Evaluation of an oil spill trajectory model using satellite-tracked, oil-spill-simulating drifters

We deployed ninety-seven oil-spill-simulating drifters over the continental shelf of the northeastern Gulf of Mexico during five hydrographic surveys conducted from 1997 through 1999. Earlier, side-by-side comparisons with spilled crude petroleum on the ocean surface had demonstrated that these drifters moved on the ocean surface like consolidated oil slicks under light to moderate winds. (Under high winds, a surface oil spill tends to be entrained into the mixed layer and Ekman transported, unlike the drifters, which remain on the sea surface and move mostly downwind.) The drifters were then deployed in the Gulf of Mexico as nonpolluting oil-spill proxies to compare their movements against results from an oil-spill trajectory model. The drifter trajectories were compared statistically to trajectories generated by the Oil-Spill Risk Analysis (OSRA) model. The model uses a variation of the 3.5-percent rule to compute the drift due to local wind forcing and superposes the prevailing ocean current on this wind-induced drift to obtain the total velocity of an oil spill on the ocean surface. The input fields are the European Center for Medium Range Weather Forecasts (ECMWF) winds and a data-assimilating hindcast of the ocean currents over the time the drifters were deployed. Scatter plots and linear regressions of the speeds and directions of simulated vs. modeled oil-spill drift show the extent to which they are different. Underlying these differences are the expected differences between the ocean current input field and the trajectories of satellite-tracked, "water-following" drifters deployed simultaneously with the oil-spill-simulating drifters. An earlier evaluation of the ECMWF winds showed better, but of course not perfect, agreement with meteorological buoys in the Gulf. The integrated effect of the errors in the input fields results in average discrepancies between the terminal ends of the simulated and modeled spill trajectories of 78, 229, 416, and 483 km after 3,10, 20, and 30 days of drift, respectively. These results are the consequence of integrating wind and ocean current fields, which are not perfect, and comparing the resultant trajectories against those of the oil-spill-simulating drifters, which themselves contain location errors and which are not considered to be perfect simulators of real oil spills. However, the results are useful to practical oils spill risk analysis through ongoing improvement of the model.