A Dynamic Machine Learning Model for Accelerated Oil Spill Remediation

134 million gallons of oil were spilled into the Gulf of Mexico after the explosion of an offshore oil rig in 2010. Known as the Deepwater Horizon spill, this event crippled marine environments spanning thousands of miles and killed countless sea creatures already deemed at risk of extinction. Over 10 years and billions of dollars later, efforts to clean up this spill continue. Rapid mitigation is necessary to prevent future incidents from spiraling out of control. After an oil spill, various organizations must decide how to remediate it. To do so, there are close to a dozen methods employed today. Each approach has its pros and cons and must be carefully selected based on spill conditions. Some techniques (such as in-situ burning of the oil slick off the water) are highly effective but have environmentally degrading effects. Choosing a suboptimal remediation tactic can lead to billions of wasted dollars, and more importantly, leftover oil that continues to harm the environment. During this study, an artificial intelligence (AI) based system using a convolutional neural network (CNN) has been developed to prescribe the most effective oil spill countermeasure. Findings were used to develop a mobile application to further expedite oil spill cleanup and recovery in real time. After being tested at various configurations, the machine learning model achieved a maximum average accuracy of 93.1 % after 16.19 seconds of training time with 10 epochs and a batch-size of 16. This work significantly enhances our ability to quickly remediate oil spills, protecting the environment from this disastrous calamity.