Anthropogenic marine debris over beaches: Spectral characterization for remote sensing applications
Anthropogenic marine debris over beaches: Spectral characterization for remote sensing applications
复制标题
DOI：
10.1016/j.rse.2018.08.008
复制
发表时间：
2018-11-01
影响因子：
13.5
通讯作者：
Mattar B, Cristian
中科院分区：
文献类型：
--
作者：
Acuna-Ruz, Tomas;Uribe, Diego;Mattar B, Cristian
人为海洋垃圾是海洋中最重要的污染物之一。数以百万吨计的碎片漂洋过海，造成了严重的环境问题。本研究提出了一种新的方法，旨在通过遥感海滩，结合AMD高光谱实验室表征和数字监督分类在高空间分辨率图像，以提高识别的大塑料。从智利奇洛埃岛海滩采集了几个样本。通过实验室工作对光谱特征样品和物理性质进行了评估。使用HyLogger 3（R）（CSIRO）、PS-300 Apogee和ASD Field Spec高光谱系统来表征每个样品。利用这些测量结果，通过处理、过滤和分类收集的光谱数据，确定用于数字分类的独特光谱带，生成了光谱库。通过使用该光谱库，对World-View 3图像实施了数字分类方法，覆盖了选定为试验地点的三个海滩。采用了不同的分类方法和地理空间分析来确定土地覆盖物的组成，目的是检测聚苯乙烯泡沫塑料和其他人为海洋废弃物。进行了四次实地活动，以验证AMD分类和大规模检索，在300多个地面点进行。AMD高光谱数据库已成功地应用于卫星图像的AMD数字分类。支持向量机方法表现出最好的性能，导致相当于88%的总体准确性和超过50吨的碎片估计试点海滩。这些结果证明了通过整合高光谱实验室测量和遥感图像来量化宏观AMD的可行性，从而可以估计人类活动对自然生态系统的影响，并为进一步开发方法和可持续AMD管理提供有价值的信息。
Anthropogenic Marine Debris (AMD) is one of the most important pollutants in the oceans. Millions of tons of debris across oceans have created a critical environmental problem. This study presents a novel method aimed to improve the identification of macroplastics through remote sensing over beaches, combining AMD hyperspectral laboratory characterization and digital supervised classification in high spatial resolution imagery. Several samples were collected from the Chiloe Island beaches, Chile. Spectral signature samples and physical properties were assessed through laboratory work. HyLogger3 (R) (CSIRO), PS-300 Apogee and ASD Field Spec hyperspectral systems were used to characterize each sample. Using those measurements, a spectral library was generated by processing, filtering and sorting the spectral data gathered, determining distinctive spectral bands for digital classification. By using this spectral library, a digital classification method was implemented over World-View 3 imagery, covering the three beaches selected as test sites. Distinct classification methods and geospatial analyses were applied to determine land cover composition, aimed for the detection of Styrofoam and the rest of anthropogenic marine debris. Four field campaigns were carried out to validate the AMD classification and mass retrievals, performed on > 300 ground based points. The AMD hyperspectral library was successfully applied for an AMD digital classification in satellite imagery. Support Vector Machine method showed the best performance, resulting in an overall accuracy equivalent to 88% and over 50 tons of debris estimated on the pilot beaches. These results prove the feasibility of quantifying macro-AMD through the integration of hyperspectral laboratory measurements and remote sensing imagery, allowing to estimate anthropogenic influence on natural ecosystems and providing valuable information for further development of the methodology and sustainable AMD management.