Evolving from atmospheric and oceanic sciences that focused on single components of the earth system, significant advancements have been made in Earth system science in the last 40 years. Responding to the societal need for predictions from weather to century time scales, there is an increasing need for improving understanding and modeling of Earth as a fully coupled system of many interactive components. Observations have provided the cornerstone for process understanding, modeling, and prediction. USGCRP has leveraged and contributed to field campaigns, process measurements, and remote sensing observations to advance its objective of Earth system science and predictions. This presentation will discuss examples of improved understanding, modeling, and predictions enabled by observations, with an emphasis on regional and global water cycle processes. The latter represent major gaps in Earth system science and predictions with immense implications ranging from weather extremes to long-term climate projections.