Return Pre Home Next
GLOBAL COUPLED OCEAN-ATMOSPHERE GENERAL CIRCULATION MODEL IN LASG/IAP 
YU Yongqiang, ZHANG Xuehong, GUO Yufu and YU Rucong
LASG, Institute of Atmospheric Physics, P.O. Box 9804, Beijing 100029
Many scientists have been contributing their effort to develop global oceanic general circulation model (OGCM) and global coupled ocean-atmosphere general circulation model (CGCM) in State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of Atmospheric Physics (IAP) since the end of 1980s. From the original fluxes anomaly-coupling model developed in the beginning of 1990s to the latest directly coupling model, LASG scientists have finished four global coupled GCMs. This study mainly describes development and application of the third and the fourth generation of coupled GCMs.
I. INTRODUCTIONIn order to simulate the Earth's climate and its variations on intra-seasonal to decadal time scales, a hierarchy of comprehensive climate system models is needed. The most important components in such models are considered to be the atmosphere, the oceans, sea ice, land and its features. The development of such models has been undergoing a step-by-step process: from stand-alone atmosphere models to coupled ocean-atmosphere models, and so as to multi-sphere interactive models. The process is far from finished for climate modeling communities. The global coupled GCMs described in this study may represent an early stage toward the development of more complete climate system models. During the past decade, climate-modeling activities in the world underwent an important transition from using atmosphere-alone models to interactive ocean-atmosphere models. Rapid progress in the development of coupled ocean-atmosphere general circulation models (CGCMs) has been made since the late 1980s. The impetus for such rapid progress comes from the recognition as best summarized in the project of Climate Variability and Predictability (CLIVAR): “The major tools used in simulating, understanding, and predicting climate variations on all time scales will increasingly be computer models of the global coupled ocean-atmosphere-land system”.
The effort in developing numerical climate models in the State Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of Atmospheric Physics (IAP), was initiated in the middle of 1980s. The first global coupled GCM was developed in LASG based on a two-layer atmospheric general circulation model (AGCM) and a four-layer oceanic general circulation model (OGCM) in by Zhang et al. (1992). The model was integrated for forty years and showed some ENSO-like interannual variability dominated by westward propagating coupled mode.
Since the early 1990s, much efforts have been put on developing multi-layer AGCMs and OGCMs at LASG. A twenty-layer and 4o×5o OGCM with a thermodynamic sea ice model incorporated was developed first (Chen, 1994; Zhang et al., 1996). With the enhanced vertical resolution, the model is capable of not only reproducing some large-scale features of the upper ocean circulation but also depicting an acceptable configuration of the thermohaline circulation. The ML20 was coupled with the two-level AGCM to study the mean climate of the coupled system (Guo et al., 1996) and the time-dependent climate change due to increasing atmospheric CO2 concentration (Chen et al., 1997; IPCC, 1996). Meanwhile, a version of the nine-level spectral AGCM with a rhomboidal truncated wave number of 15 originated from Simmonds (1995) was introduced and reconstructed by Wu et al. (1996, hereafter referred to as L9R15). On the basis of such efforts, a new CGCM with its component OGCM of ML20 and AGCM of L9R15 was preliminaryily presented in 1995 (Liu et al., 1996). This may be regarded as the original version of the IAP/LASG global ocean-atmosphere-land system model (GOALS) (Wu et al., 1997). Five versions of the IAP GOALS model have been successively presented based on the above mentioned versions of the AGCM and OGCM. All these versions of the GOALS model (except the GOALS-0) were integrated for 200 years or longer without finding serious climate drift. The data have been used in evaluating the model's ability in climate simulation and in studying some selected aspects of climate diagnoses and analyses (Liu et al., 1998; Zhang et al., 1998; Yu et al., 2000; Meng and Wu et al., 2000).
