不同滤波方法对揭示全球海洋条带结构的比较

海洋条带结构是近年物理海洋学研究的一个新热点. 在海洋中, 条带结构往往被大尺度环流过程所掩盖. 把这种隐蔽的海水运动现象显现出来的办法是对时间平均的速度场进行空间滤波. 利用全球简单海洋资料同化分析系统资料和中国科学院大气物理研究所大气科学和地球流体力学数值模拟国家重点实验室的气候系统海洋模式模拟数据对三种一维滤波方法进行了比较, 分别是常用的高斯和汉宁滤波方法, 以及本文引入的切比雪夫滤波方法. 结果表明, 尽管三种方法均可获得条带结构, 但以切比雪夫方法为最佳; 另外, 设计高通滤波器时需设定截断频率, 而它的选定取决于对具体数据的频谱分析, 当选取的归一化的截断频率值在0.1和0.4之间时, 可以有效地揭示出条带结构在全球海域内的分布. 因此本文的研究方法为海洋条带结构的深入研究提供了一个有力工具.

Intercomparison of one-dimensional detecting methods of unveiling the global ocean striations

Striation in the ocean is a research frontier in physical oceanography. Interestingly, it has some “sisters and brothers” in Mother Nature, such as the Jovian belts, subtropical jet streams in the atmosphere, and zonal flows in plasma. This meso-scale oceanic phenomenon is, however, concomitant with but covered up by the macro-scale ocean currents or circulations. In order to unveil such zonal jet-like structures, a spatial filtering must be applied to the commonly available time-average data. Previous studies mostly focused on prominent features of striations, such as banded structures, and the generation mechanism; however, the differences revealed by applying different types of filtering methods have not received enough attention. In this paper we present a comprehensive study on the effectiveness of the different detection approaches to unveiling the striations. Three one-dimensional filtering methods: Gaussian smoothing, Hanning and Chebyshev high-pass filtering, are used to analyze SODA data and LICOM model outputs. The first two methods have been used in many previous studies; on the other hand, the Chebyshev filter is a newcomer for this purpose. Our results show that all three methods can reveal ocean banded structures, but the Chebyshev filtering is the best choice. The Gaussian smoothing is not a high pass filter, and it can merely bring regional striations, such as those in the Eastern Pacific, to light. The Hanning high pass filter can introduce a northward shifting of stripes, so it is not as good as the Chebyshev filter. In addition, a cutoff frequency is often needed in applying the high-pass filter, and this frequency depends on the spectrum analysis of the original data. In this paper, we discuss the filtering output and its spatial power spectra of three normalized cutoff-frequencies, 0.1, 0.3 and 0.7. When the cutoff-frequency is too low, the filtering is insufficient; on the other hand, if the cut-off frequency is too high, excessive filtering can happen. Our study shows that for analyzing the global ocean striations, the best normalized cutoff frequency domain is between 0.1 and 0.4. In addition, the bandwidth of striation for using the Chebyshev high pass filter to analyze the SODA data in a depth of 300 m is 150-300 km. In the general case, we propose to use the Chebyshev filter in lieu of Hanning or other methods for investigating ocean striations.