September 2025 | Oceanography
DEMOCRATIZE THE DATA
A NEW WAY TO ANALYZE AND DESIGN OCEAN MODELS
By Thomas W.N. Haine
INTRODUCTION
Simulation of ocean currents by numerical models has been revo-
lutionized by information technology advances in the last 50 years.
New discoveries have resulted from improved observing tech-
nologies, such as the global Argo network of autonomous profil-
ing floats (Riser et al., 2016; Argo, 2020) and satellite observations
of sea level (Lee et al., 2010; Vinogradova et al., 2025). Improved
ocean circulation models have also resulted in new discoveries
(Fox-Kemper et al., 2019; Haine et al., 2021), particularly those
based on better model grid resolution. The growth in ocean cir-
culation model fidelity brings challenges, however. One chal-
lenge concerns the difficulty of providing access to the very large
volumes of data ocean circulation models produce, and another
concerns the priorities for future cutting-edge ocean circulation
model simulations.
This commentary introduces and explains these topics and out-
lines some possible ways ahead. Developments in cloud storage
and cloud computing are providing open cyberinfrastructure plat-
forms that lower the barrier to data access. Open discussion on
future circulation model priorities is also beginning. These ser-
vices for, and engagement with, the oceanographic community
aim to make cutting-edge ocean current simulations as widely
accessible and as useful as possible.
GRID CELL AND DATA GROWTH
Global ocean general circulation models (OGCMs) show expo-
nential growth in grid cell resolution. This remarkable expansion
ultimately derives from Moore’s law, which states that the density
of microelectronic devices doubles every two years (Moore, 1975).
To illustrate, Figure 1 shows the number of horizontal grid cells
used to discretize the global ocean in five cutting-edge OGCMs
since 1980 (with black dots). The number of horizontal grid cells
doubles every 2.5 yr, keeping up with Moore’s law (some of the
increase in computer power is used to refine OGCM vertical res-
olution). Nowadays, cutting-edge OGCMs have horizontal resolu-
tions of around 1 km, with hundreds of millions of grid cells cov-
ering the surface of the global ocean.
Coupled Earth system models of the kind used to project
global climate change by the Intergovernmental Panel on Climate
Change (IPCC) also show exponential refinement of the horizon-
tal grid resolution in their ocean models (Figure 1, colored dots).
For these models, the doubling time is 3.7 yr, somewhat slower
than for OGCMs because other components of the Earth system
model compete for the computer speedup.
Observations of the global ocean have been revolutionized by
information technology advances too. Figure 1 shows, for example,
the number of annual deep stations with high-quality temperature
measurements (CTD stations deeper than 1,000 m). In the early
2000s, the rate of such observations increased by a factor of 10 as
the global Argo network came online. Today, about 100,000 deep
temperature stations are reported each year.
Consider next the relative rates of growth of OGCM resolution
and deep temperature measurements. Figure 1 shows that OGCMs
outstrip the observations, so there are now around 1,000 hori-
zontal grid cells for every deep temperature station. Put another
way, the average spacing between Argo CTD profiles is 300 km,
ABSTRACT. Ocean circulation models running on the latest supercomputers can cover the globe with resolutions of a few kilome-
ters. These virtual ocean datasets are increasingly realistic and provide insight into processes at scales that are inaccessible with conven-
tional observations. Because these datasets are far too massive for individual researchers to download and analyze, new cloud-based,
open-source, cyberinfrastructure resources are being developed. These tools provide a new analysis paradigm that is scalable, accessible,
and inclusive, and that democratizes access to ocean circulation model output. They also accelerate the pace of analysis of ocean models
and thereby increase the pace of discovery in oceanography. Another challenge concerns the priorities for next-generation ocean circu-
lation models. In particular, to improve circulation model simulations, how should increased supercomputer power be spent? Input on
this question from the oceanographic community is sought.
COMMENTARY