Improving Satellite Remote Sensing Estimates of the Global Terrestrial Hydrologic Cycle via Neural Network Modeling