Variational Assimilation of HF Radar Surface Currents in the Coastal Ocean Circulation Model off Oregon

Variational Assimilation of HF Radar Surface Currents in the Coastal Ocean Circulation Model off Oregon
P. Yu1; A. L. Kurapov1; G. D. Egbert1; J. S. Allen1; M. Kosro1
1. COAS, Oregon State University, Corvallis, OR, United States.

Sea surface velocity fields observed by a set of high-frequency (HF) radars have been assimilated into a three-dimensional ocean circulation model configured along the Oregon coast for the period of June-July 2008. The nonlinear model is based on the Regional Ocean Modeling System (ROMS), and the data assimilation (DA) on the indirect representer-based variational algorithm, using tangent linear and adjoint codes AVRORA developed by our group. Daily averaged surface velocity maps used for assimilation are a blend of data from several standard- and long-range HF radars in an area that extends 150 km offshore between 42-47N. Assimilation proceeds in a series of 3-day windows. In each window, AVRORA is implemented to obtain improved initial conditions, and then the nonlinear ROMS is run for a period of 6 days (3-day analysis plus 3-day forecast). Both the analyses and forecasts are evaluated against the surface velocity data as well as satellite SST daily composites and satellite altimetry data that were not assimilated. Throughout the 60-day experiment, the DA system improves the area-averaged model-data RMS difference and correlation with respect to both the assimilated surface velocity and validation SST, for both the analysis and forecast periods. Experiments with two different initial condition error covariances are compared. The first is based on the use of a balanced operator that yields the correction to the initial condition in approximate geostrophic and thermal wind balance; the second is multivariate, but not dynamically balanced. Both experiments show similar analysis and forecast skills with respect to the sea surface velocity. However, the case with the balanced error covariance yields better model-data difference statistics for SST, particularly during the forecast periods.

Contact Information:
Peng Yu, Corvallis, Oregon, United States, 97331-5503, pyu@coas.oregonstate.edu