基于多源遥感数据的太平洋中尺度涡克里金提取方法研究

发布时间：2018-10-14 19:18

【摘要】：有别于大尺度的海洋环流和小尺度的波动，海洋中尺度的涡旋能传播热、盐、碳、营养盐和其它示踪物，对大洋能量的传输起着重要作用。中尺度涡的有效提取和追踪对于认识区域海洋环流模式意义重大，并且能够进一步地通过包括涡动能(EKE)、涡活动指数(EAI)和涡强度(EI)在内的一系列参量来研究其与厄尔尼诺-南方涛动(ENSO)现象的关系，有助于分析厄尔尼诺(El Nino)及拉尼娜(La Nina)异常事件发生的原因。近年来遥感数据在中尺度涡的提取中应用广泛，主要包括高度计、海温、叶绿素a浓度及盐度遥感产品。基于此类数据的提取方法主要有海表面高度闭合等值线法(有阈值/无阈值)、混合方法，基于海温、叶绿素a浓度数据的图像处理方法，但这些方法存在缺陷：复杂的判据、敏感的阈值、有限的适用性及高门槛的计算性能要求，因而本文致力于发明一种基于遥感数据的判据简单、高效省时及可靠性强的新方法，克里金法却因其空间最优线性无偏估计及误差可定量化的优点而被采用并进行了实验。本文针对海洋中尺度涡提出一套基于高度计遥感数据开发的，可适用于多源海洋遥感数据的中尺度涡克里金提取算法及一套识别海洋中尺度涡及其属性的程序。对于高度计遥感数据，算法运用变差函数工具计算其变差场，定义为广义振幅场，再利用泛克里金插值消除数据虚警和噪声，借助广义振幅场与实际振幅的关系通过少数几条特征等值线实现涡旋及其属性的提取。本文针对太平洋不同海区，采用泛克里金法分别进行了实验。计算结果表明：运用泛克里金方法，(1)在西北太平洋2014年4月的4期海平面高度异常数据(SLA)中计算得出743个中尺度涡，包含433个气旋涡和310个反气旋涡；(2)在北太平洋2012年4月的4期数据中共计算得到841个中尺度涡，包含450个气旋涡和391个反气旋涡；总共出现过3个双核涡(均为气旋涡)，其中有一个持续时间至少为15天(4月4日至18日)；(3)在南太平洋2012年4月的4期数据中共计算得到923个中尺度涡，包含423个气旋涡和500个反气旋涡。与其他遥感技术手段的对比显示应用该算法可使太平洋中尺度涡的成功检测率(SDR)接近90%，过度检测率普遍低于20%。结果表明：(1)算法具有省时高效性，通过对海平面高度异常场的重构创建广义振幅场，避免了等值线筛选过程，相对于海面高度闭合等值线等其他遥感方法具有判据简单及提取速度快的特性；(2)可靠性好，能够通过推导得出的特征等值线确保稳定的提取准确度；算法建立在发展成熟的等值线提取方法之上，并且有变差函数与泛克里金法的理论支撑；(3)自适应性强，可以对任意海区进行实时的涡旋检测和提取，并且除振幅值统计资料及必要基础数据外无需依赖其他辅助性数据。本文也采用了协同克里金算法将不同源海表面温度数据(NOAA-AVHRR以及NASA-GHRSST)进行数据插值融合，并在太平洋局部海域验证了协同克里金提取算法的精度和可靠性。黑潮及其延伸体海域计算结果显示，在与高度计遥感数据同日期的2012年4月四期数据中，采用协同克里金算法计算出的有效涡旋个数为111，包括77个气旋涡和34个反气旋涡，气旋涡的SDR分别为35.94%、34.78%、30.88%及12.50%，反气旋涡的SDR分别为13.43%、12.20%、14.67%及4.88%。一方面，这表明协同克里金算法对海表温度场局部涡旋的侦测具有可用性；另一方面，海温数据在中小尺度的随机变异使得本就针对局部海域的侦测变得更有难度(算法成功检测率尚且无法逾越40%)，潜在应用价值与局限性并存。同时能够明显看出，气旋涡的识别精度远高于反气旋涡。本文最后讨论了叶绿素a浓度及海面盐度月平均数据对于中尺度涡旋提取的可行性。本文利用了变差函数工具对卫星遥感海洋数据空间结构进行了分析，在今后研究中可利用克里金方法进一步揭示海洋数据场的隐含空间信息，特别是针对遥感数据的研究；本方法对于高度计数据大范围快速监测中尺度涡的情形特别适用；同时，对局部小范围海域而言，本文提出的算法目前可利用海温遥感数据快速提取中尺度涡。算法首先运用变差函数工具从海表面信号场中提取涡旋区域，，再利用克里金插值消除数据趋势和噪声，最后中尺度涡标识参量与信号场数据的对应关系，确定特征等值线，实现涡旋及其属性(涡核、极性、振幅、面积、半径)的提取。本文研究结果表明：(1)本方法相对于海面高度闭合等值线等其他遥感方法具有判据简单的特性，能够显著加快等值线筛选过程及提取速度；(2)利用变差计算配合泛克里金法在插值过程中对趋势的移除和对噪声的屏蔽，能够通过特征等值线实现较高的提取准确度；(3)具有大范围快速、实时性的优势，适用于大范围海域中尺度涡的快速监测。
[Abstract]:Unlike large-scale ocean circulation and small-scale fluctuations, the ocean meso-scale vortex can spread heat, salt, carbon, nutrients and other chemicals, playing an important role in the transmission of ocean energy. the effective extraction and tracking of the meso-scale vortex is significant for the recognition of regional ocean circulation patterns and can further be achieved by including eddy kinetic energy (evke), A series of parameters, including the eddy activity index (EAI) and the eddy intensity (EI), are used to study the relationship between the El Nino-Southern Tao (ENSO) phenomenon and contribute to the analysis of the causes of abnormal events in El Nino and La Nina. In recent years, remote sensing data is widely used in the extraction of meso-scale eddy, mainly including altimeter, SST, chlorophyll a concentration and salinity remote sensing product. The extraction method based on this kind of data mainly includes sea surface height closed isoline method (with threshold/ threshold value), mixing method, image processing method based on sea surface temperature, chlorophyll a concentration data, but these methods exist defects: complex criterion, sensitive threshold value, The invention discloses a novel method based on remote sensing data, which is simple in criterion, high in efficiency and strong in reliability, The Kriging method has been used and experimented because of its spatial optimal linear unbiased estimation and quantitative error. In this paper, a set of meso-scale eddy Kriging extraction algorithm based on altimeter remote sensing data, which can be applied to multi-source ocean remote sensing data, is presented in this paper. For the remote sensing data of the altimeter, the algorithm uses the variation function tool to calculate its variation field, defines the generalized amplitude field, and then uses the generalized Kriging interpolation to eliminate the data. Virtual police and noise, by means of the relationship between the generalized amplitude field and the actual amplitude, the vortex is realized by a few characteristic isolines. In this paper, the generalized Kriging method is applied to different sea areas in the Pacific. The results show that 743 meso-scale vortices, including 433 gas vortices and 310, are calculated by using the generalized Kriging