res case for testing now.

This should be ready to be reviewed @xylar.
@vanroekel, if you have a bit of time, it would be great if you could take a look too.

Any suggestion about how to adjust the latitude range in the plot is welcome. The latbin variable does not depend on the regionName in the MOC AM, it just goes from -90 to 90.

vanroekel · 2018-06-05T16:15:15Z

@milenaveneziani I can test this on anvil, you could also test this on the coupled run on theta.

/projects/ClimateEnergy_2/azamatm/E3SM_simulations/20180410.A_WCYCL1950_HR.ne120_oRRS18v3_ICG.theta/

milenaveneziani · 2018-06-05T17:16:54Z

ok, @vanroekel, thanks.

milenaveneziani · 2018-06-05T18:44:13Z

@vanroekel or @xylar, could either of you transfer the transect file for the RRS18to6v3 on theta?
I see these two directories where region_masks is:
/projects/ccsm/acme/mpas-analysis_data/mpas_analysis/region_masks owned by Luke
/projects/ccsm/acme/diagnostics_data/region_masks owned by Xylar,
but I don't have permission to write in any of them.

I have completed the MOC online ON test for the coupled high-res run on theta, I only have to do the test for MOC online off. Thanks.

xylar · 2018-06-05T19:31:57Z

@milenaveneziani, if you're not it the ccsm group on theta, I think you should get yourself added. I don't want to make /projects/ccsm/acme/diagnostics_data/region_masks world writable. You should definitely be using that directory rather than the one @vanroekel owns on theta. We should delete the other directory, since we don't plan to maintain it.

xylar · 2018-06-05T19:35:04Z

I think I just downloaded what you needed (oRRS18to6v3_SingleRegionAtlanticWTransportTransects_masks.nc) from https://web.lcrc.anl.gov/public/e3sm/diagnostics/mpas_analysis/region_masks/
Let me know if that's not right.

xylar · 2018-06-05T19:45:04Z

Any suggestion about how to adjust the latitude range in the plot is welcome. The latbin variable does not depend on the regionName in the MOC AM, it just goes from -90 to 90.

@milenaveneziani, I can't think of a better option than having a config option for this. My plan in #246 is to have a separate config section of each region (for now, [streamfunctionMOCGlobal] and [streamfunctionMOCAtlantic]) and these could include both the color map options and the min/max in latitude. If you'd rather just put config options in [streamfunctionMOC] for now, that's okay, too.

xylar · 2018-06-05T20:03:04Z

@milenaveneziani, I'm having trouble maintaining a connection to Edison long enough to do a full test. But I did an MOC-only test here:
http://portal.nersc.gov/project/acme/xylar/add_MOC_online/20180129.DECKv1b_piControl.ne30_oEC.edison/ocean/index.html

Everything looks good to me!

xylar · 2018-06-05T20:04:05Z

configs/anvil/config.20170926.FCT2.A_WCYCL1850S.ne30_oECv3.anvil

 ninoSubdirectory = Nino
 mhtSubdirectory = MHT

-


I think we want to keep 2 blank lines between each section so it's easier to find section. @gstreletz did a lot of work to put in this formatting.

Oops, sorry. I thought this was config.default. Ignore my comment above.

I only removed it because it wasn't consistent in the config example cases: in most cases there weren't blank lines, so I thought of just removing this one.
Instead, config.defaults has the blank lines between each section.

xylar · 2018-06-05T20:05:50Z

configs/anvil/config.20170926.FCT2.A_WCYCL1850S.ne30_oECv3.anvil

+# in its calculation, whereas the postprocessing script does.
+# NOTE: this is a temporary option that will be removed once the online
+# MOC takes into account the bolus velocity when GM is on.
+usePostprocessingScript = True


I think this is perfectly fine as a temporary solution.

milenaveneziani · 2018-06-05T20:06:13Z

Thanks for downloading the transect file. That is the right one.

