Variational Assimilation of Satellite Observations in the Coastal Ocean

Alexander L. Kurapov(1) (kurapov@coas.oregonstate.edu),
D. G. Foley(2), P. T. Strub(1), G. D. Egbert(1), J. S. Allen(1).

(1) College of Oceanic and Atmospheric Sciences, Oregon State
University, Corvallis, OR, USA
(2) NOAA-NESDIS-JIMAR, NOAA-SWFSC, Pacific Grove, CA, USA

Nonlinear ROMS is combined with the AVRORA tangent linear and adjoint codes developed at OSU, to facilitate implementation of the variational representer-based method in the coastal ocean model off Oregon (U.S. West Coast). Satellite observations of alongtrack SSH and SST maps are assimilated using the ROMS-AVRORA system at 6-km horizontal resolution, to learn about utility of satellite information in the context of coastal ocean prediction and forecasting systems. The study period is June-October 2005 when observations (AVISO absolute dynamic topography) from 3 satellites, Jason, Topex, and Envisat are available. Assimilation proceeds in a series of 6-day time windows. In each window, initial conditions are corrected using iterations of the adjoint and tangent linear AVRORA and a multivariate initial condition error covariance. The nonlinear ROMS is then run with the corrected initial conditions for the 6-day analysis period plus the 6-day forecast period. The forecast is a background solution for assimilation in the next window. Since the mean SSH in the alongtrack altimetry and the Boussinesq ROMS are not directly comparable, we assimilate the SSH slope along each track (or, scaled appropriately, the surface geostrophic velocity component in the direction normal to the tracks). SSH assimilation improves the model-data RMS difference of SSH forecasts and also corrects the geometry of the SST front, compared to the unassimilated SST. Combined SSH and SST assimilation further improves accuracy of near-surface transports and SST. The horizontal offshore transport of momentum and heat in the assimilative solution is more uniform along the coast than in the free-run prior model, in which the transport was dominated by separation at Cape Blanco (43N). Also, the assimilative solution helps reveal the existence of a front at 127N, about 200 km offshore, in addition to the inshore front that is usually clearly seen in SST images. This outer front is not reproduced by the limited-resolution prior model, and is characterized by a strong along-front current and strong subsurface horizontal gradient of temperature, but weak SST contrast.