参数化次表层上卷海温改进ENSO模拟

Improving ENSO Simulation by Parameterizing the Subsurface Entrainment Temperature

通过参数化次表层上卷海温改进了一个热带太平海洋模式的SSTA模拟。这种参数化方案通过经验方法将海洋上混合层底部海温变化与海表面起伏联系起来,从而可以方便地利用模式模拟的海表起伏描述温跃层的变化情况及其对混合层海温变化的影响。三组数值试验表明通过上述方法显著改善了SST年际变化的模拟,与观测相比,在赤道东太平洋及南美沿岸,距平相关系数由原来的0.7左右提高到0.8以上,均方根误差在赤道东太平洋由原来0.8℃降到0.6℃,在南美沿岸由1.3℃以上降为0.9℃。这表明在赤道东太平洋及南美沿岸,温跃层的变化通过夹卷过程及垂直扩散过程可以显著影响混合层的温度,OGCM对这些过程描述不足是导致SST年际变化模拟偏弱的一个重要原因,通过强调这些过程可以改善模拟效果。同时在热带西太平洋的改进也是显著的。

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 SSTA simulations, a separate SSTA submodel, in which Te is parameterized by an empirical nonlocal scheme, is embedded into a tropical Pacific OGCM. The To 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, anomaly 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 are used when constructing the T, 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.