Improving simulation of Earth system variability through weakly coupled ocean data assimilation in E3SM

Abstract. Accurate initialization of ocean states is essential for skillful prediction of Earth system variability across seasonal-to-decadal timescales. In this study, we evaluate the impact of a newly developed four-dimensional ensemble variational (4DEnVar)-based weakly coupled ocean data assimilation (WCODA) system within the DOE Energy Exascale Earth System Model version 2 (E3SMv2) on global and regional climate variability. By assimilating monthly ocean temperature and salinity from the EN4.2.1 reanalysis into the fully coupled model, we demonstrate substantial improvements in simulating both interannual and decadal climate variability. Compared with the free-running coupled simulation, the assimilation experiment exhibits markedly enhanced interannual correlations with observations for global mean surface air temperature and precipitation anomalies. The temporal variability of key climate modes, including ENSO, the Indian Ocean Dipole, and multidecadal variability in the Pacific and Atlantic Oceans, also shows markedly improved phase agreement with observations. Regional evaluation over the contiguous United States further shows enhanced skill in simulating winter surface air temperature and precipitation, particularly in the northern and southern regions, respectively, with these improvements linked to improved ENSO simulation. Additional hindcast experiments initialized from the WCODA system exhibit no appreciable initialization shock in the early years and reproduce physically coherent ENSO teleconnection patterns, suggesting the dynamical consistency of the coupled initialization framework. These findings underscore the critical role of coupled forecasts in the data assimilation cycle for propagating observational information across Earth system components. By assimilating ocean reanalysis within the fully coupled framework, the WCODA system enables cross-component information exchange among the ocean, atmosphere, and land, thereby generating dynamically consistent initial conditions that support more accurate simulations of Earth system variability and lay the foundation for seasonal-to-decadal prediction applications.