Multifractal filtering method for extraction of ocean eddies from remotely sensed imagery

1 IntroductionCurrently, the ocean eddy plays an important rolein energy flowand matter exchange in ocean. The ex-traction and analysisof the oceaneddave beeninves-tigated bylots of people in pastdecades. Currentmeth-ods to deal with the ocean eddy include static and dy-namic methods. The static method is to analyze indi-rectly ocean eddies using the spatial feature distribu-tion, change, growthand decline of the watermass (Suet al., 1996; Guo et al., 1995; Guo and Ge, 1997);the dynamic meth…     

多重分形方法在涡旋遥感信息处理提取中的应用研究

目前提取遥感影像中的涡旋信息主要用边缘探测技术,例如Canny算子和Hough算子等,然而由于涡旋信息和遥感影像本身的复杂性和模糊性,用传统的方法不能取得理想的提取结果.基于多重分形技术,利用海洋涡旋信息的物理特征时空形态的自相似性来提取涡旋,用NASA MODIS,Sea WiFS和NOAA在典型海区如边界流的卫星影像进行试验,研究结果表明,用多重分形方法获取的涡旋信息非常有利于对海洋涡旋的深刻认识和分析研究.     

基于传输函数标准形态模式的海洋涡旋温盐和压力异常三维结构交互可视化

随着海洋观测技术和海洋数值模拟技术的发展,海洋学家已经能够获取海量的时空连续的海洋数据。但目前人们对于三维海洋大数据缺乏高效的交互可视探索手段,因此,利用科学可视化技术展现海洋数据,以及更深一步地挖掘时空数据规律是建立对海洋大数据从感知到认知的关键技术桥梁。本文利用体绘制技术从传输函数特征点数量、特征颜色映射以及表现特征的线型三个方面,设计针对海洋温盐异常数据以及压力异常数据的传输函数标准形态模式,能够直观有效地表现海洋涡旋温盐和压力异常三维结构。本文优化了球面光线投射算法,在保证绘制质量的情况下,实现了大规模海洋三维涡旋实时交互展示;同时实现了三维数据的内部漫游功能,能够更加清晰且完整地表现局部海洋涡旋的三维结构。通过可视分析发现,处于黑潮延伸体附近的气旋涡(CE, Cyclone Eddy)温度异常呈“上正下负”的三维结构;由于涡旋引起的海水升降运动,部分反气旋涡(AE, Anticyclone Eddy)盐度异常呈现出“上正下负”的三维结构;相较于温盐异常结构,海洋涡旋压力异常结构轮廓更为清晰且不会出现上下不一致的情况。     

基于卫星测高数据的海洋中尺度涡流动态特征检测

为了最终实现对海洋中尺度涡流(简称中尺度涡)的自动采样,首先应该发展中尺度涡动态特征识别技术。本文基于SLA(Sea Level Anomaly)数据,实现了对中尺度涡动态特征的检测算法。主要内容是制定了一个判别相邻两组SLA数据中的涡流,是否为同一涡流子在不同时刻的状态的标准,即判别下一时刻SLA数据中是否存在涡流是由上一时刻确定的被检测涡流演化而来的。通过确定这种进化关系,可以得到被检测涡流的一系列动态状态信息,例如:面积变化速率、中心移动情况以及其他情况。本算法的计算量不大,从而可以应用到实时涡流跟踪的环境中。值得注意的是,本文中的算法不仅仅局限于应用SLA数据,SSH(Sea Surface Height)等大部分反映海洋高度的数据也可以使用。     

海洋中的天气式涡旋

将由海洋中各物理过程产生的海水特性分布非均匀性，分为小、中、天气式和全球4种时、空尺度。对比副势带急民生罗斯贝波变形半径，阐明海洋天气式涡旋小于和慢于大气中的数十倍。对比在高空锋带下方，低怪大气锋中出现地面涡，揭示海洋蛇曲中不产生高空锋带类似物及近海底锋；大洋中自由涡虽能集中于海洋上半部，便与大气中低空正、反气旋正相反。因而，锋流环和洋中涡能对局地天气起作用，流环能成为海洋子午向热传递的重要机制。     

