1. Introduction
The intent of this paper is to describe recent enhancements in the Center for Ocean–Land–Atmosphere Studies (COLA) anomaly coupled general circulation model (ACGCM; Kirtman et al. 1997) and coupled general circulation model (CGCM; Schneider et al. 1999). The ocean and atmosphere component models for both the ACGCM and the CGCM are identical—the only difference is in the coupling strategy. This provides the opportunity to make a consistent comparison between the two coupling strategies and examine how relatively large differences in the mean state impact the interannual variability. The emphasis here will be on the natural variability of the models, particularly in the tropical Pacific.
Simulating and predicting natural variability in the tropical Pacific has proven to be an ideal test bed for coupled models, and there have been a number of comparison studies that document how well the different model formulations reproduce the observed climate. Not surprisingly, the first comparison studies (McCreary and Anderson 1991; Neelin et al. 1992) indicated rather unimpressive results in terms of qualitatively simulating the mean state, the annual cycle, and interannual variability in the tropical Pacific. Typically, the CGCMs that did not use flux corrections or anomaly coupling were unable to simulate both the annual cycle and interannual variability. For example, the Met Office model (S. Ineson and M. K. Davey 1994, personal communication) had a strong annual cycle, no interannual variability, and substantial climate drift (cooling). On the other hand, the model used by Philander et al. (1992) produced realistic ENSO variability but no annual cycle. These initial difficulties led to speculation that the CGCMs were a long way from being able to capture both the interannual variability and the annual cycle. Indeed, the results reported in Neelin et al. (1992) suggested that the sophisticated CGCMs were far short of being competitive with hybrid coupled models (e.g., Barnett et al. 1993) and intermediate coupled models (e.g., Zebiak and Cane 1987).
Mechoso et al. (1995) compared the annual cycles over the tropical Pacific of a number of different CGCMs to observations. The models that Mechoso et al. (1995) compared had no flux corrections or anomaly coupling. Almost all the models in the Mechoso et al. (1995) study succeeded in approximating the equatorial Pacific east–west SST gradient, which was a clear improvement over the results presented in Neelin et al. (1992). Approximately half the models had reasonably realistic annual cycles of SST along the equator. Nevertheless, many serious flaws were still evident. In particular, almost all the models produced a strong cold tongue during boreal spring contrary to observations. There was also a tendency for the coupled models to produce an erroneous double intertropical convergence zone (ITCZ) in the eastern Pacific. Despite the problems with the CGCM climatologies and annual cycles, relatively successful predictions have been made with CGCMs without flux correction (e.g., Rosati et al. 1997; Schneider et al. 1999; Stockdale et al. 1998).
Given the difficulties with the CGCMs noted above and the encouraging results from the intermediate and hybrid coupled models, it is not surprising that several groups have adopted the anomaly coupling strategy for ENSO prediction (Leetmaa and Ji 1989; Ji et al. 1994; Kirtman et al. 1997). The intent of the anomaly coupling strategy is to prevent the rapid climate drift seen in the CGCMs. In terms of ENSO prediction, the working hypothesis has been that the initial climate drift (or initialization shock) degrades forecast skill. There are often subtle but important differences in how the anomaly coupling strategy has been implemented, and these differences as well as model and initialization differences lead to substantially different forecasts. Nevertheless, these models seem to have comparable forecast skill [see Kirtman et al. 1997 for a comparison of the COLA and the National Centers for Environmental Prediction (NCEP) model forecast skill over an extended period, and Barnston et al. 1999 for their performance during the 1997–99 ENSO cycle].
Many intermediate and hybrid coupled models employ some kind of anomaly coupling strategy, and, while the implementation may be significantly different from the GCM approach, the basic motivation remains the same. Namely, the intent is to prevent the myriad of problems associated with climate drift from overwhelming the simulation of the anomalies. These intermediate and hybrid coupled models have been used in a number of successful ENSO prediction and simulations studies (Zebiak and Cane 1987; Barnett et al. 1993; Kleeman 1993; Neelin 1990; Latif and Villwock 1990; Neelin and Jin 1993; among many others). These models have also been used in a number of mechanistic and predictability studies (Goswami and Shukla 1991; Jin et al. 1994; Kirtman 1997; Kirtman and Schopf 1998; Yeh et al. 2001; Fan et al. 2000).
Given the potential dangers (i.e., erroneous multiple equilibria) of flux corrections (Neelin and Dykstra 1995) and the wide use of anomaly strategies in intermediate models, hybrid models, and coupled GCMs, it is important to document the impact that anomaly coupling has on simulations of the natural variability of the climate system. Indeed, there have only been cursory examinations in model intercomparison studies of how well anomaly coupled GCMs capture the ENSO cycle along the equator in the tropical Pacific. For example, in the Latif et al. (2001) coupled model intercomparison study, the NCEP model was shown to produce reasonably realistic ENSO events, although they were weak in amplitude compared to the observations. The original COLA anomaly coupled model produced ENSO events that were too strong, had a nearly regular biennial period, and had anomalies that peaked during the boreal summer season. These results are equivocal in terms of how well the anomaly coupling strategy works in long simulations. In fact, given that there have only been these two attempts to couple an OGCM and an AGCM with the anomaly coupling strategy, the impact of the anomaly coupling strategy on the long-term behavior of the coupled model remains an open question. It is difficult to determine the precise impact of the anomaly coupling strategy on both the simulation and prediction of ENSO because there have been no studies where the same ocean general circulation model has been both anomaly coupled and coupled without any flux corrections. To our knowledge, this is the first time that identical component models have been both anomaly coupled and coupled without any empirical corrections at the air–sea interface.
The remainder of this paper is outlined as follows. Section 2 describes the recent enhancements to the COLA CGCM and ACGCM. The emphasis is to highlight the changes from the previous versions of the models. The annual mean and the annual cycle of both coupled models are compared to available observations in section 3. The natural variability of the CGCM and the ACGCM is analyzed in detail and compared to available observations in section 4. Section 4 also introduces a diagnostic ocean model calculation that is designed to isolate the impact of the difference in the mean state on the natural variability of the two models.