xylar · 2018-06-05T20:24:06Z

mpas_analysis/ocean/streamfunction_moc.py

            depth.description = 'depth'
            depth.units = 'meters'
-            depth[:] = refTopDepth
+            depth[:] = self.depth


Yep, fine to change this.

xylar · 2018-06-05T20:25:15Z

Oh, my goodness! MOC goes from one of the slowest tasks to one of the fastest for runs that can handle it!

xylar · 2018-06-05T20:44:18Z

QU240wLI tests with both python 2 and python 3 ran successfully with the offline MOC (because the analysis member was not enabled).

An EC60to30 test with python 3 ran just fine as well, with the analysis member (essentially same test I tried on Edison with python 2).

An EC60to30wLI test is running now (with python 3 and offline MOC) but I don't think I need to wait for it to approve the PR.

I'd suggest putting in the latitude bounds before merging but I don't think I need to approve the results since it's likely a simple fix. The code to crop the xarray data set should be something like this part of the SOSE transects:
https://github.com/MPAS-Dev/MPAS-Analysis/blob/develop/mpas_analysis/ocean/sose_transects.py#L269

        lat = ds.lat.values
        mask = numpy.logical_and(lat >= minLat, lat <= maxLat)
        indices = numpy.argwhere(mask)
        ds = ds.isel(lat=slice(indices[0][0], indices[-1][0]))

milenaveneziani · 2018-06-05T20:46:14Z

So, these are the results of my test on the coupled high-res case.

Online MOC script ON:
http://portal.nersc.gov/project/acme/milena/20180410.A_WCYCL1950_HR.ne120_oRRS18v3_ICG.theta_MOConlineON/ocean/index.html

Online MOC script OFF:
http://portal.nersc.gov/project/acme/milena/20180410.A_WCYCL1950_HR.ne120_oRRS18v3_ICG.theta_MOConlineOFF/ocean/index.html

I would have preferred a better correspondence for the Atlantic MOC between the 2 cases, but I think this is good enough. At least there is nothing wrong with the script.

Now the only thing to fix is the latitude range.

And yes, now the MOC is the fastest task.

xylar · 2018-06-05T20:49:15Z

I would have preferred a better correspondence for the Atlantic MOC between the 2 cases, but I think this is good enough. At least there is nothing wrong with the script.

Global looks great! Atlantic isn't that different and I suspect differenes are due to different masking. The online MOC doesn't use the mask from geometric_features, right? Does it use the exact same southern transect?

milenaveneziani · 2018-06-05T20:53:09Z

Yes, they use the same transect. I don't remember about the regional mask file. Maybe @vanroekel does.

milenaveneziani · 2018-06-06T03:52:51Z

mpas_analysis/ocean/streamfunction_moc.py

+                                         'latBinMax{}'.format(region))
+            indLat = np.logical_and(x >= minLat, x <= maxLat)
+            x = x[indLat]
+            z = z[:, indLat]


I do not have xarray datasets at this point, so I am subsampling x and z individually.

milenaveneziani · 2018-06-06T03:58:39Z

The latitude bounds as config options work. I just added them to the streamfunctionMOC section for now.

@vanroekel: I think between me and @xylar there was enough testing for this one. I can remove you as reviewer and merge, if that is OK with you.

vanroekel · 2018-06-06T04:19:05Z

I agree. Please feel free to remove me as a reviewer. Thanks for adding this! Sent with BlackBerry Work (www.blackberry.com) From: Milena Veneziani <[email protected]<mailto:[email protected]>> Date: Tuesday, Jun 05, 2018, 9:58 PM To: MPAS-Dev/MPAS-Analysis <[email protected]<mailto:[email protected]>> Cc: Van Roekel, Luke <[email protected]<mailto:[email protected]>>, Mention <[email protected]<mailto:[email protected]>> Subject: Re: [MPAS-Dev/MPAS-Analysis] Add script to plot online MOC (#415) The latitude bounds as config options work. I just added them to the streamfunctionMOC section for now. @vanroekel<https://github.com/vanroekel>: I think between me and @xylar<https://github.com/xylar> there was enough testing for this one. I can remove you as reviewer and merge, if that is OK with you. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#415 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ANB5UTbnme-NsELYby-j820iG4Jizhzoks5t51NwgaJpZM4UXVj8>.

