Assimilated Deep Learning to Assess Terrestrial Hydrology

Abstract

In this chapter, an assimilated deep learning approach is employed to improve understanding of the links between global climate teleconnection patterns (e.g., ENSO) and changes in terrestrial water storage (TWS). To this end, a hybrid deep learning framework was developed to reconstruct climate-driven TWS and to assess the influence of key climatic drivers on the spatio-temporal distribution of TWS. Using South America as a case study to showcase the potential and utility of this framework, the methodology, challenges, and prospect of machine learning in satellite hydrology are also detailed.