Observability of the Irminger Sea circulation using variational data assimilation

This paper explores the prospects for constraining circulation and stratification estimates in the Irminger Sea and Denmark Strait using variational data assimilation. A regional circulation model with 9 km horizontal grid spacing, seven vertical levels and open boundaries is used. The circulation it simulates is realistic in several important respects and includes a chaotic mesoscale eddy field. Pulsation of the outflow of dense water through the Denmark Strait is also reasonably realistic, although entrainment and mixing of this water downstream is crudely represented. Identical twin assimilation experiments are performed to address the observability of the system with the existing in situ and satellite observing network in the absence of systematic errors. That is, the constraints exerted on the circulation and stratification by the data assimilation system are quantified. Synthetic data come from a reasonably realistic observing network that includes in situ hydrographic and current measurements, and remotely sensed surface temperature and sea level. Over a 15-day assimilation period in summer, the system fits all the observations to the relevant noise level, largely by correcting the initial conditions. Only modest changes in the time-dependent surface forcing and negligible changes to the open boundary conditions are needed. The system can accurately estimate sea surface height and sea surface temperature, correcting errors in the locations of surface fronts and eddies. The mid-depth salinity is almost completely unconstrained, however, whereas the velocity fields are somewhat constrained by the assimilation, especially by the remote-sensing data. Assimilation of satellite data enables us to accurately track the phase and amplitude of the overflow variability at the exit of Denmark Strait in the identical twin experiments. At other model outflow transects; the assimilated solution is contaminated with spurious inertial oscillations. This is true even at a section offshore of Angmagssalik where a current-meter mooring array exists (in our experiments and in the real ocean). Nevertheless, improvement of the model dense outflow near the moorings is significant and propagates several hundred kilometres upstream and downstream.