CTcoral – CyberTaxonomic Classification and Morphological Characterisation of Cold-Water Corals

形成骨架的冷水珊瑚(CWC)是重要的栖息地工程师，它们在深海形成珊瑚礁，形成重要的生物多样性热点。随着时间的推移，这些珊瑚礁建造了巨大的海底障碍物，称为珊瑚丘，在大陆边缘形成了重要的碳汇。虽然我们对化学武器公约的环境需求、化学武器公约珊瑚礁的全球分布和珊瑚丘形成的了解不断增加，但由于缺乏独立于观察者的定量方法，我们对化学武器公约关于其珊瑚石、殖民地和骨架形态的形态可塑性及其对这些系统的影响的了解仍然非常有限。CTcoral的目标是开发一种自动化方法，从分类和形态上描述最重要的框架--从三维(3D)计算机断层扫描(CT)形成CWC。为了实现这一目标，我们将实施机器学习技术，结合基于图像和基于形状的分类的优势。由于海洋骨骼和贝壳受到生物侵蚀的影响，从而改变了它们的形状，从而影响了算法的性能，因此我们将进一步开发分割生物侵蚀的算法，并产生生物侵蚀校正的数据集来补偿这种影响。此外，由于分类和形态特征的准确性和置信度取决于珊瑚碎片的大小和CT扫描分辨率，我们将测试小碎片和低分辨率造成的信息损失可以在多大程度上通过使用一个海底样本的多个碎片和通过纳入额外的形态标准来补偿。开发的方法将应用于来自大西洋各地的370个CWC样本，以提供最重要的CWC的形态可塑性的第一个全面描述。CTcoral是一个结合了海洋地球科学家和计算机科学家的专业知识的跨学科项目。它构成了后续提案的基础，该提案将侧重于应用所开发的方法，调查特定珊瑚岩、群落和骨架形态的环境偏好，以及珊瑚群落及其各自形态对珊瑚丘形成的影响。CTcoral内部开发的方法预计将广泛适用于其他生物群，并将在迈向网络分类学的道路上作出重要贡献。

Framework-forming cold-water corals (CWC) are important habitat engineers, which form reefs in the deep sea that present significant biodiversity hotspots. Over time these reefs build large seafloor obstacles, called coral mounds, which present important carbon sinks along continental margins. While our knowledge on the environmental needs of CWC, the global distribution of CWC reefs and the formation of coral mounds continuously increases, our understanding of the morphoplasticity of CWC regarding their corallite, colony and framework morphology and its influence on these systems is still highly limited due to the lack of observer-independent quantitative methodologies. CTcoral aims to develop an automated methodology to taxonomically classify and morphologically characterise the most important framework-forming CWC from 3-dimensional (3D) computed tomography (CT) scans. To achieve this goal, we will implement machine-learning techniques that combine the advantages of image- and shape-based classifications. Since marine skeletons and shells are affected by bioerosion, which alters their shape and hence impacts the algorithm performance, we will further develop algorithms to segment bioerosion and produce bioerosion-corrected datasets to compensate this effect. Furthermore, as the accuracy and confidence of the taxonomic classification and morphological characterisation depend on coral fragment sizes and CT scan resolution, we will test to which degree the loss of information by small fragment sizes and low resolution can be compensated by using multiple fragments from one seafloor sample and by incorporating additional morphological criteria, respectively. The developed methodology will be applied to 370 CWC specimens from all over the Atlantic Ocean to provide the first comprehensive description of the morphoplasticity of the most important CWC.CTcoral is an interdisciplinary project that combines the expertise of marine geoscientists and computer scientists. It forms the basis for a follow-up proposal that will concentrate on the application of the developed methodology to investigate environmental preferences of specific corallite, colony and framework morphologies and the influences of coral communities and their respective morphology on coral mound formation. The developed methods within CTcoral are expected to be of broad applicability to other organism groups and will provide an important contribution on the road towards cybertaxonomy.