Spatial information recognizing of ocean eddies based on virtual force field and its application

A new approach to detecting ocean eddies automatically from remote sensing imageries based on the ocean eddy's eigen-pattern in remote sensing imagery and "force field-based shape extracting method" is proposed. First, the analysis on extracting eddies' edges from remote sensing imagery using conventional edge detection arithmetic operators is performed and returns digitized vector edge data as a result. Second, attraction forces and fusion forces between edge curves were analyzed and calculated based on the vector eddy edges. Thirdly, the virtual significant spatial patterns of eddy were detected automatically using iterative repetition followed by optimized rule. Finally, the spatial form auto-detection of different types of ocean eddies was done using satellite images. The study verified that this is an effective way to identify and detect the ocean eddy with a complex form.     

台风-涡旋相互作用研究进展

台风轨迹和强度的预报一直备受关注,也是大气和海洋学家最为关注的科学问题之一。近些年,随着观测技术和计算机模拟的快速发展,发现海洋中尺度过程和台风有紧密的联系。本文针对该主题进行简要回顾和总结, 主要侧重2个方面：①台风如何影响海洋中尺度过程？强调了地转过程的重要性和高度计观测对台风影响涡旋的低估;②海洋中尺度过程如何影响台风？强调了冷暖涡的不同以及位置不同对台风强度的影响。在此基础上,探讨台风-涡旋相互作用研究中尚未解决的问题：涡旋对台风的三维响应、多涡对台风的综合效应、时间尺度匹配和海气通量等。     

基于高分辨ROMS模式的黑潮延伸体次中尺度涡各向异性分析

次中尺度过程的水平空间尺度约为0.1~10km, 时间尺度约为1天, 里查森数和罗斯贝数为0(1), 能有效地从中尺度环流中汲取能量向小尺度湍流串级, 并对上层海洋物质的垂向交换有着重要影响。本文基于水平分辨率为~500m的高分辨率ROMS(regional ocean modeling system)数值模拟结果, 采用方差椭圆方法, 评估了黑潮延伸体海域上层海洋次中尺度涡旋的各向异性特征, 并探讨了涡旋各向异性值的大小与次中尺度过程特征参数的相关性。研究结果表明, 黑潮延伸体主轴强流区域的次中尺度涡旋各向异性值明显小于两侧海域, 主轴区域的次中尺度涡旋特征明显强于流轴两侧海域, 各向异性值与次中尺度过程的强弱有着较为显著的负相关关系, 表明次中尺度过程具有较小的各向异性特征(更趋各向同性)。方差椭圆表征了涡与平均流相互作用过程中的能量反馈机制, 较大的各向同性特征意味着动能更趋正向串级。     

Inversion of the three-dimensional temperature structure of mesoscale eddies in the Northwest Pacific based on deep learning

Mesoscale eddies, which are mainly caused by baroclinic effects in the ocean, are common oceanic phenomena in the Northwest Pacific Ocean and play very important roles in ocean circulation, ocean dynamics and material energy transport. The temperature structure of mesoscale eddies will lead to variations in oceanic baroclinity, which can be reflected in the sea level anomaly (SLA). Deep learning can automatically extract different features of data at multiple levels without human intervention, and find the hidden relations of data. Therefore, combining satellite SLA data with deep learning is a good way to invert the temperature structure inside eddies. This paper proposes a deep learning algorithm, eddy convolution neural network (ECN), which can train the relationship between mesoscale eddy temperature anomalies and sea level anomalies (SLAs), relying on the powerful feature extraction and learning abilities of convolutional neural networks. After obtaining the temperature structure model through ECN, according to climatic temperature data, the temperature structure of mesoscale eddies in the Northwest Pacific is retrieved with a spatial resolution of 0.25° at depths of 0–1 000 m. The overall accuracy of the ECN temperature structure is verified using Argo profiles at the locations of cyclonic and anticyclonic eddies during 2015–2016. Taking 10% error as the acceptable threshold of accuracy, 89.64% and 87.25% of the cyclonic and anticyclonic eddy temperature structures obtained by ECN met the threshold, respectively.     