2. The component models and coupling strategies
Both the ACGCM and the CGCM combined the COLA atmospheric GCM and the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model (MOM), version 3.0, ocean GCM. Brief descriptions of these models and the two coupling procedures are given below. The major differences between the ACGCM reported in Kirtman et al. (1997) and the CGCM reported in Schneider et al. (1999) and the versions presented here are summarized in Table 1.
a. Atmospheric model
A number of changes to the atmospheric model have been made since the original coupled models were developed. The dynamic core used in the National Center for Atmospheric Research (NCAR) Community Climate Model (CCM) version 3.0 has been adopted (Schneider 2001). The dynamic core is spectral (truncated at total triangular wavenumber 42) with semi-Lagrangian moisture transport. There are 18 unevenly spaced σ-coordinate vertical levels. The parameterization of the solar radiation is after Briegleb (1992) and terrestrial radiation follows Harshvardhan et al. (1987). The deep convection is an implementation of the relaxed Arakawa–Schubert scheme of Moorthi and Suarez (1992) described by DeWitt (1996). The convective cloud fraction follows the scheme used by the NCAR CCM (Kiehl et al. 1994; see DeWitt and Schneider 1996 for additional details). There is a turbulent closure scheme for the subgrid-scale exchange of heat, momentum, and moisture as in Miyakoda and Sirutis (1977) and Mellor and Yamada (1982). Additional details regarding the AGCM physics can be found in Kinter et al. (1988) and DeWitt (1996). Model documentation is given in Kinter et al. (1997).
The focus of this paper is comparing the ACGCM and the CGCM. Nevertheless, it is worth noting some of the major differences between the most recent version of the ACGCM and the version reported in Kirtman et al. (1997). Except for expanding to a global OGCM domain, the primary difference is in the atmospheric component model. The version described here uses a relaxed Arakawa–Schubert (RAS) scheme, whereas the original version used a Kuo (1965) convective parameterization and a coarser horizontal resolution (T30; see Table 1).
b. Ocean model
The ocean model is version 3 of the GFDL MOM (Pacanowski and Griffies 1998), a finite-difference treatment of the primitive equations of motion using the Boussinesq and hydrostatic approximations in spherical coordinates. The domain is that of the World Ocean between 74°S and 65°N. The coastline and bottom topography are realistic except that ocean depths less than 100 m are set to 100 m and the maximum depth is set to 6000 m. The artificial high-latitude meridional boundaries are impermeable and insulating. The zonal resolution is 1.5°. The meridional grid spacing is 0.5° between 10°S and 10°N, gradually increasing to 1.5° at 30°N and 30°S and fixed at 1.5° in the extratropics. There are 25 levels in the vertical with 17 levels in the upper 450 m. The vertical mixing scheme is the nonlocal K-profile parameterization (KPP) of Large et al. (1994). The horizontal mixing of momentum is Laplacian. The momentum mixing uses the space–time-dependent scheme of Smagorinsky (1963) and the tracer mixing uses Redi (1982) diffusion along with Gent and McWilliams' (1990) quasi-adiabatic stirring.
The major ocean model differences between the current and original ACGCM are that the original used constant horizontal mixing, Pacanowski and Philander (1981) vertical mixing (PP), and a Pacific basin only domain (see Table 1).
c. Coupling strategies
The anomaly coupling strategy is described in detail in Kirtman et al. (1997). The main idea is that the ocean and atmosphere exchange predicted anomalies, which are computed relative to their own model climatologies, while the climatology upon which the anomalies are superimposed is specified from observations or model-based estimates of observations. The anomaly coupling strategy requires atmospheric model climatologies of momentum, heat and freshwater flux, and an ocean model SST climatology. Similarly, observed climatologies of momentum, heat and freshwater flux, and SST are also required. The model climatologies are defined by separate uncoupled extended simulations of the ocean and atmospheric models. In the case of the atmosphere, the model climatology is computed from a 30-yr (1961–90) integration with observed specified SST. This SST is also used to define the observed SST climatology. The observed SST dataset is described in Smith et al. (1996). In the case of the ocean model SST climatology, an extended uncoupled ocean model simulation is made using 30 yr of 1000-mb NCEP–NCAR reanalysis winds. The NCEP–NCAR winds are converted to a wind stress following Trenberth et al. (1990). As with the SST, this observed wind stress product is used to define the observed momentum flux climatology. The heat flux and the freshwater flux in this ocean-only simulation is parameterized using damping of SST and sea surface salinity to observed conditions with a 100-day timescale. The heat and freshwater flux “observed” climatologies are then calculated from the results of the extended ocean-only simulation. The ocean and atmosphere model exchange daily mean fluxes of heat, momentum, and freshwater and SST once a day. No additional empirical corrections are applied to any of the exchanged anomalies.
3. Mean climates
Part of the motivation for the anomaly coupling strategy is that the ocean and atmosphere component models have serious systematic errors when forced with observed boundary conditions. Indeed, these systematic errors are in large part responsible for the climate drift in the CGCM. The anomaly coupling strategy does not remove the systematic errors internal to the component models; it does, however, “correct” the coupling information exchanged between the component models. For example, the mean SST simulated by the ocean component of the ACGCM will have systematic errors, and these errors are very similar to the uncoupled ocean model forced with “observed fluxes.”1 A comparison between the ocean components of the ACGCM and the CGCM provides an indication of how the systematic errors of the individual component models amplify and evolve when coupled. Similarly, a comparison between the systematic errors in the atmospheric components of the ACGCM and CGCM provides estimates of how the CGCM systematic error evolves.
Figures 1a–c show the annual mean SST simulated by the CGCM, ACGCM, and the observations, respectively. The most dramatic error in the CGCM (Fig. 1a) simulation is in the central Pacific, where it appears the warm pool is situated to the east of the date line and the cold tongue is very weak. The observations (Fig. 1c) indicate that the western Pacific warm pool is situated to the west of the date line. The CGCM-simulated SST in the tropical Indian and Atlantic Oceans is also considerably warmer than observed. The CGCM-simulated SST in the tropical Pacific also has more equatorial symmetry than observed. As expected, the ACGCM (Fig. 1b) is in better agreement with the observations; however, there are important errors worth noting. In particular, the cold tongue in the eastern tropical Pacific is too narrowly confined to the equator and extends too far to the west. The ACGCM also simulates a cold tongue in the tropical Atlantic Ocean that is more pronounced than in the observations. The ACGCM produces a somewhat better defined Gulf Stream, but the associated SST gradients are significantly weaker than observed.
