Storyboard

[wghuneke]

Suggested figures that go into paper. Add example figure, short description, suggestions for panels/what needs to be changed...

[wghuneke]

Figure 1: Experimental design (enhancing/relaxation of climatological wind field) a) climatological wind field in JRA b) historical trends in JRA and CMIP c) projected trends in CMIP d) applied perturbation

[adele-morrison]

I've also added a figure summary page here. I suggest we keep the discussion here in this issue, but can put the result of this discussion on that figure summary page, to keep a cleaner version and somewhere to have the figures in order.

[wghuneke]

Sounds good. Thought it's easier to add figures in an issue.

[julia-neme]

Figure 1: Experimental design (enhancing/relaxation of climatological wind field) a) climatological wind field in JRA b) historical trends in JRA and CMIP c) projected trends in CMIP d) applied perturbation

For c), trends for the period 2015-2100, or perhaps 2100-2080 mean minus 2015-2005 mean?

[adele-morrison]

You can add figures to the summary page just the same as issues, I've added an example one in now: https://github.com/adele157/easterlies-collaborative-project/blob/master/Figure_outline.md

This issue is good for discussion though, because we'll get email notifications and I'm guessing there will be lots of discussion to be had!

[adele-morrison]

Should we decide what journal we're aiming for as well? i.e. Are we aiming for 4 or 10 figures?

[adele-morrison]

I wonder if for the CMIP comparison, rather than focusing on the multi-model mean changes, we could perhaps focus on the scatter in the CMIP historical and future projections. ie. Make the argument that CMIP doesn't tell us much, because the models are all over the place and don't even reproduce the historical period well. So that's why we have chosen an idealised perturbation and simply amplified the existing wind pattern.

[matthew-england-unsw]

Yes agreed Adele - good call.

[julia-neme]

OK, how about then: Figure 1 a), b) JRA55-do mean fields of uas and vas for 1958 - 2015 c), d) JRA55-do trends in uas and vas for 1958 - 2015 e), f) Boxplots of CMIP6 models trends in different regions, with JRA55-do as reference.

I suggest the boxplots are a nice way of showing the spread in trends and differences with JRA, since they are highly variable depending on location. Some regions are OK in CMIP6, some are really bad. I've uploaded the figure for you to see.

[matthew-england-unsw]

Lovely plot Julia!! :-) So this is historical wind trends? What years? I wonder about a final row showing (say) the multi-model mean wind anomalies [2081-2100] or trends [2000-2100] for CMIP6 (under the high emissions scenario)? Biases in CMIP6 models comes out nicely here.
Love the circumpolar tendencies that can be clearly seen in the box plots as well.

[matthew-england-unsw]

Should we decide what journal we're aiming for as well? i.e. Are we aiming for 4 or 10 figures?

I'd vote for 10 I think. Unless we make 10 and decide that 4 tell the essential story, and then we can move the other 6 to Suppl. Info?

[StephenGriffies]

I find it easier to write the long manuscript first to be sure the story is solid and complete. Then, could cut back to shorter GRL size if warranted/desired.

[wghuneke]

I suggest we need figures that show (somewhat in this order):

• two-time scale response (e.g. using SSH): quick Ekman response in first year, reversal of SSH signal within a couple of years (we think due to sea ice export increasing the sea ice/DSW formation)
• increase in DSW formation (e.g. bottom age, bottom temp/salt, map of surface water mass transformation rates)
• https://github.com/adele157/easterlies-collaborative-project/issues/10, increase in formation, increase in export

[AndyHoggANU]

I've been thinking a bit about this figure order. I think we have two possibilities:

1. Something linear, where we build up a picture gradually -- starting with temp and salt evolution, SSH response, sea ice evolution and export, and culminating in the DSW and off-shelf age changes. (Note that we may need a second set of figures showing different experiments to distinguish between hypotheses.)
2. Show the headline result first (well, as figure 3, assuming a CMIP 6 winds + model configuration figure). This, I think, would be the DSW change/age change figure. Then, go back to investigating other quantities to disentangle hypotheses.

I quite like the second option for this paper, accepting that it is a little unusual. Any thoughts?

Also, I am not quite 100% convinced that we have fully disentangled the Ekman from the sea ice export hypothesis. Do we have a good experiment that could actually separate these two possibilities? That might be important for the storyline I am proposing...

