Four-dimensional variational assimilation of satellite temperature and sea level data in the coastal ocean and adjacent deep sea

John Wilkin, Javier Zavala-Garay, Julia Levin and W. Gordon Zhang
Institute of Marine and Coastal Sciences, Rutgers University

Incremental, Strong constraint, 4-Dimensional Variational (IS4DVAR) data assimilation with ROMS is used to initialize operational, coastal, mesoscale resolution, forecast models of continental shelf and associated boundary current regimes. In particular, we show assimilation results from the East Australia Current and the Mid-Atlantic Bight Slope Sea. In both areas the assimilation of adjacent, deep ocean data influence the coastal dynamics through remote forcing. We assimilate observations of daily satellite sea surface temperature, multi-satellite altimeter sea level anomalies, and subsurface temperature and salinity data from Volunteer Observing Ship transects and/or autonomous underwater vehicles. At the open boundary, both models use data from operational basin-scale circulation models. The atmospheric forcing is from operational weather forecast models. Control variables of the data assimilation are the initial conditions of a sequence of 3- to 7-day assimilation windows. The nonlinear model trajectory through each interval is deemed as the best-estimate analysis for initializing the subsequent forecast. We evaluate model skill from a large set of multi-day forecasts, from different initial mesoscale states. Forecast skill is enhanced and uncertainty reduced when empirical statistical subsurface pseudo-observations and/or so-called balance constraints are used to augment surface satellite data.