Simulated SST error maps are shown in Figs. 2a,b. Despite the fact that the ocean component model, when forced with observed data, has a nearly uniform cold bias (Fig. 2b), the CGCM has a warm bias throughout most of the Tropics. In the eastern tropical Pacific, the errors in the CGCM exceed 5°C. In the tropical Atlantic and Indian Oceans the errors are on the order of 1°–2°C. In general the ACGCM is too cold when compared to the observations, but the errors are generally less than 1°C. However, the excessively strong cold tongues in the equatorial Atlantic and Pacific are evident by the errors exceeding −1°C. As will be seen in the next section, many of these errors in the mean state have a direct effect on the simulated interannual variability.
The errors in the ocean heat content are shown in Figs. 3a,b. Heat content is defined as the vertically averaged temperature in the upper 300 m. While the errors in the CGCM are more substantial than in the ACGCM, there are some similarities that are worth noting. For instance, both coupled models are generally too cold in the western Pacific and too warm in the eastern Pacific. Both models also have difficulty capturing the thermocline ridge that extends across the Pacific along 7°N. The heat content error structure is also similar in the tropical Atlantic and Indian Oceans.
Figures 4a–c show the SST annual cycle along the equator for the coupled models and the observations, respectively. In presenting the annual cycle, the annual mean has been removed and it is expected that the ACGCM would perform markedly better than the CGCM. The ACGCM annual cycle in the Pacific is in better agreement with the observations than the CGCM. The CGCM annual cycle in the Pacific is too tightly confined to the eastern part of the basin and is approximately half the amplitude as the observed. There are also significant phase errors. The cold tongue appears (and terminates) too early in the calendar year. There is also a relatively cold period in the western Pacific during July and August that is not apparent in the observations. It should be noted that the annual cycle depicted in Fig. 4a is quite similar to that reported in Schneider et al. (1999) and indicates very little improvement since the Mechoso et al. (1995) study. These errors in the CGCM mean climate are also essentially that same as those documented by Schneider (2001) for an earlier version of the model. In the Atlantic Ocean, the simulated annual cycle is in better agreement with observations than the simulation in the Pacific, but is still not quite as good as in the ACGCM.
The annual mean precipitation (Figs. 5a–c) indicates that the double ITCZ problem noted in Mechoso et al. (1995) remains in this updated version of the CGCM (Fig. 5a). The observed precipitation is calculated from Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP; Xie and Arkin 1996) rainfall estimates. In the central Pacific, there are nearly symmetric ITCZs about 10° off the equator. This tendency to produce double ITCZs is largely a problem with the AGCM and can be detected in AGCM-only simulations or in the ACGCM (Fig. 5b). For example, in the ACGCM, the South Pacific convergence zone (SPCZ) is too zonal and extends too far into the eastern Pacific, and there is also an erroneous Southern Hemisphere ITCZ in the Atlantic Ocean that is amplified in the CGCM.
Figures 6a,b show the annual mean precipitation error for the CGCM and the ACGCM, respectively. The CGCM rainfall error has a very large scale structure throughout the Indo-Pacific region. In the central and eastern Pacific there are excessive amounts of rainfall, and over the Maritime Continent and in the Asian summer monsoon regions there is too little precipitation. In the eastern and southern Indian Ocean the CGCM simulation produces too much rainfall. Some elements of these large-scale errors are mirrored in the ACGCM simulation. The errors in the eastern Indian Ocean and much of the Maritime Continent are fairly similar. In the Indian summer monsoon region, however, the ACGCM produces too much rainfall in contrast to the CGCM. The errors in the Atlantic Ocean are also quite similar, again suggesting that much of the double ITCZ problem originates from an atmospheric model error. Both models underestimate the precipitation over the Amazon basin and overestimate the rainfall over the Andes Mountains, perhaps indicating a problem with the land–atmosphere interactions, which is further exacerbated by topographic effects.
4. Natural variability
In this section, extended simulations of the ACGCM and the CGCM are presented and compared to available observational data.
a. The global ENSO
In terms of amplitude, both models produce interannual variability that is comparable to that observed in all three basins. The models produce irregular ENSO events that are qualitatively similar to those observed in terms of period and amplitude. In the Atlantic, the variability in the CGCM is a little weak, and in the ACGCM the frequency of the variability is too high.
Figures 7a–c show the SST anomaly (SSTA) standard deviation from both models and observations, respectively. The standard deviation is calculated using monthly data. Both coupled models overestimate the variability in the extratropics and the problem is considerably worse with the ACGCM. Very close to the equator in the Pacific, the ACGCM has the about the right amplitude, but the anomalies extend too far into the western Pacific. In the CGCM, the variability is too strong and too confined to the eastern Pacific. Both models seriously underestimate the meridional extent of the SSTA in the tropical Pacific. This meridional extent problem is also seen in uncoupled simulations with MOM3 (and earlier versions of MOM) suggesting that the problem is an ocean-only issue. The ACGCM has too much variability in the equatorial Atlantic, the Bay of Bengal, and the Arabian Sea.
One of the key features of ENSO variability is the associated global SST teleconnection. Simulating these global teleconnections has proven to be a daunting challenge (e.g., Davey et al. 2001). For instance, Figs. 8a–c show the correlation between the Niño-3.4–averaged SSTA and the SSTA at each grid point for both models and for observations, respectively (the Niño-3.4 region is 5°S–5°N, 120°–170°W). The correlation in Figs. 8a–c is calculated using monthly data. Both models fail to capture the meridional extent of the wedge of positive correlation in Pacific and the spreading of positive correlation along the west coast of North and South America. The so-called horseshoe of negative correlation and the positive correlation in the Indian Ocean are poorly represented in both models. In the south tropical Atlantic, the CGCM overestimates the correlation with ENSO; and in the north tropical Atlantic, the model captures some of the observed features, but they are displaced too far into the central part of the basin. The ACGCM produces far too strong correlations along the eastern equatorial Atlantic and fails to capture the positive correlation in the north tropical Atlantic. The fact that both models fail to capture the global ENSO teleconnection in the Indian Ocean and the extratropical Pacific appears to be due to the fact that the atmospheric component has too much variability, which is due to internal atmospheric dynamics (see Kirtman and Shukla 2001).
b. Composite analysis
In order to provide a more detailed look at the ENSO variability in the two models, we present a composite analysis here. The composites are based on Niño-3.4-averaged SSTA centered on January and include the previous and subsequent 24 months. With this procedure a 4-yr ENSO composite is formed. The warm (cold) composites are based on all Januaries where the Niño-3.4 SSTA is greater (less) then 0.75(−0.75) × σ, where σ is the Niño-3.4 SSTA standard deviation. For the observed composites, we applied the same procedure to the Carton et al. (2000) ocean data assimilation results for 46 years, which we refer to as the observations.
