海洋环流与波动重点实验室知识库

初始误差能够导致两类厄尔尼诺事件的预报结果产生显著的不确定性。本文使用基于主成分分析的粒子群智能优化算法（PPSO智能算法）将条件非线性最优扰动（conditional nonlinear optimal perturbation，CNOP）方法应用到复杂的全球海气耦合气候模式GFDL CM2p1（Geophysical Fluid Dynamics Laboratory Climate Model version 2p1）中，分别计算了两类厄尔尼诺事件的最优前期征兆（optimal precursor，OPR）和最快增长初始误差（optimally growing initial error，OGE），分析了它们的空间分布和发展机制，进而明确初始误差对两类厄尔尼诺事件可预报性的影响，能够得出以下结论：

（1）对GFDL CM2p1模式模拟能力进行评估。首先，从海表面温度、风场、次表层海温三个方面对GFDL CM2p1模式模拟热带太平洋气候态的能力进行了评估。结果表明，和观测相比，模式虽然会出现冷舌西伸、信风过强等问题，但仍能较好地模拟出热带太平洋气候态的基本空间分布，例如西太暖池–东太冷舌结构、赤道东风带和副热带信风带、温跃层深度自西向东减小的特征。其次，分别选取模拟效果较好的五个中太型和五个东太型厄尔尼诺事件进行合成分析，从海表面温度异常和次表层海温异常两个方面与观测对比，发现模式模拟结果中虽然存在海表面温度异常中心西偏和事件偏强等问题，但确实能够呈现具有不同海表面温度异常中心以及发展过程的两类厄尔尼诺事件。因此，GFDL CM2p1模式可以用来研究初始误差对两类厄尔尼诺事件可预报性的影响。

（2）东太型厄尔尼诺事件OPR与OGE的分析。应用PPSO智能算法，在GFDL CM2p1模式中计算了两个东太型厄尔尼诺事件的OPR和OGE。结果表明，OPR及其平均都在赤道西太区域（2°N–2°S，135.5°E–165.5°E）表现出正的海表面温度扰动（sea surface temperature perturbation，SSTP）和上层（0–205米）正的次表层海温扰动（subsurface temperature perturbation，STP），在赤道东太区域（2°N–2°S，79.5°W–109.5°W）则表现为上层（0–85米）负的STP。进一步的分析表明，OPR的发展机制主要为：赤道西太区域的正SSTP激发西风扰动，该扰动会在海洋中引起下沉的开尔文波，使得赤道西太区域的正STP随之不断向东传播，最终到达赤道太平洋东边界。由于该区域的海表面温度变化以垂向过程（即温跃层反馈）为主导，因此正STP会在海表激发出正SSTP，降低了海表面温度东西梯度并减弱了沃克环流，从而进一步增强赤道西太区域的西风扰动。因此，OPR的发展主要依赖于海表面温度、风场和海洋动力过程之间建立的海气耦合正反馈机制，即Bjerknes正反馈。OGE的分析同样表明，赤道西太区域存在与OPR类似的误差结构，且其发展同样受到Bjerknes正反馈的主导。综上可知，可以发现赤道西太区域的海温初始扰动对东太型厄尔尼诺事件的发生发展有着十分重要的影响，对应初始误差的存在会降低这类厄尔尼诺事件预报的准确度。