Due to the coarse resolution of GOALS model, it is very difficult to accurately represent many important air-sea interaction processes such as ENSO events. Thus a medium resolution OGCM “L30T63” was developed in LASG by Jin et al. (1999). Then based on the ocean model L30T63, we finished a primary version of a Flexible General Circulation Model for climate system (referred to as FGCM-0) (Yu et al., 2002). The FGCM-0 is formulated based on the NCAR CSM-1 (Boville and Gent, 1998,) by replacing the CSM-1's ocean component, NCOM (the NCAR CSM Ocean Model), with the L30T63 in virtue of the CSM's flux coupler. The flux coupler is also responsible for interpolating and averaging between the different grid systems of component models while conserving local and integral properties. The coupling strategy (based on the flux coupler) allows component models to be exchanged relatively easily. In fact, it provides an efficient way to formulate a new climate system model and to test the sensitivity of the model to any of its component models. The atmosphere and land surface components are the CCM3 (the Community Climate Model, Version Three, see Kiehl et al., 1998) and the LSM1 (Bonan, 1998), respectively. Since the geography of the FGCM-0 is identical with that of the L30T63, the land-sea marks and surface types of the CCM3 have been modified to match the OGCM. Also, the horizontal grid of the sea ice model of Weatherly et al. (1998) has been modified to replace the original thermodynamic sea ice model of L30T63. The atmospheric, land and sea ice model communicates with the flux coupler every model hour, and the oceanic model every model day.
Generally speaking, LASG scientists have devoted to developing atmospheric, oceanic and climate system models since the end of 1980s. Up to now, we have finished four global coupled GCMs, and they were widely used in a lot of aspects of atmosphere and ocean sciences. In this study, we will describe the latest two global coupled GCMs—GOALS and FGCM-0.
II. GLOBAL OCEAN-ATMOSPHERE-LAND SYSTERM (GOALS) MODEL
1. Atmospheric and Land Component Models
The atmospheric component model is a nine-layer spectral model that is rhomboidally truncated at zonal wave number 15 (hereafter referred to as L9R15), which was originally from the version of Simmonds (1985). In order to make the model perform much better, some improvements are implemented in the model as follows.
·The standard stratification of temperature is introduced into the dynamical framework, and the scheme of reduction of a standard atmosphere, an idea proposed by Zeng (1963) and Phillips (1973), is used to improve the model performance. As a result, the deviation of temperature becomes a prognostic variable (Wu et al., 1996).
·A new k-distribution radiation scheme is implemented into the model. This radiation scheme can easily take the effects of trace gases, such as CO2, CH4, N2O and CFCs into account. Cloud processes are treated more reasonably in this scheme (Shi, 1981; Wang, 1996).
·A simplified biosphere and soil/snow model (SSiB, Xue et al., 1991) is implemented into the model (Liu and Wu, 1997).
·The diagnosed cloud cover and liquid water path scheme is introduced into the model (Liu et al., 1998).
·The drag coefficient for momentum surface stress over ocean is calculated according to Hellerman and Rosenstein (1983), instead of Simmonds (1985). The reason is that the Simmonds (1985) scheme much underestimates the drag coefficient over the eastern equatorial Pacific, only about 50% of that from observation.
·The diurnal variation of short and lo ng wave radiation is introduced into the model (Shao et al., 1998).
2. Oceanic and Sea Ice Component Models of GOALSBased on a four-layer OGCM by Zhang et al. (1989), the oceanic component model of GOALS has been developed since 1994. It is a twenty-layer model with the same horizontal resolution as that of the previous four-layer model but covers from the Antarctic coastline to 70oN in its original version (Chen, 1994; Zhang et al., 1996; hereafter referred to as ML20-0) and the global scope except the North Pole in its revised version (Yu, 1997). Another important difference between the ML20-0 and the ML20-1 is that the former uses the “upwind” finite-difference scheme for calculating the advections of temperature and salinity (Maier-Reimer et al., 1993) thereby discards the explicit diffusion terms in temperature and salinity equations, whereas the latter uses central finite-difference schemes for the advections and has conventional second-order diffusion terms. With the enhanced vertical resolution, both the ML20-0 and the ML20-1 are capable of reproducing some large-scale features of the upper ocean circulation and depicting an acceptable configuration of the thermohaline circulation (THC), while forced by observational atmospheric wind stress and thermal forcing as well as observational sea surface salinity (Zhang et al., 2000). The oceanic model ML20-0 was adopted in the original version of GOLAS (GOALS-0), and ML20-1 was adopted in the later five versions of GOALS.
A simple thermodynamic sea-ice model formulated based on Parkinson and Washington (1979) is incorporated into the ocean model. In the present ocean-sea ice coupling model, no attempt was made to parameterize the leads and brine-rejection processes. As a surrogate of the brine-rejection effect on the ocean thermohaline circulation (especially on the formation of AABW), an enhanced salinity forcing similar to England (1993) is imposed at the three rows adjacent to Antarctica throughout the model's integration without considering its seasonal variation.