Figures 9a–c show the warm SSTA composites from the models and observations. The amplitude of both model composites compare favorably to the observed composite, and appear to have about the right phase locking with the annual cycle, although with the ACGCM this phase locking occurs a couple of months later than in the observations. The duration of pre– and post–cold periods in both model simulations are too short. The eastward migration of the SSTA and the trapping of the anomalies near the eastern boundary is readily apparent in the CGCM. The eastern boundary trapping and the eastward-displaced cold anomalies are consistent with the tendency for the warm pool to be situated too far to the east. In contrast to the CGCM, the zonal extent of the SSTA in the ACGCM is in better agreement with the observed structure, but the amplitude in the eastern Pacific is too weak except during the boreal winter season. While the ACGCM SSTA is primarily a standing mode, there is a suggestion of eastward development of the SSTA, particularly during the peak phase of the warm event. This eastward propagation can also be detected in close examination of the individual events. This eastward development is consistent with strong Kelvin wave signals detected in the heat content (see Fig. 10b). These Kelvin wave signals are considerably stronger than any similar features detected in the observations. The observed SSTA is primarily a standing mode, but there is some suggestion of westward migration. On the other hand, some observed ENSO events have a distinct eastward propagation (i.e., 1982–83); however, the tendency for eastward propagation is too strong in both models. During the onset phase of the warm SSTA, the ACGCM is considerably weaker than the observed; this problem is particularly noticeable in the eastern Pacific. Similarly, the ACGCM SSTA is too weak during the pre-peak cold periods.
The heat content composites (Figs. 10a–c) indicate that the eastward migration of the heat content anomalies is too rapid in both models. The CGCM simulation has two distinct episodes or pulses of eastward propagation. The first pulse is excited during the January preceding the peak phase of the warm event and is associated with the onset of the warm SSTA during the boreal spring and summer. The second pulse is initiated during the boreal fall and is associated with the peak phase of the warm event. The relative strength of the two pulses are about the same, but the second appears to be excited farther to the east. The ACGCM has a weak indication of this double pulse, but, as seen below, it is not as strongly connected to the wind stress as in the CGCM. This double pulsing is not detected in the observations. During the peak phase of the warm event, the ACGCM has cold subsurface anomalies that are too strong in the western Pacific. This is likely due to the fact that the wind stress is displaced too far to the west compared to the observations. The observations also indicate a western Pacific buildup phase that is stronger than either model simulation.
In the same format as Figs. 9a–c and 10a–c, Figs. 11a–c show the zonal wind stress composites. The eastward phase propagation and the connection to the double pulse in the heat content is readily apparent in the CGCM composite. Consistent with the ACGCM SSTA and heat content composites, the wind stress composite indicates that the warm events are too sharply peaked, too phase-locked to the annual cycle, and that their duration is too short. The ACGCM warm events are particularly weak during the onset phase during the early part of the calendar year.
The spatial structure of the SSTA composites are shown in Figs. 12a–c. The figures show a three-month-averaged SSTA centered about the peak warm January. The most outstanding problem with both model simulations is that they seriously underestimate the meridional scale of the SSTA. Both models capture some hint of the spreading of the warm anomaly along the coast of North and South America, but as with the teleconnection map (Figs. 8a–c), the anomaly is not well organized and too weak. The CGCM composite in the far eastern Pacific is relatively strong compared to the ACGCM. However, there is almost no positive anomaly to the west of 140°W and the off-equatorial western Pacific cold anomalies are situated too far to the east. Again, this is consistent with the general feature of the CGCM in which the warm pool is situated about 40° of longitude too far to the east. In contrast, the ACGCM is too weak in the east and too strong to the west of 140°W. The warm anomalies in the ACGCM composite extend too far into the western Pacific and the off-equatorial cold anomalies are too strong compared to the observations.
In the same format as Figs. 12a–c, the precipitation composites are shown in Figs. 13a–c. The model composites are calculated using the same procedure as in Figs. 12a,b. The observed composite is calculated from CMAP rainfall estimates. Because of a limitation in the number of years of available CMAP data, the observed composite is based on January–February means from 1982–83, 1986–87, 1991–92, and 1997–98. As expected, the model rainfall composites are consistent with the SST composites in that the CGCM-enhanced rainfall is narrowly confined to the far eastern Pacific, and the ACGCM rainfall anomalies are strongest in the central and western Pacific. Both models produce off-equatorial negative rainfall anomalies, although only in the Northern Hemisphere in the CGCM. In the ACGCM, the rainfall anomalies are dominated by a distinct north–south structure, whereas in the observations there is a distinct east–west structure. This problem with producing too much of a north–south structure appears to be endemic to many AGCMs of approximately the same resolution (D. Anderson 2001, personal communication).
One of the current challenges in coupled modeling is to have a so-called one tier coupled system that is capable of capturing the extratropical response to ENSO. One measure of the response has typically been the 200-mb geopotential height over North America. Using the same compositing procedure as Fig. 12, Figs. 14a–c compare the 200-mb geopotential height anomaly composites from the two models and from the NCEP–NCAR reanalysis data (Kalnay et al. 1996). The response of both models is weaker than that observed and fails to capture the observed structure. The relatively poor simulations in both models is consistent with the errors in the tropical forcing (i.e., Fig. 14a). Indeed, this has been verified by forcing a nonlinear barotropic model with the mean divergence calculated from the CGCM.2
c. ENSO dynamics
The intent of the anomaly coupling strategy is to improve, albeit empirically, the mean climate of the coupled model. The motivating hypothesis is that if the mean state of the coupled model is improved, then the space–time evolution of the anomalies will also improve. While it is clear that the evolution of the ENSO events in the two coupled models differ substantially, it is also clear that some features are better simulated by the AGCM (i.e., the zonal extent of the SSTA) and some features are better simulated by the CGCM (i.e., the SSTA amplitude in the eastern Pacific). The question then arises, how much of the differences in the ENSO cycle is directly linked to the differences in the mean states (Fedorov and Philander 2000)? For example, Figs. 9a and 11a show distinct eastward migration of the coupled model SSTA and wind stress anomaly, respectively. What role does the mean state play in this eastward migration?