（3）中太型厄尔尼诺事件OPR与OGE的分析。根据中太型厄尔尼诺事件的特征修改PPSO智能算法参数进行OPR和OGE的计算。分析OPR的空间分布可以发现，赤道区域STP较小，热带外SSTP较强，主要存在两个大值区，分别为副热带北太平洋（20°N–40°N，175°E–140°W）和20°S以南。验证发现前者对于OPR的发展起着非常大的作用，后者作用较小。OPR的发展主要表现为副热带北太平洋处正SSTP通过影响局地大气场，在其南侧产生西南风扰动，减弱信风强度进而减弱蒸发。通过增加表层向下潜热通量减少了海洋向大气输送的热量，从而使正SSTP不断向赤道中太区域传播。到达赤道后，正SSTP一方面通过纬向平流反馈局地发展，另一方面通过产生局地西风扰动，激发出向东传播的下沉开尔文波，进而在东太区域通过温跃层反馈激发出正SSTP。然而由于赤道波动较弱，未能建立起有效的Bjerknes正反馈过程，最终在冬季发展成为东太SSTP较弱，中太SSTP占据主导地位的中太型厄尔尼诺事件。分析OGE的空间分布可知，随着误差约束的减小，表层误差强度减弱，副热带北太平洋关键区域以负的海表面温度误差（sea surface temperature error，SSTE）为主。对应的误差发展过程也与OPR类似，最终发展成为拉尼娜事件。根据目标函数的设定，OGE算例中正SSTE发展的初始误差同样能够具有较大的目标函数值，根据发展类型不同可将其分为东太型和中太型正初始误差。其中前者表现为赤道西太和中太区域正次表层海温误差（subsurface temperature error，STE）增强，发展过程主要依赖赤道波动过程；后者则表现为赤道区域正STE减弱，目标区域的正SSTE主要来自于热带外。根据以上结果，可以发现副热带北太平洋区域的初始扰动对中太型厄尔尼诺事件的发生发展有着十分重要的影响，此类初始误差会显著的影响中太型厄尔尼诺事件的预报效果。除此之外，赤道中西太区域的次表层海温初始误差同样也会导致中太型厄尔尼诺事件预报准确度的下降。

Initial error can cause significant amunt of uncertainties in the predictions of two types of El Niño events. In this study, the conditional nonlinear optimal perturbation (CNOP) approach is applied in the Geophysical Fluid Dynamics Laboratory Climate Model version 2p1 (GFDL CM2p1) through the principle component analysis based particle swarm optimization algorithm (PPSO). The optimal precursor (OPR) and the optimally growing initial error (OGE) of two types of El Niño are calculated to investigate the influence of initial error on the predictability of two types of El Niño. The conclusions are summarized as follows:

(1) Evaluate the simulation capability of GFDL CM2p1. Firstly the climatology of the tropical Pacific in the model simulation is evaluated from 3 aspects: sea surface temperature wind and subsurface temperature. Although the model simulation shows westward shift of the cold tongue and stronger trade wind, it is capable of capturing the fundamental character of the tropical Pacific Ocean such as the warm pool and cold tongue lacated in the western and eastern equatorial Pacific respectively, the easterly wind and trade wind in the equatorial and subtropical region and the deepened thermocline depth from east to west. Then the composite of five Central Pacific (CP) and Eastern Pacific (EP) type of El Niño events are used to verify that GFDL CM2p1 model can simulate two different types of El Niño event based on sea surface temperature anomaly (SSTA) and subsurface temperature anomaly (STA). Though the simulation shows slightly westward shift of the SSTA center and is slightly stronger in relative to observation, it is capable of simulating different types of El Niño events with separated features. Therefore, GFDL CM2p1 model can be used to study the influence of initial error on the predictability of two types of El Niño events.

(2) The OPR and OGE of EP type of El Niño are analyzed. By applying the PPSO algorithm, two OPRs and OGEs are obtained for EP type of El Niño event. The results suggest that both OPRs and their composite result show positive sea surface temperature perturbation (SSTP) and subsurface temperature perturbation (STP) in the western equatorial Pacific (2°N–2°S, 135.5°E–165.5°E) and negative STP in the upper layers (0–85m) of eastern equatorial Pacific (2°N–2°S, 79.5°W–109.5°W). The evolutions of the OPRs are listed as follows. The positive SSTP in the western equatorial Pacific induces westerly wind perturbation then triggers down-welling Kelvin wave, which then transports the positive STP eastward. As the positive STP reaches the eastern boundary of the Pacific, a positive SSTP emerges through thermocline feedback which in turn reduces the SST zonal gradient. The Walker Circulation is weakended and the westerly wind perturbation is intensified. Thus, Bjerknes positive feedback is established through SST wind and oceanic dynamic. Further analysis of the OGEs spatial patterns and evolutions also shows that the OGEs exhibit similar error structure in the western equatorial Pacific. In conclusion, for EP type of El Niño, the initial sea temperature perturbations in the western equatorial Pacific have a significant influence on it and the related initial errors can significantly lower the accuracy of the prediction of EP type of El Niño events.

