Model resolution and the included physical processes are
two of the most important factors that determine the realism or performance
of ocean model simulations. In this study, a new global surface-wave–tide–circulation coupled ocean model FIO-COM32 with a resolution of
As available computing power has been increasing rapidly, by approximately 1 order
of magnitude every 5 years, the state-of-the-art computing
ability of modern global ocean numerical models is becoming enormously
high, leading to recent achievements in high-resolution global ocean models.
The term of “high resolution” as regards present global ocean models may refer to
those with horizontal resolutions ranging from 1 to 5 km, which are well
beyond the mesoscale resolving threshold in most open oceans (Hallberg,
2013). Further improved resolution has a significant impact on the simulated
eddy activities (Thoppil et al., 2011; Sasaki and Klein, 2012; Biri et al.,
2016; Ajayi et al., 2020), the vertical mass and buoyancy fluxes (Capet et
al., 2016; Su et al., 2018; Dong et al., 2020a), and even the representation
of large-scale circulations (Lévy et al., 2010; Chassignet and Xu,
2017). These high-resolution models are not only helpful in improving our
scientific understanding of mesoscale and sub-mesoscale eddies and mixed-layer
dynamics but are also very useful when evaluating the satellite products of
both present (Amores et al., 2018) and future generations, such as SWOT
projects (Uchida et al., 2022). Table 1 summarises recent developments in
high-resolution global ocean models. Not only is the model horizontal
resolution significantly increased, which now ranges from
Recent developments in global high-resolution ocean models.
Improving the representation of physical processes in ocean models has been
the most fundamental aspect for ocean general circulation model (OGCM)
development. Since the establishment of the first OGCM (Bryan and Cox,
1967), surface wave models, tide models, ocean internal wave models and
OGCMs have been separated into different streams (Mellor and Blumberg, 2004;
Qiao et al., 2004). The most uncertain term in all OGCMs is ocean turbulence.
As a result, the vertical structures of ocean temperature and salinity
cannot be accurately simulated or predicted. For example, the simulated
mixed-layer depth (MLD) in the upper ocean is too shallow, and the sea
surface temperature (SST) overheats during summer in nearly all OGCMs.
Teixeira and Belcher (2002) investigated the interaction of background
turbulence with a monochromatic irrotational surface wave and found that
the Stokes drift associated with the surface wave tilts vertical vorticity
into the horizontal direction, subsequently stretching it into elongated
streamwise vortices, and generates vertical mixing. Qiao et al. (2004, 2016)
proposed an upper-ocean mixing scheme of
Ocean tides have been recognised as a fundamental aspect regulating the
hydrodynamic environments in shallow regions (e.g. Simpson and Hunter, 1974;
Garrett et al., 1981; Holt and Umlauf, 2008; Lin et al., 2020); thus,
ocean tides are often included in regional ocean models. The tidal mixing
controls the formation of thermal fronts in coastal regions and generates
upwelling (Lü et al., 2006, 2008, 2010). The regional surface-wave–tide–circulation coupled model of the sea around China has shown excellent
performance in revealing the three-dimensional circulation structure of Yellow
Sea cold water mass (Xia et al., 2006) and has been applied in operational
ocean forecasting systems (Wang et al., 2016). In recent years, high-resolution global OGCMs forced by atmospheric forcings begins to include
ocean tide explicitly (Arbic et al., 2010, 2012, 2018; Rocha et al., 2016).
