ASSESSING OIL SPILL IMPACTS IN THE GULF OF MEXICO: A DEEP LEARNING BASED NATURAL LANGUAGE PROCESSING INVESTIGATION FOR RISK MITIGATION

Hydrocarbons, as one of the primary sources of global energy, have been extensively extracted over the past few decades, leading to the drilling of numerous wells. These wells can be classified into three categories: active, abandoned, and orphaned. Without effective planning and control, hydrocarbon leaks can occur in any of these categories, posing a serious risk of environmental pollution. However, hydrocarbon leakage is not limited to the extraction process alone; it also occurs during transportation, storage, and consumption, exacerbating environmental pollution to a considerable extent. The relative contributions of each of these aspects of hydrocarbon production have not previously been quantified nor placed in the context of the contributions of other sectors to accidental hydrocarbon release. This presentation utilizes publicly available oil spill databases, provided by the National Oceanic and Atmospheric Administration (NOAA) from 1970 to 2023, employing a deep learning-based natural language processing approach to thoroughly mine and analyze the potential information within these databases, with a primary focus on investigating the causes of oil spills, the volume of leakage for each category, and their temporal and spatial characteristics. It is found that while incidents related to extraction activities, such as platform accidents, wellhead blowouts, and abandonment leakages, stand as significant contributors to hydrocarbon leakage, it is equally concerning that hydrocarbon leakage from storage and transportation match the scale of extraction-related incidents. Furthermore, these leakage incidents display distinct temporal and spatial distribution patterns. At the same time, it is also found that the quality of the incident report data leads to concerns, with many missing pieces of important information and a disordered forma. To address this, our findings underscore the urgent need to establish a more comprehensive, efficient, and accessible data reporting method, data platform, and database. This approach will pave the way for providing higher-quality and cleaner incident spill data, enabling more effective reporting, prediction, and control of hydrocarbon leakage due to various reasons. Such information is vital for enhancing preparedness and response procedures for civilization, industry, and governments.