Improving long-term prediction of terrestrial water storage through integration with CMIP6 decadal prediction

Accurate climate prediction is crucial for terrestrial water storage (TWS) decadal prediction, which contributes to the sustainable development of hydrological infrastructure. Although the initial memories from atmosphere, ocean and land surface are important sources of climate predictability, their impacts on the decadal hydrological prediction still remain unknown. Here, climate predictions with different initialization strategies from the sixth Coupled Model Intercomparison Project (CMIP6) are incorporated into the hydrological predictions over global major river basins through the elasticity framework. Integrations of the climate initialization and external forcings can improve the TWS prediction skill (Nash-Sutcliffe efficiency) by 0.14–0.24 over 69 % basins against a reference forecast without any climate prediction information, especially over high-latitudes at long lead time. Specifically, climate initialization results in a higher skill for TWS prediction over 62.5 % of basins, while considering the Atlantic or Pacific sea surface temperature information is benefit to the hydrological prediction over 38 %–90 % of basins at different lead times. Our findings imply that reliable hydrological decadal prediction can be achieved if skillful climate prediction that originates from initial conditions, external forcings and specific climate variability has been utilized appropriately.