September 2025

Oceanography | Vol. 38, No. 3

10

FUTURE OGCM PRIORITIES

Returning to Figure 1, notice that the LLC4320 simulation is

already a decade old. Moore’s law has continued in the years since

NASA computed LLC4320, and the time is ripe to make a new

benchmark cutting-edge calculation. Extrapolating the OGCM

red line in Figure 1 suggests that such a new simulation could

have 3 × 109 horizontal grid cells, which corresponds to a horizon-

tal grid scale of 350 m. This resolution captures part of the unex-

plored regime of submesoscale dynamics in which rotational, iner-

tial, and buoyancy effects are all of similar importance (Taylor and

Thompson, 2023), and which is very hard to observe with current

oceanographic instruments.

Alternatively, the extra computational power could be spent

on other priorities. For example, the simulation could be run for

longer than one year at the same resolution as LLC4320. Or the

initial condition could be improved to avoid transient adjustments

during the simulation. The question is, what are the most import-

ant priorities and, in particular, how should the extra computa-

tional power be spent?

This question was asked during a town hall meeting at the

2024 Ocean Sciences Meeting. Participants in the town hall

responded to an online survey that asked them to rank 11 differ-

ent priorities for designing the next cutting-edge global bench-

mark OGCM simulation. Participants could also write in their

own priorities. Figure 4 shows the results of the survey, summa-

rizing the opinions of 44 respondents (the survey is still open—

take the survey).

The survey results show no consensus for future bench-

mark OGCM solutions because all the priorities were ranked as

important by some respondents and as unimportant by others.

Nevertheless, preferences are clear overall. The most highly

ranked priorities include longer run time and better horizon-

tal and vertical resolution. These priorities are relatively easy to

implement because they require little OGCM code development

and little pre-computation before the main OGCM code is run.

Better model spin-up/initial conditions and better air-sea forc-

ing are also highly ranked. These priorities are harder to imple-

ment because they involve improvements (which need to be

precisely defined) to input data from other large, complex mod-

eling systems. The four middle-ranked priorities are: better con-

straints to observations, better model parametrizations, better

model topography, and better mean circulation and stratifica-

tion. These are desirable scientific goals that are easy to state but

hard to achieve. One reason is that they involve detailed tuning of

OGCM parameters and input data, or improvements to OGCM

software. Another reason is that these priorities are interrelated

because, for example, improving the mean circulation probably

requires better parametrizations and topography, which will inev-

itably improve agreement with observations. Two priorities were

ranked as unimportant overall, namely an ensemble of LLC4320

runs (easy to implement) and better diversity in model code

(relatively easy to implement using existing OGCM systems).

Other priorities listed by a few respondents included adding

biogeochemistry, better documentation, and better comparison

with observations.

OUTLOOK

Given the ongoing advances in computational hardware, software,

and infrastructure, the time is ripe for a new cutting-edge OGCM

solution (or more than one) to be computed. Efforts like LLC4320

and the Poseidon Project require significant resources and there-

fore need broad support from academia, industry, funding agen-

cies, and non-professional oceanographers. To date, these efforts

have been supported by government agencies and private founda-

tions with standalone projects every few years. The need to sustain

open shared cyberinfrastructure like SciServer and digital twins

like LLC4320 is widely recognized (Barker et al., 2019; Grossman,

2023; Le Moigne et al., 2023; NASEM, 2024). The future sources of

support and the pathway for migrating from research project fund-

ing to community infrastructure funding are uncertain, however.

One notable example of a stable, long-term, cloud-based data

analysis environment for ocean sciences is the Mercator Ocean

International and Copernicus Marine Service resource, funded

by the European Commission. It provides real-time global ocean

hindcasts, analyses, and forecasts using ocean circulation models,

in situ and remote observations, and data assimilation (although

not presently at the LLC4320 horizontal resolution). Their focus is

on operational oceanography and the state of the ocean for diverse

stakeholders (von Schuckmann et  al., 2024). Apart from aca-

demic users, people have applied the Copernicus Marine Service

PRIORITIES FOR FUTURE GLOBAL

BENCHMARK OGCM SIMULATIONS

FIGURE 4. Results from a 2024 Ocean Sciences Meeting survey on prior-

ities for the next benchmark global OGCM simulation. Forty-four respon-

dents ranked the priorities on the y-axis on a scale of 1 to 12 (1 is the top pri-

ority). The median value is shown with the dotted circle, the 25th and 75th

percentiles are shown with the thick bar, and the thin bars indicate maximum

and minimum values. “Other(s) (write in)” priorities included adding biogeo-

chemistry, better documentation and tutorials, and better evaluation with

observations. Take the survey.