Luke Phillipson, Imperial College London
Improving Short-Term Ocean Current Predictions with 4D-Var : Is There an Optimal Observational System?
Abstract: Accurately forecasting ocean currents for short-term (~4 days) predictions is of great importance to a variety of applications from tracking ocean debris and marine pollution to planning search and rescue deployments. The ocean off the coast of Angola/Congo is a specific example of a complicated system of ocean currents that is often poorly represented in ocean models. Improvements in these models can be achieved through the use of data assimilation; a technique that optimally combines the observational and model information. Using the regional ocean modeling system (ROMS) with 4d-varational data assimilation (4D-Var), we seek to find an optimal observation system that will improve short-term ocean current prediction when assimilated. The impact of assimilating a wide range of different observations on ocean current predictions is assessed through observational system experiments (OSE) and adjoint-based forecast sensitivity to observation (FSO). The observations assimilated include in-situ temperature and salinity (TS), satellite Sea Surface Temperature (SST), along track and gridded Sea Surface Height (SSHAT/SSHG) and ocean velocities derived from Lagrangian drifter data with a 24hr or 96hr filter applied (DRIFT24 / DRIFT96). To validate our results we use OSCAR ocean currents and simulated ROMS floats compared against drifter trajectories. Assimilating TS_SST_SSHG_DRIFT24 improved OSCAR ocean current comparisons by roughly 20 to 30 \% (RMSE reduction) as compared to a free model run (no assimilation). Simulated ROMS floats and observed drifter separations distances after 24 hrs reduced on average by 7.4 km for TS_SST_SSHG_DRIFT24 and 3.9 km for TS_SST_SSHG.