3. Air-Sea Coupling SchemeWith observed boundary conditions, AGCMs and OGCMs can reproduce, to certain extent, reasonable large-scale circulation patterns in the atmosphere and ocean. However, when an AGCM and an OGCM are directly coupled to each other, the basic climate simulated by the coupled model is usually drifted away from the observed one. It is the problem of climate drift. In general, the climate drift occurs due to errors in the fluxes at the air-sea interface, which may be amplified by some positive feedback processes in a CGCM.
In order to reduce the flux errors in CGCMs, additional terms are often required to correct the simulated fluxes at the air-sea interface, that is the so-called “flux correction” (Sausen et al., 1988). The “flux-correction” (or “flux-adjustment”) technique has been widely used in the existing CGCMs (see Meehl, 1990; Neelin et al., 1992; Murphy, 1995) before 1995.
After some coupling experiments based on a two-level AGCM (Zeng et al., 1989) and a four-layer OGCM (Zhang et al., 1989), Zhang et al. (1992) developed a prediction-correction monthly flux anomaly coupling scheme, i.e., the MFA scheme. With the MFA scheme, the coupled model (referred to as M2+4) was integrated for forty years without significant climate drift (Zhang et al., 1992). Afterwards, the MFA scheme was employed in the two successive CGCMs. One is based on the 2-L AGCM and a twenty-layer OGCM (ML20), and the coupled model is referred to as M2+20 in this study. The other is based on a nine-layer and R15 spectral AGCM and the ML20-0. It is the original version of the GOALS/LASG model, i.e., the GOALS-0. However, the performances of the MFA scheme in these two models were not fully successful. The M2+20 suffered from a considerable cooling trend in its 130-year control run (Chen, 1994) and the GOALS-0 showed a one-degree (Celsius) warming in its forty-year integration (Liu et al., 1996). In other words, the MFA coupling scheme is not "robust" in controlling the climate drift.
Yu (1997) analyzed the performance of the GOALS-0 and developed a modified MFA coupling scheme (MMFA) (Yu, 1997; Yu and Zhang, 1998). The MMFA scheme has been successfully used in three successive versions of the GOALS model. The coupling scheme used in the subsequent version, GOALS-3, is essentially the MMFA but slightly different from those used before. All of these GOALS-versions were integrated for long time periods without noticeable climate drift.
Until the GOALS-4, the coupling time step was one month so that the “prediction-correction” method (Zhang et al., 1992) was necessary for keeping the synchronization between the AGCM and OGCM. In the GOALS-4, the coupling time step was set to one day and a daily flux anomaly-coupling scheme (referred to as DFA) was used so that the prediction-correction procedure was quitted. Preliminary results of the GOALS-4 show a good performance of the DFA scheme (Fig. 1).
Fig.1. The monthly global mean SST (oC) simulated by GOALS-4 from 1st to 200th model year.
4. Mean Climatology State by GOALSThe evaluations show that, in general, four versions of GOALS model are all able to reproduce the basic characteristics of current observed climate in many aspects reasonably, especially for the large-scale features and seasonal cycle (Yu, 1997; Guo et al., 2000). However, it is more important to remember that there are many weaknesses in the coupled modelling. Some weaknesses may be common to most coupled models, and some weaknesses may be related to the parameterizations of some physical processes of the GOALS model. We notice that the introductions of SSib land model and diurnal cycle of short-wave radiation to the model improve the simulation of surface air temperature obviously. The diagnostic cloud parameterization scheme, which leads to the absence of cloudiness and the anomalous high temperature in middle latitude in northern summer, need to be improved in the future. All five versions of GOALS model simulate lower precipitation rate over global mean. Particularly, the simulated much less precipitation in summer and more precipitation in winter in Northern Hemisphere middle latitude suggest that modification to the parameterization scheme to precipitation is necessary. The GOALS model is able to reproduce the annual cycle of SST and the El Nino time scale variability at the equatorial Pacific, although the simulated amplitude is weaker and the position of maximum SST anomaly is located further west compare with the observed (Fig.2).
5. Simulations of Global Climate Change for the Past and the Next 100 YearsThe four versions of GOLAS have been widely applied in climate variabilities, air-sea interaction, monsoon etc. (Liu etl al., 1998; Yu et al., 2000; Meng and Wu, 2000). In this study, we will mainly describe simulations of global climate change by human activities with the GOALS model.
