Biogenic oil film is a natural surface slick that is mainly derived by sea flora and fauna. The often observation of the film near aquaculture facilities has raised awareness on the possibility of the linkage between the development of the film and the anthropogenic activities taking place on site (i.e. artificial feeding and liquid waste). This study aims to investigate the possibility of the detection of biogenic oil film on optical satellite images and discriminate it from other oceanographic phenomena. For the purposes of the study we have used a Sentinel-2 (S2) dataset consisted of 73 images for the year 2019 to detect the film on three aquaculture areas. An automatic procedure was developed on a Python based algorithm which included the following stages: (a) downloading images, (b) preprocessing the input data, (c) identifying dark formations in the adjacent fish farming area, (d) extracting attribute tables with the statistical characteristics of the formations (shape, area, etc.), (e) classification of formations as biogenic film or other (lookalike) and (f) extraction of biogenic film vectors. The developed algorithm was able to detect biogenic oil film successfully however some misleading results regarding the decision of true or false positive (biogenic oil film or lookalike) was evidenced. The efficiency of the algorithm was tested against manual classification with overall accuracy 82,6%. As further step the results of this study should be validated with in-situ measurements and further work is required to verify the results obtained by testing the methods in other sites.
|