HTML
--> --> -->Currently, there are large simulation spreads at centennial and longer time scales in the ocean modeling community (Griffies et al., 2009). To compare the behaviors of different ocean models and understand the origin of model biases in phase 6 of the Coupled Model Intercomparison Project (CMIP6), the Ocean Model Intercomparison Project (OMIP), an endorsed sub-project of CMIP6, was proposed (Griffies et al., 2016), which provides an experimental protocol for global ocean/sea-ice simulations forced with common atmospheric datasets. In the OMIP experiments, the ocean model is coupled with the sea-ice model forced by prescribed atmospheric fields. There are 11 model groups participating in the OMIP experiments according to Tsujino et al. (2020). The Co-ordinated Ocean–Ice Reference Experiment II (CORE II) datasets (Large and Yeager, 2009) are applied to force the global ocean/sea-ice models, which is denoted as phase 1 of the physical part of OMIP (OMIP1). Recently, a new OMIP experiment has been proposed, applying Japanese 55-year Reanalysis (JRA-55; Kobayashi et al., 2015) to drive the ocean–ice model, which is denoted as OMIP2. The period covered by the JRA-55 datasets is 1958–2018. The OMIP experiments are regarded as a fundamental test for ocean modelers (Griffies et al., 2016). Because the OMIP1 experiments are conducted by the ocean–ice coupled model component of the Chinese Academy of Sciences’ Earth System Model, version 2.0 (CAS-ESM2.0), we mainly focus on OMIP1 in this paper.
CAS-ESM2.0, which will take part in CMIP6, is a newcomer in the modeling community (Zhang et al, 2020). The ocean component of CAS-ESM2.0 is a revised version of the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics/Institute of Atmospheric Physics (LASG/IAP) Climate Ocean Model, version 2 (LICOM2.0) (Liu et al., 2012). The coupled ocean–ice component of CAS-ESM2.0 will participate in the OMIP1 experiment, and datasets will be released to CMIP6 for researchers worldwide. Here, a baseline evaluation of the coupled ocean–ice component of CAS-ESM2.0 in the OMIP1 experiment is presented.
The remainder of the paper is structured as follows: Section 2 describes the model and experimental design; section 3 gives a basic technical validation of OMIP1; a summary is provided in section 4; and section 5 offers usage notes.
| Level | Depth for T grid | Depth for W grid |
| 1 | ?5 | 0 |
| 2 | ?15 | ?10 |
| 3 | ?25 | ?20 |
| 4 | ?35 | ?30 |
| 5 | ?45 | ?40 |
| 6 | ?55 | ?50 |
| 7 | ?65 | ?60 |
| 8 | ?75 | ?70 |
| 9 | ?85 | ?80 |
| 10 | ?95 | ?90 |
| 11 | ?105 | ?100 |
| 12 | ?115 | ?110 |
| 13 | ?125 | ?120 |
| 14 | ?135 | ?130 |
| 15 | ?145 | ?140 |
| 16 | ?156.9303 | ?150 |
| 17 | ?178.4277 | ?163.8606 |
| 18 | ?222.5018 | ?192.9948 |
| 19 | ?303.1057 | ?252.0088 |
| 20 | ?432.5961 | ?354.2027 |
| 21 | ?621.1931 | ?510.9896 |
| 22 | ?876.5334 | ?731.3966 |
| 23 | ?1203.337 | ?1021.67 |
| 24 | ?1603.2 | ?1385.003 |
| 25 | ?2074.526 | ?1821.396 |
| 26 | ?2612.596 | ?2327.656 |
| 27 | ?3209.772 | ?2897.536 |
| 28 | ?3855.835 | ?3522.009 |
| 29 | ?4538.428 | ?4189.662 |
| 30 | ?5243.597 | ?4887.194 |
| 31 | ? | ?5600 |
Table1. T and W grids denote different variables locations in the vertical grid, the temperature T, salinity S and horizonal vectors (u, v) are located in T grid, and the vertical velocity W and pressure P are located in W grids. Negative value is the depth below Level 1 of W grid. (Units: m).
