海洋环流与波动重点实验室知识库

数值模拟是研究海洋及其变化的重要工具之一，但在当前的海洋与气候模式中，海温的模拟存在很大的误差，包括赤道太平洋“冷舌”模拟过冷与温跃层强度模拟偏弱等，其中一个主要原因在于海洋垂向混合过程的参数化方法中，一些关键参数的确定有很大的不确定性和人为性。因此本文系统地研究了导致误差出现的原因，并提出了一个新的优化方案，旨在减小热带太平洋温度的模拟偏差。

在本论文的第一部分，我们首先对比了两种垂向混合方案在热带太平洋海洋环流模式中的表现。一种是基于经典混合层模型KTN的Chen混合方案，另一种是基于湍流封闭模型的KPP混合方案。总体而言，Chen方案对热带太平洋SST的模拟优于KPP方案，但是，会加大次表层的暖误差。这主要是由于Chen方案在赤道外地区高估了风搅拌的混合效应。相比于KPP方案中流剪切不稳定模型，Chen方案中的Peters模型估算的垂向涡扩散系数更小，导致模拟的赤道东太平洋SST更暖。为了进一步优化KPP方案，我们将Peters模型引入到KPP方案中，结果显示KPP方案的模拟结果得到了极大的改善，“冷舌”的模拟偏差降低约30％，并且不会引起Chen方案中对次表层温度的负面影响。

在本论文的第二部分，我们讨论如何对Chen混合方案进行改进。通过影响海表面温度的变化，混合层深度在气候系统中扮演着重要的角色。为了描述混合层深度的变化，KTN整体混合层方案早在上世纪六十年代被提出，并且被许多海洋环流模式所应用。但是，KTN模型在模拟热带太平洋混合层深度时存在较大的误差。部分原因是由于在KTN模型中，风搅拌引起的混合效应表征为2m0u*3 ，而m₀的选取存在一定的不确定性。按照传统做法，m₀的取值为空间一致的常数，但这种假设不符合最近的观测研究结果。因此在本文的研究中，我们利用采用观测反演的方法，计算获得了时空变化的m₀，并将其应用于一个热带太平洋海洋环流模式中，评估时空变化的m₀对模式模拟结果的影响。结果表明，应用反演方法获得的m₀可以较大地提高模式对热带太平洋混合层深度的模拟。同时，模拟的热带太平洋浅层经向翻转流更强，造成赤道东太平洋存在更强的上升流。更强的上升流带来更多的次表层冷水，引起SST降低。本文进一步讨论了风应力与混合层深度变化的对应关系，对m₀的时空分布进行物理解释。

在本论文的第三部分中，我们对海洋模式中背景混合系数的表述进行优化，并利用海洋和气候模式研究其对热带太平洋海温场模拟的影响。用于刻画海洋内部垂向混合的背景混合系数，其取值通常采用10^-5 m²/s。然而在热带太平洋地区，近些年来的观测结果表明，背景混合系数应是10^-6 m²/s量级，远低于模式中的取值。虽然现阶段对热带地区弱背景混合系数的现象有了一定的认识，但如何将其应用于数值模拟中，以及量化其与模式误差的关系，一直是重要的研究课题。湍流微尺度观测资料匮乏，无法满足数值模拟对覆盖全海盆取值的要求。本研究采用Argo浮标资料，利用细尺度参数化方法，计算了热带太平洋背景混合系数的空间结构。并将其应用到海洋环流与气候模式中，结果表明“冷舌”与温跃层的模拟得到了很大的改善，可以有效减小25%的模式误差。这是由于背景混合系数的减小直接导致通过温跃层向海洋内区输送的热量减少。热量积累在混合层之下、温跃层之上，导致上层海洋层结降低，Ekman层厚度增加。引起赤道上升流减弱，垂向平流过程的冷却效应减弱，导致SST升高。温跃层内背景混合系数的减小直接导致次表层获得热量有所降低；同时赤道外的次表层降温可以通过副热带环流平流的作用输送到赤道区域，可以进一步加强次表层的降温过程。

本文最后综合提出集以上参数化方案于一体的优化方案。该方案集合了KPP方案和Peters流剪切不稳定模型的优势，并利用细尺度参数化方法估算了背景混合系数。将其应用于海洋环流模式的实验表明，赤道上的“冷舌”误差减小了70%左右。该方案也可以方便地应用于其他海洋和气候模式中以有效减小模拟误差。因此，本研究对于认清海洋及气候模式误差产生机理、改进模式模拟和预报能力都具有重要科学意义和应用价值。

Numerical simulation is one of the most powerful tools for studying the ocean and climate. However, model biases are still substantial in ocean and coupled ocean-atmosphere simulations in the tropical Pacific Ocean, including the too-cold tongue and too diffuse thermocline. These biases can be partly attributed to vertical mixing parameterizations in which the empirical parameters have great uncertainties. In this study, the bias problem is investigated using observations data and models, and a new scheme is proposed aiming to alleviate the temperature biases in the tropical Pacific Ocean.

