【作者】 于鑫；

【导师】 窦硕增；

【作者基本信息】 中国科学院研究生院（海洋研究所） ， 海洋生态学， 2014， 博士

【摘要】 耳石是存在于大多数硬骨鱼类内耳中的钙质沉积结构，目前耳石信息分析技术已广泛用于解决鱼类生态学中的诸多难题，如物种和群体识别、产卵场及洄游习性研究等。但在耳石形态学研究中，用以表征耳石形态特征的变量繁杂多样，尚未描述这些变量的系统归属特征、分类特点，对于各类参数的特点、应用效果以及常用体长校正方法缺乏系统对比分析和实证研究。在耳石元素指纹的LA-ICPMS分析方面，激光剥蚀点准确定位、有效剥蚀点识别及有效元素数据的甄别等关键技术尚未标准化，有待于深入研究。本研究选取我国近岸水域洄游性鱼类刀鲚(Coilianasus)、河口性鱼类凤鲚(Coiliamystus)以及定居性强的虾虎鱼科中华栉孔虾虎鱼(Ctenotrypauchen chinensis)、拉氏狼牙虾虎鱼(Odontamblyopuslacepedii)、六丝钝尾虾虎鱼(Amblychaeturichthys hexanema)、矛尾虾虎鱼(Chaeturichthys stigmatias)和斑尾刺虾虎鱼(Acanthogobius hasta)为研究对象，以矢耳石形态特征和微化学分析为手段，探讨耳石形态和微化学分析的相关方法论和技术，并实证研究它们在群体识别中的应用。主要研究结果如下：1)针对耳石形态分析中变量繁多的特点，以我国辽河口、秦皇岛和黄河口以及长江口的刀鲚群体识别为例，系统归纳和总结了耳石形态分析中涉及的变量特点，将其分为大小变量(Sizevariable)和轮廓变量(Shapevariable)两类，并结合两类变量进行了以上四个刀鲚群体的识别。结果表明，四个刀鲚群体的总体群体识别成功率相对适中(<60%)，秦皇岛群体正确识别率最低(34%-39.5%)，略高于随机分配概率，长江口群体正确识别率最高(64%-90%)2)变量筛选是耳石形态学分析中的一个关键问题，本文以刀鲚和凤鲚耳石形态学研究为例，归纳出以耳石形态学特征参数进行鱼类群体识别时参数筛选的基本原则：满足相应的统计假设以及具备潜在的群体识别能力。当参数数量相对较少时，采用层层筛选的方式得到所需参数；当参数数量较多时可结合主成分分析，以主成分代替原参数进行判别分析。3)在统计分析中非参数统计能够在参数统计无法适用的情况完成对分析主体的统计推断，大多数的参数统计过程均有对应的非参数统计过程。本文以辽东湾、渤海湾、胶州湾和长江口凤鲚群体识别研究为例，对比分析了参数统计与非参数统计的群体识别效果。以形状指数结合椭圆傅立叶系数进行的群体判别结果表明，非参数统计过程相比参数统计过程保留了更多的有效形态参数用于判别分析，最终取得的总体成功率更高(68.2%>46.2%)。4)体长校正是耳石形态对比分析中必须考虑的一个关键问题。本文以采辽东湾、渤海湾、黄河口、胶州湾和长江口的刀鲚和凤鲚样本为例，对比研究了三种常用的体长校正方法(协方差校正、残差校正和异速生长校正)，结果表明三种方法的模型拟合效果相近，未现显著的差异；异速生长校正在刀鲚群体识别中取得最高判别成功率(70.7%)，协方差校正在凤鲚群体识别中取得最高判别成功率(58.9%)，表明最适的体长校正方法因鱼种而异。同时，本研究在前人研究的基础上提出了一套耳石形态比较研究中统计分析过程的改进方法。5)表征耳石形态特征的形状指数和傅立叶系数变量自身特点及其在形态对比分析中的实际应用效果是学者普遍关注的科学问题。本文以辽东湾、渤海湾、黄河口和胶州湾的五种虾虎鱼群体为例开展了基于耳石形态学特征的鱼种和群体识别的研究。结果显示，两类变量的种间识别效果显著优于群体间识别效果(83.7%-98.6%vs.50%-75.8%)；傅立叶系数比形状指数更能提高种间或群体判别成功率,但形状指数在解释具体差异时更具优势。6)准确定位耳石核区并有效激光剥蚀测试样品是获得可靠的元素指纹ICPMS分析结果的关键技术。传统的耳石核区元素指纹单测点激光剥蚀取样分析容易因取样点的单一性和偏差而降低指纹元素分析的可靠性。本文构建了耳石核区多测点LA-ICPMS激光剥蚀取样-剥蚀点SEM显微结构验证技术，并以此技术进行了辽东湾、渤海湾、黄河口、胶州湾和长江口五个刀鲚群体识别研究。结果表明：耳石核区Sr:Ca和Ba:Ca比值是有效识别各刀鲚地理群体的元素指纹；基于两者的刀鲚群体识别总体判别成功率为72.7%，表现出了元素指纹较强的群体识别能力；多测点取样技术有效保证了证了剥蚀取样的质量和代表性，提高了群体识别的效率和准确性。综上所述，本论文以我国近海刀鲚、凤鲚和五种虾虎鱼为研究对象，研究了鱼类矢耳石形态分析和核区元素指纹分析在群体识别中的应用，确定了耳石形态参量的特点及分类，解析了耳石形态分析和元素指纹分析的技术关键方法，探讨了非参数统计在形态分析中的应用，提出了一套耳石形态比较研究中统计分析过程的改进方法，构建了耳石核区多测点激光剥蚀的元素指纹分析技术，拓展了耳石信息分析的应用，为开展后续研究提供了重要的理论基础和应用参考。更多还原

【Abstract】 Otoliths are calcified structures that exist in most teleost. Otolith informationanalysis technique has been used to solve many problems in fishery ecology, forexample, species and stock discrimination, spawning ground and migration habits.However, in otolith morphology analyses, too many variables are used to express otolithmorphological characteristics, and the effective characteristics and category of thosevariables have not been well investigated. The characteristics of various kinds ofvariables, their applications and the pros and cons of the commonly used size removalmethods have not been stuied, either. In otolith elemental fingerprinting by LA-ICPMS,the methods of precisey laser ablating, validation of ablation spots and effectivelychoosing elemental data are still not standardized and thus need to be furtherinvestigated. In this study, we investigate