STATISTICAL APPROACHES TO DE-ALIASING WITH APPLICATION TO EARTH SYSTEM MONITORING

Abstract

Space-borne altimeters typically have a repeat time of order 10 days and so alias high frequency sea level variability. State-of-the-art de-aliasing methods are presently based on tidal and atmospheric corrections from dynamical models. However, analysis shows that significant high frequency variability remains after such corrections that could cause aliasing problems. In order to further de-alias the altimetry products, a statistical de-aliasing model is designed. Three methods are designed to fit the model (i) in the lag domain, (ii) in the frequency domain, and (iii) in the time domain using the lasso to limit the number of predictors. The three methods are first tested in two simulation-based studies and shown to be both effective and interpretable. The methods are then applied to the altimetry products. The lasso-based method performs best and reduces the standard deviation of the satellite altimetry products in the Gulf of St. Lawrence from about 8 cm to 6 cm.