Keywords

1 Introduction

Global ocean general circulation models, which are indispensable components of climate system models, are useful tools for investigation of both ocean circulation and climate. The state-of-the-art climate and ocean models used in phase five of the coupled model intercomparison project (CMIP5) typically have horizontal resolutions as coarse as around 1° (approximately 110 km at the equator). Although some higher resolution basin-scale (e.g. Tokmakian and McClean 2003) or even global ocean models (e.g. Masumoto et al. 2004) already exist, they are impractical for use in integration of climate over hundreds or thousands of years owing to their huge computational costs.

However, some important oceanic processes that are crucial to the climate state of the ocean cannot be resolved explicitly at a resolution of around 1°. Of these processes, mesoscale eddies in the ocean are particularly important as they contain 90 % of the kinetic energy of the ocean. Owing to differences between the properties of seawater and those of the atmosphere, the first-order baroclinic Rossby radius of deformation is one order smaller in the ocean than in the atmosphere. Therefore, the characteristic length scale of mesoscale eddies is about O (100 km) or less. Therefore, a resolution of approximately 0.1° is required to resolve the mesoscale eddies in the ocean. The effects of the eddies on heat and salt transport are typically parameterized in coarse-resolution models using the scheme of Gent and McWilliams (1990) or similar. Recently, Hallberg and Gnanadesikan (2006) demonstrated that some physical processes are fundamentally different in models with explicitly resolved mesoscale eddies compared to those with parameterized eddies. However, the effects of mesoscale eddies on the mean circulation and its parameterization schemes must be investigated further.

Also important are the western boundary currents (WBCs), which play an important role in the transport of heat from the tropics to the high latitudes. Additionally, the WBCs also affect local climate along their path dramatically. The spatial scale of WBCs is around 100 km. The coarse-resolution models usually simulate a weak poleward heat transport, with large temperature biases along the western boundaries of the Pacific and Atlantic and overshooting of the WBCs. The WBC regions also display the greatest biases in air-sea coupled models. However, the WBCs typically receive no special treatment in ocean models and air-sea coupled models.

The horizontal resolution of the ocean model is also important for current systems in areas of complex land–sea distribution. For example, the Indonesian throughflow (ITF), which transports mass, heat, and salt from the Pacific to Indian oceans through the Indonesian seas. More than ten thousand islands exist in this region, and the main straits and passages of ITF cannot be resolved by coarse-resolution models. Previous studies have shown that coarse-resolution models tend to simulate inaccurate partition of transport through each strait or passage (e.g. Liu et al. 2005).

Furthermore, the coarse-resolution models also cannot simulate several other processes, including deep convection, frontal processes, internal waves, and tropical instability waves. Therefore, before resolving these processes completely in existing climate models, it is essential to understand the uncertainties caused by such unresolved processes.

The scientists at the State Key laboratory of numerical modeling for atmospheric sciences and geophysical fluid dynamics (LASG), institute of atmospheric physics (IAP) have been devoted to independently developing a global ocean general circulation model (OGCM) since the late 1980s. Moreover, increases in available computing resources have allowed continuous increases in the resolution of the model. Zhang et al. (1996) developed the first OGCM in China. This OGCM had a horizontal resolution of 4° × 5° and had four layers in the vertical; the second model (Jin et al. 1999) had a horizontal resolution of 1.875° and 30 levels in vertical. This model was subsequently renamed as the LASG/IAP climate system ocean model (LICOM) and its resolution was increased to 0.5° (Liu et al. 2004a). Recently, LICOM version 2 (LICOM2), which has a horizontal resolution of about 1° (Liu et al. 2012a), has been coupled with other component models and used to conduct core scientific experiments for CMIP5 (Li et al. 2013c; Bao et al. 2013). Simultaneously, an eddy-resolving model was developed at LASG based on LICOM2 (Yu et al. 2012a). A 20 year experiment was then conducted, forced by daily mean heat, and freshwater data from COREs (Large and Yeager 2004) and QuikSCAT surface wind stresses (CERSAT-IFREMER 2002).

We now believe that, although it is impossible to conduct climate simulations regularly with eddy-resolving OGCMs, it is necessary to understand the role of model resolution in simulation of the mean climate ocean state. The present study aims to briefly compare the coarse- and high-resolution simulations of LICOM2.0 with respect to the above mentioned variables. Section 38.2 describes model experiments and the observational datasets. The results are presented in Sects. 38.3 and 38.4 contains the concluding remarks.

