Dynamics and Predictability of Global Terrestrial Water Storage.

Terrestrial water storage (TWS) has a direct impact on fresh water resource, food security, and ecosystem function. It also plays a key role in the climate system and contributes to the development and persistence of extreme event such as droughts and wet spells. Therefore, understanding global TWS dynamics and predictability at seasonal or even longer time scale is critical for an improvement of climate prediction and for a sustainable management of water resources and infrastructure in the face of climate change. However, this is a grand challenge due to sparse and uneven long-term observations, and absence of systematical analysis of hydrological predictability. Here we use the Gravity Recovery and Climate Experiment data and Community Land Model (CLM4.5) simulation to analyze hydrological memory with a newly proposed metric that characterize TWS "inertia" after rainfalls. We display the global TWS memory distribution with various soil and vegetation conditions, and perform decadal TWS hindcast experiments with the Ensemble Streamflow Prediction (ESP) method and the reverse ESP (R-ESP) method. Our result shows that decadal prediction for TWS, only based on initial condition, is skillful over 1/3 land areas where deep soil moisture and aquifer have a non-negligible variability. This study provides a new perspective of TWS memory and quantifies its impact on decadal hydrological prediction.