In the first part of this dissertation, we have studied the impact of two different vertical mixing schemes on the solution of a tropical Pacific OGCM. In the conventional KPP scheme, the vertical eddy viscosity and diffusivity are determined based on a diagnosed boundary layer depth and a turbulent velocity scale. In contrast, the Chen scheme is mainly based on the KTN mixed layer model and Peters' shear instability model. Overall, the Chen scheme produces more realistic SST simulations in the eastern equatorial Pacific than KPP scheme. However, the subsurface warm bias is enlarged, mainly due to the overestimated wind stirring effects off the equator. The improved SST simulation can be attributed to the Peters' shear instability model, which generally produces a lower diffusivity than its counterpart in KPP scheme. Introducing the Peters' shear instability model into the KPP scheme can reduce the cold bias by 30%, and does not cause the deterioration in subsurface temperature simulation.

In the second part of this dissertation, we discuss the improvements in the Chen mixing scheme. As recognized, the mixed layer depth (MLD) plays an important role in the climate system through its influences on SST. The KTN bulk mixed layer model is designed for describing the MLD and has been adopted widely by many ocean modeling. However, large biases exist in the MLD simulation using the original KTN model in the tropical Pacific. This is partly due to the uncertainties in representing wind stirring effect in the model, which is scaled by a parameter (m₀). Traditionally, m₀ is taken as a constant uniformly in space. In this study, the m₀ is estimated as spatially and seasonally varying through its inverse calculation from a balance equation describing the turbulent kinetic energy budget of the mixed layer. It is illustrated that the m₀ is spatially and seasonally varying over the tropical Pacific. The derived m₀ fields are then embedded into an OGCM. Compared with the observations and the Global Ocean Data Assimilation System analyses, the MLD simulations in the OGCM with varying m₀ are substantially improved in the tropical Pacific Ocean on seasonal and interannual time scales. Additionally, the Pacific subtropical cells become intensified, accompanied with the strengthening of upwelling in the eastern equatorial Pacific; thus, more realistic simulations are obtained when using spatially and seasonally varying m₀ case compared with the constant m₀ case. As the related cooling effect from the upwelling is enhanced, the simulated SST is slightly cooled down in the eastern equatorial Pacific. Further applications and implications are also discussed.

In the third part of this dissertation, we optimize the depiction of background diffusivity and investigate its influences on tropical SST simulations. The background diffusivity, representing the integrated effects of diapycnal mixing processes in the ocean interior, is typically assigned with a constant value of 10^-5 m²/s in the current OGCMs. However, recent evidence shows that the diffusivity is reduced by about one order of magnitude in the tropics, implying that the tropical diapycnal mixing is overestimated in many ocean and climate modeling. The overestimated mixing can degrade the simulations in currents and water mass properties, but its relationships with the possible biases are not well demonstrated, mainly due to the sparsity of microstructure measurements for describing the spatial pattern of background diffusivity. In order to fill the gap, finescale parameterizations are proposed in recent studies, which provide an opportunity to estimate the background diffusivity with a global coverage by using the Argo profiles. In this study, the spatial structure for the background diffusivity in the tropical Pacific Ocean is derived based on the strain-based finescale parameterization, and is then employed into the MOM5-based ocean-only and coupled ocean-atmosphere simulations. Simulations of SST and upper-ocean temperatures are substantially improved by employing the Argo-derived background diffusivity compared with the original scheme, including a reduction in subsurface warm bias and thus a more realistic equatorial thermocline. The improved simulations in temperature can be attributed to the regulation in the currents system. By inhibiting the heat transfer into the ocean interior, the heat is accumulated below the mixed layer, resulting in the decrease in upper ocean stratification and the increase in Ekman layer depth. Meanwhile, the shallow meridional overturning circulation slows down and the related upwelling weakens, leading to the upper layer warming. The cooling effect beneath the thermocline is induced both by the reduced heat transfer from the upper layer and the convergence of the colder water from off the equator.

By combining on the KPP mixing scheme with other schemes, a new optimized approach is proposed with Argo-derived background diffusivity and Peters' shear instability model embedded. This scheme can reduce the “cold tongue” errors by about 70% in MOM5, and can also be easily applied to other ocean and climate models. Therefore, this study has important scientific significance and practical applications in model improvements and climate modeling studies.