method, (1) in the 4-stage sea level anomaly data (SLA) in the Northwest Pacific in April 2014. An anti-gas vortex; (2) 841 meso vortices are calculated in the 4-period data of April 2012 in the North Pacific Ocean, containing 450 gas vortices and 391 anti-gas vortices; a total of 3 dual-core vortices (each being a gas vortex), with a duration of at least 15 days (4-Apr-4). Up to 18 days); (3) 923 meso-scale vortices, including 423 gas vortices and 500, calculated in the April 2012 data of the South Pacific Compared with other remote sensing methods, this algorithm can make the successful detection rate (SDR) of the Mesoscale eddy close to 90%, and the over-detection rate is common. The results show that: (1) The algorithm has the high efficiency, and the generalized amplitude field is created by reconstruction of the sea level height anomaly field, so that the isoline screening process is avoided, and other remote sensing methods such as contour contour and other remote sensing methods relative to the sea surface height are simple and high in extraction speed. (2) the reliability is good, the obtained characteristic contour can be derived to ensure stable extraction accuracy; the algorithm is established on the developed mature isoline extraction method and has a variable difference function and a theoretical support of the universal Kriging method; and (3) The self-adaptability is strong, the real-time vortex detection and extraction can be carried out on any sea area, In this paper, the data of different source sea surface temperature data (NOAA-AVHRR and NASA-GHRSST) is interpolated by a cooperative Kriging algorithm, and cooperative Kriging extraction is verified in the Pacific local sea area. The accuracy and reliability of the method show that in April 2012 data of the same date as the remote sensing data of the altimeter, the effective vortex number calculated by the cooperative Kriging algorithm is 111, including 77 gas vortex and 34 anti-gas vortex, and the SDR of the gas vortex is 35. 94, respectively.%, 34. 78%, 30. 88% and 12.50%, the SDR of the anti-gas vortex was 13. 43%, 12.20%, 14. 6, respectively. 7% and 4.88%. On the one hand, this indicates that the cooperative Kriging algorithm has availability for the detection of local vortex in the sea table temperature field. On the other hand, the random variation of SST data in small and medium scale makes the detection of local sea area more difficult (algorithm success rate is high). Can't exceed 40%), potential application The value and limitation coexist. At the same time, it can be clearly seen that the identification of the gas vortex In this paper, the mean data of chlorophyll a concentration and sea surface salinity are discussed in this paper. In this paper, the spatial structure of the satellite remote sensing ocean data is analyzed by using the variation function tool, and the implicit spatial information of the ocean data field can be further revealed by the Kriging method in the future research. The method is particularly suitable for the case of fast monitoring the meso-scale vortex with large data range of altimeter data; meanwhile, the algorithm proposed in the paper can utilize the SST remotely for local small-range sea areas. In this paper, the vortex region is extracted from the sea surface signal field by using the variation function tool, the data trend and noise are eliminated by using Kriging interpolation, the corresponding relation between the final meso-scale eddy identification parameter and the signal field data is determined, and the characteristics and the like are determined. Value line to realize vortex and its properties (vortex core, pole The results of this paper show that: (1) This method has the characteristics of simple criterion with respect to other remote sensing methods such as contour line height closed isoline and other remote sensing methods. and (2) using the variation calculation to coordinate the removal of the trend and the shielding of the noise during the interpolation process by using the variation calculation, so that higher extraction accuracy can be realized through the characteristic contour line; and (3) has the advantages of large-range fast and real-time performance,
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P731.2;P715.7

【参考文献】