Global barotropic variability of the ocean in response to atmospheric forcing based on multichannel regression and Kalman filter techniques

Abstract

TOPEX/POSEIDON altimetry data are employed in the analysis of the global ocean response to atmospheric forcing. We use two different approaches to test the hypothesis that the global sea surface height variability can be adequately described by linear barotropic ocean models: the multichannel regression and the optimal smoothing techniques. We start with the simplest linear vorticity balance and continue by building a hierarchy of more complicated models by including effects of topography and time dependence. We use auto-regressive external (ARX) time-series models to test the hypothesis in all of the Pacific Ocean. We also test whether any significant residual regression on the atmospheric loading is left after the inverted barometer effect is corrected for. We find that no linear barotropic model is consistent with the data. We provide a check on the results of the multichannel regression by using a Kalman filter and optimal smoother. We use sequential estimation in the form of filteringsmoothing algorithm. We run the estimate for an area of 4000 km by 2000 km in the Northeast Pacific. We analyze model and data error structures by simulating the model without data assimilation. The results show that the model forecast on average explains 33% of the data variability. The Kalman filter updates the model very efficiently and produces an estimate which explains 76% of the data variance. The optimal smoother estimate is very similar to the Kalman filter estimate. Running the model in other regions of the Pacific produced worse fits of the model to the data. This supports the conclusion that the linear barotropic dynamics fails to describe the SSH variability.