Variational assimilation of satellite observations in a coastal ocean model off Oregon

This is the publisher's final pdf. The published article is copyrighted by American Geophysical Union and can be found at: http://www.agu.org/journals/jc/index.shtml.Satellite along-track sea surface height (SSH) and multisatellite sea surface temperature (SST) maps are assimilated in a coastal ocean circulation model off Oregon. The study period is June–October 2005, featuring intensive separation of the coastal upwelling jets in the eddy-dominated coastal transition zone (CTZ). The data assimilation (DA) system combines the nonlinear Regional Ocean Modeling System (ROMS) and the Advanced Variational Regional Ocean Representer Analyzer (AVRORA) tangent linear and adjoint codes developed by our group. The variational representer DA method is implemented in a series of 6 day time windows, with initial conditions corrected at the beginning of each window. To avoid the problem of matching the model and observed SSH mean levels, the observed SSH slope has been assimilated. Location, timing, and intensity of jets and eddies in the CTZ are constrained, to improve accuracy of nonlinear model analyses and forecasts. In the case assimilating SSH alone, the geometry of the SST front is improved. SSH assimilation results in the cross-shore transport more uniformly distributed along the coast than in the free run model. An outer front is identified in the DA analyses at a distance of 200 km from the coast. A strong subsurface horizontal temperature gradient across this front influences the depth of the thermocline in an area between the front and the continental slope. The DA correction term is comparable in magnitude to dominant terms in the volume-integrated heat equation. The time-averaged DA correction term in the volume-integrated heat balance is closer to 0 in the combined SSH-SST assimilation case, than in the case assimilating SSH alone