Post-convection spreading phase in the Northwestern Mediterranean Sea

This is a study about the spreading of newly formed deep waters following open ocean deep convection in the Northwestern Mediterranean Sea. The main results are from the SOFARGOS large scale float experiment initiated in 1994–1995. During the SOFARGOS project, CTD stations and Lagrangian observations of ocean currents were carried out in the Gulf of Lion from December 1994 to July 1995. Hydrological observations confirmed that deep water formation occurred very early during winter 1994–1995 (late December, early January) in conjunction with atmospheric cooling, deep convection penetrating down to 2000 m in the so-called Medoc area. Numerous eddies (both anticyclonic and cyclonic) drifted away from the convection area and advected newly formed deep waters far away from the source region. In particular, compact anticyclones appeared to be the most coherent (long-lived) eddies and capable of transporting newly formed Western Mediterranean Deep Waters several hundreds of kilometers away from the convection area. Characterized by an inner core of about 5 km in radius, these eddies are submesoscale features in the outer domain and appear as key elements of the open ocean convection processes. During their long journeys, these eddies interacted with larger scale features such as the Northern Boundary Current, the North Balearic Front, topographic Rossby waves, and Sardinian eddies. These interactions influenced the long-term behavior of the eddies (mean drift, composition) and represented an important part of (1) the spreading phase following deep convection and (2) the large scale thermohaline circulation.     

基于经验正交函数和机器学习的南海海面高度异常预测

海面高度异常 （SSHA） 作为重要的海洋要素，对研究海洋温盐剖面、海洋涡旋等海洋动力现象具有重要意义。然而，传统的海洋预测技术存在着预测时效过短、预测过程复杂等诸多问题，现有的机器学习预测方法也只针对几个点或区域 进行平均，忽略了很多重要信息。因此，本文提出了一种基于经验正交函数和 BP 神经网络 （一种机器学习方法） 的 SSHA预测模型 （EOF-BPNN） 来实现对起报时刻后 30 天的南海 SSHA 预测。首先，对 1993 年 1 月 1 日—2013 年 12 月 31 日的逐日 SSHA 数据进行距平归一化预处理，构建相关系数矩阵，并对该矩阵进行 EOF 分解，获取主成分。然后将主成分输入 BP神经网络进行训练，实现对主成分的预测。最后将主成分预测值与相应的空间模态结合，获取 SSHA 预测值。结果表明，相较于惯性预报和气候态预报，EOF-BPNN 模型不仅能够提供提前 30 天的较为精确的 SSHA 和相应的涡旋演化过程预报，且在整个南海区域拥有更高的 SSHA 相关系数，证明了 EOF-BPNN 模型具有较好的预测性能。     

逆式回声测量仪在全球海洋中应用及前景展望

系统总结了逆式回声测量仪（IES）的发展历史，重点介绍了其目前在世界范围内的应用情况，并阐述了CPIES（集成流速和压力传感器的IES）在我国吕宋海峡周边布放的意义及对未来应用的展望。结果表明采取不同的排列方式布放的IES阵列所获取的数据通过结合地转经验模态GEM（gravest empirical mode）以及最优插值方法能够获得观测海区的温度、盐度和流场的三维空间分布特征。基于CPIES的应用，可深入地研究多尺度的海洋动力过程，如内波、内潮和中尺度涡等，对后续大洋环流理论发展、中尺度涡产生和消亡机理、海洋中不同尺度能量串级等具有重要意义。     