2 Model Description and Datasets

The fundamental framework of LICOM2 is inherited from LICOM1. The primary features of LICOM2 include the η-coordinate system, free surface, primitive equation, and mesoscale eddy parameterization from Gent and McWilliams (1990) and similar studies. The main improvements offered by LICOM2 are as follows: increased vertical resolution in the upper 150 m; increased horizontal resolution in the tropics (10°S–10°N); introduction of the vertical mixing parameterization scheme of Canuto et al. (2001); tapering of mesoscale eddy-induced heat transport in the upper mixed layer, following Large et al. (1997); introduction of a two-exponent chlorophyll-dependent equation for solar radiation penetration; and introduction of a two-step shape-preserving advection scheme. Further details of all improvements can be found in the appendix of Liu et al. (2012a). A 500 year ocean-only experiment has been conducted with forcing by the corrected normal year forcing (NYF) data of COREs (Large and Yeager 2004). Details of this experiment and results are presented in Chap. 3 of Part I of this monograph.

Based on the LASG/IAP Climate System Ocean Model version 2.0 (LICOM2.0; Liu et al. 2012a), a quasi-global (excluding the Arctic Ocean) eddy-resolving OGCM has been built in LASG. The primary updates and improvements are as follows. (1) The horizontal grids are increased to 1/10°. (2) The number of vertical layers is increased: LICOM2 has 55 vertical layers, and the thickness of the first layer is 5 m. The upper 300 m has 36 uneven layers, and the mean thickness is less than 10 m. (3) The barotropic and baroclinic split methods are improved. (4) The parallel domain partition is changed from a one-dimensional (1D) message passing interface (MPI) meridional split to a mixed two-dimensional (2D) MPI and open multi-processing (OpenMP) methods. The OpenMP parallel algorithm is also optimized. Additional changes include the following. (1) The model domain spans 66°N–79°S, thus excluding the Arctic Ocean. (2) Biharmonic viscosity and diffusivity schemes are used for the horizontal directions in the equations of momentums and tracers (temperature and salinity), respectively. Meanwhile, the parameterization of mesoscale eddies from Gent and McWilliams (1990) is turned off in the equations of tracers.

The eddy-resolving OGCM is spun up for 12 years from zero velocity, initialized from the observed temperature and salinity from the world ocean atlas 2005 (WOA05; Locarnini et al. 2006; Antonov et al. 2006) and forced by the climatological monthly wind stress and heat fluxes. The forcing data are from the ocean model intercomparison project (OMIP; Röske 2001), which was derived from ERA15 reanalysis data (Gibson et al. 1997b). Moreover, the simulated sea surface salinity (SSS) is restored to the climatological monthly SSS, also from WOA05. Because the northern open boundary is set at 66°N, the simulated temperature and salinity are restored to the climatological monthly temperature and salinity from WOA05, whereas the solid wall boundary condition is used for velocity. After the 12 year spin-up numerical simulations, the eddy-resolving OGCM is integrated for an additional 8 year period starting from the end of the 12th year of the spin-up simulation. The model is forced by the daily QuikSCAT wind stress for 2000–2007 and by surface heat and freshwater fluxes derived from COREs (Large and Yeager 2004) for the same period. The detailed algorithm used to calculate the turbulent surface fluxes can be found in Large and Yeager (2004). Because the sea ice model is not included in LICOM2.0, the sea ice concentration from the Hadley dataset (Rayner et al. 2006) is used to calculate the surface fluxes. If the temperature of the seawater is below the freezing point (−1.8 °C), the temperature is restored to −1.8 °C. The 8 year simulations are used for analysis in the present study. For comparison with the observations, results from the same periods are selected to calculate the annual mean and “climatology” values.

The basic performance of the high-resolution LICOM2.0 has been evaluated previously by Yu et al. (2012a). The model better simulates the spatiotemporal features of mesoscale eddies and the paths and positions of western boundary currents compared to the previous version; moreover, it also reproduces the large meander of the Kuroshio Current and its interannual variability. The complex structures of equatorial Pacific currents and currents in the coastal ocean of China are also better captured owing to the increased horizontal and vertical resolution. The performance of the high-resolution LICOM2.0 in simulation of the ITF has been investigated previously by Feng et al. (2013) against the velocities at five main straits from the program of international nusantara stratification and transport (INSTANT, Sprintall et al. 2004; Gordon et al. 2010).