One of the benefits of including ocean tide in eddy-resolving global OGCMs
is that the global distribution of internal tide and mesoscale eddies can be
resolved concurrently (e.g. Arbic et al., 2010, 2012, 2018; Buijsman et al.,
2015; Shriver et al., 2012; Ansong et al., 2018; Timko et al., 2019). In
addition, inclusion of ocean tide in a global
To examine whether high-resolution OGCMs can faithfully reproduce the ocean
environment, the comparison of wavenumber and frequency spectra against
observations is widely adopted (e.g. Sasaki and Klein, 2012; Richman et al.,
2012; Chassignet and Xu, 2017, hereafter CX17; Savage et al., 2017a, b; Biri
et al., 2016). Therefore, the wavenumber spectral slope in the 70–250 km
mesoscale range becomes an important criteria for OGCMs' validation. Sasaki
and Klein (2012, hereafter SK12) noted that as the horizontal resolution
improved to
This paper aims to answer the following three key questions through
establishing a global
This paper is organised as follows. Section 2 describes the model
configurations and design of the numerical experiments. In Sect. 3.1, we
present the results of basic aspects of the new global
FIO-COM consists of a global ocean circulation model, a sea ice model
and a global ocean wave model. The ocean circulation component is based on
the Modular Ocean Model 5 (Griffies, 2012), the sea ice component is based
on the Sea Ice Simulator (Winton, 2000) and the surface wave component is
based on the MASNUM (laboratory of MArine Sciences and NUmerical Modelling
of the Ministry of Natural Resources of the People's Republic of China) ocean wave model (Yang et al., 2005; Qiao et al., 2017).
Currently, the data exchanged between the ocean wave and the ocean
circulation model are limited to
According to different research purposes, there are two different coupling strategies for the wave model component. Firstly, in the lower-resolution numerical experiments, where the computational costs are not expensive, the wave model component is coupled online with the ocean circulation and sea ice models. The real-time online data exchanges are achieved based on the subroutine version of the MASNUM wave model. In this way, the wave model component could be coupled with ocean circulation and ice components through direct calling of the MASNUM wave model as subroutines in the ocean circulation model. The ocean wave, ocean circulation and sea ice components share the same model grids. Secondly, in the numerical experiments with high computational costs and research focusing on ocean dynamics, the wave component can be turned off to save computational resources (the computational cost of the coupled ocean–ice model is approximately half that of the surface-wave–ocean–ice model). In this configuration, the surface-wave-induced mixing coefficients previously saved in data files are then read into the OGCM; the drawback is that the wave field cannot interact with the ocean circulation fields as they change.
The horizontal grid of FIO-COM32 is a tripolar grid (Murray, 1996) with a
horizontal resolution of
Ocean tide is explicitly included through introducing eight main tidal
generating potentials, i.e.
The initial conditions, including sea temperature, salinity, velocity and sea
surface height (SSH), are interpolated from outcomes of a global operational
ocean forecasting system based on FIO-COM10 with a horizontal resolution of
To investigate to what extent surface waves and tides contribute to the
newly established FIO-COM32 model and taking the extremely expensive
computational cost into consideration, two numerical experiments are
designed. In EXP1, only the OGCM is active, neither the ocean surface waves
nor the tides are included, and the simulated period is from 1 June 2016 to 31 December 2019. In EXP2, wave–tide–circulation coupling is enabled, which
means both
EXP1 starts from outcomes of the global FIO-COM10 forecast system on 1 June 2016. EXP2 branches off from EXP1 on 1 July 2017. The data analysis
focuses on the period from 1 January 2018 to 31 December 2019. The model
output frequency of three-dimensional variables is daily, and the output
frequency of SSH and steric SSH is hourly. An additional two experiments with a
horizontal resolution of
Another two numerical experiments, OGCM
In order to understand how the model resolution affects the simulated
barotropic tide, simulation results of the two global barotropic tide
models with horizontal resolutions of
In this section, the surface EKE and SSH values simulated by the
One of the most noteworthy elements of these simulations is that the EKE increases significantly as
the model horizontal resolution increases. As CX17 pointed out, this
increase is due to smaller effective horizontal viscosity and more active
mesoscale and sub-mesoscale motions of higher-resolution models. The
simulated surface EKE and satellite observations are compared here to
investigate the effects of improved horizontal resolution in FIO-COM32. The
surface EKE is calculated as follows:
Eddy kinetic energy (EKE) of CMEMS from all merged and
gridded satellite data
Time series of globally averaged EKE
Both the mean SSH and SSH standard deviation (SD) of the
Mean SSH
To better understand the effect of model resolution on the large-scale
circulations, we investigated the simulated main paths of the Kuroshio and
Gulf Stream. The annual mean SSH of
Mean SSH of the Kuroshio region. The route of the Kuroshio is
represented by the contour line at level of 0.16 m. CMEMS
The same as Fig. 4 but for the Gulf Stream region. Note that the ICRE of the Gulf stream recirculation is masked out to focus on the main path of Gulf Stream.
