Description: Oil spills rapidly spread on sea surfaces covering wide areas, assuming different appearances and thicknesses. Due to currents and winds, continuous slicks break into smaller fragments which can reach coastal ecosystems and lead to adverse environmental and socio-economic impacts. The faster the actions to detect, stop, and contain the released oil from spreading, the higher the Oil Spill Response (OSR) success rate. Clean-up effectiveness is higher over thicker oil layers, referred to as "actionable oil". Detecting these regions is crucial to guide response vessels allowing a better deployment of barriers and skimmers, enhancing mechanical oil recovery efficiency, in situ burning tasks, as well as aerial-based dispersant application.
Under this scenario, "Searching for Oil Spills on Sea Surfaces" (SOSeas) will employ the latest generation of deep learning methods for semantic segmentation to develop an artificial intelligence (AI) based system to identify relative oil thicknesses by using Synthetic Aperture Radars (SAR). Oil slick characterization is a new, promising and highly innovative research area with great perspectives owing to the availability of free and open Earth Observation products, and to the effectiveness of machine learning algorithms combined with high-performance computing infrastructure based on graphical processing units (GPUs).
A team of scientists and key stakeholders from diverse research institutes, governmental agencies, and private companies are composing the project's advisory panel. This multidisciplinary and intersectoral panel merges complementary skills and expertise within academic and operational scenarios, consolidating an important research network. It is expected that a deep learning architecture well-trained on large-scale datasets to recognize oil thickness variations has the potential to indicate the location of recoverable oil, thereby improving situational awareness, decision-making, and clean-up effectiveness.
Project website: not yet available
Partners: Collecte Localisation Satellite (CLS); Jet Propulsion Laboratory (JPL - NASA); Brazilian Institute for Space Research (INPE); Development and Innovation Center from Petrobras (CENPES - Petrobras); Brazilian Institute of Environment and Renewable Natural Resources (IBAMA).
Sponsor/funding: Marie Sklodowska-Curie Action Postdoctoral Fellowships (MSCA-PF)
ITC staff involved:
Post-doc Researcher: Dr. Patrícia Carneiro Genovez
ITC Supervisors: Prof. Dr. Claudio Persello, Dr Raian Vargas Maretto, and Dr. Ling Chang
CLS staff involved:
CLS Supervisors: Dr. Vincent Kerbaol and Dr. Guilaume Hajduch