Detection of Marine Plastic Debris in the North Pacific Ocean using Optical Satellite Imagery

Abstract

Plastic pollution is ubiquitous across marine environments, yet detection of anthropogenic debris in the global oceans is in its infancy. Here, we exploit high-resolution multispectral satellite imagery over the North Pacific Ocean and information from GPS-tracked floating plastic conglomerates to explore the potential for detecting marine plastic debris via spaceborne remote sensing platforms. Through an innovative method of estimating material abundance in mixed pixels, combined with an inverse spectral unmixing calculation, a spectral signature of aggregated plastic litter was derived from an 8-band WorldView-2 image. By leveraging the spectral characteristics of marine plastic debris in a real environment, plastic detectability was demonstrated and evaluated utilising a Spectral Angle Mapper (SAM) classification, Mixture Tuned Matched Filtering (MTMF), the Reed-Xiaoli Detector (RXD) algorithm, and spectral indices in a three-variable feature space. Results indicate that floating aggregations are detectable on sub-pixel scales, but as reliable ground truth information was restricted to a single confirmed target, detections were only validated by means of their respective spectral responses. Effects of atmospheric correction algorithms were evaluated using ACOLITE, ACOMP, and FLAASH, in which derived unbiased percentage differences ranged from 1% to 81% following a pairwise comparison. Building first steps towards an integrated marine monitoring system, the strengths and limitations of current remote sensing technology are identified and adopted to make suggestions for future improvements.