Session 5: Model development options

This topic contains discussions, questions and thoughts for Session 5 of the CMIP7 Workshop
Time and Date: Wednesday 01 March, 2023 from 9:00 - 10:45 am

Chair: Kelsey Druken

Speakers and topics:

Australia’s CMIP6 configurations: ESM1.5, CM2, OM2 + future possible configurations (Tilo Ziehn, CSIRO, 10+5 min)

OM3 (Adele Morrison, ANU, 10+5 min)

CABLE/POP (Ian Harman, 10+5 min)

Ice sheet modelling (Ben Galton-Fenzi, Australian Antarctic Division (AAD), 10+5 min)

Atmospheric Chemistry (Matt Woodhouse, CSIRO, 10+5 min)

Model evaluation (Gab Abramowitz, UNSW,10+5 min)

Tropical variability Modelling (Shayne McGregor, Monash University, 10+5 min)

Just to update Tilo’s numbers on OM2 compute requirements:
ACCESS-OM2 (1°) - 288 cores, 96 yrs/day
ACCESS-OM2-025 (0.25°) - 1824 cores, 12.2 yrs/day
ACCESS-OM2-01 (0.1°) - 12,144 cores, 3.7 yrs/day …

1 Like

Tilo Ziehn: Possible New configurations:

Adele Morrison: MOM6 vs MOM5

Adele Morrison: ACCESS-OM3 progress

1 Like

Just to briefly answer Shayne’s question on hybrid coordinates in MOM6. The way MOM6 timesteps is that the horizontal dynamics is separated from vertical dynamics. The vertical dynamics allows you to regrid to any vertical coordinate you choose, and thus gives you complete flexibility of the vertical coordinate — it’s just up to you to provide a nice algorithm on how to do that regridding. We are currently testing z-star, terrain-following, quasi-isopycnal, hybrid (i.e. z-star merged with isopycnal) and adaptive…

Ian Harman: Possible ACCESS land configurations in CMIP7:

1 Like

Comment to Ian: NUOPC has the flexibility to support a land BGC model split into 2 components coupled with different frequency

1 Like

Nice idea Andy to see if enhanced vertical OGCM resolution (or maybe you meant different vertical coordinate schemes?) will solve the biannual ENSO - easily testable with a ~20-yr simulation in coupled mode. Would be good to test. In COSIMA we spend most of our time in uncoupled mode, this is a great problem to solve across the ocean and atmosphere modelling communities. e.g. maybe it’s the AGCM convection scheme? Or both that and OGCM resolution (etc).

Re. ice sheet processes and meltwater, in case you’re wondering if this matters for CMIP projections, here’s the impact of added MW (high end scenario) in the GFDL projections (taken from Change in future climate due to Antarctic meltwater | Nature):

1 Like

Ben Galton-Fenzi: Ice sheets and climate interactions:

1 Like

Ben Galton-Fenzi: Concluding remarks:

Matt Woodhouse: Is there a definition of “nudging”, and how it differs from “full data assimilation”? Or are they same thing with a different name?

Matt Woodhouse: Chemistry and Aerosol in ACCESS

1 Like

@gab recommended reading these papers

@gab mentioned unit testing: my comment HALLELUJAH!

@gab - nice talk. I think that just because we can make a part better, should we make that part better or move elsewhere? Obvious and substantive errors should be fixed of course, or avoided in the first place. I think about your approach as trying to optimise the model but “How good is good enough?” is perhaps something we should also be asking more often and of each other? This helps define the measure of how we can efficiency.

1 Like

Gab Abramowitz: Model evaluation:

1 Like

Re. Gab’s talk - FYI this is the testing approach being taken in CICE:

Changes, additions and updates to CICE fall into four categories: (I) BFB [bit-for-bit] with no further assessment required; (II) non-BFB but unlikely to be climate changing; (III) non-BFB and climate changing; and (IV) a new model configuration option requiring separate scientific assessment. [This paper] describes the automated methods used to flag the first three categories