海洋生态与环境科学重点实验室知识库

  群体是渔业资源管理和濒危物种保护的基本单元。群体判别方法有很多，其中耳石形态分析具有简单、高效和成本低等优点，并且可以用于不同时空尺度下的群体结构研究。傅里叶变换和形状指数是两种常用的耳石形态分析方法。而小波变换是一种常用的信号处理方法，它不仅可以分析耳石的精细特征，还能同时定位特征区域，但是尚未广泛用于相关研究中。此外，传统的耳石形态分析集中于耳石的整体轮廓，而对于耳石上的一些特征结构的形态研究较少。其中，听沟是耳石内侧面上一条横向的凹槽结构，也是耳石与听觉纤毛相互作用的主要区域，其形态特征因种类或群体而异，也是群体识别的潜在指标。目前听沟形态学分析只应用于少量的鱼种识别等研究，用于鱼类群体判别的研究则更少，并且所采用的形态学指标也较简单。耳石或听沟的各种形态分析方法有其优缺点，应用效率各异。本研究以中国近海小黄鱼（Larimichthys polyactis）、黄姑鱼（Nibea albiflora）和白姑鱼（Pennahia argentata）的不同地理群体为研究对象，利用形状指数、椭圆傅里叶系数和离散小波系数，对比不同耳石形态分析方法的群体判别效率，并探索性地研究了听沟形态分析在鱼类群体判别中的应用，以期为科学地选择耳石形态分析方法、提高群体判别效率提供新方法和科学依据。

  主要研究结果如下：

  1）总体而言，无论单独使用还是与形状指数相结合，椭圆傅里叶系数和离散小波系数得到的群体判别成功率都相近（差异≤3.0%），并且均高于基于形状指数所得到的群体判别成功率（≥16.1%）。这表明椭圆傅里叶变换和离散小波变换具有相同水平的提取耳石形态信息的能力，且提取的有效耳石信息比形状指数多。此外，当不同参数结合时（如椭圆傅里叶系数或离散小波系数与形状指数相结合），其群体判别成功率通常会高于单独使用一种参数。不同类型的参数提取的耳石信息不同，具有一定的互补性，二者的结合能获取更多的耳石信息，更有效地描述耳石形态，从而提高了其群体判别效率。

  2）基于听沟形态分析（椭圆傅里叶系数与形状指数相结合）的三种石首鱼的群体判别成功率达到70.8%-80.4%，表明听沟形态分析与耳石形态分析类似，是一种有效的群体判别方法。二者在黄姑鱼（73.8% vs 73.8%）和白姑鱼群体（80.4% vs 86.5%）判别中的判别成功率相近，但在小黄鱼的群体判别成功率相差较大（70.8% vs 88.2%）。获取的听沟轮廓精度是影响群体判别成功率的一个重要因素。听沟与耳石其它部分的颜色差异小，且听沟轮廓需要借助软件手动描绘，这在一定程度上影响了提取听沟形态信息的精细度。因此，开发高效的听沟轮廓提取方法是提高基于听沟形态分析的群体判别效率的基础。

  3）不同群体的耳石平均轮廓图和离散小波系数的组间差异图表明，耳石平均轮廓差异较大的地方也对应着小波系数较大的组间差异。本研究主要分析了第4层和第5层小波系数的组间差异。其中，第4层小波系数组间差异较大的地方通常能在耳石轮廓上找到相对应的区域；而少数第5层小波系数组间差异较大的地方在耳石轮廓上找不到相对应的区域。这可能是由小波变换本身的特性所决定的，即精细尺度（第5层）的小波系数更容易受到“噪声信号”的干扰。在本研究中，干扰信号表现为同一群体内不同个体间的耳石轮廓差异。总体上，小波变换可以很好地定位耳石的特征区域，而这是傅里叶变换所不具备的能力。

  4）基于椭圆傅里叶系数的耳石形态分析中，三种石首鱼的所保留的傅里叶谐值的组数分别为10（小黄鱼）、10（黄姑鱼）和9（白姑鱼），这些谐值的傅里叶总功效均达到了99.99%。通过椭圆傅里叶谐值来重建耳石的轮廓，可以清楚地发现每组谐值给耳石轮廓带来的变化，并且所保留的这几组傅里叶谐值都能精细地刻画耳石轮廓特征。相比较而言，在听沟形态分析中，三种石首鱼的所保留的椭圆傅里叶谐值的组数要稍多些，分别为12（小黄鱼）、9（黄姑鱼）和14（白姑鱼），这表明三种石首鱼的听沟轮廓比耳石轮廓更为复杂。通过重建耳石和听沟的轮廓，不仅可以验证所保留的椭圆傅里叶谐值的组数是否合理，还能更为形象地刻画耳石和听沟的轮廓特征。

  5）耳石和听沟的形态测量参数（重量、长度、宽度、周长、面积和相对面积比）在三种石首鱼不同群体间的差异性显著（P<0.05）。总体来说，长江口群体的形态测量参数值最大。耳石的沉积受到遗传和环境因素的影响。因此，传统的耳石和听沟形态测量参数能为群体间差异性分析提供具有生物学意义的指标，同时在一定程度上反映不同群体生活环境的差异，相关形态学参数在群体判别中也具有重要的应用价值。