As the resolution increases from
Figures 6 and 7 show the simulated relative vorticity of sea surface current
in the
Relative vorticity (
The same as Fig. 6 but for winter (1 March 2019).
Reasonable global tide simulation is an important prerequisite for tides and
OGCMs to be able to be coupled, since tuning global barotropic tide can also yield
better model topography settings, especially as some numerically unstable
topography features can be fixed through this practice. The accuracy of
simulated global barotropic tide is quite sensitive to the model resolution
(Egbert et al., 2004). Here we show the effects of improving horizontal
resolutions from
For the
The simulated global tide of EXP2, which is a global baroclinic tide model,
is shown in Fig. 8f. The overall pattern of the M
In order to investigate whether the tide of the
Incoherent tide amplitude of semi-diurnal tidal band of the
Prior to examining the effects of surface-wave-induced mixing in the newly
established FIO-COM32 model, the accuracy of the ocean wave model is
evaluated. Figure 10 shows the comparisons of satellite data from Jason-3 and model-simulated significant wave height (SWH). From the comparison, the model
results are interpolated onto the satellite ground tracks. Both the
distributions and seasonal cycle can be reproduced well by the global
Along-track seasonal mean significant wave height of Jason-3
The ocean upper mixed layer is a crucial layer located between the ocean
interior and atmosphere. The MLD determines the heat and momentum content of
the upper-ocean boundary layer and is a key factor that reflects the
ability of an ocean model. Figure 11 shows the
derived summer MLD based on Argo observations (Fig. 11a) and EXP1 (Fig. 11b) and EXP2 (Fig. 11c)
simulations. The summer is the mean of July, August and September (JAS) for
the Northern Hemisphere and the mean of January, February and March (JFM)
for the Southern Hemisphere. The MLD is defined as the depth
at which potential density increases by 0.125 kg m
Summer (JAS and JFM for the Northern Hemisphere and Southern Hemisphere,
respectively) MLD based on Argo observations
Both EXP1 and EXP2 are able to reproduce the general patterns of summer MLD
distributions. However, the MLD of EXP1 without
Length scale of summer (JAS and JFM for the Northern Hemisphere and Southern Hemisphere,
respectively) surface mixed-layer instability (
As the model resolution is increased and global tide is explicitly included
in EXP2, the model is able to simulate global internal tide (IT), which is
activated when tidal currents flow over rough topography. Since a large
portion of barotropic-to-baroclinic energy transfer can be explicitly
resolved in EXP2, we disable the parameterised topography drag that has been
used in the barotropic model. Figure 13 compares the global internal tide
fields of satellite MIOST (Multivariate Inversion of Ocean Surface
Topography) (Ubelmann et al., 2022) and model simulations. We apply the
radial filter with a radius of 4
Previous studies (SK12, CX17) reported that the high-resolution model-simulated SSH wavenumber spectral slope in the 70–250 km mesoscale domain
is quite different from that of satellite observations. In the previous
models, the slopes range between
The SSH wavenumber spectral slope in the 70–250 km range is calculated as
follows. First, the model SSH data are interpolated onto Jason-3 ground
tracks. Second, the spectral slopes are calculated on a 1
SSH wavenumber spectrum slope of 70–250 km in the Jason-3 along-track
data
Figure 14a, b and c show SSH wavenumber spectral slopes of 70–250 km range of Jason-3 along-track filtered data for EXP1 and EXP2. Figure 14a resembles the results of Xu and Fu (2012), which shows the strong latitudinal variability in wavenumber spectral slopes in the satellite observations. Figure 14b shows the wavenumber spectral slopes of EXP1 have much weaker latitudinal variability, which resemble that of SK12 and CX17. The wavenumber spectral slopes of EXP2 show substantially improved agreement with those of satellite observations shown in Fig. 14a. The flattened wavenumber spectral slopes of EXP2 in low-EKE regions result from internal tide-induced SSH undulations that can be clearly observed in SSH snapshots (see Figs. 14g and h and the Hovmöller diagram in Fig. A5). These internal tide signals have spatial scales of tens to hundreds of kilometres and become nontrivial where the background eddy and circulation field is relatively weak. Figure 14d, e and f show wavenumber spectra at three representative sites. Three locations are chosen for representing typical conditions: (i) high EKE and