Soil moisture and temperature have long been recognized to provide important sources of predictability for weather and climate through their inherent memory and local land-atmosphere interactions. However, through diabatic heating and changes in atmospheric circulations, anomalies of soil moisture and temperature may also influence the remote oceans, thereby providing predictability at timescales beyond their own memory. Here a weakly coupled land data assimilation system (WCLDA) is implemented in two global climate models to investigate the role of soil moisture and temperature in subseasonal-to-decadal climate predictability. Hindcast experiments initialized by the atmosphere, land, and ocean states taken from fully coupled simulations in which soil moisture and temperature are continuously assimilated through the WCLDA show substantial skill in predicting surface temperature on interannual timescale in many land regions worldwide. Interestingly, hindcasts initialized in 2001 reproduce the observed summertime atmospheric circulation and surface anomalies associated with the 2003 European summer heatwave. Sensitivity experiments isolate the dominant influence of surface heating over the Tibetan Plateau on sea surface temperatures in the tropical oceans and North Atlantic which contribute to the extended predictability of the heatwave. Our results highlight the synergistic roles of land and ocean through land-atmosphere-ocean interactions in predictability of climate including extreme climatic events and motivate research to better understand subseasonal-to-decadal predictability beyond the dominant view of ocean’s influence.