Computing Lagrangian statistics from tracer observations and a model output

A stable method of low computational cost is suggested for estimating Lagrangian trajectories by combining observations of a continuously distributed conservative tracer and a circulation model output (background velocity field). A theoretical error analysis for a homogeneous flow is provided while for a more sophisticated synthetic flow the error is evaluated numerically. The reduction in error compared to the background varies in the range 1–34% depending on the flow structure. Finally, the method is applied for estimating the absolute and relative dispersion in a gyre perturbed by periodic or stochastic fluctuations.