Journal of the Meteorological Society of Japan, Vol. 89, No. 5, pp. 399–411, 2011 399 DOI:10.2151/jmsj.2011-501 ITCZ Splitting and the Influence of Large-Scale Eddy Fields on the Tropical Mean State Eileen DAHMS Max Planck Institute for Meteorology, Hamburg, Germany Meteorological Institute, KlimaCampus, University of Hamburg, Germany Hartmut BORTH, Frank LUNKEIT and Klaus FRAEDRICH Meteorological Institute, KlimaCampus, University of Hamburg, Germany (Manuscript received 21 April 2010, in final form 12 December 2010) Abstract A spectral aqua-planet atmospheric general circulation model (AGCM) is forced with a series of zonally constant sea surface temperature (SST) distributions which are symmetric about the equator. For every oceanic forcing, the AGCM is run twice; a first time keeping all spectral modes and a second time with only the zonally symmetric ones. Parameterizations and boundary conditions remain the same in all cases thus allowing a consistent comparison of 3-D and 2-D flows. The comparative study shows that the structure of the tropical mean state of the full model is basically captured by the zonally symmetric model and that eddy fields merely modify this structure. This shows that the structure of the tropical mean state is mainly determined by the shape of the effective SST forcing. We confirm previous studies where the shape and strength of the Hadley circulation is comparable in the 3-D and 2-D experiments for cases with a well pronounced single ITCZ. So the underestimation of the Hadley circulation often found in idealized zonally symmetric models is not only due to the neglect of large-scale eddy fields. A new result is that both the full as well as the zonally symmetric model show the phenomenon of ITCZ splitting if the SST distribution gets flat enough at the equator. So ITCZ splitting can be explained by purely zonally symmetric mechanisms and is not necessarily induced by eddy fields. Eddies seem to stabilize single ITCZ circulation regimes. For SST distributions where ITCZ splitting occurs, multiple states exist in the zonally symmetric model. The atmosphere switches, for the same SST, between a single and double ITCZ state introducing a long-term variability in the tropics without the influence of mid-latitudes and without atmosphere-ocean coupling. 1. Introduction cations. In a series of follow up investigations, as for ex- ample in Held and Hou (1980), Satoh (1994), and Fang and Tung (1996), it was shown that the basic charac- The question about the importance of large-scale ed- teristics of the tropical mean state, such as the Hadley dies for the tropical mean state has a long tradition. Al- cell and subtropical jets, can be nicely reproduced by ready in the work of Schneider and Lindzen (1977), the purely zonally symmetric models. An important draw- question is raised if eddies are absolutely crucial in pro- back of the above models is underestimation by an order ducing the observed time and zonally averaged mean of magnitude of the transport of the Hadley circulation. circulation or if they merely produce qualitative modifi- Several improvements to the zonally symmetric approach have been achieved by poleward shifting Corresponding author: Eileen Dahms, Klima Campus (MPI- (Lindzen and Hou 1988) or by concentrating (Hou and M), Grindelberg 5, 20144 Hamburg, Germany. Lindzen 1992) the convective heat source. The investi- E-mail: [email protected] gations of Held and Phillips (1990) show that the feed- c 2011, Meteorological Society of Japan 400 Journal of the Meteorological Society of Japan Vol. 89, No. 5 back between a single Rossby wave and the Hadley cir- culation in a shallow water model exhibits highly non- trivial feedback behavior. Starting from the first satellite observations, it was obvious that the tropical circulation was highly variable in space and time (see e.g., Hubert et al. 1969). Since then, the question about position and structure of the ITCZ has become an important issue. Numerical atmo- spheric general circulation models are very sensitive in the deep tropics and often have problems in reproducing the correct structure of the ITCZ, also on climatological time scales (see e.g., Zhang and Wang 2006 and refer- Fig. 1. Idealized zonally symmetric SST distri- ences therein.) For conceptual models, there is the fun- butions: peak, control, qobs, and flat (northern damental question about the processes which determine hemisphere only). the position of the ITCZ and the connected phenomenon of ITCZ splitting. Two different paradigms have been established. One paradigm which goes back to at least the pioneering ing the same method. In contrast to us, they used a dry work of Charney (1971) is based on purely zonally sym- primitive equation model in which the strength and lo- metric mechanisms. The other paradigm put forward by cation of the tropical heat