[adele-morrison]

I like this suggestion Andy, but I would be tempted to show the SSH time series before the DSW plot. That way we could start off by saying that we might naively to expect the easterlies to pile up water on the shelf, and indeed we do see that initially, but it is quickly overwhelmed by a larger opposing response by the DSW. Then go into DSW mechanisms. I agree I don’t think we’ve distangled the mechanisms completely yet, let’s discuss today what analysis / simulations we need to do that properly.

[matthew-england-unsw]

I think that’s a great idea Andy if the aim is to make the paper headline about the DSW response. But over the wider ocean there’s a largely coherent circumpolar (and counter-intuitive) response that is pretty impressive too, that remains when we pulled away the DSW formation region wind anomalies. To me that’s the headline story, and the DSW changes are kind of secondary or at least consequent on that. But that could just be me, DSW enthusiasts would be taken in by the storyline you propose. :-)

[wghuneke]

Thanks for updating Fig 1 @julia-neme, looks good!

Just a few minor things: Can you change the colorbar ticks (control) to a spacing of 1? Or maybe even 2 and for the anomaly 0.5 and then you can increase the font size for all the ticks and the other text. And maybe show a few less arrows, I find it looks slightly busy, but that might just be me.

[wghuneke]

Also adding a short summary on what we discussed today (mostly for @StephenGriffies ):

We continued our discussion about the paper outline, Adele and Andy updated the Figure_outline.md document accordingly.

We have three hypotheses for different mechanisms that could explain the observed response: A local wind (katabatics) mechanism, B Upwelling mechanism, C Sea ice mechanism. We can rule out A with the additional perturbation experiment we did. The next steps are to do a bit more analysis/run additional simulations to disentangle B and C. Atm, we think C is more relevant.

Adele is currently running two more simulations for the UP case: a "meridional winds only" and a "zonal winds only" case, where only one component is perturbed and the other is kept as in the control run. The idea is that it will change the curl, hence Ekman upwelling, but not so much the sea ice advection.

[matthew-england-unsw]

Thanks @wghuneke great summary from today. 👍

[matthew-england-unsw]

PS: @julia-neme given that the wind experiments are simply +10% and -10% south of the purple line, the 2 x anomaly plots on the right are maybe not needed, and so instead you could use the full 'real estate' of the page width to devote to a single circumpolar wind vector plot....? I'm also curious how this looks without uniform vectors (which can be deceiving in the 'blank' low wind regions). e.g. if taking up the full page width and using a proper vector scale, we will maybe see clearly the full wind field and then knowing the expts are +/- 10% anomalies on this mean field could be sufficient?

[adele-morrison]

Yes agreed Matt, I think we could lose the 2 right panels. Julia, what do you think about using the model land/sea mask too, which cuts out some more of the Ross and Weddell?

[julia-neme]

Yeah! Will update based on all these comments :) Thanks!

[julia-neme]

Fig 1 updated

With input from @adele157 and @matthew-england-unsw, how about this Fig. 1:

a) Wind streamlines colorcoded by wind velocity, 1000m isobath in black contour, zero divide line poleward of which we apply the perturbation in dark blue and 1000m, 3000m depth contours in light grey. The boxes enclose the masked out regions for the additional experiment. b) Zonal and c) meridional components of the wind, with 1000m isobath in black.

I've opted for streamlines since the quiver plots are a bit messy and difficult to interpret. The advantage of having the 3 panels is that a) makes it easier to understand the wind field, b) and c) allow to focus on the zonal/meridional components separately.

[julia-neme]

I had a go at Figure 2: Top panel: monthly SSH anomalies. Bottom panel: monthly magnitude of the zonally integrated overturning streamfunction at 55S.

I'm not sure who wins the bet based on panel 2! Perhaps this is not the best way to measure bottom water transport? I'm currently working on trying to separate thermosteric from halosteric contributions (#44) to add to the top panel.

[adele-morrison]

Yep, I think we may need to look at DSW transport across the 1000m isobath instead! I guess the transport change just isn't making it that far north in this time. Is this 55S?

[AndyHoggANU]

OK, or maybe 2000m isobath?

[matthew-england-unsw]

(Thanks Julia!!). Ummm.... the values look a bit strange? e.g. there's a strong semi-annual cycle from what I can tell? I count 8 cycles per 4 yrs. And that range from max to min within a year is min O(10 Sv) shooting up to O(30-40 Sv) max values? But not from summer to winter: it's 2 peaks per annum..... ?