中国南海区1992-2012年的中尺度涡旋演化过程表达与统计分析

Automated identification and tracking of mesoscale ocean eddies has recently become one research hotspot in physical oceanography. Several methods have been developed and applied to survey the general kinetic and geometric characteristics of the ocean eddies in the South China Sea（SCS）. However, very few studies attempt to examine eddies＇ internal evolution processes. In this study, we reported a hybrid method to trace eddies＇ propagation in the SCS based on their internal structures, which are characterized by eddy centers, footprint borders, and composite borders. Eddy identification and tracking results were represented by a GIS-based spatiotemporal model. Information on instant states, dynamic evolution processes, and events of disappearance, reappearance, split, and mergence is stored in a GIS database. Results were validated by comparing against the ten Dongsha Cyclonic Eddies（DCEs） and the three long-lived anticyclonic eddies（ACEs） in the northern SCS, which were reported in previous literature. Our study confirmed the development of these eddies. Furthermore, we found more DCE-like and ACE-like eddies in these areas from 2005 to 2012 in our database. Spatial distribution analysis of disappearing, reappearing, splitting, and merging activities shows that eddies in the SCS tend to cluster to the northwest of Luzon Island, southwest of Luzon Strait, and around the marginal sea of Vietnam. Kuroshio intrusions and the complex sea floor topography in these areas are the possible factors that lead to these spatial clusters.     

Anticyclonic eddies in the Norwegian Sea; their generation,evolution and impact on primary production

Altimetry and ocean color observations are used in combination with a coupled physical-primary production ocean model to investigate anticyclonic eddies at two locations in the Norwegian Sea. Of particular interest are the formation of the anticyclonic eddies, and their influence on primary production. The formation of these anticyclonic eddies are due to baroclinic instabilities set up by shifts in the wind in north/south direction, leading to simultaneously formation of eddies throughout the area. After a density stratification develops in the upper 100 m of the water column, the anticyclones become a subsurface lens of well mixed water with the characteristics of intra-thermocline eddies. The deep mixed layer inside anticyclonic eddies delay phytoplankton bloom by approximately two weeks compared to the surrounding areas. As the mixed layer within the anticyclones become smaller than the critical depth, the combination of this and sufficiently high nutrient levels support a phytoplankton bloom. From the satellite observations, there is an evidence of phytoplankton being advected toward the center of the eddies, but also of isolated phytoplankton blooms within the intra-thermocline eddies. The combined use of a numerical model and satellite observations provides three-dimensional information on the structure and properties of both eddies and primary production. The presented model is particularly useful in cloud-covered areas where ocean color images are frequently unavailable.     

Eddy mass transport for the Southern Ocean in an eddy-permitting global ocean model

The eddy-induced mass transport is diagnosed for the Southern Ocean in an eddy-permitting global ocean model (OCCAM). The focus is on the transport by transient eddies in the deep ocean. The transport streamfunction is calculated in four different combinations of coordinate system. Depending on the coordinate system employed, the strength of transient eddy transport varies from 6 Sv meridional transport in latitude-density coordinates to 20 Sv across-streamline transport in streamline-depth coordinates. It is shown that transient eddies as well as standing eddies are necessary for cancelling the Deacon cell.In the Antarctic bottom water density layer, the major contribution of the transient eddies towards net equatorward transport occurs (a) as a strong transport over the narrow Drake Passage and (b) as a weaker but systematic transport over a broader region in the southeast Pacific where the Antarctic circumpolar current breaks up into multiple jets. In contrast, in the North Atlantic deep water density layer the net poleward eddy transport is spread out almost everywhere. This suggests that attention to eddies should not be restricted to places where the eddy transport has large magnitude.     