Fig.2. The simulated SST (oC) anomaly averaged from 2oS to 2oN by GOALS-4 as a function of longitude (X axis) and time (Y axis) from 50th to 70th model year.
Although the GOALS model shows pretty good simulations of present climate, this does not necessarily guarantee that the response to a perturbation remains credible. Therefore, a lot of work firstly assesses the performance of the model in simulating the climate over the 20th century. Especially since the pioneer experiments conducted at the Hadley Centre for Climate Prediction and Research (Mitchell et al., 1995) and at the Deutsche Klimarechenzentrum (DKRZ) (Hasselmann et al., 1995), the simulations of the global climate change during the past 100 years become a standard experiment for coupled models (Boer et al., 2000; Emori et al., 1999; Haywood et al., 1997). Ma (2002) carried out simulations of the 20th century with GOALS including four experiments: control run, sensitivity run forced by greenhouse gases only, sensitivity run forced by greenhouse gases and solar radiation, sensitivity run forced by greenhouse gases, solar radiation and sulphate aerosols. The simulated global mean surface air temperature anomaly (Fig.3) suggested that the observed global warming 0.6oC should be mainly attributed to greenhouse gases during the past 100 years, but the effects of solar radiation and sulphate aerosols can not be ignored. When all forcings including greenhouse gases, solar radiation and sulphate aerosols are considered in the coupled model, the simulated global warming is 0.65oC during 20th century, which agrees well with the observed value.
Fig. 3. The global mean SST anomaly (relative to 1960-1990 mean value) during 20th century, GHG denotes the experiment considering greenhouse gases only, GHG+aerosol denotes the experiment considering greenhouse gases and sulphate aerosols, and GHG+aerosol+solar denotes the experiment considering greenhouse gases, sulphate aerosols and solar radiation.
Based on the simulations of 20th century, Guo et al. (2001) finished two simulations, one for the control run and the other for the perturbation run with GOALS-4 to invesgitae the global warming, with much detailed emphasis on East Asia. Results indicate that there is no climate drift in the control run and at the time of CO2 doubling the global temperature increases about 1.65 (Fig. 4). The GOALS model is able to simulate the observed spatial distribution and annual cycles of temperature and precipitation for East Asia quite well. But, in general, the model underestimates temperature and overestimates rainfall amount for regional annual average. For the climate change in East Asia, the temperature and precipitation in East Asia increase 2.1% and 5% respectively, and the maximum warming occurs at middle-latitude continent and the maximum precipitation increase occurs around 25°N with reduced precipitation in the tropical western Pacific.
Fig.4. The global mean surface air temperature simulated by GOALS-4 for the control run (solid line) and CO2 doubling run (dashed line).
III. THE FLEXIBLE GLOBAL COUPLED GCM VERSION 0—FGCM0
1. Basic Configuration of FGCM0
The coupled ocean-atmosphere general circulation model (CGCM) used in this study is the primary version of a Flexible General Circulation Model for climate system (referred to as FGCM-0). The FGCM-0 is formulated based on the NCAR CSM-1 (Boville and Gent, 1998) by replacing the CSM-1's ocean component, NCOM (the NCAR CSM Ocean Model), with the L30T63 in virtue of the CSM's flux coupler. The functions of the flux coupler are: 1) controlling the time coordination of all the component models of the climate system model; 2) calculating most of interfacial fluxes; 3) communicating with component models for exchanging fluxes and some control parameters (Fig.5). The flux coupler is also responsible for interpolating and averaging between the different grid systems of component models while conserving local and integral properties. As pointed out already in BG98, the coupling strategy (based on the flux coupler) allows component models to be exchanged relatively easily. In fact, it provides an efficient way to formulate a new climate system model and to test the sensitivity of the model to any of its component models. The FGCM-0 represents just the first step towards a fully developed FGCM, of which the oceanic component is the OGCM, L30T63 (without using its thermodynamic sea-ice component), and the others are almost the same as those in the CSM-1. The atmosphere and land surface components are the CCM3 (the Community Climate Model, Version Three, see Kiehl et al., 1998) and the LSM1 (Bonan, 1998), respectively. Since the geography of the FGCM-0 is identical with that of the L30T63, the land-sea marks and surface types of the CCM3 have been modified to match the OGCM. Also, the horizontal grid of the sea ice model of Weatherly et al. (1998) has been modified to replace that in the original L30T63. The atmospheric, land and sea ice model communicates with the flux coupler every model hour, and the oceanic model every model day. To diminish the initial shock in the coupling process, a spin-up procedure is adopted before running the FGCM-0.