 (3) The OPR and OGE of CP type of El Niño are analyzed. The parameters of the PPSO algorithm are modified based on the characteristics of the CP type of El Niño events in order to obtain the OPRs and OGEs for it. Analyzing the spatial pattern of the OPRs and their composite, the following conclusions can be drawn: the STP in the tropical Pacific is small comparing to the SSTP outside the tropical Pacific; the large SSTPs are mainly located in two separated area: the subtropical northern Pacific (20°N–40°N，175°E–140°W) and south of the 20°S. The importance of these two SSTPs is tested and the results show that the first one is crucial in the evolution of the CP type of El Niño event and the second one is less important. Further analysis of the evolution of the OPRs shows that the positive SSTP in the northern subtropical Pacific induces local atmospheric perturbation and triggers southwesterly wind perturbation. Then the magnitude of the trade wind decreases and evaporation is reduced resulting less heat transferred from the ocean to the atmosphere. Therefore, the positive SSTP moves equatorward. After the positive SSTP reaches the central equatorial Pacific it evolves locally through the zonal advection feedback. The positive SSTP in the central equatorial Pacific also triggers local westerly wind perturbation and induces the down-welling Kelvin wave which then induces positive SSTP in the eastern equatorial Pacific through the thermocline feedback. However, because the strength of the down-welling Kelvin wave is rather small, the positive SSTP in the east can not form a basin wide Bjerknes feedback and slowly dissipates. Eventually the OPRs evolve into a mature CP type of El Niño event. Analysis of the OGEs error structure shows that the magnitude of the sea surface temperature error (SSTE) decreases as the constraint reduces and the SSTE located in the northern subtropical Pacific shows a negative sign. The OGEs show similar evolution as the OPRs but with an opposite sign. According to the algorithm, the positive SSTE also can have large objective function value. Based on the result of the control experiment, there are two types of positive results: CP type and EP type. The EP type of positive results show similar error structure and evolution as the OPRs for the EP type of El Niño event while the CP type of positive results show similar evolution as the OPRs the CP type of El Niño event. In conclusion, the initial perturbation in the subtropical northern Pacific plays an important role in the evolution of the CP type of El Niño event and the related initial errors can significantly lower the accuracy of its prediction. In addition to that, the subsurface initial error in the western and central equatorial Pacific can also result in an unreliable prediction of the CP type of El Niño event.

目  录

第1章  绪论. 1

1.1  研究背景及意义. 1

1.2  国内外研究现状. 6

1.2.1  两类厄尔尼诺事件物理机制研究进展. 6

1.2.2  两类厄尔尼诺事件物理模拟与预测. 9

1.2.3  两类厄尔尼诺事件可预报性研究现状. 10

1.3  论文主要内容以及章节安排. 12

第2章  方法介绍. 15

2.1  条件非线性最优扰动（CNOP）方法. 15

2.2  PPSO智能算法计算CNOP的试验步骤. 16

2.3  小结. 20

第3章  模式介绍以及模拟能力验证. 21

3.1  GFDL CM2p1模式简介. 21

3.2  GFDL CM2p1模式对热带太平洋气候态的模拟能力. 22

3.3  GFDL CM2p1模式对两类厄尔尼诺事件模拟能力. 24

3.4  小结. 32

第4章  东太型厄尔尼诺事件算例分析. 35

4.1  引言. 35

4.2  试验设置. 37

4.3  东太型厄尔尼诺事件最优前期征兆空间分布. 42

4.4  东太型厄尔尼诺事件最优前期征兆发展物理机制. 47

4.5  东太型厄尔尼诺事件最快增长初始误差空间分布. 51

4.6  东太型厄尔尼诺事件最快增长初始误差发展物理机制. 55

4.7  小结与讨论. 58

第5章  中太型厄尔尼诺事件算例分析. 61

5.1  引言. 61

5.2  试验设置方案. 62

5.3  中太型厄尔尼诺事件最优前期征兆空间分布. 67

5.4  中太型厄尔尼诺事件最优前期征兆发展物理机制. 70

5.5  中太型厄尔尼诺事件最快增长初始误差空间分布. 77

5.6  中太型厄尔尼诺事件最快增长初始误差发展物理机制. 82

5.7  小结与讨论. 91

第6章  总结与讨论. 93

6.1  主要内容. 93

6.2  主要结论. 93

6.3  主要创新点. 95

6.4  讨论. 96

参考文献. 97

致谢. 109

作者简历及攻读学位期间发表的学术论文与研究成果. 111