If NumericalEarth.jl is the coupler, what is the model?

[taimoorsohail]

Hi all,

I recently gave a talk and came across some confusion in the audience.

In the Australian ocean modelling community, the model used is called ACCESS-OM3, which is:

ACCESS-OM3 = MOM6 + JRA55do prescribed atmosphere + CICE, coupled via NUOPC

Similarly, for ORCA:

ORCA1 = NEMO + prescribed atmosphere, coupled via OASIS

This begs the question - if NumericalEarth.jl is the coupler, what should we call our various model configurations. For example, what should I call:

Oceananigans.jl + JRA55do prescribed atmosphere + ClimaSeaIce.jl, coupled via NumericalEarth.jl ?

What about:

Breeze.jl + Veros, coupled via NumericalEarth.jl?

I think it would make sense to have a unified naming system for the system of models that come out of this...

[glwagner]

One small thing is that in my opinion, we should indeed think of NumericalEarth.EarthSystemModel as "the model". In other words, the coupled model is a first class citizen here. There isn't really anything you could call the "coupler", except maybe ComponentInterfaces. Which I think is nice! (Curious whether you agree with this idea!)

However, I think your question still stands, because it would be nice to have a way to communicate about different configurations / flavors of EarthSystemModel. In reality, even if all the components are the same, the climate that's simulated by a model depends specifically on every aspect of how the models are set up / configured / parameterized. So maybe, the person / people who do the work to make the configuration get to name it.

[milankl]

I don't think a consistent naming scheme is possible that's also brief and intuitively understandable. I like Greg's philosophy of lifting the coupler up to be the model itself. In same way as we talk about e.g. "Oceanangians with WENO advection" or simply "Oceananigans" for brevity maybe we should just say "NumericalEarth" or to be specific "NumericalEarth with Oceananigans" or "NumericalEarth without sea ice" wherever more detail is needed. If some configurations are very common, e.g. "Oceananigans with Breeze in an Aquaplanet configuration with a meridional land strip and Drake passage" maybe it'll be briefer to say "Aquaplanet NumericalEarth" but I'd leave that to the person coming up with the configuration!

[francispoulin]

I agree with Greg, having NumericalEarth as the coupler can include any or all of it's components. I realize that might not be satisfying because it's not very precise. I would think that in any context people would specify what they are using, for example Oceananigans + Breeze + ClimaSeaIce. But I don't think that changing to name is necessarily going to be more clear.

[glwagner]

For example, what should I call:

Oceananigans.jl + JRA55do prescribed atmosphere + ClimaSeaIce.jl, coupled via NumericalEarth.jl ?

I propose "SuperMelb Icy Ocean ultra-model (v1)" !

[vtamsitt]

Just my two cents... I agree with Greg that I think of NumericalEarth.jl as "the model" and it's the specific configurations (or in some cases you could call them "products", once they have a set of experiments that have output available, validated, particularly in the case of data assimilative state estimates) that we don't currently have names for.
I think it's super valuable to have a simple naming structure for specific configurations that are validated/may be used repeatedly by the community e.g. global OMIP runs, or BGC runs that are valuable to run a long RYF spin-up once and share that to run multiple simulations. But I don't think it's that important for names to be very descriptive (short and memorable is nice!) because that's all captured in the configuration itself and it's more important to have really good metadata on the configuration. I do like how ACCESS for example uses consistent naming for different resolutions of otherwise similar model configurations, so it can be useful to capture some relevant config information in the name if there are multiple similar configurations.

[taimoorsohail]

I agree in principle with giving the task of "naming" the specific configuration of model to the creator. But I also am aware that, from a "marketing" perspective, we want to be able to quickly and easily communicate our specific configuration (or classes of models making up the configuration). So that if I say, "we are running a version of the SuperIcyMelb model", people intuitively think "some combination of Oceananigans.jl + ClimaSeaIce.jl". Similar to how ORCA1 invokes "NEMO+other stuff"?

Even some guidelines around naming various configurations would be nice, so that there is enough overlap between names and there is some central "signal" to audiences that this comes from the body of modelling work centred around/drawing from these repos. By the way this comment I guess is coming from previous practices by modelling groups, so we don't have to do it. But it has helped me and I think helps with take-up and understanding by ECRs etc.

[glwagner]

I thought that names for Earth system models are chosen by the organization / group that creates the model. For example CESM was created by NCAR. But "CESM" doesn't actually refer to specific model components, rather CESM2, CESM3 have the same name because its produced by the same group, but the model components are completely different between them. Similarly NEMO-IFS are coupled in the "EC-Earth" model, but also in DESTIN-E (arguably the latter is a very special project).

So I guess if this principle should hold, then any Earth system model that couples NEMO to IFS should be called EC-Earth, is that right?

That said I think if you want to get together with other people running Oceananigans+ClimaSeaIce and decide on a name for that, it could be helpful. Especially if you also run with the same configuration / parameterizations, then there is a family of models that differs only by resolution and that could be deserving of a name to communicate it.