Estimating mussel mound distribution and geometric properties in coastal salt marshes by using UAV-Lidar point clouds
Estimating mussel mound distribution and geometric properties in coastal salt marshes by using UAV-Lidar point clouds
复制标题
利用无人机激光雷达点云估计沿海盐沼中的贻贝丘分布和几何特性
DOI：
10.1016/j.scitotenv.2023.163707
复制
发表时间：
2023
影响因子：
9.8
通讯作者：
Wilkinson, Benjamin
中科院分区：
文献类型：
--
作者：
Pinton, Daniele;Canestrelli, Alberto;Williams, Sydney;Angelini, Christine;Wilkinson, Benjamin
大西洋肋贻贝（Geukensia demissa）常见于美国东南部的盐沼，在那里它们形成密集的聚集体（土丘），在靠近潮沟头的沼泽平台上以最高的密度和大小出现。在这些沼泽中，贻贝通过有机和无机物质的生物沉积帮助建立沼泽海拔，刺激占主导地位的基础物种灯心草（互花米草）的生长，并创造无脊椎动物生物多样性，营养循环和干旱恢复的热点。由于其强大的作用，有越来越多的兴趣在评估贻贝土丘分布的自然变化，并利用这些信息来指导沼泽保护和恢复战略。然而，收集这类信息具有挑战性，因为土丘的尺寸很小（1米），而且上面有植被覆盖，因此难以量化沼泽地上土丘的分布情况。因此，本研究提出了一个新的程序来计算分布，高度，半径，体积和距离的土丘在沼泽环境中使用遥感。一个高分辨率的无人机激光雷达点云已经收集了高度植被盐沼在格鲁吉亚，美国，使用定制的激光扫描仪系统。一个原始的检测算法，基于随机森林分类器，已被实施，以确定从点云的土丘。该算法已被训练和测试的调查土丘，并提供其位置和几何属性。结果表明，该分类器可以区分贻贝土丘从非贻贝丘位置的准确率为95%。分类器识别出108000个土丘，占研究区域的10%，体积（贝壳+粪便/假粪便）为680 m3。该方法是非常有用的努力，监测贻贝土丘随着时间的推移和规模扩大，以评估土丘跨网站，提供宝贵的数据，为未来的研究有关的沼泽地貌演变海平面上升和选址沼泽保护和增强项目。
The Atlantic ribbed mussel (Geukensia demissa) is common in southeastern US salt marshes, where they form dense aggregations (mounds), that occur in the highest densities and sizes on the marsh platform close to the tidal creeks' heads. Within these marshes, mussels help build marsh elevation via their biodeposition of organic and inorganic material, stimulate the growth of the dominant foundation species cordgrass (Spartinaalterniflora), and create hotspots of invertebrate biodiversity, nutrient cycling, and drought resilience. Given their powerful role, there is rising interest in assessing natural variation in the distribution of mussel mounds and using such information to guide marsh conservation and restoration strategies. However, gathering such information is challenging, because the small dimension (∼1 m) of the mounds and the presence of overlying vegetation make it difficult to quantify mound distribution on the marsh. Therefore, this study presents a new procedure to compute the distribution, height, radius, volume, and distance of mounds in marsh environments using remote sensing. A high-resolution UAV-Lidar point cloud has been collected over a highly vegetated salt marsh in Georgia, USA, using a custom-built laser scanner system. An original detection algorithm, based on a Random Forest classifier, has been implemented to identify the mounds from the point cloud. The algorithm has been trained and tested on surveyed mounds and provides their location and geometric properties. Results indicate that the classifier can distinguish mussel mounds from non-mussel mound locations with an accuracy of 95 %. The classifier identified ∼8000 mounds, which occupy 10 % of the study domain, and a volume (shells+feces/pseudofeces) of 680 m3. The method is highly useful in efforts to monitor mussel mounds over time and scale up to assess mounds across sites, providing invaluable data for future studies related to the geomorphic evolution of marshes to sea level rise and siting marsh conservation and enhancement projects.