Aim 3: emulate ocean heat content from ACCESS-om2-01

[PaulSpence]

Since emulating SST from SAT doesn't seem that challenging, we would like to try to autoregressively emulate ACCESS-OM2’s vertically integrated ocean heat content evolution given surface forcing (basically, emulate Huguenin et al. 2022; https://www.nature.com/articles/s41467-022-32540-5 Nat Comms.)

Ocean state that we would like to predict: OHC(x,y,t) (with the option of also including SST(x,y,t)) Atmospheric forcing provided as input data: U, V, Ta, with also the option of providing humidity and radiation (SW, LW). I.e. at each auto-regressive time-step the model takes OHC, U, V, Ta... (x,y,t) and is trained to output OHC(x,y,t+dt), with some time step dt (Dheeshjith et al. used dt = 1 day).

Possible training /// testing datasets:

• OMIP-2 ACCESS-OM2-1 IAF run /// Maurice’s historical
• Maurice's RDF /// Maurice's historical
• Early part of Maurice’s historical /// late part of Maurice’s historical
• Maurice's historical /// wind and thermal forcing perturbation experiments
• Maurice's historical /// Qian’s run (or something like it)

Metrics to test: Global OHC and spatial structure of OHC (SO average…).

[rmholmes]

@PaulSpence I edited the above. Looks good, but lets start with the others.