On a nonlinear model for arctic gyres