The utilization of satellite altimetry poses real challenges and must rely on sophisticated inversion techniques to infer subsurface information from measurements of the ocean surface. Besides the difficult specification of background error covariances, the presence of systematic model errors--especially in the Eastern Pacific--can not be ignored.

The NASA GMAO uses a massively parallel ensemble Kalman filter (EnKF) to process satellite altimetry and in situ temperature measurements with its global OGCM. Covariance localization addresses low-rank issues and facilitates the decomposition of the analysis between processors. Recently, the EnKF has been modified to account for the presence of a forecast-model bias. The online bias correction (OBC) algorithm uses the same ensemble of model state vectors to estimate biased-error and unbiased-error covariance matrices. Covariance localization is also used for the bias but the bias covariances have different localization scales than the unbiased-error covariances. The application discussed in this talk involves a high-resolution version of the OGCM with over 30 million prognostic state variables.

Experiments in which Topex/Poseidon altimeter anomalies are assimilated into the OGCM show that the OBC reduces the RMS observation minus forecast difference for sea-surface height over a similar EnKF run in which OBC is not used. Independent in situ temperature observations show that the temperature field is also improved. The talk will conclude with an examination of the impact that the simultaneous processing of temperature and altimeter observations with the EnKF into the OGCM has on the seasonal forecast skill of the GMAO coupled model.