ABSTRACT
This chapter describes a PROcess-guided deep learning and DAta-driven modeling (PRODA) approach to optimize parameterization of Earth system models (ESMs) using spatio-temporal datasets. PRODA involves both data assimilation to estimate parameter values and deep learning to predict spatial and temporal distributions of parameter values so as to optimize ESM prediction. An application to the Community Land Model version 5 (CLM5) using soil organic carbon (SOC) distributions in the conterminous United States illustrates the potential and utility of the PRODA approach.
