Evaluation of sea surface temperature and heat flux in an assimilative Gulf of Mexico ocean model

Variational assimilation of sea surface temperature (SST) guides analyses and forecasts toward agreement with measured conditions. 3DVar assimilation assumes that forecast mismatches are due to errors in the initial state, while weak constraint 4DVar balances assumed error levels in the initial state, lateral boundary conditions, model, observations, and forcing terms to best adjust the model trajectory. We examine the skill of forecasts from the Naval Coastal Ocean Model (NCOM) implemented in the Gulf of Mexico and assimilating various satellite observations. The impacts of various satellite data streams and alternative assimilation methodologies are evaluated by comparing model analyses and forecasts to unassimilated ship and buoy observations. Seasonal and diurnal trends in the forecast errors suggest biases in the heat flux which may be better addressed in a 4DVar approach.