In March 2015, an oil spill incident occurred in Northern Ontario due to a train derailment. The derailment involved 35 crude oil cars, all of which ruptured and leaked some amount of oil to the ground surface or directly to the Makami River. At the beginning of the project, real-time numerical oil spill modelling was performed to predict the extent of the oil plume in the Makami River and Minisinakwa Lake and provide guidelines for environmental monitoring. An open source hydrodynamic and water quality model, TELEMAC-2D, was used to build the oil spill model. TELEMAC-2D solves the depth-averaged shallow water equations, and calculates flow velocity and depth at each node of an unstructured mesh using the finite element method. TELEMAC’s oil spill module simulates oil slicks using a large number of hydrocarbon particles, consisting of both soluble and unsoluble components. The module accounts for advection, diffusion, evaporation, wind effects, and dissolution processes.
The key input in numerical spill modeling is the spill scenario (instantaneous and/or continuous). At the time of the incident little information was available to define the spill scenario, which made the real-time spill modelling challenging. Additionally, the majority of the free-phase oil was contained at the spill location based on the ice conditions in the river and the strategic deployment of oil containment booms and ice slots. Therefore, a trial-and-error calibration process was used to match available water quality measurements by adjusting the spill scenario. It was assumed that the free phase oil was contained at the spill site during the emergency response, and therefore only the downstream propagation of the soluble component of the oil was modelled. Some of the initially spilled volume was trapped under the river ice and this was also considered in the spill scenario. The calibrated oil spill model was used to demonstrate the propagation of oil in the river and the lake and provide outputs for environmental studies. The main limitations of this study was limited model input (bathymetry and model boundary conditions) and calibration data (flow velocity, water levels).