[wghuneke]

I started working on the bottom age plot (paper Fig 3) and created a list with items we need to agree upon:

• Which projection do we want to use? Stereo is nice, but I find it hard to see details on the continental slope. (compare my two version of the age plot)
• 1 km isobath: We should use the same definition
• Naming of experiments
• Which years do we want to show (if it's not a time series)? Average over the last year or last 5 years? (I found it weird to average the age over 5 years, but maybe that's ok.)

[wghuneke]

Here two version of the bottom age plot for the last year of the model run (year 15). The control shows the normalised (length of model simulation) age and the anomalies are the anomaly from the control (not normalised).

[matthew-england-unsw]

Nice Wilma! I agree - polar-stereographic is not so ideal here, given our shelf focus. So the bottom set gets my vote. And I agree - age differences averaged over last 5 yrs is not as intuitive as age for last year, but I suppose either/or is fine and maybe what's important is to match up with whatever time window we're using to examine hydrographic property anomalies?

[matthew-england-unsw]

Experiment names? How about:

[julia-neme]

Maybe we should also decide on the colors for time series for each experiment?

[adele-morrison]

Sample Figure 4:

[wghuneke]

Nice! Interesting that Adelie went up for the no katabatics case. The plot also shows how DSW formation changes less in the Ross Sea than, e.g., in the Weddell Sea.

[AndyHoggANU]

I have been working a little on Figure 5 -- which is designed to show the net buoyancy fluxes. I decided there is little purpose in showing maps, and that instead we should restrict ourselves to shelf-averaged or shelf-summed timeseries or climatologies.

Here (a) and (d) show the annual cycle of net heat flux and net FW flux averaged over the shelf, along with the up and down anomalies. One point here is that FW flux is almost entirely melt + river, the latter of which is constant, so for anomaly plots (e and f) I just sum the melt (to give a flux in Sv).

I decided that it may be worth including the zonal and meridional expts in the remaining panels, which show the anomaly in climatology for each experiment (b,e) and the 12-month running mean (c,f). If we retain these lines here, that may change the order we explain things in the paper...

.

[wghuneke]

Update on conventions

Naming and colours: CONTROL -> 'k' WIND+ -> 'C0' WIND$-$ -> 'C1' WIND+$zonal$ -> 'C2' WIND+$merid$ -> 'C3' WIND+_$noDSW$ -> 'C4' (might change the naming)

Length of experiments:

• all experiments for 10 years

[adele-morrison]

Andy, a comment on Figure 5. I think the point here is to show that for the UP and DOWN simulations: a) heat flux changes are insignificant compared to freshwater flux changes, and b) the changes in sea ice formation (i.e. melt diagnostic) dominate the change in the freshwater flux change. I think we could make this point most clearly if all 3 of these things have the same units, and then could we show them all on the same plot?

Probably also the changes in melt (i.e. ice export) could be kept for later when we use the zonal/meridional experiments to prove the sea ice mechanism?

[adele-morrison]

I did some editing of the Draft figure outline page with a suggestion of when to introduce the zonal/meridional experiments.

Basically, I think even before we introduce the new zonal/meridional experiments, we could first show the Ekman pumping Figure 6 and argue that the increase in upwelling is driven by zonal winds, and then the sea ice advection Figure 7 and argue that the sea ice advection is driven by meridional winds. So on these figures we wouldn't actually use the results from those experiments.

Then Figure 8 shows the SSH and DSW transport changes from the zonal/meridional experiments and after that we make the connection with the meridional winds driving the DSW changes.

Possibly this is actually what we said last meeting. But today a lot of the plots before Figure 8 already had the zonal/meridional experiments on them.

[wghuneke]

Here is a time series of sea ice volume (circumpolar and annual integral). The idea was to show this together with Andy's Fig5f): WIND+ and WIND+merid have more sea ice formation (neg melting), but sea ice volume is lower for these experiments. This means that there must be more sea ice export to the north.

[adele-morrison]

I guess I was thinking rather than showing this as a separate plot, can we combine this with the melt diagnostic to more accurately compute the sea ice export off the shelf? Or at the very least can we use the changes in sea ice volume over time here to quantify what the error is in assuming that the change in formation minus melt (i.e. the melt diagnostic) = change in export?