(1) A new sea surface salinity (SSS) boundary condition has been introduced based on the physical process of air–sea flux exchange at the actual sea–air interface, rather than the commonly used virtual salt flux scheme. The new scheme can reproduce the magnitude of Atlantic Meridional Overturning Circulation (AMOC) better than the original one (Jin et al., 2017).
(2) The diurnal variation of sea surface temperature (SST) is resolved by coupling the atmospheric and oceanic model components once every two hours.
(3) A new formulation of turbulent air–sea fluxes (Fairall et al., 2003) has been introduced, which improves the latent heat flux and wind stress simulations. In the new scheme, the effects of gustiness are parameterized, and the thermal stability of the marine atmosphere boundary layer is considered in detail (Pelletier et al., 2018).
The sea-ice model in CAS-ESM2.0 is an improved Los Alamos sea-ice model, version 4.0 (Hunke and Lipscomb, 2008), using the same grid as the oceanic model. The model solves the dynamic and thermodynamic equations for five ice thickness categories, with one snow layer and four ice layers. For the dynamic component, the model uses elastic–viscous–plastic rheology (Hunke and Dukowicz, 1997), a mechanical redistribution scheme (Lipscomb et al., 2007), and an incremental remapping advection scheme (Lipscomb and Hunke, 2004). For the thermodynamic component, a parameterization with a relatively more realistic sea-ice salinity budget (Liu, 2010) is used. Also employed is an improved CCSM3 radiation scheme for the albedo and radiative fluxes at the sea-ice surface, which works by incorporating explicit melt ponds and their impact on albedo, expanding the spectrum from two to four bands, distinguishing direct and diffuse short waves, and considering different albedos for wet and dry snow, as well as sea-ice and snow thickness effects on albedo.
The OMIP1 experiment follows the CORE-II protocol, and the initial conditions of the coupled ocean–ice component of CAS-ESM2.0 are from the World Ocean Atlas (Antonov et al., 2006, 2010) temperature and salinity and state of rest. The SSS is restored to monthly climatological observations from the World Ocean Atlas (Antonov et al., 2006, 2010), with a restoring velocity of 50 m (4 yr)?1 for the sea surface and 50 m (30 d)?1 for the sea ice.
The forcing fields in OMIP1 are CORE-II datasets, which are mainly derived from the NCEP–NCAR atmospheric reanalysis. The forcing dataset covers the 62-year period from 1948 to 2009, and the temporal resolution of CORE-II is 6 hours and the horizontal resolution is T62. The atmospheric forcing from the CORE-II datasets, including atmospheric wind vectors at 10 m, surface air temperature at 10 m, specific humidity at 10 m, air density at 10 m, precipitation including rain and snow, surface downward shortwave and longwave radiation, sea level pressure, and runoff, is prescribed. To derive quasi-equilibrium and stable results, the ocean–sea ice coupled model is integrated for 372 years (six cycles of the CORE-II period). In the following section, we present a baseline evaluation of the OMIP1 simulation by the coupled ocean–ice component of CAS-ESM2.0.
2
3.1. Global mean
First, to evaluate the stability of the ocean model integration, the time series of global annual mean temperature and salinity at the surface and throughout the whole ocean are shown in Fig. 1. The model integration covers 372 years, including 6 cycles of the observational period of 62 years (1948–2009). This cyclic feature is clear in the global mean SST and SSS. Signatures of interannual variations are also apparent at the sea surface, with temperature (salinity) fluctuating at approximately 18.40°C (34.7–34.8 psu). The global mean SST and SSS at the mean time of the final cycle is 18.41°C and 34.73 psu, respectively, and the corresponding observational values are 18.39°C and 34.73 psu. For the global volume-averaged ocean temperature (VOT), the simulated VOT shows a near stationary state (~3.83°C) throughout the entire integration, indicating that the model reaches quasi-equilibrium and that the drift of sea surface heat flux can be negligible.
