Application of novel 3D imaging techniques to quantify biomass and secondary production associated with North Sea artificial structures.

人为结构被部署在海洋环境中，以支持世界各地的工业活动。固着的附生生物群迅速在海洋中定居，进而吸引移动的无脊椎动物、鱼类和顶级捕食者。了解海洋环境中越来越多的人造建筑物对生态系统的影响是研究的优先事项，也是支持全球可持续安装和退役做法的必要条件。二次生产是衡量通过食物网的能量流动的指标，与生态系统功能直接相关，因此二次生产是生态系统功能的替代。为了了解次生生产与更广泛的生态系统过程(例如，大型移动动物的行为)之间的关系，我们需要准确地预测次生生产(本提案的重点)，并将这些数据提供给生态系统建模人员。由于进行专门的生态调查的时间/成本限制，获取与近海结构有关的二手生产的定制数据受到限制。海上能源运营商使用遥控车辆(ROV)获取基础设施的视频，用于维护目的。这些视频涵盖了所有建筑类型、年代和地点。“运动结构摄影测量”的最新进展意味着，现在可以从这段视频片段中生成表观生物群的3D图像，并使用这些3D图像来估计表生生物群的生物体积。通过应用特定于分类群的转换因子，生物量可以转换为生物量，然后转换为次生产品。通过将3D成像与有监督的机器学习算法相结合来自动识别分类群(然后应用特定于分类群的转换)，可以快速处理大量的ROV数据，以产生整个构造/生产盆地的高分辨率次生生产量估计。在之前的可行性研究中，我们率先在温带和热带水域对人造结构进行了3D成像，并使用这些图像来估计表观生物群的生物量。我们已经开发和应用了协议，通过分类单元特定的校准曲线将生物体积转换为生物量。在这里，我们建议使用现有的ROV胶片，为位于北海和更广泛的英国水域的85个人造建筑物生成3D图像。从图像中，我们将估计观察到的类群的生物体积。然后，我们将开发/改进机器学习算法，以自动识别3D图像中的分类群，并应用特定于分类群的体积-质量校准曲线。我们将把这些发展综合起来，以估计85个人造建筑物的次生生产力，并建立一个随建筑物位置、类型和年龄的函数的次生生产力统计模型，该模型可应用于其他建筑物。我们的新方法将使我们能够(1)首次对整个北海生态系统内所有近海能源结构的二次生产量进行估计，(2)预测一系列安装/退役情景对生态系统功能的变化，以及(3)交叉验证/比较我们的估计与墨西哥湾、澳大利亚和泰国湾的自然珊瑚礁栖息地和结构，这些地区正在应用类似的技术。我们的研究针对INSITE2挑战2和3，将大大促进我们对人造结构所发挥的生态作用的理解，并作为支持对安装和拆除结构的生态系统规模后果的地方、区域和全球评估的证据基础。通过开发3D成像和AUTO-ID，我们还将提供一种新的监测工具，促进对近海结构物的监测/监管的战略性全系统方法。这样的工具可以很容易地应用于生态(和工程)应用的历史行业数据。

Anthropogenic structures are deployed in marine environments to support industrial activities worldwide. Sessile epibiota rapidly colonise structures in the sea, in turn attracting mobile invertebrates, fish and top predators. Understanding the ecosystem effects of the increasing number of man-made structures in marine environments is a priority for research, and necessary to support sustainable installation and decommissioning practices worldwide.Secondary production is a measure of energy flow through the food-web, and relates directly to ecosystem function, thus secondary production is a proxy for ecosystem function. In order to understand the relationship between secondary production and wider ecosystem processes (e.g. mobile mega fauna behaviour), we need to accurately predict secondary production (the focus of this proposal), and make this data available to ecosystem modellers. Obtaining bespoke data on secondary production associated with offshore structures is limited by the time/cost constraints of conducting dedicated ecological surveys. Offshore energy operators use remotely operated vehicles (ROVs) to obtain videos of infrastructure for maintenance purposes. These videos cover all structures types, ages and locations. Recent advances in "Structure from Motion Photogrammetry" mean that it is now possible to generate 3D images of epibiota from this video footage, and use the 3D images to estimate the biovolume of epibiota. Biovolume can be converted to biomass, then to secondary production, by applying taxa-specific conversion factors. By pairing 3D imaging with supervised machine learning algorithms to automatically identify taxa, (and then apply the taxa-specific conversions), large volumes of ROV data can be rapidly processed to produce high-resolution estimates of secondary production for entire structures /production basins. In a previous feasibility study, we pioneered 3D imaging of man-made structures in temperate and tropical waters, and used these images to estimate epibiota biovolumes. We have developed and applied protocols to convert biovolumes into biomass via taxon-specific calibration curves. Here, we propose to generate 3D images for 85 man-made structures located in the North Sea, and wider UK waters, using existing ROV footage. From the images, we will estimate the biovolume of the observed taxa. We will then develop/refine machine learning algorithms to automatically identify the taxa within the 3D images, and apply taxa-specific volume-to-mass calibration curves. We will bring these developments together to estimate secondary production on the 85 man-made structures, and develop a statistical model of secondary production as a function of structure location, type and age, which can be applied to other structures. Our novel approach will enable us to (1) generate, for the first time, an estimate of secondary production across all offshore energy structures within the whole North Sea ecosystem,(2) predict changes to ecosystem function stemming from a range of installation/decommissioning scenarios, and (3) cross-validate/compare our estimates to natural reef habitats and structures in Gulf of Mexico, Australia and the Gulf of Thailand, where similar techniques are being applied. Our research, which addresses INSITE2 Challenges 2 and 3, will significantly advance our understanding of the ecological role played by man-made structures, and serve as an evidence base to support local, regional and global assessments of the ecosystem-scale consequences of installing and removing structures. Through development of 3D imaging and auto-ID, we will also deliver a novel monitoring tool that facilitates a strategic whole-system approach to the monitoring/regulation of offshore structures. Such a tool could be readily applied to historic industry data for ecological (and engineering) applications.