On Thu, 4 Nov 2021 at 12:52, Wilma Huneke @.***> wrote:

Here is a time series of sea ice volume (circumpolar and annual integral). The idea was to show this together with Andy's Fig5f): WIND+ and WIND+merid have more sea ice formation (neg melting), but sea ice volume is lower for these experiments. This means that there must be more sea ice export to the north.

[image: Screenshot_SeaIceVolume_CircumpolarIntegral_AnnualIntegral] https://user-images.githubusercontent.com/15355753/140244039-6332cda4-aaf9-4b73-8f2c-58559fe2af8a.png

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/adele157/easterlies-collaborative-project/issues/37#issuecomment-960369086, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACA44U6YGZ5N77JHVOY57NTUKHRNNANCNFSM5DV6WWMA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[AndyHoggANU]

Hi @adele157 - I've been working on the relative buoyancy fluxes and have struck a flaw in our logic. All the calculations I have done indicate that buoyancy fluxes for heat are greater than for salt. OK, so it's most likely I have made an error. But, at it's most basic, I can calculate alpha*net_surface_heat_flux/C_p and beta*SSS*net_fw_flux to be both in kg/s/m^2, and I get something like this (for the 15-year running-mean): Any thoughts?

[adele-morrison]

Hmmm that does seem problematic. Where’s the code for this?

On Fri, Nov 5, 2021 at 2:50 PM, Andy Hogg @.***> wrote:

Hi @adele157 https://github.com/adele157 - I've been working on the relative buoyancy fluxes and have struck a flaw in our logic. All the calculations I have done indicate that buoyancy fluxes for heat are greater than for salt. OK, so it's most likely I have made an error. But, at it's most basic, I can calculate alphanet_surface_heat_flux/C_p and betaSSS*net_fw_flux to be both in kg/s/m^2, and I get something like this (for the 15-year running-mean): [image: Screen Shot 2021-11-05 at 2 46 27 pm] https://user-images.githubusercontent.com/26753100/140455283-b02b854d-0c22-4a0a-9118-b225fd402147.png Any thoughts?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/adele157/easterlies-collaborative-project/issues/37#issuecomment-961607209, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACA44U26TMTFVKP6B7PRVEDUKNIBPANCNFSM5DV6WWMA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[AndyHoggANU]

It's a bit scratchy -- I didn't calculate local fluxes, I just got an estimate for alpha and beta using averages, and then converted to see what it looked like. See

https://github.com/adele157/easterlies-collaborative-project/blob/master/notebooks/buoyancy_fluxes/Heat-FW-Fluxes-Figure.ipynb

[StephenGriffies]

@AndyHoggANU did you also compute the contribution from salt fluxes? I would presume that salt from the ice model is important too, as well as the SSS restoring.

sfc_salt_flux_total = sfc_salt_flux_restore + sfc_salt_flux_coupler

[adele-morrison]

Yes, I think there's something wrong with your code @AndyHoggANU. Not sure exactly what, but when I use monthly alpha and beta, and also a slightly different shelf mask to yours, I find that the freshwater fluxes are ~5 times larger than the heat fluxes over the shelf, see here: https://github.com/adele157/easterlies-collaborative-project/blob/Adele/notebooks/buoyancy_fluxes.ipynb

[AndyHoggANU]

Yep, right you are. I had forgotten to convert temperature to Celsius. Will fix and see what happens.

[AndyHoggANU]

OK, take 2 for Fig. 5 .

[adele-morrison]

Looks good Andy. Do we also want to show that the buoyancy flux anomaly is dominated by sea ice? i.e. adding another line to the anomaly plot for melt (in same buoyancy units)? Or another panel if it's too busy? I know early on Matt kept wondering if the salinity changes could be driven by evaporation, so might be nice to rule that out here.

On Sat, 6 Nov 2021 at 15:20, Andy Hogg @.***> wrote:

OK, take 2 for Fig. 5 [image: Figure5(1)] https://user-images.githubusercontent.com/26753100/140597625-0886eef6-0fcd-4231-a2f5-19a332106f70.png .

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/adele157/easterlies-collaborative-project/issues/37#issuecomment-962391943, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACA44UZ5QPKFDPQM2EI4SITUKSUG3ANCNFSM5DV6WWMA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[AndyHoggANU]

Yeah, I was thinking that they would just be too similar, but perhaps we can do it this way (I think it is self-explanatory, but I can explain if you want).

[matthew-england-unsw]

Thanks Andy. I think if it’s that similar it can also just be noted in the caption / ms text.