Figure1. Time series of global annual mean (a) SST, (b) VOT, (c) SSS, and (d) AMOC maximum values at 26.5°N in the CAS-ESM2.0 OMIP1 experiment. The six cycles are separated by vertical gray lines.We also provide the time series of the AMOC maximum at 26.5°N in the full integration. The time series exhibits an evident period of 62 years, similar to the time series of SST and SSS. This feature can be reasonably reproduced by the model in each atmospheric forcing cycle. The drift is very small after the adjustment of the first cycle, which further confirms the stability of the model integration. The simulated magnitude of the AMOC maximum at 26.5°N at the mean time of the final cycle is approximately 17.7 Sv (Sv = 106 m3 s?1).
2
3.2. Spatial patterns
Figure 2 shows the spatial patterns of SST and SSS biases compared with observations. The last 30 years of integration is shown. In general, the model can reasonably reproduce the global patterns of SST and SSS. The global mean biases of SST and SSS are 0.07°C and 0.04 psu, respectively. The simulated SST bias exhibits common anomalous patterns in many climate models (Golaz et al., 2019; Held et al., 2019), with opposite signs between sub-polar regions (a prominent colder bias in the Arctic Ocean and warmer bias in the Southern Ocean). Overall warm biases are found in coastal upwelling regions, such as the western coasts of the continents of America and Africa, and the eastern equatorial Pacific, which may be associated with poorly resolved coastal upwelling due to coarse grid resolution. Cold anomalies are mainly concentrated in subtropical regions, especially the margins of western currents. For the SSS bias, the freshwater bias is in the high latitudes and western equatorial Pacific, while the saline water SSS bias is located in the tropical Atlantic Ocean and Beaufort Sea in the Arctic Ocean. The root-mean-square errors of SST and SSS are 0.65°C and 0.49 psu, respectively.
Figure2. Spatial pattern of simulated bias in (a) SST (units: °C) and (b) SSS (units: psu); the contours in (a, b) are simulated SST and SSS values, respectively. The simulation time mean is taken from the last 30 years of integration.The observed and simulated climatological annual mean SSH are shown in Fig. 3. The simulation can reasonably capture the spatial features of SSH from observations, with a large pattern correlation coefficient of 0.96 between them. In both the simulation and observation, the high SSHs are located around the western boundary in the low and middle latitudes of the Pacific and Atlantic oceans in both hemispheres, while the low ones are close to the eastern boundary, which is associated with the westward ocean currents and trade winds. In addition, a salient asymmetric feature exists; namely, a ridge of SSH in the northern western Pacific that is higher than its counterpart in the Southern Hemisphere. In the Indian Ocean, the east–west contrast reverses in the tropical region, with higher SSH located in the east and lower in the west. In high latitudes, the relatively low SSHs are located in the subpolar gyre in the Northern Hemisphere. The lowest SSHs in the global ocean are located around the Antarctic Circumpolar Current in the Southern Ocean.
Figure3. Climatological annual mean SSH (units: m) in the (a) observations and (b) CAS-ESM2.0 OMIP1 simulation. The pattern correlation coefficient between the observations and simulation is shown in the top-right corner in panel (b).2
3.3. Zonal mean
We further show the distribution of the zonal mean ocean temperature and salinity biases of different basins in Fig. 4, derived from the last 30 years of integration. The global ocean is divided into three basins: the Southern Ocean (78°S–35°N); the Indo-Pacific Ocean (35°S–65°N, 22°–134°E); and the Arctic and Atlantic oceans (35°S–90°N). In the Southern Ocean, the zonal mean temperature bias exhibits a south-warm–north-cold dipole pattern in the upper 1000 m. Below 1000 m, the model simulation shows a warm bias. In the Indo-Pacific basin, the pattern of the warm bias in the North Pacific has its largest bias at depths of 100–200 m, and the surface cold bias in its southern region expands downward and dominates the entire basin below 1000 m. With respect to the Atlantic–Arctic Ocean basin, it generally shows an overly warm bias, especially from the depth of 600 to 3000 m, with freshwater biases in the Arctic Ocean in the upper 300 m (Fig. 4b). Compared to that in the Atlantic–Arctic Ocean basin, the salinity bias in the Southern Ocean and Indo-Pacific Ocean basin is minimal. Most ocean areas in the Southern Ocean and Indo-Pacific Ocean have positive salinity biases, except in the upper northern region in the Southern Ocean and subsurface tropical Indo-Pacific regions, where cold biases occur (Fig. 4b).
