search

abby-baskind / seniorthesis

0 stars 2 forks source link

on-campus thesis-ing soon! woo! #16

Open abby-baskind opened 3 years ago

abby-baskind commented 3 years ago

I'm making this issue for 2 reasons. First, I am in hour 2 of waiting for code to run, and I need something to do. And second, today is probably the last day I'm working before traversing the country to get back to Pton, and I need to somehow document what i've been working on, what I've been struggling with, and what's up next on my to-do list, so I don't forget everything before my next in person meeting.

1. PCO2 calculation: It's super slow going. The kernel is not shutting down, but it is taking forever to run. The specifics of this are in issue #15.

2. fgco2 plots: I implemented some minor changes in the plots I already had, namely changing the units and adjusting the x-axis extent. fgco2_projections

I'm just now noticing this plot only goes to 40°S, so I'll need to change that once more so it goes to 30°S.

Screen Shot 2021-08-20 at 10 46 36 AM

I also made a bonus plot that's just the ensemble mean and an error band. I don't think it's necessarily more effective than the previous line plot, but it is simpler. What is interesting about this plot is that uncertainty looks to be greatest where CO2 is being outgassed (between 50 and 60°S). But CNRM-ESM2-1 could be to blame for that.

Screen Shot 2021-08-20 at 10 46 44 AM

I also messed around a bit with adding isolines of density to the projections of fgco2. I got it to work for a single gr model, and it looks great. However, when I tried to do it for all of the projections, the kernel shut down. I haven't done much to debug that yet.

Screen Shot 2021-08-20 at 10 20 41 AM

3. areacello: areacello is not available for every model. It's only available for CESM2, CESM2-FV2, CESM2-WACCM-FV2, GFDL-ESM4, IPSL-CM6A-LR, CanESM5, CNRM-ESM2-1, and CanESM5-CanOE. I'm hoping that i can reuse areacello from some models for other models with the same dimensions (e.g. using areacello from GFDL-ESM4 for GFDL-CM4).

4. Histograms: I haven't added weights to my histograms yet, so I'm behind on this. Partially because I had to spend some time thinking about the areacello issue, partially because my WiFi has been unreliable, and partially because in the fleeting moments with Wifi I've been dealing with PCO2.

gmacgilchrist commented 3 years ago
  1. This may be the sort of thing where you do the calculation once and save the necessary output (as long as it does not exceed 10Gb). Saving an array with the 20-year mean PCO2 for each of the models might not exceed this limit.
  2. Great plots. I would suggest making the colorbar a bit smaller and the maps themselves bigger. You can also explore just doing one colorbar, since it is the same for all plots. This will maximize the space for the maps, where the interesting information is. Take a look at the first answer here for one possible way of doing that. I would also suggest narrowing the limits of the colormap, to reveal more of the detail and differences between the models. In the line plot, I would recommend going from -80 to -30 along the x axis. Also, line plots like this often look better with a smaller aspect ratio - i.e. more rectangle than square, longer along x-axis than the y-axis. The plot with the error shading is cool, but I would agree that it's not much more effective - I personally like seeing all the models. I agree though that it looks like the maximum uncertainty is right at the outgassing region, which is neat, and good motivation.
  3. That's annoying but not a big issue. As you say, the grid (and therefore areacello) should be exactly the same for all models. It may actually be more efficient to just save a single grid file in your home directory and use that with each of the models - this would avoid loading it each time from the individual model files.
  4. All good.