Improving ENSO Simulation by Parameterizing the Subsurface Entrainment Temperature
Standard oceanic general circulation model (OGCM) simulations indicate that the simulated SST variability is underestimated in the eastern equatorial Pacific and along the coast of South America. The bias is common to the current OGCMs, which has been ascribed to the model deficiencies in the parameterization of entrainment and vertical mixing process. On the other hand, since the temperature of subsurface water entrained into the mixed layer (Te) is associated with these two terms: entrainment and vertical mixing process, it can be one of the major error sources for SST anomalies (SSTA) simulations in ocean and coupled ocean-atmosphere models. In order to improve SST A simulations, a separate SSTA submodel, in which Te is parameterized by an empirical nonlocal scheme, is embedded into a tropical Pacific OGCM. The Te parameterization scheme is developed in two steps. Firstly, an inverse modeling approach is adopted to estimate Te anomaly from the SSTA equation, using observed SST and modeled currents. Secondly, an EOF technique is used to build an empirical relationship between the modeled sea level elevation anomaly and the inversed Te anomaly, in which a multiple linear regression analysis is conducted on the EOF subspace. Therefore, the SSTA is simulated by the separate SSTA submodel in which Te a-nomaly is parameterized via the modeled sea level elevation anomaly. Three numerical experiments are carried out to examine the simulations. The first one is the standard OGCM run (control run). The second one is a SSTA submodel-embedded run in which the above Te parameterization is conducted, and will be referred to as "dependent run" hereafter because the whole monthly mean data during the period of 1961 -2000 are used to construct the Te parameterization. The third one is same as the second except that the model data for the simulated year in 1961 -2000 are excluded and the other 39 -year data arc used when constructing the Te parameterization (hereafter referred to as "independent run"). The comparative analysis shows that the SSTA simulation is indeed obviously improved due to the optimized empirical Te parameterization: the correlations in the eastern tropical Pacific and the south American coast are increased from about 0. 7 to above 0. 8 in comparison with the observation, and the root mean square errors are decreased from 0. 8 ℃ to 0. 6 ℃ in the eastern equatorial Pacific and from 1. 3 ℃ to 0. 9℃ along the south American coast, respectively. This indicates that Te, in association with entrainment/mixing process is an important factor affecting the SST anomaly in the eastern tropical Pacific which is underestimated in the current OGCMs. In addition, the simulation in western Pacific is also improved notably.
国家科技期刊平台
NSTL
万方数据
维普