Fig.5. The configuration of FGCM-0 component models.
Firstly, the AGCM, CCM3, and the land surface model, LSM1, are integrated for five years by using the observed climatological sea surface temperatures (SST), sea-ice distributions based on Shea et al. (1990) and the same land-sea mask as the ocean component model L30T63. This five-year integration is referred to as "Run 1" in this study. Daily data from the last four years of Run 1 are archived for state variables and for radiation flux at the lowest model level.
Secondly, the OGCM, L30T63, in association with the thermodynamic sea ice model of Weatherly et al. (1998) is integrated for seventy years, started from the year 1160 of the ocean model's basic run mentioned above. In the course of the integration, the surface wind stress and thermal forcing were taken from the daily state variables and radiation flux for the last four years of the aforementioned CCM3 run. For the surface salinity, the restoring condition is still used, without considering the fresh water flux calculated based on the AGCM and OGCM themselves. This seventy-year integration of the ocean model will be referred to as "Run 2" or "Spinup Run".
Following the spinup step, the fully coupled model, FGCM-0, was integrated for sixty years. The initial conditions for atmosphere, land surface, ocean and sea ice models are taken from the end of the Run 1 and Run 2, respectively. As in the CSM-1, the atmosphere and ocean-sea ice models in the FGCM-0 were directly coupled by exchanging the heat and momentum fluxes themselves, rather than the flux-anomalies as in the previous IAP/LASG CGCMs. Different from the CSM-1, the fresh water exchange was not included; instead, the relaxation condition of salinity was still used in the FGCM-0. Also in order to diminish the initial shock, the dynamic processes of sea ice were not included until the end of the first seven years of the coupled integration.
2. Oceanic Component Model of FGCM-0The oceanic component model used in the coupled model is the third-generation global OGCM developed at IAP by Jin et al. (1999). Its horizontal grid is just the same as that of a T63 spectral atmospheric general circulation model (AGCM) with the grid size of about 1.875°´1.875°. There are thirty layers in the vertical, of which twelve equal depth layers are placed in the upper 300 m for better depicting the equatorial thermocline. To be short, the model is often called “L30T63”. Some fairly mature parameterizations are adapted to the model, including the penetration of solar radiation (Rosati et al., 1988), the "PP" scheme for the upper ocean vertical mixing (Pacanowski et al., 1981), and the isopycnal mixing scheme proposed by Gent and MacWilliams (1990). A thermodynamic sea-ice model based on Parkinson and Washington (1979) is also incorporated into the ocean model. The model was first integrated for 1160 years with the wind stress forcing of Hellerman and Rosenstein (1983) and the thermal forcing required in a Haney-type formula for heat-flux (Haney, 1971), taken from the COADS (Comprehensive Ocean-Atmosphere Data Set) (da Silva et al., 1994). The model's surface salinity was simply relaxed to the climatological annual cycle of Levitus et al. (1994). By the end of the integration, the model reaches a quasi-equilibrium state, of which both the wind-driven circulation and the thermohaline circulation are reasonably simulated (Jin et al., 1999). This may be seen as the L30T63's basic run that provides initial conditions for the coupled model spin up run. Forced by the month-by-month wind stress over the tropical Pacific Ocean from the ECMWF reanalysis (ERA) data for the time period from 1980 to 1989, the OGCM reproduces the reasonable interannual variability in the tropical Pacific Ocean (Yu et al., 2001).
3. Mean Climatology by FGCM-0The flexible coupled ocean-atmosphere general circulation model version 0 (FGCM-0) has been integrated 60 years successfully. Although the flux correction is not employed in the coupled model FGCM-0, the model doesn't show the obvious climate drift. This is because that all component models, NCAR CCM3, land model, sea ice model, as well as IAP OGCM, show good ability to depict dynamical and physical processes in the climate system, in the meantime the flux coupler, which assures the conservation of energy and mass at the interfaces of model components, also plays a very important role in controlling the climate drift.