source (ITCZ) is prescribed Holton et al. (1971) is based on equatorial waves which a priori. A similar method was used in Kim and Lee are by definition not zonally symmetric. Which of the (2001a, b), but their model is a dry primitive model two paradigms is the more appropriate one is up to now driven by Newtonian cooling without an explicit hydro- under discussion. On the one hand, the dynamics of the logical cycle. So the mean circulation, eddy fields, and ITCZ are dominated by waves (see e.g., Wheeler and hydrological cycle are not free to adjust. Kiladis 1999). On the other hand, the climatological In a more recent study, Walker and Schneider (2006) circulation in the tropics is captured, to a large extent, investigate the influence of eddies on the Hadley cell by the stationary mean, as already described in Lorenz using an idealized aqua-planet AGCM forced by New- (1967). tonian cooling. Changing the rotation rate and the plan- In our investigation we quantify the influence of eddy etary radius, they derive scaling laws for the strength fields by comparing two runs of the same model, one and the extent of the Hadley cell. Modifying the ref- time with all spectral modes and another time only with erence temperature and relaxation times produces dif- the zonally symmetric ones. We use an AGCM with ferent eddy regimes. In their study, eddy fields are not parameterized radiation, moisture processes and turbu- suppressed in a controllable way and no hydrological lent processes. In our investigations only the boundary cycle is included in the model. conditions are prescribed, so that the circulation, eddy- Moist Hadley cell dynamics have been analyzed in fields and diabatic heating are free to adjust to a statis- previous studies. Numaguti (1993, 1995) focuses in his tically equilibrated state. This means that the structure studies on the role of surface evaporation and the influ- of the ITCZ and baroclinic zones change from case to ence of the position of the maximum SST on the Hadley case. circulation and the ITCZ in a series of numerical exper- Satoh et al. (1995) did similar experiments showing iments. Frierson (2007) compares different idealized that the Hadley cell had comparable width and strength convection schemes and their role for the zonal mean in the 3-D and 2-D cases for a large range of rotation circulation of the tropics. rates. We confirm this result in our investigations, sup- Our article is organized as follows: In Section 2 the porting the view that the result is of a more general na- model and experimental set-up are described, in Sec- ture since their model and our model have two different tion 3 the zonally symmetric model is compared to the sets of parameterizations and two different temperature climatological mean fields of the full model. In addition differences between equator and pole. Our analysis is the structural variability is discussed. A summary of the extended to the hydrological cycle and focuses more on results and a conclusion are presented in Section 4. the phenomenon of ITCZ splitting. Becker et al. (1997) have also investigated the feed- back between eddy fields and the Hadley circulation us- October 2011 E. DAHMS et al. 401 2. Model and experimental design Table 1. Experiments carried out. Experiment SST Model The numerical model applied for this study is the 1 peaked full peaked full Planet Simulator (Fraedrich et al. 2005), a model of in- 2 peaked sym peaked zonally symmetric termediate complexity (MIC). It is freely available un- 3 control full control full der http://www.mi.uni-hamburg.de/plasim. The dynam- 4 control sym control zonally symmetric ical core is based on the Portable University Model of 5 qobs full qobs full the Atmosphere (PUMA; Fraedrich et al. 1998). The 6 qobs sym qobs zonally symmetric primi tive equations are solved by the spectral trans- 7 flat full flat full form method (Eliasen et al. 1970, Orszag 1970). Un- 8 flat sym flat zonally symmetric resolved processes are parameterized. The parameteri- zation packet includes long- (Sasamori 1968) and short- (Lacis and Hansen 1974) wave radiation with interac- tive clouds (Stephens 1978; Stephens et al. 1984; Slingo the simulated circulation remains zonally symmetric if and Slingo 1991). A horizontal diffusion according to the model is initialized in a zonally symmetric state and Laursen and Eliasen (1989) is applied. Formulations is driven by zonally symmetric boundary conditions. for boundary layer fluxes of latent and sensible heat and Using this feature, two experiments are implemented for for vertical diffusion follow Louis (1979), Louis et al. each of the four SST profiles: One in the zonally sym- (1982), and Roeckner et al. (1992). Stratiform precip- metric mode and one in the full mode (i.e., including the itation is generated in supersaturated states. The Kuo full eddy spectrum).