The purpose of the diagnostic analysis presented here is to determine how much of the differences in the ENSO evolution in the two coupled models are directly attributable to the differences in mean state of the ocean. For this analysis, we employ a simple diagnostic anomaly ocean model (Zebiak and Cane 1987) where the mean state of the ocean is prescribed. Four separate ocean-only experiments are made. In the first experiment, both the ocean mean state and the prescribed wind stress anomalies are calculated from the long simulation of the CGCM. Similarly, the second experiment uses data that are calculated from the ACGCM simulation. The third experiment is to force the ocean model with wind stress anomalies from the CGCM simulation, but specify the ocean mean state from ACGCM simulation. Finally, in the fourth experiment the CGCM mean state and the ACGCM wind stress anomalies are used. The four experiments are summarized in Table 2. By comparing these four simulations, the direct effect of the differences in the mean state can easily be determined. Details regarding how the mean states were calculated are provided in the appendix.
Based on a number of experiments we have concluded that the two key elements of the mean state that are relevant here are the annual cycle of thermocline depth and upwelling. The time–longitude evolution of these annual cycle fields are shown in Figs. 15a–d. In comparing these fields, the most striking difference is that the CGCM has a pronounced annual cycle, whereas the ACGCM annual cycle is comparatively weak. The CGCM annual cycle in both the thermocline depth and the upwelling has a strong semiannual component that migrates eastward. Very little propagation can be detected in the ACGCM mean states. The ACGCM mean thermocline is deeper (shallower) in the west (east). Overall, the upwelling in the ACGCM is stronger than in the CGCM, except in the far eastern Pacific. The ACGCM has upwelling year round along the equator, whereas the CGCM has mean downwelling in the boreal spring and fall. Not surprisingly, the ACGCM mean annual cycle is in better agreement with observational estimates.
The four diagnostic ocean model simulations were made over the exact same time periods as the CGCM and the ACGCM simulations. Composites were calculated using exactly the same years as the CGCM and ACGCM composites. Figures 16a–d show the SSTA composites from all four experiments. How well the diagnostic model recaptures the CGCM and ACGCM composites can be assessed by comparing Fig. 16a to Fig. 9a and Fig. 16b to Fig. 9b, respectively. Qualitatively, the diagnostic model captures the main features of the CGCM and ACGCM composites. In particular, the eastward migration and the eastern boundary trapping of the CGCM SSTA composite is apparent in Fig. 16a. During the peak phase of the CGCM warm events, cold anomalies can be detected in the western Pacific. This feature is also simulated by the diagnostic model. The diagnostic simulation of the ACGCM composite captures the broader zonal extent of the SSTA and the distinct boreal winter intensification of the anomalies. The diagnostic model, however, does not capture the fact that the ACGCM SSTA composite is maximum around 150°W. Despite this shortcoming, the diagnostic model simulates the main differences between the two coupled model simulations.
The third experiment (i.e., mean state from ACGCM and wind stress anomaly from CGCM) SSTA composite is shown in Fig. 16c. Using the ACGCM ocean mean state has a substantial impact on the SSTA composite. To the west of the date line, the third experiment composite and the second composite experiment are in close agreement (Figs. 16c and 16b). By substituting in the ACGCM climatology, the cold SSTA in the western Pacific has been eliminated, the pre-peak phase eastward migration is not detected, and the zonal extent of the anomalies during the peak phase of the warm event is increased. On the other hand, to the east of the date line, the third experiment agrees with the first experiment (Figs. 16c and 16a). This implies that the strong CGCM response in the eastern Pacific is primarily due to the fact that the wind stress anomalies are situated much farther to the east compared to the ACGCM wind stress anomaly (see Figs. 11a and 11b). The lack of an SSTA response in the western Pacific in Fig. 16c is because the mean thermocline is considerably deeper so that any local anomalous upwelling has little effect on the SSTA. The eastward migration in Fig. 16a is strongly connected to the eastward-migrating CGCM mean upwelling (Fig. 15c) and relative minimum in thermocline depth (Fig. 15a).
The importance of the mean state in simulating the eastward migration and the eastern Pacific trapping is further emphasized by the results of the fourth experiment (Fig. 16d). In this case, the mean state is prescribed from the CGCM and the anomalous wind stress forcing is given by the ACGCM results. Despite the fact that there is no detectable eastward propagation in the ACGCM zonal wind stress anomalies (see Fig. 11b), pre-peak phase eastward propagation can be detected in the diagnostic model SSTA composite. The eastern Pacific trapping and the relatively strong western Pacific anomalies are also apparent in this last simulation. In addition, the relatively weaker response in the east Pacific in Fig. 16d further confirms the importance of the CGCM wind stress anomalies in the east.
5. Concluding remarks
The purpose of this paper was to describe recent enhancements to the COLA CGCM and ACGCM. The models have been unified in the sense that the atmospheric and oceanic component models are identical—the only difference between the two coupled models is the coupling strategy. Therefore, when the coupled models behave differently, we can eliminate differences in the formulation of the component models as being the cause. The ACGCM employs the anomaly-coupling strategy wherein the climatological fluxes from the atmosphere are corrected before they are felt by the ocean model so that they agree with model-based estimates of the observed annual mean and annual cycle of the fluxes. Similarly, the SST climatology simulated by the ocean component is corrected so that the atmospheric component experiences the simulated SSTA superimposed on the observed SST climatology. The motivating hypothesis for the anomaly coupling strategy is that by “correcting” the coupled model climatology, the simulation and prediction of the anomalies will also improve. The CGCM does not employ any empirical corrections at the air–sea interface.