Compared to the observed climatology, although the model FGCM-0 reproduces east-west gradient of SST in the equatorial Pacific, the model's major errors are the simulated the double ITCZ and the associated SST pattern (Fig.6). As described by a lot of studies (Mechoso et al., 1995; Boville and Gent, 1998), they are common features for the coupled models without flux correction. The equatorial warm pool simulated by the FGCM-0 differs substantially from observation, reflecting the temperature averaged over upper 100 m there is about three degrees colder than observation. With the exaggerated role of ocean dynamics in the thermal equilibrium, the “warm pool” simulated by the FGCM-0 resembles, to a certain extent, the cold tongue. Moreover, the FGCM-0-simulated thermocline in most of the tropical South Pacific is tens-to-hundred meters shallower than observation. As a consequence, the thermocline tends to be latitudinally symmetrical about the equator. This may represent an image of the “double ITCZ” mode in the ocean component of a coupled ocean-atmosphere model. The severe biases in the simulated thermocline may be attributed to the systematic errors in the surface wind stress, and to the sharply decrease of the vertical mixing with depth. It is found that the meridional overturning circulation related to the unrealistic cold water in the middle- and high-latitude North Pacific Ocean may be favorable to maintain the clod biases in the tropical thermocline.
Fig.6 . Simulated sea surface temperatures (oC) by FGCM0 averaged from 11th to 60th model year for (a) March and (b) September.
4. ENSO and IOD by FGCM-0Forced with the observed wind stress, the oceanic component model has shown good ability in reproducing El Niño and La Niña events during 1980s (Yu et al., 2001). When the oceanic model is coupled to the atmospheric, land and sea ice models through the flux coupler, the interannual variation of the averaged SST over Niño 3.4 region (170o—120oW, 5oS—5oN) is produced automatically without any external forcing except for solar radiation, implying that the coupled model can simulate the “ENSO-like” phenomena (Fig.7). The coupled model produces very similar amplitude of Niño index to the observational one, but shows a quasi-biennial oscillation only instead of a considerable wide period from 2 to 7 years as in the real world. The further analyses indicate that the physical mechanism for the simulated ENSO-like events is very similar to “Delayed Oscillator” proposed by Schopf and Suares (1988) and Battisti and Hirst (1989).
Fig.7. Simulated sea surface temperature anomalies (oC) by FGCM-0 averaged over Niño 3.4 region (170oW- 120oW, 5oS-5oN).
In the tropical Indian Ocean, the coupled model FGCM-0 shows a similar dipole mode pattern to the observational one (Fig.8). Figure 8a shows the simulated dipole mode index (DMI), which is defined as Saji et al. (1999), and the correlation coefficient between the DMI and the Niño3.4 index is 0.44, which is not very strong but still significant. It implies that the simulated dipole mode in tropical Indian Ocean is associated with the ENSO event in the coupled model. Figures 8b-8d is the correlations between DMI and SST, zonal wind stress and upper ocean heat content, respectively. Thus, the simulated dipole mode pattern isn't only shown in SST, but also in VAT, especially there is close correlation between dipole mode index and the zonal wind stress in the central Indian Ocean, which implies that the dipole mode pattern is a coupled mode resulted from the air-sea interaction in the tropical Indian Ocean.
Fig.8. (a) The simulated dipole mode index in Indian Ocean, correlation coefficients of DMI with (b) SST, (c) vertically averaged temperatures in the upper 300m, (d) zonal wind stress.
Fig.8. (to be continued)
5. Paleo-Climate Modeling by FGCM-0Using the present, 14 MaBP and 6 MaBP topography of the ocean in a global OGCM and a coupled ocean-atmosphere model, respectively, a series of numerical experiments are implemented in order to investigate the paleoclimate effect of opened Isthmus of Panama and Indonesian passage on the oceanic and atmospheric circulation. The numerical experiments of the individual OGCM forced by the modern atmospheric circulation and coupled ocean-atmosphere model show the similar results, which indicate that the close of Indonesian passage results in warming in the Pacific Ocean and cooling in the Indian Ocean; furthermore, the Indonesian Through Flow (ITF) mainly originates from the southern Pacific at 14 MaBP, but it mainly originates from the northern Pacific now. Meanwhile, the close of isthmus of panama causes the strong upwelling in the eastern equatorial Pacific and isolates heat exchange between Pacific and Atlantic Oceans, and eventually results in cooling in SST and uplift of thermocline in the eastern Pacific and increases the contrast of heat content between the western and eastern Pacific (Fig.9).
Fig. 9. Simulated SST averaged from 2oS-2oN as a function of longitude (X axis) and depth (Y axis, unit: m), for control experiment (modern topography), sensitivity experiment 1 (opening Indonesian passage as 6MaBP), sensitivity experiment 2 (opening Indonesian passage as 14 MaBP) and sensitivity experiment 3 (opening Indonesian Passage and the Isthum of Panama as 14 MaBP).