active internal tides, (ii) typical equatorial regions, and (iii) high EKE but inactive internal tides. In the high-latitude and high-EKE regions, the spectra are less affected by the internal tide signals. However, in the low-latitude and low-EKE regions, the presence of internal tide substantially alters the spectra and yields much more realistic simulations. The textures of SSH in EXP1 and EXP2 shown in Fig. 14g and h manifest many differences that mainly result from the internal tide signals, and EXP2 would be much closer to reality. Therefore, for high-resolution global ocean models, tide–circulation coupling is important for more detailed and realistic SSH simulations. Although the slopes of EXP2 (Fig. 14c) are more in agreement with the observations than those of EXP1 (Fig. 14b), there are still some discrepancies, especially in the low-latitude regions of the Atlantic. In this region, the slope of EXP2 is apparently more flat than the observations, which may indicate that the internal tide is stronger here than normal. A more realistic internal tidal field may further improve the agreement between the model and observations.
SSH variance spectrum amplitude [m
To further explore how the tide–circulation coupling and the increased resolution affect the simulations of different processes (including the rotational processes of sub-mesoscale turbulence and divergent processes of inertial gravity waves, IGWs), the wavenumber–frequency spectra of SSH variance are calculated from the hourly outputs of EXP1, EXP2 and EXP2Low simulations. Focusing on the high-frequency sub-mesoscale processes, the wavenumber–frequency spectra are calculated every 30 d and average the results to obtain a spectral estimate. Before calculating the discrete Fourier transform (DFT), we remove linear trends and multiply the data using a three-dimensional Hanning window. Meanwhile, the dispersion curves of IGWs are estimated from the World Ocean Atlas 2013 climatological temperature–salinity profiles (Qiu et al., 2018).
With respect to sub-mesoscale IGWs, these are shown by clear discrete beams
above the 10th vertical normal mode curve in EXP2 (Fig. 15a–c),
especially in the subtropical gyre of the western Pacific, which is
invisible in EXP1 (Fig. 15d–f). By comparing the SSH wavenumber–frequency spectrum of EXP1 and EXP2, it shows that SSH wavenumber–frequency spectra are more energetic after including
tidal forcing in the
near-inertial, diurnal and semi-diurnal tidal frequency bands and
along the dispersion curves of discrete vertical modes of inertial-gravity
waves. This can further explain how the substantially enhanced unbalanced
motions (according to Chereskin et al., 2019, “unbalanced motions”
refers to the ageostrophic motions, but here the term mainly refers to internal tides and
inertia-gravity waves) are responsible for the better match of EXP2 to
Jason-3 than EXP1 in Fig. 14. As shown in Fig. 15a–b, in the regions
with active IGWs, the energy distribution extends continuously from the
sub-mesoscale regime to the IGW regime, which may indicate a strong
non-linear interaction between them. However, in the low-resolution experiments
shown in Fig. 15g–h there is either a gap between the sub-mesoscale and IGW regimes
or insufficient sub-mesoscale activities, which may
limit the non-linear interaction between them. It is obvious that SSH variance density is more energetic in the
sub-mesoscale regime as resolution increases (
In this paper, for the first time a global
In future work, the effects of sea surface currents (SSCs) on surface wave simulation need to be taken into consideration for a fully coupled wave–circulation model, as this phenomenon has been shown to have impact on surface wave simulation (Ardhuin et al., 2017). As has been stated in this paper, the online coupled surface-wave–ocean circulation version of the FIO-COM32 is turned off pro tempore and is ready to be implemented when the demands for greater computational resources are met in the near future, and at that time the effects of ocean currents on the surface waves will be explored in depth. Figure 16 shows an example of the effect of SSCs on the simulated significant wave height. The hourly SSC output from FIO-COM32 in the western Pacific is fed into the MASNUM wave model to test SSC effects on surface waves. It is clear that SSCs have a notable influence on the simulated significant wave height, especially around the Kuroshio. In MASNUM wave model, SSCs influence the surface wave simulations through advection, refraction and the energy source function due to interaction between surface waves and SSC shear (Yang et al., 2005).
