Eddy detection inverted from Argo profiles to surface altimetry

Abstract

Abstract Argo floats are widely used to characterize vertical structures of ocean eddies, yet their capability to invert sea-surface features of eddies, especially those overlooked by available altimeters, has not been explored. In this paper, we propose an “interior-to-surface” inversion algorithm to effectively expand the capacity of eddy detection by estimating altimeter-missed eddies’ surface attributes from their Argo-derived potential density anomaly profiles, given that interior property and surface signature of eddies are highly correlated. An altimeter-calibrated machine learning ensemble is employed for the inversion training based on the joint altimeter-Argo eddy data and shows promising performance with mean absolute errors of 5.4 km, 0.5 cm, and 14.3 cm2/s2 for eddy radius, amplitude, and kinetic energy. Then, the trained ensemble model is applied to independently invert the properties of eddies captured by an Argo-alone detection scheme, which yields a high spatiotemporal consistency with their altimeter-captured counterparts. In particular, a portion of Argo-alone eddies is ~25% smaller than altimeter-derived ones, indicating Argo’s unique capability of profiling weaker submesoscale eddies. Sea surface temperature and chlorophyll data are further applied to validate the reliability of eddies identified and characterized by the Argo-only algorithm. This new methodology effectively complements that of altimetry in eddy detecting and can be expanded to estimate other physical/biochemical eddy variables from a variety of in-situ observations.