IV. SUMMARYTwo global coupled GCMs GOALS and FGCM-0 have been developed in LASG in the past five years, and they have been widely employed in simulating natural climate variability from seasonal to decadal time scales and anthropogenic climate change. Although there are a lot of uncertainties in all state-of-art coupled GCMs including GOALS and FGCM-0, the coupled models show the better and better ability to reproduce the fundamental features of climate system since 1980s. In fact, after scientists in LASG contributed much effort during the past decades, considerable progresses were made in the development and validation of climate models as follows.
(1) Based on the individual ocean, atmosphere, land and sea ice models, to put forward the MMFA and DFA coupling schemes and develop a coupled GCM GOALS model. In fact, the GOASL model is the first physical climate system model in LASG.
(2) Four versions of GOALS model have been integrated at least 200 years for control run, and especially several additional extended integrations are carried out for investigating climate change induced by human activities. Almost all model output data have been analyzed for evaluating model, investigating physical processes of climate system etc. Especially, the second and the forth versions of GOALS (GOALS-2 and GOALS-4) joined the Couple Model Intercomparisons Project phase 1 and phase 2 (CMIP1 and CMIP2), and simulations of global climate change are used to estimate future climate change (IPCC, 1996; IPCC, 2001) .
(3) Although there still be a lot of apparent shortcomings in GOALS and the other earlier coupled models developed in LASG, LASG scientists acquired the most important knowledge, experiences through developing and improving them so that the latest developed coupled model FGCM-0 come to higher level than othe earlier coupled models. FGCM-0 is a flexible GCM, the word “flexible” implies that it is very easy to replace any parts of FGCM-0 including physical parameterization schemes or component models with new one. Thus it is possible to compare individual GCMs or parameterization schemes under the same framework of climate system model, and then it is easy to improve and develop model.
Besides FGCM-0 and GOALS, we developed another coupled GCM-LASG/NCC T63AOGCM through coupling IAP/LASG L30T63 OGCM and NCC T63 AGCM with the DFA coupling scheme for short-range climate prediction in 2000 (Yu et al., 2000). The model T63AOGCM's ocean component is the same as FGCM-0, but its atmospheric model is developed based on a medium-range weather forecast model from ECMWF by Ye et al. (2000) in National Climate Center (NCC). Yu et al. (personal communication) indicated that FGCM-0 and T63AOGCM show very great difference in simulating mean climatology state and ENSO events, which implies that AGCM plays a key role in coupled model. The model T63AOGCM shows somewhat skill in short-range climate prediction during the past three years (Ding et al., 2000).
Acknowlegements: LASG coupled GCMs are not able to finished without support from the former director of LASG Dr. WU Guoxiong and the director of LASG Dr. WANG Bin. We also wish to appreciate all scientists participating development, evaluation and application of LASG coupled models, especially Dr. ZHOU Tianjun, Dr. LI Wei, Dr. LIU Hui, Dr. JIN Xiangze, Dr. CHEN Keming, Dr. LIU Hailong, Dr. WANG Zaizhi, Dr. WANG Biao and Dr. SHI Guangyu.
Battisti, D.S. and A.C. Hisrt, 1989, Interannual variability in a tropical atmosphere-ocean model: influnce of the basic state, ocean geometry and nonlinearity, J.A.S., 46, 1687-1712.
Boer, G.J., G. Flato, M.C. Reader and D. Ramsden, 2000, A transient climate change simulation with greenhouse gas and aerosol forcing: experimental design and comparison with the instrumental record for the 20th century, Clim. Dyn., 16, 405-426.
Bonan, G.B., 1998, The Land surface climatology of the NCAR land surface model coupled to the NCAR community climate model, Journal of Climate, 11, 1307-1326.
Boville, B.A. and P.R. Gent , 1998, The NCAR climate system model, Version One, Journal of Climate, 11, 1115-1130.
Chen, K., X. Zhang, and X. Jin, 1997, A coupled ocean-atmosphere general circulation model for studies of global climate changes, I. Formulation and performance of the model, Acta Oceanologia Sinica, 19 (3), 21-32 (in Chinese).
