Can an eddy-resolving general circulation model adequately represent the Labrador Sea deep convection cycle?

Geoffrey Gebbie
Advisor: Carl Wunsch
Collaborators: Julio Sheinbaum (CICESE), Detlef Stammer (SIO)

April 10, 2003

Abstract:

Due to the importance of small scales of motion in the Labrador Sea, an observational study of the dynamics which includes all scales must combine many different forms of complementary data. Synthesis of the data collected in the Labrador Sea may be achieved through a state estimation procedure which combines information from both data and numerical model. The first step in any state estimation procedure is to assess the ability of the model to produce an first-order picture that is consistent with observations. A model-data comparison is shown here between a Labrador Sea MIT general circulation model run with 14 kilometer resolution and 1) TOPEX/POSEIDON altimetry, 2) WOCE hydrography, 3) current meters, and 4) PALACE floats. The deepest convection in the model is to a depth of approximately 1500 meters in mid-March in the southwest side of the interior Labrador Sea, consistent with observation. The slow deepening and rapid restratification of the model's mixed layer show that the nature of convection is also realistically represented here. The mean circulation and density structure of the model agrees with observation except in a thin surface layer in part due to a lack of mixed layer physics. The model has more variability than a previous model study (Stammer et al., 1996) and probably is the best effort to date in producing eddy variability. SSH variability and eddy kinetic energy have the correct spatial structure but are too low in energy at most timescales. However, the model does show signs of producing enough variability at periods between 100-300 days. The vertical structure of kinetic energy is dominated by the barotropic and 1st baroclinic modes, which is consistent with high latitude observations, and the shape of vertical kinetic energy profiles is reasonable with a small sample of current meter observations. Methods for improvement of the model already exist in many areas.