CTD Camera images analysis - collected from the TEMPO voyage of the RV Investigator, 2021

This dataset contains raw and annotated images from the CTD camera deployed during the TEMPO voyage. Raw images were annotated by Maddie Brasier in Biigle as per the Antarctic Seafloor Annotated Imagery Database (AS-AID) (Gros et al. 2023, Jansen et al. 2023). Two annotation catalogues were used including the ‘CATAMI – mobile and indicator species’ in which labels were based on the CATAMI classification scheme (Althaus et al., 2015) and, ‘VME morpho-taxa’ based on the CCAMLR guide to VME indicator taxa (CCCAMLR, 2009).Counts and percentage cover data of the identified species per image are given. Metadata files required for Biigle uploads included. Summary document of image processing included.Most krill studies focus on the epipelagic zone (surface to ~250m), but recent observations indicate that they can also exist within the abyssal zone. As a part of the TEMPO voyage a deep sea camera mounted to the CTD rosette was used to observe krill at the ocean floor. As a result seafloor images were collected during every full depth CTD deployment during the voyage. In these analyses the images were used to identify benthic fauna, the taxa present will be compared to the corresponding eDNA samples collected during the same CTD deployment. Taken from the summary document of image processing:Images were collected during CTD deployments via a Canon HFG10 mounted on to frame of the rosette. The camera and lights were switched on 25 m above the seafloor during the CTD decent and captured a still image every 10 seconds. At this point the focus was set to infinity. The CTD was lowered to 3-5 m from the seabed, after 2 minutes the focus was adjusted to 1 m. For each CTD deployment the highest quality image/s were selected for image analysis. Out of focus images were discarded, for some deployments multiple images were used as the if they captured different areas of the seabed and encompass as must potential benthic diversity as possible. Whilst the ship was as stationary as possible there was enough movement of the to capture different sections of the seabed with the roll of the ship and subsequent movement on the CTD cable. On average 2 images were used per station (Table 1). Whilst the best image/s were selected for each station in some instances there were no fauna present, or the image was out of focus (Table 1). During data analysis I suggest we remove the ‘unscorable’ out of focus images to prevent false negatives (i.e. a result that suggested no fauna present when actually no fauna scored because of out of focus image). Examples of high quality images and range images in Figure 1. Images were annotated using the web-based software Biigle (Langenkämper et al. 2017) and were uploaded together with their metadata including date, time, latitude, longitude, water depth, height above seabed and image area. Height above seabed was determined by the altimeter reading from the time each image was taken during the CTD deployment. The width and height of each image was calculated using a lens field calculator (https://www.deepsea.com/lens-field-calculator/), using the heigh above seabed and the sensor size of 1/3 inch and a native focal length of 4.25mm of the camera. From this the area calculated for each image and included in the metadata. Images were annotated as per the Antarctic Seafloor Annotated Imagery Database (AS-AID) (Gros et al. 2023, Jansen et al. 2023). Two annotation catalogues were used including the ‘CATAMI – mobile and indicator species’ in which labels were based on the CATAMI classification scheme (Althaus et al., 2015) and, ‘VME morpho-taxa’ based on the CCAMLR guide to VME indicator taxa (CCCAMLR, 2009). A single scorer annotated all images (MB) any ambiguous fauna were reviewed by independent experts (Charley Gros (IMAS) and Melanie Mackenzie (Museum Victoria)) but in some cases were still unidentifiable. Some key taxa were not listed in either catalogue and were labelled as either ‘indicator species’ or ‘no id’ to be edited in post processing, these labels were also used for some ambiguous taxa and notes taken (Table 2).

本数据集包含TEMPO航行期间部署的CTD相机（CTD camera）采集的原始图像及带注释图像。原始图像由Maddie Brasier在Biigle软件中按照南极海底带注释影像数据库（Antarctic Seafloor Annotated Imagery Database，AS-AID）（Gros等，2023；Jansen等，2023）进行注释。注释采用两个分类目录：其一为基于CATAMI分类方案（CATAMI classification scheme）的“CATAMI——移动与指示物种”（Althaus等，2015）；其二为基于南极海洋生物资源养护委员会（CCAMLR）VME指示分类群指南的“VME形态分类群”（CCAMLR，2009）。数据集提供每幅图像中已识别物种的计数及覆盖百分比数据，包含Biigle软件上传所需的元数据文件（metadata files）及图像处理总结文档。

大多数磷虾研究聚焦于上层水域（epipelagic zone，海面至约250米），但近期观测表明磷虾也可存在于深海区（abyssal zone）。作为TEMPO航行的一部分，研究人员使用安装于CTD采水器（CTD rosette）上的深海相机观测海底磷虾，因此航行期间每次全深度CTD部署均收集海底图像。分析中，这些图像用于识别底栖生物（benthic fauna），并将存在的分类群与同一CTD部署期间采集的对应环境DNA（eDNA）样本进行比较。

引自图像处理总结文档：

图像通过安装于采水器框架上的Canon HFG10相机在CTD部署期间采集。CTD下降过程中，相机及灯光在距海底25米处开启，每10秒捕获一张静态图像，此时焦点设为无限远。CTD降至距海底3-5米后，2分钟内将焦点调整为1米。每次CTD部署均选取最高质量的图像进行分析，失焦图像予以丢弃；对于部分部署，若多张图像捕捉到海底不同区域且能尽可能涵盖潜在底栖生物多样性，则采用多张图像。尽管船只尽可能保持静止，但船体摇晃及CTD电缆的后续移动仍足以捕获海底不同区域。每个站点平均使用2张图像（表1）；尽管每个站点均选取最优图像，但部分情况下无生物存在或图像失焦（表1）。数据分析期间，建议移除“无法评分”的失焦图像，以防出现假阴性结果（即因图像失焦未检测到生物，却错误显示无生物存在）。高质量图像及各类图像示例见附图1。

图像采用基于网络的Biigle软件（Langenkämper等，2017）进行注释，并与包含日期、时间、纬度、经度、水深、距海底高度及图像区域的元数据一同上传。距海底高度由CTD部署期间每张图像拍摄时的高度计读数确定。每幅图像的宽度与高度通过镜头视野计算器（https://www.deepsea.com/lens-field-calculator/）计算，输入参数包括距海底高度、1/3英寸传感器尺寸及相机原生焦距4.25毫米，据此算出每幅图像的区域并纳入元数据。

图像注释遵循南极海底带注释影像数据库（AS-AID）规范。部分关键分类群未列入任一目录，故标记为“指示物种”或“未鉴定（no id）”以便后期处理时编辑；这些标签亦用于部分模糊分类群，并记录相关说明（表2）。