Improving oil slick trajectory simulations with Bayesian optimization

Accurate simulations of oil spill trajectories are essential for supporting practitioners' response and mitigating environmental and socioeconomic impacts. Numerical models, such as MEDSLIK-II, simulate advection, dispersion, and transformation processes of oil particles, but their accuracy depends strongly on the correct tuning of physical parameters, often relying on manual calibration and expert knowledge. This approach is suboptimal, especially in dynamic and uncertain environmental conditions. To overcome these limitations, we couple the MEDSLIK-II oil spill model with a Bayesian optimization framework to iteratively estimate the optimal values of key parameters, such as the horizontal diffusivity, wind angle and wind drag, in order to obtain simulation closer to satellite observations of the slick. We adopt a stochastic parameterization strategy, which probabilistically explores the parameter space to enhance simulation skill. To this end, the Fraction Skill Score (FSS) is maximized to evaluate spatial-temporal overlap between simulated and observed oil distributions. The framework is validated for the Baniyas oil incident that occurred in Syria between August 23 and September 4, 2021, which released over 12,000 m of oil. The approach improves FSS from 7. 97 % to 20. 66 %, on average, compared to control simulations initialized with default parameters. Results demonstrate consistent improvements across time steps, highlighting the method's robustness and suitability for operational oil spill modeling under uncertainty.

Accurate simulations of oil spill trajectories are essential for supporting practitioners' response and mitigating environmental and socioeconomic impacts. Numerical models, such as MEDSLIK-II, simulate advection, dispersion, and transformation processes of oil particles, but their accuracy depends strongly on the correct tuning of physical parameters, often relying on manual calibration and expert knowledge. This approach is suboptimal, especially in dynamic and uncertain environmental conditions. To overcome these limitations, we couple the MEDSLIK-II oil spill model with a Bayesian optimization framework to iteratively estimate the optimal values of key parameters, such as the horizontal diffusivity, wind angle and wind drag, in order to obtain simulation closer to satellite observations of the slick. We adopt a stochastic parameterization strategy, which probabilistically explores the parameter space to enhance simulation skill. To this end, the Fraction Skill Score (FSS) is maximized to evaluate spatial-temporal overlap between simulated and observed oil distributions. The framework is validated for the Baniyas oil incident that occurred in Syria between August 23 and September 4, 2021, which released over 12,000 m3 of oil. The approach improves FSS from 7. 97 % to 20. 66 %, on average, compared to control simulations initialized with default parameters. Results demonstrate consistent improvements across time steps, highlighting the method's robustness and suitability for operational oil spill modeling under uncertainty.