基于密度峰值聚类的中尺度涡轨迹自动追踪方法

中尺度涡信息的提取包括涡旋的识别和轨迹追踪，其自动识别与追踪对于基于海量数据的中尺度涡分析十分重要。传统涡旋轨迹自动追踪方法一般需要预先设定搜索半径的阈值，存在一定的主观性。针对传统中尺度涡轨迹追踪方法存在的问题，论文从聚类的角度出发，提出基于密度峰值聚类算法实现对涡旋轨迹的自动追踪，并以南海中尺度涡追踪为例，将基于聚类的追踪算法与传统的相似度追踪算法进行比较分析。结果表明:（1）基于密度峰值聚类算法，可实现对海洋中尺度涡的自动追踪，该算法涡旋追踪准确率优于传统相似度算法；（2）该涡旋追踪算法对资料的完整性依赖度较低，特别是对于存在部分缺损数据的情况仍能较准确追踪；（3）该追踪算法克服了传统涡旋追踪算法需要预先设定搜索半径阈值的问题，自适应性更强。     

基于深度学习的中尺度涡检测技术及其在声场中的应用

针对海洋中尺度涡的检测与参数提取问题,本文使用中尺度涡SAR图像数据集,提出基于深度学习的EddyYolo目标检测模型进行中尺度涡的涡旋中心和涡旋水平尺度的多目标检测,并且提取涡心位置和涡旋水平半径等参数.实验结果表明:本文提出的EddyYolo模型实现了涡旋中心和涡旋水平尺度的多目标检测,检测准确率达到94%.在此基础上,结合二维高斯涡模型和三维中尺度涡模型,本文提出了基于卫星遥感与声学对中尺度涡的联合建模方法.     

深度学习在海洋信息探测中的应用:现状与展望

深度学习可以通过深度神经网络,使机器理解学习数据,从而提高数据分类效果和预测结果的准确性,因此在海洋信息探测中应用越来越受到重视.作者基于深度学习的基本原理,阐述了海洋上常用的深度学习神经网络模型,并结合海洋信息探测要素,对温度、盐度、风场、有效波高和海冰等进行了海洋环境信息因子的预测分析;同时,对船舶、溢油和涡旋等进...     

面向典型海洋现象观测的水下滑翔机应用综述

水下滑翔机是一种浮力驱动的无人水下自治观探测平台, 具有持续性、鲁棒性、自主性等特点, 是构建海洋观测技术体系重要的水下航行器之一。文章从水下滑翔机发展历程、运行方式和专用传感器展开, 基于国内外海洋学文献综述研究, 重点对水下滑翔机在典型海洋现象观测应用方面进行总结。水下滑翔机的运行稳定性、长时续, 观测高分辨率、低成本的特点, 满足了对海洋中尺度涡、内波、湍流、环流/边界流和锋面的长时间、精细化立体观测要求; 鲁棒性、自主性的特点使其能够适应台风过境、地震、原油泄漏前后等紧急恶劣海洋状况的监测; 运行平稳、低噪声的特点使其成为声学监测与探测的良好平台。对比分析国内外的水下滑翔机(组网)应用状况可以发现, 国内水下滑翔机在传感器、数据处理和组网技术方面都需要进一步完善。最后总结并展望了国内外水下滑翔机面向海洋现象观测的应用。     

涡分辨率模式模拟的南海中尺度涡旋

Mesoscale eddies(MEs) in the South China Sea(SCS) simulated by a quasi-global eddy-resolving ocean general circulation model are evaluated against satellite data during 1993–2007. The modeled ocean data show more activity than shown by the satellite data and reproduces more eddies in the SCS. A total of 345(428) cyclonic eddies(CEs) and 330(371) anti-cyclonic eddies(AEs) generated for satellite(model) data are identified during the study period, showing increase of ~24% and ~12% for the model data, respectively. Compared with eddies in satellite, the simulated eddies tend to have smaller radii, larger amplitudes, a slightly longer lifetime, faster movement and rotation speed, a slightly larger nonlinear properties(U/c) in the model. However, the spatial distribution of generated eddies appears to be inhomogeneous, with more CEs in the northern part of SCS and fewer AEs in the southern part. This is attributed to the exaggerated Kuroshio intrusion in the model because the small islands in the Luzon Strait are still not well resolved although the horizontal resolution reaches(1/10)°. The seasonal variability in the number and the amplitude of eddies generated is also investigated.