  Understanding of stock is fundamental to fisheries management and endangered species conservation. Various methods have been applied to stock identification. Among them, otolith morphology analysis has its unique advantages including convenience, high efficiency and low cost. Additionally, otolith morphology analysis can be used for stock identification at different tempo-spatial scales. Wavelet transform is commonly used in signal processing. When adopted in otolith morphology, it can not only extract fine otolith information, but also locate the feature region. Compared to other two methods, Fourier transform and shape indices, wavelet transform has not been widely adopted in otolith morphology analysis. Traditional otolith morphology analysis commonly focuses on the whole otolith outlines, whereas some characteristic structures are less concerned. The sulcus is a longitudinal depression on the medial side of the otolith. It is also the major area where otolith interacts with auditory cilia. The morphology of sulcus differs among species or geographic stocks, and therefore has potentials in discriminating stocks. So far, there are few studies concerning sulcus morphology analysis, which generally deal with species identification. Moreover, the adopted parameters for sulcus morphology are also simpler than otolith morphology. Different methods of otolith or sulcus have their own merits or demerits, and thus the efficiency of these methods in stock discrimination also varies. The present study analyzed and evaluated the efficiency of different otolith morphology analyses in stock discrimination of three sciaenids along Chinese coast: the small yellow croaker (Larimichthys polyactis), the yellow drum (Nibea albiflora) and the white croaker (Pennahia argentata). They included shape indices, elliptic Fourier coefficients and discrete wavelet coefficients. In addition, the feasibility of adopting sulcus morphology analysis for stock discrimination was investigated and evaluated. The main goal of this study was to provide new insights into selecting different methods of otolith morpholy analysis and to improve stock discrimination.

  The main results are as follows:

  1) Whether using it alone or combining it with shape indices, the stock discrimination rate of elliptic Fourier coefficients or discrete wavelet coefficients was on the same level (differences ≤3.0%), which is higher than that obtained by using shape indices alone (≥16.1%). This indicated that elliptic Fourier transform and discrete wavelet transform could extract fine information of otolith morphology at comparable levels, whereas shape indices had less ability to describe the fine structure of otolith morphology. When combining different parameters, the stock discrimination rates were commonly promoted. Different types of parameters could extract different otolith information, which might be complementary to each other. Combining them could extract more and finer otolith morphology information than using them alone, and thus promoted the stock discrimination efficiency.

  2) When adopting sulcus morphology analysis (combing elliptic Fourier coefficients and shape indices) for discriminating stocks of the three sciaenids, the overall discrimination rates reach 70.8%-80.4%. This indicated that sulcus morphology analysis can discriminate stocks as effective as otolith morphology analysis. The efficiency of these two methods in discriminating stocks of yellow drum (73.8% vs 73.8%) and white croaker (80.4% vs 86.5%) were similar, whereas they showed relatively large difference in the small yellow croaker (70.8% vs 88.2%). The accuracy of acquiring sulcus outline was a decisive factor that could affect stock discrimination rate. Since the contrast between sulcus and other parts of otolith was usually low, it was sometimes difficult to identify sulcus outlines automatically by the software. Manually identifying and depicting the outlines could limit the accuracy of sulcus outline. Therefore, it is crucial to develop effective methods for extracting fine sulcus outline, which could improve the stock discrimination efficiency using sulcus morphology analysis.

  3) When comparing the reconstructed mean otolith outlines and wavelet coefficients variance among stocks, it was found that major differences along otolith outlines corresponded to large variance of wavelet coefficients. There were two levels of wavelet coefficients that were adopted to compare the variance among stocks. The variances that existed in level four wavelet coefficients were consistent with those in otolith outlines. But in level five, several small inconsistences were found between them. These might be determined by the property of wavelet transform: finer scale (level 5) of wavelet coefficients were more susceptible to “noise signals”. In this study, the noise signals could be induced by variances of otolith outlines among different individuals in each stock. Overall, wavelet transform could locate otolith feature regions well, and this could not be achieved by Fourier transform.

  4) In the otolith morphology analysis by elliptic Fourier coefficients, the number of Fourier harmonics adopted in each of three sciaenids was 10 (small yellow croaker), 10 (yellow drum) and 9 (white croaker). These Fourier harmonics explained more than 99.99% of the total Fourier power. By reconstructing the otolith outline based on elliptic Fourier harmonics, the change of otolith outline caused by adding each harmonic was obvious. It could also be found that the adopted Fourier harmonics could describe the otolith outline delicately. In the sulcus morphology analysis, the number of elliptic Fourier harmonincs adopted in each of three sciaenids was 12 (small yellow croaker), 9 (yellow drum) and 14 (white croaker). This indicated that the sulcus outlines of three sciaenids were more complicated than the corresponding otolith outlines. Generally, it could be validated whether the number of Fourier harmonics adopted in the study was appropriate by reconstructing the otolith and sulcus outlines. Moreover, the features along otolith and sulcus outlines could also be accurately depicted.

  5) The differences of otolith and sulcus morphometrics (weight, length, width, perimeter, area and relative area ratio) were significant (P<0.05) among the stocks of three sciaenids. Overall, the morphometrics of Changjiang Estuary stock were usually the largest. The deposition of otolith is regulated by genetics but are also influcenced by environmental factors. The conventional morphometrics of otolith and sulcus could provide biologically significant indicators for describing differences among geographical stocks and reflect habitat differences. This makes otolith and sulcus morphometrics useful parameters for stock discrimination.