A repeated sampling method for oil spill impact uncertainty and interpolation

Abstract

The Deepwater Horizon disaster and other oil spill events have motivated the continued development of spill impact estimation methods and models. Spills are simulated, impacts are estimated and prescriptions are made for improving response and associated mitigation efforts. However, there is significant uncertainty in oil spill models due to the stochastic nature of the ocean and the representation of a plume as points in space. Furthermore, large scale analyses, while useful, may fail to recognize and characterize the micro- or meso-scale impacts of a spill. This paper presents an innovative application of a repeated sampling procedure to mitigate elements of uncertainty in oil spill models by capturing and characterizing where the oiling is likely to beach and providing probability estimates of the associated predictions. Specifically, we use a kriging interpolation method to model the oiled coastline as a continuous surface to better match actual oil landfall observed in reality and then use it to provide a more robust estimation of the smaller scale impacts. Through two measures of validation this work finds that the repeated sampling procedure does provide a more robust estimate of oil impact when compared to estimations from a single simulation of a spill