milenaveneziani · 2018-06-06T04:20:01Z

Thanks Luke.

milenaveneziani · 2018-06-06T04:23:05Z

This was easier than expected, thanks to @vanroekel notebook script.

xylar · 2018-06-06T06:47:00Z

Congratulations, @milenaveneziani. This seems to have gone quite smoothly.

milenaveneziani requested a review from xylar June 1, 2018 21:25

milenaveneziani self-assigned this Jun 1, 2018

milenaveneziani requested a review from vanroekel June 1, 2018 21:26

milenaveneziani added enhancement in progress priority work in progress labels Jun 1, 2018

milenaveneziani commented Jun 1, 2018

View reviewed changes

milenaveneziani commented Jun 4, 2018

View reviewed changes

milenaveneziani commented Jun 5, 2018

View reviewed changes

xylar reviewed Jun 5, 2018

View reviewed changes

xylar approved these changes Jun 5, 2018

View reviewed changes

milenaveneziani added 4 commits June 5, 2018 20:12

Add script for online MOC

a23da8f

Removed use of bolus as option if GM is on

a321f7f

Added tmp config option 'usePostprocessingScript'

1d83048

Add latitude min/max for better plotting of MOC

e918458

milenaveneziani force-pushed the add_MOC_online branch from 0057825 to e918458 Compare June 6, 2018 03:13

milenaveneziani commented Jun 6, 2018

View reviewed changes

milenaveneziani changed the title ~~WIP: Add script to plot online MOC~~ Add script to plot online MOC Jun 6, 2018

milenaveneziani removed in progress work in progress labels Jun 6, 2018

milenaveneziani removed the request for review from vanroekel June 6, 2018 04:20

milenaveneziani merged commit e918458 into MPAS-Dev:develop Jun 6, 2018

milenaveneziani deleted the add_MOC_online branch June 6, 2018 04:23

xylar mentioned this pull request Jun 6, 2018

Checklist for v1.0 release #347

Closed

9 tasks

milenaveneziani mentioned this pull request Jun 6, 2018

Add support for online MOC #348

Closed

		# }}}

		def _compute_moc_time_series_analysismember(self): # {{{

Add script to plot online MOC #415

Add script to plot online MOC #415

Uh oh!

Conversation

milenaveneziani commented Jun 1, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

milenaveneziani commented Jun 4, 2018

Uh oh!

milenaveneziani commented Jun 4, 2018

Uh oh!

milenaveneziani commented Jun 4, 2018

Uh oh!

milenaveneziani commented Jun 5, 2018

Uh oh!

milenaveneziani commented Jun 5, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

milenaveneziani commented Jun 5, 2018

Uh oh!

xylar commented Jun 5, 2018

Uh oh!

milenaveneziani commented Jun 5, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vanroekel commented Jun 5, 2018

Uh oh!

milenaveneziani commented Jun 5, 2018

Uh oh!

milenaveneziani commented Jun 5, 2018

Uh oh!

xylar commented Jun 5, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

xylar commented Jun 5, 2018

Uh oh!

xylar commented Jun 5, 2018

Uh oh!

xylar commented Jun 5, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

milenaveneziani commented Jun 5, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xylar commented Jun 5, 2018

Uh oh!

xylar commented Jun 5, 2018

milenaveneziani commented Jun 1, 2018 •

edited

Loading

milenaveneziani commented Jun 5, 2018 •

edited

Loading

xylar commented Jun 5, 2018 •

edited

Loading