Four datasets are used here for model validation. The first is the weekly altimeter-derived sea surface height anomaly (SSHA) dataset obtained from the maps of sea level anomalies (MSLA) product for 2000–2007. The horizontal resolution of this dataset is 1/4°. The second dataset comprises the upper layer currents of version 2 of the simple ocean data assimilation (SODA) reanalysis (Carton and Giese 2008). The third dataset includes the zonal velocities in the equatorial Pacific Ocean, obtained from Johnson et al. (2002). The final dataset comprises the volume transports of five main straits in the Indonesian seas during January 2004 to December 2006; these data were measured as part of a multinational field program known as INSTANT program (Sprintall et al. 2004; Gordon et al. 2010).

3 Results

3.1 Mesoscale Eddies

The models investigated here, which have horizontal resolutions of around 0.1° or more, are typically referred to as eddy-resolving models, as they can explicitly resolve ocean mesoscale eddies. It is important to understand how such models actually simulate mesoscale eddies. Owing to their typically small scales, it was impossible to capture ocean mesoscale eddies until the development of remote sensing technology in the 1990s. Usually, the rootmeansquare (RMS) of the SSHA from satellite altimetry is used to measure the magnitude of ocean mesoscale eddies. Figure 38.1 illustrates the RMSs of the SSHA derived from satellite altimeters, eddy-resolving and coarse-resolution OGCMs. The altimeter SSHA was derived from the weekly MSLA product for 2000–2007, which has a horizontal resolution of 1/4°. The daily mean model results are also interpolated onto the same grid and averaged on a weekly basis. High RMSs, which indicate active eddies, can be found along the western boundary of all basins and always track along the WBCs and their extensions in the observations. Eddies can also be seen in the vicinity of the antarctic circumpolar current (ACC), which exhibits its greatest magnitudes along the coast of southern Africa.

Fig. 38.1
figure 1

RMS of weekly SSHA (cm) from a satellite altimeter from 2000-01-02 to 2007-12-19 (MSLA) and from eddy-resolving LICOM2 (1/10°) and coarse-resolution LICOM (1°) between 2000-01-01 and 2007-12-31. The daily SSHA is also used in the coarse-resolution OGCM. This figure was taken from Yu et al. (2012)

Full size image

The coarse-resolution model detects only few signs of eddies. The RMSs in the Kuroshio Current and Gulf Stream regions are only around one third the magnitudes of those in the observations. Compared with the coarse-resolution model, the eddy-resolving version of LICOM2.0 reproduces the distribution of eddies very well. In fact, it captures not only the most active regions (i.e. the western boundaries of all basins and the ACC region), but also the less active regions such as the region surrounding the Subtropical Countercurrent (20–30°N) in the Pacific. The magnitudes of the RMSs of SSHA in the high-resolution LICOM2.0 are slightly larger than those in the observations; this is a common problem in eddy-resolving models and is usually attributed to improper use of horizontal viscosity schemes or coefficients.

Yu et al. (2012) investigated the spatial scale of mesoscale eddies in the region of the WBCs and their extensions using the transient potential temperature at 50 m and found both cyclonic and anticyclonic eddies. The radii of these eddies were all around 100 km, corresponding to the scale of those in the satellite observations.

3.2 Western Boundary Currents

The Kuroshio Current in the North Pacific and the Gulf Stream in the North Atlantic are two of the most important WBC systems. They not only affect local climate, but also transport heat from the tropics to the high latitudes. Therefore, they play a crucial role in the global heat budget. Figure 38.2 illustrates these currents averaged over the upper 100 m for the Kuroshio Current (left column) and the Gulf Stream (right column) regions for the SODA product and the high- and low-resolution versions of LICOM2.0. In the reanalysis data, these two strong narrow currents flow northeastward along the boundary and then turn eastward at about 40°N, with magnitudes often exceeding 1 ms−1.