Figure4. Latitude–depth plot of the zonal mean (a) ocean temperature bias (units: °C) and (b) salinity bias (unit: psu) in the Southern Ocean, Indo-Pacific Ocean, and Atlantic–Arctic Ocean basins. The simulation time mean is taken from the last 30 years of integration. The unit on the y-axis is km.2
3.4. AMOC
In the Atlantic, the most active thermohaline circulation occurs in the global ocean, which can be represented by the AMOC. The AMOC can portray the transport of volume or mass in the Atlantic. AMOC simulation is one of the key indicators used to measure the performance of an ocean model. Figure 5a shows the AMOC simulation in the ocean component of CAS-ESM2.0 over the final decade (2000–09) mean of the final cycle. The model can reasonably reproduce the North Atlantic Deep Water (NADW) between 500 m and 3000 m north of 35°S, and the Antarctic Bottom Water below 3000 m. Compared with the observations at 26.5°N, the model can simulate the peak value at the right depth of approximately 1000 m, with magnitudes up to 18.6 Sv, which is slightly greater than the 18.3 Sv inferred from observations for 2004–09 (Cunningham et al., 2007; Smeed et al., 2018). The simulated AMOC depth is largely limited to the top 3000 m, which is slightly shallower than the 4300 m in observational estimates (McCarthy et al., 2015).
Figure5. (a) The AMOC stream function (units: Sv) averaged over the last 10 years of integration. (b) Vertical profiles of the time-averaged circulation at 26.5°N in the simulation (red) and observations (black).2
3.5. Sea ice
In the OMIP experiment, the sea-ice model is coupled with the ocean model and can interact with the ocean model; thus, the sea-ice concentration (SIC) is another metric used to evaluate the model results. Figure 6 shows the spatial pattern and temporal evolution of the simulated sea-ice area in the Arctic and Antarctic. The coupled ocean and sea-ice model can reasonably reproduce the SIC in both the Northern and Southern hemispheres (Figs. 6a–d). For the wintertime in both hemispheres (March in the Northern Hemisphere and September in the Southern Hemisphere), the SIC is overestimated in the Iceland and Bering seas in the Northern Hemisphere and almost the entire Antarctic region in the Southern Hemisphere, as indicated by the difference in the black and purple contours that represent the threshold of 15% coverage. Similarly, the slightly underestimated SIC in summertime in both hemispheres (September in the Northern Hemisphere and March in the Southern Hemisphere) is mainly located in the Chukchi Sea, East Siberian Sea, Laptev Sea, Canada basin, Ross Sea, and Weddell Sea. Figure 6e provides the time series of the maximum and minimum sea-ice areas in both hemispheres, which is also an important measure to show the stability of the model during integration. The simulated sea-ice area in the two hemispheres drifts minimally in the seasonal cycle of the sea-ice area during the whole integration, with interannual variation and 62-year atmospheric forcing cycles. In addition, it is obvious that the difference between the maximum and minimum sea-ice areas is larger in the Antarctic than in the Arctic, which indicates a larger seasonal cycle in the Antarctic than in the Arctic.
