DATA ANALYTICS AND OPTIMIZATION MODELLING BASED DECISION SUPPORT TOOL FOR CANADIAN ARCTIC OIL SPILLS RESPONSE

Abstract

Oil spills adversely affect marine ecosystems. Hence, strategic planning for the development and operation of marine oil spill response facilities, along with resource allocation, holds significant importance for decision-makers. However, an integrated Decision Support Tool (DST) which can estimate marine oil spills size, rank response technologies, recommend facility location and resource allocation, and capture uncertainty is yet to be developed for the harsh conditions of the Canadian Arctic. To close this gap, this thesis develops a data analytics and optimization-based DST to support strategic decision-making regarding oil spill pollution preparedness and response risk management in the Canadian Arctic. The DST framework consists of four models. First, an Artificial Intelligence inspired Deep Neural Network model is developed to predict spill size in ship accidents. The results underscore the model's capacity to provide accurate and computationally efficient estimates. Second, a Bayesian Inference model ranks response technologies in Arctic marine oil spills, for given oil spill information and Arctic environmental factors. The findings show that this model can accurately select the best available response technologies among mechanical containment and recovery, chemical dispersant use, and in-situ burning, which is useful in risk management. Third, a facility location and resource allocation model maximizes spill coverage and minimizes cost in the Canadian Arctic for deterministic cases. The problem is mathematically formulated using Mixed Integer Programming. Spill size and ranking of technologies are key input parameters of this optimization model. The findings elucidate the optimal configuration of spill response facilities and the strategic allocation of associated resources, thereby supporting decision-making. However, input parameters such as spill size, location, response time are uncertain in real-world situations. Therefore, fourth, a stochastic location-allocation model, which accounts for parametric uncertainty, is developed by extending the prior deterministic model. The proposed DST significantly advances the existing literature on model-based strategic oil spill preparedness and response planning, by providing data analytics and optimization models. This DST presents an innovative approach to strategic planning, thereby benefiting various stakeholders involved in Arctic oil spill preparedness and response.