I tried to simulate the trajectory of an drifting object in the oceans by using the data of the OSCAR project http://www.oscar.noaa.gov/. The dataset actually used consist of grid sampled mean 2d current vectors averaged on a monthly interval indicating the speed and direction of water near the sea surface.
However, moving an object by the gradients shows that the gradient field is riddled with attractors and repellors. There may be several reasons for this, like the water flowing vertically or the mean operation introducing artifacts.
Thus moving an object along trajectories made up by a static snapshot of the dataset isn't very useful, as it usually get stuck in one of the hundreds of sink attractors. This contradicts the usual knowledge of drifting particles to accumulate in very large vortices and eventually reach almost every coastal point on earth.
So how should the mean current be interpreted in respect to drifting object movement? Is there a simple solution to get a coarse drifting simulation that qualitative resembles the expected behavior ?