Extended simulations were made with both coupled models and the annual mean, and the mean annual cycle was compared to available observations. Not surprisingly, the CGCM had substantial climate drift. The mean SST error in the east Pacific was over 5°C. This error is somewhat unusual in that most CGCMs have a cold bias in the equatorial east Pacific (Delecluse et al. 1998). Consistent with the mean SST error, the CGCM simulation has nearly symmetric double ITCZs straddling the equator in the Pacific. In some sense the ACGCM also has a mean SST error; however, the anomaly coupling strategy is designed to remove this error so that it is not felt by the atmospheric component model. This ocean component model error that is “corrected” is on the order of −1°C throughout most of the domain. The ACGCM also has significant precipitation errors; in fact, the tendency to produce the double ITCZs is clearly evident in the ACGCM simulation.
As expected, the mean annual cycle of the SST along the equator in the Pacific in ACGCM is in better agreement with the observations than the CGCM. In addition to phase errors, the CGCM SST annual cycle is weaker than observations. This is interesting because the CGCM mean annual cycle of wind stress, thermocline depth, and upwelling is considerably too strong and has an erroneous eastward migration/propagation. These errors in the mean annual cycle have direct implications on the simulation of the ENSO cycle in the CGCM. The annual cycle of thermocline depth and upwelling in the ACGCM are in better agreement with observations and are comparatively much weaker than in the CGCM.
Both models simulate robust interannual variability in the tropical Pacific. The ENSO events in the CGCM have a distinct eastward migration, and during the peak phase of the events, the SSTAs are trapped in the eastern Pacific. Conversely, the ACGCM ENSO events are dominated by a standing mode, and during their peak phase there is a tendency for the anomaly to extend too far to the west. Both models seriously underestimate the meridional extent of the SSTA. Both models also fail to capture much of the global SSTA teleconnection associated with ENSO, particularly in the Indian Ocean and subtropical Pacific.
A series of simple ocean model diagnostic calculations were presented in order to determine how the differences in the ocean annual cycle impacted the ENSO variability. These calculations suggest that much of the eastward migration of the ENSO events in the CGCM was intimately linked to the mean annual cycle in the west Pacific. In fact, when using wind stress anomalies from the ACGCM (i.e., no eastward migration) and the mean state from the CGCM, eastward migration in the simulated SSTA was still detected in the west Pacific. In the east Pacific, the anomalous wind stress appears to be the main factor contributing to the differences between the simulated ENSO events in the models. These diagnostic calculations emphasize the importance of accurately simulating the mean annual cycle in order to capture the observed ENSO cycle. It should be emphasized that the problem with the CGCM mean state is not necessarily due to the ocean model, but is a coupled model problem, the source of which has not been identified.
The results presented here indicate that the anomaly coupling approach is a viable strategy for simulating and understanding the natural variability of the climate system on a global scale (the previous COLA and NCEP anomaly coupled models only had active coupling in the tropical Pacific). While both the ACGCM and the CGCM have their relative strengths and weaknesses, it is our assertion that the anomaly coupled model is better suited for seasonal-to-interannual prediction. We acknowledge that given the results presented here, this assertion is debatable. However, our experience with retrospective forecasts (to be reported in another paper) supports this assertion. It is also our contention that the systematic errors in current state-of-the-art CGCMs are so large that simulating the interannual variability of the global circulation and rainfall is often extremely difficult. In terms of global seasonal-to-interannual prediction, there are operational efforts that use CGCMs and the systematic error is removed a posteriori. However, it is becoming increasingly clear that ameliorating the systematic error a priori can give rather impressive improvement in forecast skill, particularly for midlatitude circulation and rainfall (Shukla et al. 2000). Given the current state of CGCMs, it is also likely that the anomaly coupling approach will be useful for some time to come. Moreover, as the component models improve, we expect the ACGCM to also improve.
Acknowledgments
The authors are grateful to J. Shukla and P. Schopf for their constructive comments. Anonymous reviewers provided many suggestions that have greatly improved this manuscript. The authors acknowledge support from Grants NSF ATM-0122859, NSF ATM-9814295, NOAA NA96-GP0056, and NASA NAG5-8202.
REFERENCES
Barnett, T. P., M. Latif, N. Graham, M. Flugel, S. Pazan, and W. White, 1993: ENSO and ENSO-related predictability. Part I: Prediction of equatorial Pacific sea surface temperature with a hybrid coupled ocean–atmosphere model. J. Climate, 6 , 1545–1566.
Barnston, A. G., M. Glantz, and Y. He, 1999: Predictive skill of statistical and dynamical climate models in SST forecasts during the 1997–98 El Niño and the 1998 La Niña onset. Bull. Amer. Meteor. Soc., 80 , 217–243.
Briegleb, B. P., 1992: Delta-Eddington approximation for solar radiation in the NCAR community climate model. J. Geophys. Res., 97 , 7603–7612.
Carton, J. A., G. Chepurin, X. Cao, and B. Giese, 2000: A simple ocean data assimilation analysis of the global upper ocean 1950–95. Part I: Methodology. J. Phys. Oceanogr., 30 , 294–309.
Conkright, M. E., and Coauthors. 1998: World Ocean Database 1998: Documentation and quality control. NODC Internal Rept. 14, 43 pp.
Davey, M. K., and Coauthors. 2001: STOIC: A study of coupled GCM climatology and variability in tropical ocean regions. Climate Dyn., in press.
Delecluse, P., M. K. Davey, Y. Kitamura, S. G. H. Philander, M. Suarez, and L. Bengtsson, 1998: Coupled general circulation modeling of the tropical Pacific. J. Geophys. Res., 103 , 14357–14374.
DeWitt, D. G., 1996: The effect of the cumulus convection on the climate of the COLA general circulation model. COLA Tech. Rep. 27, 69 pp. [Available from COLA, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].
DeWitt, D. G., and E. K. Schneider, 1996: The Earth radiation budget as simulated by the COLA GCM. COLA Tech. Rep. 35, 39 pp. [Available from COLA, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].
Fan, Y., M. R. Allen, D. L. T. Anderson, and M. A. Balmaseda, 2000: How predictability depends on the nature of uncertainty in initial conditions in a coupled model of ENSO. J. Climate, 13 , 3298–3313.
Fedorov, A. V., and S. G. Philander, 2000: Is El Niño changing? Science, 288 , 1997–2002.
