Dimensionality-Reduction of Climate Data using Deep Autoencoders.

Ocean surface current is one of the main oceanographic variables. To estimate and track these currents, we use satellite measurements of Sea Surface Height (SSH), but these data are sparse in space and time, as they are collected along altimeter tracks. However, Sea Surface Temperature (SST) observations are much more complete in both space and time, and so the covariance of SST and SSH can be exploited to use SST datasets to help fill in the missing information about ocean currents where SSH data are lacking. Here, we test a new data- driven methodology combining SST and SSH information to estimate the ocean surface currents in the Agulhas current.