Lagrangian predictability in the DWH region from HF radar observations and model output

Abstract

Predictability of surface drifter trajectories in the Deep Water Horizon oil spill region is used as a criterion for optimizing the parameters of the 2d variational (2dVar) interpolation of high-frequency radar (HFR) data, and assessing the accuracy of the surface currents' simulations by regional models. It is shown that penalizing the magnitude and enforcing smoothness of the divergence field significantly increases the Lagrangian predictability of the 2dVar output at the forecast times of 3–9 days while preserving it at the shorter forecast times. Applying preliminary gap-filling technique based on the analysis of spatial correlations of the radial velocities adds an extra 1–2% to the 2dVar forecast skill. Comparison of the forecast skills provided by the 2dVar interpolation of the HFR data and the assimilative solutions of the Navy Coastal Ocean Model demonstrates 25–30% better skill of the 2dVar product, indicating potential benefits of assimilating HFR data into regional models.