search

dhruvbhagtani / Atlantic-meridional-heat-transport-heatfunction

Jupyter notebooks to reproduce figures for the paper:
GNU General Public License v3.0
1 stars 0 forks source link

Re-running experiments: Choices #2

Open dhruvbhagtani opened 2 weeks ago

dhruvbhagtani commented 2 weeks ago

This issue discusses all choices we will have to agree on to re-run the wind and surface heat flux contrast experiments. They are aimed at addressing the reviewers concerns regarding:

  1. Model drift, and
  2. Atlantic-only vs global perturbations,
  3. (Add other points that may become clear as we discuss the model setup).
dhruvbhagtani commented 2 weeks ago

Going through the overall picture:

  1. We will begin by running a control ACCESS-OM2 simulation (~0.8 MSU) for some period, and take some of the last years of this control to generate our surface boundary fluxes.

Choices here: a. Do we need to run this at all? We have an 4 year old version lying in ik11 (025deg_jra55_ryf9091_gadi). I'll need its exe and other config files to extend it to generate high-frequency variables for surface boundary fluxes. b. How many years do we run this control? Original setup: 200 years. c. How many years do we take for creating the surface boundary fluxes? Original setup: 20 years d. Frequency of surface boundary fluxes: Original setup: 3 hourly

My take: We re-run a fresh ACCESS-OM2 control for 80 years, followed by 20 years of outputting the original 3-hourly data.

  1. We begin our flux-forced control (ideally from the same MOM5 used in ACCESS-OM2; re-running ACCESS-OM2 control will also more easily ensure we take the same version for both).

Choices here: a. How many years do we run it for? Original setup: 100 years (0.7 MSU + 0.7 MSU)

My take: Let's see how the simulation progresses. If it's getting equilibrated quicker than 100 years, we can start our perturbations from an earlier date.

  1. We create anomalies for both sets of experiments.

Choices here: a. Since these are Atlantic-only perturbations, how do we tone the anomalies at the edges? What are the edges?

My take: One standard approach to tone the anomalies at the edges is by applying a tan-hyperbolic function. We can set the southern edge as tropics, and northward edge as 60-70 N depending on the range of latitudes that the AMOC spans. Keeping 60-70 N will also partially deal with extra/deficit frazil formation in perturbation experiments.

  1. Original setup for wind stress anomalies: 0.5 and 1.5 times the control setup (0.7 x 2 = 1.4 MSU)

Choices here: a. Do we keep the same perturbation map, i.e., what's the physical justification? b. If the answer to above is yes, then do we keep the same 0.5 and 1.5 magnitudes?

My take: I would prefer changing the perturbation map to something more realistic (NAO-like perturbation haha, but that is an oscillation rather than a long-term trend). But happy to keep the same setup if we can think of a physical reasoning for it (can't think of any). Answer to (4b) will come once we solve (4a).

  1. Original setup for surface heat flux contrast experiments: -15, -7.5, +7.5, +15 (0.7 x 4 = 2.8 MSU)

Choices here: a. Do we keep the same perturbation map, i.e., what's the physical justification? b. If the answer to above is yes, then do we keep the same 0.5 and 1.5 magnitudes?

My take: I would prefer changing the perturbation map to something more realistic. Last time, we mentioned that polar amplification will lead to reduced surface buoyancy flux contrast. But then, we should run freshwater flux perturbation experiments. I'll think more about perturbation maps that we can justify more easily.

Total estimated cost: 6.4 MSU! (Another point to discuss)

Anything else I missed?

navidcy commented 2 weeks ago
  1. We will begin by running a control ACCESS-OM2 simulation (~0.8 MSU) for some period, and take some of the last years of this control to generate our surface boundary fluxes.

I don't see why we need to rerun the ACCESS-OM2 control simulation and not use the one we run already? What would be the difference? If it makes things simpler for technical reasons then go for it!!

We begin our flux-forced control (ideally from the same MOM5 used in ACCESS-OM2; re-running ACCESS-OM2 control will also more easily ensure we take the same version for both).

That's a good enough reason for me to rerun the ACCESS-OM2 control!

  1. flux-forced control

We begin our flux-forced control (ideally from the same MOM5 used in ACCESS-OM2; re-running ACCESS-OM2 control will also more easily ensure we take the same version for both). Choices here: a. How many years do we run it for? Original setup: 100 years (0.7 MSU + 0.7 MSU)

At least the same? Why would it equilibrate faster? You didn't mention that you'd do something different... But over Zoom Andy was mentioning about something regarding making the frazil balance better? I missed that...

  1. We create anomalies for both sets of experiments. Choices here: a. Since these are Atlantic-only perturbations, how do we tone the anomalies at the edges? What are the edges?

What does "tone" mean? I admit I can't really follow the discussion without a figure... my brain struggles to visualize what you are suggesting. But let me say the following:

Regarding buoyancy fluxes, I'd but the mask you did the previous time but cut the mask somewhere in North America and In Europe. This way no sharp Δbuoyancy are introduced in ocean regions. Is what I'm suggesting making sense?

For the winds I'm not sure... perhaps just change the winds everywhere as we did before but only in the North Hemisphere?