the methods of otolith morphometrics andelemental fingprinting and their applications in stock discrimination of fish thatcommonly inhabit Chinese coastal waters. These fish include anadromous Coilia nasus,estuarine Coilia mystus and five highly localized gobies (Ctenotrypauchen chinensis,Odontamblyopus lacepedii, Amblychaeturichthys hexanema, Chaeturichthysstigmatias and Acanthogobius hasta). The main findings are as follows:1) Because there are too many otolith morphological variables that can be used inotolith morphology analysis, reasonable classification of the variables is essentialwhen applying them to stock discrimination. In this study, we classified the otolithmorphological variables into size variables and shape variables (i.e. Fouriercoefficients and shape indices), which were used for discriminating among stocksof Coilia nasus (the Liaohe River Estuary, the Qinhuang Island, the Yellow Riverestuary and the Yangtze River estuary). The results showed that the overallclassification success was moderate (<60%), with the lowest (34-39.5%) inQinhuang Island stock (slightly better than randomly assigned) and the highest (64-90%) in the Yangtze River estuary stock.2) Selecting the appropriate morphological variables in stock discrimination is a keytechnique in otolith morphology analysis. From the results of discriminating amongstocks of C. nasus and C. mystus in this study, some basic principles for selectingotolith morphological variable were suggested. Firstly, the variables should strictlymeet the corresponding statistic assumptions and have potentials for discrimination among stocks. Secondly, when the number of variables is relatively small, effectivevariables can be selected according to step-by-step statistics. However, when thereare too many variables, principle components analysis should be first applied to geteffective variables from the original ones before they are run for further statisticalanalyses such as discrimination analysis.3) The nonparametric test is an option when the data assumptions for parametric testare not met, while most parametric tests undergo the corresponding procedures fornonparametric tests. Comparisons between parametric tests and nonparametrictests running on the otolith morphometrics of four C. mystus stocks (the LiaodongBay, the Bohai Bay, the Jiaozhou Bay and the Yangtze River estuary) wereinvestigated. Results of the stock discrimination using the combination of shapeindices and elliptic Fourier coefficients showed that nonparametric tests could keepmore effective morphological variables for stock discrimination, which producedhigher overall classification success (68.2%) than parametric tests (46.2%).4) Size adjustment is another important technique in otolith morphology analysis.Three methods (ANCOVA adjustment, residual adjustment and allometricadjustment) were compared to remove fish length effects, which were applied tostock discrimination of C. nasus and C. mystus (the Liaodong Bay, the Bohai Bay,the Yellow River Estuary, the Jiaozhou Bay and the Yangtze River Estuary). Theresults showed that the fitting effects of the models in the three methods weresimilar. The allometric adjustment obtained the highest classification success(70.7%) in C. nasus, while the ANCOVA adjustment got the highest classificationsuccess (58.9%) in C. mystus. This suggested that the effectiveness of each sizeadjustment might be species-specific. Moreover, a procedure that can improve thestatistical process in comparing otolith morphology analysis was proposed in thisstudy.5) It is of interest to fishery researchers how to identify effective otolithmorphological variables and apply them to species and stock discrimination. Thisstudy compared the shape indices and elliptic Fourier coefficients as well as theirapplications to species and stock discrimination in five goby species that inhabitedin the Liaodong Bay, the Bohai Bay, the Yellow River estuary and the JiaozhouBay (Ctenotrypauchen chinensis, Odontamblyopus lacepedii,Amblychaeturichthys hexanema, Chaeturichthys stigmatias and Acanthogobiushasta). The results showed that the classification success of species discrimination (83.7-98.6%) were significantly higher than stock discrimination (50-75.8%). Theelliptic Fourier coefficients performed better in improving classification successthan shape indices in stock discrimination.6) Accurately locating otolith nuclei and ensuring the effectiveness of laser ablationsare key techniques to acquire reliable LA-ICPMS analysis. The traditional singlelaser ablating of otolith nuclei may weaken the reliability of elementalfingerprinting due to the unicity and bias of the ablating spots. A multi-ablatingotolith nuclei technique for elemental fingerprint was established and was appliedto discriminating among stocks of C. nasus that were collected from the LiaodongBay, the Bohai Bay, the Yellow River estuary, the Jiaozhou Bay and the YangtzeRiver estuary. The results showed that Sr:Ca and Ba:Ca ratios were effective atelementally fingerprinting the stocks. The overall classification success was at arelatively high rate of72.7%, indicating that this technique could ensure highquality of laser ablating and thus improve the efficiency and precision of the stockdiscrimination.In conclusion, a number of methods of otolith morphology analysis and elementalfingerprinting for species or stock discrimination were compared and investigated inthis study. These included size adjustment, selection of effective morphologicalvariables, nonparametric test versus parametric test, shape or size variables versuselliptic Fourier analysis and multi-laser ablating in otolith nuclei for otolithfingerprinting etc.. A case study was conducted to validate each methodology, whichsuggested that each appropriate analysis methods could improve the effectiveness ofotolith analysis for species or stock discrimination. This study could help betterunderstand the fundamental theories of otolith information analysis and theirapplications in fishery ecology.更多还原