Fig. 38.2
figure 2

Annual mean speed (shaded, ms−1) and currents (vectors, ms−1) averaged over the upper 100 m for (a, d) the SODA product, (b, e) the eddy-resolving LICOM2, and (c, f) the medium-resolution LICOM2 in (a–c) the North Pacific western boundary region and (d–f) the North Atlantic western boundary region

Full size image

The coarse-resolution models typically result in large errors when reproducing the WBCs. In particular, such models tend to produce wide and weak currents and omit detailed features such as meanders and steady eddies. Moreover, the currents in the coarse-resolution models tend to overshoot those in the observations, typically by 2–3° (further north). Such errors in surface currents can induce large errors in temperatures and salinities (not shown). These aspects are all considerably improved in the eddy-resolving model, such that the width and strength of the simulated currents are both comparable with those of the SODA data. LICOM can even reproduce the detailed characteristics of the currents, including the meander along the coast of southern Japan and two meanders in its extension, among others. The separation points of both the Kuroshio Current and the Gulf Stream occur at about 40°N in the high-resolution LICOM2.0 run; this also corresponds closely to the SODA data, indicating that model resolution is one of the important factors controlling the location of current separation points.

Figure 38.2 also demonstrates clearly that the high-resolution LICOM2.0 overestimates the mesoscale eddy in these two regions, as discussed above. Further analyses and more sensitive experiments are required to clarify model performance in this regard.

3.3 Equatorial Pacific Currents

High-resolution models are also required in equatorial regions to resolve both the waveguide and the narrow (around 3–10°) zonal current belts in the upper layers. Typically, the meridional resolution is increased in the equatorial region. As in the low-resolution LICOM2.0, the resolution in the meridional direction is increased to 0.5° between 10°S and 10°N. The zonal currents in the equatorial Pacific at two longitudes (165°E and 150°W) are shown in Fig. 38.3, based on the observations of Johnson et al. (2002) and the high- and low-resolution LICOM2.0.

Fig. 38.3
figure 3

Zonal current along 165°E (left) and 155°W (right) for (a, d) the observations (b, e) the eddy-resolving LICOM2, and (c, f) the medium-resolution LICOM2 in the equatorial Pacific (units: ms−1)

Full size image

In the surface layer, the south equatorial current (SEC) and north equatorial countercurrent (NECC) are clearly exhibited in both sections for the observation data (Figs. 38.3a, b). Below the surface, the equatorial undercurrent (EUC), south subsurface countercurrent (SSCC), north subsurface countercurrent (NSCC), and equatorial intermediate current (EIC) are clear in Figs. 38.3a, b. These currents can be simulated by the coarse-resolution model, although with broader structures and weakened magnitudes compared to those in the high-resolution model. For instance, the SSCC and NSCC in the coarse-resolution LICOM2.0 are about 4° wide and have magnitudes below 0.1 ms−1, whereas their widths are only 2° in the observations.

With increasing horizontal and vertical resolution, the eddy-resolving model is able to more realistically simulate the complex structure and magnitude of currents in the upper equatorial Pacific. The meridional scales of the zonal current belts were all simulated well (Figs. 38.3b, e except for the NECC, which was not simulated in the western equatorial Pacific and is much weaker than the observation in the eastern equatorial Pacific. Wu et al. (2012) attributed this problem to biases in the QuikSCAT wind stress. The results also indicate that the horizontal resolution is crucial for simulation of the currents below the tropical thermocline, such as the EIC, SSCC, and NSCC. Furthermore, the magnitude of the EUC simulated by the eddy-resolving model is 0.2 ms−1 greater than that of the observations. It has been shown that the magnitude of the EUC is very sensitive to horizontal viscosity (Maes et al. 1997), such that reductions in the horizontal viscosity coefficient in the high-resolution model lead to strengthening of the EUC.

