The NASA Seasonal-to-Interannual Prediction Project (NSIPP) uses data assimilation to initialize the ocean component of its coupled ocean/atmosphere/land general circulation model (CGCM) to produce routine seasonal forecasts. The forecasts are produced by means of ensemble integrations of the CGCM.

At present, univariate optimal interpolation (UOI) is used to assimilate in situ temperature observations into the ocean model. However, a substantial effort has been put into the design, implementation and testing of a massively parallel multivariate ensemble Kalman filter (MvEnKF) which is intended to replace the UOI in the ocean model initialization. The MvEnKF has been tested in experiments involving the assimilation of in situ temperature observations and remotely sensed sea surface height (SSH) measurements, first in a Pacific basin configuration of the ocean GCM and, more recently, in a global 27-layer, 2/3-degree configuration of the ocean model.

This paper will summarize the progress made by NSIPP with the MvEnKF. Specifically, the results of experiments in which sea-surface height observations from TOPEX/Poseidon are assimilated into the ocean GCM and independent current and temperature data are used for cross validation will be presented.