Chen, K. 1994, Improvement of IAP global coupled ocean-atmosphere general circulation model and numerical simulation of climate change induced by the enhanced greenhouse effect, Ph.D. Thesis, Institute of Atmospheric Physics, Chinese Academy of Sciences, 145pp. (in Chinese).
da Silva, A. M., C. C. Young and S. Levitus, 1994, Atlas of Surface Marine Data 1994, Volume 2: Anomalies of Directly Observed Quantities. NOAA Atlas NESDIS 7.
Ding, Y. and Y. Liu, 2002, Study and experiments of short-range climate prediction dynamical system in China, 7, 236-246 (in Chinese).
Emori, S., T. Nozawa, A. Abe-Ouchi, A. Numaguti, M. Kimoto and T. Nakajima, 1999, Coupled ocean-atmosphere model experiments of future climate change with an explicit representation of sulfate aerosol scattering. J. Met. Soc. Japan, 77, 1299-1307.
England, M.H., 1993, Representing the global-scale water masses in ocean circulation models, J. Phys.Oceanogr., 23, 1523-1552.
Gent P R, and McWilliams J C. 1990, Isopycnal mixing in ocean circulation models, J. Phys. Oc eanogr., 20, 150-155.
Guo, Y., Y. Yu, K. Chen, X. Jin, and X. Zhang, 1996, Mean climate state simulated by a coupled ocean-atmosphere general circulation model, Theor. and Appl. Climatol., (1-4), 99-112.
Guo, Y., Y. Yu, T. Zhang, 2000, Evaluation of IAP/LASG GOALS model, In IAP Global Ocean-Atmosphere-Land System Model, Edited by Zhang et al., Science Press, Beijing, New York, 252pp.
Guo, Y., Y. Yu, X. Liu and X. Zhang, 2001, Simulation of climate change induced by CO2 increasing for East Asia with IAP/LASG GOALS Model, Adv. Atmos. Sci., 18, 53-66.
Hasselmann, K., L. Bengtsson, U. Cubasch, G.C. Hegerl, H. Rodhe, E. Roeckner, H. von Storch, R. Voss and J. Waszkewitz, 1995, Detection of anthropogenic climate change using a fingerprint method. In: Proceedings of “Modern Dynamical Meteorology”, Symposium in honor of Aksel Wiin-Nielsen, 1995, P. Ditlevsen (ed.), ECMWF Press, 1995.
Haywood, J.M., R.J. Stouffer, R.T. Wetherald, S. Manabe and V. Ramaswamy, 1997, Transient response of a coupled model to estimated changes in greenhouse gas and sulfate concentrations. Geophys. Res. Lett., 24, 1335-1338.
Hellerman, S., and M. Rosenstein, 1983, Normal monthly wind stress over the world ocean with error estimates, J. Phys. Oceanagr., 13, 1093-1104.
IPCC, 1996, Climate Change 1996, The Science of Climate Change, T. J. Houghton, L. G. Meira Filho, B. A. Callander, N. Harris, A. Kattenberg and K. Maskell (eds.), Cambridge University Press, Cambridge, U. K. 572 pp.
IPCC, 2001, Climate Change 2001 : The Scientific Basis, R. Watson J. Houghton and Y.Ding (eds.), Cambridge University Press, Cambridge, U. K. 882 pp.
Jin, X., X. Zhang and T. Zhou, 1999, Fundamental framework and experiments of the third generation of IAP/LASG world ocean general circulation model, Adv. Atmos. Sci., 16, 197-215.
Jin, X., X. Zhang and Y. Yu 2000, Oceanic General Circulation and Sea-Ice Model, In IAP Global Ocean-Atmosphere-Land System Model, Edited by Zhang et al., Science Press, Beijing, New York, 252pp.
Levitus S., R. Burgett and T.P. Boyer, 1994, World Ocean Atlas 1994 Volume 3, Salinity, NOAA Atlas NESDIS 3, U.S. Department of Commerce, Washington, D.C. 99 pp.
Liu, H., X. Jin, X. Zhang and G. Wu, 1996, A coupling experiment of an atmosphere and an ocean model with a monthly anomaly exchange scheme, Adv. Atmos. Sci., 13, 133-146.
Liu, H., and G. Wu, 1997, Impacts of land surface on climate of July and onset of summer monsoon: A study with an AGCM plus SSiB, Adv. Atmos. Sci., 14, 289-308.
Liu, H., X. Zhang, and G. Wu, 1998, Cloud feedback on SST variability in western equatorial Pacific in a CGCM, Adv. Atmos. Sci. 15(3), 410