Gent, P. R., and J. C. McWilliams, 1990: Isopycnal mixing in ocean circulation models. J. Phys. Oceanogr., 20 , 150–155.
Goswami, B. N., and J. Shukla, 1991: Predictability of a coupled ocean–atmosphere model. J. Climate, 4 , 1–22.
Harshvardhan, R. Davis, D. A. Randall, and T. G. Corsetti, 1987: A fast radiation parameterization for general circulation models. J. Geophys. Res., 92 , 1009–1016.
Huang, B., and T. Shukla, 1997: An examination of the AGCM simulated surface wind stress and low-level winds over the tropical Pacific Ocean. Mon. Wea. Rev., 125 , 985–998.
Ji, M., A. Kumar, and A. Leetmaa, 1994: An experimental coupled forecast system at the National Meteorological Center: Some early results. Tellus, 46A , 398–418.
Jin, F. F., J. D. Neelin, and M. Ghil, 1994: El Niño on the devil's staircase: Annual subharmonic steps to chaos. Science, 264 , 70–72.
Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Kiehl, J. T., J. J. Hack, and B. P. Briegleb, 1994: The simulated Earth radiation budget of the National Center for Atmospheric Research community climate model CCM2 and comparisons with the Earth Radiation Budget Experiment (ERBE). J. Geophys. Res., 99 , 20815–20827.
Kinter, J. L. I. I. I., J. Shukla, L. Marx, and E. K. Schneider, 1988: A simulation of winter and summer circulations with the NMC global spectral model. J. Atmos. Sci., 45 , 2468–2522.
Kinter, J. L. I. I. I., and Coauthors. 1997: The COLA Atmosphere–Biosphere General Circulation Model. Volume 1: Formulation. COLA Tech. Rep. 51, 46 pp. [Available from COLA 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].
Kirtman, B. P., 1997: Oceanic Rossby wave dynamics and the ENSO period in a coupled model. J. Climate, 10 , 1690–1704.
Kirtman, B. P., and D. G. DeWitt, 1997: Comparison of atmospheric model wind stress with three different convective parameterizations: Sensitivity of tropical Pacific Ocean simulations. Mon. Wea. Rev., 125 , 1231–1250.
Kirtman, B. P., and P. S. Schopf, 1998: Decadal variability in ENSO prediction and predictability. J. Climate, 11 , 2804–2822.
Kirtman, B. P., and J. Shukla, 2001: Interactive coupled ensemble: A new strategy for coupling CGCMs. COLA Tech. Rep. 105, 24 pp. [Available from COLA, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].
Kirtman, B. P., B. Huang, Z. Zhu, and E. K. Schneider, 1997: Multiseasonal prediction with a coupled tropical ocean global atmosphere system. Mon. Wea. Rev., 125 , 789–808.
Kirtman, B. P., D. A. Paolino, J. L. Kinter III, and D. M. Straus, 2001: Impact of tropical subseasonal SST variability on seasonal mean climate simulations. Mon. Wea. Rev., 129 , 853–868.
Kleeman, R., 1993: On the dependence of hindcast skill on ocean thermodynamics in a coupled ocean–atmosphere model. J. Climate, 6 , 2012–2033.
Kuo, H. L., 1965: On the formation and intensification of tropical cyclones through latent heat release by cumulus convection. J. Atmos. Sci., 22 , 40–63.
Large, W. G., J. C. McWilliams, and S. C. Doney, 1994: Oceanic vertical mixing: A review and a model with a nonlocal boundary layer parameterization. Rev. Geophys., 32 , 363–403.
Latif, M., and A. Villwock, 1990: Interannual variability as simulated in coupled ocean–atmosphere models. J. Mar. Syst., 1 , 51–60.
Latif, M., and Coauthors. 2001: ENSIP: The El Niño simulation intercomparison project. Climate Dyn., 18 , 255–276.
Leetmaa, A., and M. Ji, 1989: Operational hindcasting of the tropical Pacific. Dyn. Atmos. Oceans, 13 , 465–490.
McCreary, J. P., and D. L. T. Anderson, 1991: An overview of coupled ocean–atmosphere models of El Niño and the Southern Oscillation. J. Geophys. Res., 96 , 3125–3150.
Mechoso, C. R., and Coauthors. 1995: The seasonal cycle over the tropical Pacific in general circulation model. Mon. Wea. Rev., 123 , 2825–2838.
Mellor, G. L., and T. Yamada, 1982: Development of a turbulence closure model for geophysical fluid problems. Rev. Geophys. Space Phys., 20 , 851–875.
Miyakoda, K., and J. Sirutis, 1977: Comparative integrations of global spectral models with various parameterized processes of sub-grid scale vertical transports. Beitr. Phys. Atmos., 50 , 445–480.
Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120 , 978–1002.
Neelin, J. D., 1990: A hybrid coupled general circulation model for El Niño studies. J. Atmos. Sci., 47 , 674–693.
Neelin, J. D., and F-F. Jin, 1993: Modes of interannual tropical ocean–atmosphere interaction—A unified view. Part II: Analytical results in the weak coupling limit. J. Atmos. Sci., 50 , 3504–3522.
Neelin, J. D., and H. A. Dykstra, 1995: Ocean–atmosphere interaction and the tropical climatology. Part I: The dangers of flux correction. J. Climate, 8 , 1325–1342.
Neelin, J. D., and Coauthors. 1992: Tropical air–sea interaction in general circulation models. Climate Dyn., 7 , 73–104.
Pacanowski, R. C., and S. G. H. Philander, 1981: Parameterization of the vertical mixing in numerical models of tropical oceans. J. Phys. Oceanogr., 11 , 1443–1451.
Pacanowski, R. C., and S. M. Griffies, 1998: MOM 3.0 Manual. NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, NJ, 668 pp.
Philander, S. G. H., R. C. Pacanowski, N. C. Lau, and M. J. Nath, 1992: Simulation of ENSO with a global atmospheric GCM coupled to a high resolution tropical Pacific ocean GCM. J. Climate, 5 , 308–329.
Redi, M. H., 1982: Oceanic isopychnal mixing by coordinate rotation. J. Phys. Oceanogr., 12 , 1155–1158.