3.4 Indonesian Throughflow

The ITF connects the western equatorial Pacific and the eastern Indian Ocean through the Indonesian seas. Because there are more than ten thousands of islands in this region, the paths of the ITF are extremely complex. The water masses from the Pacific Ocean enter the Indonesian seas along two main paths: the western route through the Makassar Strait, and the eastern route through the Lifamatola Strait. After mixing induced by the strong tidal mixing in this region, the ITF exits the Indonesian seas through the Lombok Strait, the Ombai Strait, and the Timor Passage. Figure 38.4 illustrates the land–sea distribution in this region. The red lines represent the sections for which the volume transports in LICOM2.0 are computed. The numbers represent the volume transports at the five straits for the observations (former) and the high- (middle) and low-resolution (latter) LICOM. The star represents no transport. The observation data were obtained from the INSTANT program (Gordon et al. 2010), which is an international field program that provides the first continual simultaneous observation of the five major pathways of ITF for a 3 year period (2004–2006). Because only data below 1,250 m are available for the Lifamatola Strait for INSTANT, model transports were also computed below 1,250 m. The volume transports of the inflow straits are −11.6 and −2.5 Sv for the Makassar and Lifamatola straits, respectively. The total inflow is 14.1 Sv, and the Makassar Strait receives 82 % of the total inflow. The total outflow is 15.0 Sv, with outflow of −2.6, −4.9, and −7.5 Sv for the Lombok Strait, the Ombai Strait, and the Timor Passage, respectively. Thus, an imbalance of about 1 Sv exists in the observations; this has been attributed to processes in straits other than those mentioned here (Feng et al. 2013).

Fig. 38.4
figure 4

The land–sea distribution of the eddy-resolving LICOM2 in the Indonesian seas. The values of volume transports for INSTANT (left), the eddy-resolving LICOM2 (middle), and the medium-resolution LICOM2 (right) are marked in each strait or passage. The red lines represent the locations of volume transports computed in the model (units: Sv)

Full size image

Because of the northward transport in the Lifamatola Strait in the coarse-resolution LICOM2.0, the total inflow in this version is about half that in the observations. The total outflow is 13.9 Sv in the coarse resolution model, which is slightly less than that in the observations. Moreover, the closing of the Lombok Strait leads to increased transport in the Ombai Strait, indicating that the partitions for both inflow and outflow of the straits are incorrect.

All inflow and outflow transports were simulated well by LICOM2.0. Only 0.1 Sv difference in inflow transports is present between the simulation and the observations: inflow transport was 14.0 Sv for LICOM2.0 and 14.1 Sv for INSTANT observations. Both LICOM2.0 and INSTANT exhibit outflow transport of 15.0 Sv. The transport partitions were also simulated well for each individual strait. With regard to inflow, the Makassar Strait receives around 77 % of the total inflow transport, which represents 82 % of the total observed inflow. For the outflow straits, the high-resolution model reproduced the transport ratios well, with transport of −2.7, −3.1 and −9.2 Sv for the Lombok Strait, Ombai Strait, and Timor Passage, respectively.

4 Concluding Remarks

Recently, an eddy-resolving model was developed at LASG based on the LICOM2.0. A 20 year experiment has been conducted, forced by the daily mean COREs heat and freshwater data and QuikSCAT surface wind stresses. In the present study, the results from high- and coarse-resolution runs of LICOM2.0 are compared. The effects of horizontal resolution on the simulation of the mesoscale eddy, the western boundary currents, the equatorial Pacific currents, and the Indonesian throughflow are also investigated.

The simulated results improve in almost all aspects considered with increasing horizontal resolution. The major results can be summarized as follows. (1) The distributions and magnitudes of the mesoscale eddies, which were barely simulated in the coarse-resolution model, were reproduced well in the high-resolution model. (2) Improvements are obvious in the mean general circulation, such as in the WBCs and the equatorial zonal currents. The horizontal scales of the currents are all close to the observations. The high-resolution model also correctly simulated the separation points of the WBCs, which always overshoot in the coarse-resolution model. (3) The high-resolution model also simulated the circulation well in regions with complex land–sea distribution, such as in the Indonesian seas. In particular, the partition of volume transports through each strait was better simulated.

However, some problems remain in the eddy-resolving model. First, the model seems to be too energetic. Both eddies and the mean circulation are stronger than those observed. It is well known that the magnitudes of currents in models are sensitive to the adopted viscosity schemes or coefficients, representing a subgrid process in models. This suggests that the viscosity parameterization must be selected carefully in high-resolution models such as that described here. Furthermore, the topography of the model requires further modification, particularly for some key passages and straits. In particular, the depths of the deep straits in the Indonesian seas were all underestimated; such underestimation could lead to unrealistic vertical structure or volume transport. Finally, the NECC is weakened in both versions of LICOM2.0. The errors in the high-resolution model can be attributed partly to the wind stresses used to force the model. However, the errors in the coarse-resolution model cannot be explained. Therefore, the dynamics of the equatorial current in the Pacific require further investigation in LICOM2.0.