51. Refined Analysis of RADARSAT-2 Measurements to Discriminate Two Petrogenic Oil-Slick Categories: Seeps versus Spills
- Author
-
Luiz Landau, Gustavo de Araújo Carvalho, Eduardo Tavares Paes, Peter J. Minnett, and Fernando Pellon de Miranda
- Subjects
Synthetic aperture radar ,010504 meteorology & atmospheric sciences ,petrogenic oil-slick category ,0211 other engineering and technologies ,Data transformation (statistics) ,Ocean Engineering ,Feature selection ,man-made oil spills ,02 engineering and technology ,Information repository ,01 natural sciences ,lcsh:Oceanography ,naturally-occurring oil seeps ,remote sensing ,lcsh:VM1-989 ,oil-slick discrimination algorithm ,lcsh:GC1-1581 ,021101 geological & geomatics engineering ,0105 earth and related environmental sciences ,Water Science and Technology ,Civil and Structural Engineering ,Gulf of Mexico ,RADARSAT ,lcsh:Naval architecture. Shipbuilding. Marine engineering ,exploratory data analysis ,Racing slick ,Exploratory data analysis ,Oil spill ,Environmental science ,Surface expression ,Campeche Bay ,Cartography ,synthetic aperture radar - Abstract
Our research focuses on refining the ability to discriminate two petrogenic oil-slick categories: the sea surface expression of naturally-occurring oil seeps and man-made oil spills. For that, a long-term RADARSAT-2 dataset (244 scenes imaged between 2008 and 2012) is analyzed to investigate oil slicks (4562) observed in the Gulf of Mexico (Campeche Bay, Mexico). As the scientific literature on the use of satellite-derived measurements to discriminate the oil-slick category is sparse, our research addresses this gap by extending our previous investigations aimed at discriminating seeps from spills. To reveal hidden traits of the available satellite information and to evaluate an existing Oil-Slick Discrimination Algorithm, distinct processing segments methodically inspect the data at several levels: input data repository, data transformation, attribute selection, and multivariate data analysis. Different attribute selection strategies similarly excel at the seep-spill differentiation. The combination of different Oil-Slick Information Descriptors presents comparable discrimination accuracies. Among 8 non-linear transformations, the Logarithm and Cube Root normalizations disclose the most effective discrimination power of almost 70%. Our refined analysis corroborates and consolidates our earlier findings, providing a firmer basis and useful accuracies of the seep-spill discrimination practice using information acquired with space-borne surveillance systems based on Synthetic Aperture Radars.
- Published
- 2018
- Full Text
- View/download PDF