Since the internal tide can be resolved in the FIO-COM32 model, finding a more
robust representation of energy dissipation of internal tide propagation is
still an open task that needs to be addressed in the future. A proper
dissipation scheme of the internal tide and its adaption to the
traditional viscosity schemes for OGCM is still a daunting challenge. We
have conducted numerical experiments (not shown) with an inordinately large
background vertical viscosity of 1.0 m
A better prediction of ocean is the inexhaustible goal of the ocean model community. As the ocean, especially the upper ocean, plays a dominant role in the climate system, ocean model improvement can shed light on new generations of climate model development. Surface waves, tides and ocean circulations are often separately simulated by different numerical models. In this paper, we show that surface wave–tide–circulation coupling can dramatically improve simulations. Therefore, it is time for us to regard the ocean as a fully coupled dynamic system through the key conjunction of turbulence and develop fully coupled surface-wave–internal-wave–tide–circulation models for the next generation of ocean model development.
Since the time span of the numerical experiments is not long (3.5 years),
the model drift is generally quite small. The model drift in temperature
shows a strong seasonal cycle in the upper ocean and a warming trend in the
sub-surface layer, and the maximum drift in temperature at the end is about
0.2
Time series of simulated total kinetic energy of EXP1.
Figure A1 shows the total kinetic energy of EXP1 since the beginning of the
numerical experiment. The
Model drift of temperature
Hovmöller diagram of surface relative vorticity along a section in the Kuroshio Extension region, marked as black line shown in Fig. 6a.
Temperature along the section in the Kuroshio Extension region of
Hovmöller diagram of steric SSH along the section of
135
The exact version of the model, input data used to produce the results in
this paper and data to produce the plots are archived on Zenodo
(
BX was responsible for model development and drafted the manuscript. FQ designed the road map of this research and supervised the writing of the paper. All authors contributed to discussions, data analysis and writing of the paper.
The contact author has declared that none of the authors has any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research is jointly supported by the National Natural Science Foundation of China under grant no. 41821004 and the Marine S&T Fund of Shandong Province for the Pilot National Laboratory for Marine Science and Technology (Qingdao) (grant no. 2018SDKJ0106-1). This work is a contribution to the UN Decade of Ocean Science for Sustainable Development (2021–2030) through both the Decade Collaborative Centre on Ocean-Climate Nexus and Coordination Amongst Decade Implementing Partners in P.R. China (DCC-OCC) and the approved programme of the Ocean to climate Seamless Forecasting system (OSF).
This research is jointly supported by the National Natural Science Foundation of China under grant no. 41821004 and the Marine S&T Fund of Shandong Province for the Pilot National Laboratory for Marine Science and Technology (Qingdao) (grant no. 2018SDKJ0106-1).
This paper was edited by Riccardo Farneti and reviewed by two anonymous referees.