The Seasonal Cycle in Surface Heat Fluxes Over the Tropical Pacific in a Cou pled GCM

A.W. Robertson, C.R. Mechoso, and C.-C. Ma

Introduction: Work with simple models suggests that if a coupled general circulation model (GC M) is to simulate the El Ni\026o-Southern Oscillation (ENSO) phenomenon realisti cally, it must also reproduce the observed mean climate and seasonal cycle in th e tropical Pacific. The surface heat flux is obviously a crucial component of th e ocean heat balance. We examine it here in the UCLA CGCM, using the Oberhuber ( 1988) observed climatology which is based on COADS data, as our bench mark. The model consists of the UCLA atmospheric GCM (4o latitude x 5o longitude x 9 layer s), coupled to the GFDL ocean GCM (tropical Pacific version: 1/3o-2o latitude x 1o longitude x 27 layers). We consider the annual mean and annual harmonic in a 10-year coupled simulation. For brevity we omit discussion of the semi-annual ha rmonic, whose amplitude is comparatively small.

The Annual Mean: In the annual mean, the model captures the observed equatorial east-west asymmet ry and "cold tongue" in sea surface temperature (SST), though SSTs are 1-2oC too cold over most of the tropical Pacific (Fig. 1a,b). The associated net heat flu x into the ocean (Q) is simulated remarkably well to first order (Fig. 1c,d): th e model captures its maximum over the cold tongue\121a negative feedback on SST. Figure 2 shows a meridional section of Q and its components, averaged over the eastern Pacific (150oW-100oW). Although the simulated magnitudes of Q agree with Oberhuber's estimates, the simulated short-wave and latent components are consi derably overestimated. The observed meridional asymmetry about the equator in Q is not captured by the model, and this is traceable to deficiencies in the laten t and short-wave components. The latter is consistent with model shortcomings in annual-mean cloud cover (not shown)\121the CGCM produces a double ITCZ straddli ng the equator in the annual mean, so that the observed north-south asymmetry in cloud cover is absent. This is also accompanied by spurious equatorial doldrums in the surface winds, and thus a pronounced minimum in evaporation at the equat or. These deficiencies are common to several other coupled GCMs. The AGCM alone, when forced with observed SSTs, produces a picture midway between Fig. 2a & b.

The Annual Harmonic: The annual harmonic in SST and Q from 30oS to 30oN across the eastern Pacific (F ig. 3) has an anti-symmetric modal structure with respect to the equator. The ph ase quadrature between SST and Q indicates a forcing by the latter. SST is gener ally well simulated, but the model clearly overemphasizes the anti-symmetry. The secondary maximum in SST just south of the equator (associated with ocean trans port) is underestimated by the model, whose south-east Trades are too weak. Nort h of the equator there are errors in SST that are consistent with the model's fa ilure to reproduce the annual mean cloudiness maximum north of the equator, conc urrent with the observed annual-mean position of the ITCZ. The model's ITCZ actu ally migrates south of the equator, and errors in the annual harmonics of the la tent and short-wave fluxes (not shown) suggest that local air-sea interaction en hances this defect (via evaporation and PBL stratus respectively).

Figure 4 shows the annual harmonic along the equator in SST and Q. In contrast t o Fig. 3, the equatorial cold tongue in the eastern equatorial Pacific is clearl y not forced locally by Q. The model succeeds in simulating SST quite well (if s lightly weak and one month premature in phase), despite large errors in Q. In th e Oberhuber climatology, PBL stratocumulus near 90oW accompanies and amplifies t he cold tongue. Enhanced latent heat fluxes associated with the seasonal south-e ast Trades (that mechanically force the cold tongue) also act to amplify the ann ual harmonic in SST. Neither of these effects are caught by the model where wind anomalies are weak and stratocumulus does not develop over the cold tongue, who se character is thus more associated with ocean transport than observed.

Discussion: Errors in surface heat flux are a sensitive indicator of the performance of a co upled model. How the model's heat flux errors\121both local and remote\121affect SST, and whether air-sea interaction amplifies SST errors are important issues. The following points arise from this study:

Over the eastern Pacific, our CGCM's simulation of convective cloud and PBL stra tus is poor. This is a source of local errors in SST in both the annual mean and annual harmonic. Although the local forcing of the equatorial cold tongue by Q is secondary, defects in the seasonal displacement of the model's ITCZ do seem t o be partially diabatic in origin. Local feedbacks involving evaporation and clo ud cover appear to play a role.

Acknowledgments: Model development was supported by ONR (N00014-89J-1845), by NS F and DARPA with CNRI (NCR-8919038), and by INCOR. The model was run at the San Diego Supercomputer Center.

Reference: Oberhuber, J.M., 1988: MPI Report No. 15, Bundesstr. 55, 2000 Hamburg 13, Germany.

Figure 1: Annual mean maps of (a) observed SST, (b) simulated SST, (c) Oberhube r 's estimate of the net heat flux into the ocean (Q), and (d) simulated Q. Maps are 50oS-50oN, 120oE-70oW. Units of (c) and (d) are W m-2.

Figure 2: Eastern Pacific (150oW-100oW) averages of annual mean Q and its compon ents (positive down), from 30oS-30oN, in W m-2. (a) simulated, (b) observed (Obe rhuber 1988). Key: thick solid\120Q, thin solid\120shortwave, dash\120latent, do t\120sensible, dot-dash\120longwave.

Figure 3: Annual harmonic for same section as Fig. 2. (a) model SST, (b) observe d SST, (c) model Q, (d) observed (Oberhuber 1988) Q. Contour intervals: 0.5K (a, b), 20 Wm-2 (c,d).

Figure 4: Annual harmonic of SST and Q along the equator. Contour intervals: 0.2 5K (a,b), 5 Wm-2 (c,d).Otherwise same as Fig. 3.