Rosati, A., R. Gudgel, and K. Miyakoda, 1997: The impact of ocean initial conditions on ENSO forecasting with a coupled model. Mon. Wea. Rev., 125 , 754–772.
Schneider, E. K., 2001: Understanding differences between the equatorial Pacific as simulated by two coupled GCM's. COLA Tech. Rep. 98, 46 pp. [Available from COLA, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].
Schneider, E. K., B. Huang, Z. Zhu, D. G. DeWitt, J. L. Kinter III, B. P. Kirtman, and J. Shukla, 1999: Ocean data assimilation, initialization, and prediction of ENSO with a coupled GCM. Mon. Wea. Rev., 127 , 1187–1207.
Shukla, J., and Coauthors. 2000: Dynamical seasonal prediction. Bull. Amer. Meteor. Soc., 81 , 2591–2606.
Smagorinsky, J., 1963: General circulation experiments with the primitive equations. I. The basic experiment. Mon. Wea. Rev., 91 , 99–164.
Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. C. Stokes, 1996: Reconstruction of historical sea surface temperature using empirical orthogonal functions. J. Climate, 9 , 1403–1420.
Stockdale, T. N., D. L. T. Anderson, J. O. S. Alves, and M. A. Balmaseda, 1998: Global seasonal rainfall forecasts using a coupled ocean–atmosphere model. Nature, 392 , 370–373.
Trenberth, K. E., W. G. Large, and J. G. Olson, 1990: The mean annual cycle in global ocean wind stress. J. Phys. Oceanogr., 20 , 1742–1760.
Xie, P., and P. Arkin, 1996: Analysis of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840–858.
Yeh, S. W., B. Dewitte, J-G. Jhun, and I-S. Kang, 2001: The characteristic oscillation induced by coupled processes between oceanic vertical modes and atmospheric modes in the tropical Pacific. Geophys. Res. Lett., 28 , 2847–2850.
Zebiak, S. E., and M. A. Cane, 1987: A model of El Niño and the Southern Oscillation. Mon. Wea. Rev., 115 , 2262–2278.
APPENDIX
Diagnostic Ocean Model
The equations governing the ocean component model are given in detail in Zebiak and Cane (1987; hereafter referred to as ZC). The only changes to the model that we have made are in the specified mean surface currents, upwelling, and thermocline depth. The original ZC model used mean surface currents and upwelling that depended on latitude, longitude, and month of the year, whereas the mean thermocline was independent of time and latitude. In the mean states we specify here, the currents, upwelling, and thermocline depth all depend on latitude, longitude, and month of the year. We found that including the annual cycle of the mean thermocline depth was important for capturing the CGCM simulation with sufficient fidelity. This is, perhaps, obvious from the results shown in Figs. 15a and 15c.
Due to the paucity of marine observations, the mean ocean fields that ZC used were calculated diagnostically. In other words, the mean state ocean fields were spun up by forcing the ZC ocean model with the annual cycle of the observed wind stress. In our case, the mean annual cycle from the ACGCM and CGCM is well known and easy to calculate. The mean annual cycle of the thermocline depth is calculated from the depth of the 22°C isotherm for both the CGCM and the ACGCM. The mean annual cycle of upwelling is calculated from the CGCM- and ACGCM-simulated vertical velocity at 40 m. The mean surface currents are also directly calculated from the CGCM and ACGCM simulations. The CGCM and ACGCM fields were interpolated to the ZC ocean model grid via bilinear interpolation. The mean vertical temperature gradient was found to only have a small effect and was not modified from the original values used by ZC.
Annual mean SST from (a) the CGCM, (b) the ACGCM, and (c) observations. The contour interval is 1°C.
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Annual mean SST error from (a) the CGCM and (b) the ACGCM. The contour interval is 1°C
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Vertically averaged (0–300 m) temperature errors [model minus Levitus climatology (Conkright et al. 1998)] for the (a) GCM and the (b) ACGCM. The contour interval is 1°C
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
SST mean annual cycle (with the annual mean removed) along the equator for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 0.5°C
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Annual mean precipitation from (a) the CGCM, (b) the ACGCM, and (c) CMAP observations. The contour interval is 2 mm day−1
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Annual mean precipitation error (model − CMAP observations) for (a) the CGCM and (b) for the ACGCM. Contour interval is 2 mm day−1
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Sea surface temperature std dev for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 0.2°C
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Global SSTA point correlation with Niño-3.4 index for (a) the ACGCM, (b) the CGCM, and (c) the observations. The contour interval is 0.1
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Time–longitude section along the equator of SST composites for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 0.25°C
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Time–longitude section along the equator of vertically averaged temperature (0–300 m) composites for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 0.25°C
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Time–longitude section along the equator of zonal wind stress anomaly composites for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 0.05 dyn cm−2 (0.005 Pa)
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Dec–Feb average SST composites for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 0.25°C.
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Dec–Feb average precipitation composites for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 1 mm day−1
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Dec–Feb average geopotential height composites for (a) the CGCM, (b) the ACGCM, and (c) the observations. The contour interval is 10 m
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Time–longitude cross section of the annual cycle along the equator of the (a) CGCM thermocline depth, (b) ACGCM thermocline depth, (c) CGCM upwelling, and (d) the ACGCM upwelling. The contour interval in (a) and (b) is 0.1 m corresponding to the 10 m. The contour interval in (c) and (d) is 0.25 × 10−4 cm s−1
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Time–longitude cross section along the equator of SST composites for the diagnostic model. The results shown are for (a) experiment 1, (b) experiment 2, (c) experiment 3, and (d) experiment 4. The contour interval is 0.25°C. For a description of the experiments see the text
Citation: Journal of Climate 15, 17; 10.1175/1520-0442(2002)015<2301:TCGCAA>2.0.CO;2
Comparison of new and old COLA coupled models. See text for additional details
Description of ocean model diagnostic experiments. See text for additional details
It is important to point out that the SST errors of the ACGCM are not necessarily the same as with the uncoupled ocean. It is possible for some mean anomaly to develop in, for example, the wind stress, leading to differences in the climatologies. We have compared the ACGCM SST climate with the uncoupled ocean climate and found only minor differences. The ACGCM is generally colder by about 0.15°–0.2°C throughout most of the Tropics.
This is the same diagnostic procedure employed by Kirtman et al. (2001).
