Report of talk given at CGSG/DICE workshop in Paris 24-27 June 1996

An Intercomparison of the Surface Fluxes over the Tropical Pacific in 9 Coupled GCMs

Andrew Robertson & Roberto Mechoso


In Mechoso, Robertson et al. ( 1995; Mon. Wea. Rev.), the mean state and seasonal cycle over the tropical Pacific were examined in 11 coupled ocean-atmosphere GCMs, none of which used flux correction. Certain systematic errors were found to be shared, almost without exception: (i) The equatorial cold tongue was found to be too intense, too narrow, and to extend too far west along the equator in October. Even in April, when the cold tongue is absent in observed SST, most of the models showed a remnant of it. (ii) SSTs were overestimated in a broad region off the coast of Peru, and in a narrow band near 10S. (iii) The models were generally characterized by a double ITCZ, or an ITCZ that migrated south of the equator in boreal spring. In general, the SST climatology was too symmetrical both meridionally about the equator in the eastern Pacific, and also in longitude across the Pacific.

The aim of this note is to assess progress since the MWR paper, by examining the simulated surface fluxes of heat and momentum together with SST (a) east-west along the equator (130E-80W), and (b) north-south (30S-30N), averaged across the eastern Pacific between 100W-150W. By examining the individual flux components, we hope to distinguish to what extent shared systematic errors arise from similar mechanisms. The models included are ARPEGE/OPA, COLA1 (Dewitt), COLA2 (Schneider), LMD/OPA, MPI, MRI, Tokyo University, UCLA and UKMO. All the models are newer versions than those in presented in the MWR paper, except UKMO; the UCLA model is essentially unchanged. Observed estimates (bold curves) of SST and heat flux are from Oberhuber (1988), with wind stresses from FSU (Cd=1.5e-3).

The Figures

Each figure shows all the model simulations of a single surface-flux component: the top two panels are the equatorial east-west sections (EW), with the meridional sections across the eastern Pacific (SN) below. We plot April (left) and October (right), the monthly extremes of the seasonal cycle in SST. The surface fluxes--are plotted on the grid of the atmospheric GCM, with SST plotted on a 2o-grid. The flux components considered are both components of the momentum flux, tx, ty, together with the net shortwave heat flux into the ocean (SW), and the latent heat flux (LE). The fluxes of longwave radiation and sensible heat are relatively small in magnitude. See key to line styles.

Fig. 1, SST: The cold tongue is still generally too intense in the central Pacific, with SSTs being too warm in the extreme east, so that the cold tongue in shifted westward. The gradient and its seasonal variation are generally reasonable in the central Pacific. The inter-model spread is large, with a significant improvement in certain models over the MWR paper. In the north-south, the spread is smaller, and reasonable meridional asymmetry about the equator is captured in October. The intense narrow cold tongue in the simulations contrasts with the Oberhuber estimates in the north-south plots, although the latter may be unrealistically smooth.

Fig 1

Fig. 2, tx: Magnitudes of tx are greatly underestimated at the equator over the central Pacific, so that there is an apparent mismatch with SST, whose equatorial gradient is realistic. However, SST is plotted on the OGCM grid and this narrow tongue is likely to be invisible to the AGCM. In all the models, the easterly wind stress is overestimated in the western equatorial Pacific, consistent with the unrealistic westward extension of the equatorial cold tongue in most of the models. The subtle zonal shift of the tx-maximum along the equator with season is not well captured by the models, and, for example, the MPI model simulates a much larger seasonal change in amplitude for these two months. In the north-south section, the underestimation of the south-east trades in April is clear; the spread is large in October.

Fig 2

Fig. 3, ty: The meridional component exhibits considerable spread between models. Some do capture the phase reversal along the equator in April associated with the ITCZ in the east and the SPCZ in the west, although no model captures the amplitude. Similarly in October, several models capture the northward shift of the pattern, but the amplitude tends to be weak over the eastern Pacific. In the north-south, there is still a tendency for many models to have an ITCZ south of the equator in April. The equatorward intensification of the southerlies is poorly caught of the eastern Pacific in October

Fig 3

Fig. 4, SW: The short-wave heat flux into the ocean (SW) is overestimated in many of the models, both along the equator, and over the eastern tropical Pacific. This systematic error is smaller south of the equator in April, where realistic stratus or an unrealistic southward-displaced ITCZ may contribute. Since the error in this region is larger in October, the latter is suspected. The inter-model spread in the short wave is large, with some models being quite realistic

Fig 4

Fig. 5, LE: The latent heat flux (LE) is quite realistic along the equator, except in the UCLA and UKMO models. This is despite the generally underestimated wind stress amplitudes, suggesting a compensation of errors. Indeed, away from the equator where the winds are more realistic, evaporation does tend to be underestimated by most models.

Fig 5

In summary, simulated SST and the surface heat and momentum flux components still exhibit substantial systematic errors, and the 9 models examined exhibit substantial spread.