An ensemble trajectory prediction model for maritime search and rescue and oil spill based on sub-grid velocity model

Abstract

An important part of establishing a prediction model for maritime search and rescue and oil spill drift is to quantify the uncertainty in particle motion simulation. A regional Sub-grid velocity model based on drifting buoy data is proposed to simulate the unsolved velocity which is composed of turbulence and advection simulation errors (TASE) in the ocean model. Contrary to most of the traditional drift trajectory prediction models, the presented method divided the experimental area into 69 custom grids. The standard deviation of TASE velocity, TASE time scale, and TASE diffusion coefficients of each grid were obtained by using the fitting velocity difference of buoy velocity, current velocity and wind velocity, and autocorrelation analysis of time series. Two Sub-grid velocity models, the random flight model and the random walk model, were built respectively and were combined with the Lagrangian particle tracking model to simulate the drifting buoy trajectory. In the process of Lagrange particle tracking, fourth-order Runge-Kutta was used to interpolate the velocity and kernel density estimate method was used to calculate the predicted range of 95% confidence interval for evaluation and validation. The experimental results of several different settings indicated that the prediction performance of the random flight model was better than that of the random walk model when the TASE diffusion coefficient was determined by this method, and the prediction performance of models with different diffusion coefficients is better than that of models with the same diffusion coefficient. This work also proved the value of using the trajectory data of drifting buoys to build a Sub-grid velocity model for trajectory prediction.