An evaluation of semi-automated methods for collecting ecosystem-level data in temperate marine systems.
An evaluation of semi-automated methods for collecting ecosystem-level data in temperate marine systems.
复制标题
DOI：
10.1002/ece3.3041
复制
发表时间：
2017-07
影响因子：
2.6
通讯作者：
Johnston EL
中科院分区：
文献类型：
--
作者：
Griffin KJ;Hedge LH;González-Rivero M;Hoegh-Guldberg OI;Johnston EL
从历史上看，海洋生态学家一直缺乏有效的工具，能够捕捉详细的物种分布数据在大面积。高分辨率成像和相关的机器学习图像评分软件等新兴技术正在为绘制海洋大面积物种地图提供新的工具。在这里，我们将联合收割机与新型潜水推进器（DPV）成像系统和免费使用的机器学习软件相结合，半自动生成超过5,000平方米温带珊瑚礁的栖息地形成藻类的密集和广泛的丰度记录。我们采用可复制的空间技术来测试传统潜水员采样的有效性，并更好地了解一个关键藻类物种的分布和空间排列。我们发现，传统调查的有效性取决于空间结构的水平，通常需要10-20个样带（50 × 1 m）才能获得可靠的结果。这代表了2-20倍的复制比在以前的研究中收集。此外，我们证明了高分辨率分布模型在多个空间尺度上理解冠层藻类覆盖模式的有用性，并讨论了对其他海洋栖息地的应用。我们的分析表明，半自动数据收集和处理方法比描述海景尺度栖息地结构的传统方法提供了更准确的结果，因此代表了理解和管理海洋海景的极大改进的技术。
Historically, marine ecologists have lacked efficient tools that are capable of capturing detailed species distribution data over large areas. Emerging technologies such as high‐resolution imaging and associated machine‐learning image‐scoring software are providing new tools to map species over large areas in the ocean. Here, we combine a novel diver propulsion vehicle (DPV) imaging system with free‐to‐use machine‐learning software to semi‐automatically generate dense and widespread abundance records of a habitat‐forming algae over ~5,000 m2 of temperate reef. We employ replicable spatial techniques to test the effectiveness of traditional diver‐based sampling, and better understand the distribution and spatial arrangement of one key algal species. We found that the effectiveness of a traditional survey depended on the level of spatial structuring, and generally 10–20 transects (50 × 1 m) were required to obtain reliable results. This represents 2–20 times greater replication than have been collected in previous studies. Furthermore, we demonstrate the usefulness of fine‐resolution distribution modeling for understanding patterns in canopy algae cover at multiple spatial scales, and discuss applications to other marine habitats. Our analyses demonstrate that semi‐automated methods of data gathering and processing provide more accurate results than traditional methods for describing habitat structure at seascape scales, and therefore represent vastly improved techniques for understanding and managing marine seascapes.