Figure6. (a–d) Spatial patterns of the simulated SIC in (a, c) March and (b, d) September, in (a, b) the Northern Hemisphere and (c, d) the Southern Hemisphere. The thick black (purple) line denotes the 15% contour line in the observations (simulation). The simulation time mean is taken from the last 30 years of integration. (e) Time series of the maximum and minimum of the total sea-ice area for the Arctic and Antarctic in the simulation (units: 106 km2). The black (red) line denotes the Arctic (Antarctic). The six cycles are separated by vertical gray lines.2
3.6. Comparison of coupled model behaviors
According to the above results, the behaviors of the coupled ocean–ice model forced by prescribed CORE-II data can reasonably reproduce the preliminary performances of the ocean. In addition to the OMIP1 experiment, CAS-ESM2.0 also conducted the CMIP6 historical experiment (see footnote). Understanding the behaviors in standalone component models is an essential part of investigating the bias of the fully coupled model. Actually, on the whole, the ocean–ice model has similar bias patterns as the coupled model simulation, indicating to some extent that the coupled simulation bias has some origin in the ocean model. Taking the zonal mean temperature simulation as an example, the coupled model also has a warm bias throughout the South Indian Ocean, in the upper level of the North Pacific Ocean, and in the NADW, and a cold bias at depths of ~800 m in the North Indian Ocean and tropical Pacific Ocean, as the ocean model shows in Fig. 4. In some respects, the bias in the ocean model is decreased compared with the results from the coupled model, e.g., the subsurface cold biases in the Pacific Ocean and Indian Ocean are substantially reduced in the ocean model compared with the coupled model, suggesting that the bias in the atmospheric variables in the coupled model can enlarge the ocean model bias through air–sea coupling. Systematic evaluation of the coupled simulation of CAS-ESM2.0 has been provided in Zhang et al. (2020).| Name | Description | Units |
| hfbasin | Northward ocean heat transport | W |
| hfbasinpmadv | Northward ocean heat transport due to parameterized mesoscale advection | W |
| hfds | Downward heat flux at sea water surface | W m?2 |
| mlotst | Ocean mixed-layer thickness defined by sigma T | m |
| msftbarot | Ocean barotropic mass streamfunction | Sv = 106 m3 s?1 |
| msftmz | Ocean meridional overturning mass streamfunction | Sv = 106 m3 s?1 |
| obvfsq | Square of Brunt–Vaisala frequency in sea water | s?2 |
| pbo | Seawater pressure at sea floor | Pa |
| so | Seawater salinity | psu |
| sob | Seawater salinity at sea floor | psu |
| soga | Global mean sea water salinity | psu |
| sos | SSS | psu |
| sosga | Global average SSS | psu |
| thetao | Seawater potential temperature | °C |
| thetaoga | Global average sea water potential temperature | °C |
| tob | Seawater potential temperature at sea floor | °C |
| tos | SST | °C |
| tosga | Global average SST | °C |
| uo | Seawater X velocity | m s?1 |
| umo | Ocean mass X transport | kg s?1 |
| vo | Sea water Y velocity | m s?1 |
| wo | Seawater vertical velocity | m s?1 |
| wmo | Upward ocean mass transport | kg s?1 |
| zos | SSH above geoid | m |
Table2. Descriptions of Priority 1 dataset variables for OMIP.
Acknowledgements. We greatly appreciate the constructive comments and suggestions from the two reviewers and Executive Editors-in-Chief. This work was jointly supported by the National Natural Science Foundation of China (Grant Nos. 41706036 and 41706028), the Key Research Program of Frontier Sciences, Chinese Academy of Sciences (Grant No. QYZDY-SSW-DQC002), the National Key R&D Program for Developing Basic Sciences (Grant Nos. 2016YFC1401401 2016YFC1401601 and 2016YFB0200804), the National Key Scientific and Technological Infrastructure project entitled “Earth System Science Numerical Simulator Facility” (EarthLab) and key operation construction projects of Chongqing Meteorological Bureau-"Construction of chongqing short-term climate numerical prediction platform”.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (
