A new phylogeny of ichthyosaurs (Reptilia: Diapsida)

The largest phylogenetic analysis of ichthyosaurs to date is presented, with 114 ingroup taxa coded at species level. Completeness of the taxa included varied from &gt; 98% to &lt; 2%; 10 taxa were removed <i>a priori</i> using Concatabominations due to incompleteness and taxonomic uncertainty. The data were analysed using three widely used optimization criteria, maximum parsimony, maximum likelihood and Bayesian inference; while similar, each criterion produced different topologies, support and levels of resolution. Maximum parsimony found a poorly-resolved consensus tree with moderate improvement from <i>a posteriori</i> pruning of unstable taxa; however, general support remained low. Tree resolution was reduced more by taxa that lacked codings from phylogenetically important regions of the tree, rather than by those that simply lacked many codings. Resolution present in the most likely tree is poorly supported; sister relationships cannot be confirmed, although similarities are found to the most parsimonious tree. Bayesian inference found poorly resolved consensus trees. While more resolved, an equal-distribution rate prior is significantly worse than the null gamma-distribution rate prior for morphological data, but suggests rate heterogeneity across ichthyosaur phylogeny. Tree comparisons under each analytical criterion failed to select a single best tree; however, the Bayesian inference tree with gamma-distribution rate prior is selected as the best tree based on recent analyses showing improved accuracy using this criterion. Unequivocally resolved clades include Ichthyopterygia, Ichthyosauria, Shastasauria, Euichthyosauria, Parvipelvia and Neoichthyosauria, but with variation in their taxonomic components. Mixosauridae and Ophthalmosauridae are similarly recovered, but their definitions are modified to stem-based definitions to prevent substantial variation of included taxa. Several genera are not monophyletic: <i>Brachypterygius, Leptonectes, Mixosaurus, Ophthalmosaurus</i>, <i>Paraophthalmosaurus</i>, <i>Phalarodon</i>, <i>Platypterygius</i>, <i>Stenopterygius</i>, <i>Temnodontosaurus</i> and <i>Undorosaurus</i>. Complex and variable relationships suggest the need for new characters and a re-evaluation of the state of ichthyosaur phylogenetics.

本研究呈现了迄今为止规模最大的鱼龙类（ichthyosaurs）系统发育分析，共纳入114个内群分类单元，均以物种水平进行编码。所纳入分类单元的完整性占比介于98%以上至2%以下；有10个分类单元因数据不完整及分类学存疑，在先验（a priori）分析阶段通过Concatabominations工具予以剔除。

本研究采用三种广泛使用的优化准则开展数据分析，分别为最大简约法（maximum parsimony）、最大似然法（maximum likelihood）与贝叶斯推断（Bayesian inference）；尽管三种方法的分析结果具有一定相似性，但各自生成的拓扑结构、支持率与解析度均存在差异。最大简约法得到一棵解析度较差的合意树，通过后验（a posteriori）剪除不稳定分类单元后，合意树的解析度得到一定程度提升，但整体支持率依旧偏低。相较于仅存在大量编码缺失的分类单元，那些缺失系统发育关键区域编码的分类单元，对树结构解析度的削弱作用更为显著。最大似然树所呈现的解析节点支持率普遍偏低，尽管其与最大简约树存在一定相似性，但无法确认其姊妹群关系。

贝叶斯推断得到的合意树整体解析度同样不佳。尽管使用均等分布速率先验的分析能得到解析度更高的树，但相较于零假设伽马分布速率先验，其对形态学数据的拟合效果显著更劣；不过两类先验均表明鱼龙类系统发育过程中存在置换速率异质性。基于各分析准则的树结构比较均无法选出唯一最优树，但结合近期研究表明该先验下的贝叶斯推断树可提升分析精度，因此将带有伽马分布速率先验的贝叶斯推断树定为最优树。

本研究明确解析的演化支包括鱼龙形超目（Ichthyopterygia）、鱼龙目（Ichthyosauria）、萨斯特鱼龙类（Shastasauria）、真鱼龙类（Euichthyosauria）、狭鳍类（Parvipelvia）与新鱼龙类（Neoichthyosauria），但各演化支的分类组成存在差异。混鱼龙科（Mixosauridae）与大眼鱼龙科（Ophthalmosauridae）同样被成功恢复，但为避免纳入类群出现大幅变动，二者的定义被修订为基干类群定义（stem-based definitions）。多个现用属并非单系群，包括短鳍鱼龙属（Brachypterygius）、细龙属（Leptonectes）、混鱼龙属（Mixosaurus）、大眼鱼龙属（Ophthalmosaurus）、副大眼鱼龙属（Paraophthalmosaurus）、狭颌龙属（Phalarodon）、扁鳍鱼龙属（Platypterygius）、狭翼鱼龙属（Stenopterygius）、切齿鱼龙属（Temnodontosaurus）与温德姆鱼龙属（Undorosaurus）。类群间关系复杂且多变，这表明亟需开发新的形态学特征，并重新评估鱼龙类系统发育研究的现状。